* [RFC] dmaengine: assorted patches and Freescale SDMA support
@ 2010-08-16 11:07 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 11:07 UTC (permalink / raw)
To: linux-kernel; +Cc: Linus Walleij, Dan Williams, linux-arm-kernel
Hi all,
This is the first attempt to implement the Freescale i.MX SDMA engine
as part of the dmaengine API.
The first patch adds support for cyclic transfers usable for audio. There
is a similar approach implemented for the dw_dmac, this patch instead adds
the cyclic transfers to the dmaengine API.
Other patches not posted here are necessary to actually make use of the
SDMA engine. Please pull the following branch if you want to test these
patches. The branch includes mxcmmc support for the i.MX31 based on SDMA
and i.MX35-3stack audio support (sgtl5000 codec):
http://git.pengutronix.de/git/imx/linux-2.6.git sdma-sdmaengine
For generating firmware images for the SDMA engine look at
http://git.pengutronix.de/git/imx/sdma-firmware.git
Sascha
^ permalink raw reply [flat|nested] 78+ messages in thread
* [RFC] dmaengine: assorted patches and Freescale SDMA support
@ 2010-08-16 11:07 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 11:07 UTC (permalink / raw)
To: linux-arm-kernel
Hi all,
This is the first attempt to implement the Freescale i.MX SDMA engine
as part of the dmaengine API.
The first patch adds support for cyclic transfers usable for audio. There
is a similar approach implemented for the dw_dmac, this patch instead adds
the cyclic transfers to the dmaengine API.
Other patches not posted here are necessary to actually make use of the
SDMA engine. Please pull the following branch if you want to test these
patches. The branch includes mxcmmc support for the i.MX31 based on SDMA
and i.MX35-3stack audio support (sgtl5000 codec):
http://git.pengutronix.de/git/imx/linux-2.6.git sdma-sdmaengine
For generating firmware images for the SDMA engine look at
http://git.pengutronix.de/git/imx/sdma-firmware.git
Sascha
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-16 11:07 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 11:07 UTC (permalink / raw)
To: linux-kernel
Cc: Linus Walleij, Dan Williams, linux-arm-kernel, Sascha Hauer,
Haavard Skinnemoen
Cyclic transfers are useful for audio where a single buffer divided
in periods has to be transfered endlessly until stopped. After being
prepared the transfer is started using the dma_async_descriptor->tx_submit
function. dma_async_descriptor->callback is called after each period.
The transfer is stopped using the DMA_TERMINATE_ALL callback.
While being used for cyclic transfers the channel cannot be used
for other transfer types.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
---
drivers/dma/dmaengine.c | 2 ++
include/linux/dmaengine.h | 6 +++++-
2 files changed, 7 insertions(+), 1 deletions(-)
diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 9d31d5e..e5e79ce 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -692,6 +692,8 @@ int dma_async_device_register(struct dma_device *device)
!device->device_prep_dma_interrupt);
BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
!device->device_prep_slave_sg);
+ BUG_ON(dma_has_cap(DMA_CYCLIC, device->cap_mask) &&
+ !device->device_prep_dma_cyclic);
BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
!device->device_control);
diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
index c61d4ca..0df7864 100644
--- a/include/linux/dmaengine.h
+++ b/include/linux/dmaengine.h
@@ -67,10 +67,11 @@ enum dma_transaction_type {
DMA_PRIVATE,
DMA_ASYNC_TX,
DMA_SLAVE,
+ DMA_CYCLIC,
};
/* last transaction type for creation of the capabilities mask */
-#define DMA_TX_TYPE_END (DMA_SLAVE + 1)
+#define DMA_TX_TYPE_END (DMA_CYCLIC + 1)
/**
@@ -478,6 +479,9 @@ struct dma_device {
struct dma_chan *chan, struct scatterlist *sgl,
unsigned int sg_len, enum dma_data_direction direction,
unsigned long flags);
+ struct dma_async_tx_descriptor *(*device_prep_dma_cyclic)(
+ struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction);
int (*device_control)(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
unsigned long arg);
--
1.7.1
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
@ 2010-08-16 11:07 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 11:07 UTC (permalink / raw)
To: linux-arm-kernel
Cyclic transfers are useful for audio where a single buffer divided
in periods has to be transfered endlessly until stopped. After being
prepared the transfer is started using the dma_async_descriptor->tx_submit
function. dma_async_descriptor->callback is called after each period.
The transfer is stopped using the DMA_TERMINATE_ALL callback.
While being used for cyclic transfers the channel cannot be used
for other transfer types.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
---
drivers/dma/dmaengine.c | 2 ++
include/linux/dmaengine.h | 6 +++++-
2 files changed, 7 insertions(+), 1 deletions(-)
diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 9d31d5e..e5e79ce 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -692,6 +692,8 @@ int dma_async_device_register(struct dma_device *device)
!device->device_prep_dma_interrupt);
BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
!device->device_prep_slave_sg);
+ BUG_ON(dma_has_cap(DMA_CYCLIC, device->cap_mask) &&
+ !device->device_prep_dma_cyclic);
BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
!device->device_control);
diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
index c61d4ca..0df7864 100644
--- a/include/linux/dmaengine.h
+++ b/include/linux/dmaengine.h
@@ -67,10 +67,11 @@ enum dma_transaction_type {
DMA_PRIVATE,
DMA_ASYNC_TX,
DMA_SLAVE,
+ DMA_CYCLIC,
};
/* last transaction type for creation of the capabilities mask */
-#define DMA_TX_TYPE_END (DMA_SLAVE + 1)
+#define DMA_TX_TYPE_END (DMA_CYCLIC + 1)
/**
@@ -478,6 +479,9 @@ struct dma_device {
struct dma_chan *chan, struct scatterlist *sgl,
unsigned int sg_len, enum dma_data_direction direction,
unsigned long flags);
+ struct dma_async_tx_descriptor *(*device_prep_dma_cyclic)(
+ struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction);
int (*device_control)(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
unsigned long arg);
--
1.7.1
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-16 11:07 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 11:07 UTC (permalink / raw)
To: linux-kernel; +Cc: Linus Walleij, Dan Williams, linux-arm-kernel, Sascha Hauer
Currently dmaengine users have to explicitely dereference function
pointers in struct dma_device. For the convenience of drivers and
to be more flexible when changing the dmaengine later add static
inline wrapper functions for the dma commands.
This patch is not complete yet. If there's consensus on this patch
I'll provide an updated patch with the missing functions.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
include/linux/dmaengine.h | 41 +++++++++++++++++++++++++++++++++++++++++
1 files changed, 41 insertions(+), 0 deletions(-)
diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
index 0df7864..635c60b 100644
--- a/include/linux/dmaengine.h
+++ b/include/linux/dmaengine.h
@@ -491,6 +491,47 @@ struct dma_device {
void (*device_issue_pending)(struct dma_chan *chan);
};
+static inline int dmaengine_device_control(struct dma_chan *chan,
+ enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ return chan->device->device_control(chan, cmd, arg);
+}
+
+static inline int dmaengine_slave_config(struct dma_chan *chan,
+ struct dma_slave_config *config)
+{
+ return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
+ (unsigned long)config);
+}
+
+static inline int dmaengine_terminate_all(struct dma_chan *chan)
+{
+ return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
+}
+
+static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
+ flags);
+}
+
+static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
+ struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
+ period_len, direction);
+}
+
+static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
+{
+ return desc->tx_submit(desc);
+}
+
static inline bool dmaengine_check_align(u8 align, size_t off1, size_t off2, size_t len)
{
size_t mask;
--
1.7.1
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
@ 2010-08-16 11:07 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 11:07 UTC (permalink / raw)
To: linux-arm-kernel
Currently dmaengine users have to explicitely dereference function
pointers in struct dma_device. For the convenience of drivers and
to be more flexible when changing the dmaengine later add static
inline wrapper functions for the dma commands.
This patch is not complete yet. If there's consensus on this patch
I'll provide an updated patch with the missing functions.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
include/linux/dmaengine.h | 41 +++++++++++++++++++++++++++++++++++++++++
1 files changed, 41 insertions(+), 0 deletions(-)
diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
index 0df7864..635c60b 100644
--- a/include/linux/dmaengine.h
+++ b/include/linux/dmaengine.h
@@ -491,6 +491,47 @@ struct dma_device {
void (*device_issue_pending)(struct dma_chan *chan);
};
+static inline int dmaengine_device_control(struct dma_chan *chan,
+ enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ return chan->device->device_control(chan, cmd, arg);
+}
+
+static inline int dmaengine_slave_config(struct dma_chan *chan,
+ struct dma_slave_config *config)
+{
+ return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
+ (unsigned long)config);
+}
+
+static inline int dmaengine_terminate_all(struct dma_chan *chan)
+{
+ return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
+}
+
+static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
+ flags);
+}
+
+static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
+ struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
+ period_len, direction);
+}
+
+static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
+{
+ return desc->tx_submit(desc);
+}
+
static inline bool dmaengine_check_align(u8 align, size_t off1, size_t off2, size_t len)
{
size_t mask;
--
1.7.1
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-16 11:07 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 11:07 UTC (permalink / raw)
To: linux-kernel; +Cc: Linus Walleij, Dan Williams, linux-arm-kernel, Sascha Hauer
This patch adds support for the Freescale i.MX SDMA engine.
The SDMA engine is a scatter/gather DMA engine which is implemented
as a seperate coprocessor. SDMA needs its own firmware which is
requested using the standard request_firmware mechanism. The firmware
has different entry points for each peripheral type, so drivers
have to pass the peripheral type to the DMA engine which in turn
picks the correct firmware entry point from a table contained in
the firmware image itself.
The original Freescale code also supports support for transfering
data to the internal SRAM which needs different entry points to
the firmware. Support for this is currently not implemented. Also,
support for the ASRC (asymmetric sample rate converter) is skipped.
I took a very simple approach to implement dmaengine support. Only
a single descriptor is statically assigned to a each channel. This
means that transfers can't be queued up but only a single transfer
is in progress. This simplifies implementation a lot and is sufficient
for the usual device/memory transfers.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
arch/arm/mach-mx3/Kconfig | 2 +
arch/arm/plat-mxc/Kconfig | 10 +
arch/arm/plat-mxc/include/mach/dma.h | 64 ++
arch/arm/plat-mxc/include/mach/sdma.h | 8 +
drivers/dma/Kconfig | 8 +
drivers/dma/Makefile | 1 +
drivers/dma/imx-sdma.c | 1383 +++++++++++++++++++++++++++++++
8 files changed, 1478 insertions(+), 6 deletions(-)
create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
create mode 100644 drivers/dma/imx-sdma.c
diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
index 287431c..ac6fd71 100644
--- a/arch/arm/mach-imx/include/mach/dma-v1.h
+++ b/arch/arm/mach-imx/include/mach/dma-v1.h
@@ -27,6 +27,8 @@
#define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
+#include <mach/dma.h>
+
#define IMX_DMA_CHANNELS 16
#define DMA_MODE_READ 0
@@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
void imx_dma_free(int channel);
-enum imx_dma_prio {
- DMA_PRIO_HIGH = 0,
- DMA_PRIO_MEDIUM = 1,
- DMA_PRIO_LOW = 2
-};
-
int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
#endif /* __MACH_DMA_V1_H__ */
diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
index 85beece..301375c 100644
--- a/arch/arm/mach-mx3/Kconfig
+++ b/arch/arm/mach-mx3/Kconfig
@@ -3,12 +3,14 @@ if ARCH_MX3
config ARCH_MX31
select ARCH_HAS_RNGA
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
bool
config ARCH_MX35
bool
select ARCH_MXC_IOMUX_V3
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
comment "MX3 platforms:"
diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
index 0527e65..6741625 100644
--- a/arch/arm/plat-mxc/Kconfig
+++ b/arch/arm/plat-mxc/Kconfig
@@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
config ARCH_MXC_AUDMUX_V2
bool
+config IMX_HAVE_SDMA
+ bool
+
+config IMX_SDMA
+ depends on IMX_HAVE_SDMA
+ tristate "Enable SDMA support"
+ help
+ Include support for the SDMA engine. The SDMA engine needs additional
+ firmware support. SDMA can be compiled as a module to support loading
+ the firmware when a rootfs is present.
endif
diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
new file mode 100644
index 0000000..94cab29
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/dma.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARCH_MXC_DMA_H__
+#define __ASM_ARCH_MXC_DMA_H__
+
+#include <linux/scatterlist.h>
+
+/*
+ * This enumerates peripheral types. Used for SDMA.
+ */
+typedef enum {
+ IMX_DMATYPE_SSI, /* MCU domain SSI */
+ IMX_DMATYPE_SSI_SP, /* Shared SSI */
+ IMX_DMATYPE_MMC, /* MMC */
+ IMX_DMATYPE_SDHC, /* SDHC */
+ IMX_DMATYPE_UART, /* MCU domain UART */
+ IMX_DMATYPE_UART_SP, /* Shared UART */
+ IMX_DMATYPE_FIRI, /* FIRI */
+ IMX_DMATYPE_CSPI, /* MCU domain CSPI */
+ IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
+ IMX_DMATYPE_SIM, /* SIM */
+ IMX_DMATYPE_ATA, /* ATA */
+ IMX_DMATYPE_CCM, /* CCM */
+ IMX_DMATYPE_EXT, /* External peripheral */
+ IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
+ IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
+ IMX_DMATYPE_DSP, /* DSP */
+ IMX_DMATYPE_MEMORY, /* Memory */
+ IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
+ IMX_DMATYPE_SPDIF, /* SPDIF */
+ IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
+ IMX_DMATYPE_ASRC, /* ASRC */
+ IMX_DMATYPE_ESAI, /* ESAI */
+} sdma_peripheral_type;
+
+enum imx_dma_prio {
+ DMA_PRIO_HIGH = 0,
+ DMA_PRIO_MEDIUM = 1,
+ DMA_PRIO_LOW = 2
+};
+
+struct imx_dma_data {
+ int dma_request; /* DMA request line */
+ sdma_peripheral_type peripheral_type;
+ int priority;
+};
+
+static inline int imx_dma_is_ipu(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "ipu-core");
+}
+
+static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "imx-sdma");
+}
+
+#endif
diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
new file mode 100644
index 0000000..5d542b8
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/sdma.h
@@ -0,0 +1,8 @@
+#ifndef __MACH_MXC_SDMA_H__
+#define __MACH_MXC_SDMA_H__
+
+struct sdma_platform_data {
+ int sdma_version;
+};
+
+#endif /* __MACH_MXC_SDMA_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9520cf0..f76bda9 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -195,6 +195,14 @@ config PCH_DMA
help
Enable support for the Topcliff PCH DMA engine.
+config IMX_SDMA
+ tristate "Atmel AHB DMA support"
+ depends on ARCH_MXC
+ select DMA_ENGINE
+ help
+ Support the i.MX SDMA engine. This engine is integrated into
+ Freescale i.MX25/31/35/51 chips.
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 72bd703..14d7a1b 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
obj-$(CONFIG_PCH_DMA) += pch_dma.o
+obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
new file mode 100644
index 0000000..3ba7905
--- /dev/null
+++ b/drivers/dma/imx-sdma.c
@@ -0,0 +1,1383 @@
+/*
+ * drivers/dma/imx-sdma.c
+ *
+ * This file contains a driver for the Freescale Smart DMA engine
+ *
+ * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
+ *
+ * Based on code from Freescale:
+ *
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * The code contained herein is licensed under the GNU General Public
+ * License. You may obtain a copy of the GNU General Public License
+ * Version 2 or later at the following locations:
+ *
+ * http://www.opensource.org/licenses/gpl-license.html
+ * http://www.gnu.org/copyleft/gpl.html
+ */
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <linux/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/dmaengine.h>
+
+#include <asm/irq.h>
+#include <mach/sdma.h>
+#include <mach/dma.h>
+#include <mach/hardware.h>
+
+/* SDMA registers */
+#define SDMA_H_C0PTR (sdma_base + 0x000)
+#define SDMA_H_INTR (sdma_base + 0x004)
+#define SDMA_H_STATSTOP (sdma_base + 0x008)
+#define SDMA_H_START (sdma_base + 0x00c)
+#define SDMA_H_EVTOVR (sdma_base + 0x010)
+#define SDMA_H_DSPOVR (sdma_base + 0x014)
+#define SDMA_H_HOSTOVR (sdma_base + 0x018)
+#define SDMA_H_EVTPEND (sdma_base + 0x01c)
+#define SDMA_H_DSPENBL (sdma_base + 0x020)
+#define SDMA_H_RESET (sdma_base + 0x024)
+#define SDMA_H_EVTERR (sdma_base + 0x028)
+#define SDMA_H_INTRMSK (sdma_base + 0x02c)
+#define SDMA_H_PSW (sdma_base + 0x030)
+#define SDMA_H_EVTERRDBG (sdma_base + 0x034)
+#define SDMA_H_CONFIG (sdma_base + 0x038)
+#define SDMA_ONCE_ENB (sdma_base + 0x040)
+#define SDMA_ONCE_DATA (sdma_base + 0x044)
+#define SDMA_ONCE_INSTR (sdma_base + 0x048)
+#define SDMA_ONCE_STAT (sdma_base + 0x04c)
+#define SDMA_ONCE_CMD (sdma_base + 0x050)
+#define SDMA_EVT_MIRROR (sdma_base + 0x054)
+#define SDMA_ILLINSTADDR (sdma_base + 0x058)
+#define SDMA_CHN0ADDR (sdma_base + 0x05c)
+#define SDMA_ONCE_RTB (sdma_base + 0x060)
+#define SDMA_XTRIG_CONF1 (sdma_base + 0x070)
+#define SDMA_XTRIG_CONF2 (sdma_base + 0x074)
+#define SDMA_CHNENBL_0 (sdma_base + (sdma_version == 2 ? 0x200 : 0x80))
+#define SDMA_CHNPRI_0 (sdma_base + 0x100)
+
+/*
+ * Buffer descriptor status values.
+ */
+#define BD_DONE 0x01
+#define BD_WRAP 0x02
+#define BD_CONT 0x04
+#define BD_INTR 0x08
+#define BD_RROR 0x10
+#define BD_LAST 0x20
+#define BD_EXTD 0x80
+
+/*
+ * Data Node descriptor status values.
+ */
+#define DND_END_OF_FRAME 0x80
+#define DND_END_OF_XFER 0x40
+#define DND_DONE 0x20
+#define DND_UNUSED 0x01
+
+/*
+ * IPCV2 descriptor status values.
+ */
+#define BD_IPCV2_END_OF_FRAME 0x40
+
+#define IPCV2_MAX_NODES 50
+/*
+ * Error bit set in the CCB status field by the SDMA,
+ * in setbd routine, in case of a transfer error
+ */
+#define DATA_ERROR 0x10000000
+
+/*
+ * Buffer descriptor commands.
+ */
+#define C0_ADDR 0x01
+#define C0_LOAD 0x02
+#define C0_DUMP 0x03
+#define C0_SETCTX 0x07
+#define C0_GETCTX 0x03
+#define C0_SETDM 0x01
+#define C0_SETPM 0x04
+#define C0_GETDM 0x02
+#define C0_GETPM 0x08
+/*
+ * Change endianness indicator in the BD command field
+ */
+#define CHANGE_ENDIANNESS 0x80
+
+/*
+ * Mode/Count of data node descriptors - IPCv2
+ */
+#ifdef __BIG_ENDIAN
+struct sdma_mode_count {
+ u32 command : 8; /* command mostlky used for channel 0 */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 count : 16; /* size of the buffer pointed by this BD */
+};
+#else
+struct sdma_mode_count {
+ u32 count : 16; /* size of the buffer pointed by this BD */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 command : 8; /* command mostlky used for channel 0 */
+};
+#endif
+
+/*
+ * Buffer descriptor
+ */
+struct sdma_buffer_descriptor {
+ struct sdma_mode_count mode;
+ u32 buffer_addr; /* address of the buffer described */
+ u32 ext_buffer_addr; /* extended buffer address */
+};
+
+/*
+ * Channel control Block
+ */
+struct sdma_channel_control {
+ u32 current_bd_ptr; /* current buffer descriptor processed */
+ u32 base_bd_ptr; /* first element of buffer descriptor array */
+ void *unused;
+ void *unused1;
+};
+
+/**
+ * Context structure.
+ */
+#ifdef __BIG_ENDIAN
+struct sdma_state_registers {
+ u32 sf : 1; /* source fault while loading data */
+ u32 unused0: 1;
+ u32 rpc :14; /* return program counter */
+ u32 t : 1; /* test bit:status of arithmetic & test instruction*/
+ u32 unused1: 1;
+ u32 pc :14; /* program counter */
+ u32 lm : 2; /* loop mode */
+ u32 epc :14; /* loop end program counter */
+ u32 df : 1; /* destination fault while storing data */
+ u32 unused2: 1;
+ u32 spc :14; /* loop start program counter */
+};
+#else
+struct sdma_state_registers {
+ u32 pc :14; /* program counter */
+ u32 unused1: 1;
+ u32 t : 1; /* test bit: status of arithmetic & test instruction*/
+ u32 rpc :14; /* return program counter */
+ u32 unused0: 1;
+ u32 sf : 1; /* source fault while loading data */
+ u32 spc :14; /* loop start program counter */
+ u32 unused2: 1;
+ u32 df : 1; /* destination fault while storing data */
+ u32 epc :14; /* loop end program counter */
+ u32 lm : 2; /* loop mode */
+};
+#endif
+
+struct sdma_context_data {
+ struct sdma_state_registers channel_state; /* channel state bits */
+ u32 gReg[8]; /* general registers */
+ u32 mda; /* burst dma destination address register */
+ u32 msa; /* burst dma source address register */
+ u32 ms; /* burst dma status register */
+ u32 md; /* burst dma data register */
+ u32 pda; /* peripheral dma destination address register */
+ u32 psa; /* peripheral dma source address register */
+ u32 ps; /* peripheral dma status register */
+ u32 pd; /* peripheral dma data register */
+ u32 ca; /* CRC polynomial register */
+ u32 cs; /* CRC accumulator register */
+ u32 dda; /* dedicated core destination address register */
+ u32 dsa; /* dedicated core source address register */
+ u32 ds; /* dedicated core status register */
+ u32 dd; /* dedicated core data register */
+ u32 scratch0;
+ u32 scratch1;
+ u32 scratch2;
+ u32 scratch3;
+ u32 scratch4;
+ u32 scratch5;
+ u32 scratch6;
+ u32 scratch7;
+};
+
+struct sdma_channel {
+ /* Channel number */
+ int channel;
+ /* Transfer type. Needed for setting SDMA script */
+ enum dma_data_direction direction;
+ /* Peripheral type. Needed for setting SDMA script */
+ sdma_peripheral_type peripheral_type;
+ /* Peripheral event id */
+ int event_id;
+ /* Peripheral event id2 (for channels that use 2 events) */
+ int event_id2;
+ /* SDMA data access word size */
+ unsigned long word_size;
+
+ /* ID of the buffer that was processed */
+ unsigned int buf_tail;
+
+ wait_queue_head_t waitq; /* channel completion waitqeue */
+
+ int num_bd;
+
+ struct sdma_buffer_descriptor *bd;
+ dma_addr_t bd_phys;
+
+ int pc_from_device, pc_to_device;
+
+ unsigned long flags;
+ dma_addr_t per_address;
+
+ uint32_t event_mask1, event_mask2;
+ uint32_t watermark_level;
+ uint32_t shp_addr, per_addr;
+
+ /* DMA-Engine Channel */
+ struct dma_chan chan;
+
+ spinlock_t lock;
+ struct dma_async_tx_descriptor desc;
+ dma_cookie_t last_completed;
+ int busy;
+};
+
+#define IMX_DMA_SG_LOOP (1 << 0)
+
+#define MAX_DMA_CHANNELS 32
+#define MXC_SDMA_DEFAULT_PRIORITY 1
+#define MXC_SDMA_MIN_PRIORITY 1
+#define MXC_SDMA_MAX_PRIORITY 7
+
+/*
+ * This enumerates transfer types
+ */
+typedef enum {
+ emi_2_per = 0, /* EMI memory to peripheral */
+ emi_2_int, /* EMI memory to internal RAM */
+ emi_2_emi, /* EMI memory to EMI memory */
+ emi_2_dsp, /* EMI memory to DSP memory */
+ per_2_int, /* Peripheral to internal RAM */
+ per_2_emi, /* Peripheral to internal EMI memory */
+ per_2_dsp, /* Peripheral to DSP memory */
+ per_2_per, /* Peripheral to Peripheral */
+ int_2_per, /* Internal RAM to peripheral */
+ int_2_int, /* Internal RAM to Internal RAM */
+ int_2_emi, /* Internal RAM to EMI memory */
+ int_2_dsp, /* Internal RAM to DSP memory */
+ dsp_2_per, /* DSP memory to peripheral */
+ dsp_2_int, /* DSP memory to internal RAM */
+ dsp_2_emi, /* DSP memory to EMI memory */
+ dsp_2_dsp, /* DSP memory to DSP memory */
+ emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
+ dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
+ dvfs_pll, /* DVFS script with PLL change */
+ dvfs_pdr /* DVFS script without PLL change */
+} sdma_transfer_type;
+
+/*
+ * Structure containing sdma request parameters.
+ */
+struct sdma_script_start_addrs {
+ int ap_2_ap_addr;
+ int ap_2_bp_addr;
+ int ap_2_ap_fixed_addr;
+ int bp_2_ap_addr;
+ int loopback_on_dsp_side_addr;
+ int mcu_interrupt_only_addr;
+
+ int firi_2_per_addr;
+ int firi_2_mcu_addr;
+ int per_2_firi_addr;
+ int mcu_2_firi_addr;
+
+ int uart_2_per_addr;
+ int uart_2_mcu_addr;
+ int per_2_app_addr;
+ int mcu_2_app_addr;
+ int per_2_per_addr;
+
+ int uartsh_2_per_addr;
+ int uartsh_2_mcu_addr;
+ int per_2_shp_addr;
+ int mcu_2_shp_addr;
+
+ int ata_2_mcu_addr;
+ int mcu_2_ata_addr;
+
+ int app_2_per_addr;
+ int app_2_mcu_addr;
+ int shp_2_per_addr;
+ int shp_2_mcu_addr;
+
+ int mshc_2_mcu_addr;
+ int mcu_2_mshc_addr;
+
+ int spdif_2_mcu_addr;
+ int mcu_2_spdif_addr;
+
+ int asrc_2_mcu_addr;
+
+ int ext_mem_2_ipu_addr;
+
+ int descrambler_addr;
+
+ int dptc_dvfs_addr;
+
+ int utra_addr;
+
+ int ram_code_start_addr;
+};
+
+#define SDMA_FIRMWARE_MAGIC 0x414d4453
+
+struct sdma_firmware_header {
+ uint32_t magic; /* "SDMA" */
+ uint32_t version_major; /* increased whenever layout of struct sdma_script_start_addrs changes */
+ uint32_t version_minor; /* firmware version */
+ uint32_t script_addrs_start; /* offset of struct sdma_script_start_addrs in this image */
+ uint32_t num_script_addrs; /* Number of script addresses in this image */
+ uint32_t ram_code_start; /* offset of SDMA ram image in this firmware image */
+ uint32_t ram_code_size; /* size of SDMA ram image */
+};
+
+static struct sdma_channel sdma_data[MAX_DMA_CHANNELS];
+static struct sdma_channel_control *channel_control;
+static void __iomem *sdma_base;
+static int sdma_version;
+static int sdma_num_events;
+static struct sdma_context_data *sdma_context;
+dma_addr_t sdma_context_phys;
+static struct dma_device __sdma_dma_device;
+static struct dma_device *sdma_dma_device = &__sdma_dma_device;
+
+#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
+#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
+#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
+#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
+
+static int sdma_config_ownership(int channel, int event_override,
+ int mcu_verride, int dsp_override)
+{
+ u32 evt, mcu, dsp;
+
+ if (event_override && mcu_verride && dsp_override)
+ return -EINVAL;
+
+ evt = readl(SDMA_H_EVTOVR);
+ mcu = readl(SDMA_H_HOSTOVR);
+ dsp = readl(SDMA_H_DSPOVR);
+
+ if (dsp_override)
+ dsp &= ~(1 << channel);
+ else
+ dsp |= (1 << channel);
+
+ if (event_override)
+ evt &= ~(1 << channel);
+ else
+ evt |= (1 << channel);
+
+ if (mcu_verride)
+ mcu &= ~(1 << channel);
+ else
+ mcu |= (1 << channel);
+
+ writel(evt, SDMA_H_EVTOVR);
+ writel(mcu, SDMA_H_HOSTOVR);
+ writel(dsp, SDMA_H_DSPOVR);
+
+ return 0;
+}
+
+/*
+ * sdma_run_channel - run a channel and wait till it's done
+ */
+static int sdma_run_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ int ret;
+
+ writel(1 << channel, SDMA_H_START);
+
+ ret = wait_event_interruptible(sdma->waitq,
+ !(readl(SDMA_H_STATSTOP) & (1 << channel)));
+ return ret;
+}
+
+static int sdma_load_script(void *buf, int size, u32 address)
+{
+ struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
+ void *buf_virt;
+ dma_addr_t buf_phys;
+ int ret;
+
+ buf_virt = dma_alloc_coherent(NULL,
+ size,
+ &buf_phys, GFP_KERNEL);
+ if (!buf_virt)
+ return -ENOMEM;
+
+ bd0->mode.command = C0_SETPM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = size / 2;
+ bd0->buffer_addr = buf_phys;
+ bd0->ext_buffer_addr = address;
+
+ memcpy(buf_virt, buf, size);
+
+ ret = sdma_run_channel(0);
+
+ dma_free_coherent(NULL, size, buf_virt, buf_phys);
+
+ return ret;
+}
+
+static void sdma_event_enable(int channel, int event)
+{
+ u32 val;
+
+ val = readl(SDMA_CHNENBL_0 + event * 4);
+ val |= (1 << channel);
+ writel(val, SDMA_CHNENBL_0 + event * 4);
+}
+
+static void sdma_event_disable(int channel, int event)
+{
+ u32 val;
+
+ val = readl(SDMA_CHNENBL_0 + event * 4);
+ val &= ~(1 << channel);
+ writel(val, SDMA_CHNENBL_0 + event * 4);
+}
+
+static void mxc_sdma_handle_channel_loop(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ struct sdma_buffer_descriptor *bd;
+ int error = 0;
+
+ /*
+ * loop mode. Iterate over descriptors, re-setup them and
+ * call callback function.
+ */
+ while (1) {
+ bd = &sdma->bd[sdma->buf_tail];
+
+ if (bd->mode.status & BD_DONE)
+ break;
+
+ if (bd->mode.status & BD_RROR)
+ error = -EIO;
+
+ bd->mode.status |= BD_DONE;
+ sdma->buf_tail++;
+ sdma->buf_tail %= sdma->num_bd;
+
+ if (sdma->desc.callback)
+ sdma->desc.callback(sdma->desc.callback_param);
+ }
+}
+
+static void mxc_sdma_handle_channel_normal(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ struct sdma_buffer_descriptor *bd;
+ int i, error = 0;
+
+ /*
+ * non loop mode. Iterate over all descriptors, collect
+ * errors and call callback function
+ */
+ for (i = 0; i < sdma->num_bd; i++) {
+ bd = &sdma->bd[i];
+
+ if (bd->mode.status & (BD_DONE | BD_RROR))
+ error = -EIO;
+ }
+
+ if (sdma->desc.callback)
+ sdma->desc.callback(sdma->desc.callback_param);
+ sdma->last_completed = sdma->desc.cookie;
+
+ sdma->busy = 0;
+}
+
+static void mxc_sdma_handle_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+
+ wake_up_interruptible(&sdma->waitq);
+
+ /* not interested in channel 0 interrupts */
+ if (!channel)
+ return;
+
+ if (sdma->flags & IMX_DMA_SG_LOOP)
+ mxc_sdma_handle_channel_loop(channel);
+ else
+ mxc_sdma_handle_channel_normal(channel);
+}
+
+static irqreturn_t sdma_int_handler(int irq, void *dev_id)
+{
+ u32 stat;
+
+ stat = readl(SDMA_H_INTR);
+ writel(stat, SDMA_H_INTR);
+
+ while (stat) {
+ int channel = fls(stat) - 1;
+
+ mxc_sdma_handle_channel(channel);
+
+ stat &= ~(1 << channel);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static struct clk *sdma_clk;
+
+/*
+ * Stores the start address of the SDMA scripts
+ */
+static struct sdma_script_start_addrs __sdma_script_addrs;
+static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
+
+/*
+ * sets the pc of SDMA script according to the peripheral type
+ */
+static void sdma_get_pc(struct sdma_channel *sdma,
+ sdma_peripheral_type peripheral_type)
+{
+ int res = 0;
+ int per_2_emi = 0, emi_2_per = 0;
+ int per_2_int = 0, int_2_per = 0;
+ int per_2_per = 0, emi_2_emi = 0;
+
+ sdma->pc_from_device = 0;
+ sdma->pc_to_device = 0;
+
+ switch (peripheral_type) {
+ case IMX_DMATYPE_MEMORY:
+ emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
+ break;
+ case IMX_DMATYPE_DSP:
+ emi_2_per = sdma_script_addrs->bp_2_ap_addr;
+ per_2_emi = sdma_script_addrs->ap_2_bp_addr;
+ break;
+ case IMX_DMATYPE_FIRI:
+ per_2_int = sdma_script_addrs->firi_2_per_addr;
+ per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_firi_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
+ break;
+ case IMX_DMATYPE_UART:
+ per_2_int = sdma_script_addrs->uart_2_per_addr;
+ per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_app_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_UART_SP:
+ per_2_int = sdma_script_addrs->uartsh_2_per_addr;
+ per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_shp_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ATA:
+ per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
+ break;
+ case IMX_DMATYPE_CSPI:
+ case IMX_DMATYPE_EXT:
+ case IMX_DMATYPE_SSI:
+ per_2_int = sdma_script_addrs->app_2_per_addr;
+ per_2_emi = sdma_script_addrs->app_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_app_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_SSI_SP:
+ case IMX_DMATYPE_MMC:
+ case IMX_DMATYPE_SDHC:
+ case IMX_DMATYPE_CSPI_SP:
+ case IMX_DMATYPE_ESAI:
+ case IMX_DMATYPE_MSHC_SP:
+ per_2_int = sdma_script_addrs->shp_2_per_addr;
+ per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_shp_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ASRC:
+ per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
+ per_2_per = sdma_script_addrs->per_2_per_addr;
+ break;
+ case IMX_DMATYPE_MSHC:
+ per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
+ break;
+ case IMX_DMATYPE_CCM:
+ per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
+ break;
+ case IMX_DMATYPE_FIFO_MEMORY:
+ res = sdma_script_addrs->ap_2_ap_fixed_addr;
+ break;
+ case IMX_DMATYPE_SPDIF:
+ per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
+ break;
+ case IMX_DMATYPE_IPU_MEMORY:
+ emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
+ break;
+ default:
+ break;
+ }
+
+ sdma->pc_from_device = per_2_emi;
+ sdma->pc_to_device = emi_2_per;
+}
+
+static int sdma_load_context(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ int load_address;
+ struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
+ int ret;
+
+ if (sdma->direction == DMA_FROM_DEVICE) {
+ load_address = sdma->pc_from_device;
+ } else {
+ load_address = sdma->pc_to_device;
+ }
+
+ if (load_address < 0)
+ return load_address;
+
+ pr_debug("%s: load_address = %d\n", __func__, load_address);
+ pr_debug("%s: wml = 0x%08x\n", __func__, sdma->watermark_level);
+ pr_debug("%s: shp_addr = 0x%08x\n", __func__, sdma->shp_addr);
+ pr_debug("%s: per_addr = 0x%08x\n", __func__, sdma->per_addr);
+ pr_debug("%s: event_mask1 = 0x%08x\n", __func__, sdma->event_mask1);
+ pr_debug("%s: event_mask2 = 0x%08x\n", __func__, sdma->event_mask2);
+
+ memset(sdma_context, 0, sizeof(*sdma_context));
+ sdma_context->channel_state.pc = load_address;
+
+ /* Send by context the event mask,base address for peripheral
+ * and watermark level
+ */
+ sdma_context->gReg[0] = sdma->event_mask2;
+ sdma_context->gReg[1] = sdma->event_mask1;
+ sdma_context->gReg[2] = sdma->per_addr;
+ sdma_context->gReg[6] = sdma->shp_addr;
+ sdma_context->gReg[7] = sdma->watermark_level;
+
+ bd0->mode.command = C0_SETDM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = sizeof(*sdma_context) / 4;
+ bd0->buffer_addr = sdma_context_phys;
+ bd0->ext_buffer_addr = 2048 + (sizeof(*sdma_context) / 4) * channel;
+
+ ret = sdma_run_channel(0);
+
+ return ret;
+}
+
+static void sdma_disable_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+
+ writel(1 << channel, SDMA_H_STATSTOP);
+ sdma->busy = 0;
+}
+
+static int sdma_config_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ int ret;
+
+ sdma_disable_channel(channel);
+
+ sdma->event_mask1 = 0;
+ sdma->event_mask2 = 0;
+ sdma->shp_addr = 0;
+ sdma->per_addr = 0;
+
+ if (sdma->event_id)
+ sdma_event_enable(channel, sdma->event_id);
+
+ switch (sdma->peripheral_type) {
+ case IMX_DMATYPE_DSP:
+ sdma_config_ownership(channel, 0, 1, 1);
+ break;
+ case IMX_DMATYPE_MEMORY:
+ sdma_config_ownership(channel, 0, 1, 0);
+ break;
+ default:
+ sdma_config_ownership(channel, 1, 1, 0);
+ break;
+ }
+
+ sdma_get_pc(sdma, sdma->peripheral_type);
+
+ if ((sdma->peripheral_type != IMX_DMATYPE_MEMORY) &&
+ (sdma->peripheral_type != IMX_DMATYPE_DSP)) {
+ /* Handle multiple event channels differently */
+ if (sdma->event_id2) {
+ sdma->event_mask2 = 1 << (sdma->event_id2 % 32);
+ if (sdma->event_id2 > 31)
+ sdma->watermark_level |= 1 << 31;
+ sdma->event_mask1 = 1 << (sdma->event_id % 32);
+ if (sdma->event_id > 31)
+ sdma->watermark_level |= 1 << 30;
+ } else {
+ sdma->event_mask1 = 1 << sdma->event_id;
+ sdma->event_mask2 = 1 << (sdma->event_id - 32);
+ }
+ /* Watermark Level */
+ sdma->watermark_level |= sdma->watermark_level;
+ /* Address */
+ sdma->shp_addr = sdma->per_address;
+ } else {
+ sdma->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
+ }
+
+ ret = sdma_load_context(channel);
+
+ return ret;
+}
+
+static int sdma_set_channel_priority(unsigned int channel, unsigned int priority)
+{
+ if (priority < MXC_SDMA_MIN_PRIORITY
+ || priority > MXC_SDMA_MAX_PRIORITY) {
+ return -EINVAL;
+ }
+
+ writel(priority, SDMA_CHNPRI_0 + 4 * channel);
+
+ return 0;
+}
+
+static int sdma_request_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ int ret = -EBUSY;
+
+ sdma->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdma->bd_phys, GFP_KERNEL);
+ if (!sdma->bd) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memset(sdma->bd, 0, PAGE_SIZE);
+
+ channel_control[channel].base_bd_ptr = sdma->bd_phys;
+ channel_control[channel].current_bd_ptr = sdma->bd_phys;
+
+ clk_enable(sdma_clk);
+
+ sdma_set_channel_priority(channel, MXC_SDMA_DEFAULT_PRIORITY);
+
+ init_waitqueue_head(&sdma->waitq);
+
+ sdma->buf_tail = 0;
+
+ return 0;
+out:
+
+ return ret;
+}
+
+static void sdma_enable_channel(int channel)
+{
+ writel(1 << channel, SDMA_H_START);
+}
+
+static int __init sdma_init(unsigned long phys_base, int irq, int version,
+ void *ram_code,
+ int ram_code_size)
+{
+ int i, ret;
+ int channel;
+ dma_addr_t ccb_phys;
+
+ sdma_version = version;
+ switch (sdma_version) {
+ case 1:
+ sdma_num_events = 32;
+ break;
+ case 2:
+ sdma_num_events = 48;
+ break;
+ default:
+ pr_err("SDMA: Unknown version %d. aborting\n", sdma_version);
+ return -ENODEV;
+ }
+
+ clk_enable(sdma_clk);
+
+ sdma_base = ioremap(phys_base, 4096);
+ if (!sdma_base) {
+ ret = -ENOMEM;
+ goto err_ioremap;
+ }
+
+ /* Initialize SDMA private data */
+ memset(sdma_data, 0, sizeof(struct sdma_channel) * MAX_DMA_CHANNELS);
+
+ for (channel = 0; channel < MAX_DMA_CHANNELS; channel++)
+ sdma_data[channel].channel = channel;
+
+ ret = request_irq(irq, sdma_int_handler, 0, "sdma", NULL);
+ if (ret)
+ goto err_request_irq;
+
+ /* Be sure SDMA has not started yet */
+ writel(0, SDMA_H_C0PTR);
+
+ channel_control = dma_alloc_coherent(NULL,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
+ sizeof(struct sdma_context_data),
+ &ccb_phys, GFP_KERNEL);
+
+ if (!channel_control) {
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ sdma_context = (void *)channel_control +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+ sdma_context_phys = ccb_phys +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+
+ /* Zero-out the CCB structures array just allocated */
+ memset(channel_control, 0,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
+
+ /* disable all channels */
+ for (i = 0; i < sdma_num_events; i++)
+ writel(0, SDMA_CHNENBL_0 + i * 4);
+
+ /* All channels have priority 0 */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++)
+ writel(0, SDMA_CHNPRI_0 + i * 4);
+
+ ret = sdma_request_channel(0);
+ if (ret)
+ goto err_dma_alloc;
+
+ sdma_config_ownership(0, 0, 1, 0);
+
+ /* Set Command Channel (Channel Zero) */
+ writel(0x4050, SDMA_CHN0ADDR);
+
+ /* Set bits of CONFIG register but with static context switching */
+ /* FIXME: Check whether to set ACR bit depending on clock ratios */
+ writel(0, SDMA_H_CONFIG);
+
+ writel(ccb_phys, SDMA_H_C0PTR);
+
+ /* download the RAM image for SDMA */
+ sdma_load_script(ram_code,
+ ram_code_size,
+ sdma_script_addrs->ram_code_start_addr);
+
+ /* Set bits of CONFIG register with given context switching mode */
+ writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
+
+ /* Initializes channel's priorities */
+ sdma_set_channel_priority(0, 7);
+
+ clk_disable(sdma_clk);
+
+ return 0;
+
+err_dma_alloc:
+ free_irq(irq, NULL);
+err_request_irq:
+ iounmap(sdma_base);
+err_ioremap:
+ clk_disable(sdma_clk);
+ pr_err("%s failed with %d\n", __func__, ret);
+ return ret;
+}
+
+static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
+{
+ dma_cookie_t cookie = sdma->chan.cookie;
+
+ if (++cookie < 0)
+ cookie = 1;
+
+ sdma->chan.cookie = cookie;
+ sdma->desc.cookie = cookie;
+
+ return cookie;
+}
+
+static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sdma_channel, chan);
+}
+
+static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sdma_channel *sdma = to_sdma_chan(tx->chan);
+ dma_cookie_t cookie;
+
+ spin_lock_irq(&sdma->lock);
+
+ cookie = sdma_assign_cookie(sdma);
+
+ sdma_enable_channel(tx->chan->chan_id);
+
+ spin_unlock_irq(&sdma->lock);
+
+ return cookie;
+}
+
+static int sdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ struct imx_dma_data *data = chan->private;
+ int prio, ret;
+
+ /* No need to execute this for internal channel 0 */
+ if (!chan->chan_id)
+ return 0;
+
+ if (!data)
+ return -EINVAL;
+
+ switch (data->priority) {
+ case DMA_PRIO_HIGH:
+ prio = 3;
+ break;
+ case DMA_PRIO_MEDIUM:
+ prio = 2;
+ break;
+ case DMA_PRIO_LOW:
+ default:
+ prio = 1;
+ break;
+ }
+
+ sdma->peripheral_type = data->peripheral_type;
+ sdma->event_id = data->dma_request;
+ ret = sdma_set_channel_priority(chan->chan_id, prio);
+ if (ret)
+ return ret;
+
+ if (chan->chan_id) {
+ ret = sdma_request_channel(chan->chan_id);
+ if (ret)
+ return ret;
+ }
+
+ dma_async_tx_descriptor_init(&sdma->desc, chan);
+ sdma->desc.tx_submit = sdma_tx_submit;
+ /* txd.flags will be overwritten in prep funcs */
+ sdma->desc.flags = DMA_CTRL_ACK;
+
+ return 0;
+}
+
+static void sdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ int channel = chan->chan_id;
+
+ sdma_disable_channel(channel);
+
+ if (sdma->event_id)
+ sdma_event_disable(channel, sdma->event_id);
+ if (sdma->event_id2)
+ sdma_event_disable(channel, sdma->event_id2);
+
+ sdma->event_id = 0;
+ sdma->event_id2 = 0;
+
+ sdma_set_channel_priority(channel, 0);
+
+ dma_free_coherent(NULL, PAGE_SIZE, sdma->bd, sdma->bd_phys);
+
+ clk_disable(sdma_clk);
+}
+
+#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
+
+static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ int ret, i, count;
+ int channel = chan->chan_id;
+ struct scatterlist *sg;
+
+ if (sdma->busy)
+ return NULL;
+ sdma->busy = 1;
+
+ sdma->flags = 0;
+
+ pr_debug("SDMA: setting up %d entries for channel %d.\n",
+ sg_len, channel);
+
+ sdma->direction = direction;
+ ret = sdma_load_context(channel);
+ if (ret)
+ goto err_out;
+
+ if (sg_len > NUM_BD) {
+ pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, sg_len, NUM_BD);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ struct sdma_buffer_descriptor *bd = &sdma->bd[i];
+ int param;
+
+ bd->buffer_addr = sgl->dma_address;
+
+ count = sg->length;
+
+ if (count > 0xffff) {
+ pr_err("SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
+ channel, count, 0xffff);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ bd->mode.count = count;
+
+ if (sdma->word_size > 4) {
+ ret = -EINVAL;
+ goto err_out;
+ }
+ if (sdma->word_size == 4)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdma->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT;
+
+ if (sdma->flags & IMX_DMA_SG_LOOP) {
+ param |= BD_INTR;
+ if (i + 1 == sg_len)
+ param |= BD_WRAP;
+ }
+
+ if (i + 1 == sg_len)
+ param |= BD_INTR;
+
+ pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, count, sg->dma_address,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+ }
+
+ sdma->num_bd = sg_len;
+ channel_control[channel].current_bd_ptr = sdma->bd_phys;
+
+ return &sdma->desc;
+err_out:
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ int num_periods = buf_len / period_len;
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ int channel = chan->chan_id;
+ int ret, i = 0, buf = 0;
+
+ pr_debug("%s channel: %d\n", __func__, channel);
+
+ if (sdma->busy)
+ return NULL;
+
+ sdma->busy = 1;
+
+ sdma->flags |= IMX_DMA_SG_LOOP;
+ sdma->direction = direction;
+ ret = sdma_load_context(channel);
+ if (ret)
+ goto err_out;
+
+ if (num_periods > NUM_BD) {
+ pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, num_periods, NUM_BD);
+ goto err_out;
+ }
+
+ if (period_len > 0xffff) {
+ pr_err("SDMA channel %d: maximum period size exceeded: %d > %d\n",
+ channel, period_len, 0xffff);
+ goto err_out;
+ }
+
+ while (buf < buf_len) {
+ struct sdma_buffer_descriptor *bd = &sdma->bd[i];
+ int param;
+
+ bd->buffer_addr = dma_addr;
+
+ bd->mode.count = period_len;
+
+ if (sdma->word_size > 4)
+ goto err_out;
+ if (sdma->word_size == 4)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdma->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
+ if (i + 1 == num_periods)
+ param |= BD_WRAP;
+
+ pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, period_len, dma_addr,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+
+ dma_addr += period_len;
+ buf += period_len;
+
+ i++;
+ }
+
+ sdma->num_bd = num_periods;
+ channel_control[channel].current_bd_ptr = sdma->bd_phys;
+
+ return &sdma->desc;
+err_out:
+ sdma->busy = 0;
+ return NULL;
+}
+
+static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ struct dma_slave_config *dmaengine_cfg = (void *)arg;
+
+ switch (cmd) {
+ case DMA_TERMINATE_ALL:
+ sdma_disable_channel(chan->chan_id);
+ return 0;
+ case DMA_SLAVE_CONFIG:
+ if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
+ sdma->per_address = dmaengine_cfg->src_addr;
+ sdma->watermark_level = dmaengine_cfg->src_maxburst;
+ sdma->word_size = dmaengine_cfg->src_addr_width;
+ } else {
+ sdma->per_address = dmaengine_cfg->dst_addr;
+ sdma->watermark_level = dmaengine_cfg->dst_maxburst;
+ sdma->word_size = dmaengine_cfg->dst_addr_width;
+ }
+ return sdma_config_channel(chan->chan_id);
+ default:
+ return -ENOSYS;
+ }
+
+ return -EINVAL;
+}
+
+static enum dma_status sdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ dma_cookie_t last_used;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+
+ ret = dma_async_is_complete(cookie, sdma->last_completed, last_used);
+ dma_set_tx_state(txstate, sdma->last_completed, last_used, 0);
+
+ return ret;
+}
+
+static void sdma_issue_pending(struct dma_chan *chan)
+{
+ /*
+ * Nothing to do. We only have a single descriptor
+ */
+}
+
+static int __devinit sdma_probe(struct platform_device *pdev)
+{
+ int ret;
+ const struct firmware *fw;
+ const struct sdma_firmware_header *header;
+ const struct sdma_script_start_addrs *addr;
+ int irq;
+ unsigned short *ram_code;
+ struct resource *iores;
+ struct sdma_platform_data *pdata = pdev->dev.platform_data;
+ int version;
+ char *cpustr, *fwname;
+ int i;
+ dma_cap_mask_t mask;
+
+ /* there can be only one */
+ BUG_ON(sdma_base);
+
+ iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ irq = platform_get_irq(pdev, 0);
+ if (!iores || irq < 0 || !pdata)
+ return -EINVAL;
+
+ sdma_clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(sdma_clk)) {
+ ret = PTR_ERR(sdma_clk);
+ goto err_clk;
+ }
+
+ if (cpu_is_mx31()) {
+ cpustr = "imx31";
+ version = mx31_revision() >> 4;
+ } else if (cpu_is_mx35()) {
+ cpustr = "imx35";
+/* FIXME: version = mx35_revision(); */
+ version = 2;
+ } else {
+ ret = -EINVAL;
+ goto err_cputype;
+ }
+
+ fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin", cpustr, version);
+ if (!fwname) {
+ ret = -ENOMEM;
+ goto err_cputype;
+ }
+
+ ret = request_firmware(&fw, fwname, &pdev->dev);
+ if (ret) {
+ dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
+ fwname, ret);
+ kfree(fwname);
+ goto err_cputype;
+ }
+ kfree(fwname);
+
+ if (fw->size < sizeof(*header))
+ goto err_firmware;
+
+ header = (struct sdma_firmware_header *)fw->data;
+
+ if (header->magic != SDMA_FIRMWARE_MAGIC)
+ goto err_firmware;
+ if (header->ram_code_start + header->ram_code_size > fw->size)
+ goto err_firmware;
+
+ addr = (void *)header + header->script_addrs_start;
+ ram_code = (void *)header + header->ram_code_start;
+ memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
+
+ ret = sdma_init(iores->start, irq, pdata->sdma_version,
+ ram_code, header->ram_code_size);
+ if (ret)
+ goto err_firmware;
+
+ INIT_LIST_HEAD(&sdma_dma_device->channels);
+
+ /* Initialize channel parameters */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++) {
+ struct sdma_channel *sdma = &sdma_data[i];
+
+ spin_lock_init(&sdma->lock);
+
+ dma_cap_set(DMA_SLAVE, sdma_dma_device->cap_mask);
+ dma_cap_set(DMA_CYCLIC, sdma_dma_device->cap_mask);
+
+ sdma->chan.device = sdma_dma_device;
+ sdma->chan.chan_id = i;
+
+ /* Add the channel to the DMAC list */
+ list_add_tail(&sdma->chan.device_node, &sdma_dma_device->channels);
+ }
+
+ sdma_dma_device->dev = &pdev->dev;
+
+ sdma_dma_device->device_alloc_chan_resources = sdma_alloc_chan_resources;
+ sdma_dma_device->device_free_chan_resources = sdma_free_chan_resources;
+ sdma_dma_device->device_tx_status = sdma_tx_status;
+ sdma_dma_device->device_prep_slave_sg = sdma_prep_slave_sg;
+ sdma_dma_device->device_prep_dma_cyclic = sdma_prep_dma_cyclic;
+ sdma_dma_device->device_control = sdma_control;
+ sdma_dma_device->device_issue_pending = sdma_issue_pending;
+
+ ret = dma_async_device_register(sdma_dma_device);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register DMAC\n");
+ goto err_firmware;
+ }
+
+ dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
+ header->version_major,
+ header->version_minor);
+
+ /* request channel 0. This is an internal control channel
+ * to the SDMA engine and not available to clients.
+ */
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_request_channel(mask, NULL, NULL);
+
+ release_firmware(fw);
+
+ return 0;
+
+err_firmware:
+ release_firmware(fw);
+err_cputype:
+ clk_put(sdma_clk);
+err_clk:
+ return 0;
+}
+
+static int __devexit sdma_remove(struct platform_device *pdev)
+{
+ return -EBUSY;
+}
+
+static struct platform_driver sdma_driver = {
+ .driver = {
+ .name = "imx-sdma",
+ },
+ .probe = sdma_probe,
+ .remove = __devexit_p(sdma_remove),
+};
+
+static int __init sdma_module_init(void)
+{
+ return platform_driver_register(&sdma_driver);
+}
+subsys_initcall(sdma_module_init);
+
+MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("i.MX SDMA driver");
+MODULE_LICENSE("GPL");
--
1.7.1
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-16 11:07 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 11:07 UTC (permalink / raw)
To: linux-arm-kernel
This patch adds support for the Freescale i.MX SDMA engine.
The SDMA engine is a scatter/gather DMA engine which is implemented
as a seperate coprocessor. SDMA needs its own firmware which is
requested using the standard request_firmware mechanism. The firmware
has different entry points for each peripheral type, so drivers
have to pass the peripheral type to the DMA engine which in turn
picks the correct firmware entry point from a table contained in
the firmware image itself.
The original Freescale code also supports support for transfering
data to the internal SRAM which needs different entry points to
the firmware. Support for this is currently not implemented. Also,
support for the ASRC (asymmetric sample rate converter) is skipped.
I took a very simple approach to implement dmaengine support. Only
a single descriptor is statically assigned to a each channel. This
means that transfers can't be queued up but only a single transfer
is in progress. This simplifies implementation a lot and is sufficient
for the usual device/memory transfers.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
arch/arm/mach-mx3/Kconfig | 2 +
arch/arm/plat-mxc/Kconfig | 10 +
arch/arm/plat-mxc/include/mach/dma.h | 64 ++
arch/arm/plat-mxc/include/mach/sdma.h | 8 +
drivers/dma/Kconfig | 8 +
drivers/dma/Makefile | 1 +
drivers/dma/imx-sdma.c | 1383 +++++++++++++++++++++++++++++++
8 files changed, 1478 insertions(+), 6 deletions(-)
create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
create mode 100644 drivers/dma/imx-sdma.c
diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
index 287431c..ac6fd71 100644
--- a/arch/arm/mach-imx/include/mach/dma-v1.h
+++ b/arch/arm/mach-imx/include/mach/dma-v1.h
@@ -27,6 +27,8 @@
#define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
+#include <mach/dma.h>
+
#define IMX_DMA_CHANNELS 16
#define DMA_MODE_READ 0
@@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
void imx_dma_free(int channel);
-enum imx_dma_prio {
- DMA_PRIO_HIGH = 0,
- DMA_PRIO_MEDIUM = 1,
- DMA_PRIO_LOW = 2
-};
-
int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
#endif /* __MACH_DMA_V1_H__ */
diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
index 85beece..301375c 100644
--- a/arch/arm/mach-mx3/Kconfig
+++ b/arch/arm/mach-mx3/Kconfig
@@ -3,12 +3,14 @@ if ARCH_MX3
config ARCH_MX31
select ARCH_HAS_RNGA
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
bool
config ARCH_MX35
bool
select ARCH_MXC_IOMUX_V3
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
comment "MX3 platforms:"
diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
index 0527e65..6741625 100644
--- a/arch/arm/plat-mxc/Kconfig
+++ b/arch/arm/plat-mxc/Kconfig
@@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
config ARCH_MXC_AUDMUX_V2
bool
+config IMX_HAVE_SDMA
+ bool
+
+config IMX_SDMA
+ depends on IMX_HAVE_SDMA
+ tristate "Enable SDMA support"
+ help
+ Include support for the SDMA engine. The SDMA engine needs additional
+ firmware support. SDMA can be compiled as a module to support loading
+ the firmware when a rootfs is present.
endif
diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
new file mode 100644
index 0000000..94cab29
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/dma.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARCH_MXC_DMA_H__
+#define __ASM_ARCH_MXC_DMA_H__
+
+#include <linux/scatterlist.h>
+
+/*
+ * This enumerates peripheral types. Used for SDMA.
+ */
+typedef enum {
+ IMX_DMATYPE_SSI, /* MCU domain SSI */
+ IMX_DMATYPE_SSI_SP, /* Shared SSI */
+ IMX_DMATYPE_MMC, /* MMC */
+ IMX_DMATYPE_SDHC, /* SDHC */
+ IMX_DMATYPE_UART, /* MCU domain UART */
+ IMX_DMATYPE_UART_SP, /* Shared UART */
+ IMX_DMATYPE_FIRI, /* FIRI */
+ IMX_DMATYPE_CSPI, /* MCU domain CSPI */
+ IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
+ IMX_DMATYPE_SIM, /* SIM */
+ IMX_DMATYPE_ATA, /* ATA */
+ IMX_DMATYPE_CCM, /* CCM */
+ IMX_DMATYPE_EXT, /* External peripheral */
+ IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
+ IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
+ IMX_DMATYPE_DSP, /* DSP */
+ IMX_DMATYPE_MEMORY, /* Memory */
+ IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
+ IMX_DMATYPE_SPDIF, /* SPDIF */
+ IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
+ IMX_DMATYPE_ASRC, /* ASRC */
+ IMX_DMATYPE_ESAI, /* ESAI */
+} sdma_peripheral_type;
+
+enum imx_dma_prio {
+ DMA_PRIO_HIGH = 0,
+ DMA_PRIO_MEDIUM = 1,
+ DMA_PRIO_LOW = 2
+};
+
+struct imx_dma_data {
+ int dma_request; /* DMA request line */
+ sdma_peripheral_type peripheral_type;
+ int priority;
+};
+
+static inline int imx_dma_is_ipu(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "ipu-core");
+}
+
+static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "imx-sdma");
+}
+
+#endif
diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
new file mode 100644
index 0000000..5d542b8
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/sdma.h
@@ -0,0 +1,8 @@
+#ifndef __MACH_MXC_SDMA_H__
+#define __MACH_MXC_SDMA_H__
+
+struct sdma_platform_data {
+ int sdma_version;
+};
+
+#endif /* __MACH_MXC_SDMA_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9520cf0..f76bda9 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -195,6 +195,14 @@ config PCH_DMA
help
Enable support for the Topcliff PCH DMA engine.
+config IMX_SDMA
+ tristate "Atmel AHB DMA support"
+ depends on ARCH_MXC
+ select DMA_ENGINE
+ help
+ Support the i.MX SDMA engine. This engine is integrated into
+ Freescale i.MX25/31/35/51 chips.
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 72bd703..14d7a1b 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
obj-$(CONFIG_PCH_DMA) += pch_dma.o
+obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
new file mode 100644
index 0000000..3ba7905
--- /dev/null
+++ b/drivers/dma/imx-sdma.c
@@ -0,0 +1,1383 @@
+/*
+ * drivers/dma/imx-sdma.c
+ *
+ * This file contains a driver for the Freescale Smart DMA engine
+ *
+ * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
+ *
+ * Based on code from Freescale:
+ *
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * The code contained herein is licensed under the GNU General Public
+ * License. You may obtain a copy of the GNU General Public License
+ * Version 2 or later at the following locations:
+ *
+ * http://www.opensource.org/licenses/gpl-license.html
+ * http://www.gnu.org/copyleft/gpl.html
+ */
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <linux/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/dmaengine.h>
+
+#include <asm/irq.h>
+#include <mach/sdma.h>
+#include <mach/dma.h>
+#include <mach/hardware.h>
+
+/* SDMA registers */
+#define SDMA_H_C0PTR (sdma_base + 0x000)
+#define SDMA_H_INTR (sdma_base + 0x004)
+#define SDMA_H_STATSTOP (sdma_base + 0x008)
+#define SDMA_H_START (sdma_base + 0x00c)
+#define SDMA_H_EVTOVR (sdma_base + 0x010)
+#define SDMA_H_DSPOVR (sdma_base + 0x014)
+#define SDMA_H_HOSTOVR (sdma_base + 0x018)
+#define SDMA_H_EVTPEND (sdma_base + 0x01c)
+#define SDMA_H_DSPENBL (sdma_base + 0x020)
+#define SDMA_H_RESET (sdma_base + 0x024)
+#define SDMA_H_EVTERR (sdma_base + 0x028)
+#define SDMA_H_INTRMSK (sdma_base + 0x02c)
+#define SDMA_H_PSW (sdma_base + 0x030)
+#define SDMA_H_EVTERRDBG (sdma_base + 0x034)
+#define SDMA_H_CONFIG (sdma_base + 0x038)
+#define SDMA_ONCE_ENB (sdma_base + 0x040)
+#define SDMA_ONCE_DATA (sdma_base + 0x044)
+#define SDMA_ONCE_INSTR (sdma_base + 0x048)
+#define SDMA_ONCE_STAT (sdma_base + 0x04c)
+#define SDMA_ONCE_CMD (sdma_base + 0x050)
+#define SDMA_EVT_MIRROR (sdma_base + 0x054)
+#define SDMA_ILLINSTADDR (sdma_base + 0x058)
+#define SDMA_CHN0ADDR (sdma_base + 0x05c)
+#define SDMA_ONCE_RTB (sdma_base + 0x060)
+#define SDMA_XTRIG_CONF1 (sdma_base + 0x070)
+#define SDMA_XTRIG_CONF2 (sdma_base + 0x074)
+#define SDMA_CHNENBL_0 (sdma_base + (sdma_version == 2 ? 0x200 : 0x80))
+#define SDMA_CHNPRI_0 (sdma_base + 0x100)
+
+/*
+ * Buffer descriptor status values.
+ */
+#define BD_DONE 0x01
+#define BD_WRAP 0x02
+#define BD_CONT 0x04
+#define BD_INTR 0x08
+#define BD_RROR 0x10
+#define BD_LAST 0x20
+#define BD_EXTD 0x80
+
+/*
+ * Data Node descriptor status values.
+ */
+#define DND_END_OF_FRAME 0x80
+#define DND_END_OF_XFER 0x40
+#define DND_DONE 0x20
+#define DND_UNUSED 0x01
+
+/*
+ * IPCV2 descriptor status values.
+ */
+#define BD_IPCV2_END_OF_FRAME 0x40
+
+#define IPCV2_MAX_NODES 50
+/*
+ * Error bit set in the CCB status field by the SDMA,
+ * in setbd routine, in case of a transfer error
+ */
+#define DATA_ERROR 0x10000000
+
+/*
+ * Buffer descriptor commands.
+ */
+#define C0_ADDR 0x01
+#define C0_LOAD 0x02
+#define C0_DUMP 0x03
+#define C0_SETCTX 0x07
+#define C0_GETCTX 0x03
+#define C0_SETDM 0x01
+#define C0_SETPM 0x04
+#define C0_GETDM 0x02
+#define C0_GETPM 0x08
+/*
+ * Change endianness indicator in the BD command field
+ */
+#define CHANGE_ENDIANNESS 0x80
+
+/*
+ * Mode/Count of data node descriptors - IPCv2
+ */
+#ifdef __BIG_ENDIAN
+struct sdma_mode_count {
+ u32 command : 8; /* command mostlky used for channel 0 */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 count : 16; /* size of the buffer pointed by this BD */
+};
+#else
+struct sdma_mode_count {
+ u32 count : 16; /* size of the buffer pointed by this BD */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 command : 8; /* command mostlky used for channel 0 */
+};
+#endif
+
+/*
+ * Buffer descriptor
+ */
+struct sdma_buffer_descriptor {
+ struct sdma_mode_count mode;
+ u32 buffer_addr; /* address of the buffer described */
+ u32 ext_buffer_addr; /* extended buffer address */
+};
+
+/*
+ * Channel control Block
+ */
+struct sdma_channel_control {
+ u32 current_bd_ptr; /* current buffer descriptor processed */
+ u32 base_bd_ptr; /* first element of buffer descriptor array */
+ void *unused;
+ void *unused1;
+};
+
+/**
+ * Context structure.
+ */
+#ifdef __BIG_ENDIAN
+struct sdma_state_registers {
+ u32 sf : 1; /* source fault while loading data */
+ u32 unused0: 1;
+ u32 rpc :14; /* return program counter */
+ u32 t : 1; /* test bit:status of arithmetic & test instruction*/
+ u32 unused1: 1;
+ u32 pc :14; /* program counter */
+ u32 lm : 2; /* loop mode */
+ u32 epc :14; /* loop end program counter */
+ u32 df : 1; /* destination fault while storing data */
+ u32 unused2: 1;
+ u32 spc :14; /* loop start program counter */
+};
+#else
+struct sdma_state_registers {
+ u32 pc :14; /* program counter */
+ u32 unused1: 1;
+ u32 t : 1; /* test bit: status of arithmetic & test instruction*/
+ u32 rpc :14; /* return program counter */
+ u32 unused0: 1;
+ u32 sf : 1; /* source fault while loading data */
+ u32 spc :14; /* loop start program counter */
+ u32 unused2: 1;
+ u32 df : 1; /* destination fault while storing data */
+ u32 epc :14; /* loop end program counter */
+ u32 lm : 2; /* loop mode */
+};
+#endif
+
+struct sdma_context_data {
+ struct sdma_state_registers channel_state; /* channel state bits */
+ u32 gReg[8]; /* general registers */
+ u32 mda; /* burst dma destination address register */
+ u32 msa; /* burst dma source address register */
+ u32 ms; /* burst dma status register */
+ u32 md; /* burst dma data register */
+ u32 pda; /* peripheral dma destination address register */
+ u32 psa; /* peripheral dma source address register */
+ u32 ps; /* peripheral dma status register */
+ u32 pd; /* peripheral dma data register */
+ u32 ca; /* CRC polynomial register */
+ u32 cs; /* CRC accumulator register */
+ u32 dda; /* dedicated core destination address register */
+ u32 dsa; /* dedicated core source address register */
+ u32 ds; /* dedicated core status register */
+ u32 dd; /* dedicated core data register */
+ u32 scratch0;
+ u32 scratch1;
+ u32 scratch2;
+ u32 scratch3;
+ u32 scratch4;
+ u32 scratch5;
+ u32 scratch6;
+ u32 scratch7;
+};
+
+struct sdma_channel {
+ /* Channel number */
+ int channel;
+ /* Transfer type. Needed for setting SDMA script */
+ enum dma_data_direction direction;
+ /* Peripheral type. Needed for setting SDMA script */
+ sdma_peripheral_type peripheral_type;
+ /* Peripheral event id */
+ int event_id;
+ /* Peripheral event id2 (for channels that use 2 events) */
+ int event_id2;
+ /* SDMA data access word size */
+ unsigned long word_size;
+
+ /* ID of the buffer that was processed */
+ unsigned int buf_tail;
+
+ wait_queue_head_t waitq; /* channel completion waitqeue */
+
+ int num_bd;
+
+ struct sdma_buffer_descriptor *bd;
+ dma_addr_t bd_phys;
+
+ int pc_from_device, pc_to_device;
+
+ unsigned long flags;
+ dma_addr_t per_address;
+
+ uint32_t event_mask1, event_mask2;
+ uint32_t watermark_level;
+ uint32_t shp_addr, per_addr;
+
+ /* DMA-Engine Channel */
+ struct dma_chan chan;
+
+ spinlock_t lock;
+ struct dma_async_tx_descriptor desc;
+ dma_cookie_t last_completed;
+ int busy;
+};
+
+#define IMX_DMA_SG_LOOP (1 << 0)
+
+#define MAX_DMA_CHANNELS 32
+#define MXC_SDMA_DEFAULT_PRIORITY 1
+#define MXC_SDMA_MIN_PRIORITY 1
+#define MXC_SDMA_MAX_PRIORITY 7
+
+/*
+ * This enumerates transfer types
+ */
+typedef enum {
+ emi_2_per = 0, /* EMI memory to peripheral */
+ emi_2_int, /* EMI memory to internal RAM */
+ emi_2_emi, /* EMI memory to EMI memory */
+ emi_2_dsp, /* EMI memory to DSP memory */
+ per_2_int, /* Peripheral to internal RAM */
+ per_2_emi, /* Peripheral to internal EMI memory */
+ per_2_dsp, /* Peripheral to DSP memory */
+ per_2_per, /* Peripheral to Peripheral */
+ int_2_per, /* Internal RAM to peripheral */
+ int_2_int, /* Internal RAM to Internal RAM */
+ int_2_emi, /* Internal RAM to EMI memory */
+ int_2_dsp, /* Internal RAM to DSP memory */
+ dsp_2_per, /* DSP memory to peripheral */
+ dsp_2_int, /* DSP memory to internal RAM */
+ dsp_2_emi, /* DSP memory to EMI memory */
+ dsp_2_dsp, /* DSP memory to DSP memory */
+ emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
+ dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
+ dvfs_pll, /* DVFS script with PLL change */
+ dvfs_pdr /* DVFS script without PLL change */
+} sdma_transfer_type;
+
+/*
+ * Structure containing sdma request parameters.
+ */
+struct sdma_script_start_addrs {
+ int ap_2_ap_addr;
+ int ap_2_bp_addr;
+ int ap_2_ap_fixed_addr;
+ int bp_2_ap_addr;
+ int loopback_on_dsp_side_addr;
+ int mcu_interrupt_only_addr;
+
+ int firi_2_per_addr;
+ int firi_2_mcu_addr;
+ int per_2_firi_addr;
+ int mcu_2_firi_addr;
+
+ int uart_2_per_addr;
+ int uart_2_mcu_addr;
+ int per_2_app_addr;
+ int mcu_2_app_addr;
+ int per_2_per_addr;
+
+ int uartsh_2_per_addr;
+ int uartsh_2_mcu_addr;
+ int per_2_shp_addr;
+ int mcu_2_shp_addr;
+
+ int ata_2_mcu_addr;
+ int mcu_2_ata_addr;
+
+ int app_2_per_addr;
+ int app_2_mcu_addr;
+ int shp_2_per_addr;
+ int shp_2_mcu_addr;
+
+ int mshc_2_mcu_addr;
+ int mcu_2_mshc_addr;
+
+ int spdif_2_mcu_addr;
+ int mcu_2_spdif_addr;
+
+ int asrc_2_mcu_addr;
+
+ int ext_mem_2_ipu_addr;
+
+ int descrambler_addr;
+
+ int dptc_dvfs_addr;
+
+ int utra_addr;
+
+ int ram_code_start_addr;
+};
+
+#define SDMA_FIRMWARE_MAGIC 0x414d4453
+
+struct sdma_firmware_header {
+ uint32_t magic; /* "SDMA" */
+ uint32_t version_major; /* increased whenever layout of struct sdma_script_start_addrs changes */
+ uint32_t version_minor; /* firmware version */
+ uint32_t script_addrs_start; /* offset of struct sdma_script_start_addrs in this image */
+ uint32_t num_script_addrs; /* Number of script addresses in this image */
+ uint32_t ram_code_start; /* offset of SDMA ram image in this firmware image */
+ uint32_t ram_code_size; /* size of SDMA ram image */
+};
+
+static struct sdma_channel sdma_data[MAX_DMA_CHANNELS];
+static struct sdma_channel_control *channel_control;
+static void __iomem *sdma_base;
+static int sdma_version;
+static int sdma_num_events;
+static struct sdma_context_data *sdma_context;
+dma_addr_t sdma_context_phys;
+static struct dma_device __sdma_dma_device;
+static struct dma_device *sdma_dma_device = &__sdma_dma_device;
+
+#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
+#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
+#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
+#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
+
+static int sdma_config_ownership(int channel, int event_override,
+ int mcu_verride, int dsp_override)
+{
+ u32 evt, mcu, dsp;
+
+ if (event_override && mcu_verride && dsp_override)
+ return -EINVAL;
+
+ evt = readl(SDMA_H_EVTOVR);
+ mcu = readl(SDMA_H_HOSTOVR);
+ dsp = readl(SDMA_H_DSPOVR);
+
+ if (dsp_override)
+ dsp &= ~(1 << channel);
+ else
+ dsp |= (1 << channel);
+
+ if (event_override)
+ evt &= ~(1 << channel);
+ else
+ evt |= (1 << channel);
+
+ if (mcu_verride)
+ mcu &= ~(1 << channel);
+ else
+ mcu |= (1 << channel);
+
+ writel(evt, SDMA_H_EVTOVR);
+ writel(mcu, SDMA_H_HOSTOVR);
+ writel(dsp, SDMA_H_DSPOVR);
+
+ return 0;
+}
+
+/*
+ * sdma_run_channel - run a channel and wait till it's done
+ */
+static int sdma_run_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ int ret;
+
+ writel(1 << channel, SDMA_H_START);
+
+ ret = wait_event_interruptible(sdma->waitq,
+ !(readl(SDMA_H_STATSTOP) & (1 << channel)));
+ return ret;
+}
+
+static int sdma_load_script(void *buf, int size, u32 address)
+{
+ struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
+ void *buf_virt;
+ dma_addr_t buf_phys;
+ int ret;
+
+ buf_virt = dma_alloc_coherent(NULL,
+ size,
+ &buf_phys, GFP_KERNEL);
+ if (!buf_virt)
+ return -ENOMEM;
+
+ bd0->mode.command = C0_SETPM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = size / 2;
+ bd0->buffer_addr = buf_phys;
+ bd0->ext_buffer_addr = address;
+
+ memcpy(buf_virt, buf, size);
+
+ ret = sdma_run_channel(0);
+
+ dma_free_coherent(NULL, size, buf_virt, buf_phys);
+
+ return ret;
+}
+
+static void sdma_event_enable(int channel, int event)
+{
+ u32 val;
+
+ val = readl(SDMA_CHNENBL_0 + event * 4);
+ val |= (1 << channel);
+ writel(val, SDMA_CHNENBL_0 + event * 4);
+}
+
+static void sdma_event_disable(int channel, int event)
+{
+ u32 val;
+
+ val = readl(SDMA_CHNENBL_0 + event * 4);
+ val &= ~(1 << channel);
+ writel(val, SDMA_CHNENBL_0 + event * 4);
+}
+
+static void mxc_sdma_handle_channel_loop(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ struct sdma_buffer_descriptor *bd;
+ int error = 0;
+
+ /*
+ * loop mode. Iterate over descriptors, re-setup them and
+ * call callback function.
+ */
+ while (1) {
+ bd = &sdma->bd[sdma->buf_tail];
+
+ if (bd->mode.status & BD_DONE)
+ break;
+
+ if (bd->mode.status & BD_RROR)
+ error = -EIO;
+
+ bd->mode.status |= BD_DONE;
+ sdma->buf_tail++;
+ sdma->buf_tail %= sdma->num_bd;
+
+ if (sdma->desc.callback)
+ sdma->desc.callback(sdma->desc.callback_param);
+ }
+}
+
+static void mxc_sdma_handle_channel_normal(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ struct sdma_buffer_descriptor *bd;
+ int i, error = 0;
+
+ /*
+ * non loop mode. Iterate over all descriptors, collect
+ * errors and call callback function
+ */
+ for (i = 0; i < sdma->num_bd; i++) {
+ bd = &sdma->bd[i];
+
+ if (bd->mode.status & (BD_DONE | BD_RROR))
+ error = -EIO;
+ }
+
+ if (sdma->desc.callback)
+ sdma->desc.callback(sdma->desc.callback_param);
+ sdma->last_completed = sdma->desc.cookie;
+
+ sdma->busy = 0;
+}
+
+static void mxc_sdma_handle_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+
+ wake_up_interruptible(&sdma->waitq);
+
+ /* not interested in channel 0 interrupts */
+ if (!channel)
+ return;
+
+ if (sdma->flags & IMX_DMA_SG_LOOP)
+ mxc_sdma_handle_channel_loop(channel);
+ else
+ mxc_sdma_handle_channel_normal(channel);
+}
+
+static irqreturn_t sdma_int_handler(int irq, void *dev_id)
+{
+ u32 stat;
+
+ stat = readl(SDMA_H_INTR);
+ writel(stat, SDMA_H_INTR);
+
+ while (stat) {
+ int channel = fls(stat) - 1;
+
+ mxc_sdma_handle_channel(channel);
+
+ stat &= ~(1 << channel);
+ }
+
+ return IRQ_HANDLED;
+}
+
+static struct clk *sdma_clk;
+
+/*
+ * Stores the start address of the SDMA scripts
+ */
+static struct sdma_script_start_addrs __sdma_script_addrs;
+static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
+
+/*
+ * sets the pc of SDMA script according to the peripheral type
+ */
+static void sdma_get_pc(struct sdma_channel *sdma,
+ sdma_peripheral_type peripheral_type)
+{
+ int res = 0;
+ int per_2_emi = 0, emi_2_per = 0;
+ int per_2_int = 0, int_2_per = 0;
+ int per_2_per = 0, emi_2_emi = 0;
+
+ sdma->pc_from_device = 0;
+ sdma->pc_to_device = 0;
+
+ switch (peripheral_type) {
+ case IMX_DMATYPE_MEMORY:
+ emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
+ break;
+ case IMX_DMATYPE_DSP:
+ emi_2_per = sdma_script_addrs->bp_2_ap_addr;
+ per_2_emi = sdma_script_addrs->ap_2_bp_addr;
+ break;
+ case IMX_DMATYPE_FIRI:
+ per_2_int = sdma_script_addrs->firi_2_per_addr;
+ per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_firi_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
+ break;
+ case IMX_DMATYPE_UART:
+ per_2_int = sdma_script_addrs->uart_2_per_addr;
+ per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_app_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_UART_SP:
+ per_2_int = sdma_script_addrs->uartsh_2_per_addr;
+ per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_shp_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ATA:
+ per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
+ break;
+ case IMX_DMATYPE_CSPI:
+ case IMX_DMATYPE_EXT:
+ case IMX_DMATYPE_SSI:
+ per_2_int = sdma_script_addrs->app_2_per_addr;
+ per_2_emi = sdma_script_addrs->app_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_app_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_SSI_SP:
+ case IMX_DMATYPE_MMC:
+ case IMX_DMATYPE_SDHC:
+ case IMX_DMATYPE_CSPI_SP:
+ case IMX_DMATYPE_ESAI:
+ case IMX_DMATYPE_MSHC_SP:
+ per_2_int = sdma_script_addrs->shp_2_per_addr;
+ per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
+ int_2_per = sdma_script_addrs->per_2_shp_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ASRC:
+ per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
+ per_2_per = sdma_script_addrs->per_2_per_addr;
+ break;
+ case IMX_DMATYPE_MSHC:
+ per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
+ break;
+ case IMX_DMATYPE_CCM:
+ per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
+ break;
+ case IMX_DMATYPE_FIFO_MEMORY:
+ res = sdma_script_addrs->ap_2_ap_fixed_addr;
+ break;
+ case IMX_DMATYPE_SPDIF:
+ per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
+ break;
+ case IMX_DMATYPE_IPU_MEMORY:
+ emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
+ break;
+ default:
+ break;
+ }
+
+ sdma->pc_from_device = per_2_emi;
+ sdma->pc_to_device = emi_2_per;
+}
+
+static int sdma_load_context(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ int load_address;
+ struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
+ int ret;
+
+ if (sdma->direction == DMA_FROM_DEVICE) {
+ load_address = sdma->pc_from_device;
+ } else {
+ load_address = sdma->pc_to_device;
+ }
+
+ if (load_address < 0)
+ return load_address;
+
+ pr_debug("%s: load_address = %d\n", __func__, load_address);
+ pr_debug("%s: wml = 0x%08x\n", __func__, sdma->watermark_level);
+ pr_debug("%s: shp_addr = 0x%08x\n", __func__, sdma->shp_addr);
+ pr_debug("%s: per_addr = 0x%08x\n", __func__, sdma->per_addr);
+ pr_debug("%s: event_mask1 = 0x%08x\n", __func__, sdma->event_mask1);
+ pr_debug("%s: event_mask2 = 0x%08x\n", __func__, sdma->event_mask2);
+
+ memset(sdma_context, 0, sizeof(*sdma_context));
+ sdma_context->channel_state.pc = load_address;
+
+ /* Send by context the event mask,base address for peripheral
+ * and watermark level
+ */
+ sdma_context->gReg[0] = sdma->event_mask2;
+ sdma_context->gReg[1] = sdma->event_mask1;
+ sdma_context->gReg[2] = sdma->per_addr;
+ sdma_context->gReg[6] = sdma->shp_addr;
+ sdma_context->gReg[7] = sdma->watermark_level;
+
+ bd0->mode.command = C0_SETDM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = sizeof(*sdma_context) / 4;
+ bd0->buffer_addr = sdma_context_phys;
+ bd0->ext_buffer_addr = 2048 + (sizeof(*sdma_context) / 4) * channel;
+
+ ret = sdma_run_channel(0);
+
+ return ret;
+}
+
+static void sdma_disable_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+
+ writel(1 << channel, SDMA_H_STATSTOP);
+ sdma->busy = 0;
+}
+
+static int sdma_config_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ int ret;
+
+ sdma_disable_channel(channel);
+
+ sdma->event_mask1 = 0;
+ sdma->event_mask2 = 0;
+ sdma->shp_addr = 0;
+ sdma->per_addr = 0;
+
+ if (sdma->event_id)
+ sdma_event_enable(channel, sdma->event_id);
+
+ switch (sdma->peripheral_type) {
+ case IMX_DMATYPE_DSP:
+ sdma_config_ownership(channel, 0, 1, 1);
+ break;
+ case IMX_DMATYPE_MEMORY:
+ sdma_config_ownership(channel, 0, 1, 0);
+ break;
+ default:
+ sdma_config_ownership(channel, 1, 1, 0);
+ break;
+ }
+
+ sdma_get_pc(sdma, sdma->peripheral_type);
+
+ if ((sdma->peripheral_type != IMX_DMATYPE_MEMORY) &&
+ (sdma->peripheral_type != IMX_DMATYPE_DSP)) {
+ /* Handle multiple event channels differently */
+ if (sdma->event_id2) {
+ sdma->event_mask2 = 1 << (sdma->event_id2 % 32);
+ if (sdma->event_id2 > 31)
+ sdma->watermark_level |= 1 << 31;
+ sdma->event_mask1 = 1 << (sdma->event_id % 32);
+ if (sdma->event_id > 31)
+ sdma->watermark_level |= 1 << 30;
+ } else {
+ sdma->event_mask1 = 1 << sdma->event_id;
+ sdma->event_mask2 = 1 << (sdma->event_id - 32);
+ }
+ /* Watermark Level */
+ sdma->watermark_level |= sdma->watermark_level;
+ /* Address */
+ sdma->shp_addr = sdma->per_address;
+ } else {
+ sdma->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
+ }
+
+ ret = sdma_load_context(channel);
+
+ return ret;
+}
+
+static int sdma_set_channel_priority(unsigned int channel, unsigned int priority)
+{
+ if (priority < MXC_SDMA_MIN_PRIORITY
+ || priority > MXC_SDMA_MAX_PRIORITY) {
+ return -EINVAL;
+ }
+
+ writel(priority, SDMA_CHNPRI_0 + 4 * channel);
+
+ return 0;
+}
+
+static int sdma_request_channel(int channel)
+{
+ struct sdma_channel *sdma = &sdma_data[channel];
+ int ret = -EBUSY;
+
+ sdma->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdma->bd_phys, GFP_KERNEL);
+ if (!sdma->bd) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memset(sdma->bd, 0, PAGE_SIZE);
+
+ channel_control[channel].base_bd_ptr = sdma->bd_phys;
+ channel_control[channel].current_bd_ptr = sdma->bd_phys;
+
+ clk_enable(sdma_clk);
+
+ sdma_set_channel_priority(channel, MXC_SDMA_DEFAULT_PRIORITY);
+
+ init_waitqueue_head(&sdma->waitq);
+
+ sdma->buf_tail = 0;
+
+ return 0;
+out:
+
+ return ret;
+}
+
+static void sdma_enable_channel(int channel)
+{
+ writel(1 << channel, SDMA_H_START);
+}
+
+static int __init sdma_init(unsigned long phys_base, int irq, int version,
+ void *ram_code,
+ int ram_code_size)
+{
+ int i, ret;
+ int channel;
+ dma_addr_t ccb_phys;
+
+ sdma_version = version;
+ switch (sdma_version) {
+ case 1:
+ sdma_num_events = 32;
+ break;
+ case 2:
+ sdma_num_events = 48;
+ break;
+ default:
+ pr_err("SDMA: Unknown version %d. aborting\n", sdma_version);
+ return -ENODEV;
+ }
+
+ clk_enable(sdma_clk);
+
+ sdma_base = ioremap(phys_base, 4096);
+ if (!sdma_base) {
+ ret = -ENOMEM;
+ goto err_ioremap;
+ }
+
+ /* Initialize SDMA private data */
+ memset(sdma_data, 0, sizeof(struct sdma_channel) * MAX_DMA_CHANNELS);
+
+ for (channel = 0; channel < MAX_DMA_CHANNELS; channel++)
+ sdma_data[channel].channel = channel;
+
+ ret = request_irq(irq, sdma_int_handler, 0, "sdma", NULL);
+ if (ret)
+ goto err_request_irq;
+
+ /* Be sure SDMA has not started yet */
+ writel(0, SDMA_H_C0PTR);
+
+ channel_control = dma_alloc_coherent(NULL,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
+ sizeof(struct sdma_context_data),
+ &ccb_phys, GFP_KERNEL);
+
+ if (!channel_control) {
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ sdma_context = (void *)channel_control +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+ sdma_context_phys = ccb_phys +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+
+ /* Zero-out the CCB structures array just allocated */
+ memset(channel_control, 0,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
+
+ /* disable all channels */
+ for (i = 0; i < sdma_num_events; i++)
+ writel(0, SDMA_CHNENBL_0 + i * 4);
+
+ /* All channels have priority 0 */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++)
+ writel(0, SDMA_CHNPRI_0 + i * 4);
+
+ ret = sdma_request_channel(0);
+ if (ret)
+ goto err_dma_alloc;
+
+ sdma_config_ownership(0, 0, 1, 0);
+
+ /* Set Command Channel (Channel Zero) */
+ writel(0x4050, SDMA_CHN0ADDR);
+
+ /* Set bits of CONFIG register but with static context switching */
+ /* FIXME: Check whether to set ACR bit depending on clock ratios */
+ writel(0, SDMA_H_CONFIG);
+
+ writel(ccb_phys, SDMA_H_C0PTR);
+
+ /* download the RAM image for SDMA */
+ sdma_load_script(ram_code,
+ ram_code_size,
+ sdma_script_addrs->ram_code_start_addr);
+
+ /* Set bits of CONFIG register with given context switching mode */
+ writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
+
+ /* Initializes channel's priorities */
+ sdma_set_channel_priority(0, 7);
+
+ clk_disable(sdma_clk);
+
+ return 0;
+
+err_dma_alloc:
+ free_irq(irq, NULL);
+err_request_irq:
+ iounmap(sdma_base);
+err_ioremap:
+ clk_disable(sdma_clk);
+ pr_err("%s failed with %d\n", __func__, ret);
+ return ret;
+}
+
+static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
+{
+ dma_cookie_t cookie = sdma->chan.cookie;
+
+ if (++cookie < 0)
+ cookie = 1;
+
+ sdma->chan.cookie = cookie;
+ sdma->desc.cookie = cookie;
+
+ return cookie;
+}
+
+static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sdma_channel, chan);
+}
+
+static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sdma_channel *sdma = to_sdma_chan(tx->chan);
+ dma_cookie_t cookie;
+
+ spin_lock_irq(&sdma->lock);
+
+ cookie = sdma_assign_cookie(sdma);
+
+ sdma_enable_channel(tx->chan->chan_id);
+
+ spin_unlock_irq(&sdma->lock);
+
+ return cookie;
+}
+
+static int sdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ struct imx_dma_data *data = chan->private;
+ int prio, ret;
+
+ /* No need to execute this for internal channel 0 */
+ if (!chan->chan_id)
+ return 0;
+
+ if (!data)
+ return -EINVAL;
+
+ switch (data->priority) {
+ case DMA_PRIO_HIGH:
+ prio = 3;
+ break;
+ case DMA_PRIO_MEDIUM:
+ prio = 2;
+ break;
+ case DMA_PRIO_LOW:
+ default:
+ prio = 1;
+ break;
+ }
+
+ sdma->peripheral_type = data->peripheral_type;
+ sdma->event_id = data->dma_request;
+ ret = sdma_set_channel_priority(chan->chan_id, prio);
+ if (ret)
+ return ret;
+
+ if (chan->chan_id) {
+ ret = sdma_request_channel(chan->chan_id);
+ if (ret)
+ return ret;
+ }
+
+ dma_async_tx_descriptor_init(&sdma->desc, chan);
+ sdma->desc.tx_submit = sdma_tx_submit;
+ /* txd.flags will be overwritten in prep funcs */
+ sdma->desc.flags = DMA_CTRL_ACK;
+
+ return 0;
+}
+
+static void sdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ int channel = chan->chan_id;
+
+ sdma_disable_channel(channel);
+
+ if (sdma->event_id)
+ sdma_event_disable(channel, sdma->event_id);
+ if (sdma->event_id2)
+ sdma_event_disable(channel, sdma->event_id2);
+
+ sdma->event_id = 0;
+ sdma->event_id2 = 0;
+
+ sdma_set_channel_priority(channel, 0);
+
+ dma_free_coherent(NULL, PAGE_SIZE, sdma->bd, sdma->bd_phys);
+
+ clk_disable(sdma_clk);
+}
+
+#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
+
+static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ int ret, i, count;
+ int channel = chan->chan_id;
+ struct scatterlist *sg;
+
+ if (sdma->busy)
+ return NULL;
+ sdma->busy = 1;
+
+ sdma->flags = 0;
+
+ pr_debug("SDMA: setting up %d entries for channel %d.\n",
+ sg_len, channel);
+
+ sdma->direction = direction;
+ ret = sdma_load_context(channel);
+ if (ret)
+ goto err_out;
+
+ if (sg_len > NUM_BD) {
+ pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, sg_len, NUM_BD);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ struct sdma_buffer_descriptor *bd = &sdma->bd[i];
+ int param;
+
+ bd->buffer_addr = sgl->dma_address;
+
+ count = sg->length;
+
+ if (count > 0xffff) {
+ pr_err("SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
+ channel, count, 0xffff);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ bd->mode.count = count;
+
+ if (sdma->word_size > 4) {
+ ret = -EINVAL;
+ goto err_out;
+ }
+ if (sdma->word_size == 4)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdma->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT;
+
+ if (sdma->flags & IMX_DMA_SG_LOOP) {
+ param |= BD_INTR;
+ if (i + 1 == sg_len)
+ param |= BD_WRAP;
+ }
+
+ if (i + 1 == sg_len)
+ param |= BD_INTR;
+
+ pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, count, sg->dma_address,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+ }
+
+ sdma->num_bd = sg_len;
+ channel_control[channel].current_bd_ptr = sdma->bd_phys;
+
+ return &sdma->desc;
+err_out:
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ int num_periods = buf_len / period_len;
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ int channel = chan->chan_id;
+ int ret, i = 0, buf = 0;
+
+ pr_debug("%s channel: %d\n", __func__, channel);
+
+ if (sdma->busy)
+ return NULL;
+
+ sdma->busy = 1;
+
+ sdma->flags |= IMX_DMA_SG_LOOP;
+ sdma->direction = direction;
+ ret = sdma_load_context(channel);
+ if (ret)
+ goto err_out;
+
+ if (num_periods > NUM_BD) {
+ pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, num_periods, NUM_BD);
+ goto err_out;
+ }
+
+ if (period_len > 0xffff) {
+ pr_err("SDMA channel %d: maximum period size exceeded: %d > %d\n",
+ channel, period_len, 0xffff);
+ goto err_out;
+ }
+
+ while (buf < buf_len) {
+ struct sdma_buffer_descriptor *bd = &sdma->bd[i];
+ int param;
+
+ bd->buffer_addr = dma_addr;
+
+ bd->mode.count = period_len;
+
+ if (sdma->word_size > 4)
+ goto err_out;
+ if (sdma->word_size == 4)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdma->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
+ if (i + 1 == num_periods)
+ param |= BD_WRAP;
+
+ pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, period_len, dma_addr,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+
+ dma_addr += period_len;
+ buf += period_len;
+
+ i++;
+ }
+
+ sdma->num_bd = num_periods;
+ channel_control[channel].current_bd_ptr = sdma->bd_phys;
+
+ return &sdma->desc;
+err_out:
+ sdma->busy = 0;
+ return NULL;
+}
+
+static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ struct dma_slave_config *dmaengine_cfg = (void *)arg;
+
+ switch (cmd) {
+ case DMA_TERMINATE_ALL:
+ sdma_disable_channel(chan->chan_id);
+ return 0;
+ case DMA_SLAVE_CONFIG:
+ if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
+ sdma->per_address = dmaengine_cfg->src_addr;
+ sdma->watermark_level = dmaengine_cfg->src_maxburst;
+ sdma->word_size = dmaengine_cfg->src_addr_width;
+ } else {
+ sdma->per_address = dmaengine_cfg->dst_addr;
+ sdma->watermark_level = dmaengine_cfg->dst_maxburst;
+ sdma->word_size = dmaengine_cfg->dst_addr_width;
+ }
+ return sdma_config_channel(chan->chan_id);
+ default:
+ return -ENOSYS;
+ }
+
+ return -EINVAL;
+}
+
+static enum dma_status sdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sdma_channel *sdma = to_sdma_chan(chan);
+ dma_cookie_t last_used;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+
+ ret = dma_async_is_complete(cookie, sdma->last_completed, last_used);
+ dma_set_tx_state(txstate, sdma->last_completed, last_used, 0);
+
+ return ret;
+}
+
+static void sdma_issue_pending(struct dma_chan *chan)
+{
+ /*
+ * Nothing to do. We only have a single descriptor
+ */
+}
+
+static int __devinit sdma_probe(struct platform_device *pdev)
+{
+ int ret;
+ const struct firmware *fw;
+ const struct sdma_firmware_header *header;
+ const struct sdma_script_start_addrs *addr;
+ int irq;
+ unsigned short *ram_code;
+ struct resource *iores;
+ struct sdma_platform_data *pdata = pdev->dev.platform_data;
+ int version;
+ char *cpustr, *fwname;
+ int i;
+ dma_cap_mask_t mask;
+
+ /* there can be only one */
+ BUG_ON(sdma_base);
+
+ iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ irq = platform_get_irq(pdev, 0);
+ if (!iores || irq < 0 || !pdata)
+ return -EINVAL;
+
+ sdma_clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(sdma_clk)) {
+ ret = PTR_ERR(sdma_clk);
+ goto err_clk;
+ }
+
+ if (cpu_is_mx31()) {
+ cpustr = "imx31";
+ version = mx31_revision() >> 4;
+ } else if (cpu_is_mx35()) {
+ cpustr = "imx35";
+/* FIXME: version = mx35_revision(); */
+ version = 2;
+ } else {
+ ret = -EINVAL;
+ goto err_cputype;
+ }
+
+ fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin", cpustr, version);
+ if (!fwname) {
+ ret = -ENOMEM;
+ goto err_cputype;
+ }
+
+ ret = request_firmware(&fw, fwname, &pdev->dev);
+ if (ret) {
+ dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
+ fwname, ret);
+ kfree(fwname);
+ goto err_cputype;
+ }
+ kfree(fwname);
+
+ if (fw->size < sizeof(*header))
+ goto err_firmware;
+
+ header = (struct sdma_firmware_header *)fw->data;
+
+ if (header->magic != SDMA_FIRMWARE_MAGIC)
+ goto err_firmware;
+ if (header->ram_code_start + header->ram_code_size > fw->size)
+ goto err_firmware;
+
+ addr = (void *)header + header->script_addrs_start;
+ ram_code = (void *)header + header->ram_code_start;
+ memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
+
+ ret = sdma_init(iores->start, irq, pdata->sdma_version,
+ ram_code, header->ram_code_size);
+ if (ret)
+ goto err_firmware;
+
+ INIT_LIST_HEAD(&sdma_dma_device->channels);
+
+ /* Initialize channel parameters */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++) {
+ struct sdma_channel *sdma = &sdma_data[i];
+
+ spin_lock_init(&sdma->lock);
+
+ dma_cap_set(DMA_SLAVE, sdma_dma_device->cap_mask);
+ dma_cap_set(DMA_CYCLIC, sdma_dma_device->cap_mask);
+
+ sdma->chan.device = sdma_dma_device;
+ sdma->chan.chan_id = i;
+
+ /* Add the channel to the DMAC list */
+ list_add_tail(&sdma->chan.device_node, &sdma_dma_device->channels);
+ }
+
+ sdma_dma_device->dev = &pdev->dev;
+
+ sdma_dma_device->device_alloc_chan_resources = sdma_alloc_chan_resources;
+ sdma_dma_device->device_free_chan_resources = sdma_free_chan_resources;
+ sdma_dma_device->device_tx_status = sdma_tx_status;
+ sdma_dma_device->device_prep_slave_sg = sdma_prep_slave_sg;
+ sdma_dma_device->device_prep_dma_cyclic = sdma_prep_dma_cyclic;
+ sdma_dma_device->device_control = sdma_control;
+ sdma_dma_device->device_issue_pending = sdma_issue_pending;
+
+ ret = dma_async_device_register(sdma_dma_device);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register DMAC\n");
+ goto err_firmware;
+ }
+
+ dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
+ header->version_major,
+ header->version_minor);
+
+ /* request channel 0. This is an internal control channel
+ * to the SDMA engine and not available to clients.
+ */
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_request_channel(mask, NULL, NULL);
+
+ release_firmware(fw);
+
+ return 0;
+
+err_firmware:
+ release_firmware(fw);
+err_cputype:
+ clk_put(sdma_clk);
+err_clk:
+ return 0;
+}
+
+static int __devexit sdma_remove(struct platform_device *pdev)
+{
+ return -EBUSY;
+}
+
+static struct platform_driver sdma_driver = {
+ .driver = {
+ .name = "imx-sdma",
+ },
+ .probe = sdma_probe,
+ .remove = __devexit_p(sdma_remove),
+};
+
+static int __init sdma_module_init(void)
+{
+ return platform_driver_register(&sdma_driver);
+}
+subsys_initcall(sdma_module_init);
+
+MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("i.MX SDMA driver");
+MODULE_LICENSE("GPL");
--
1.7.1
^ permalink raw reply related [flat|nested] 78+ messages in thread
* Re: [PATCH 1/3] dmaengine: add possibility for cyclic transfers
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-16 11:56 ` Lothar Waßmann
-1 siblings, 0 replies; 78+ messages in thread
From: Lothar Waßmann @ 2010-08-16 11:56 UTC (permalink / raw)
To: Sascha Hauer
Cc: linux-kernel, Linus Walleij, Dan Williams, Haavard Skinnemoen,
linux-arm-kernel
Hi,
Sascha Hauer writes:
> Cyclic transfers are useful for audio where a single buffer divided
> in periods has to be transfered endlessly until stopped. After being
> prepared the transfer is started using the dma_async_descriptor->tx_submit
> function. dma_async_descriptor->callback is called after each period.
> The transfer is stopped using the DMA_TERMINATE_ALL callback.
> While being used for cyclic transfers the channel cannot be used
> for other transfer types.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
> ---
> drivers/dma/dmaengine.c | 2 ++
> include/linux/dmaengine.h | 6 +++++-
> 2 files changed, 7 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index 9d31d5e..e5e79ce 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -692,6 +692,8 @@ int dma_async_device_register(struct dma_device *device)
> !device->device_prep_dma_interrupt);
> BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> !device->device_prep_slave_sg);
> + BUG_ON(dma_has_cap(DMA_CYCLIC, device->cap_mask) &&
> + !device->device_prep_dma_cyclic);
> BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> !device->device_control);
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index c61d4ca..0df7864 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -67,10 +67,11 @@ enum dma_transaction_type {
> DMA_PRIVATE,
> DMA_ASYNC_TX,
> DMA_SLAVE,
> + DMA_CYCLIC,
> };
>
> /* last transaction type for creation of the capabilities mask */
> -#define DMA_TX_TYPE_END (DMA_SLAVE + 1)
> +#define DMA_TX_TYPE_END (DMA_CYCLIC + 1)
>
>
> /**
> @@ -478,6 +479,9 @@ struct dma_device {
> struct dma_chan *chan, struct scatterlist *sgl,
> unsigned int sg_len, enum dma_data_direction direction,
> unsigned long flags);
> + struct dma_async_tx_descriptor *(*device_prep_dma_cyclic)(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction);
> int (*device_control)(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> unsigned long arg);
>
> --
> 1.7.1
>
Why not implement this feature using cyclic SG lists (created with
sg_chain())? This would give you endless DMA transfers without any
special DMA API extensions.
Lothar Waßmann
--
___________________________________________________________
Ka-Ro electronics GmbH | Pascalstraße 22 | D - 52076 Aachen
Phone: +49 2408 1402-0 | Fax: +49 2408 1402-10
Geschäftsführer: Matthias Kaussen
Handelsregistereintrag: Amtsgericht Aachen, HRB 4996
www.karo-electronics.de | info@karo-electronics.de
___________________________________________________________
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
@ 2010-08-16 11:56 ` Lothar Waßmann
0 siblings, 0 replies; 78+ messages in thread
From: Lothar Waßmann @ 2010-08-16 11:56 UTC (permalink / raw)
To: linux-arm-kernel
Hi,
Sascha Hauer writes:
> Cyclic transfers are useful for audio where a single buffer divided
> in periods has to be transfered endlessly until stopped. After being
> prepared the transfer is started using the dma_async_descriptor->tx_submit
> function. dma_async_descriptor->callback is called after each period.
> The transfer is stopped using the DMA_TERMINATE_ALL callback.
> While being used for cyclic transfers the channel cannot be used
> for other transfer types.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
> ---
> drivers/dma/dmaengine.c | 2 ++
> include/linux/dmaengine.h | 6 +++++-
> 2 files changed, 7 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index 9d31d5e..e5e79ce 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -692,6 +692,8 @@ int dma_async_device_register(struct dma_device *device)
> !device->device_prep_dma_interrupt);
> BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> !device->device_prep_slave_sg);
> + BUG_ON(dma_has_cap(DMA_CYCLIC, device->cap_mask) &&
> + !device->device_prep_dma_cyclic);
> BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> !device->device_control);
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index c61d4ca..0df7864 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -67,10 +67,11 @@ enum dma_transaction_type {
> DMA_PRIVATE,
> DMA_ASYNC_TX,
> DMA_SLAVE,
> + DMA_CYCLIC,
> };
>
> /* last transaction type for creation of the capabilities mask */
> -#define DMA_TX_TYPE_END (DMA_SLAVE + 1)
> +#define DMA_TX_TYPE_END (DMA_CYCLIC + 1)
>
>
> /**
> @@ -478,6 +479,9 @@ struct dma_device {
> struct dma_chan *chan, struct scatterlist *sgl,
> unsigned int sg_len, enum dma_data_direction direction,
> unsigned long flags);
> + struct dma_async_tx_descriptor *(*device_prep_dma_cyclic)(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction);
> int (*device_control)(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> unsigned long arg);
>
> --
> 1.7.1
>
Why not implement this feature using cyclic SG lists (created with
sg_chain())? This would give you endless DMA transfers without any
special DMA API extensions.
Lothar Wa?mann
--
___________________________________________________________
Ka-Ro electronics GmbH | Pascalstra?e 22 | D - 52076 Aachen
Phone: +49 2408 1402-0 | Fax: +49 2408 1402-10
Gesch?ftsf?hrer: Matthias Kaussen
Handelsregistereintrag: Amtsgericht Aachen, HRB 4996
www.karo-electronics.de | info at karo-electronics.de
___________________________________________________________
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-16 12:21 ` Linus Walleij
-1 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-16 12:21 UTC (permalink / raw)
To: Sascha Hauer; +Cc: linux-kernel, Dan Williams, linux-arm-kernel
2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> This patch adds support for the Freescale i.MX SDMA engine.
I like it!
> The SDMA engine is a scatter/gather DMA engine which is implemented
> as a seperate coprocessor. SDMA needs its own firmware which is
> requested using the standard request_firmware mechanism. The firmware
> has different entry points for each peripheral type, so drivers
> have to pass the peripheral type to the DMA engine which in turn
> picks the correct firmware entry point from a table contained in
> the firmware image itself.
Quite fun, if the spec for the microcode is open this opens up
for dynamic firmware generation for specific DMA jobs does it
not?
> I took a very simple approach to implement dmaengine support. Only
> a single descriptor is statically assigned to a each channel. This
> means that transfers can't be queued up but only a single transfer
> is in progress. This simplifies implementation a lot and is sufficient
> for the usual device/memory transfers.
If you want to add memcpy() capability later you're gonna need
this I think, but you can take that when that need arise.
>(...)
> +++ b/arch/arm/plat-mxc/include/mach/dma.h
> @@ -0,0 +1,64 @@
> +/*
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 as
> + * published by the Free Software Foundation.
> + */
> +
> +#ifndef __ASM_ARCH_MXC_DMA_H__
> +#define __ASM_ARCH_MXC_DMA_H__
> +
> +#include <linux/scatterlist.h>
> +
> +/*
> + * This enumerates peripheral types. Used for SDMA.
> + */
> +typedef enum {
The kernel is not really happy about typedefs, can't this be a
regular enum?
> + IMX_DMATYPE_SSI, /* MCU domain SSI */
> + IMX_DMATYPE_SSI_SP, /* Shared SSI */
> + IMX_DMATYPE_MMC, /* MMC */
> + IMX_DMATYPE_SDHC, /* SDHC */
> + IMX_DMATYPE_UART, /* MCU domain UART */
> + IMX_DMATYPE_UART_SP, /* Shared UART */
> + IMX_DMATYPE_FIRI, /* FIRI */
> + IMX_DMATYPE_CSPI, /* MCU domain CSPI */
> + IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
> + IMX_DMATYPE_SIM, /* SIM */
> + IMX_DMATYPE_ATA, /* ATA */
> + IMX_DMATYPE_CCM, /* CCM */
> + IMX_DMATYPE_EXT, /* External peripheral */
> + IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
> + IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
> + IMX_DMATYPE_DSP, /* DSP */
> + IMX_DMATYPE_MEMORY, /* Memory */
> + IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
> + IMX_DMATYPE_SPDIF, /* SPDIF */
> + IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
> + IMX_DMATYPE_ASRC, /* ASRC */
> + IMX_DMATYPE_ESAI, /* ESAI */
> +} sdma_peripheral_type;
> +
> +enum imx_dma_prio {
> + DMA_PRIO_HIGH = 0,
> + DMA_PRIO_MEDIUM = 1,
> + DMA_PRIO_LOW = 2
> +};
> +
> +struct imx_dma_data {
> + int dma_request; /* DMA request line */
Can this be negative and what is the range? I would
suspect something like u8 or u16 would surely be more
apropriate...
> + sdma_peripheral_type peripheral_type;
> + int priority;
Isn't this an enum imx_dma_prio?
> +};
> +
> +static inline int imx_dma_is_ipu(struct dma_chan *chan)
> +{
> + return !strcmp(dev_name(chan->device->dev), "ipu-core");
> +}
> +
> +static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
> +{
> + return !strcmp(dev_name(chan->device->dev), "imx-sdma");
> +}
> +
> +#endif
> diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
> new file mode 100644
> index 0000000..5d542b8
> --- /dev/null
> +++ b/arch/arm/plat-mxc/include/mach/sdma.h
> @@ -0,0 +1,8 @@
> +#ifndef __MACH_MXC_SDMA_H__
> +#define __MACH_MXC_SDMA_H__
> +
> +struct sdma_platform_data {
> + int sdma_version;
Do you have negative versions or can it be unsigned?
> +};
> +
> +#endif /* __MACH_MXC_SDMA_H__ */
> diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
> index 9520cf0..f76bda9 100644
> --- a/drivers/dma/Kconfig
> +++ b/drivers/dma/Kconfig
> @@ -195,6 +195,14 @@ config PCH_DMA
> help
> Enable support for the Topcliff PCH DMA engine.
>
> +config IMX_SDMA
> + tristate "Atmel AHB DMA support"
> + depends on ARCH_MXC
> + select DMA_ENGINE
> + help
> + Support the i.MX SDMA engine. This engine is integrated into
> + Freescale i.MX25/31/35/51 chips.
> +
> config DMA_ENGINE
> bool
>
> diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
> index 72bd703..14d7a1b 100644
> --- a/drivers/dma/Makefile
> +++ b/drivers/dma/Makefile
> @@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
> obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
> obj-$(CONFIG_PL330_DMA) += pl330.o
> obj-$(CONFIG_PCH_DMA) += pch_dma.o
> +obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
> diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> new file mode 100644
> index 0000000..3ba7905
> --- /dev/null
> +++ b/drivers/dma/imx-sdma.c
> @@ -0,0 +1,1383 @@
> +/*
> + * drivers/dma/imx-sdma.c
> + *
> + * This file contains a driver for the Freescale Smart DMA engine
> + *
> + * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
> + *
> + * Based on code from Freescale:
> + *
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * The code contained herein is licensed under the GNU General Public
> + * License. You may obtain a copy of the GNU General Public License
> + * Version 2 or later at the following locations:
> + *
> + * http://www.opensource.org/licenses/gpl-license.html
> + * http://www.gnu.org/copyleft/gpl.html
> + */
> +#include <linux/init.h>
> +#include <linux/types.h>
> +#include <linux/mm.h>
> +#include <linux/interrupt.h>
> +#include <linux/clk.h>
> +#include <linux/semaphore.h>
> +#include <linux/spinlock.h>
> +#include <linux/device.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/firmware.h>
> +#include <linux/slab.h>
> +#include <linux/platform_device.h>
> +#include <linux/dmaengine.h>
> +
> +#include <asm/irq.h>
> +#include <mach/sdma.h>
> +#include <mach/dma.h>
> +#include <mach/hardware.h>
> +
> +/* SDMA registers */
> +#define SDMA_H_C0PTR (sdma_base + 0x000)
> +#define SDMA_H_INTR (sdma_base + 0x004)
> +#define SDMA_H_STATSTOP (sdma_base + 0x008)
> +#define SDMA_H_START (sdma_base + 0x00c)
> +#define SDMA_H_EVTOVR (sdma_base + 0x010)
> +#define SDMA_H_DSPOVR (sdma_base + 0x014)
> +#define SDMA_H_HOSTOVR (sdma_base + 0x018)
> +#define SDMA_H_EVTPEND (sdma_base + 0x01c)
> +#define SDMA_H_DSPENBL (sdma_base + 0x020)
> +#define SDMA_H_RESET (sdma_base + 0x024)
> +#define SDMA_H_EVTERR (sdma_base + 0x028)
> +#define SDMA_H_INTRMSK (sdma_base + 0x02c)
> +#define SDMA_H_PSW (sdma_base + 0x030)
> +#define SDMA_H_EVTERRDBG (sdma_base + 0x034)
> +#define SDMA_H_CONFIG (sdma_base + 0x038)
> +#define SDMA_ONCE_ENB (sdma_base + 0x040)
> +#define SDMA_ONCE_DATA (sdma_base + 0x044)
> +#define SDMA_ONCE_INSTR (sdma_base + 0x048)
> +#define SDMA_ONCE_STAT (sdma_base + 0x04c)
> +#define SDMA_ONCE_CMD (sdma_base + 0x050)
> +#define SDMA_EVT_MIRROR (sdma_base + 0x054)
> +#define SDMA_ILLINSTADDR (sdma_base + 0x058)
> +#define SDMA_CHN0ADDR (sdma_base + 0x05c)
> +#define SDMA_ONCE_RTB (sdma_base + 0x060)
> +#define SDMA_XTRIG_CONF1 (sdma_base + 0x070)
> +#define SDMA_XTRIG_CONF2 (sdma_base + 0x074)
> +#define SDMA_CHNENBL_0 (sdma_base + (sdma_version == 2 ? 0x200 : 0x80))
> +#define SDMA_CHNPRI_0 (sdma_base + 0x100)
All these rely on a fixed sdma_base which makes the driver
a singleton. This is not so good if you imagine the situation with a
platform with two SDMA engines on different addresses.
Can't you create a runtime allocated stateholder to hold
the base and access relative to the offset?
> +
> +/*
> + * Buffer descriptor status values.
> + */
> +#define BD_DONE 0x01
> +#define BD_WRAP 0x02
> +#define BD_CONT 0x04
> +#define BD_INTR 0x08
> +#define BD_RROR 0x10
> +#define BD_LAST 0x20
> +#define BD_EXTD 0x80
> +
> +/*
> + * Data Node descriptor status values.
> + */
> +#define DND_END_OF_FRAME 0x80
> +#define DND_END_OF_XFER 0x40
> +#define DND_DONE 0x20
> +#define DND_UNUSED 0x01
> +
> +/*
> + * IPCV2 descriptor status values.
> + */
> +#define BD_IPCV2_END_OF_FRAME 0x40
> +
> +#define IPCV2_MAX_NODES 50
> +/*
> + * Error bit set in the CCB status field by the SDMA,
> + * in setbd routine, in case of a transfer error
> + */
> +#define DATA_ERROR 0x10000000
> +
> +/*
> + * Buffer descriptor commands.
> + */
> +#define C0_ADDR 0x01
> +#define C0_LOAD 0x02
> +#define C0_DUMP 0x03
> +#define C0_SETCTX 0x07
> +#define C0_GETCTX 0x03
> +#define C0_SETDM 0x01
> +#define C0_SETPM 0x04
> +#define C0_GETDM 0x02
> +#define C0_GETPM 0x08
> +/*
> + * Change endianness indicator in the BD command field
> + */
> +#define CHANGE_ENDIANNESS 0x80
> +
> +/*
> + * Mode/Count of data node descriptors - IPCv2
> + */
> +#ifdef __BIG_ENDIAN
> +struct sdma_mode_count {
> + u32 command : 8; /* command mostlky used for channel 0 */
There are a lot of inline commented struct members, please
use kerneldoc, that's simple. (Applies all over the patch...)
Documentation/kernel-doc-nano-HOWTO
> + u32 status : 8; /* E,R,I,C,W,D status bits stored here */
> + u32 count : 16; /* size of the buffer pointed by this BD */
> +};
> +#else
> +struct sdma_mode_count {
> + u32 count : 16; /* size of the buffer pointed by this BD */
> + u32 status : 8; /* E,R,I,C,W,D status bits stored here */
> + u32 command : 8; /* command mostlky used for channel 0 */
> +};
> +#endif
This use of #ifdef is odd to me but others are probably more
experienced. Anyway, the way it is used with different
:n suffixes makes me believe that you need a packed
compiler directive for this layout to be explicitly coherent.
Atleast add some comment on what this #ifdef construction
does so guys like me can understand what's going on.
> +
> +/*
> + * Buffer descriptor
> + */
> +struct sdma_buffer_descriptor {
> + struct sdma_mode_count mode;
> + u32 buffer_addr; /* address of the buffer described */
> + u32 ext_buffer_addr; /* extended buffer address */
Shouldn't these be dma_addr_t? OK that's probably u32
anyway but just to make a marker...
> +};
> +
> +/*
> + * Channel control Block
> + */
> +struct sdma_channel_control {
> + u32 current_bd_ptr; /* current buffer descriptor processed */
> + u32 base_bd_ptr; /* first element of buffer descriptor array */
> + void *unused;
> + void *unused1;
Hm, can you comment on what these unused things are for...?
> +};
> +
> +/**
> + * Context structure.
> + */
> +#ifdef __BIG_ENDIAN
> +struct sdma_state_registers {
> + u32 sf : 1; /* source fault while loading data */
> + u32 unused0: 1;
> + u32 rpc :14; /* return program counter */
> + u32 t : 1; /* test bit:status of arithmetic & test instruction*/
> + u32 unused1: 1;
> + u32 pc :14; /* program counter */
> + u32 lm : 2; /* loop mode */
> + u32 epc :14; /* loop end program counter */
> + u32 df : 1; /* destination fault while storing data */
> + u32 unused2: 1;
> + u32 spc :14; /* loop start program counter */
> +};
> +#else
> +struct sdma_state_registers {
> + u32 pc :14; /* program counter */
> + u32 unused1: 1;
> + u32 t : 1; /* test bit: status of arithmetic & test instruction*/
> + u32 rpc :14; /* return program counter */
> + u32 unused0: 1;
> + u32 sf : 1; /* source fault while loading data */
> + u32 spc :14; /* loop start program counter */
> + u32 unused2: 1;
> + u32 df : 1; /* destination fault while storing data */
> + u32 epc :14; /* loop end program counter */
> + u32 lm : 2; /* loop mode */
> +};
> +#endif
Again this is odd to me...
> +
> +struct sdma_context_data {
> + struct sdma_state_registers channel_state; /* channel state bits */
> + u32 gReg[8]; /* general registers */
> + u32 mda; /* burst dma destination address register */
> + u32 msa; /* burst dma source address register */
> + u32 ms; /* burst dma status register */
> + u32 md; /* burst dma data register */
> + u32 pda; /* peripheral dma destination address register */
> + u32 psa; /* peripheral dma source address register */
> + u32 ps; /* peripheral dma status register */
> + u32 pd; /* peripheral dma data register */
> + u32 ca; /* CRC polynomial register */
> + u32 cs; /* CRC accumulator register */
> + u32 dda; /* dedicated core destination address register */
> + u32 dsa; /* dedicated core source address register */
> + u32 ds; /* dedicated core status register */
> + u32 dd; /* dedicated core data register */
> + u32 scratch0;
> + u32 scratch1;
> + u32 scratch2;
> + u32 scratch3;
> + u32 scratch4;
> + u32 scratch5;
> + u32 scratch6;
> + u32 scratch7;
> +};
> +
> +struct sdma_channel {
> + /* Channel number */
> + int channel;
Unsigned?
> + /* Transfer type. Needed for setting SDMA script */
> + enum dma_data_direction direction;
> + /* Peripheral type. Needed for setting SDMA script */
> + sdma_peripheral_type peripheral_type;
> + /* Peripheral event id */
> + int event_id;
Unsigned?
> + /* Peripheral event id2 (for channels that use 2 events) */
> + int event_id2;
Unsigned?
> + /* SDMA data access word size */
> + unsigned long word_size;
Is this in bits, bytes etc? Isn't e.g. an u8 enough to hold this,
and further, isn't it possible to recycle enum dma_slave_buswidth
from dmaengine.h instead?
> +
> + /* ID of the buffer that was processed */
> + unsigned int buf_tail;
> +
> + wait_queue_head_t waitq; /* channel completion waitqeue */
> +
> + int num_bd;
Unsigned? Range?
> +
> + struct sdma_buffer_descriptor *bd;
> + dma_addr_t bd_phys;
> +
> + int pc_from_device, pc_to_device;
Unsigned?
> +
> + unsigned long flags;
Is this an u32?
> + dma_addr_t per_address;
> +
> + uint32_t event_mask1, event_mask2;
> + uint32_t watermark_level;
> + uint32_t shp_addr, per_addr;
> +
> + /* DMA-Engine Channel */
> + struct dma_chan chan;
> +
> + spinlock_t lock;
> + struct dma_async_tx_descriptor desc;
> + dma_cookie_t last_completed;
> + int busy;
Shouldn't this be a bool?
> +};
> +
> +#define IMX_DMA_SG_LOOP (1 << 0)
> +
> +#define MAX_DMA_CHANNELS 32
> +#define MXC_SDMA_DEFAULT_PRIORITY 1
> +#define MXC_SDMA_MIN_PRIORITY 1
> +#define MXC_SDMA_MAX_PRIORITY 7
> +
> +/*
> + * This enumerates transfer types
> + */
> +typedef enum {
Again a typedef, please plain enum is fine.
> + emi_2_per = 0, /* EMI memory to peripheral */
> + emi_2_int, /* EMI memory to internal RAM */
> + emi_2_emi, /* EMI memory to EMI memory */
> + emi_2_dsp, /* EMI memory to DSP memory */
> + per_2_int, /* Peripheral to internal RAM */
> + per_2_emi, /* Peripheral to internal EMI memory */
> + per_2_dsp, /* Peripheral to DSP memory */
> + per_2_per, /* Peripheral to Peripheral */
> + int_2_per, /* Internal RAM to peripheral */
> + int_2_int, /* Internal RAM to Internal RAM */
> + int_2_emi, /* Internal RAM to EMI memory */
> + int_2_dsp, /* Internal RAM to DSP memory */
> + dsp_2_per, /* DSP memory to peripheral */
> + dsp_2_int, /* DSP memory to internal RAM */
> + dsp_2_emi, /* DSP memory to EMI memory */
> + dsp_2_dsp, /* DSP memory to DSP memory */
> + emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
> + dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
> + dvfs_pll, /* DVFS script with PLL change */
> + dvfs_pdr /* DVFS script without PLL change */
> +} sdma_transfer_type;
> +
> +/*
> + * Structure containing sdma request parameters.
> + */
> +struct sdma_script_start_addrs {
> + int ap_2_ap_addr;
> + int ap_2_bp_addr;
> + int ap_2_ap_fixed_addr;
> + int bp_2_ap_addr;
> + int loopback_on_dsp_side_addr;
> + int mcu_interrupt_only_addr;
> +
> + int firi_2_per_addr;
> + int firi_2_mcu_addr;
> + int per_2_firi_addr;
> + int mcu_2_firi_addr;
> +
> + int uart_2_per_addr;
> + int uart_2_mcu_addr;
> + int per_2_app_addr;
> + int mcu_2_app_addr;
> + int per_2_per_addr;
> +
> + int uartsh_2_per_addr;
> + int uartsh_2_mcu_addr;
> + int per_2_shp_addr;
> + int mcu_2_shp_addr;
> +
> + int ata_2_mcu_addr;
> + int mcu_2_ata_addr;
> +
> + int app_2_per_addr;
> + int app_2_mcu_addr;
> + int shp_2_per_addr;
> + int shp_2_mcu_addr;
> +
> + int mshc_2_mcu_addr;
> + int mcu_2_mshc_addr;
> +
> + int spdif_2_mcu_addr;
> + int mcu_2_spdif_addr;
> +
> + int asrc_2_mcu_addr;
> +
> + int ext_mem_2_ipu_addr;
> +
> + int descrambler_addr;
> +
> + int dptc_dvfs_addr;
> +
> + int utra_addr;
> +
> + int ram_code_start_addr;
All these addresses, are they really integers with
valid negative values... Aren't they dma_addr_t or
atleast u32?
> +};
> +
> +#define SDMA_FIRMWARE_MAGIC 0x414d4453
> +
> +struct sdma_firmware_header {
> + uint32_t magic; /* "SDMA" */
> + uint32_t version_major; /* increased whenever layout of struct sdma_script_start_addrs changes */
> + uint32_t version_minor; /* firmware version */
> + uint32_t script_addrs_start; /* offset of struct sdma_script_start_addrs in this image */
> + uint32_t num_script_addrs; /* Number of script addresses in this image */
> + uint32_t ram_code_start; /* offset of SDMA ram image in this firmware image */
> + uint32_t ram_code_size; /* size of SDMA ram image */
Please use u32. uint32_t is not the preferred kernel type.
(Still I've seen people use it in some cases so I might be wrong,
feel welcome to bit back on this.)
> +};
> +
> +static struct sdma_channel sdma_data[MAX_DMA_CHANNELS];
> +static struct sdma_channel_control *channel_control;
> +static void __iomem *sdma_base;
> +static int sdma_version;
Unsigned?
> +static int sdma_num_events;
Unsigned?
> +static struct sdma_context_data *sdma_context;
> +dma_addr_t sdma_context_phys;
> +static struct dma_device __sdma_dma_device;
> +static struct dma_device *sdma_dma_device = &__sdma_dma_device;
This is what I suspected: local variables making the entire driver
a singleton, which means you can never have more than one
SDMA. Atleast collect all of these in a struct, call it
"struct sdma" simply (if you ask me) and use as a stateholder.
This makes it easier to kzalloc() that struct later if you
want to support non-singletons.
I know this require some work but I've done it to several drivers
(always asked on mailinglists to do this) and I don't regret a single
rewrite. Last time was for the PL18x DMAengine driver actually.
> +
> +#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
> +#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
> +#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
> +#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
> +
> +static int sdma_config_ownership(int channel, int event_override,
> + int mcu_verride, int dsp_override)
> +{
> + u32 evt, mcu, dsp;
> +
> + if (event_override && mcu_verride && dsp_override)
> + return -EINVAL;
> +
> + evt = readl(SDMA_H_EVTOVR);
> + mcu = readl(SDMA_H_HOSTOVR);
> + dsp = readl(SDMA_H_DSPOVR);
> +
> + if (dsp_override)
> + dsp &= ~(1 << channel);
> + else
> + dsp |= (1 << channel);
> +
> + if (event_override)
> + evt &= ~(1 << channel);
> + else
> + evt |= (1 << channel);
> +
> + if (mcu_verride)
> + mcu &= ~(1 << channel);
> + else
> + mcu |= (1 << channel);
> +
> + writel(evt, SDMA_H_EVTOVR);
> + writel(mcu, SDMA_H_HOSTOVR);
> + writel(dsp, SDMA_H_DSPOVR);
> +
> + return 0;
> +}
> +
> +/*
> + * sdma_run_channel - run a channel and wait till it's done
> + */
> +static int sdma_run_channel(int channel)
> +{
> + struct sdma_channel *sdma = &sdma_data[channel];
> + int ret;
> +
> + writel(1 << channel, SDMA_H_START);
> +
> + ret = wait_event_interruptible(sdma->waitq,
> + !(readl(SDMA_H_STATSTOP) & (1 << channel)));
OK not the biggest thing in the world, but can't you use a
completion for this? (I'm not so clever with waitqueues so
forgive me if this is malinformed.)
> + return ret;
> +}
> +
> +static int sdma_load_script(void *buf, int size, u32 address)
> +{
> + struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> + void *buf_virt;
> + dma_addr_t buf_phys;
> + int ret;
> +
> + buf_virt = dma_alloc_coherent(NULL,
> + size,
> + &buf_phys, GFP_KERNEL);
> + if (!buf_virt)
> + return -ENOMEM;
> +
> + bd0->mode.command = C0_SETPM;
> + bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> + bd0->mode.count = size / 2;
> + bd0->buffer_addr = buf_phys;
> + bd0->ext_buffer_addr = address;
> +
> + memcpy(buf_virt, buf, size);
> +
> + ret = sdma_run_channel(0);
> +
> + dma_free_coherent(NULL, size, buf_virt, buf_phys);
> +
> + return ret;
> +}
> +
> +static void sdma_event_enable(int channel, int event)
> +{
> + u32 val;
> +
> + val = readl(SDMA_CHNENBL_0 + event * 4);
This use indicates that event should probably be
unsigned, and probably not greater than u16 atleast.
I suspect it is never more than an u8 really.
> + val |= (1 << channel);
> + writel(val, SDMA_CHNENBL_0 + event * 4);
> +}
> +
> +static void sdma_event_disable(int channel, int event)
> +{
> + u32 val;
> +
> + val = readl(SDMA_CHNENBL_0 + event * 4);
> + val &= ~(1 << channel);
> + writel(val, SDMA_CHNENBL_0 + event * 4);
Same comment here.
> +}
> +
> +static void mxc_sdma_handle_channel_loop(int channel)
> +{
> + struct sdma_channel *sdma = &sdma_data[channel];
This indicates that channel should be unsigned.
> + struct sdma_buffer_descriptor *bd;
> + int error = 0;
Unused variable?
> +
> + /*
> + * loop mode. Iterate over descriptors, re-setup them and
> + * call callback function.
> + */
> + while (1) {
> + bd = &sdma->bd[sdma->buf_tail];
> +
> + if (bd->mode.status & BD_DONE)
> + break;
> +
> + if (bd->mode.status & BD_RROR)
> + error = -EIO;
> +
> + bd->mode.status |= BD_DONE;
> + sdma->buf_tail++;
> + sdma->buf_tail %= sdma->num_bd;
> +
> + if (sdma->desc.callback)
> + sdma->desc.callback(sdma->desc.callback_param);
> + }
> +}
> +
> +static void mxc_sdma_handle_channel_normal(int channel)
> +{
> + struct sdma_channel *sdma = &sdma_data[channel];
> + struct sdma_buffer_descriptor *bd;
> + int i, error = 0;
> +
> + /*
> + * non loop mode. Iterate over all descriptors, collect
> + * errors and call callback function
> + */
> + for (i = 0; i < sdma->num_bd; i++) {
> + bd = &sdma->bd[i];
> +
> + if (bd->mode.status & (BD_DONE | BD_RROR))
> + error = -EIO;
> + }
> +
> + if (sdma->desc.callback)
> + sdma->desc.callback(sdma->desc.callback_param);
> + sdma->last_completed = sdma->desc.cookie;
> +
> + sdma->busy = 0;
= true if you switch this to bool..
> +}
> +
> +static void mxc_sdma_handle_channel(int channel)
> +{
> + struct sdma_channel *sdma = &sdma_data[channel];
> +
> + wake_up_interruptible(&sdma->waitq);
> +
> + /* not interested in channel 0 interrupts */
> + if (!channel)
> + return;
> +
> + if (sdma->flags & IMX_DMA_SG_LOOP)
> + mxc_sdma_handle_channel_loop(channel);
> + else
> + mxc_sdma_handle_channel_normal(channel);
> +}
> +
> +static irqreturn_t sdma_int_handler(int irq, void *dev_id)
> +{
> + u32 stat;
> +
> + stat = readl(SDMA_H_INTR);
> + writel(stat, SDMA_H_INTR);
> +
> + while (stat) {
> + int channel = fls(stat) - 1;
> +
> + mxc_sdma_handle_channel(channel);
> +
> + stat &= ~(1 << channel);
> + }
> +
> + return IRQ_HANDLED;
> +}
> +
> +static struct clk *sdma_clk;
> +
> +/*
> + * Stores the start address of the SDMA scripts
> + */
> +static struct sdma_script_start_addrs __sdma_script_addrs;
> +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
> +
> +/*
> + * sets the pc of SDMA script according to the peripheral type
> + */
> +static void sdma_get_pc(struct sdma_channel *sdma,
> + sdma_peripheral_type peripheral_type)
> +{
> + int res = 0;
> + int per_2_emi = 0, emi_2_per = 0;
> + int per_2_int = 0, int_2_per = 0;
> + int per_2_per = 0, emi_2_emi = 0;
> +
> + sdma->pc_from_device = 0;
> + sdma->pc_to_device = 0;
There are a *lot* of local variables here, and only two of them
are used eventually, at the end of the function. I cannot quite
follow this, what is going on?
Some like emi_2_emi seem to be totally unused.
The types here look like some kind of enum or other
similar construction is really what's being asked for
here.
> +
> + switch (peripheral_type) {
> + case IMX_DMATYPE_MEMORY:
> + emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
> + break;
> + case IMX_DMATYPE_DSP:
> + emi_2_per = sdma_script_addrs->bp_2_ap_addr;
> + per_2_emi = sdma_script_addrs->ap_2_bp_addr;
> + break;
> + case IMX_DMATYPE_FIRI:
> + per_2_int = sdma_script_addrs->firi_2_per_addr;
> + per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
> + int_2_per = sdma_script_addrs->per_2_firi_addr;
> + emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
> + break;
> + case IMX_DMATYPE_UART:
> + per_2_int = sdma_script_addrs->uart_2_per_addr;
> + per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
> + int_2_per = sdma_script_addrs->per_2_app_addr;
> + emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> + break;
> + case IMX_DMATYPE_UART_SP:
> + per_2_int = sdma_script_addrs->uartsh_2_per_addr;
> + per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
> + int_2_per = sdma_script_addrs->per_2_shp_addr;
> + emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> + break;
> + case IMX_DMATYPE_ATA:
> + per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
> + emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
> + break;
> + case IMX_DMATYPE_CSPI:
> + case IMX_DMATYPE_EXT:
> + case IMX_DMATYPE_SSI:
> + per_2_int = sdma_script_addrs->app_2_per_addr;
> + per_2_emi = sdma_script_addrs->app_2_mcu_addr;
> + int_2_per = sdma_script_addrs->per_2_app_addr;
> + emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> + break;
> + case IMX_DMATYPE_SSI_SP:
> + case IMX_DMATYPE_MMC:
> + case IMX_DMATYPE_SDHC:
> + case IMX_DMATYPE_CSPI_SP:
> + case IMX_DMATYPE_ESAI:
> + case IMX_DMATYPE_MSHC_SP:
> + per_2_int = sdma_script_addrs->shp_2_per_addr;
> + per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
> + int_2_per = sdma_script_addrs->per_2_shp_addr;
> + emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> + break;
> + case IMX_DMATYPE_ASRC:
> + per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
> + emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
> + per_2_per = sdma_script_addrs->per_2_per_addr;
> + break;
> + case IMX_DMATYPE_MSHC:
> + per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
> + emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
> + break;
> + case IMX_DMATYPE_CCM:
> + per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
> + break;
> + case IMX_DMATYPE_FIFO_MEMORY:
> + res = sdma_script_addrs->ap_2_ap_fixed_addr;
res? This thing is never used.
> + break;
> + case IMX_DMATYPE_SPDIF:
> + per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
> + emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
> + break;
> + case IMX_DMATYPE_IPU_MEMORY:
> + emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
> + break;
> + default:
> + break;
> + }
> +
> + sdma->pc_from_device = per_2_emi;
> + sdma->pc_to_device = emi_2_per;
Return res? You're assigning it a value in some cases.
> +}
> +
> +static int sdma_load_context(int channel)
> +{
> + struct sdma_channel *sdma = &sdma_data[channel];
> + int load_address;
> + struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> + int ret;
> +
> + if (sdma->direction == DMA_FROM_DEVICE) {
> + load_address = sdma->pc_from_device;
> + } else {
> + load_address = sdma->pc_to_device;
> + }
> +
> + if (load_address < 0)
> + return load_address;
> +
> + pr_debug("%s: load_address = %d\n", __func__, load_address);
> + pr_debug("%s: wml = 0x%08x\n", __func__, sdma->watermark_level);
> + pr_debug("%s: shp_addr = 0x%08x\n", __func__, sdma->shp_addr);
> + pr_debug("%s: per_addr = 0x%08x\n", __func__, sdma->per_addr);
> + pr_debug("%s: event_mask1 = 0x%08x\n", __func__, sdma->event_mask1);
> + pr_debug("%s: event_mask2 = 0x%08x\n", __func__, sdma->event_mask2);
Surely it must be possible to get the struct device * pointer for the
channels host and use dev_dbg() instead?
> +
> + memset(sdma_context, 0, sizeof(*sdma_context));
> + sdma_context->channel_state.pc = load_address;
> +
> + /* Send by context the event mask,base address for peripheral
> + * and watermark level
> + */
> + sdma_context->gReg[0] = sdma->event_mask2;
> + sdma_context->gReg[1] = sdma->event_mask1;
> + sdma_context->gReg[2] = sdma->per_addr;
> + sdma_context->gReg[6] = sdma->shp_addr;
> + sdma_context->gReg[7] = sdma->watermark_level;
> +
> + bd0->mode.command = C0_SETDM;
> + bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> + bd0->mode.count = sizeof(*sdma_context) / 4;
> + bd0->buffer_addr = sdma_context_phys;
> + bd0->ext_buffer_addr = 2048 + (sizeof(*sdma_context) / 4) * channel;
> +
> + ret = sdma_run_channel(0);
> +
> + return ret;
> +}
> +
> +static void sdma_disable_channel(int channel)
> +{
> + struct sdma_channel *sdma = &sdma_data[channel];
> +
> + writel(1 << channel, SDMA_H_STATSTOP);
> + sdma->busy = 0;
> +}
> +
> +static int sdma_config_channel(int channel)
> +{
> + struct sdma_channel *sdma = &sdma_data[channel];
> + int ret;
> +
> + sdma_disable_channel(channel);
> +
> + sdma->event_mask1 = 0;
> + sdma->event_mask2 = 0;
> + sdma->shp_addr = 0;
> + sdma->per_addr = 0;
> +
> + if (sdma->event_id)
> + sdma_event_enable(channel, sdma->event_id);
> +
> + switch (sdma->peripheral_type) {
> + case IMX_DMATYPE_DSP:
> + sdma_config_ownership(channel, 0, 1, 1);
The parameters here makes yoy believe that the types should
be bool rather than int...
> + break;
> + case IMX_DMATYPE_MEMORY:
> + sdma_config_ownership(channel, 0, 1, 0);
> + break;
> + default:
> + sdma_config_ownership(channel, 1, 1, 0);
> + break;
> + }
> +
> + sdma_get_pc(sdma, sdma->peripheral_type);
> +
> + if ((sdma->peripheral_type != IMX_DMATYPE_MEMORY) &&
> + (sdma->peripheral_type != IMX_DMATYPE_DSP)) {
> + /* Handle multiple event channels differently */
> + if (sdma->event_id2) {
> + sdma->event_mask2 = 1 << (sdma->event_id2 % 32);
> + if (sdma->event_id2 > 31)
> + sdma->watermark_level |= 1 << 31;
> + sdma->event_mask1 = 1 << (sdma->event_id % 32);
> + if (sdma->event_id > 31)
> + sdma->watermark_level |= 1 << 30;
> + } else {
> + sdma->event_mask1 = 1 << sdma->event_id;
> + sdma->event_mask2 = 1 << (sdma->event_id - 32);
> + }
> + /* Watermark Level */
> + sdma->watermark_level |= sdma->watermark_level;
> + /* Address */
> + sdma->shp_addr = sdma->per_address;
> + } else {
> + sdma->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
> + }
> +
> + ret = sdma_load_context(channel);
> +
> + return ret;
> +}
> +
> +static int sdma_set_channel_priority(unsigned int channel, unsigned int priority)
> +{
> + if (priority < MXC_SDMA_MIN_PRIORITY
> + || priority > MXC_SDMA_MAX_PRIORITY) {
> + return -EINVAL;
> + }
> +
> + writel(priority, SDMA_CHNPRI_0 + 4 * channel);
> +
> + return 0;
> +}
> +
> +static int sdma_request_channel(int channel)
> +{
> + struct sdma_channel *sdma = &sdma_data[channel];
> + int ret = -EBUSY;
> +
> + sdma->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdma->bd_phys, GFP_KERNEL);
> + if (!sdma->bd) {
> + ret = -ENOMEM;
> + goto out;
> + }
> +
> + memset(sdma->bd, 0, PAGE_SIZE);
> +
> + channel_control[channel].base_bd_ptr = sdma->bd_phys;
> + channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> + clk_enable(sdma_clk);
Aha you're enabling it once for every channel and rely on
clk reference counting that's clever!
> +
> + sdma_set_channel_priority(channel, MXC_SDMA_DEFAULT_PRIORITY);
> +
> + init_waitqueue_head(&sdma->waitq);
> +
> + sdma->buf_tail = 0;
> +
> + return 0;
> +out:
> +
> + return ret;
> +}
> +
> +static void sdma_enable_channel(int channel)
> +{
> + writel(1 << channel, SDMA_H_START);
> +}
> +
> +static int __init sdma_init(unsigned long phys_base, int irq, int version,
> + void *ram_code,
> + int ram_code_size)
> +{
> + int i, ret;
> + int channel;
> + dma_addr_t ccb_phys;
> +
> + sdma_version = version;
> + switch (sdma_version) {
> + case 1:
> + sdma_num_events = 32;
> + break;
> + case 2:
> + sdma_num_events = 48;
> + break;
> + default:
> + pr_err("SDMA: Unknown version %d. aborting\n", sdma_version);
> + return -ENODEV;
> + }
> +
> + clk_enable(sdma_clk);
> +
> + sdma_base = ioremap(phys_base, 4096);
Use SZ_4K instead of 4096.
> + if (!sdma_base) {
> + ret = -ENOMEM;
> + goto err_ioremap;
> + }
> +
> + /* Initialize SDMA private data */
> + memset(sdma_data, 0, sizeof(struct sdma_channel) * MAX_DMA_CHANNELS);
> +
> + for (channel = 0; channel < MAX_DMA_CHANNELS; channel++)
> + sdma_data[channel].channel = channel;
> +
> + ret = request_irq(irq, sdma_int_handler, 0, "sdma", NULL);
> + if (ret)
> + goto err_request_irq;
> +
> + /* Be sure SDMA has not started yet */
> + writel(0, SDMA_H_C0PTR);
> +
> + channel_control = dma_alloc_coherent(NULL,
> + MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
> + sizeof(struct sdma_context_data),
> + &ccb_phys, GFP_KERNEL);
> +
> + if (!channel_control) {
> + ret = -ENOMEM;
> + goto err_dma_alloc;
> + }
> +
> + sdma_context = (void *)channel_control +
> + MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> + sdma_context_phys = ccb_phys +
> + MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> +
> + /* Zero-out the CCB structures array just allocated */
> + memset(channel_control, 0,
> + MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
> +
> + /* disable all channels */
> + for (i = 0; i < sdma_num_events; i++)
> + writel(0, SDMA_CHNENBL_0 + i * 4);
> +
> + /* All channels have priority 0 */
> + for (i = 0; i < MAX_DMA_CHANNELS; i++)
> + writel(0, SDMA_CHNPRI_0 + i * 4);
> +
> + ret = sdma_request_channel(0);
> + if (ret)
> + goto err_dma_alloc;
> +
> + sdma_config_ownership(0, 0, 1, 0);
> +
> + /* Set Command Channel (Channel Zero) */
> + writel(0x4050, SDMA_CHN0ADDR);
> +
> + /* Set bits of CONFIG register but with static context switching */
> + /* FIXME: Check whether to set ACR bit depending on clock ratios */
> + writel(0, SDMA_H_CONFIG);
> +
> + writel(ccb_phys, SDMA_H_C0PTR);
> +
> + /* download the RAM image for SDMA */
> + sdma_load_script(ram_code,
> + ram_code_size,
> + sdma_script_addrs->ram_code_start_addr);
> +
> + /* Set bits of CONFIG register with given context switching mode */
> + writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
> +
> + /* Initializes channel's priorities */
> + sdma_set_channel_priority(0, 7);
> +
> + clk_disable(sdma_clk);
> +
> + return 0;
> +
> +err_dma_alloc:
> + free_irq(irq, NULL);
> +err_request_irq:
> + iounmap(sdma_base);
> +err_ioremap:
> + clk_disable(sdma_clk);
> + pr_err("%s failed with %d\n", __func__, ret);
> + return ret;
> +}
> +
> +static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
> +{
> + dma_cookie_t cookie = sdma->chan.cookie;
> +
> + if (++cookie < 0)
> + cookie = 1;
> +
> + sdma->chan.cookie = cookie;
> + sdma->desc.cookie = cookie;
> +
> + return cookie;
> +}
> +
> +static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
> +{
> + return container_of(chan, struct sdma_channel, chan);
> +}
> +
> +static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
> +{
> + struct sdma_channel *sdma = to_sdma_chan(tx->chan);
> + dma_cookie_t cookie;
> +
> + spin_lock_irq(&sdma->lock);
> +
> + cookie = sdma_assign_cookie(sdma);
> +
> + sdma_enable_channel(tx->chan->chan_id);
> +
> + spin_unlock_irq(&sdma->lock);
> +
> + return cookie;
> +}
> +
> +static int sdma_alloc_chan_resources(struct dma_chan *chan)
> +{
> + struct sdma_channel *sdma = to_sdma_chan(chan);
> + struct imx_dma_data *data = chan->private;
> + int prio, ret;
> +
> + /* No need to execute this for internal channel 0 */
> + if (!chan->chan_id)
> + return 0;
> +
> + if (!data)
> + return -EINVAL;
> +
> + switch (data->priority) {
> + case DMA_PRIO_HIGH:
> + prio = 3;
Wait, aren't these enumerated?
Add some enum sdma_channel_prio {}..
> + break;
> + case DMA_PRIO_MEDIUM:
> + prio = 2;
> + break;
> + case DMA_PRIO_LOW:
> + default:
> + prio = 1;
> + break;
> + }
> +
> + sdma->peripheral_type = data->peripheral_type;
> + sdma->event_id = data->dma_request;
> + ret = sdma_set_channel_priority(chan->chan_id, prio);
> + if (ret)
> + return ret;
> +
> + if (chan->chan_id) {
> + ret = sdma_request_channel(chan->chan_id);
> + if (ret)
> + return ret;
> + }
> +
> + dma_async_tx_descriptor_init(&sdma->desc, chan);
> + sdma->desc.tx_submit = sdma_tx_submit;
> + /* txd.flags will be overwritten in prep funcs */
> + sdma->desc.flags = DMA_CTRL_ACK;
> +
> + return 0;
> +}
> +
> +static void sdma_free_chan_resources(struct dma_chan *chan)
> +{
> + struct sdma_channel *sdma = to_sdma_chan(chan);
> + int channel = chan->chan_id;
> +
> + sdma_disable_channel(channel);
> +
> + if (sdma->event_id)
> + sdma_event_disable(channel, sdma->event_id);
> + if (sdma->event_id2)
> + sdma_event_disable(channel, sdma->event_id2);
> +
> + sdma->event_id = 0;
> + sdma->event_id2 = 0;
> +
> + sdma_set_channel_priority(channel, 0);
> +
> + dma_free_coherent(NULL, PAGE_SIZE, sdma->bd, sdma->bd_phys);
> +
> + clk_disable(sdma_clk);
> +}
> +
> +#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
> +
> +static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
> + struct dma_chan *chan, struct scatterlist *sgl,
> + unsigned int sg_len, enum dma_data_direction direction,
> + unsigned long flags)
> +{
> + struct sdma_channel *sdma = to_sdma_chan(chan);
> + int ret, i, count;
> + int channel = chan->chan_id;
> + struct scatterlist *sg;
> +
> + if (sdma->busy)
> + return NULL;
> + sdma->busy = 1;
> +
> + sdma->flags = 0;
What are those flags anyway? I think you will need some
#define:s for them.
> +
> + pr_debug("SDMA: setting up %d entries for channel %d.\n",
> + sg_len, channel);
> +
> + sdma->direction = direction;
> + ret = sdma_load_context(channel);
> + if (ret)
> + goto err_out;
> +
> + if (sg_len > NUM_BD) {
> + pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> + channel, sg_len, NUM_BD);
> + ret = -EINVAL;
> + goto err_out;
> + }
> +
> + for_each_sg(sgl, sg, sg_len, i) {
> + struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> + int param;
> +
> + bd->buffer_addr = sgl->dma_address;
> +
> + count = sg->length;
> +
> + if (count > 0xffff) {
> + pr_err("SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
> + channel, count, 0xffff);
> + ret = -EINVAL;
> + goto err_out;
> + }
> +
> + bd->mode.count = count;
> +
> + if (sdma->word_size > 4) {
> + ret = -EINVAL;
> + goto err_out;
> + }
> + if (sdma->word_size == 4)
> + bd->mode.command = 0;
> + else
> + bd->mode.command = sdma->word_size;
> +
> + param = BD_DONE | BD_EXTD | BD_CONT;
> +
> + if (sdma->flags & IMX_DMA_SG_LOOP) {
> + param |= BD_INTR;
> + if (i + 1 == sg_len)
> + param |= BD_WRAP;
> + }
> +
> + if (i + 1 == sg_len)
> + param |= BD_INTR;
> +
> + pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> + i, count, sg->dma_address,
> + param & BD_WRAP ? "wrap" : "",
> + param & BD_INTR ? " intr" : "");
> +
> + bd->mode.status = param;
> + }
> +
> + sdma->num_bd = sg_len;
> + channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> + return &sdma->desc;
> +err_out:
> + return NULL;
> +}
> +
> +static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
> + struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction)
> +{
> + int num_periods = buf_len / period_len;
> + struct sdma_channel *sdma = to_sdma_chan(chan);
> + int channel = chan->chan_id;
> + int ret, i = 0, buf = 0;
> +
> + pr_debug("%s channel: %d\n", __func__, channel);
Must be possible to find struct device * and use dev_dbg()
> +
> + if (sdma->busy)
> + return NULL;
> +
> + sdma->busy = 1;
> +
> + sdma->flags |= IMX_DMA_SG_LOOP;
> + sdma->direction = direction;
> + ret = sdma_load_context(channel);
> + if (ret)
> + goto err_out;
> +
> + if (num_periods > NUM_BD) {
> + pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> + channel, num_periods, NUM_BD);
> + goto err_out;
> + }
> +
> + if (period_len > 0xffff) {
> + pr_err("SDMA channel %d: maximum period size exceeded: %d > %d\n",
> + channel, period_len, 0xffff);
> + goto err_out;
> + }
> +
> + while (buf < buf_len) {
> + struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> + int param;
> +
> + bd->buffer_addr = dma_addr;
> +
> + bd->mode.count = period_len;
> +
> + if (sdma->word_size > 4)
> + goto err_out;
> + if (sdma->word_size == 4)
> + bd->mode.command = 0;
> + else
> + bd->mode.command = sdma->word_size;
> +
> + param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
> + if (i + 1 == num_periods)
> + param |= BD_WRAP;
> +
> + pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> + i, period_len, dma_addr,
> + param & BD_WRAP ? "wrap" : "",
> + param & BD_INTR ? " intr" : "");
> +
> + bd->mode.status = param;
> +
> + dma_addr += period_len;
> + buf += period_len;
> +
> + i++;
> + }
> +
> + sdma->num_bd = num_periods;
> + channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> + return &sdma->desc;
> +err_out:
> + sdma->busy = 0;
> + return NULL;
> +}
> +
> +static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> + unsigned long arg)
> +{
> + struct sdma_channel *sdma = to_sdma_chan(chan);
> + struct dma_slave_config *dmaengine_cfg = (void *)arg;
> +
> + switch (cmd) {
> + case DMA_TERMINATE_ALL:
> + sdma_disable_channel(chan->chan_id);
> + return 0;
> + case DMA_SLAVE_CONFIG:
> + if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
> + sdma->per_address = dmaengine_cfg->src_addr;
> + sdma->watermark_level = dmaengine_cfg->src_maxburst;
> + sdma->word_size = dmaengine_cfg->src_addr_width;
> + } else {
> + sdma->per_address = dmaengine_cfg->dst_addr;
> + sdma->watermark_level = dmaengine_cfg->dst_maxburst;
> + sdma->word_size = dmaengine_cfg->dst_addr_width;
> + }
> + return sdma_config_channel(chan->chan_id);
> + default:
> + return -ENOSYS;
> + }
> +
> + return -EINVAL;
> +}
> +
> +static enum dma_status sdma_tx_status(struct dma_chan *chan,
> + dma_cookie_t cookie,
> + struct dma_tx_state *txstate)
> +{
> + struct sdma_channel *sdma = to_sdma_chan(chan);
> + dma_cookie_t last_used;
> + enum dma_status ret;
> +
> + last_used = chan->cookie;
> +
> + ret = dma_async_is_complete(cookie, sdma->last_completed, last_used);
> + dma_set_tx_state(txstate, sdma->last_completed, last_used, 0);
> +
> + return ret;
> +}
> +
> +static void sdma_issue_pending(struct dma_chan *chan)
> +{
> + /*
> + * Nothing to do. We only have a single descriptor
> + */
> +}
> +
> +static int __devinit sdma_probe(struct platform_device *pdev)
> +{
> + int ret;
> + const struct firmware *fw;
> + const struct sdma_firmware_header *header;
> + const struct sdma_script_start_addrs *addr;
> + int irq;
> + unsigned short *ram_code;
> + struct resource *iores;
> + struct sdma_platform_data *pdata = pdev->dev.platform_data;
> + int version;
> + char *cpustr, *fwname;
> + int i;
> + dma_cap_mask_t mask;
> +
> + /* there can be only one */
> + BUG_ON(sdma_base);
> +
> + iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + irq = platform_get_irq(pdev, 0);
> + if (!iores || irq < 0 || !pdata)
> + return -EINVAL;
> +
> + sdma_clk = clk_get(&pdev->dev, NULL);
> + if (IS_ERR(sdma_clk)) {
> + ret = PTR_ERR(sdma_clk);
> + goto err_clk;
> + }
> +
> + if (cpu_is_mx31()) {
> + cpustr = "imx31";
> + version = mx31_revision() >> 4;
> + } else if (cpu_is_mx35()) {
> + cpustr = "imx35";
> +/* FIXME: version = mx35_revision(); */
> + version = 2;
> + } else {
> + ret = -EINVAL;
> + goto err_cputype;
> + }
> +
> + fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin", cpustr, version);
> + if (!fwname) {
> + ret = -ENOMEM;
> + goto err_cputype;
> + }
> +
> + ret = request_firmware(&fw, fwname, &pdev->dev);
> + if (ret) {
> + dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
> + fwname, ret);
> + kfree(fwname);
> + goto err_cputype;
> + }
> + kfree(fwname);
> +
> + if (fw->size < sizeof(*header))
> + goto err_firmware;
> +
> + header = (struct sdma_firmware_header *)fw->data;
> +
> + if (header->magic != SDMA_FIRMWARE_MAGIC)
> + goto err_firmware;
> + if (header->ram_code_start + header->ram_code_size > fw->size)
> + goto err_firmware;
> +
> + addr = (void *)header + header->script_addrs_start;
> + ram_code = (void *)header + header->ram_code_start;
> + memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
> +
> + ret = sdma_init(iores->start, irq, pdata->sdma_version,
> + ram_code, header->ram_code_size);
> + if (ret)
> + goto err_firmware;
> +
> + INIT_LIST_HEAD(&sdma_dma_device->channels);
> +
> + /* Initialize channel parameters */
> + for (i = 0; i < MAX_DMA_CHANNELS; i++) {
> + struct sdma_channel *sdma = &sdma_data[i];
> +
> + spin_lock_init(&sdma->lock);
> +
> + dma_cap_set(DMA_SLAVE, sdma_dma_device->cap_mask);
> + dma_cap_set(DMA_CYCLIC, sdma_dma_device->cap_mask);
> +
> + sdma->chan.device = sdma_dma_device;
> + sdma->chan.chan_id = i;
> +
> + /* Add the channel to the DMAC list */
> + list_add_tail(&sdma->chan.device_node, &sdma_dma_device->channels);
> + }
> +
> + sdma_dma_device->dev = &pdev->dev;
> +
> + sdma_dma_device->device_alloc_chan_resources = sdma_alloc_chan_resources;
> + sdma_dma_device->device_free_chan_resources = sdma_free_chan_resources;
> + sdma_dma_device->device_tx_status = sdma_tx_status;
> + sdma_dma_device->device_prep_slave_sg = sdma_prep_slave_sg;
> + sdma_dma_device->device_prep_dma_cyclic = sdma_prep_dma_cyclic;
> + sdma_dma_device->device_control = sdma_control;
> + sdma_dma_device->device_issue_pending = sdma_issue_pending;
> +
> + ret = dma_async_device_register(sdma_dma_device);
> + if (ret) {
> + dev_err(&pdev->dev, "unable to register DMAC\n");
SDMAC even?
> + goto err_firmware;
> + }
> +
> + dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
> + header->version_major,
> + header->version_minor);
> +
> + /* request channel 0. This is an internal control channel
> + * to the SDMA engine and not available to clients.
> + */
> + dma_cap_zero(mask);
> + dma_cap_set(DMA_SLAVE, mask);
> + dma_request_channel(mask, NULL, NULL);
> +
> + release_firmware(fw);
> +
> + return 0;
> +
> +err_firmware:
> + release_firmware(fw);
> +err_cputype:
> + clk_put(sdma_clk);
> +err_clk:
> + return 0;
> +}
> +
> +static int __devexit sdma_remove(struct platform_device *pdev)
> +{
> + return -EBUSY;
> +}
> +
> +static struct platform_driver sdma_driver = {
> + .driver = {
> + .name = "imx-sdma",
> + },
> + .probe = sdma_probe,
> + .remove = __devexit_p(sdma_remove),
> +};
> +
> +static int __init sdma_module_init(void)
> +{
> + return platform_driver_register(&sdma_driver);
> +}
> +subsys_initcall(sdma_module_init);
> +
> +MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
> +MODULE_DESCRIPTION("i.MX SDMA driver");
> +MODULE_LICENSE("GPL");
> --
> 1.7.1
Thanks for using this API
Sascha!
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-16 12:21 ` Linus Walleij
0 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-16 12:21 UTC (permalink / raw)
To: linux-arm-kernel
2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> This patch adds support for the Freescale i.MX SDMA engine.
I like it!
> The SDMA engine is a scatter/gather DMA engine which is implemented
> as a seperate coprocessor. SDMA needs its own firmware which is
> requested using the standard request_firmware mechanism. The firmware
> has different entry points for each peripheral type, so drivers
> have to pass the peripheral type to the DMA engine which in turn
> picks the correct firmware entry point from a table contained in
> the firmware image itself.
Quite fun, if the spec for the microcode is open this opens up
for dynamic firmware generation for specific DMA jobs does it
not?
> I took a very simple approach to implement dmaengine support. Only
> a single descriptor is statically assigned to a each channel. This
> means that transfers can't be queued up but only a single transfer
> is in progress. This simplifies implementation a lot and is sufficient
> for the usual device/memory transfers.
If you want to add memcpy() capability later you're gonna need
this I think, but you can take that when that need arise.
>(...)
> +++ b/arch/arm/plat-mxc/include/mach/dma.h
> @@ -0,0 +1,64 @@
> +/*
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 as
> + * published by the Free Software Foundation.
> + */
> +
> +#ifndef __ASM_ARCH_MXC_DMA_H__
> +#define __ASM_ARCH_MXC_DMA_H__
> +
> +#include <linux/scatterlist.h>
> +
> +/*
> + * This enumerates peripheral types. Used for SDMA.
> + */
> +typedef enum {
The kernel is not really happy about typedefs, can't this be a
regular enum?
> + ? ? ? IMX_DMATYPE_SSI, ? ? ? ?/* MCU domain SSI */
> + ? ? ? IMX_DMATYPE_SSI_SP, ? ? /* Shared SSI */
> + ? ? ? IMX_DMATYPE_MMC, ? ? ? ?/* MMC */
> + ? ? ? IMX_DMATYPE_SDHC, ? ? ? /* SDHC */
> + ? ? ? IMX_DMATYPE_UART, ? ? ? /* MCU domain UART */
> + ? ? ? IMX_DMATYPE_UART_SP, ? ?/* Shared UART */
> + ? ? ? IMX_DMATYPE_FIRI, ? ? ? /* FIRI */
> + ? ? ? IMX_DMATYPE_CSPI, ? ? ? /* MCU domain CSPI */
> + ? ? ? IMX_DMATYPE_CSPI_SP, ? ?/* Shared CSPI */
> + ? ? ? IMX_DMATYPE_SIM, ? ? ? ?/* SIM */
> + ? ? ? IMX_DMATYPE_ATA, ? ? ? ?/* ATA */
> + ? ? ? IMX_DMATYPE_CCM, ? ? ? ?/* CCM */
> + ? ? ? IMX_DMATYPE_EXT, ? ? ? ?/* External peripheral */
> + ? ? ? IMX_DMATYPE_MSHC, ? ? ? /* Memory Stick Host Controller */
> + ? ? ? IMX_DMATYPE_MSHC_SP, ? ?/* Shared Memory Stick Host Controller */
> + ? ? ? IMX_DMATYPE_DSP, ? ? ? ?/* DSP */
> + ? ? ? IMX_DMATYPE_MEMORY, ? ? /* Memory */
> + ? ? ? IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
> + ? ? ? IMX_DMATYPE_SPDIF, ? ? ?/* SPDIF */
> + ? ? ? IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
> + ? ? ? IMX_DMATYPE_ASRC, ? ? ? /* ASRC */
> + ? ? ? IMX_DMATYPE_ESAI, ? ? ? /* ESAI */
> +} sdma_peripheral_type;
> +
> +enum imx_dma_prio {
> + ? ? ? DMA_PRIO_HIGH = 0,
> + ? ? ? DMA_PRIO_MEDIUM = 1,
> + ? ? ? DMA_PRIO_LOW = 2
> +};
> +
> +struct imx_dma_data {
> + ? ? ? int dma_request; /* DMA request line */
Can this be negative and what is the range? I would
suspect something like u8 or u16 would surely be more
apropriate...
> + ? ? ? sdma_peripheral_type peripheral_type;
> + ? ? ? int priority;
Isn't this an enum imx_dma_prio?
> +};
> +
> +static inline int imx_dma_is_ipu(struct dma_chan *chan)
> +{
> + ? ? ? return !strcmp(dev_name(chan->device->dev), "ipu-core");
> +}
> +
> +static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
> +{
> + ? ? ? return !strcmp(dev_name(chan->device->dev), "imx-sdma");
> +}
> +
> +#endif
> diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
> new file mode 100644
> index 0000000..5d542b8
> --- /dev/null
> +++ b/arch/arm/plat-mxc/include/mach/sdma.h
> @@ -0,0 +1,8 @@
> +#ifndef __MACH_MXC_SDMA_H__
> +#define __MACH_MXC_SDMA_H__
> +
> +struct sdma_platform_data {
> + ? ? ? int sdma_version;
Do you have negative versions or can it be unsigned?
> +};
> +
> +#endif /* __MACH_MXC_SDMA_H__ */
> diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
> index 9520cf0..f76bda9 100644
> --- a/drivers/dma/Kconfig
> +++ b/drivers/dma/Kconfig
> @@ -195,6 +195,14 @@ config PCH_DMA
> ? ? ? ?help
> ? ? ? ? ?Enable support for the Topcliff PCH DMA engine.
>
> +config IMX_SDMA
> + ? ? ? tristate "Atmel AHB DMA support"
> + ? ? ? depends on ARCH_MXC
> + ? ? ? select DMA_ENGINE
> + ? ? ? help
> + ? ? ? ? Support the i.MX SDMA engine. This engine is integrated into
> + ? ? ? ? Freescale i.MX25/31/35/51 chips.
> +
> ?config DMA_ENGINE
> ? ? ? ?bool
>
> diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
> index 72bd703..14d7a1b 100644
> --- a/drivers/dma/Makefile
> +++ b/drivers/dma/Makefile
> @@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
> ?obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
> ?obj-$(CONFIG_PL330_DMA) += pl330.o
> ?obj-$(CONFIG_PCH_DMA) += pch_dma.o
> +obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
> diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> new file mode 100644
> index 0000000..3ba7905
> --- /dev/null
> +++ b/drivers/dma/imx-sdma.c
> @@ -0,0 +1,1383 @@
> +/*
> + * drivers/dma/imx-sdma.c
> + *
> + * This file contains a driver for the Freescale Smart DMA engine
> + *
> + * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
> + *
> + * Based on code from Freescale:
> + *
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * The code contained herein is licensed under the GNU General Public
> + * License. You may obtain a copy of the GNU General Public License
> + * Version 2 or later at the following locations:
> + *
> + * http://www.opensource.org/licenses/gpl-license.html
> + * http://www.gnu.org/copyleft/gpl.html
> + */
> +#include <linux/init.h>
> +#include <linux/types.h>
> +#include <linux/mm.h>
> +#include <linux/interrupt.h>
> +#include <linux/clk.h>
> +#include <linux/semaphore.h>
> +#include <linux/spinlock.h>
> +#include <linux/device.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/firmware.h>
> +#include <linux/slab.h>
> +#include <linux/platform_device.h>
> +#include <linux/dmaengine.h>
> +
> +#include <asm/irq.h>
> +#include <mach/sdma.h>
> +#include <mach/dma.h>
> +#include <mach/hardware.h>
> +
> +/* SDMA registers */
> +#define SDMA_H_C0PTR ? ? ? ? ? (sdma_base + 0x000)
> +#define SDMA_H_INTR ? ? ? ? ? ?(sdma_base + 0x004)
> +#define SDMA_H_STATSTOP ? ? ? ? ? ? ? ?(sdma_base + 0x008)
> +#define SDMA_H_START ? ? ? ? ? (sdma_base + 0x00c)
> +#define SDMA_H_EVTOVR ? ? ? ? ?(sdma_base + 0x010)
> +#define SDMA_H_DSPOVR ? ? ? ? ?(sdma_base + 0x014)
> +#define SDMA_H_HOSTOVR ? ? ? ? (sdma_base + 0x018)
> +#define SDMA_H_EVTPEND ? ? ? ? (sdma_base + 0x01c)
> +#define SDMA_H_DSPENBL ? ? ? ? (sdma_base + 0x020)
> +#define SDMA_H_RESET ? ? ? ? ? (sdma_base + 0x024)
> +#define SDMA_H_EVTERR ? ? ? ? ?(sdma_base + 0x028)
> +#define SDMA_H_INTRMSK ? ? ? ? (sdma_base + 0x02c)
> +#define SDMA_H_PSW ? ? ? ? ? ? (sdma_base + 0x030)
> +#define SDMA_H_EVTERRDBG ? ? ? (sdma_base + 0x034)
> +#define SDMA_H_CONFIG ? ? ? ? ?(sdma_base + 0x038)
> +#define SDMA_ONCE_ENB ? ? ? ? ?(sdma_base + 0x040)
> +#define SDMA_ONCE_DATA ? ? ? ? (sdma_base + 0x044)
> +#define SDMA_ONCE_INSTR ? ? ? ? ? ? ? ?(sdma_base + 0x048)
> +#define SDMA_ONCE_STAT ? ? ? ? (sdma_base + 0x04c)
> +#define SDMA_ONCE_CMD ? ? ? ? ?(sdma_base + 0x050)
> +#define SDMA_EVT_MIRROR ? ? ? ? ? ? ? ?(sdma_base + 0x054)
> +#define SDMA_ILLINSTADDR ? ? ? (sdma_base + 0x058)
> +#define SDMA_CHN0ADDR ? ? ? ? ?(sdma_base + 0x05c)
> +#define SDMA_ONCE_RTB ? ? ? ? ?(sdma_base + 0x060)
> +#define SDMA_XTRIG_CONF1 ? ? ? (sdma_base + 0x070)
> +#define SDMA_XTRIG_CONF2 ? ? ? (sdma_base + 0x074)
> +#define SDMA_CHNENBL_0 ? ? ? ? (sdma_base + (sdma_version == 2 ? 0x200 : 0x80))
> +#define SDMA_CHNPRI_0 ? ? ? ? ?(sdma_base + 0x100)
All these rely on a fixed sdma_base which makes the driver
a singleton. This is not so good if you imagine the situation with a
platform with two SDMA engines on different addresses.
Can't you create a runtime allocated stateholder to hold
the base and access relative to the offset?
> +
> +/*
> + * Buffer descriptor status values.
> + */
> +#define BD_DONE ?0x01
> +#define BD_WRAP ?0x02
> +#define BD_CONT ?0x04
> +#define BD_INTR ?0x08
> +#define BD_RROR ?0x10
> +#define BD_LAST ?0x20
> +#define BD_EXTD ?0x80
> +
> +/*
> + * Data Node descriptor status values.
> + */
> +#define DND_END_OF_FRAME ?0x80
> +#define DND_END_OF_XFER ? 0x40
> +#define DND_DONE ? ? ? ? ?0x20
> +#define DND_UNUSED ? ? ? ?0x01
> +
> +/*
> + * IPCV2 descriptor status values.
> + */
> +#define BD_IPCV2_END_OF_FRAME ?0x40
> +
> +#define IPCV2_MAX_NODES ? ? ? ?50
> +/*
> + * Error bit set in the CCB status field by the SDMA,
> + * in setbd routine, in case of a transfer error
> + */
> +#define DATA_ERROR ?0x10000000
> +
> +/*
> + * Buffer descriptor commands.
> + */
> +#define C0_ADDR ? ? ? ? ? ? 0x01
> +#define C0_LOAD ? ? ? ? ? ? 0x02
> +#define C0_DUMP ? ? ? ? ? ? 0x03
> +#define C0_SETCTX ? ? ? ? ? 0x07
> +#define C0_GETCTX ? ? ? ? ? 0x03
> +#define C0_SETDM ? ? ? ? ? ?0x01
> +#define C0_SETPM ? ? ? ? ? ?0x04
> +#define C0_GETDM ? ? ? ? ? ?0x02
> +#define C0_GETPM ? ? ? ? ? ?0x08
> +/*
> + * Change endianness indicator in the BD command field
> + */
> +#define CHANGE_ENDIANNESS ? 0x80
> +
> +/*
> + * Mode/Count of data node descriptors - IPCv2
> + */
> +#ifdef __BIG_ENDIAN
> +struct sdma_mode_count {
> + ? ? ? u32 command : ?8; /* command mostlky used for channel 0 */
There are a lot of inline commented struct members, please
use kerneldoc, that's simple. (Applies all over the patch...)
Documentation/kernel-doc-nano-HOWTO
> + ? ? ? u32 status ?: ?8; /* E,R,I,C,W,D status bits stored here */
> + ? ? ? u32 count ? : 16; /* size of the buffer pointed by this BD */
> +};
> +#else
> +struct sdma_mode_count {
> + ? ? ? u32 count ? : 16; /* size of the buffer pointed by this BD */
> + ? ? ? u32 status ?: ?8; /* E,R,I,C,W,D status bits stored here */
> + ? ? ? u32 command : ?8; /* command mostlky used for channel 0 */
> +};
> +#endif
This use of #ifdef is odd to me but others are probably more
experienced. Anyway, the way it is used with different
:n suffixes makes me believe that you need a packed
compiler directive for this layout to be explicitly coherent.
Atleast add some comment on what this #ifdef construction
does so guys like me can understand what's going on.
> +
> +/*
> + * Buffer descriptor
> + */
> +struct sdma_buffer_descriptor {
> + ? ? ? struct sdma_mode_count ?mode;
> + ? ? ? u32 buffer_addr; ? ?/* address of the buffer described */
> + ? ? ? u32 ext_buffer_addr; /* extended buffer address */
Shouldn't these be dma_addr_t? OK that's probably u32
anyway but just to make a marker...
> +};
> +
> +/*
> + * Channel control Block
> + */
> +struct sdma_channel_control {
> + ? ? ? u32 current_bd_ptr; /* current buffer descriptor processed */
> + ? ? ? u32 base_bd_ptr; ? ?/* first element of buffer descriptor array */
> + ? ? ? void *unused;
> + ? ? ? void *unused1;
Hm, can you comment on what these unused things are for...?
> +};
> +
> +/**
> + * Context structure.
> + */
> +#ifdef __BIG_ENDIAN
> +struct sdma_state_registers {
> + ? ? ? u32 sf ? ? : 1; /* source fault while loading data */
> + ? ? ? u32 unused0: 1;
> + ? ? ? u32 rpc ? ?:14; /* return program counter */
> + ? ? ? u32 t ? ? ?: 1; /* test bit:status of arithmetic & test instruction*/
> + ? ? ? u32 unused1: 1;
> + ? ? ? u32 pc ? ? :14; /* program counter */
> + ? ? ? u32 lm ? ? : 2; /* loop mode */
> + ? ? ? u32 epc ? ?:14; /* loop end program counter */
> + ? ? ? u32 df ? ? : 1; /* destination fault while storing data */
> + ? ? ? u32 unused2: 1;
> + ? ? ? u32 spc ? ?:14; /* loop start program counter */
> +};
> +#else
> +struct sdma_state_registers {
> + ? ? ? u32 pc ? ? :14; /* program counter */
> + ? ? ? u32 unused1: 1;
> + ? ? ? u32 t ? ? ?: 1; /* test bit: status of arithmetic & test instruction*/
> + ? ? ? u32 rpc ? ?:14; /* return program counter */
> + ? ? ? u32 unused0: 1;
> + ? ? ? u32 sf ? ? : 1; /* source fault while loading data */
> + ? ? ? u32 spc ? ?:14; /* loop start program counter */
> + ? ? ? u32 unused2: 1;
> + ? ? ? u32 df ? ? : 1; /* destination fault while storing data */
> + ? ? ? u32 epc ? ?:14; /* loop end program counter */
> + ? ? ? u32 lm ? ? : 2; /* loop mode */
> +};
> +#endif
Again this is odd to me...
> +
> +struct sdma_context_data {
> + ? ? ? struct sdma_state_registers ?channel_state; /* channel state bits */
> + ? ? ? u32 ?gReg[8]; /* general registers */
> + ? ? ? u32 ?mda; /* burst dma destination address register */
> + ? ? ? u32 ?msa; /* burst dma source address register */
> + ? ? ? u32 ?ms; ?/* burst dma status ?register */
> + ? ? ? u32 ?md; ?/* burst dma data ? ?register */
> + ? ? ? u32 ?pda; /* peripheral dma destination address register */
> + ? ? ? u32 ?psa; /* peripheral dma source address register */
> + ? ? ? u32 ?ps; ?/* peripheral dma ?status ?register */
> + ? ? ? u32 ?pd; ?/* peripheral dma ?data ? ?register */
> + ? ? ? u32 ?ca; ?/* CRC polynomial ?register */
> + ? ? ? u32 ?cs; ?/* CRC accumulator register */
> + ? ? ? u32 ?dda; /* dedicated core destination address register */
> + ? ? ? u32 ?dsa; /* dedicated core source address register */
> + ? ? ? u32 ?ds; ?/* dedicated core status ?register */
> + ? ? ? u32 ?dd; ?/* dedicated core data ? ?register */
> + ? ? ? u32 ?scratch0;
> + ? ? ? u32 ?scratch1;
> + ? ? ? u32 ?scratch2;
> + ? ? ? u32 ?scratch3;
> + ? ? ? u32 ?scratch4;
> + ? ? ? u32 ?scratch5;
> + ? ? ? u32 ?scratch6;
> + ? ? ? u32 ?scratch7;
> +};
> +
> +struct sdma_channel {
> + ? ? ? /* Channel number */
> + ? ? ? int channel;
Unsigned?
> + ? ? ? /* Transfer type. Needed for setting SDMA script */
> + ? ? ? enum dma_data_direction direction;
> + ? ? ? /* Peripheral type. Needed for setting SDMA script */
> + ? ? ? sdma_peripheral_type peripheral_type;
> + ? ? ? /* Peripheral event id */
> + ? ? ? int event_id;
Unsigned?
> + ? ? ? /* Peripheral event id2 (for channels that use 2 events) */
> + ? ? ? int event_id2;
Unsigned?
> + ? ? ? /* SDMA data access word size */
> + ? ? ? unsigned long word_size;
Is this in bits, bytes etc? Isn't e.g. an u8 enough to hold this,
and further, isn't it possible to recycle enum dma_slave_buswidth
from dmaengine.h instead?
> +
> + ? ? ? /* ID of the buffer that was processed */
> + ? ? ? unsigned int buf_tail;
> +
> + ? ? ? wait_queue_head_t waitq; ? ? ? ?/* channel completion waitqeue */
> +
> + ? ? ? int num_bd;
Unsigned? Range?
> +
> + ? ? ? struct sdma_buffer_descriptor *bd;
> + ? ? ? dma_addr_t ? ? ?bd_phys;
> +
> + ? ? ? int pc_from_device, pc_to_device;
Unsigned?
> +
> + ? ? ? unsigned long flags;
Is this an u32?
> + ? ? ? dma_addr_t per_address;
> +
> + ? ? ? uint32_t event_mask1, event_mask2;
> + ? ? ? uint32_t watermark_level;
> + ? ? ? uint32_t shp_addr, per_addr;
> +
> + ? ? ? /* DMA-Engine Channel */
> + ? ? ? struct dma_chan chan;
> +
> + ? ? ? spinlock_t ? ? ? ? ? ? ?lock;
> + ? ? ? struct dma_async_tx_descriptor desc;
> + ? ? ? dma_cookie_t ? ? ? ? ? ?last_completed;
> + ? ? ? int busy;
Shouldn't this be a bool?
> +};
> +
> +#define IMX_DMA_SG_LOOP ? ? ? ? ? ? ? ?(1 << 0)
> +
> +#define MAX_DMA_CHANNELS 32
> +#define MXC_SDMA_DEFAULT_PRIORITY 1
> +#define MXC_SDMA_MIN_PRIORITY 1
> +#define MXC_SDMA_MAX_PRIORITY 7
> +
> +/*
> + * This enumerates transfer types
> + */
> +typedef enum {
Again a typedef, please plain enum is fine.
> + ? ? ? emi_2_per = 0, ? ? ? ? ?/* EMI memory to peripheral */
> + ? ? ? emi_2_int, ? ? ? ? ? ? ?/* EMI memory to internal RAM */
> + ? ? ? emi_2_emi, ? ? ? ? ? ? ?/* EMI memory to EMI memory */
> + ? ? ? emi_2_dsp, ? ? ? ? ? ? ?/* EMI memory to DSP memory */
> + ? ? ? per_2_int, ? ? ? ? ? ? ?/* Peripheral to internal RAM */
> + ? ? ? per_2_emi, ? ? ? ? ? ? ?/* Peripheral to internal EMI memory */
> + ? ? ? per_2_dsp, ? ? ? ? ? ? ?/* Peripheral to DSP memory */
> + ? ? ? per_2_per, ? ? ? ? ? ? ?/* Peripheral to Peripheral */
> + ? ? ? int_2_per, ? ? ? ? ? ? ?/* Internal RAM to peripheral */
> + ? ? ? int_2_int, ? ? ? ? ? ? ?/* Internal RAM to Internal RAM */
> + ? ? ? int_2_emi, ? ? ? ? ? ? ?/* Internal RAM to EMI memory */
> + ? ? ? int_2_dsp, ? ? ? ? ? ? ?/* Internal RAM to DSP memory */
> + ? ? ? dsp_2_per, ? ? ? ? ? ? ?/* DSP memory to peripheral */
> + ? ? ? dsp_2_int, ? ? ? ? ? ? ?/* DSP memory to internal RAM */
> + ? ? ? dsp_2_emi, ? ? ? ? ? ? ?/* DSP memory to EMI memory */
> + ? ? ? dsp_2_dsp, ? ? ? ? ? ? ?/* DSP memory to DSP memory */
> + ? ? ? emi_2_dsp_loop, ? ? ? ? /* EMI memory to DSP memory loopback */
> + ? ? ? dsp_2_emi_loop, ? ? ? ? /* DSP memory to EMI memory loopback */
> + ? ? ? dvfs_pll, ? ? ? ? ? ? ? /* DVFS script with PLL change ? ? ? */
> + ? ? ? dvfs_pdr ? ? ? ? ? ? ? ?/* DVFS script without PLL change ? ?*/
> +} sdma_transfer_type;
> +
> +/*
> + * Structure containing sdma request ?parameters.
> + */
> +struct sdma_script_start_addrs {
> + ? ? ? int ap_2_ap_addr;
> + ? ? ? int ap_2_bp_addr;
> + ? ? ? int ap_2_ap_fixed_addr;
> + ? ? ? int bp_2_ap_addr;
> + ? ? ? int loopback_on_dsp_side_addr;
> + ? ? ? int mcu_interrupt_only_addr;
> +
> + ? ? ? int firi_2_per_addr;
> + ? ? ? int firi_2_mcu_addr;
> + ? ? ? int per_2_firi_addr;
> + ? ? ? int mcu_2_firi_addr;
> +
> + ? ? ? int uart_2_per_addr;
> + ? ? ? int uart_2_mcu_addr;
> + ? ? ? int per_2_app_addr;
> + ? ? ? int mcu_2_app_addr;
> + ? ? ? int per_2_per_addr;
> +
> + ? ? ? int uartsh_2_per_addr;
> + ? ? ? int uartsh_2_mcu_addr;
> + ? ? ? int per_2_shp_addr;
> + ? ? ? int mcu_2_shp_addr;
> +
> + ? ? ? int ata_2_mcu_addr;
> + ? ? ? int mcu_2_ata_addr;
> +
> + ? ? ? int app_2_per_addr;
> + ? ? ? int app_2_mcu_addr;
> + ? ? ? int shp_2_per_addr;
> + ? ? ? int shp_2_mcu_addr;
> +
> + ? ? ? int mshc_2_mcu_addr;
> + ? ? ? int mcu_2_mshc_addr;
> +
> + ? ? ? int spdif_2_mcu_addr;
> + ? ? ? int mcu_2_spdif_addr;
> +
> + ? ? ? int asrc_2_mcu_addr;
> +
> + ? ? ? int ext_mem_2_ipu_addr;
> +
> + ? ? ? int descrambler_addr;
> +
> + ? ? ? int dptc_dvfs_addr;
> +
> + ? ? ? int utra_addr;
> +
> + ? ? ? int ram_code_start_addr;
All these addresses, are they really integers with
valid negative values... Aren't they dma_addr_t or
atleast u32?
> +};
> +
> +#define SDMA_FIRMWARE_MAGIC 0x414d4453
> +
> +struct sdma_firmware_header {
> + ? ? ? uint32_t ? ? ? ?magic; /* "SDMA" */
> + ? ? ? uint32_t ? ? ? ?version_major; ?/* increased whenever layout of struct sdma_script_start_addrs changes */
> + ? ? ? uint32_t ? ? ? ?version_minor; ?/* firmware version */
> + ? ? ? uint32_t ? ? ? ?script_addrs_start; /* offset of struct sdma_script_start_addrs in this image */
> + ? ? ? uint32_t ? ? ? ?num_script_addrs; /* Number of script addresses in this image */
> + ? ? ? uint32_t ? ? ? ?ram_code_start; /* offset of SDMA ram image in this firmware image */
> + ? ? ? uint32_t ? ? ? ?ram_code_size; /* size of SDMA ram image */
Please use u32. uint32_t is not the preferred kernel type.
(Still I've seen people use it in some cases so I might be wrong,
feel welcome to bit back on this.)
> +};
> +
> +static struct sdma_channel sdma_data[MAX_DMA_CHANNELS];
> +static struct sdma_channel_control *channel_control;
> +static void __iomem *sdma_base;
> +static int sdma_version;
Unsigned?
> +static int sdma_num_events;
Unsigned?
> +static struct sdma_context_data *sdma_context;
> +dma_addr_t sdma_context_phys;
> +static struct dma_device __sdma_dma_device;
> +static struct dma_device *sdma_dma_device = &__sdma_dma_device;
This is what I suspected: local variables making the entire driver
a singleton, which means you can never have more than one
SDMA. Atleast collect all of these in a struct, call it
"struct sdma" simply (if you ask me) and use as a stateholder.
This makes it easier to kzalloc() that struct later if you
want to support non-singletons.
I know this require some work but I've done it to several drivers
(always asked on mailinglists to do this) and I don't regret a single
rewrite. Last time was for the PL18x DMAengine driver actually.
> +
> +#define SDMA_H_CONFIG_DSPDMA ? (1 << 12) /* indicates if the DSPDMA is used */
> +#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
> +#define SDMA_H_CONFIG_ACR ? ? ?(1 << 4) ?/* indicates if AHB freq /core freq = 2 or 1 */
> +#define SDMA_H_CONFIG_CSM ? ? ?(3) ? ? ? /* indicates which context switch mode is selected*/
> +
> +static int sdma_config_ownership(int channel, int event_override,
> + ? ? ? ? ? ? ? ? ?int mcu_verride, int dsp_override)
> +{
> + ? ? ? u32 evt, mcu, dsp;
> +
> + ? ? ? if (event_override && mcu_verride && dsp_override)
> + ? ? ? ? ? ? ? return -EINVAL;
> +
> + ? ? ? evt = readl(SDMA_H_EVTOVR);
> + ? ? ? mcu = readl(SDMA_H_HOSTOVR);
> + ? ? ? dsp = readl(SDMA_H_DSPOVR);
> +
> + ? ? ? if (dsp_override)
> + ? ? ? ? ? ? ? dsp &= ~(1 << channel);
> + ? ? ? else
> + ? ? ? ? ? ? ? dsp |= (1 << channel);
> +
> + ? ? ? if (event_override)
> + ? ? ? ? ? ? ? evt &= ~(1 << channel);
> + ? ? ? else
> + ? ? ? ? ? ? ? evt |= (1 << channel);
> +
> + ? ? ? if (mcu_verride)
> + ? ? ? ? ? ? ? mcu &= ~(1 << channel);
> + ? ? ? else
> + ? ? ? ? ? ? ? mcu |= (1 << channel);
> +
> + ? ? ? writel(evt, SDMA_H_EVTOVR);
> + ? ? ? writel(mcu, SDMA_H_HOSTOVR);
> + ? ? ? writel(dsp, SDMA_H_DSPOVR);
> +
> + ? ? ? return 0;
> +}
> +
> +/*
> + * sdma_run_channel - run a channel and wait till it's done
> + */
> +static int sdma_run_channel(int channel)
> +{
> + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> + ? ? ? int ret;
> +
> + ? ? ? writel(1 << channel, SDMA_H_START);
> +
> + ? ? ? ret = wait_event_interruptible(sdma->waitq,
> + ? ? ? ? ? ? ? ? ? ? ? !(readl(SDMA_H_STATSTOP) & (1 << channel)));
OK not the biggest thing in the world, but can't you use a
completion for this? (I'm not so clever with waitqueues so
forgive me if this is malinformed.)
> + ? ? ? return ret;
> +}
> +
> +static int sdma_load_script(void *buf, int size, u32 address)
> +{
> + ? ? ? struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> + ? ? ? void *buf_virt;
> + ? ? ? dma_addr_t buf_phys;
> + ? ? ? int ret;
> +
> + ? ? ? buf_virt = dma_alloc_coherent(NULL,
> + ? ? ? ? ? ? ? ? ? ? ? size,
> + ? ? ? ? ? ? ? ? ? ? ? &buf_phys, GFP_KERNEL);
> + ? ? ? if (!buf_virt)
> + ? ? ? ? ? ? ? return -ENOMEM;
> +
> + ? ? ? bd0->mode.command = C0_SETPM;
> + ? ? ? bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> + ? ? ? bd0->mode.count = size / 2;
> + ? ? ? bd0->buffer_addr = buf_phys;
> + ? ? ? bd0->ext_buffer_addr = address;
> +
> + ? ? ? memcpy(buf_virt, buf, size);
> +
> + ? ? ? ret = sdma_run_channel(0);
> +
> + ? ? ? dma_free_coherent(NULL, size, buf_virt, buf_phys);
> +
> + ? ? ? return ret;
> +}
> +
> +static void sdma_event_enable(int channel, int event)
> +{
> + ? ? ? u32 val;
> +
> + ? ? ? val = readl(SDMA_CHNENBL_0 + event * 4);
This use indicates that event should probably be
unsigned, and probably not greater than u16 atleast.
I suspect it is never more than an u8 really.
> + ? ? ? val |= (1 << channel);
> + ? ? ? writel(val, SDMA_CHNENBL_0 + event * 4);
> +}
> +
> +static void sdma_event_disable(int channel, int event)
> +{
> + ? ? ? u32 val;
> +
> + ? ? ? val = readl(SDMA_CHNENBL_0 + event * 4);
> + ? ? ? val &= ~(1 << channel);
> + ? ? ? writel(val, SDMA_CHNENBL_0 + event * 4);
Same comment here.
> +}
> +
> +static void mxc_sdma_handle_channel_loop(int channel)
> +{
> + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
This indicates that channel should be unsigned.
> + ? ? ? struct sdma_buffer_descriptor *bd;
> + ? ? ? int error = 0;
Unused variable?
> +
> + ? ? ? /*
> + ? ? ? ?* loop mode. Iterate over descriptors, re-setup them and
> + ? ? ? ?* call callback function.
> + ? ? ? ?*/
> + ? ? ? while (1) {
> + ? ? ? ? ? ? ? bd = &sdma->bd[sdma->buf_tail];
> +
> + ? ? ? ? ? ? ? if (bd->mode.status & BD_DONE)
> + ? ? ? ? ? ? ? ? ? ? ? break;
> +
> + ? ? ? ? ? ? ? if (bd->mode.status & BD_RROR)
> + ? ? ? ? ? ? ? ? ? ? ? error = -EIO;
> +
> + ? ? ? ? ? ? ? bd->mode.status |= BD_DONE;
> + ? ? ? ? ? ? ? sdma->buf_tail++;
> + ? ? ? ? ? ? ? sdma->buf_tail %= sdma->num_bd;
> +
> + ? ? ? ? ? ? ? if (sdma->desc.callback)
> + ? ? ? ? ? ? ? ? ? ? ? sdma->desc.callback(sdma->desc.callback_param);
> + ? ? ? }
> +}
> +
> +static void mxc_sdma_handle_channel_normal(int channel)
> +{
> + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> + ? ? ? struct sdma_buffer_descriptor *bd;
> + ? ? ? int i, error = 0;
> +
> + ? ? ? /*
> + ? ? ? ?* non loop mode. Iterate over all descriptors, collect
> + ? ? ? ?* errors and call callback function
> + ? ? ? ?*/
> + ? ? ? for (i = 0; i < sdma->num_bd; i++) {
> + ? ? ? ? ? ? ? bd = &sdma->bd[i];
> +
> + ? ? ? ? ? ? ? ?if (bd->mode.status & (BD_DONE | BD_RROR))
> + ? ? ? ? ? ? ? ? ? ? ? error = -EIO;
> + ? ? ? }
> +
> + ? ? ? if (sdma->desc.callback)
> + ? ? ? ? ? ? ? sdma->desc.callback(sdma->desc.callback_param);
> + ? ? ? sdma->last_completed = sdma->desc.cookie;
> +
> + ? ? ? sdma->busy = 0;
= true if you switch this to bool..
> +}
> +
> +static void mxc_sdma_handle_channel(int channel)
> +{
> + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> +
> + ? ? ? wake_up_interruptible(&sdma->waitq);
> +
> + ? ? ? /* not interested in channel 0 interrupts */
> + ? ? ? if (!channel)
> + ? ? ? ? ? ? ? return;
> +
> + ? ? ? if (sdma->flags & IMX_DMA_SG_LOOP)
> + ? ? ? ? ? ? ? mxc_sdma_handle_channel_loop(channel);
> + ? ? ? else
> + ? ? ? ? ? ? ? mxc_sdma_handle_channel_normal(channel);
> +}
> +
> +static irqreturn_t sdma_int_handler(int irq, void *dev_id)
> +{
> + ? ? ? u32 stat;
> +
> + ? ? ? stat = readl(SDMA_H_INTR);
> + ? ? ? writel(stat, SDMA_H_INTR);
> +
> + ? ? ? while (stat) {
> + ? ? ? ? ? ? ? int channel = fls(stat) - 1;
> +
> + ? ? ? ? ? ? ? mxc_sdma_handle_channel(channel);
> +
> + ? ? ? ? ? ? ? stat &= ~(1 << channel);
> + ? ? ? }
> +
> + ? ? ? return IRQ_HANDLED;
> +}
> +
> +static struct clk *sdma_clk;
> +
> +/*
> + * Stores the start address of the SDMA scripts
> + */
> +static struct sdma_script_start_addrs __sdma_script_addrs;
> +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
> +
> +/*
> + * sets the pc of SDMA script according to the peripheral type
> + */
> +static void sdma_get_pc(struct sdma_channel *sdma,
> + ? ? ? ? ? ? ? sdma_peripheral_type peripheral_type)
> +{
> + ? ? ? int res = 0;
> + ? ? ? int per_2_emi = 0, emi_2_per = 0;
> + ? ? ? int per_2_int = 0, int_2_per = 0;
> + ? ? ? int per_2_per = 0, emi_2_emi = 0;
> +
> + ? ? ? sdma->pc_from_device = 0;
> + ? ? ? sdma->pc_to_device = 0;
There are a *lot* of local variables here, and only two of them
are used eventually, at the end of the function. I cannot quite
follow this, what is going on?
Some like emi_2_emi seem to be totally unused.
The types here look like some kind of enum or other
similar construction is really what's being asked for
here.
> +
> + ? ? ? switch (peripheral_type) {
> + ? ? ? case IMX_DMATYPE_MEMORY:
> + ? ? ? ? ? ? ? emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_DSP:
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->bp_2_ap_addr;
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->ap_2_bp_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_FIRI:
> + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->firi_2_per_addr;
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
> + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_firi_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_UART:
> + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->uart_2_per_addr;
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
> + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_app_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_UART_SP:
> + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->uartsh_2_per_addr;
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
> + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_shp_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_ATA:
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_CSPI:
> + ? ? ? case IMX_DMATYPE_EXT:
> + ? ? ? case IMX_DMATYPE_SSI:
> + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->app_2_per_addr;
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->app_2_mcu_addr;
> + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_app_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_SSI_SP:
> + ? ? ? case IMX_DMATYPE_MMC:
> + ? ? ? case IMX_DMATYPE_SDHC:
> + ? ? ? case IMX_DMATYPE_CSPI_SP:
> + ? ? ? case IMX_DMATYPE_ESAI:
> + ? ? ? case IMX_DMATYPE_MSHC_SP:
> + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->shp_2_per_addr;
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
> + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_shp_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_ASRC:
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
> + ? ? ? ? ? ? ? per_2_per = sdma_script_addrs->per_2_per_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_MSHC:
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_CCM:
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_FIFO_MEMORY:
> + ? ? ? ? ? ? ? res = sdma_script_addrs->ap_2_ap_fixed_addr;
res? This thing is never used.
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_SPDIF:
> + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_IPU_MEMORY:
> + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
> + ? ? ? ? ? ? ? break;
> + ? ? ? default:
> + ? ? ? ? ? ? ? break;
> + ? ? ? }
> +
> + ? ? ? sdma->pc_from_device = per_2_emi;
> + ? ? ? sdma->pc_to_device = emi_2_per;
Return res? You're assigning it a value in some cases.
> +}
> +
> +static int sdma_load_context(int channel)
> +{
> + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> + ? ? ? int load_address;
> + ? ? ? struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> + ? ? ? int ret;
> +
> + ? ? ? if (sdma->direction == DMA_FROM_DEVICE) {
> + ? ? ? ? ? ? ? load_address = sdma->pc_from_device;
> + ? ? ? } else {
> + ? ? ? ? ? ? ? load_address = sdma->pc_to_device;
> + ? ? ? }
> +
> + ? ? ? if (load_address < 0)
> + ? ? ? ? ? ? ? return load_address;
> +
> + ? ? ? pr_debug("%s: load_address = %d\n", __func__, load_address);
> + ? ? ? pr_debug("%s: wml = 0x%08x\n", __func__, sdma->watermark_level);
> + ? ? ? pr_debug("%s: shp_addr = 0x%08x\n", __func__, sdma->shp_addr);
> + ? ? ? pr_debug("%s: per_addr = 0x%08x\n", __func__, sdma->per_addr);
> + ? ? ? pr_debug("%s: event_mask1 = 0x%08x\n", __func__, sdma->event_mask1);
> + ? ? ? pr_debug("%s: event_mask2 = 0x%08x\n", __func__, sdma->event_mask2);
Surely it must be possible to get the struct device * pointer for the
channels host and use dev_dbg() instead?
> +
> + ? ? ? memset(sdma_context, 0, sizeof(*sdma_context));
> + ? ? ? sdma_context->channel_state.pc = load_address;
> +
> + ? ? ? /* Send by context the event mask,base address for peripheral
> + ? ? ? ?* and watermark level
> + ? ? ? ?*/
> + ? ? ? sdma_context->gReg[0] = sdma->event_mask2;
> + ? ? ? sdma_context->gReg[1] = sdma->event_mask1;
> + ? ? ? sdma_context->gReg[2] = sdma->per_addr;
> + ? ? ? sdma_context->gReg[6] = sdma->shp_addr;
> + ? ? ? sdma_context->gReg[7] = sdma->watermark_level;
> +
> + ? ? ? bd0->mode.command = C0_SETDM;
> + ? ? ? bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> + ? ? ? bd0->mode.count = sizeof(*sdma_context) / 4;
> + ? ? ? bd0->buffer_addr = sdma_context_phys;
> + ? ? ? bd0->ext_buffer_addr = 2048 + (sizeof(*sdma_context) / 4) * channel;
> +
> + ? ? ? ret = sdma_run_channel(0);
> +
> + ? ? ? return ret;
> +}
> +
> +static void sdma_disable_channel(int channel)
> +{
> + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> +
> + ? ? ? writel(1 << channel, SDMA_H_STATSTOP);
> + ? ? ? sdma->busy = 0;
> +}
> +
> +static int sdma_config_channel(int channel)
> +{
> + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> + ? ? ? int ret;
> +
> + ? ? ? sdma_disable_channel(channel);
> +
> + ? ? ? sdma->event_mask1 = 0;
> + ? ? ? sdma->event_mask2 = 0;
> + ? ? ? sdma->shp_addr = 0;
> + ? ? ? sdma->per_addr = 0;
> +
> + ? ? ? if (sdma->event_id)
> + ? ? ? ? ? ? ? sdma_event_enable(channel, sdma->event_id);
> +
> + ? ? ? switch (sdma->peripheral_type) {
> + ? ? ? case IMX_DMATYPE_DSP:
> + ? ? ? ? ? ? ? sdma_config_ownership(channel, 0, 1, 1);
The parameters here makes yoy believe that the types should
be bool rather than int...
> + ? ? ? ? ? ? ? break;
> + ? ? ? case IMX_DMATYPE_MEMORY:
> + ? ? ? ? ? ? ? sdma_config_ownership(channel, 0, 1, 0);
> + ? ? ? ? ? ? ? break;
> + ? ? ? default:
> + ? ? ? ? ? ? ? sdma_config_ownership(channel, 1, 1, 0);
> + ? ? ? ? ? ? ? break;
> + ? ? ? }
> +
> + ? ? ? sdma_get_pc(sdma, sdma->peripheral_type);
> +
> + ? ? ? if ((sdma->peripheral_type != IMX_DMATYPE_MEMORY) &&
> + ? ? ? ? ? ? ? ? ? ? ? (sdma->peripheral_type != IMX_DMATYPE_DSP)) {
> + ? ? ? ? ? ? ? /* Handle multiple event channels differently */
> + ? ? ? ? ? ? ? if (sdma->event_id2) {
> + ? ? ? ? ? ? ? ? ? ? ? sdma->event_mask2 = 1 << (sdma->event_id2 % 32);
> + ? ? ? ? ? ? ? ? ? ? ? if (sdma->event_id2 > 31)
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? sdma->watermark_level |= 1 << 31;
> + ? ? ? ? ? ? ? ? ? ? ? sdma->event_mask1 = 1 << (sdma->event_id % 32);
> + ? ? ? ? ? ? ? ? ? ? ? if (sdma->event_id > 31)
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? sdma->watermark_level |= 1 << 30;
> + ? ? ? ? ? ? ? } else {
> + ? ? ? ? ? ? ? ? ? ? ? sdma->event_mask1 = 1 << sdma->event_id;
> + ? ? ? ? ? ? ? ? ? ? ? sdma->event_mask2 = 1 << (sdma->event_id - 32);
> + ? ? ? ? ? ? ? }
> + ? ? ? ? ? ? ? /* Watermark Level */
> + ? ? ? ? ? ? ? sdma->watermark_level |= sdma->watermark_level;
> + ? ? ? ? ? ? ? /* Address */
> + ? ? ? ? ? ? ? sdma->shp_addr = sdma->per_address;
> + ? ? ? } else {
> + ? ? ? ? ? ? ? sdma->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
> + ? ? ? }
> +
> + ? ? ? ret = sdma_load_context(channel);
> +
> + ? ? ? return ret;
> +}
> +
> +static int sdma_set_channel_priority(unsigned int channel, unsigned int priority)
> +{
> + ? ? ? if (priority < MXC_SDMA_MIN_PRIORITY
> + ? ? ? ? ? || priority > MXC_SDMA_MAX_PRIORITY) {
> + ? ? ? ? ? ? ? return -EINVAL;
> + ? ? ? }
> +
> + ? ? ? writel(priority, SDMA_CHNPRI_0 + 4 * channel);
> +
> + ? ? ? return 0;
> +}
> +
> +static int sdma_request_channel(int channel)
> +{
> + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> + ? ? ? int ret = -EBUSY;
> +
> + ? ? ? sdma->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdma->bd_phys, GFP_KERNEL);
> + ? ? ? if (!sdma->bd) {
> + ? ? ? ? ? ? ? ret = -ENOMEM;
> + ? ? ? ? ? ? ? goto out;
> + ? ? ? }
> +
> + ? ? ? memset(sdma->bd, 0, PAGE_SIZE);
> +
> + ? ? ? channel_control[channel].base_bd_ptr = sdma->bd_phys;
> + ? ? ? channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> + ? ? ? clk_enable(sdma_clk);
Aha you're enabling it once for every channel and rely on
clk reference counting that's clever!
> +
> + ? ? ? sdma_set_channel_priority(channel, MXC_SDMA_DEFAULT_PRIORITY);
> +
> + ? ? ? init_waitqueue_head(&sdma->waitq);
> +
> + ? ? ? sdma->buf_tail = 0;
> +
> + ? ? ? return 0;
> +out:
> +
> + ? ? ? return ret;
> +}
> +
> +static void sdma_enable_channel(int channel)
> +{
> + ? ? ? writel(1 << channel, SDMA_H_START);
> +}
> +
> +static int __init sdma_init(unsigned long phys_base, int irq, int version,
> + ? ? ? ? ? ? ? void *ram_code,
> + ? ? ? ? ? ? ? int ram_code_size)
> +{
> + ? ? ? int i, ret;
> + ? ? ? int channel;
> + ? ? ? dma_addr_t ccb_phys;
> +
> + ? ? ? sdma_version = version;
> + ? ? ? switch (sdma_version) {
> + ? ? ? case 1:
> + ? ? ? ? ? ? ? sdma_num_events = 32;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case 2:
> + ? ? ? ? ? ? ? sdma_num_events = 48;
> + ? ? ? ? ? ? ? break;
> + ? ? ? default:
> + ? ? ? ? ? ? ? pr_err("SDMA: Unknown version %d. aborting\n", sdma_version);
> + ? ? ? ? ? ? ? return -ENODEV;
> + ? ? ? }
> +
> + ? ? ? clk_enable(sdma_clk);
> +
> + ? ? ? sdma_base = ioremap(phys_base, 4096);
Use SZ_4K instead of 4096.
> + ? ? ? if (!sdma_base) {
> + ? ? ? ? ? ? ? ret = -ENOMEM;
> + ? ? ? ? ? ? ? goto err_ioremap;
> + ? ? ? }
> +
> + ? ? ? /* Initialize SDMA private data */
> + ? ? ? memset(sdma_data, 0, sizeof(struct sdma_channel) * MAX_DMA_CHANNELS);
> +
> + ? ? ? for (channel = 0; channel < MAX_DMA_CHANNELS; channel++)
> + ? ? ? ? ? ? ? sdma_data[channel].channel = channel;
> +
> + ? ? ? ret = request_irq(irq, sdma_int_handler, 0, "sdma", NULL);
> + ? ? ? if (ret)
> + ? ? ? ? ? ? ? goto err_request_irq;
> +
> + ? ? ? /* Be sure SDMA has not started yet */
> + ? ? ? writel(0, SDMA_H_C0PTR);
> +
> + ? ? ? channel_control = dma_alloc_coherent(NULL,
> + ? ? ? ? ? ? ? ? ? ? ? MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
> + ? ? ? ? ? ? ? ? ? ? ? sizeof(struct sdma_context_data),
> + ? ? ? ? ? ? ? ? ? ? ? &ccb_phys, GFP_KERNEL);
> +
> + ? ? ? if (!channel_control) {
> + ? ? ? ? ? ? ? ret = -ENOMEM;
> + ? ? ? ? ? ? ? goto err_dma_alloc;
> + ? ? ? }
> +
> + ? ? ? sdma_context = (void *)channel_control +
> + ? ? ? ? ? ? ? MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> + ? ? ? sdma_context_phys = ccb_phys +
> + ? ? ? ? ? ? ? MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> +
> + ? ? ? /* Zero-out the CCB structures array just allocated */
> + ? ? ? memset(channel_control, 0,
> + ? ? ? ? ? ? ? ? ? ? ? MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
> +
> + ? ? ? /* disable all channels */
> + ? ? ? for (i = 0; i < sdma_num_events; i++)
> + ? ? ? ? ? ? ? writel(0, SDMA_CHNENBL_0 + i * 4);
> +
> + ? ? ? /* All channels have priority 0 */
> + ? ? ? for (i = 0; i < MAX_DMA_CHANNELS; i++)
> + ? ? ? ? ? ? ? writel(0, SDMA_CHNPRI_0 + i * 4);
> +
> + ? ? ? ret = sdma_request_channel(0);
> + ? ? ? if (ret)
> + ? ? ? ? ? ? ? goto err_dma_alloc;
> +
> + ? ? ? sdma_config_ownership(0, 0, 1, 0);
> +
> + ? ? ? /* Set Command Channel (Channel Zero) */
> + ? ? ? writel(0x4050, SDMA_CHN0ADDR);
> +
> + ? ? ? /* Set bits of CONFIG register but with static context switching */
> + ? ? ? /* FIXME: Check whether to set ACR bit depending on clock ratios */
> + ? ? ? writel(0, SDMA_H_CONFIG);
> +
> + ? ? ? writel(ccb_phys, SDMA_H_C0PTR);
> +
> + ? ? ? /* download the RAM image for SDMA */
> + ? ? ? sdma_load_script(ram_code,
> + ? ? ? ? ? ? ? ? ? ? ? ram_code_size,
> + ? ? ? ? ? ? ? ? ? ? ? sdma_script_addrs->ram_code_start_addr);
> +
> + ? ? ? /* Set bits of CONFIG register with given context switching mode */
> + ? ? ? writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
> +
> + ? ? ? /* Initializes channel's priorities */
> + ? ? ? sdma_set_channel_priority(0, 7);
> +
> + ? ? ? clk_disable(sdma_clk);
> +
> + ? ? ? return 0;
> +
> +err_dma_alloc:
> + ? ? ? free_irq(irq, NULL);
> +err_request_irq:
> + ? ? ? iounmap(sdma_base);
> +err_ioremap:
> + ? ? ? clk_disable(sdma_clk);
> + ? ? ? pr_err("%s failed with %d\n", __func__, ret);
> + ? ? ? return ret;
> +}
> +
> +static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
> +{
> + ? ? ? dma_cookie_t cookie = sdma->chan.cookie;
> +
> + ? ? ? if (++cookie < 0)
> + ? ? ? ? ? ? ? cookie = 1;
> +
> + ? ? ? sdma->chan.cookie = cookie;
> + ? ? ? sdma->desc.cookie = cookie;
> +
> + ? ? ? return cookie;
> +}
> +
> +static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
> +{
> + ? ? ? return container_of(chan, struct sdma_channel, chan);
> +}
> +
> +static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
> +{
> + ? ? ? struct sdma_channel *sdma = to_sdma_chan(tx->chan);
> + ? ? ? dma_cookie_t cookie;
> +
> + ? ? ? spin_lock_irq(&sdma->lock);
> +
> + ? ? ? cookie = sdma_assign_cookie(sdma);
> +
> + ? ? ? sdma_enable_channel(tx->chan->chan_id);
> +
> + ? ? ? spin_unlock_irq(&sdma->lock);
> +
> + ? ? ? return cookie;
> +}
> +
> +static int sdma_alloc_chan_resources(struct dma_chan *chan)
> +{
> + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> + ? ? ? struct imx_dma_data *data = chan->private;
> + ? ? ? int prio, ret;
> +
> + ? ? ? /* No need to execute this for internal channel 0 */
> + ? ? ? if (!chan->chan_id)
> + ? ? ? ? ? ? ? return 0;
> +
> + ? ? ? if (!data)
> + ? ? ? ? ? ? ? return -EINVAL;
> +
> + ? ? ? switch (data->priority) {
> + ? ? ? case DMA_PRIO_HIGH:
> + ? ? ? ? ? ? ? prio = 3;
Wait, aren't these enumerated?
Add some enum sdma_channel_prio {}..
> + ? ? ? ? ? ? ? break;
> + ? ? ? case DMA_PRIO_MEDIUM:
> + ? ? ? ? ? ? ? prio = 2;
> + ? ? ? ? ? ? ? break;
> + ? ? ? case DMA_PRIO_LOW:
> + ? ? ? default:
> + ? ? ? ? ? ? ? prio = 1;
> + ? ? ? ? ? ? ? break;
> + ? ? ? }
> +
> + ? ? ? sdma->peripheral_type = data->peripheral_type;
> + ? ? ? sdma->event_id = data->dma_request;
> + ? ? ? ret = sdma_set_channel_priority(chan->chan_id, prio);
> + ? ? ? if (ret)
> + ? ? ? ? ? ? ? return ret;
> +
> + ? ? ? if (chan->chan_id) {
> + ? ? ? ? ? ? ? ret = sdma_request_channel(chan->chan_id);
> + ? ? ? ? ? ? ? if (ret)
> + ? ? ? ? ? ? ? ? ? ? ? return ret;
> + ? ? ? }
> +
> + ? ? ? dma_async_tx_descriptor_init(&sdma->desc, chan);
> + ? ? ? sdma->desc.tx_submit = sdma_tx_submit;
> + ? ? ? /* txd.flags will be overwritten in prep funcs */
> + ? ? ? sdma->desc.flags = DMA_CTRL_ACK;
> +
> + ? ? ? return 0;
> +}
> +
> +static void sdma_free_chan_resources(struct dma_chan *chan)
> +{
> + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> + ? ? ? int channel = chan->chan_id;
> +
> + ? ? ? sdma_disable_channel(channel);
> +
> + ? ? ? if (sdma->event_id)
> + ? ? ? ? ? ? ? sdma_event_disable(channel, sdma->event_id);
> + ? ? ? if (sdma->event_id2)
> + ? ? ? ? ? ? ? sdma_event_disable(channel, sdma->event_id2);
> +
> + ? ? ? sdma->event_id = 0;
> + ? ? ? sdma->event_id2 = 0;
> +
> + ? ? ? sdma_set_channel_priority(channel, 0);
> +
> + ? ? ? dma_free_coherent(NULL, PAGE_SIZE, sdma->bd, sdma->bd_phys);
> +
> + ? ? ? clk_disable(sdma_clk);
> +}
> +
> +#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
> +
> +static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
> + ? ? ? ? ? ? ? struct dma_chan *chan, struct scatterlist *sgl,
> + ? ? ? ? ? ? ? unsigned int sg_len, enum dma_data_direction direction,
> + ? ? ? ? ? ? ? unsigned long flags)
> +{
> + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> + ? ? ? int ret, i, count;
> + ? ? ? int channel = chan->chan_id;
> + ? ? ? struct scatterlist *sg;
> +
> + ? ? ? if (sdma->busy)
> + ? ? ? ? ? ? ? return NULL;
> + ? ? ? sdma->busy = 1;
> +
> + ? ? ? sdma->flags = 0;
What are those flags anyway? I think you will need some
#define:s for them.
> +
> + ? ? ? pr_debug("SDMA: setting up %d entries for channel %d.\n",
> + ? ? ? ? ? ? ? ? ? ? ? sg_len, channel);
> +
> + ? ? ? sdma->direction = direction;
> + ? ? ? ret = sdma_load_context(channel);
> + ? ? ? if (ret)
> + ? ? ? ? ? ? ? goto err_out;
> +
> + ? ? ? if (sg_len > NUM_BD) {
> + ? ? ? ? ? ? ? pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? channel, sg_len, NUM_BD);
> + ? ? ? ? ? ? ? ret = -EINVAL;
> + ? ? ? ? ? ? ? goto err_out;
> + ? ? ? }
> +
> + ? ? ? for_each_sg(sgl, sg, sg_len, i) {
> + ? ? ? ? ? ? ? struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> + ? ? ? ? ? ? ? int param;
> +
> + ? ? ? ? ? ? ? bd->buffer_addr = sgl->dma_address;
> +
> + ? ? ? ? ? ? ? count = sg->length;
> +
> + ? ? ? ? ? ? ? if (count > 0xffff) {
> + ? ? ? ? ? ? ? ? ? ? ? pr_err("SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? channel, count, 0xffff);
> + ? ? ? ? ? ? ? ? ? ? ? ret = -EINVAL;
> + ? ? ? ? ? ? ? ? ? ? ? goto err_out;
> + ? ? ? ? ? ? ? }
> +
> + ? ? ? ? ? ? ? bd->mode.count = count;
> +
> + ? ? ? ? ? ? ? if (sdma->word_size > 4) {
> + ? ? ? ? ? ? ? ? ? ? ? ret = ?-EINVAL;
> + ? ? ? ? ? ? ? ? ? ? ? goto err_out;
> + ? ? ? ? ? ? ? }
> + ? ? ? ? ? ? ? if (sdma->word_size == 4)
> + ? ? ? ? ? ? ? ? ? ? ? bd->mode.command = 0;
> + ? ? ? ? ? ? ? else
> + ? ? ? ? ? ? ? ? ? ? ? bd->mode.command = sdma->word_size;
> +
> + ? ? ? ? ? ? ? param = BD_DONE | BD_EXTD | BD_CONT;
> +
> + ? ? ? ? ? ? ? if (sdma->flags & IMX_DMA_SG_LOOP) {
> + ? ? ? ? ? ? ? ? ? ? ? param |= BD_INTR;
> + ? ? ? ? ? ? ? ? ? ? ? if (i + 1 == sg_len)
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param |= BD_WRAP;
> + ? ? ? ? ? ? ? }
> +
> + ? ? ? ? ? ? ? if (i + 1 == sg_len)
> + ? ? ? ? ? ? ? ? ? ? ? param |= BD_INTR;
> +
> + ? ? ? ? ? ? ? pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? i, count, sg->dma_address,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param & BD_WRAP ? "wrap" : "",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param & BD_INTR ? " intr" : "");
> +
> + ? ? ? ? ? ? ? bd->mode.status = param;
> + ? ? ? }
> +
> + ? ? ? sdma->num_bd = sg_len;
> + ? ? ? channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> + ? ? ? return &sdma->desc;
> +err_out:
> + ? ? ? return NULL;
> +}
> +
> +static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
> + ? ? ? ? ? ? ? struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
> + ? ? ? ? ? ? ? size_t period_len, enum dma_data_direction direction)
> +{
> + ? ? ? int num_periods = buf_len / period_len;
> + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> + ? ? ? int channel = chan->chan_id;
> + ? ? ? int ret, i = 0, buf = 0;
> +
> + ? ? ? pr_debug("%s channel: %d\n", __func__, channel);
Must be possible to find struct device * and use dev_dbg()
> +
> + ? ? ? if (sdma->busy)
> + ? ? ? ? ? ? ? return NULL;
> +
> + ? ? ? sdma->busy = 1;
> +
> + ? ? ? sdma->flags |= IMX_DMA_SG_LOOP;
> + ? ? ? sdma->direction = direction;
> + ? ? ? ret = sdma_load_context(channel);
> + ? ? ? if (ret)
> + ? ? ? ? ? ? ? goto err_out;
> +
> + ? ? ? if (num_periods > NUM_BD) {
> + ? ? ? ? ? ? ? pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? channel, num_periods, NUM_BD);
> + ? ? ? ? ? ? ? goto err_out;
> + ? ? ? }
> +
> + ? ? ? if (period_len > 0xffff) {
> + ? ? ? ? ? ? ? pr_err("SDMA channel %d: maximum period size exceeded: %d > %d\n",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? channel, period_len, 0xffff);
> + ? ? ? ? ? ? ? goto err_out;
> + ? ? ? }
> +
> + ? ? ? while (buf < buf_len) {
> + ? ? ? ? ? ? ? struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> + ? ? ? ? ? ? ? int param;
> +
> + ? ? ? ? ? ? ? bd->buffer_addr = dma_addr;
> +
> + ? ? ? ? ? ? ? bd->mode.count = period_len;
> +
> + ? ? ? ? ? ? ? if (sdma->word_size > 4)
> + ? ? ? ? ? ? ? ? ? ? ? goto err_out;
> + ? ? ? ? ? ? ? if (sdma->word_size == 4)
> + ? ? ? ? ? ? ? ? ? ? ? bd->mode.command = 0;
> + ? ? ? ? ? ? ? else
> + ? ? ? ? ? ? ? ? ? ? ? bd->mode.command = sdma->word_size;
> +
> + ? ? ? ? ? ? ? param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
> + ? ? ? ? ? ? ? if (i + 1 == num_periods)
> + ? ? ? ? ? ? ? ? ? ? ? param |= BD_WRAP;
> +
> + ? ? ? ? ? ? ? pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? i, period_len, dma_addr,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param & BD_WRAP ? "wrap" : "",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param & BD_INTR ? " intr" : "");
> +
> + ? ? ? ? ? ? ? bd->mode.status = param;
> +
> + ? ? ? ? ? ? ? dma_addr += period_len;
> + ? ? ? ? ? ? ? buf += period_len;
> +
> + ? ? ? ? ? ? ? i++;
> + ? ? ? }
> +
> + ? ? ? sdma->num_bd = num_periods;
> + ? ? ? channel_control[channel].current_bd_ptr = sdma->bd_phys;
> +
> + ? ? ? return &sdma->desc;
> +err_out:
> + ? ? ? sdma->busy = 0;
> + ? ? ? return NULL;
> +}
> +
> +static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> + ? ? ? ? ? ? ? unsigned long arg)
> +{
> + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> + ? ? ? struct dma_slave_config *dmaengine_cfg = (void *)arg;
> +
> + ? ? ? switch (cmd) {
> + ? ? ? case DMA_TERMINATE_ALL:
> + ? ? ? ? ? ? ? sdma_disable_channel(chan->chan_id);
> + ? ? ? ? ? ? ? return 0;
> + ? ? ? case DMA_SLAVE_CONFIG:
> + ? ? ? ? ? ? ? if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
> + ? ? ? ? ? ? ? ? ? ? ? sdma->per_address = dmaengine_cfg->src_addr;
> + ? ? ? ? ? ? ? ? ? ? ? sdma->watermark_level = dmaengine_cfg->src_maxburst;
> + ? ? ? ? ? ? ? ? ? ? ? sdma->word_size = dmaengine_cfg->src_addr_width;
> + ? ? ? ? ? ? ? } else {
> + ? ? ? ? ? ? ? ? ? ? ? sdma->per_address = dmaengine_cfg->dst_addr;
> + ? ? ? ? ? ? ? ? ? ? ? sdma->watermark_level = dmaengine_cfg->dst_maxburst;
> + ? ? ? ? ? ? ? ? ? ? ? sdma->word_size = dmaengine_cfg->dst_addr_width;
> + ? ? ? ? ? ? ? }
> + ? ? ? ? ? ? ? return sdma_config_channel(chan->chan_id);
> + ? ? ? default:
> + ? ? ? ? ? ? ? return -ENOSYS;
> + ? ? ? }
> +
> + ? ? ? return -EINVAL;
> +}
> +
> +static enum dma_status sdma_tx_status(struct dma_chan *chan,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? dma_cookie_t cookie,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? struct dma_tx_state *txstate)
> +{
> + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> + ? ? ? dma_cookie_t last_used;
> + ? ? ? enum dma_status ret;
> +
> + ? ? ? last_used = chan->cookie;
> +
> + ? ? ? ret = dma_async_is_complete(cookie, sdma->last_completed, last_used);
> + ? ? ? dma_set_tx_state(txstate, sdma->last_completed, last_used, 0);
> +
> + ? ? ? return ret;
> +}
> +
> +static void sdma_issue_pending(struct dma_chan *chan)
> +{
> + ? ? ? /*
> + ? ? ? ?* Nothing to do. We only have a single descriptor
> + ? ? ? ?*/
> +}
> +
> +static int __devinit sdma_probe(struct platform_device *pdev)
> +{
> + ? ? ? int ret;
> + ? ? ? const struct firmware *fw;
> + ? ? ? const struct sdma_firmware_header *header;
> + ? ? ? const struct sdma_script_start_addrs *addr;
> + ? ? ? int irq;
> + ? ? ? unsigned short *ram_code;
> + ? ? ? struct resource *iores;
> + ? ? ? struct sdma_platform_data *pdata = pdev->dev.platform_data;
> + ? ? ? int version;
> + ? ? ? char *cpustr, *fwname;
> + ? ? ? int i;
> + ? ? ? dma_cap_mask_t mask;
> +
> + ? ? ? /* there can be only one */
> + ? ? ? BUG_ON(sdma_base);
> +
> + ? ? ? iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + ? ? ? irq = platform_get_irq(pdev, 0);
> + ? ? ? if (!iores || irq < 0 || !pdata)
> + ? ? ? ? ? ? ? return -EINVAL;
> +
> + ? ? ? sdma_clk = clk_get(&pdev->dev, NULL);
> + ? ? ? if (IS_ERR(sdma_clk)) {
> + ? ? ? ? ? ? ? ret = PTR_ERR(sdma_clk);
> + ? ? ? ? ? ? ? goto err_clk;
> + ? ? ? }
> +
> + ? ? ? if (cpu_is_mx31()) {
> + ? ? ? ? ? ? ? cpustr = "imx31";
> + ? ? ? ? ? ? ? version = mx31_revision() >> 4;
> + ? ? ? } else if (cpu_is_mx35()) {
> + ? ? ? ? ? ? ? cpustr = "imx35";
> +/* FIXME: ? ? ?version = mx35_revision(); */
> + ? ? ? ? ? ? ? version = 2;
> + ? ? ? } else {
> + ? ? ? ? ? ? ? ret = -EINVAL;
> + ? ? ? ? ? ? ? goto err_cputype;
> + ? ? ? }
> +
> + ? ? ? fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin", cpustr, version);
> + ? ? ? if (!fwname) {
> + ? ? ? ? ? ? ? ret = -ENOMEM;
> + ? ? ? ? ? ? ? goto err_cputype;
> + ? ? ? }
> +
> + ? ? ? ret = request_firmware(&fw, fwname, &pdev->dev);
> + ? ? ? if (ret) {
> + ? ? ? ? ? ? ? dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? fwname, ret);
> + ? ? ? ? ? ? ? kfree(fwname);
> + ? ? ? ? ? ? ? goto err_cputype;
> + ? ? ? }
> + ? ? ? kfree(fwname);
> +
> + ? ? ? if (fw->size < sizeof(*header))
> + ? ? ? ? ? ? ? goto err_firmware;
> +
> + ? ? ? header = (struct sdma_firmware_header *)fw->data;
> +
> + ? ? ? if (header->magic != SDMA_FIRMWARE_MAGIC)
> + ? ? ? ? ? ? ? goto err_firmware;
> + ? ? ? if (header->ram_code_start + header->ram_code_size > fw->size)
> + ? ? ? ? ? ? ? goto err_firmware;
> +
> + ? ? ? addr = (void *)header + header->script_addrs_start;
> + ? ? ? ram_code = (void *)header + header->ram_code_start;
> + ? ? ? memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
> +
> + ? ? ? ret = sdma_init(iores->start, irq, pdata->sdma_version,
> + ? ? ? ? ? ? ? ? ? ? ? ram_code, header->ram_code_size);
> + ? ? ? if (ret)
> + ? ? ? ? ? ? ? goto err_firmware;
> +
> + ? ? ? INIT_LIST_HEAD(&sdma_dma_device->channels);
> +
> + ? ? ? /* Initialize channel parameters */
> + ? ? ? for (i = 0; i < MAX_DMA_CHANNELS; i++) {
> + ? ? ? ? ? ? ? struct sdma_channel *sdma = &sdma_data[i];
> +
> + ? ? ? ? ? ? ? spin_lock_init(&sdma->lock);
> +
> + ? ? ? ? ? ? ? dma_cap_set(DMA_SLAVE, sdma_dma_device->cap_mask);
> + ? ? ? ? ? ? ? dma_cap_set(DMA_CYCLIC, sdma_dma_device->cap_mask);
> +
> + ? ? ? ? ? ? ? sdma->chan.device = sdma_dma_device;
> + ? ? ? ? ? ? ? sdma->chan.chan_id = i;
> +
> + ? ? ? ? ? ? ? /* Add the channel to the DMAC list */
> + ? ? ? ? ? ? ? list_add_tail(&sdma->chan.device_node, &sdma_dma_device->channels);
> + ? ? ? }
> +
> + ? ? ? sdma_dma_device->dev = &pdev->dev;
> +
> + ? ? ? sdma_dma_device->device_alloc_chan_resources = sdma_alloc_chan_resources;
> + ? ? ? sdma_dma_device->device_free_chan_resources = sdma_free_chan_resources;
> + ? ? ? sdma_dma_device->device_tx_status = sdma_tx_status;
> + ? ? ? sdma_dma_device->device_prep_slave_sg = sdma_prep_slave_sg;
> + ? ? ? sdma_dma_device->device_prep_dma_cyclic = sdma_prep_dma_cyclic;
> + ? ? ? sdma_dma_device->device_control = sdma_control;
> + ? ? ? sdma_dma_device->device_issue_pending = sdma_issue_pending;
> +
> + ? ? ? ret = dma_async_device_register(sdma_dma_device);
> + ? ? ? if (ret) {
> + ? ? ? ? ? ? ? dev_err(&pdev->dev, "unable to register DMAC\n");
SDMAC even?
> + ? ? ? ? ? ? ? goto err_firmware;
> + ? ? ? }
> +
> + ? ? ? dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
> + ? ? ? ? ? ? ? ? ? ? ? header->version_major,
> + ? ? ? ? ? ? ? ? ? ? ? header->version_minor);
> +
> + ? ? ? /* request channel 0. This is an internal control channel
> + ? ? ? ?* to the SDMA engine and not available to clients.
> + ? ? ? ?*/
> + ? ? ? dma_cap_zero(mask);
> + ? ? ? dma_cap_set(DMA_SLAVE, mask);
> + ? ? ? dma_request_channel(mask, NULL, NULL);
> +
> + ? ? ? release_firmware(fw);
> +
> + ? ? ? return 0;
> +
> +err_firmware:
> + ? ? ? release_firmware(fw);
> +err_cputype:
> + ? ? ? clk_put(sdma_clk);
> +err_clk:
> + ? ? ? return 0;
> +}
> +
> +static int __devexit sdma_remove(struct platform_device *pdev)
> +{
> + ? ? ? return -EBUSY;
> +}
> +
> +static struct platform_driver sdma_driver = {
> + ? ? ? .driver ? ? ? ? = {
> + ? ? ? ? ? ? ? .name ? = "imx-sdma",
> + ? ? ? },
> + ? ? ? .probe ? ? ? ? ?= sdma_probe,
> + ? ? ? .remove ? ? ? ? = __devexit_p(sdma_remove),
> +};
> +
> +static int __init sdma_module_init(void)
> +{
> + ? ? ? return platform_driver_register(&sdma_driver);
> +}
> +subsys_initcall(sdma_module_init);
> +
> +MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
> +MODULE_DESCRIPTION("i.MX SDMA driver");
> +MODULE_LICENSE("GPL");
> --
> 1.7.1
Thanks for using this API
Sascha!
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 1/3] dmaengine: add possibility for cyclic transfers
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-16 12:22 ` Linus Walleij
-1 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-16 12:22 UTC (permalink / raw)
To: Sascha Hauer
Cc: linux-kernel, Dan Williams, linux-arm-kernel, Haavard Skinnemoen
2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> Cyclic transfers are useful for audio where a single buffer divided
> in periods has to be transfered endlessly until stopped. After being
> prepared the transfer is started using the dma_async_descriptor->tx_submit
> function. dma_async_descriptor->callback is called after each period.
> The transfer is stopped using the DMA_TERMINATE_ALL callback.
> While being used for cyclic transfers the channel cannot be used
> for other transfer types.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Looks good to me.
Acked-by: Linus Walleij <linus.walleij@stericsson.com>
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
@ 2010-08-16 12:22 ` Linus Walleij
0 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-16 12:22 UTC (permalink / raw)
To: linux-arm-kernel
2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> Cyclic transfers are useful for audio where a single buffer divided
> in periods has to be transfered endlessly until stopped. After being
> prepared the transfer is started using the dma_async_descriptor->tx_submit
> function. dma_async_descriptor->callback is called after each period.
> The transfer is stopped using the DMA_TERMINATE_ALL callback.
> While being used for cyclic transfers the channel cannot be used
> for other transfer types.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Looks good to me.
Acked-by: Linus Walleij <linus.walleij@stericsson.com>
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 1/3] dmaengine: add possibility for cyclic transfers
2010-08-16 11:56 ` Lothar Waßmann
@ 2010-08-16 12:27 ` Linus Walleij
-1 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-16 12:27 UTC (permalink / raw)
To: Lothar Waßmann
Cc: Sascha Hauer, linux-kernel, Dan Williams, Haavard Skinnemoen,
linux-arm-kernel
2010/8/16 Lothar Waßmann <LW@karo-electronics.de>:
> Why not implement this feature using cyclic SG lists (created with
> sg_chain())? This would give you endless DMA transfers without any
> special DMA API extensions.
That would be elegant...
The driver will have to detect that the sglist is chained like an
ouroboros to program the DMAC apropriately, is is easy to
detect if an sglist is chained onto itself?
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
@ 2010-08-16 12:27 ` Linus Walleij
0 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-16 12:27 UTC (permalink / raw)
To: linux-arm-kernel
2010/8/16 Lothar Wa?mann <LW@karo-electronics.de>:
> Why not implement this feature using cyclic SG lists (created with
> sg_chain())? This would give you endless DMA transfers without any
> special DMA API extensions.
That would be elegant...
The driver will have to detect that the sglist is chained like an
ouroboros to program the DMAC apropriately, is is easy to
detect if an sglist is chained onto itself?
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 1/3] dmaengine: add possibility for cyclic transfers
2010-08-16 11:56 ` Lothar Waßmann
@ 2010-08-16 12:32 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 12:32 UTC (permalink / raw)
To: Lothar Waßmann
Cc: linux-kernel, Linus Walleij, Dan Williams, Haavard Skinnemoen,
linux-arm-kernel
On Mon, Aug 16, 2010 at 01:56:34PM +0200, Lothar Waßmann wrote:
> Hi,
>
> Sascha Hauer writes:
> > Cyclic transfers are useful for audio where a single buffer divided
> > in periods has to be transfered endlessly until stopped. After being
> > prepared the transfer is started using the dma_async_descriptor->tx_submit
> > function. dma_async_descriptor->callback is called after each period.
> > The transfer is stopped using the DMA_TERMINATE_ALL callback.
> > While being used for cyclic transfers the channel cannot be used
> > for other transfer types.
> >
> > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> > Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
> > ---
> > drivers/dma/dmaengine.c | 2 ++
> > include/linux/dmaengine.h | 6 +++++-
> > 2 files changed, 7 insertions(+), 1 deletions(-)
> >
> > diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> > index 9d31d5e..e5e79ce 100644
> > --- a/drivers/dma/dmaengine.c
> > +++ b/drivers/dma/dmaengine.c
> > @@ -692,6 +692,8 @@ int dma_async_device_register(struct dma_device *device)
> > !device->device_prep_dma_interrupt);
> > BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> > !device->device_prep_slave_sg);
> > + BUG_ON(dma_has_cap(DMA_CYCLIC, device->cap_mask) &&
> > + !device->device_prep_dma_cyclic);
> > BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> > !device->device_control);
> >
> > diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> > index c61d4ca..0df7864 100644
> > --- a/include/linux/dmaengine.h
> > +++ b/include/linux/dmaengine.h
> > @@ -67,10 +67,11 @@ enum dma_transaction_type {
> > DMA_PRIVATE,
> > DMA_ASYNC_TX,
> > DMA_SLAVE,
> > + DMA_CYCLIC,
> > };
> >
> > /* last transaction type for creation of the capabilities mask */
> > -#define DMA_TX_TYPE_END (DMA_SLAVE + 1)
> > +#define DMA_TX_TYPE_END (DMA_CYCLIC + 1)
> >
> >
> > /**
> > @@ -478,6 +479,9 @@ struct dma_device {
> > struct dma_chan *chan, struct scatterlist *sgl,
> > unsigned int sg_len, enum dma_data_direction direction,
> > unsigned long flags);
> > + struct dma_async_tx_descriptor *(*device_prep_dma_cyclic)(
> > + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> > + size_t period_len, enum dma_data_direction direction);
> > int (*device_control)(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> > unsigned long arg);
> >
> > --
> > 1.7.1
> >
> Why not implement this feature using cyclic SG lists (created with
> sg_chain())? This would give you endless DMA transfers without any
> special DMA API extensions.
Been there, done that. In the end it just seemed like adding additional
overhead to create the sg list and using the sg for something for which
it is not really designed. Still you need extensions to the DMA API to
signal that you want to have a callback for every sg entry. Normally you
only need a callback on the end of the list.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
@ 2010-08-16 12:32 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 12:32 UTC (permalink / raw)
To: linux-arm-kernel
On Mon, Aug 16, 2010 at 01:56:34PM +0200, Lothar Wa?mann wrote:
> Hi,
>
> Sascha Hauer writes:
> > Cyclic transfers are useful for audio where a single buffer divided
> > in periods has to be transfered endlessly until stopped. After being
> > prepared the transfer is started using the dma_async_descriptor->tx_submit
> > function. dma_async_descriptor->callback is called after each period.
> > The transfer is stopped using the DMA_TERMINATE_ALL callback.
> > While being used for cyclic transfers the channel cannot be used
> > for other transfer types.
> >
> > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> > Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
> > ---
> > drivers/dma/dmaengine.c | 2 ++
> > include/linux/dmaengine.h | 6 +++++-
> > 2 files changed, 7 insertions(+), 1 deletions(-)
> >
> > diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> > index 9d31d5e..e5e79ce 100644
> > --- a/drivers/dma/dmaengine.c
> > +++ b/drivers/dma/dmaengine.c
> > @@ -692,6 +692,8 @@ int dma_async_device_register(struct dma_device *device)
> > !device->device_prep_dma_interrupt);
> > BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> > !device->device_prep_slave_sg);
> > + BUG_ON(dma_has_cap(DMA_CYCLIC, device->cap_mask) &&
> > + !device->device_prep_dma_cyclic);
> > BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> > !device->device_control);
> >
> > diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> > index c61d4ca..0df7864 100644
> > --- a/include/linux/dmaengine.h
> > +++ b/include/linux/dmaengine.h
> > @@ -67,10 +67,11 @@ enum dma_transaction_type {
> > DMA_PRIVATE,
> > DMA_ASYNC_TX,
> > DMA_SLAVE,
> > + DMA_CYCLIC,
> > };
> >
> > /* last transaction type for creation of the capabilities mask */
> > -#define DMA_TX_TYPE_END (DMA_SLAVE + 1)
> > +#define DMA_TX_TYPE_END (DMA_CYCLIC + 1)
> >
> >
> > /**
> > @@ -478,6 +479,9 @@ struct dma_device {
> > struct dma_chan *chan, struct scatterlist *sgl,
> > unsigned int sg_len, enum dma_data_direction direction,
> > unsigned long flags);
> > + struct dma_async_tx_descriptor *(*device_prep_dma_cyclic)(
> > + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> > + size_t period_len, enum dma_data_direction direction);
> > int (*device_control)(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> > unsigned long arg);
> >
> > --
> > 1.7.1
> >
> Why not implement this feature using cyclic SG lists (created with
> sg_chain())? This would give you endless DMA transfers without any
> special DMA API extensions.
Been there, done that. In the end it just seemed like adding additional
overhead to create the sg list and using the sg for something for which
it is not really designed. Still you need extensions to the DMA API to
signal that you want to have a callback for every sg entry. Normally you
only need a callback on the end of the list.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-16 12:21 ` Linus Walleij
@ 2010-08-16 14:15 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 14:15 UTC (permalink / raw)
To: Linus Walleij; +Cc: linux-kernel, Dan Williams, linux-arm-kernel
Hi Linus,
Thank you for the review. Sorry for so many trival mistakes in the code,
but I'm staring at this code in many different variants for some time
now and I'm getting blind on it.
On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> 2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
>
> > This patch adds support for the Freescale i.MX SDMA engine.
>
> I like it!
>
> > The SDMA engine is a scatter/gather DMA engine which is implemented
> > as a seperate coprocessor. SDMA needs its own firmware which is
> > requested using the standard request_firmware mechanism. The firmware
> > has different entry points for each peripheral type, so drivers
> > have to pass the peripheral type to the DMA engine which in turn
> > picks the correct firmware entry point from a table contained in
> > the firmware image itself.
>
> Quite fun, if the spec for the microcode is open this opens up
> for dynamic firmware generation for specific DMA jobs does it
> not?
Unfortunately the specs are not open, so we are sticked to the binary
microcode from Freescale. I'm pretty sure though that the SDMA engine
could do at least a device_prep_dma_xor operation.
>
> > I took a very simple approach to implement dmaengine support. Only
> > a single descriptor is statically assigned to a each channel. This
> > means that transfers can't be queued up but only a single transfer
> > is in progress. This simplifies implementation a lot and is sufficient
> > for the usual device/memory transfers.
>
> If you want to add memcpy() capability later you're gonna need
> this I think, but you can take that when that need arise.
Yes, I left this as an exercise for those who want to have this feature ;)
I think it's better to have tested code in the Kernel than having a
complicated list handling which is completely untested for anything
other than a single entry in the list.
>
> >(...)
> > +++ b/arch/arm/plat-mxc/include/mach/dma.h
> > @@ -0,0 +1,64 @@
> > +/*
> > + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> > + *
> > + * This program is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License version 2 as
> > + * published by the Free Software Foundation.
> > + */
> > +
> > +#ifndef __ASM_ARCH_MXC_DMA_H__
> > +#define __ASM_ARCH_MXC_DMA_H__
> > +
> > +#include <linux/scatterlist.h>
> > +
> > +/*
> > + * This enumerates peripheral types. Used for SDMA.
> > + */
> > +typedef enum {
>
> The kernel is not really happy about typedefs, can't this be a
> regular enum?
>
> > + IMX_DMATYPE_SSI, /* MCU domain SSI */
> > + IMX_DMATYPE_SSI_SP, /* Shared SSI */
> > + IMX_DMATYPE_MMC, /* MMC */
> > + IMX_DMATYPE_SDHC, /* SDHC */
> > + IMX_DMATYPE_UART, /* MCU domain UART */
> > + IMX_DMATYPE_UART_SP, /* Shared UART */
> > + IMX_DMATYPE_FIRI, /* FIRI */
> > + IMX_DMATYPE_CSPI, /* MCU domain CSPI */
> > + IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
> > + IMX_DMATYPE_SIM, /* SIM */
> > + IMX_DMATYPE_ATA, /* ATA */
> > + IMX_DMATYPE_CCM, /* CCM */
> > + IMX_DMATYPE_EXT, /* External peripheral */
> > + IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
> > + IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
> > + IMX_DMATYPE_DSP, /* DSP */
> > + IMX_DMATYPE_MEMORY, /* Memory */
> > + IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
> > + IMX_DMATYPE_SPDIF, /* SPDIF */
> > + IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
> > + IMX_DMATYPE_ASRC, /* ASRC */
> > + IMX_DMATYPE_ESAI, /* ESAI */
> > +} sdma_peripheral_type;
> > +
> > +enum imx_dma_prio {
> > + DMA_PRIO_HIGH = 0,
> > + DMA_PRIO_MEDIUM = 1,
> > + DMA_PRIO_LOW = 2
> > +};
> > +
> > +struct imx_dma_data {
> > + int dma_request; /* DMA request line */
>
> Can this be negative and what is the range? I would
> suspect something like u8 or u16 would surely be more
> apropriate...
>
> > + sdma_peripheral_type peripheral_type;
> > + int priority;
>
> Isn't this an enum imx_dma_prio?
>
> > +};
> > +
> > +static inline int imx_dma_is_ipu(struct dma_chan *chan)
> > +{
> > + return !strcmp(dev_name(chan->device->dev), "ipu-core");
> > +}
> > +
> > +static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
> > +{
> > + return !strcmp(dev_name(chan->device->dev), "imx-sdma");
> > +}
> > +
> > +#endif
> > diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
> > new file mode 100644
> > index 0000000..5d542b8
> > --- /dev/null
> > +++ b/arch/arm/plat-mxc/include/mach/sdma.h
> > @@ -0,0 +1,8 @@
> > +#ifndef __MACH_MXC_SDMA_H__
> > +#define __MACH_MXC_SDMA_H__
> > +
> > +struct sdma_platform_data {
> > + int sdma_version;
>
> Do you have negative versions or can it be unsigned?
nope, will change this to an unsigned type.
>
> > +};
> > +
> > +#endif /* __MACH_MXC_SDMA_H__ */
> > diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
> > index 9520cf0..f76bda9 100644
> > --- a/drivers/dma/Kconfig
> > +++ b/drivers/dma/Kconfig
> > @@ -195,6 +195,14 @@ config PCH_DMA
> > help
> > Enable support for the Topcliff PCH DMA engine.
> >
> > +config IMX_SDMA
> > + tristate "Atmel AHB DMA support"
> > + depends on ARCH_MXC
> > + select DMA_ENGINE
> > + help
> > + Support the i.MX SDMA engine. This engine is integrated into
> > + Freescale i.MX25/31/35/51 chips.
> > +
> > config DMA_ENGINE
> > bool
> >
> > diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
> > index 72bd703..14d7a1b 100644
> > --- a/drivers/dma/Makefile
> > +++ b/drivers/dma/Makefile
> > @@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
> > obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
> > obj-$(CONFIG_PL330_DMA) += pl330.o
> > obj-$(CONFIG_PCH_DMA) += pch_dma.o
> > +obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
> > diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> > new file mode 100644
> > index 0000000..3ba7905
> > --- /dev/null
> > +++ b/drivers/dma/imx-sdma.c
> > @@ -0,0 +1,1383 @@
> > +/*
> > + * drivers/dma/imx-sdma.c
> > + *
> > + * This file contains a driver for the Freescale Smart DMA engine
> > + *
> > + * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
> > + *
> > + * Based on code from Freescale:
> > + *
> > + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> > + *
> > + * The code contained herein is licensed under the GNU General Public
> > + * License. You may obtain a copy of the GNU General Public License
> > + * Version 2 or later at the following locations:
> > + *
> > + * http://www.opensource.org/licenses/gpl-license.html
> > + * http://www.gnu.org/copyleft/gpl.html
> > + */
> > +#include <linux/init.h>
> > +#include <linux/types.h>
> > +#include <linux/mm.h>
> > +#include <linux/interrupt.h>
> > +#include <linux/clk.h>
> > +#include <linux/semaphore.h>
> > +#include <linux/spinlock.h>
> > +#include <linux/device.h>
> > +#include <linux/dma-mapping.h>
> > +#include <linux/firmware.h>
> > +#include <linux/slab.h>
> > +#include <linux/platform_device.h>
> > +#include <linux/dmaengine.h>
> > +
> > +#include <asm/irq.h>
> > +#include <mach/sdma.h>
> > +#include <mach/dma.h>
> > +#include <mach/hardware.h>
> > +
> > +/* SDMA registers */
> > +#define SDMA_H_C0PTR (sdma_base + 0x000)
> > +#define SDMA_H_INTR (sdma_base + 0x004)
> > +#define SDMA_H_STATSTOP (sdma_base + 0x008)
> > +#define SDMA_H_START (sdma_base + 0x00c)
> > +#define SDMA_H_EVTOVR (sdma_base + 0x010)
> > +#define SDMA_H_DSPOVR (sdma_base + 0x014)
> > +#define SDMA_H_HOSTOVR (sdma_base + 0x018)
> > +#define SDMA_H_EVTPEND (sdma_base + 0x01c)
> > +#define SDMA_H_DSPENBL (sdma_base + 0x020)
> > +#define SDMA_H_RESET (sdma_base + 0x024)
> > +#define SDMA_H_EVTERR (sdma_base + 0x028)
> > +#define SDMA_H_INTRMSK (sdma_base + 0x02c)
> > +#define SDMA_H_PSW (sdma_base + 0x030)
> > +#define SDMA_H_EVTERRDBG (sdma_base + 0x034)
> > +#define SDMA_H_CONFIG (sdma_base + 0x038)
> > +#define SDMA_ONCE_ENB (sdma_base + 0x040)
> > +#define SDMA_ONCE_DATA (sdma_base + 0x044)
> > +#define SDMA_ONCE_INSTR (sdma_base + 0x048)
> > +#define SDMA_ONCE_STAT (sdma_base + 0x04c)
> > +#define SDMA_ONCE_CMD (sdma_base + 0x050)
> > +#define SDMA_EVT_MIRROR (sdma_base + 0x054)
> > +#define SDMA_ILLINSTADDR (sdma_base + 0x058)
> > +#define SDMA_CHN0ADDR (sdma_base + 0x05c)
> > +#define SDMA_ONCE_RTB (sdma_base + 0x060)
> > +#define SDMA_XTRIG_CONF1 (sdma_base + 0x070)
> > +#define SDMA_XTRIG_CONF2 (sdma_base + 0x074)
> > +#define SDMA_CHNENBL_0 (sdma_base + (sdma_version == 2 ? 0x200 : 0x80))
> > +#define SDMA_CHNPRI_0 (sdma_base + 0x100)
>
> All these rely on a fixed sdma_base which makes the driver
> a singleton. This is not so good if you imagine the situation with a
> platform with two SDMA engines on different addresses.
>
> Can't you create a runtime allocated stateholder to hold
> the base and access relative to the offset?
This could be done since a dma channel is a pointer now. Originally
a channel was referenced by its number only. Doing this would only
be for the beauty of the code though since I don't think there will
be ever more than one SDMA engine in one SoC. Famous last words...
>
> > +
> > +/*
> > + * Buffer descriptor status values.
> > + */
> > +#define BD_DONE 0x01
> > +#define BD_WRAP 0x02
> > +#define BD_CONT 0x04
> > +#define BD_INTR 0x08
> > +#define BD_RROR 0x10
> > +#define BD_LAST 0x20
> > +#define BD_EXTD 0x80
> > +
> > +/*
> > + * Data Node descriptor status values.
> > + */
> > +#define DND_END_OF_FRAME 0x80
> > +#define DND_END_OF_XFER 0x40
> > +#define DND_DONE 0x20
> > +#define DND_UNUSED 0x01
> > +
> > +/*
> > + * IPCV2 descriptor status values.
> > + */
> > +#define BD_IPCV2_END_OF_FRAME 0x40
> > +
> > +#define IPCV2_MAX_NODES 50
> > +/*
> > + * Error bit set in the CCB status field by the SDMA,
> > + * in setbd routine, in case of a transfer error
> > + */
> > +#define DATA_ERROR 0x10000000
> > +
> > +/*
> > + * Buffer descriptor commands.
> > + */
> > +#define C0_ADDR 0x01
> > +#define C0_LOAD 0x02
> > +#define C0_DUMP 0x03
> > +#define C0_SETCTX 0x07
> > +#define C0_GETCTX 0x03
> > +#define C0_SETDM 0x01
> > +#define C0_SETPM 0x04
> > +#define C0_GETDM 0x02
> > +#define C0_GETPM 0x08
> > +/*
> > + * Change endianness indicator in the BD command field
> > + */
> > +#define CHANGE_ENDIANNESS 0x80
> > +
> > +/*
> > + * Mode/Count of data node descriptors - IPCv2
> > + */
> > +#ifdef __BIG_ENDIAN
> > +struct sdma_mode_count {
> > + u32 command : 8; /* command mostlky used for channel 0 */
>
> There are a lot of inline commented struct members, please
> use kerneldoc, that's simple. (Applies all over the patch...)
> Documentation/kernel-doc-nano-HOWTO
Ok.
>
> > + u32 status : 8; /* E,R,I,C,W,D status bits stored here */
> > + u32 count : 16; /* size of the buffer pointed by this BD */
> > +};
> > +#else
> > +struct sdma_mode_count {
> > + u32 count : 16; /* size of the buffer pointed by this BD */
> > + u32 status : 8; /* E,R,I,C,W,D status bits stored here */
> > + u32 command : 8; /* command mostlky used for channel 0 */
> > +};
> > +#endif
>
> This use of #ifdef is odd to me but others are probably more
> experienced. Anyway, the way it is used with different
> :n suffixes makes me believe that you need a packed
> compiler directive for this layout to be explicitly coherent.
>
> Atleast add some comment on what this #ifdef construction
> does so guys like me can understand what's going on.
This is a direct copy from the Freescale code. Since Linux does not
support i.MX SoCs in big endian modes I think we can remove the ifdef
completely. Adding this again will be the smallest problem when we want
to add big endian mode in the future.
>
> > +
> > +/*
> > + * Buffer descriptor
> > + */
> > +struct sdma_buffer_descriptor {
> > + struct sdma_mode_count mode;
> > + u32 buffer_addr; /* address of the buffer described */
> > + u32 ext_buffer_addr; /* extended buffer address */
>
> Shouldn't these be dma_addr_t? OK that's probably u32
> anyway but just to make a marker...
>
> > +};
> > +
> > +/*
> > + * Channel control Block
> > + */
> > +struct sdma_channel_control {
> > + u32 current_bd_ptr; /* current buffer descriptor processed */
> > + u32 base_bd_ptr; /* first element of buffer descriptor array */
> > + void *unused;
> > + void *unused1;
>
> Hm, can you comment on what these unused things are for...?
The SDMA engine expects an array of these structures (one for each
channel). The unused fields are just to make the structure the correct
size. They should be of type u32 though.
>
> > +};
> > +
> > +/**
> > + * Context structure.
> > + */
> > +#ifdef __BIG_ENDIAN
> > +struct sdma_state_registers {
> > + u32 sf : 1; /* source fault while loading data */
> > + u32 unused0: 1;
> > + u32 rpc :14; /* return program counter */
> > + u32 t : 1; /* test bit:status of arithmetic & test instruction*/
> > + u32 unused1: 1;
> > + u32 pc :14; /* program counter */
> > + u32 lm : 2; /* loop mode */
> > + u32 epc :14; /* loop end program counter */
> > + u32 df : 1; /* destination fault while storing data */
> > + u32 unused2: 1;
> > + u32 spc :14; /* loop start program counter */
> > +};
> > +#else
> > +struct sdma_state_registers {
> > + u32 pc :14; /* program counter */
> > + u32 unused1: 1;
> > + u32 t : 1; /* test bit: status of arithmetic & test instruction*/
> > + u32 rpc :14; /* return program counter */
> > + u32 unused0: 1;
> > + u32 sf : 1; /* source fault while loading data */
> > + u32 spc :14; /* loop start program counter */
> > + u32 unused2: 1;
> > + u32 df : 1; /* destination fault while storing data */
> > + u32 epc :14; /* loop end program counter */
> > + u32 lm : 2; /* loop mode */
> > +};
> > +#endif
>
> Again this is odd to me...
>
> > +
> > +struct sdma_context_data {
> > + struct sdma_state_registers channel_state; /* channel state bits */
> > + u32 gReg[8]; /* general registers */
> > + u32 mda; /* burst dma destination address register */
> > + u32 msa; /* burst dma source address register */
> > + u32 ms; /* burst dma status register */
> > + u32 md; /* burst dma data register */
> > + u32 pda; /* peripheral dma destination address register */
> > + u32 psa; /* peripheral dma source address register */
> > + u32 ps; /* peripheral dma status register */
> > + u32 pd; /* peripheral dma data register */
> > + u32 ca; /* CRC polynomial register */
> > + u32 cs; /* CRC accumulator register */
> > + u32 dda; /* dedicated core destination address register */
> > + u32 dsa; /* dedicated core source address register */
> > + u32 ds; /* dedicated core status register */
> > + u32 dd; /* dedicated core data register */
> > + u32 scratch0;
> > + u32 scratch1;
> > + u32 scratch2;
> > + u32 scratch3;
> > + u32 scratch4;
> > + u32 scratch5;
> > + u32 scratch6;
> > + u32 scratch7;
> > +};
> > +
> > +struct sdma_channel {
> > + /* Channel number */
> > + int channel;
>
> Unsigned?
>
> > + /* Transfer type. Needed for setting SDMA script */
> > + enum dma_data_direction direction;
> > + /* Peripheral type. Needed for setting SDMA script */
> > + sdma_peripheral_type peripheral_type;
> > + /* Peripheral event id */
> > + int event_id;
>
> Unsigned?
>
> > + /* Peripheral event id2 (for channels that use 2 events) */
> > + int event_id2;
>
> Unsigned?
Ok for all the 'unsigned'. Will change.
>
> > + /* SDMA data access word size */
> > + unsigned long word_size;
>
> Is this in bits, bytes etc? Isn't e.g. an u8 enough to hold this,
> and further, isn't it possible to recycle enum dma_slave_buswidth
> from dmaengine.h instead?
Yes, will change.
>
> > +
> > + /* ID of the buffer that was processed */
> > + unsigned int buf_tail;
> > +
> > + wait_queue_head_t waitq; /* channel completion waitqeue */
> > +
> > + int num_bd;
>
> Unsigned? Range?
>
> > +
> > + struct sdma_buffer_descriptor *bd;
> > + dma_addr_t bd_phys;
> > +
> > + int pc_from_device, pc_to_device;
>
> Unsigned?
>
> > +
> > + unsigned long flags;
>
> Is this an u32?
There is no need to tie this to a particular size.
>
> > + dma_addr_t per_address;
> > +
> > + uint32_t event_mask1, event_mask2;
> > + uint32_t watermark_level;
> > + uint32_t shp_addr, per_addr;
> > +
> > + /* DMA-Engine Channel */
> > + struct dma_chan chan;
> > +
> > + spinlock_t lock;
> > + struct dma_async_tx_descriptor desc;
> > + dma_cookie_t last_completed;
> > + int busy;
>
> Shouldn't this be a bool?
ok
>
> > +};
> > +
> > +#define IMX_DMA_SG_LOOP (1 << 0)
> > +
> > +#define MAX_DMA_CHANNELS 32
> > +#define MXC_SDMA_DEFAULT_PRIORITY 1
> > +#define MXC_SDMA_MIN_PRIORITY 1
> > +#define MXC_SDMA_MAX_PRIORITY 7
> > +
> > +/*
> > + * This enumerates transfer types
> > + */
> > +typedef enum {
>
> Again a typedef, please plain enum is fine.
>
> > + emi_2_per = 0, /* EMI memory to peripheral */
> > + emi_2_int, /* EMI memory to internal RAM */
> > + emi_2_emi, /* EMI memory to EMI memory */
> > + emi_2_dsp, /* EMI memory to DSP memory */
> > + per_2_int, /* Peripheral to internal RAM */
> > + per_2_emi, /* Peripheral to internal EMI memory */
> > + per_2_dsp, /* Peripheral to DSP memory */
> > + per_2_per, /* Peripheral to Peripheral */
> > + int_2_per, /* Internal RAM to peripheral */
> > + int_2_int, /* Internal RAM to Internal RAM */
> > + int_2_emi, /* Internal RAM to EMI memory */
> > + int_2_dsp, /* Internal RAM to DSP memory */
> > + dsp_2_per, /* DSP memory to peripheral */
> > + dsp_2_int, /* DSP memory to internal RAM */
> > + dsp_2_emi, /* DSP memory to EMI memory */
> > + dsp_2_dsp, /* DSP memory to DSP memory */
> > + emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
> > + dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
> > + dvfs_pll, /* DVFS script with PLL change */
> > + dvfs_pdr /* DVFS script without PLL change */
> > +} sdma_transfer_type;
> > +
> > +/*
> > + * Structure containing sdma request parameters.
> > + */
> > +struct sdma_script_start_addrs {
> > + int ap_2_ap_addr;
> > + int ap_2_bp_addr;
> > + int ap_2_ap_fixed_addr;
> > + int bp_2_ap_addr;
> > + int loopback_on_dsp_side_addr;
> > + int mcu_interrupt_only_addr;
> > +
> > + int firi_2_per_addr;
> > + int firi_2_mcu_addr;
> > + int per_2_firi_addr;
> > + int mcu_2_firi_addr;
> > +
> > + int uart_2_per_addr;
> > + int uart_2_mcu_addr;
> > + int per_2_app_addr;
> > + int mcu_2_app_addr;
> > + int per_2_per_addr;
> > +
> > + int uartsh_2_per_addr;
> > + int uartsh_2_mcu_addr;
> > + int per_2_shp_addr;
> > + int mcu_2_shp_addr;
> > +
> > + int ata_2_mcu_addr;
> > + int mcu_2_ata_addr;
> > +
> > + int app_2_per_addr;
> > + int app_2_mcu_addr;
> > + int shp_2_per_addr;
> > + int shp_2_mcu_addr;
> > +
> > + int mshc_2_mcu_addr;
> > + int mcu_2_mshc_addr;
> > +
> > + int spdif_2_mcu_addr;
> > + int mcu_2_spdif_addr;
> > +
> > + int asrc_2_mcu_addr;
> > +
> > + int ext_mem_2_ipu_addr;
> > +
> > + int descrambler_addr;
> > +
> > + int dptc_dvfs_addr;
> > +
> > + int utra_addr;
> > +
> > + int ram_code_start_addr;
>
> All these addresses, are they really integers with
> valid negative values... Aren't they dma_addr_t or
> atleast u32?
Since this struct must match the layout of the firmware, they should be
u32, yes. They are no dma_addr_t since it's the sdma controller address
space described here.
>
> > +};
> > +
> > +#define SDMA_FIRMWARE_MAGIC 0x414d4453
> > +
> > +struct sdma_firmware_header {
> > + uint32_t magic; /* "SDMA" */
> > + uint32_t version_major; /* increased whenever layout of struct sdma_script_start_addrs changes */
> > + uint32_t version_minor; /* firmware version */
> > + uint32_t script_addrs_start; /* offset of struct sdma_script_start_addrs in this image */
> > + uint32_t num_script_addrs; /* Number of script addresses in this image */
> > + uint32_t ram_code_start; /* offset of SDMA ram image in this firmware image */
> > + uint32_t ram_code_size; /* size of SDMA ram image */
>
> Please use u32. uint32_t is not the preferred kernel type.
> (Still I've seen people use it in some cases so I might be wrong,
> feel welcome to bit back on this.)
At least one type should be consequently used in a driver. Changed them
all to u32.
>
> > +};
> > +
> > +static struct sdma_channel sdma_data[MAX_DMA_CHANNELS];
> > +static struct sdma_channel_control *channel_control;
> > +static void __iomem *sdma_base;
> > +static int sdma_version;
>
> Unsigned?
>
> > +static int sdma_num_events;
>
> Unsigned?
>
> > +static struct sdma_context_data *sdma_context;
> > +dma_addr_t sdma_context_phys;
> > +static struct dma_device __sdma_dma_device;
> > +static struct dma_device *sdma_dma_device = &__sdma_dma_device;
>
> This is what I suspected: local variables making the entire driver
> a singleton, which means you can never have more than one
> SDMA. Atleast collect all of these in a struct, call it
> "struct sdma" simply (if you ask me) and use as a stateholder.
> This makes it easier to kzalloc() that struct later if you
> want to support non-singletons.
>
> I know this require some work but I've done it to several drivers
> (always asked on mailinglists to do this) and I don't regret a single
> rewrite. Last time was for the PL18x DMAengine driver actually.
I've done it myself often enough. Ok, will change.
>
> > +
> > +#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
> > +#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
> > +#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
> > +#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
> > +
> > +static int sdma_config_ownership(int channel, int event_override,
> > + int mcu_verride, int dsp_override)
> > +{
> > + u32 evt, mcu, dsp;
> > +
> > + if (event_override && mcu_verride && dsp_override)
> > + return -EINVAL;
> > +
> > + evt = readl(SDMA_H_EVTOVR);
> > + mcu = readl(SDMA_H_HOSTOVR);
> > + dsp = readl(SDMA_H_DSPOVR);
> > +
> > + if (dsp_override)
> > + dsp &= ~(1 << channel);
> > + else
> > + dsp |= (1 << channel);
> > +
> > + if (event_override)
> > + evt &= ~(1 << channel);
> > + else
> > + evt |= (1 << channel);
> > +
> > + if (mcu_verride)
> > + mcu &= ~(1 << channel);
> > + else
> > + mcu |= (1 << channel);
> > +
> > + writel(evt, SDMA_H_EVTOVR);
> > + writel(mcu, SDMA_H_HOSTOVR);
> > + writel(dsp, SDMA_H_DSPOVR);
> > +
> > + return 0;
> > +}
> > +
> > +/*
> > + * sdma_run_channel - run a channel and wait till it's done
> > + */
> > +static int sdma_run_channel(int channel)
> > +{
> > + struct sdma_channel *sdma = &sdma_data[channel];
> > + int ret;
> > +
> > + writel(1 << channel, SDMA_H_START);
> > +
> > + ret = wait_event_interruptible(sdma->waitq,
> > + !(readl(SDMA_H_STATSTOP) & (1 << channel)));
>
> OK not the biggest thing in the world, but can't you use a
> completion for this? (I'm not so clever with waitqueues so
> forgive me if this is malinformed.)
Ok.
>
> > + return ret;
> > +}
> > +
> > +static int sdma_load_script(void *buf, int size, u32 address)
> > +{
> > + struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> > + void *buf_virt;
> > + dma_addr_t buf_phys;
> > + int ret;
> > +
> > + buf_virt = dma_alloc_coherent(NULL,
> > + size,
> > + &buf_phys, GFP_KERNEL);
> > + if (!buf_virt)
> > + return -ENOMEM;
> > +
> > + bd0->mode.command = C0_SETPM;
> > + bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> > + bd0->mode.count = size / 2;
> > + bd0->buffer_addr = buf_phys;
> > + bd0->ext_buffer_addr = address;
> > +
> > + memcpy(buf_virt, buf, size);
> > +
> > + ret = sdma_run_channel(0);
> > +
> > + dma_free_coherent(NULL, size, buf_virt, buf_phys);
> > +
> > + return ret;
> > +}
> > +
> > +static void sdma_event_enable(int channel, int event)
> > +{
> > + u32 val;
> > +
> > + val = readl(SDMA_CHNENBL_0 + event * 4);
>
> This use indicates that event should probably be
> unsigned, and probably not greater than u16 atleast.
> I suspect it is never more than an u8 really.
>
> > + val |= (1 << channel);
> > + writel(val, SDMA_CHNENBL_0 + event * 4);
> > +}
> > +
> > +static void sdma_event_disable(int channel, int event)
> > +{
> > + u32 val;
> > +
> > + val = readl(SDMA_CHNENBL_0 + event * 4);
> > + val &= ~(1 << channel);
> > + writel(val, SDMA_CHNENBL_0 + event * 4);
>
> Same comment here.
Ok, changed to unsigned and added a check for valid values in
sdma_config_channel.
>
> > +}
> > +
> > +static void mxc_sdma_handle_channel_loop(int channel)
> > +{
> > + struct sdma_channel *sdma = &sdma_data[channel];
>
> This indicates that channel should be unsigned.
>
> > + struct sdma_buffer_descriptor *bd;
> > + int error = 0;
>
> Unused variable?
Originally the callback function had a status parameter where errors
were signalled. I assume device_tx_status is the function to which this
error should be passed, right?
>
> > +
> > + /*
> > + * loop mode. Iterate over descriptors, re-setup them and
> > + * call callback function.
> > + */
> > + while (1) {
> > + bd = &sdma->bd[sdma->buf_tail];
> > +
> > + if (bd->mode.status & BD_DONE)
> > + break;
> > +
> > + if (bd->mode.status & BD_RROR)
> > + error = -EIO;
> > +
> > + bd->mode.status |= BD_DONE;
> > + sdma->buf_tail++;
> > + sdma->buf_tail %= sdma->num_bd;
> > +
> > + if (sdma->desc.callback)
> > + sdma->desc.callback(sdma->desc.callback_param);
> > + }
> > +}
> > +
> > +static void mxc_sdma_handle_channel_normal(int channel)
> > +{
> > + struct sdma_channel *sdma = &sdma_data[channel];
> > + struct sdma_buffer_descriptor *bd;
> > + int i, error = 0;
> > +
> > + /*
> > + * non loop mode. Iterate over all descriptors, collect
> > + * errors and call callback function
> > + */
> > + for (i = 0; i < sdma->num_bd; i++) {
> > + bd = &sdma->bd[i];
> > +
> > + if (bd->mode.status & (BD_DONE | BD_RROR))
> > + error = -EIO;
> > + }
> > +
> > + if (sdma->desc.callback)
> > + sdma->desc.callback(sdma->desc.callback_param);
> > + sdma->last_completed = sdma->desc.cookie;
> > +
> > + sdma->busy = 0;
>
> = true if you switch this to bool..
>
> > +}
> > +
> > +static void mxc_sdma_handle_channel(int channel)
> > +{
> > + struct sdma_channel *sdma = &sdma_data[channel];
> > +
> > + wake_up_interruptible(&sdma->waitq);
> > +
> > + /* not interested in channel 0 interrupts */
> > + if (!channel)
> > + return;
> > +
> > + if (sdma->flags & IMX_DMA_SG_LOOP)
> > + mxc_sdma_handle_channel_loop(channel);
> > + else
> > + mxc_sdma_handle_channel_normal(channel);
> > +}
> > +
> > +static irqreturn_t sdma_int_handler(int irq, void *dev_id)
> > +{
> > + u32 stat;
> > +
> > + stat = readl(SDMA_H_INTR);
> > + writel(stat, SDMA_H_INTR);
> > +
> > + while (stat) {
> > + int channel = fls(stat) - 1;
> > +
> > + mxc_sdma_handle_channel(channel);
> > +
> > + stat &= ~(1 << channel);
> > + }
> > +
> > + return IRQ_HANDLED;
> > +}
> > +
> > +static struct clk *sdma_clk;
> > +
> > +/*
> > + * Stores the start address of the SDMA scripts
> > + */
> > +static struct sdma_script_start_addrs __sdma_script_addrs;
> > +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
> > +
> > +/*
> > + * sets the pc of SDMA script according to the peripheral type
> > + */
> > +static void sdma_get_pc(struct sdma_channel *sdma,
> > + sdma_peripheral_type peripheral_type)
> > +{
> > + int res = 0;
> > + int per_2_emi = 0, emi_2_per = 0;
> > + int per_2_int = 0, int_2_per = 0;
> > + int per_2_per = 0, emi_2_emi = 0;
> > +
> > + sdma->pc_from_device = 0;
> > + sdma->pc_to_device = 0;
>
> There are a *lot* of local variables here, and only two of them
> are used eventually, at the end of the function. I cannot quite
> follow this, what is going on?
'per' is for peripheral (like sdhc, ssi unit or similar)
'int' is for internal SRAM
'emi' is for SDRAM
currently we only support transfers from per to emi and emi to per and
so the other variables are unused. We could make this function simpler
by removing the unused variables, but I suggest to keep them for making
it easier to support other transfer types later. I can add a comment
what these variables are for and why they are unused.
>
> Some like emi_2_emi seem to be totally unused.
>
> The types here look like some kind of enum or other
> similar construction is really what's being asked for
> here.
>
> > +
> > + switch (peripheral_type) {
> > + case IMX_DMATYPE_MEMORY:
> > + emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
> > + break;
> > + case IMX_DMATYPE_DSP:
> > + emi_2_per = sdma_script_addrs->bp_2_ap_addr;
> > + per_2_emi = sdma_script_addrs->ap_2_bp_addr;
> > + break;
> > + case IMX_DMATYPE_FIRI:
> > + per_2_int = sdma_script_addrs->firi_2_per_addr;
> > + per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
> > + int_2_per = sdma_script_addrs->per_2_firi_addr;
> > + emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
> > + break;
> > + case IMX_DMATYPE_UART:
> > + per_2_int = sdma_script_addrs->uart_2_per_addr;
> > + per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
> > + int_2_per = sdma_script_addrs->per_2_app_addr;
> > + emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> > + break;
> > + case IMX_DMATYPE_UART_SP:
> > + per_2_int = sdma_script_addrs->uartsh_2_per_addr;
> > + per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
> > + int_2_per = sdma_script_addrs->per_2_shp_addr;
> > + emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> > + break;
> > + case IMX_DMATYPE_ATA:
> > + per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
> > + emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
> > + break;
> > + case IMX_DMATYPE_CSPI:
> > + case IMX_DMATYPE_EXT:
> > + case IMX_DMATYPE_SSI:
> > + per_2_int = sdma_script_addrs->app_2_per_addr;
> > + per_2_emi = sdma_script_addrs->app_2_mcu_addr;
> > + int_2_per = sdma_script_addrs->per_2_app_addr;
> > + emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> > + break;
> > + case IMX_DMATYPE_SSI_SP:
> > + case IMX_DMATYPE_MMC:
> > + case IMX_DMATYPE_SDHC:
> > + case IMX_DMATYPE_CSPI_SP:
> > + case IMX_DMATYPE_ESAI:
> > + case IMX_DMATYPE_MSHC_SP:
> > + per_2_int = sdma_script_addrs->shp_2_per_addr;
> > + per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
> > + int_2_per = sdma_script_addrs->per_2_shp_addr;
> > + emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> > + break;
> > + case IMX_DMATYPE_ASRC:
> > + per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
> > + emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
> > + per_2_per = sdma_script_addrs->per_2_per_addr;
> > + break;
> > + case IMX_DMATYPE_MSHC:
> > + per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
> > + emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
> > + break;
> > + case IMX_DMATYPE_CCM:
> > + per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
> > + break;
> > + case IMX_DMATYPE_FIFO_MEMORY:
> > + res = sdma_script_addrs->ap_2_ap_fixed_addr;
>
> res? This thing is never used.
I have no idea what DMATYPE_FIFO_MEMORY is. Will remove this.
>
> > + break;
> > + case IMX_DMATYPE_SPDIF:
> > + per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
> > + emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
> > + break;
> > + case IMX_DMATYPE_IPU_MEMORY:
> > + emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
> > + break;
> > + default:
> > + break;
> > + }
> > +
> > + sdma->pc_from_device = per_2_emi;
> > + sdma->pc_to_device = emi_2_per;
>
> Return res? You're assigning it a value in some cases.
>
> > +}
> > +
> > +static int sdma_load_context(int channel)
> > +{
> > + struct sdma_channel *sdma = &sdma_data[channel];
> > + int load_address;
> > + struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> > + int ret;
> > +
> > + if (sdma->direction == DMA_FROM_DEVICE) {
> > + load_address = sdma->pc_from_device;
> > + } else {
> > + load_address = sdma->pc_to_device;
> > + }
> > +
> > + if (load_address < 0)
> > + return load_address;
> > +
> > + pr_debug("%s: load_address = %d\n", __func__, load_address);
> > + pr_debug("%s: wml = 0x%08x\n", __func__, sdma->watermark_level);
> > + pr_debug("%s: shp_addr = 0x%08x\n", __func__, sdma->shp_addr);
> > + pr_debug("%s: per_addr = 0x%08x\n", __func__, sdma->per_addr);
> > + pr_debug("%s: event_mask1 = 0x%08x\n", __func__, sdma->event_mask1);
> > + pr_debug("%s: event_mask2 = 0x%08x\n", __func__, sdma->event_mask2);
>
> Surely it must be possible to get the struct device * pointer for the
> channels host and use dev_dbg() instead?
Ok, will do
>
> > +
> > + memset(sdma_context, 0, sizeof(*sdma_context));
> > + sdma_context->channel_state.pc = load_address;
> > +
> > + /* Send by context the event mask,base address for peripheral
> > + * and watermark level
> > + */
> > + sdma_context->gReg[0] = sdma->event_mask2;
> > + sdma_context->gReg[1] = sdma->event_mask1;
> > + sdma_context->gReg[2] = sdma->per_addr;
> > + sdma_context->gReg[6] = sdma->shp_addr;
> > + sdma_context->gReg[7] = sdma->watermark_level;
> > +
> > + bd0->mode.command = C0_SETDM;
> > + bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> > + bd0->mode.count = sizeof(*sdma_context) / 4;
> > + bd0->buffer_addr = sdma_context_phys;
> > + bd0->ext_buffer_addr = 2048 + (sizeof(*sdma_context) / 4) * channel;
> > +
> > + ret = sdma_run_channel(0);
> > +
> > + return ret;
> > +}
> > +
> > +static void sdma_disable_channel(int channel)
> > +{
> > + struct sdma_channel *sdma = &sdma_data[channel];
> > +
> > + writel(1 << channel, SDMA_H_STATSTOP);
> > + sdma->busy = 0;
> > +}
> > +
> > +static int sdma_config_channel(int channel)
> > +{
> > + struct sdma_channel *sdma = &sdma_data[channel];
> > + int ret;
> > +
> > + sdma_disable_channel(channel);
> > +
> > + sdma->event_mask1 = 0;
> > + sdma->event_mask2 = 0;
> > + sdma->shp_addr = 0;
> > + sdma->per_addr = 0;
> > +
> > + if (sdma->event_id)
> > + sdma_event_enable(channel, sdma->event_id);
> > +
> > + switch (sdma->peripheral_type) {
> > + case IMX_DMATYPE_DSP:
> > + sdma_config_ownership(channel, 0, 1, 1);
>
> The parameters here makes yoy believe that the types should
> be bool rather than int...
ok.
>
> > + break;
> > + case IMX_DMATYPE_MEMORY:
> > + sdma_config_ownership(channel, 0, 1, 0);
> > + break;
> > + default:
> > + sdma_config_ownership(channel, 1, 1, 0);
> > + break;
> > + }
> > +
> > + sdma_get_pc(sdma, sdma->peripheral_type);
> > +
> > + if ((sdma->peripheral_type != IMX_DMATYPE_MEMORY) &&
> > + (sdma->peripheral_type != IMX_DMATYPE_DSP)) {
> > + /* Handle multiple event channels differently */
> > + if (sdma->event_id2) {
> > + sdma->event_mask2 = 1 << (sdma->event_id2 % 32);
> > + if (sdma->event_id2 > 31)
> > + sdma->watermark_level |= 1 << 31;
> > + sdma->event_mask1 = 1 << (sdma->event_id % 32);
> > + if (sdma->event_id > 31)
> > + sdma->watermark_level |= 1 << 30;
> > + } else {
> > + sdma->event_mask1 = 1 << sdma->event_id;
> > + sdma->event_mask2 = 1 << (sdma->event_id - 32);
> > + }
> > + /* Watermark Level */
> > + sdma->watermark_level |= sdma->watermark_level;
> > + /* Address */
> > + sdma->shp_addr = sdma->per_address;
> > + } else {
> > + sdma->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
> > + }
> > +
> > + ret = sdma_load_context(channel);
> > +
> > + return ret;
> > +}
> > +
> > +static int sdma_set_channel_priority(unsigned int channel, unsigned int priority)
> > +{
> > + if (priority < MXC_SDMA_MIN_PRIORITY
> > + || priority > MXC_SDMA_MAX_PRIORITY) {
> > + return -EINVAL;
> > + }
> > +
> > + writel(priority, SDMA_CHNPRI_0 + 4 * channel);
> > +
> > + return 0;
> > +}
> > +
> > +static int sdma_request_channel(int channel)
> > +{
> > + struct sdma_channel *sdma = &sdma_data[channel];
> > + int ret = -EBUSY;
> > +
> > + sdma->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdma->bd_phys, GFP_KERNEL);
> > + if (!sdma->bd) {
> > + ret = -ENOMEM;
> > + goto out;
> > + }
> > +
> > + memset(sdma->bd, 0, PAGE_SIZE);
> > +
> > + channel_control[channel].base_bd_ptr = sdma->bd_phys;
> > + channel_control[channel].current_bd_ptr = sdma->bd_phys;
> > +
> > + clk_enable(sdma_clk);
>
> Aha you're enabling it once for every channel and rely on
> clk reference counting that's clever!
>
> > +
> > + sdma_set_channel_priority(channel, MXC_SDMA_DEFAULT_PRIORITY);
> > +
> > + init_waitqueue_head(&sdma->waitq);
> > +
> > + sdma->buf_tail = 0;
> > +
> > + return 0;
> > +out:
> > +
> > + return ret;
> > +}
> > +
> > +static void sdma_enable_channel(int channel)
> > +{
> > + writel(1 << channel, SDMA_H_START);
> > +}
> > +
> > +static int __init sdma_init(unsigned long phys_base, int irq, int version,
> > + void *ram_code,
> > + int ram_code_size)
> > +{
> > + int i, ret;
> > + int channel;
> > + dma_addr_t ccb_phys;
> > +
> > + sdma_version = version;
> > + switch (sdma_version) {
> > + case 1:
> > + sdma_num_events = 32;
> > + break;
> > + case 2:
> > + sdma_num_events = 48;
> > + break;
> > + default:
> > + pr_err("SDMA: Unknown version %d. aborting\n", sdma_version);
> > + return -ENODEV;
> > + }
> > +
> > + clk_enable(sdma_clk);
> > +
> > + sdma_base = ioremap(phys_base, 4096);
>
> Use SZ_4K instead of 4096.
Or even better resouce_size(iores)
>
> > + if (!sdma_base) {
> > + ret = -ENOMEM;
> > + goto err_ioremap;
> > + }
> > +
> > + /* Initialize SDMA private data */
> > + memset(sdma_data, 0, sizeof(struct sdma_channel) * MAX_DMA_CHANNELS);
> > +
> > + for (channel = 0; channel < MAX_DMA_CHANNELS; channel++)
> > + sdma_data[channel].channel = channel;
> > +
> > + ret = request_irq(irq, sdma_int_handler, 0, "sdma", NULL);
> > + if (ret)
> > + goto err_request_irq;
> > +
> > + /* Be sure SDMA has not started yet */
> > + writel(0, SDMA_H_C0PTR);
> > +
> > + channel_control = dma_alloc_coherent(NULL,
> > + MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
> > + sizeof(struct sdma_context_data),
> > + &ccb_phys, GFP_KERNEL);
> > +
> > + if (!channel_control) {
> > + ret = -ENOMEM;
> > + goto err_dma_alloc;
> > + }
> > +
> > + sdma_context = (void *)channel_control +
> > + MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> > + sdma_context_phys = ccb_phys +
> > + MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> > +
> > + /* Zero-out the CCB structures array just allocated */
> > + memset(channel_control, 0,
> > + MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
> > +
> > + /* disable all channels */
> > + for (i = 0; i < sdma_num_events; i++)
> > + writel(0, SDMA_CHNENBL_0 + i * 4);
> > +
> > + /* All channels have priority 0 */
> > + for (i = 0; i < MAX_DMA_CHANNELS; i++)
> > + writel(0, SDMA_CHNPRI_0 + i * 4);
> > +
> > + ret = sdma_request_channel(0);
> > + if (ret)
> > + goto err_dma_alloc;
> > +
> > + sdma_config_ownership(0, 0, 1, 0);
> > +
> > + /* Set Command Channel (Channel Zero) */
> > + writel(0x4050, SDMA_CHN0ADDR);
> > +
> > + /* Set bits of CONFIG register but with static context switching */
> > + /* FIXME: Check whether to set ACR bit depending on clock ratios */
> > + writel(0, SDMA_H_CONFIG);
> > +
> > + writel(ccb_phys, SDMA_H_C0PTR);
> > +
> > + /* download the RAM image for SDMA */
> > + sdma_load_script(ram_code,
> > + ram_code_size,
> > + sdma_script_addrs->ram_code_start_addr);
> > +
> > + /* Set bits of CONFIG register with given context switching mode */
> > + writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
> > +
> > + /* Initializes channel's priorities */
> > + sdma_set_channel_priority(0, 7);
> > +
> > + clk_disable(sdma_clk);
> > +
> > + return 0;
> > +
> > +err_dma_alloc:
> > + free_irq(irq, NULL);
> > +err_request_irq:
> > + iounmap(sdma_base);
> > +err_ioremap:
> > + clk_disable(sdma_clk);
> > + pr_err("%s failed with %d\n", __func__, ret);
> > + return ret;
> > +}
> > +
> > +static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
> > +{
> > + dma_cookie_t cookie = sdma->chan.cookie;
> > +
> > + if (++cookie < 0)
> > + cookie = 1;
> > +
> > + sdma->chan.cookie = cookie;
> > + sdma->desc.cookie = cookie;
> > +
> > + return cookie;
> > +}
> > +
> > +static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
> > +{
> > + return container_of(chan, struct sdma_channel, chan);
> > +}
> > +
> > +static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
> > +{
> > + struct sdma_channel *sdma = to_sdma_chan(tx->chan);
> > + dma_cookie_t cookie;
> > +
> > + spin_lock_irq(&sdma->lock);
> > +
> > + cookie = sdma_assign_cookie(sdma);
> > +
> > + sdma_enable_channel(tx->chan->chan_id);
> > +
> > + spin_unlock_irq(&sdma->lock);
> > +
> > + return cookie;
> > +}
> > +
> > +static int sdma_alloc_chan_resources(struct dma_chan *chan)
> > +{
> > + struct sdma_channel *sdma = to_sdma_chan(chan);
> > + struct imx_dma_data *data = chan->private;
> > + int prio, ret;
> > +
> > + /* No need to execute this for internal channel 0 */
> > + if (!chan->chan_id)
> > + return 0;
> > +
> > + if (!data)
> > + return -EINVAL;
> > +
> > + switch (data->priority) {
> > + case DMA_PRIO_HIGH:
> > + prio = 3;
>
> Wait, aren't these enumerated?
> Add some enum sdma_channel_prio {}..
Hm, The SDMA engine has priorities from 1 to 7 from which we happen to use
the lowest priorities only. I think this should not be an enum.
(The DMA_PRIO_* are only used in an attempt to provide the same API for
the i.MX2 SoCs which have a different DMA engine which is not so
flexible)
>
>
> > + break;
> > + case DMA_PRIO_MEDIUM:
> > + prio = 2;
> > + break;
> > + case DMA_PRIO_LOW:
> > + default:
> > + prio = 1;
> > + break;
> > + }
> > +
> > + sdma->peripheral_type = data->peripheral_type;
> > + sdma->event_id = data->dma_request;
> > + ret = sdma_set_channel_priority(chan->chan_id, prio);
> > + if (ret)
> > + return ret;
> > +
> > + if (chan->chan_id) {
> > + ret = sdma_request_channel(chan->chan_id);
> > + if (ret)
> > + return ret;
> > + }
> > +
> > + dma_async_tx_descriptor_init(&sdma->desc, chan);
> > + sdma->desc.tx_submit = sdma_tx_submit;
> > + /* txd.flags will be overwritten in prep funcs */
> > + sdma->desc.flags = DMA_CTRL_ACK;
> > +
> > + return 0;
> > +}
> > +
> > +static void sdma_free_chan_resources(struct dma_chan *chan)
> > +{
> > + struct sdma_channel *sdma = to_sdma_chan(chan);
> > + int channel = chan->chan_id;
> > +
> > + sdma_disable_channel(channel);
> > +
> > + if (sdma->event_id)
> > + sdma_event_disable(channel, sdma->event_id);
> > + if (sdma->event_id2)
> > + sdma_event_disable(channel, sdma->event_id2);
> > +
> > + sdma->event_id = 0;
> > + sdma->event_id2 = 0;
> > +
> > + sdma_set_channel_priority(channel, 0);
> > +
> > + dma_free_coherent(NULL, PAGE_SIZE, sdma->bd, sdma->bd_phys);
> > +
> > + clk_disable(sdma_clk);
> > +}
> > +
> > +#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
> > +
> > +static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
> > + struct dma_chan *chan, struct scatterlist *sgl,
> > + unsigned int sg_len, enum dma_data_direction direction,
> > + unsigned long flags)
> > +{
> > + struct sdma_channel *sdma = to_sdma_chan(chan);
> > + int ret, i, count;
> > + int channel = chan->chan_id;
> > + struct scatterlist *sg;
> > +
> > + if (sdma->busy)
> > + return NULL;
> > + sdma->busy = 1;
> > +
> > + sdma->flags = 0;
>
> What are those flags anyway? I think you will need some
> #define:s for them.
There's only one currently: IMX_DMA_SG_LOOP indicating that we are
doing cyclic transfers.
>
> > +
> > + pr_debug("SDMA: setting up %d entries for channel %d.\n",
> > + sg_len, channel);
> > +
> > + sdma->direction = direction;
> > + ret = sdma_load_context(channel);
> > + if (ret)
> > + goto err_out;
> > +
> > + if (sg_len > NUM_BD) {
> > + pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> > + channel, sg_len, NUM_BD);
> > + ret = -EINVAL;
> > + goto err_out;
> > + }
> > +
> > + for_each_sg(sgl, sg, sg_len, i) {
> > + struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> > + int param;
> > +
> > + bd->buffer_addr = sgl->dma_address;
> > +
> > + count = sg->length;
> > +
> > + if (count > 0xffff) {
> > + pr_err("SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
> > + channel, count, 0xffff);
> > + ret = -EINVAL;
> > + goto err_out;
> > + }
> > +
> > + bd->mode.count = count;
> > +
> > + if (sdma->word_size > 4) {
> > + ret = -EINVAL;
> > + goto err_out;
> > + }
> > + if (sdma->word_size == 4)
> > + bd->mode.command = 0;
> > + else
> > + bd->mode.command = sdma->word_size;
> > +
> > + param = BD_DONE | BD_EXTD | BD_CONT;
> > +
> > + if (sdma->flags & IMX_DMA_SG_LOOP) {
> > + param |= BD_INTR;
> > + if (i + 1 == sg_len)
> > + param |= BD_WRAP;
> > + }
> > +
> > + if (i + 1 == sg_len)
> > + param |= BD_INTR;
> > +
> > + pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> > + i, count, sg->dma_address,
> > + param & BD_WRAP ? "wrap" : "",
> > + param & BD_INTR ? " intr" : "");
> > +
> > + bd->mode.status = param;
> > + }
> > +
> > + sdma->num_bd = sg_len;
> > + channel_control[channel].current_bd_ptr = sdma->bd_phys;
> > +
> > + return &sdma->desc;
> > +err_out:
> > + return NULL;
> > +}
> > +
> > +static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
> > + struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
> > + size_t period_len, enum dma_data_direction direction)
> > +{
> > + int num_periods = buf_len / period_len;
> > + struct sdma_channel *sdma = to_sdma_chan(chan);
> > + int channel = chan->chan_id;
> > + int ret, i = 0, buf = 0;
> > +
> > + pr_debug("%s channel: %d\n", __func__, channel);
>
> Must be possible to find struct device * and use dev_dbg()
>
> > +
> > + if (sdma->busy)
> > + return NULL;
> > +
> > + sdma->busy = 1;
> > +
> > + sdma->flags |= IMX_DMA_SG_LOOP;
> > + sdma->direction = direction;
> > + ret = sdma_load_context(channel);
> > + if (ret)
> > + goto err_out;
> > +
> > + if (num_periods > NUM_BD) {
> > + pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> > + channel, num_periods, NUM_BD);
> > + goto err_out;
> > + }
> > +
> > + if (period_len > 0xffff) {
> > + pr_err("SDMA channel %d: maximum period size exceeded: %d > %d\n",
> > + channel, period_len, 0xffff);
> > + goto err_out;
> > + }
> > +
> > + while (buf < buf_len) {
> > + struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> > + int param;
> > +
> > + bd->buffer_addr = dma_addr;
> > +
> > + bd->mode.count = period_len;
> > +
> > + if (sdma->word_size > 4)
> > + goto err_out;
> > + if (sdma->word_size == 4)
> > + bd->mode.command = 0;
> > + else
> > + bd->mode.command = sdma->word_size;
> > +
> > + param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
> > + if (i + 1 == num_periods)
> > + param |= BD_WRAP;
> > +
> > + pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> > + i, period_len, dma_addr,
> > + param & BD_WRAP ? "wrap" : "",
> > + param & BD_INTR ? " intr" : "");
> > +
> > + bd->mode.status = param;
> > +
> > + dma_addr += period_len;
> > + buf += period_len;
> > +
> > + i++;
> > + }
> > +
> > + sdma->num_bd = num_periods;
> > + channel_control[channel].current_bd_ptr = sdma->bd_phys;
> > +
> > + return &sdma->desc;
> > +err_out:
> > + sdma->busy = 0;
> > + return NULL;
> > +}
> > +
> > +static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> > + unsigned long arg)
> > +{
> > + struct sdma_channel *sdma = to_sdma_chan(chan);
> > + struct dma_slave_config *dmaengine_cfg = (void *)arg;
> > +
> > + switch (cmd) {
> > + case DMA_TERMINATE_ALL:
> > + sdma_disable_channel(chan->chan_id);
> > + return 0;
> > + case DMA_SLAVE_CONFIG:
> > + if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
> > + sdma->per_address = dmaengine_cfg->src_addr;
> > + sdma->watermark_level = dmaengine_cfg->src_maxburst;
> > + sdma->word_size = dmaengine_cfg->src_addr_width;
> > + } else {
> > + sdma->per_address = dmaengine_cfg->dst_addr;
> > + sdma->watermark_level = dmaengine_cfg->dst_maxburst;
> > + sdma->word_size = dmaengine_cfg->dst_addr_width;
> > + }
> > + return sdma_config_channel(chan->chan_id);
> > + default:
> > + return -ENOSYS;
> > + }
> > +
> > + return -EINVAL;
> > +}
> > +
> > +static enum dma_status sdma_tx_status(struct dma_chan *chan,
> > + dma_cookie_t cookie,
> > + struct dma_tx_state *txstate)
> > +{
> > + struct sdma_channel *sdma = to_sdma_chan(chan);
> > + dma_cookie_t last_used;
> > + enum dma_status ret;
> > +
> > + last_used = chan->cookie;
> > +
> > + ret = dma_async_is_complete(cookie, sdma->last_completed, last_used);
> > + dma_set_tx_state(txstate, sdma->last_completed, last_used, 0);
> > +
> > + return ret;
> > +}
> > +
> > +static void sdma_issue_pending(struct dma_chan *chan)
> > +{
> > + /*
> > + * Nothing to do. We only have a single descriptor
> > + */
> > +}
> > +
> > +static int __devinit sdma_probe(struct platform_device *pdev)
> > +{
> > + int ret;
> > + const struct firmware *fw;
> > + const struct sdma_firmware_header *header;
> > + const struct sdma_script_start_addrs *addr;
> > + int irq;
> > + unsigned short *ram_code;
> > + struct resource *iores;
> > + struct sdma_platform_data *pdata = pdev->dev.platform_data;
> > + int version;
> > + char *cpustr, *fwname;
> > + int i;
> > + dma_cap_mask_t mask;
> > +
> > + /* there can be only one */
> > + BUG_ON(sdma_base);
> > +
> > + iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > + irq = platform_get_irq(pdev, 0);
> > + if (!iores || irq < 0 || !pdata)
> > + return -EINVAL;
> > +
> > + sdma_clk = clk_get(&pdev->dev, NULL);
> > + if (IS_ERR(sdma_clk)) {
> > + ret = PTR_ERR(sdma_clk);
> > + goto err_clk;
> > + }
> > +
> > + if (cpu_is_mx31()) {
> > + cpustr = "imx31";
> > + version = mx31_revision() >> 4;
> > + } else if (cpu_is_mx35()) {
> > + cpustr = "imx35";
> > +/* FIXME: version = mx35_revision(); */
> > + version = 2;
> > + } else {
> > + ret = -EINVAL;
> > + goto err_cputype;
> > + }
> > +
> > + fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin", cpustr, version);
> > + if (!fwname) {
> > + ret = -ENOMEM;
> > + goto err_cputype;
> > + }
> > +
> > + ret = request_firmware(&fw, fwname, &pdev->dev);
> > + if (ret) {
> > + dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
> > + fwname, ret);
> > + kfree(fwname);
> > + goto err_cputype;
> > + }
> > + kfree(fwname);
> > +
> > + if (fw->size < sizeof(*header))
> > + goto err_firmware;
> > +
> > + header = (struct sdma_firmware_header *)fw->data;
> > +
> > + if (header->magic != SDMA_FIRMWARE_MAGIC)
> > + goto err_firmware;
> > + if (header->ram_code_start + header->ram_code_size > fw->size)
> > + goto err_firmware;
> > +
> > + addr = (void *)header + header->script_addrs_start;
> > + ram_code = (void *)header + header->ram_code_start;
> > + memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
> > +
> > + ret = sdma_init(iores->start, irq, pdata->sdma_version,
> > + ram_code, header->ram_code_size);
> > + if (ret)
> > + goto err_firmware;
> > +
> > + INIT_LIST_HEAD(&sdma_dma_device->channels);
> > +
> > + /* Initialize channel parameters */
> > + for (i = 0; i < MAX_DMA_CHANNELS; i++) {
> > + struct sdma_channel *sdma = &sdma_data[i];
> > +
> > + spin_lock_init(&sdma->lock);
> > +
> > + dma_cap_set(DMA_SLAVE, sdma_dma_device->cap_mask);
> > + dma_cap_set(DMA_CYCLIC, sdma_dma_device->cap_mask);
> > +
> > + sdma->chan.device = sdma_dma_device;
> > + sdma->chan.chan_id = i;
> > +
> > + /* Add the channel to the DMAC list */
> > + list_add_tail(&sdma->chan.device_node, &sdma_dma_device->channels);
> > + }
> > +
> > + sdma_dma_device->dev = &pdev->dev;
> > +
> > + sdma_dma_device->device_alloc_chan_resources = sdma_alloc_chan_resources;
> > + sdma_dma_device->device_free_chan_resources = sdma_free_chan_resources;
> > + sdma_dma_device->device_tx_status = sdma_tx_status;
> > + sdma_dma_device->device_prep_slave_sg = sdma_prep_slave_sg;
> > + sdma_dma_device->device_prep_dma_cyclic = sdma_prep_dma_cyclic;
> > + sdma_dma_device->device_control = sdma_control;
> > + sdma_dma_device->device_issue_pending = sdma_issue_pending;
> > +
> > + ret = dma_async_device_register(sdma_dma_device);
> > + if (ret) {
> > + dev_err(&pdev->dev, "unable to register DMAC\n");
>
> SDMAC even?
Better just "unable to register". The name of the device will give
enough information.
>
> > + goto err_firmware;
> > + }
> > +
> > + dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
> > + header->version_major,
> > + header->version_minor);
> > +
> > + /* request channel 0. This is an internal control channel
> > + * to the SDMA engine and not available to clients.
> > + */
> > + dma_cap_zero(mask);
> > + dma_cap_set(DMA_SLAVE, mask);
> > + dma_request_channel(mask, NULL, NULL);
> > +
> > + release_firmware(fw);
> > +
> > + return 0;
> > +
> > +err_firmware:
> > + release_firmware(fw);
> > +err_cputype:
> > + clk_put(sdma_clk);
> > +err_clk:
> > + return 0;
> > +}
> > +
> > +static int __devexit sdma_remove(struct platform_device *pdev)
> > +{
> > + return -EBUSY;
> > +}
> > +
> > +static struct platform_driver sdma_driver = {
> > + .driver = {
> > + .name = "imx-sdma",
> > + },
> > + .probe = sdma_probe,
> > + .remove = __devexit_p(sdma_remove),
> > +};
> > +
> > +static int __init sdma_module_init(void)
> > +{
> > + return platform_driver_register(&sdma_driver);
> > +}
> > +subsys_initcall(sdma_module_init);
> > +
> > +MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
> > +MODULE_DESCRIPTION("i.MX SDMA driver");
> > +MODULE_LICENSE("GPL");
> > --
> > 1.7.1
>
> Thanks for using this API
> Sascha!
>
> Yours,
> Linus Walleij
>
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-16 14:15 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-16 14:15 UTC (permalink / raw)
To: linux-arm-kernel
Hi Linus,
Thank you for the review. Sorry for so many trival mistakes in the code,
but I'm staring at this code in many different variants for some time
now and I'm getting blind on it.
On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> 2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
>
> > This patch adds support for the Freescale i.MX SDMA engine.
>
> I like it!
>
> > The SDMA engine is a scatter/gather DMA engine which is implemented
> > as a seperate coprocessor. SDMA needs its own firmware which is
> > requested using the standard request_firmware mechanism. The firmware
> > has different entry points for each peripheral type, so drivers
> > have to pass the peripheral type to the DMA engine which in turn
> > picks the correct firmware entry point from a table contained in
> > the firmware image itself.
>
> Quite fun, if the spec for the microcode is open this opens up
> for dynamic firmware generation for specific DMA jobs does it
> not?
Unfortunately the specs are not open, so we are sticked to the binary
microcode from Freescale. I'm pretty sure though that the SDMA engine
could do at least a device_prep_dma_xor operation.
>
> > I took a very simple approach to implement dmaengine support. Only
> > a single descriptor is statically assigned to a each channel. This
> > means that transfers can't be queued up but only a single transfer
> > is in progress. This simplifies implementation a lot and is sufficient
> > for the usual device/memory transfers.
>
> If you want to add memcpy() capability later you're gonna need
> this I think, but you can take that when that need arise.
Yes, I left this as an exercise for those who want to have this feature ;)
I think it's better to have tested code in the Kernel than having a
complicated list handling which is completely untested for anything
other than a single entry in the list.
>
> >(...)
> > +++ b/arch/arm/plat-mxc/include/mach/dma.h
> > @@ -0,0 +1,64 @@
> > +/*
> > + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> > + *
> > + * This program is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License version 2 as
> > + * published by the Free Software Foundation.
> > + */
> > +
> > +#ifndef __ASM_ARCH_MXC_DMA_H__
> > +#define __ASM_ARCH_MXC_DMA_H__
> > +
> > +#include <linux/scatterlist.h>
> > +
> > +/*
> > + * This enumerates peripheral types. Used for SDMA.
> > + */
> > +typedef enum {
>
> The kernel is not really happy about typedefs, can't this be a
> regular enum?
>
> > + ? ? ? IMX_DMATYPE_SSI, ? ? ? ?/* MCU domain SSI */
> > + ? ? ? IMX_DMATYPE_SSI_SP, ? ? /* Shared SSI */
> > + ? ? ? IMX_DMATYPE_MMC, ? ? ? ?/* MMC */
> > + ? ? ? IMX_DMATYPE_SDHC, ? ? ? /* SDHC */
> > + ? ? ? IMX_DMATYPE_UART, ? ? ? /* MCU domain UART */
> > + ? ? ? IMX_DMATYPE_UART_SP, ? ?/* Shared UART */
> > + ? ? ? IMX_DMATYPE_FIRI, ? ? ? /* FIRI */
> > + ? ? ? IMX_DMATYPE_CSPI, ? ? ? /* MCU domain CSPI */
> > + ? ? ? IMX_DMATYPE_CSPI_SP, ? ?/* Shared CSPI */
> > + ? ? ? IMX_DMATYPE_SIM, ? ? ? ?/* SIM */
> > + ? ? ? IMX_DMATYPE_ATA, ? ? ? ?/* ATA */
> > + ? ? ? IMX_DMATYPE_CCM, ? ? ? ?/* CCM */
> > + ? ? ? IMX_DMATYPE_EXT, ? ? ? ?/* External peripheral */
> > + ? ? ? IMX_DMATYPE_MSHC, ? ? ? /* Memory Stick Host Controller */
> > + ? ? ? IMX_DMATYPE_MSHC_SP, ? ?/* Shared Memory Stick Host Controller */
> > + ? ? ? IMX_DMATYPE_DSP, ? ? ? ?/* DSP */
> > + ? ? ? IMX_DMATYPE_MEMORY, ? ? /* Memory */
> > + ? ? ? IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
> > + ? ? ? IMX_DMATYPE_SPDIF, ? ? ?/* SPDIF */
> > + ? ? ? IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
> > + ? ? ? IMX_DMATYPE_ASRC, ? ? ? /* ASRC */
> > + ? ? ? IMX_DMATYPE_ESAI, ? ? ? /* ESAI */
> > +} sdma_peripheral_type;
> > +
> > +enum imx_dma_prio {
> > + ? ? ? DMA_PRIO_HIGH = 0,
> > + ? ? ? DMA_PRIO_MEDIUM = 1,
> > + ? ? ? DMA_PRIO_LOW = 2
> > +};
> > +
> > +struct imx_dma_data {
> > + ? ? ? int dma_request; /* DMA request line */
>
> Can this be negative and what is the range? I would
> suspect something like u8 or u16 would surely be more
> apropriate...
>
> > + ? ? ? sdma_peripheral_type peripheral_type;
> > + ? ? ? int priority;
>
> Isn't this an enum imx_dma_prio?
>
> > +};
> > +
> > +static inline int imx_dma_is_ipu(struct dma_chan *chan)
> > +{
> > + ? ? ? return !strcmp(dev_name(chan->device->dev), "ipu-core");
> > +}
> > +
> > +static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
> > +{
> > + ? ? ? return !strcmp(dev_name(chan->device->dev), "imx-sdma");
> > +}
> > +
> > +#endif
> > diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
> > new file mode 100644
> > index 0000000..5d542b8
> > --- /dev/null
> > +++ b/arch/arm/plat-mxc/include/mach/sdma.h
> > @@ -0,0 +1,8 @@
> > +#ifndef __MACH_MXC_SDMA_H__
> > +#define __MACH_MXC_SDMA_H__
> > +
> > +struct sdma_platform_data {
> > + ? ? ? int sdma_version;
>
> Do you have negative versions or can it be unsigned?
nope, will change this to an unsigned type.
>
> > +};
> > +
> > +#endif /* __MACH_MXC_SDMA_H__ */
> > diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
> > index 9520cf0..f76bda9 100644
> > --- a/drivers/dma/Kconfig
> > +++ b/drivers/dma/Kconfig
> > @@ -195,6 +195,14 @@ config PCH_DMA
> > ? ? ? ?help
> > ? ? ? ? ?Enable support for the Topcliff PCH DMA engine.
> >
> > +config IMX_SDMA
> > + ? ? ? tristate "Atmel AHB DMA support"
> > + ? ? ? depends on ARCH_MXC
> > + ? ? ? select DMA_ENGINE
> > + ? ? ? help
> > + ? ? ? ? Support the i.MX SDMA engine. This engine is integrated into
> > + ? ? ? ? Freescale i.MX25/31/35/51 chips.
> > +
> > ?config DMA_ENGINE
> > ? ? ? ?bool
> >
> > diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
> > index 72bd703..14d7a1b 100644
> > --- a/drivers/dma/Makefile
> > +++ b/drivers/dma/Makefile
> > @@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
> > ?obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
> > ?obj-$(CONFIG_PL330_DMA) += pl330.o
> > ?obj-$(CONFIG_PCH_DMA) += pch_dma.o
> > +obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
> > diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> > new file mode 100644
> > index 0000000..3ba7905
> > --- /dev/null
> > +++ b/drivers/dma/imx-sdma.c
> > @@ -0,0 +1,1383 @@
> > +/*
> > + * drivers/dma/imx-sdma.c
> > + *
> > + * This file contains a driver for the Freescale Smart DMA engine
> > + *
> > + * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
> > + *
> > + * Based on code from Freescale:
> > + *
> > + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> > + *
> > + * The code contained herein is licensed under the GNU General Public
> > + * License. You may obtain a copy of the GNU General Public License
> > + * Version 2 or later at the following locations:
> > + *
> > + * http://www.opensource.org/licenses/gpl-license.html
> > + * http://www.gnu.org/copyleft/gpl.html
> > + */
> > +#include <linux/init.h>
> > +#include <linux/types.h>
> > +#include <linux/mm.h>
> > +#include <linux/interrupt.h>
> > +#include <linux/clk.h>
> > +#include <linux/semaphore.h>
> > +#include <linux/spinlock.h>
> > +#include <linux/device.h>
> > +#include <linux/dma-mapping.h>
> > +#include <linux/firmware.h>
> > +#include <linux/slab.h>
> > +#include <linux/platform_device.h>
> > +#include <linux/dmaengine.h>
> > +
> > +#include <asm/irq.h>
> > +#include <mach/sdma.h>
> > +#include <mach/dma.h>
> > +#include <mach/hardware.h>
> > +
> > +/* SDMA registers */
> > +#define SDMA_H_C0PTR ? ? ? ? ? (sdma_base + 0x000)
> > +#define SDMA_H_INTR ? ? ? ? ? ?(sdma_base + 0x004)
> > +#define SDMA_H_STATSTOP ? ? ? ? ? ? ? ?(sdma_base + 0x008)
> > +#define SDMA_H_START ? ? ? ? ? (sdma_base + 0x00c)
> > +#define SDMA_H_EVTOVR ? ? ? ? ?(sdma_base + 0x010)
> > +#define SDMA_H_DSPOVR ? ? ? ? ?(sdma_base + 0x014)
> > +#define SDMA_H_HOSTOVR ? ? ? ? (sdma_base + 0x018)
> > +#define SDMA_H_EVTPEND ? ? ? ? (sdma_base + 0x01c)
> > +#define SDMA_H_DSPENBL ? ? ? ? (sdma_base + 0x020)
> > +#define SDMA_H_RESET ? ? ? ? ? (sdma_base + 0x024)
> > +#define SDMA_H_EVTERR ? ? ? ? ?(sdma_base + 0x028)
> > +#define SDMA_H_INTRMSK ? ? ? ? (sdma_base + 0x02c)
> > +#define SDMA_H_PSW ? ? ? ? ? ? (sdma_base + 0x030)
> > +#define SDMA_H_EVTERRDBG ? ? ? (sdma_base + 0x034)
> > +#define SDMA_H_CONFIG ? ? ? ? ?(sdma_base + 0x038)
> > +#define SDMA_ONCE_ENB ? ? ? ? ?(sdma_base + 0x040)
> > +#define SDMA_ONCE_DATA ? ? ? ? (sdma_base + 0x044)
> > +#define SDMA_ONCE_INSTR ? ? ? ? ? ? ? ?(sdma_base + 0x048)
> > +#define SDMA_ONCE_STAT ? ? ? ? (sdma_base + 0x04c)
> > +#define SDMA_ONCE_CMD ? ? ? ? ?(sdma_base + 0x050)
> > +#define SDMA_EVT_MIRROR ? ? ? ? ? ? ? ?(sdma_base + 0x054)
> > +#define SDMA_ILLINSTADDR ? ? ? (sdma_base + 0x058)
> > +#define SDMA_CHN0ADDR ? ? ? ? ?(sdma_base + 0x05c)
> > +#define SDMA_ONCE_RTB ? ? ? ? ?(sdma_base + 0x060)
> > +#define SDMA_XTRIG_CONF1 ? ? ? (sdma_base + 0x070)
> > +#define SDMA_XTRIG_CONF2 ? ? ? (sdma_base + 0x074)
> > +#define SDMA_CHNENBL_0 ? ? ? ? (sdma_base + (sdma_version == 2 ? 0x200 : 0x80))
> > +#define SDMA_CHNPRI_0 ? ? ? ? ?(sdma_base + 0x100)
>
> All these rely on a fixed sdma_base which makes the driver
> a singleton. This is not so good if you imagine the situation with a
> platform with two SDMA engines on different addresses.
>
> Can't you create a runtime allocated stateholder to hold
> the base and access relative to the offset?
This could be done since a dma channel is a pointer now. Originally
a channel was referenced by its number only. Doing this would only
be for the beauty of the code though since I don't think there will
be ever more than one SDMA engine in one SoC. Famous last words...
>
> > +
> > +/*
> > + * Buffer descriptor status values.
> > + */
> > +#define BD_DONE ?0x01
> > +#define BD_WRAP ?0x02
> > +#define BD_CONT ?0x04
> > +#define BD_INTR ?0x08
> > +#define BD_RROR ?0x10
> > +#define BD_LAST ?0x20
> > +#define BD_EXTD ?0x80
> > +
> > +/*
> > + * Data Node descriptor status values.
> > + */
> > +#define DND_END_OF_FRAME ?0x80
> > +#define DND_END_OF_XFER ? 0x40
> > +#define DND_DONE ? ? ? ? ?0x20
> > +#define DND_UNUSED ? ? ? ?0x01
> > +
> > +/*
> > + * IPCV2 descriptor status values.
> > + */
> > +#define BD_IPCV2_END_OF_FRAME ?0x40
> > +
> > +#define IPCV2_MAX_NODES ? ? ? ?50
> > +/*
> > + * Error bit set in the CCB status field by the SDMA,
> > + * in setbd routine, in case of a transfer error
> > + */
> > +#define DATA_ERROR ?0x10000000
> > +
> > +/*
> > + * Buffer descriptor commands.
> > + */
> > +#define C0_ADDR ? ? ? ? ? ? 0x01
> > +#define C0_LOAD ? ? ? ? ? ? 0x02
> > +#define C0_DUMP ? ? ? ? ? ? 0x03
> > +#define C0_SETCTX ? ? ? ? ? 0x07
> > +#define C0_GETCTX ? ? ? ? ? 0x03
> > +#define C0_SETDM ? ? ? ? ? ?0x01
> > +#define C0_SETPM ? ? ? ? ? ?0x04
> > +#define C0_GETDM ? ? ? ? ? ?0x02
> > +#define C0_GETPM ? ? ? ? ? ?0x08
> > +/*
> > + * Change endianness indicator in the BD command field
> > + */
> > +#define CHANGE_ENDIANNESS ? 0x80
> > +
> > +/*
> > + * Mode/Count of data node descriptors - IPCv2
> > + */
> > +#ifdef __BIG_ENDIAN
> > +struct sdma_mode_count {
> > + ? ? ? u32 command : ?8; /* command mostlky used for channel 0 */
>
> There are a lot of inline commented struct members, please
> use kerneldoc, that's simple. (Applies all over the patch...)
> Documentation/kernel-doc-nano-HOWTO
Ok.
>
> > + ? ? ? u32 status ?: ?8; /* E,R,I,C,W,D status bits stored here */
> > + ? ? ? u32 count ? : 16; /* size of the buffer pointed by this BD */
> > +};
> > +#else
> > +struct sdma_mode_count {
> > + ? ? ? u32 count ? : 16; /* size of the buffer pointed by this BD */
> > + ? ? ? u32 status ?: ?8; /* E,R,I,C,W,D status bits stored here */
> > + ? ? ? u32 command : ?8; /* command mostlky used for channel 0 */
> > +};
> > +#endif
>
> This use of #ifdef is odd to me but others are probably more
> experienced. Anyway, the way it is used with different
> :n suffixes makes me believe that you need a packed
> compiler directive for this layout to be explicitly coherent.
>
> Atleast add some comment on what this #ifdef construction
> does so guys like me can understand what's going on.
This is a direct copy from the Freescale code. Since Linux does not
support i.MX SoCs in big endian modes I think we can remove the ifdef
completely. Adding this again will be the smallest problem when we want
to add big endian mode in the future.
>
> > +
> > +/*
> > + * Buffer descriptor
> > + */
> > +struct sdma_buffer_descriptor {
> > + ? ? ? struct sdma_mode_count ?mode;
> > + ? ? ? u32 buffer_addr; ? ?/* address of the buffer described */
> > + ? ? ? u32 ext_buffer_addr; /* extended buffer address */
>
> Shouldn't these be dma_addr_t? OK that's probably u32
> anyway but just to make a marker...
>
> > +};
> > +
> > +/*
> > + * Channel control Block
> > + */
> > +struct sdma_channel_control {
> > + ? ? ? u32 current_bd_ptr; /* current buffer descriptor processed */
> > + ? ? ? u32 base_bd_ptr; ? ?/* first element of buffer descriptor array */
> > + ? ? ? void *unused;
> > + ? ? ? void *unused1;
>
> Hm, can you comment on what these unused things are for...?
The SDMA engine expects an array of these structures (one for each
channel). The unused fields are just to make the structure the correct
size. They should be of type u32 though.
>
> > +};
> > +
> > +/**
> > + * Context structure.
> > + */
> > +#ifdef __BIG_ENDIAN
> > +struct sdma_state_registers {
> > + ? ? ? u32 sf ? ? : 1; /* source fault while loading data */
> > + ? ? ? u32 unused0: 1;
> > + ? ? ? u32 rpc ? ?:14; /* return program counter */
> > + ? ? ? u32 t ? ? ?: 1; /* test bit:status of arithmetic & test instruction*/
> > + ? ? ? u32 unused1: 1;
> > + ? ? ? u32 pc ? ? :14; /* program counter */
> > + ? ? ? u32 lm ? ? : 2; /* loop mode */
> > + ? ? ? u32 epc ? ?:14; /* loop end program counter */
> > + ? ? ? u32 df ? ? : 1; /* destination fault while storing data */
> > + ? ? ? u32 unused2: 1;
> > + ? ? ? u32 spc ? ?:14; /* loop start program counter */
> > +};
> > +#else
> > +struct sdma_state_registers {
> > + ? ? ? u32 pc ? ? :14; /* program counter */
> > + ? ? ? u32 unused1: 1;
> > + ? ? ? u32 t ? ? ?: 1; /* test bit: status of arithmetic & test instruction*/
> > + ? ? ? u32 rpc ? ?:14; /* return program counter */
> > + ? ? ? u32 unused0: 1;
> > + ? ? ? u32 sf ? ? : 1; /* source fault while loading data */
> > + ? ? ? u32 spc ? ?:14; /* loop start program counter */
> > + ? ? ? u32 unused2: 1;
> > + ? ? ? u32 df ? ? : 1; /* destination fault while storing data */
> > + ? ? ? u32 epc ? ?:14; /* loop end program counter */
> > + ? ? ? u32 lm ? ? : 2; /* loop mode */
> > +};
> > +#endif
>
> Again this is odd to me...
>
> > +
> > +struct sdma_context_data {
> > + ? ? ? struct sdma_state_registers ?channel_state; /* channel state bits */
> > + ? ? ? u32 ?gReg[8]; /* general registers */
> > + ? ? ? u32 ?mda; /* burst dma destination address register */
> > + ? ? ? u32 ?msa; /* burst dma source address register */
> > + ? ? ? u32 ?ms; ?/* burst dma status ?register */
> > + ? ? ? u32 ?md; ?/* burst dma data ? ?register */
> > + ? ? ? u32 ?pda; /* peripheral dma destination address register */
> > + ? ? ? u32 ?psa; /* peripheral dma source address register */
> > + ? ? ? u32 ?ps; ?/* peripheral dma ?status ?register */
> > + ? ? ? u32 ?pd; ?/* peripheral dma ?data ? ?register */
> > + ? ? ? u32 ?ca; ?/* CRC polynomial ?register */
> > + ? ? ? u32 ?cs; ?/* CRC accumulator register */
> > + ? ? ? u32 ?dda; /* dedicated core destination address register */
> > + ? ? ? u32 ?dsa; /* dedicated core source address register */
> > + ? ? ? u32 ?ds; ?/* dedicated core status ?register */
> > + ? ? ? u32 ?dd; ?/* dedicated core data ? ?register */
> > + ? ? ? u32 ?scratch0;
> > + ? ? ? u32 ?scratch1;
> > + ? ? ? u32 ?scratch2;
> > + ? ? ? u32 ?scratch3;
> > + ? ? ? u32 ?scratch4;
> > + ? ? ? u32 ?scratch5;
> > + ? ? ? u32 ?scratch6;
> > + ? ? ? u32 ?scratch7;
> > +};
> > +
> > +struct sdma_channel {
> > + ? ? ? /* Channel number */
> > + ? ? ? int channel;
>
> Unsigned?
>
> > + ? ? ? /* Transfer type. Needed for setting SDMA script */
> > + ? ? ? enum dma_data_direction direction;
> > + ? ? ? /* Peripheral type. Needed for setting SDMA script */
> > + ? ? ? sdma_peripheral_type peripheral_type;
> > + ? ? ? /* Peripheral event id */
> > + ? ? ? int event_id;
>
> Unsigned?
>
> > + ? ? ? /* Peripheral event id2 (for channels that use 2 events) */
> > + ? ? ? int event_id2;
>
> Unsigned?
Ok for all the 'unsigned'. Will change.
>
> > + ? ? ? /* SDMA data access word size */
> > + ? ? ? unsigned long word_size;
>
> Is this in bits, bytes etc? Isn't e.g. an u8 enough to hold this,
> and further, isn't it possible to recycle enum dma_slave_buswidth
> from dmaengine.h instead?
Yes, will change.
>
> > +
> > + ? ? ? /* ID of the buffer that was processed */
> > + ? ? ? unsigned int buf_tail;
> > +
> > + ? ? ? wait_queue_head_t waitq; ? ? ? ?/* channel completion waitqeue */
> > +
> > + ? ? ? int num_bd;
>
> Unsigned? Range?
>
> > +
> > + ? ? ? struct sdma_buffer_descriptor *bd;
> > + ? ? ? dma_addr_t ? ? ?bd_phys;
> > +
> > + ? ? ? int pc_from_device, pc_to_device;
>
> Unsigned?
>
> > +
> > + ? ? ? unsigned long flags;
>
> Is this an u32?
There is no need to tie this to a particular size.
>
> > + ? ? ? dma_addr_t per_address;
> > +
> > + ? ? ? uint32_t event_mask1, event_mask2;
> > + ? ? ? uint32_t watermark_level;
> > + ? ? ? uint32_t shp_addr, per_addr;
> > +
> > + ? ? ? /* DMA-Engine Channel */
> > + ? ? ? struct dma_chan chan;
> > +
> > + ? ? ? spinlock_t ? ? ? ? ? ? ?lock;
> > + ? ? ? struct dma_async_tx_descriptor desc;
> > + ? ? ? dma_cookie_t ? ? ? ? ? ?last_completed;
> > + ? ? ? int busy;
>
> Shouldn't this be a bool?
ok
>
> > +};
> > +
> > +#define IMX_DMA_SG_LOOP ? ? ? ? ? ? ? ?(1 << 0)
> > +
> > +#define MAX_DMA_CHANNELS 32
> > +#define MXC_SDMA_DEFAULT_PRIORITY 1
> > +#define MXC_SDMA_MIN_PRIORITY 1
> > +#define MXC_SDMA_MAX_PRIORITY 7
> > +
> > +/*
> > + * This enumerates transfer types
> > + */
> > +typedef enum {
>
> Again a typedef, please plain enum is fine.
>
> > + ? ? ? emi_2_per = 0, ? ? ? ? ?/* EMI memory to peripheral */
> > + ? ? ? emi_2_int, ? ? ? ? ? ? ?/* EMI memory to internal RAM */
> > + ? ? ? emi_2_emi, ? ? ? ? ? ? ?/* EMI memory to EMI memory */
> > + ? ? ? emi_2_dsp, ? ? ? ? ? ? ?/* EMI memory to DSP memory */
> > + ? ? ? per_2_int, ? ? ? ? ? ? ?/* Peripheral to internal RAM */
> > + ? ? ? per_2_emi, ? ? ? ? ? ? ?/* Peripheral to internal EMI memory */
> > + ? ? ? per_2_dsp, ? ? ? ? ? ? ?/* Peripheral to DSP memory */
> > + ? ? ? per_2_per, ? ? ? ? ? ? ?/* Peripheral to Peripheral */
> > + ? ? ? int_2_per, ? ? ? ? ? ? ?/* Internal RAM to peripheral */
> > + ? ? ? int_2_int, ? ? ? ? ? ? ?/* Internal RAM to Internal RAM */
> > + ? ? ? int_2_emi, ? ? ? ? ? ? ?/* Internal RAM to EMI memory */
> > + ? ? ? int_2_dsp, ? ? ? ? ? ? ?/* Internal RAM to DSP memory */
> > + ? ? ? dsp_2_per, ? ? ? ? ? ? ?/* DSP memory to peripheral */
> > + ? ? ? dsp_2_int, ? ? ? ? ? ? ?/* DSP memory to internal RAM */
> > + ? ? ? dsp_2_emi, ? ? ? ? ? ? ?/* DSP memory to EMI memory */
> > + ? ? ? dsp_2_dsp, ? ? ? ? ? ? ?/* DSP memory to DSP memory */
> > + ? ? ? emi_2_dsp_loop, ? ? ? ? /* EMI memory to DSP memory loopback */
> > + ? ? ? dsp_2_emi_loop, ? ? ? ? /* DSP memory to EMI memory loopback */
> > + ? ? ? dvfs_pll, ? ? ? ? ? ? ? /* DVFS script with PLL change ? ? ? */
> > + ? ? ? dvfs_pdr ? ? ? ? ? ? ? ?/* DVFS script without PLL change ? ?*/
> > +} sdma_transfer_type;
> > +
> > +/*
> > + * Structure containing sdma request ?parameters.
> > + */
> > +struct sdma_script_start_addrs {
> > + ? ? ? int ap_2_ap_addr;
> > + ? ? ? int ap_2_bp_addr;
> > + ? ? ? int ap_2_ap_fixed_addr;
> > + ? ? ? int bp_2_ap_addr;
> > + ? ? ? int loopback_on_dsp_side_addr;
> > + ? ? ? int mcu_interrupt_only_addr;
> > +
> > + ? ? ? int firi_2_per_addr;
> > + ? ? ? int firi_2_mcu_addr;
> > + ? ? ? int per_2_firi_addr;
> > + ? ? ? int mcu_2_firi_addr;
> > +
> > + ? ? ? int uart_2_per_addr;
> > + ? ? ? int uart_2_mcu_addr;
> > + ? ? ? int per_2_app_addr;
> > + ? ? ? int mcu_2_app_addr;
> > + ? ? ? int per_2_per_addr;
> > +
> > + ? ? ? int uartsh_2_per_addr;
> > + ? ? ? int uartsh_2_mcu_addr;
> > + ? ? ? int per_2_shp_addr;
> > + ? ? ? int mcu_2_shp_addr;
> > +
> > + ? ? ? int ata_2_mcu_addr;
> > + ? ? ? int mcu_2_ata_addr;
> > +
> > + ? ? ? int app_2_per_addr;
> > + ? ? ? int app_2_mcu_addr;
> > + ? ? ? int shp_2_per_addr;
> > + ? ? ? int shp_2_mcu_addr;
> > +
> > + ? ? ? int mshc_2_mcu_addr;
> > + ? ? ? int mcu_2_mshc_addr;
> > +
> > + ? ? ? int spdif_2_mcu_addr;
> > + ? ? ? int mcu_2_spdif_addr;
> > +
> > + ? ? ? int asrc_2_mcu_addr;
> > +
> > + ? ? ? int ext_mem_2_ipu_addr;
> > +
> > + ? ? ? int descrambler_addr;
> > +
> > + ? ? ? int dptc_dvfs_addr;
> > +
> > + ? ? ? int utra_addr;
> > +
> > + ? ? ? int ram_code_start_addr;
>
> All these addresses, are they really integers with
> valid negative values... Aren't they dma_addr_t or
> atleast u32?
Since this struct must match the layout of the firmware, they should be
u32, yes. They are no dma_addr_t since it's the sdma controller address
space described here.
>
> > +};
> > +
> > +#define SDMA_FIRMWARE_MAGIC 0x414d4453
> > +
> > +struct sdma_firmware_header {
> > + ? ? ? uint32_t ? ? ? ?magic; /* "SDMA" */
> > + ? ? ? uint32_t ? ? ? ?version_major; ?/* increased whenever layout of struct sdma_script_start_addrs changes */
> > + ? ? ? uint32_t ? ? ? ?version_minor; ?/* firmware version */
> > + ? ? ? uint32_t ? ? ? ?script_addrs_start; /* offset of struct sdma_script_start_addrs in this image */
> > + ? ? ? uint32_t ? ? ? ?num_script_addrs; /* Number of script addresses in this image */
> > + ? ? ? uint32_t ? ? ? ?ram_code_start; /* offset of SDMA ram image in this firmware image */
> > + ? ? ? uint32_t ? ? ? ?ram_code_size; /* size of SDMA ram image */
>
> Please use u32. uint32_t is not the preferred kernel type.
> (Still I've seen people use it in some cases so I might be wrong,
> feel welcome to bit back on this.)
At least one type should be consequently used in a driver. Changed them
all to u32.
>
> > +};
> > +
> > +static struct sdma_channel sdma_data[MAX_DMA_CHANNELS];
> > +static struct sdma_channel_control *channel_control;
> > +static void __iomem *sdma_base;
> > +static int sdma_version;
>
> Unsigned?
>
> > +static int sdma_num_events;
>
> Unsigned?
>
> > +static struct sdma_context_data *sdma_context;
> > +dma_addr_t sdma_context_phys;
> > +static struct dma_device __sdma_dma_device;
> > +static struct dma_device *sdma_dma_device = &__sdma_dma_device;
>
> This is what I suspected: local variables making the entire driver
> a singleton, which means you can never have more than one
> SDMA. Atleast collect all of these in a struct, call it
> "struct sdma" simply (if you ask me) and use as a stateholder.
> This makes it easier to kzalloc() that struct later if you
> want to support non-singletons.
>
> I know this require some work but I've done it to several drivers
> (always asked on mailinglists to do this) and I don't regret a single
> rewrite. Last time was for the PL18x DMAengine driver actually.
I've done it myself often enough. Ok, will change.
>
> > +
> > +#define SDMA_H_CONFIG_DSPDMA ? (1 << 12) /* indicates if the DSPDMA is used */
> > +#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
> > +#define SDMA_H_CONFIG_ACR ? ? ?(1 << 4) ?/* indicates if AHB freq /core freq = 2 or 1 */
> > +#define SDMA_H_CONFIG_CSM ? ? ?(3) ? ? ? /* indicates which context switch mode is selected*/
> > +
> > +static int sdma_config_ownership(int channel, int event_override,
> > + ? ? ? ? ? ? ? ? ?int mcu_verride, int dsp_override)
> > +{
> > + ? ? ? u32 evt, mcu, dsp;
> > +
> > + ? ? ? if (event_override && mcu_verride && dsp_override)
> > + ? ? ? ? ? ? ? return -EINVAL;
> > +
> > + ? ? ? evt = readl(SDMA_H_EVTOVR);
> > + ? ? ? mcu = readl(SDMA_H_HOSTOVR);
> > + ? ? ? dsp = readl(SDMA_H_DSPOVR);
> > +
> > + ? ? ? if (dsp_override)
> > + ? ? ? ? ? ? ? dsp &= ~(1 << channel);
> > + ? ? ? else
> > + ? ? ? ? ? ? ? dsp |= (1 << channel);
> > +
> > + ? ? ? if (event_override)
> > + ? ? ? ? ? ? ? evt &= ~(1 << channel);
> > + ? ? ? else
> > + ? ? ? ? ? ? ? evt |= (1 << channel);
> > +
> > + ? ? ? if (mcu_verride)
> > + ? ? ? ? ? ? ? mcu &= ~(1 << channel);
> > + ? ? ? else
> > + ? ? ? ? ? ? ? mcu |= (1 << channel);
> > +
> > + ? ? ? writel(evt, SDMA_H_EVTOVR);
> > + ? ? ? writel(mcu, SDMA_H_HOSTOVR);
> > + ? ? ? writel(dsp, SDMA_H_DSPOVR);
> > +
> > + ? ? ? return 0;
> > +}
> > +
> > +/*
> > + * sdma_run_channel - run a channel and wait till it's done
> > + */
> > +static int sdma_run_channel(int channel)
> > +{
> > + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> > + ? ? ? int ret;
> > +
> > + ? ? ? writel(1 << channel, SDMA_H_START);
> > +
> > + ? ? ? ret = wait_event_interruptible(sdma->waitq,
> > + ? ? ? ? ? ? ? ? ? ? ? !(readl(SDMA_H_STATSTOP) & (1 << channel)));
>
> OK not the biggest thing in the world, but can't you use a
> completion for this? (I'm not so clever with waitqueues so
> forgive me if this is malinformed.)
Ok.
>
> > + ? ? ? return ret;
> > +}
> > +
> > +static int sdma_load_script(void *buf, int size, u32 address)
> > +{
> > + ? ? ? struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> > + ? ? ? void *buf_virt;
> > + ? ? ? dma_addr_t buf_phys;
> > + ? ? ? int ret;
> > +
> > + ? ? ? buf_virt = dma_alloc_coherent(NULL,
> > + ? ? ? ? ? ? ? ? ? ? ? size,
> > + ? ? ? ? ? ? ? ? ? ? ? &buf_phys, GFP_KERNEL);
> > + ? ? ? if (!buf_virt)
> > + ? ? ? ? ? ? ? return -ENOMEM;
> > +
> > + ? ? ? bd0->mode.command = C0_SETPM;
> > + ? ? ? bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> > + ? ? ? bd0->mode.count = size / 2;
> > + ? ? ? bd0->buffer_addr = buf_phys;
> > + ? ? ? bd0->ext_buffer_addr = address;
> > +
> > + ? ? ? memcpy(buf_virt, buf, size);
> > +
> > + ? ? ? ret = sdma_run_channel(0);
> > +
> > + ? ? ? dma_free_coherent(NULL, size, buf_virt, buf_phys);
> > +
> > + ? ? ? return ret;
> > +}
> > +
> > +static void sdma_event_enable(int channel, int event)
> > +{
> > + ? ? ? u32 val;
> > +
> > + ? ? ? val = readl(SDMA_CHNENBL_0 + event * 4);
>
> This use indicates that event should probably be
> unsigned, and probably not greater than u16 atleast.
> I suspect it is never more than an u8 really.
>
> > + ? ? ? val |= (1 << channel);
> > + ? ? ? writel(val, SDMA_CHNENBL_0 + event * 4);
> > +}
> > +
> > +static void sdma_event_disable(int channel, int event)
> > +{
> > + ? ? ? u32 val;
> > +
> > + ? ? ? val = readl(SDMA_CHNENBL_0 + event * 4);
> > + ? ? ? val &= ~(1 << channel);
> > + ? ? ? writel(val, SDMA_CHNENBL_0 + event * 4);
>
> Same comment here.
Ok, changed to unsigned and added a check for valid values in
sdma_config_channel.
>
> > +}
> > +
> > +static void mxc_sdma_handle_channel_loop(int channel)
> > +{
> > + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
>
> This indicates that channel should be unsigned.
>
> > + ? ? ? struct sdma_buffer_descriptor *bd;
> > + ? ? ? int error = 0;
>
> Unused variable?
Originally the callback function had a status parameter where errors
were signalled. I assume device_tx_status is the function to which this
error should be passed, right?
>
> > +
> > + ? ? ? /*
> > + ? ? ? ?* loop mode. Iterate over descriptors, re-setup them and
> > + ? ? ? ?* call callback function.
> > + ? ? ? ?*/
> > + ? ? ? while (1) {
> > + ? ? ? ? ? ? ? bd = &sdma->bd[sdma->buf_tail];
> > +
> > + ? ? ? ? ? ? ? if (bd->mode.status & BD_DONE)
> > + ? ? ? ? ? ? ? ? ? ? ? break;
> > +
> > + ? ? ? ? ? ? ? if (bd->mode.status & BD_RROR)
> > + ? ? ? ? ? ? ? ? ? ? ? error = -EIO;
> > +
> > + ? ? ? ? ? ? ? bd->mode.status |= BD_DONE;
> > + ? ? ? ? ? ? ? sdma->buf_tail++;
> > + ? ? ? ? ? ? ? sdma->buf_tail %= sdma->num_bd;
> > +
> > + ? ? ? ? ? ? ? if (sdma->desc.callback)
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->desc.callback(sdma->desc.callback_param);
> > + ? ? ? }
> > +}
> > +
> > +static void mxc_sdma_handle_channel_normal(int channel)
> > +{
> > + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> > + ? ? ? struct sdma_buffer_descriptor *bd;
> > + ? ? ? int i, error = 0;
> > +
> > + ? ? ? /*
> > + ? ? ? ?* non loop mode. Iterate over all descriptors, collect
> > + ? ? ? ?* errors and call callback function
> > + ? ? ? ?*/
> > + ? ? ? for (i = 0; i < sdma->num_bd; i++) {
> > + ? ? ? ? ? ? ? bd = &sdma->bd[i];
> > +
> > + ? ? ? ? ? ? ? ?if (bd->mode.status & (BD_DONE | BD_RROR))
> > + ? ? ? ? ? ? ? ? ? ? ? error = -EIO;
> > + ? ? ? }
> > +
> > + ? ? ? if (sdma->desc.callback)
> > + ? ? ? ? ? ? ? sdma->desc.callback(sdma->desc.callback_param);
> > + ? ? ? sdma->last_completed = sdma->desc.cookie;
> > +
> > + ? ? ? sdma->busy = 0;
>
> = true if you switch this to bool..
>
> > +}
> > +
> > +static void mxc_sdma_handle_channel(int channel)
> > +{
> > + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> > +
> > + ? ? ? wake_up_interruptible(&sdma->waitq);
> > +
> > + ? ? ? /* not interested in channel 0 interrupts */
> > + ? ? ? if (!channel)
> > + ? ? ? ? ? ? ? return;
> > +
> > + ? ? ? if (sdma->flags & IMX_DMA_SG_LOOP)
> > + ? ? ? ? ? ? ? mxc_sdma_handle_channel_loop(channel);
> > + ? ? ? else
> > + ? ? ? ? ? ? ? mxc_sdma_handle_channel_normal(channel);
> > +}
> > +
> > +static irqreturn_t sdma_int_handler(int irq, void *dev_id)
> > +{
> > + ? ? ? u32 stat;
> > +
> > + ? ? ? stat = readl(SDMA_H_INTR);
> > + ? ? ? writel(stat, SDMA_H_INTR);
> > +
> > + ? ? ? while (stat) {
> > + ? ? ? ? ? ? ? int channel = fls(stat) - 1;
> > +
> > + ? ? ? ? ? ? ? mxc_sdma_handle_channel(channel);
> > +
> > + ? ? ? ? ? ? ? stat &= ~(1 << channel);
> > + ? ? ? }
> > +
> > + ? ? ? return IRQ_HANDLED;
> > +}
> > +
> > +static struct clk *sdma_clk;
> > +
> > +/*
> > + * Stores the start address of the SDMA scripts
> > + */
> > +static struct sdma_script_start_addrs __sdma_script_addrs;
> > +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
> > +
> > +/*
> > + * sets the pc of SDMA script according to the peripheral type
> > + */
> > +static void sdma_get_pc(struct sdma_channel *sdma,
> > + ? ? ? ? ? ? ? sdma_peripheral_type peripheral_type)
> > +{
> > + ? ? ? int res = 0;
> > + ? ? ? int per_2_emi = 0, emi_2_per = 0;
> > + ? ? ? int per_2_int = 0, int_2_per = 0;
> > + ? ? ? int per_2_per = 0, emi_2_emi = 0;
> > +
> > + ? ? ? sdma->pc_from_device = 0;
> > + ? ? ? sdma->pc_to_device = 0;
>
> There are a *lot* of local variables here, and only two of them
> are used eventually, at the end of the function. I cannot quite
> follow this, what is going on?
'per' is for peripheral (like sdhc, ssi unit or similar)
'int' is for internal SRAM
'emi' is for SDRAM
currently we only support transfers from per to emi and emi to per and
so the other variables are unused. We could make this function simpler
by removing the unused variables, but I suggest to keep them for making
it easier to support other transfer types later. I can add a comment
what these variables are for and why they are unused.
>
> Some like emi_2_emi seem to be totally unused.
>
> The types here look like some kind of enum or other
> similar construction is really what's being asked for
> here.
>
> > +
> > + ? ? ? switch (peripheral_type) {
> > + ? ? ? case IMX_DMATYPE_MEMORY:
> > + ? ? ? ? ? ? ? emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_DSP:
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->bp_2_ap_addr;
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->ap_2_bp_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_FIRI:
> > + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->firi_2_per_addr;
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
> > + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_firi_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_UART:
> > + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->uart_2_per_addr;
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
> > + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_app_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_UART_SP:
> > + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->uartsh_2_per_addr;
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
> > + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_shp_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_ATA:
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_CSPI:
> > + ? ? ? case IMX_DMATYPE_EXT:
> > + ? ? ? case IMX_DMATYPE_SSI:
> > + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->app_2_per_addr;
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->app_2_mcu_addr;
> > + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_app_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_app_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_SSI_SP:
> > + ? ? ? case IMX_DMATYPE_MMC:
> > + ? ? ? case IMX_DMATYPE_SDHC:
> > + ? ? ? case IMX_DMATYPE_CSPI_SP:
> > + ? ? ? case IMX_DMATYPE_ESAI:
> > + ? ? ? case IMX_DMATYPE_MSHC_SP:
> > + ? ? ? ? ? ? ? per_2_int = sdma_script_addrs->shp_2_per_addr;
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
> > + ? ? ? ? ? ? ? int_2_per = sdma_script_addrs->per_2_shp_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_ASRC:
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
> > + ? ? ? ? ? ? ? per_2_per = sdma_script_addrs->per_2_per_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_MSHC:
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_CCM:
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_FIFO_MEMORY:
> > + ? ? ? ? ? ? ? res = sdma_script_addrs->ap_2_ap_fixed_addr;
>
> res? This thing is never used.
I have no idea what DMATYPE_FIFO_MEMORY is. Will remove this.
>
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_SPDIF:
> > + ? ? ? ? ? ? ? per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_IPU_MEMORY:
> > + ? ? ? ? ? ? ? emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? default:
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? }
> > +
> > + ? ? ? sdma->pc_from_device = per_2_emi;
> > + ? ? ? sdma->pc_to_device = emi_2_per;
>
> Return res? You're assigning it a value in some cases.
>
> > +}
> > +
> > +static int sdma_load_context(int channel)
> > +{
> > + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> > + ? ? ? int load_address;
> > + ? ? ? struct sdma_buffer_descriptor *bd0 = sdma_data[0].bd;
> > + ? ? ? int ret;
> > +
> > + ? ? ? if (sdma->direction == DMA_FROM_DEVICE) {
> > + ? ? ? ? ? ? ? load_address = sdma->pc_from_device;
> > + ? ? ? } else {
> > + ? ? ? ? ? ? ? load_address = sdma->pc_to_device;
> > + ? ? ? }
> > +
> > + ? ? ? if (load_address < 0)
> > + ? ? ? ? ? ? ? return load_address;
> > +
> > + ? ? ? pr_debug("%s: load_address = %d\n", __func__, load_address);
> > + ? ? ? pr_debug("%s: wml = 0x%08x\n", __func__, sdma->watermark_level);
> > + ? ? ? pr_debug("%s: shp_addr = 0x%08x\n", __func__, sdma->shp_addr);
> > + ? ? ? pr_debug("%s: per_addr = 0x%08x\n", __func__, sdma->per_addr);
> > + ? ? ? pr_debug("%s: event_mask1 = 0x%08x\n", __func__, sdma->event_mask1);
> > + ? ? ? pr_debug("%s: event_mask2 = 0x%08x\n", __func__, sdma->event_mask2);
>
> Surely it must be possible to get the struct device * pointer for the
> channels host and use dev_dbg() instead?
Ok, will do
>
> > +
> > + ? ? ? memset(sdma_context, 0, sizeof(*sdma_context));
> > + ? ? ? sdma_context->channel_state.pc = load_address;
> > +
> > + ? ? ? /* Send by context the event mask,base address for peripheral
> > + ? ? ? ?* and watermark level
> > + ? ? ? ?*/
> > + ? ? ? sdma_context->gReg[0] = sdma->event_mask2;
> > + ? ? ? sdma_context->gReg[1] = sdma->event_mask1;
> > + ? ? ? sdma_context->gReg[2] = sdma->per_addr;
> > + ? ? ? sdma_context->gReg[6] = sdma->shp_addr;
> > + ? ? ? sdma_context->gReg[7] = sdma->watermark_level;
> > +
> > + ? ? ? bd0->mode.command = C0_SETDM;
> > + ? ? ? bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
> > + ? ? ? bd0->mode.count = sizeof(*sdma_context) / 4;
> > + ? ? ? bd0->buffer_addr = sdma_context_phys;
> > + ? ? ? bd0->ext_buffer_addr = 2048 + (sizeof(*sdma_context) / 4) * channel;
> > +
> > + ? ? ? ret = sdma_run_channel(0);
> > +
> > + ? ? ? return ret;
> > +}
> > +
> > +static void sdma_disable_channel(int channel)
> > +{
> > + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> > +
> > + ? ? ? writel(1 << channel, SDMA_H_STATSTOP);
> > + ? ? ? sdma->busy = 0;
> > +}
> > +
> > +static int sdma_config_channel(int channel)
> > +{
> > + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> > + ? ? ? int ret;
> > +
> > + ? ? ? sdma_disable_channel(channel);
> > +
> > + ? ? ? sdma->event_mask1 = 0;
> > + ? ? ? sdma->event_mask2 = 0;
> > + ? ? ? sdma->shp_addr = 0;
> > + ? ? ? sdma->per_addr = 0;
> > +
> > + ? ? ? if (sdma->event_id)
> > + ? ? ? ? ? ? ? sdma_event_enable(channel, sdma->event_id);
> > +
> > + ? ? ? switch (sdma->peripheral_type) {
> > + ? ? ? case IMX_DMATYPE_DSP:
> > + ? ? ? ? ? ? ? sdma_config_ownership(channel, 0, 1, 1);
>
> The parameters here makes yoy believe that the types should
> be bool rather than int...
ok.
>
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case IMX_DMATYPE_MEMORY:
> > + ? ? ? ? ? ? ? sdma_config_ownership(channel, 0, 1, 0);
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? default:
> > + ? ? ? ? ? ? ? sdma_config_ownership(channel, 1, 1, 0);
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? }
> > +
> > + ? ? ? sdma_get_pc(sdma, sdma->peripheral_type);
> > +
> > + ? ? ? if ((sdma->peripheral_type != IMX_DMATYPE_MEMORY) &&
> > + ? ? ? ? ? ? ? ? ? ? ? (sdma->peripheral_type != IMX_DMATYPE_DSP)) {
> > + ? ? ? ? ? ? ? /* Handle multiple event channels differently */
> > + ? ? ? ? ? ? ? if (sdma->event_id2) {
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->event_mask2 = 1 << (sdma->event_id2 % 32);
> > + ? ? ? ? ? ? ? ? ? ? ? if (sdma->event_id2 > 31)
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? sdma->watermark_level |= 1 << 31;
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->event_mask1 = 1 << (sdma->event_id % 32);
> > + ? ? ? ? ? ? ? ? ? ? ? if (sdma->event_id > 31)
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? sdma->watermark_level |= 1 << 30;
> > + ? ? ? ? ? ? ? } else {
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->event_mask1 = 1 << sdma->event_id;
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->event_mask2 = 1 << (sdma->event_id - 32);
> > + ? ? ? ? ? ? ? }
> > + ? ? ? ? ? ? ? /* Watermark Level */
> > + ? ? ? ? ? ? ? sdma->watermark_level |= sdma->watermark_level;
> > + ? ? ? ? ? ? ? /* Address */
> > + ? ? ? ? ? ? ? sdma->shp_addr = sdma->per_address;
> > + ? ? ? } else {
> > + ? ? ? ? ? ? ? sdma->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
> > + ? ? ? }
> > +
> > + ? ? ? ret = sdma_load_context(channel);
> > +
> > + ? ? ? return ret;
> > +}
> > +
> > +static int sdma_set_channel_priority(unsigned int channel, unsigned int priority)
> > +{
> > + ? ? ? if (priority < MXC_SDMA_MIN_PRIORITY
> > + ? ? ? ? ? || priority > MXC_SDMA_MAX_PRIORITY) {
> > + ? ? ? ? ? ? ? return -EINVAL;
> > + ? ? ? }
> > +
> > + ? ? ? writel(priority, SDMA_CHNPRI_0 + 4 * channel);
> > +
> > + ? ? ? return 0;
> > +}
> > +
> > +static int sdma_request_channel(int channel)
> > +{
> > + ? ? ? struct sdma_channel *sdma = &sdma_data[channel];
> > + ? ? ? int ret = -EBUSY;
> > +
> > + ? ? ? sdma->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdma->bd_phys, GFP_KERNEL);
> > + ? ? ? if (!sdma->bd) {
> > + ? ? ? ? ? ? ? ret = -ENOMEM;
> > + ? ? ? ? ? ? ? goto out;
> > + ? ? ? }
> > +
> > + ? ? ? memset(sdma->bd, 0, PAGE_SIZE);
> > +
> > + ? ? ? channel_control[channel].base_bd_ptr = sdma->bd_phys;
> > + ? ? ? channel_control[channel].current_bd_ptr = sdma->bd_phys;
> > +
> > + ? ? ? clk_enable(sdma_clk);
>
> Aha you're enabling it once for every channel and rely on
> clk reference counting that's clever!
>
> > +
> > + ? ? ? sdma_set_channel_priority(channel, MXC_SDMA_DEFAULT_PRIORITY);
> > +
> > + ? ? ? init_waitqueue_head(&sdma->waitq);
> > +
> > + ? ? ? sdma->buf_tail = 0;
> > +
> > + ? ? ? return 0;
> > +out:
> > +
> > + ? ? ? return ret;
> > +}
> > +
> > +static void sdma_enable_channel(int channel)
> > +{
> > + ? ? ? writel(1 << channel, SDMA_H_START);
> > +}
> > +
> > +static int __init sdma_init(unsigned long phys_base, int irq, int version,
> > + ? ? ? ? ? ? ? void *ram_code,
> > + ? ? ? ? ? ? ? int ram_code_size)
> > +{
> > + ? ? ? int i, ret;
> > + ? ? ? int channel;
> > + ? ? ? dma_addr_t ccb_phys;
> > +
> > + ? ? ? sdma_version = version;
> > + ? ? ? switch (sdma_version) {
> > + ? ? ? case 1:
> > + ? ? ? ? ? ? ? sdma_num_events = 32;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case 2:
> > + ? ? ? ? ? ? ? sdma_num_events = 48;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? default:
> > + ? ? ? ? ? ? ? pr_err("SDMA: Unknown version %d. aborting\n", sdma_version);
> > + ? ? ? ? ? ? ? return -ENODEV;
> > + ? ? ? }
> > +
> > + ? ? ? clk_enable(sdma_clk);
> > +
> > + ? ? ? sdma_base = ioremap(phys_base, 4096);
>
> Use SZ_4K instead of 4096.
Or even better resouce_size(iores)
>
> > + ? ? ? if (!sdma_base) {
> > + ? ? ? ? ? ? ? ret = -ENOMEM;
> > + ? ? ? ? ? ? ? goto err_ioremap;
> > + ? ? ? }
> > +
> > + ? ? ? /* Initialize SDMA private data */
> > + ? ? ? memset(sdma_data, 0, sizeof(struct sdma_channel) * MAX_DMA_CHANNELS);
> > +
> > + ? ? ? for (channel = 0; channel < MAX_DMA_CHANNELS; channel++)
> > + ? ? ? ? ? ? ? sdma_data[channel].channel = channel;
> > +
> > + ? ? ? ret = request_irq(irq, sdma_int_handler, 0, "sdma", NULL);
> > + ? ? ? if (ret)
> > + ? ? ? ? ? ? ? goto err_request_irq;
> > +
> > + ? ? ? /* Be sure SDMA has not started yet */
> > + ? ? ? writel(0, SDMA_H_C0PTR);
> > +
> > + ? ? ? channel_control = dma_alloc_coherent(NULL,
> > + ? ? ? ? ? ? ? ? ? ? ? MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
> > + ? ? ? ? ? ? ? ? ? ? ? sizeof(struct sdma_context_data),
> > + ? ? ? ? ? ? ? ? ? ? ? &ccb_phys, GFP_KERNEL);
> > +
> > + ? ? ? if (!channel_control) {
> > + ? ? ? ? ? ? ? ret = -ENOMEM;
> > + ? ? ? ? ? ? ? goto err_dma_alloc;
> > + ? ? ? }
> > +
> > + ? ? ? sdma_context = (void *)channel_control +
> > + ? ? ? ? ? ? ? MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> > + ? ? ? sdma_context_phys = ccb_phys +
> > + ? ? ? ? ? ? ? MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
> > +
> > + ? ? ? /* Zero-out the CCB structures array just allocated */
> > + ? ? ? memset(channel_control, 0,
> > + ? ? ? ? ? ? ? ? ? ? ? MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
> > +
> > + ? ? ? /* disable all channels */
> > + ? ? ? for (i = 0; i < sdma_num_events; i++)
> > + ? ? ? ? ? ? ? writel(0, SDMA_CHNENBL_0 + i * 4);
> > +
> > + ? ? ? /* All channels have priority 0 */
> > + ? ? ? for (i = 0; i < MAX_DMA_CHANNELS; i++)
> > + ? ? ? ? ? ? ? writel(0, SDMA_CHNPRI_0 + i * 4);
> > +
> > + ? ? ? ret = sdma_request_channel(0);
> > + ? ? ? if (ret)
> > + ? ? ? ? ? ? ? goto err_dma_alloc;
> > +
> > + ? ? ? sdma_config_ownership(0, 0, 1, 0);
> > +
> > + ? ? ? /* Set Command Channel (Channel Zero) */
> > + ? ? ? writel(0x4050, SDMA_CHN0ADDR);
> > +
> > + ? ? ? /* Set bits of CONFIG register but with static context switching */
> > + ? ? ? /* FIXME: Check whether to set ACR bit depending on clock ratios */
> > + ? ? ? writel(0, SDMA_H_CONFIG);
> > +
> > + ? ? ? writel(ccb_phys, SDMA_H_C0PTR);
> > +
> > + ? ? ? /* download the RAM image for SDMA */
> > + ? ? ? sdma_load_script(ram_code,
> > + ? ? ? ? ? ? ? ? ? ? ? ram_code_size,
> > + ? ? ? ? ? ? ? ? ? ? ? sdma_script_addrs->ram_code_start_addr);
> > +
> > + ? ? ? /* Set bits of CONFIG register with given context switching mode */
> > + ? ? ? writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
> > +
> > + ? ? ? /* Initializes channel's priorities */
> > + ? ? ? sdma_set_channel_priority(0, 7);
> > +
> > + ? ? ? clk_disable(sdma_clk);
> > +
> > + ? ? ? return 0;
> > +
> > +err_dma_alloc:
> > + ? ? ? free_irq(irq, NULL);
> > +err_request_irq:
> > + ? ? ? iounmap(sdma_base);
> > +err_ioremap:
> > + ? ? ? clk_disable(sdma_clk);
> > + ? ? ? pr_err("%s failed with %d\n", __func__, ret);
> > + ? ? ? return ret;
> > +}
> > +
> > +static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
> > +{
> > + ? ? ? dma_cookie_t cookie = sdma->chan.cookie;
> > +
> > + ? ? ? if (++cookie < 0)
> > + ? ? ? ? ? ? ? cookie = 1;
> > +
> > + ? ? ? sdma->chan.cookie = cookie;
> > + ? ? ? sdma->desc.cookie = cookie;
> > +
> > + ? ? ? return cookie;
> > +}
> > +
> > +static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
> > +{
> > + ? ? ? return container_of(chan, struct sdma_channel, chan);
> > +}
> > +
> > +static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
> > +{
> > + ? ? ? struct sdma_channel *sdma = to_sdma_chan(tx->chan);
> > + ? ? ? dma_cookie_t cookie;
> > +
> > + ? ? ? spin_lock_irq(&sdma->lock);
> > +
> > + ? ? ? cookie = sdma_assign_cookie(sdma);
> > +
> > + ? ? ? sdma_enable_channel(tx->chan->chan_id);
> > +
> > + ? ? ? spin_unlock_irq(&sdma->lock);
> > +
> > + ? ? ? return cookie;
> > +}
> > +
> > +static int sdma_alloc_chan_resources(struct dma_chan *chan)
> > +{
> > + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> > + ? ? ? struct imx_dma_data *data = chan->private;
> > + ? ? ? int prio, ret;
> > +
> > + ? ? ? /* No need to execute this for internal channel 0 */
> > + ? ? ? if (!chan->chan_id)
> > + ? ? ? ? ? ? ? return 0;
> > +
> > + ? ? ? if (!data)
> > + ? ? ? ? ? ? ? return -EINVAL;
> > +
> > + ? ? ? switch (data->priority) {
> > + ? ? ? case DMA_PRIO_HIGH:
> > + ? ? ? ? ? ? ? prio = 3;
>
> Wait, aren't these enumerated?
> Add some enum sdma_channel_prio {}..
Hm, The SDMA engine has priorities from 1 to 7 from which we happen to use
the lowest priorities only. I think this should not be an enum.
(The DMA_PRIO_* are only used in an attempt to provide the same API for
the i.MX2 SoCs which have a different DMA engine which is not so
flexible)
>
>
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case DMA_PRIO_MEDIUM:
> > + ? ? ? ? ? ? ? prio = 2;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? case DMA_PRIO_LOW:
> > + ? ? ? default:
> > + ? ? ? ? ? ? ? prio = 1;
> > + ? ? ? ? ? ? ? break;
> > + ? ? ? }
> > +
> > + ? ? ? sdma->peripheral_type = data->peripheral_type;
> > + ? ? ? sdma->event_id = data->dma_request;
> > + ? ? ? ret = sdma_set_channel_priority(chan->chan_id, prio);
> > + ? ? ? if (ret)
> > + ? ? ? ? ? ? ? return ret;
> > +
> > + ? ? ? if (chan->chan_id) {
> > + ? ? ? ? ? ? ? ret = sdma_request_channel(chan->chan_id);
> > + ? ? ? ? ? ? ? if (ret)
> > + ? ? ? ? ? ? ? ? ? ? ? return ret;
> > + ? ? ? }
> > +
> > + ? ? ? dma_async_tx_descriptor_init(&sdma->desc, chan);
> > + ? ? ? sdma->desc.tx_submit = sdma_tx_submit;
> > + ? ? ? /* txd.flags will be overwritten in prep funcs */
> > + ? ? ? sdma->desc.flags = DMA_CTRL_ACK;
> > +
> > + ? ? ? return 0;
> > +}
> > +
> > +static void sdma_free_chan_resources(struct dma_chan *chan)
> > +{
> > + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> > + ? ? ? int channel = chan->chan_id;
> > +
> > + ? ? ? sdma_disable_channel(channel);
> > +
> > + ? ? ? if (sdma->event_id)
> > + ? ? ? ? ? ? ? sdma_event_disable(channel, sdma->event_id);
> > + ? ? ? if (sdma->event_id2)
> > + ? ? ? ? ? ? ? sdma_event_disable(channel, sdma->event_id2);
> > +
> > + ? ? ? sdma->event_id = 0;
> > + ? ? ? sdma->event_id2 = 0;
> > +
> > + ? ? ? sdma_set_channel_priority(channel, 0);
> > +
> > + ? ? ? dma_free_coherent(NULL, PAGE_SIZE, sdma->bd, sdma->bd_phys);
> > +
> > + ? ? ? clk_disable(sdma_clk);
> > +}
> > +
> > +#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
> > +
> > +static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
> > + ? ? ? ? ? ? ? struct dma_chan *chan, struct scatterlist *sgl,
> > + ? ? ? ? ? ? ? unsigned int sg_len, enum dma_data_direction direction,
> > + ? ? ? ? ? ? ? unsigned long flags)
> > +{
> > + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> > + ? ? ? int ret, i, count;
> > + ? ? ? int channel = chan->chan_id;
> > + ? ? ? struct scatterlist *sg;
> > +
> > + ? ? ? if (sdma->busy)
> > + ? ? ? ? ? ? ? return NULL;
> > + ? ? ? sdma->busy = 1;
> > +
> > + ? ? ? sdma->flags = 0;
>
> What are those flags anyway? I think you will need some
> #define:s for them.
There's only one currently: IMX_DMA_SG_LOOP indicating that we are
doing cyclic transfers.
>
> > +
> > + ? ? ? pr_debug("SDMA: setting up %d entries for channel %d.\n",
> > + ? ? ? ? ? ? ? ? ? ? ? sg_len, channel);
> > +
> > + ? ? ? sdma->direction = direction;
> > + ? ? ? ret = sdma_load_context(channel);
> > + ? ? ? if (ret)
> > + ? ? ? ? ? ? ? goto err_out;
> > +
> > + ? ? ? if (sg_len > NUM_BD) {
> > + ? ? ? ? ? ? ? pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? channel, sg_len, NUM_BD);
> > + ? ? ? ? ? ? ? ret = -EINVAL;
> > + ? ? ? ? ? ? ? goto err_out;
> > + ? ? ? }
> > +
> > + ? ? ? for_each_sg(sgl, sg, sg_len, i) {
> > + ? ? ? ? ? ? ? struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> > + ? ? ? ? ? ? ? int param;
> > +
> > + ? ? ? ? ? ? ? bd->buffer_addr = sgl->dma_address;
> > +
> > + ? ? ? ? ? ? ? count = sg->length;
> > +
> > + ? ? ? ? ? ? ? if (count > 0xffff) {
> > + ? ? ? ? ? ? ? ? ? ? ? pr_err("SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? channel, count, 0xffff);
> > + ? ? ? ? ? ? ? ? ? ? ? ret = -EINVAL;
> > + ? ? ? ? ? ? ? ? ? ? ? goto err_out;
> > + ? ? ? ? ? ? ? }
> > +
> > + ? ? ? ? ? ? ? bd->mode.count = count;
> > +
> > + ? ? ? ? ? ? ? if (sdma->word_size > 4) {
> > + ? ? ? ? ? ? ? ? ? ? ? ret = ?-EINVAL;
> > + ? ? ? ? ? ? ? ? ? ? ? goto err_out;
> > + ? ? ? ? ? ? ? }
> > + ? ? ? ? ? ? ? if (sdma->word_size == 4)
> > + ? ? ? ? ? ? ? ? ? ? ? bd->mode.command = 0;
> > + ? ? ? ? ? ? ? else
> > + ? ? ? ? ? ? ? ? ? ? ? bd->mode.command = sdma->word_size;
> > +
> > + ? ? ? ? ? ? ? param = BD_DONE | BD_EXTD | BD_CONT;
> > +
> > + ? ? ? ? ? ? ? if (sdma->flags & IMX_DMA_SG_LOOP) {
> > + ? ? ? ? ? ? ? ? ? ? ? param |= BD_INTR;
> > + ? ? ? ? ? ? ? ? ? ? ? if (i + 1 == sg_len)
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param |= BD_WRAP;
> > + ? ? ? ? ? ? ? }
> > +
> > + ? ? ? ? ? ? ? if (i + 1 == sg_len)
> > + ? ? ? ? ? ? ? ? ? ? ? param |= BD_INTR;
> > +
> > + ? ? ? ? ? ? ? pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? i, count, sg->dma_address,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param & BD_WRAP ? "wrap" : "",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param & BD_INTR ? " intr" : "");
> > +
> > + ? ? ? ? ? ? ? bd->mode.status = param;
> > + ? ? ? }
> > +
> > + ? ? ? sdma->num_bd = sg_len;
> > + ? ? ? channel_control[channel].current_bd_ptr = sdma->bd_phys;
> > +
> > + ? ? ? return &sdma->desc;
> > +err_out:
> > + ? ? ? return NULL;
> > +}
> > +
> > +static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
> > + ? ? ? ? ? ? ? struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
> > + ? ? ? ? ? ? ? size_t period_len, enum dma_data_direction direction)
> > +{
> > + ? ? ? int num_periods = buf_len / period_len;
> > + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> > + ? ? ? int channel = chan->chan_id;
> > + ? ? ? int ret, i = 0, buf = 0;
> > +
> > + ? ? ? pr_debug("%s channel: %d\n", __func__, channel);
>
> Must be possible to find struct device * and use dev_dbg()
>
> > +
> > + ? ? ? if (sdma->busy)
> > + ? ? ? ? ? ? ? return NULL;
> > +
> > + ? ? ? sdma->busy = 1;
> > +
> > + ? ? ? sdma->flags |= IMX_DMA_SG_LOOP;
> > + ? ? ? sdma->direction = direction;
> > + ? ? ? ret = sdma_load_context(channel);
> > + ? ? ? if (ret)
> > + ? ? ? ? ? ? ? goto err_out;
> > +
> > + ? ? ? if (num_periods > NUM_BD) {
> > + ? ? ? ? ? ? ? pr_err("SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? channel, num_periods, NUM_BD);
> > + ? ? ? ? ? ? ? goto err_out;
> > + ? ? ? }
> > +
> > + ? ? ? if (period_len > 0xffff) {
> > + ? ? ? ? ? ? ? pr_err("SDMA channel %d: maximum period size exceeded: %d > %d\n",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? channel, period_len, 0xffff);
> > + ? ? ? ? ? ? ? goto err_out;
> > + ? ? ? }
> > +
> > + ? ? ? while (buf < buf_len) {
> > + ? ? ? ? ? ? ? struct sdma_buffer_descriptor *bd = &sdma->bd[i];
> > + ? ? ? ? ? ? ? int param;
> > +
> > + ? ? ? ? ? ? ? bd->buffer_addr = dma_addr;
> > +
> > + ? ? ? ? ? ? ? bd->mode.count = period_len;
> > +
> > + ? ? ? ? ? ? ? if (sdma->word_size > 4)
> > + ? ? ? ? ? ? ? ? ? ? ? goto err_out;
> > + ? ? ? ? ? ? ? if (sdma->word_size == 4)
> > + ? ? ? ? ? ? ? ? ? ? ? bd->mode.command = 0;
> > + ? ? ? ? ? ? ? else
> > + ? ? ? ? ? ? ? ? ? ? ? bd->mode.command = sdma->word_size;
> > +
> > + ? ? ? ? ? ? ? param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
> > + ? ? ? ? ? ? ? if (i + 1 == num_periods)
> > + ? ? ? ? ? ? ? ? ? ? ? param |= BD_WRAP;
> > +
> > + ? ? ? ? ? ? ? pr_debug("entry %d: count: %d dma: 0x%08x %s%s\n",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? i, period_len, dma_addr,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param & BD_WRAP ? "wrap" : "",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? param & BD_INTR ? " intr" : "");
> > +
> > + ? ? ? ? ? ? ? bd->mode.status = param;
> > +
> > + ? ? ? ? ? ? ? dma_addr += period_len;
> > + ? ? ? ? ? ? ? buf += period_len;
> > +
> > + ? ? ? ? ? ? ? i++;
> > + ? ? ? }
> > +
> > + ? ? ? sdma->num_bd = num_periods;
> > + ? ? ? channel_control[channel].current_bd_ptr = sdma->bd_phys;
> > +
> > + ? ? ? return &sdma->desc;
> > +err_out:
> > + ? ? ? sdma->busy = 0;
> > + ? ? ? return NULL;
> > +}
> > +
> > +static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> > + ? ? ? ? ? ? ? unsigned long arg)
> > +{
> > + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> > + ? ? ? struct dma_slave_config *dmaengine_cfg = (void *)arg;
> > +
> > + ? ? ? switch (cmd) {
> > + ? ? ? case DMA_TERMINATE_ALL:
> > + ? ? ? ? ? ? ? sdma_disable_channel(chan->chan_id);
> > + ? ? ? ? ? ? ? return 0;
> > + ? ? ? case DMA_SLAVE_CONFIG:
> > + ? ? ? ? ? ? ? if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->per_address = dmaengine_cfg->src_addr;
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->watermark_level = dmaengine_cfg->src_maxburst;
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->word_size = dmaengine_cfg->src_addr_width;
> > + ? ? ? ? ? ? ? } else {
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->per_address = dmaengine_cfg->dst_addr;
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->watermark_level = dmaengine_cfg->dst_maxburst;
> > + ? ? ? ? ? ? ? ? ? ? ? sdma->word_size = dmaengine_cfg->dst_addr_width;
> > + ? ? ? ? ? ? ? }
> > + ? ? ? ? ? ? ? return sdma_config_channel(chan->chan_id);
> > + ? ? ? default:
> > + ? ? ? ? ? ? ? return -ENOSYS;
> > + ? ? ? }
> > +
> > + ? ? ? return -EINVAL;
> > +}
> > +
> > +static enum dma_status sdma_tx_status(struct dma_chan *chan,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? dma_cookie_t cookie,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? struct dma_tx_state *txstate)
> > +{
> > + ? ? ? struct sdma_channel *sdma = to_sdma_chan(chan);
> > + ? ? ? dma_cookie_t last_used;
> > + ? ? ? enum dma_status ret;
> > +
> > + ? ? ? last_used = chan->cookie;
> > +
> > + ? ? ? ret = dma_async_is_complete(cookie, sdma->last_completed, last_used);
> > + ? ? ? dma_set_tx_state(txstate, sdma->last_completed, last_used, 0);
> > +
> > + ? ? ? return ret;
> > +}
> > +
> > +static void sdma_issue_pending(struct dma_chan *chan)
> > +{
> > + ? ? ? /*
> > + ? ? ? ?* Nothing to do. We only have a single descriptor
> > + ? ? ? ?*/
> > +}
> > +
> > +static int __devinit sdma_probe(struct platform_device *pdev)
> > +{
> > + ? ? ? int ret;
> > + ? ? ? const struct firmware *fw;
> > + ? ? ? const struct sdma_firmware_header *header;
> > + ? ? ? const struct sdma_script_start_addrs *addr;
> > + ? ? ? int irq;
> > + ? ? ? unsigned short *ram_code;
> > + ? ? ? struct resource *iores;
> > + ? ? ? struct sdma_platform_data *pdata = pdev->dev.platform_data;
> > + ? ? ? int version;
> > + ? ? ? char *cpustr, *fwname;
> > + ? ? ? int i;
> > + ? ? ? dma_cap_mask_t mask;
> > +
> > + ? ? ? /* there can be only one */
> > + ? ? ? BUG_ON(sdma_base);
> > +
> > + ? ? ? iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > + ? ? ? irq = platform_get_irq(pdev, 0);
> > + ? ? ? if (!iores || irq < 0 || !pdata)
> > + ? ? ? ? ? ? ? return -EINVAL;
> > +
> > + ? ? ? sdma_clk = clk_get(&pdev->dev, NULL);
> > + ? ? ? if (IS_ERR(sdma_clk)) {
> > + ? ? ? ? ? ? ? ret = PTR_ERR(sdma_clk);
> > + ? ? ? ? ? ? ? goto err_clk;
> > + ? ? ? }
> > +
> > + ? ? ? if (cpu_is_mx31()) {
> > + ? ? ? ? ? ? ? cpustr = "imx31";
> > + ? ? ? ? ? ? ? version = mx31_revision() >> 4;
> > + ? ? ? } else if (cpu_is_mx35()) {
> > + ? ? ? ? ? ? ? cpustr = "imx35";
> > +/* FIXME: ? ? ?version = mx35_revision(); */
> > + ? ? ? ? ? ? ? version = 2;
> > + ? ? ? } else {
> > + ? ? ? ? ? ? ? ret = -EINVAL;
> > + ? ? ? ? ? ? ? goto err_cputype;
> > + ? ? ? }
> > +
> > + ? ? ? fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin", cpustr, version);
> > + ? ? ? if (!fwname) {
> > + ? ? ? ? ? ? ? ret = -ENOMEM;
> > + ? ? ? ? ? ? ? goto err_cputype;
> > + ? ? ? }
> > +
> > + ? ? ? ret = request_firmware(&fw, fwname, &pdev->dev);
> > + ? ? ? if (ret) {
> > + ? ? ? ? ? ? ? dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? fwname, ret);
> > + ? ? ? ? ? ? ? kfree(fwname);
> > + ? ? ? ? ? ? ? goto err_cputype;
> > + ? ? ? }
> > + ? ? ? kfree(fwname);
> > +
> > + ? ? ? if (fw->size < sizeof(*header))
> > + ? ? ? ? ? ? ? goto err_firmware;
> > +
> > + ? ? ? header = (struct sdma_firmware_header *)fw->data;
> > +
> > + ? ? ? if (header->magic != SDMA_FIRMWARE_MAGIC)
> > + ? ? ? ? ? ? ? goto err_firmware;
> > + ? ? ? if (header->ram_code_start + header->ram_code_size > fw->size)
> > + ? ? ? ? ? ? ? goto err_firmware;
> > +
> > + ? ? ? addr = (void *)header + header->script_addrs_start;
> > + ? ? ? ram_code = (void *)header + header->ram_code_start;
> > + ? ? ? memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
> > +
> > + ? ? ? ret = sdma_init(iores->start, irq, pdata->sdma_version,
> > + ? ? ? ? ? ? ? ? ? ? ? ram_code, header->ram_code_size);
> > + ? ? ? if (ret)
> > + ? ? ? ? ? ? ? goto err_firmware;
> > +
> > + ? ? ? INIT_LIST_HEAD(&sdma_dma_device->channels);
> > +
> > + ? ? ? /* Initialize channel parameters */
> > + ? ? ? for (i = 0; i < MAX_DMA_CHANNELS; i++) {
> > + ? ? ? ? ? ? ? struct sdma_channel *sdma = &sdma_data[i];
> > +
> > + ? ? ? ? ? ? ? spin_lock_init(&sdma->lock);
> > +
> > + ? ? ? ? ? ? ? dma_cap_set(DMA_SLAVE, sdma_dma_device->cap_mask);
> > + ? ? ? ? ? ? ? dma_cap_set(DMA_CYCLIC, sdma_dma_device->cap_mask);
> > +
> > + ? ? ? ? ? ? ? sdma->chan.device = sdma_dma_device;
> > + ? ? ? ? ? ? ? sdma->chan.chan_id = i;
> > +
> > + ? ? ? ? ? ? ? /* Add the channel to the DMAC list */
> > + ? ? ? ? ? ? ? list_add_tail(&sdma->chan.device_node, &sdma_dma_device->channels);
> > + ? ? ? }
> > +
> > + ? ? ? sdma_dma_device->dev = &pdev->dev;
> > +
> > + ? ? ? sdma_dma_device->device_alloc_chan_resources = sdma_alloc_chan_resources;
> > + ? ? ? sdma_dma_device->device_free_chan_resources = sdma_free_chan_resources;
> > + ? ? ? sdma_dma_device->device_tx_status = sdma_tx_status;
> > + ? ? ? sdma_dma_device->device_prep_slave_sg = sdma_prep_slave_sg;
> > + ? ? ? sdma_dma_device->device_prep_dma_cyclic = sdma_prep_dma_cyclic;
> > + ? ? ? sdma_dma_device->device_control = sdma_control;
> > + ? ? ? sdma_dma_device->device_issue_pending = sdma_issue_pending;
> > +
> > + ? ? ? ret = dma_async_device_register(sdma_dma_device);
> > + ? ? ? if (ret) {
> > + ? ? ? ? ? ? ? dev_err(&pdev->dev, "unable to register DMAC\n");
>
> SDMAC even?
Better just "unable to register". The name of the device will give
enough information.
>
> > + ? ? ? ? ? ? ? goto err_firmware;
> > + ? ? ? }
> > +
> > + ? ? ? dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
> > + ? ? ? ? ? ? ? ? ? ? ? header->version_major,
> > + ? ? ? ? ? ? ? ? ? ? ? header->version_minor);
> > +
> > + ? ? ? /* request channel 0. This is an internal control channel
> > + ? ? ? ?* to the SDMA engine and not available to clients.
> > + ? ? ? ?*/
> > + ? ? ? dma_cap_zero(mask);
> > + ? ? ? dma_cap_set(DMA_SLAVE, mask);
> > + ? ? ? dma_request_channel(mask, NULL, NULL);
> > +
> > + ? ? ? release_firmware(fw);
> > +
> > + ? ? ? return 0;
> > +
> > +err_firmware:
> > + ? ? ? release_firmware(fw);
> > +err_cputype:
> > + ? ? ? clk_put(sdma_clk);
> > +err_clk:
> > + ? ? ? return 0;
> > +}
> > +
> > +static int __devexit sdma_remove(struct platform_device *pdev)
> > +{
> > + ? ? ? return -EBUSY;
> > +}
> > +
> > +static struct platform_driver sdma_driver = {
> > + ? ? ? .driver ? ? ? ? = {
> > + ? ? ? ? ? ? ? .name ? = "imx-sdma",
> > + ? ? ? },
> > + ? ? ? .probe ? ? ? ? ?= sdma_probe,
> > + ? ? ? .remove ? ? ? ? = __devexit_p(sdma_remove),
> > +};
> > +
> > +static int __init sdma_module_init(void)
> > +{
> > + ? ? ? return platform_driver_register(&sdma_driver);
> > +}
> > +subsys_initcall(sdma_module_init);
> > +
> > +MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
> > +MODULE_DESCRIPTION("i.MX SDMA driver");
> > +MODULE_LICENSE("GPL");
> > --
> > 1.7.1
>
> Thanks for using this API
> Sascha!
>
> Yours,
> Linus Walleij
>
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-16 14:15 ` Sascha Hauer
@ 2010-08-17 4:36 ` Baruch Siach
-1 siblings, 0 replies; 78+ messages in thread
From: Baruch Siach @ 2010-08-17 4:36 UTC (permalink / raw)
To: Sascha Hauer; +Cc: Linus Walleij, Dan Williams, linux-kernel, linux-arm-kernel
Hi Sascha,
On Mon, Aug 16, 2010 at 04:15:40PM +0200, Sascha Hauer wrote:
> On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> > 2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> > > The SDMA engine is a scatter/gather DMA engine which is implemented
> > > as a seperate coprocessor. SDMA needs its own firmware which is
> > > requested using the standard request_firmware mechanism. The firmware
> > > has different entry points for each peripheral type, so drivers
> > > have to pass the peripheral type to the DMA engine which in turn
> > > picks the correct firmware entry point from a table contained in
> > > the firmware image itself.
> >
> > Quite fun, if the spec for the microcode is open this opens up
> > for dynamic firmware generation for specific DMA jobs does it
> > not?
>
> Unfortunately the specs are not open, so we are sticked to the binary
> microcode from Freescale. I'm pretty sure though that the SDMA engine
> could do at least a device_prep_dma_xor operation.
Chapter 38 in the i.MX25 Reference Manual seems to include almost everything
there is to know about the SDMA. Isn't this enough for writing custom SDMA
microcodes?
baruch
--
~. .~ Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
- baruch@tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-17 4:36 ` Baruch Siach
0 siblings, 0 replies; 78+ messages in thread
From: Baruch Siach @ 2010-08-17 4:36 UTC (permalink / raw)
To: linux-arm-kernel
Hi Sascha,
On Mon, Aug 16, 2010 at 04:15:40PM +0200, Sascha Hauer wrote:
> On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> > 2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> > > The SDMA engine is a scatter/gather DMA engine which is implemented
> > > as a seperate coprocessor. SDMA needs its own firmware which is
> > > requested using the standard request_firmware mechanism. The firmware
> > > has different entry points for each peripheral type, so drivers
> > > have to pass the peripheral type to the DMA engine which in turn
> > > picks the correct firmware entry point from a table contained in
> > > the firmware image itself.
> >
> > Quite fun, if the spec for the microcode is open this opens up
> > for dynamic firmware generation for specific DMA jobs does it
> > not?
>
> Unfortunately the specs are not open, so we are sticked to the binary
> microcode from Freescale. I'm pretty sure though that the SDMA engine
> could do at least a device_prep_dma_xor operation.
Chapter 38 in the i.MX25 Reference Manual seems to include almost everything
there is to know about the SDMA. Isn't this enough for writing custom SDMA
microcodes?
baruch
--
~. .~ Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
- baruch at tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-17 4:36 ` Baruch Siach
@ 2010-08-17 6:47 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-17 6:47 UTC (permalink / raw)
To: Baruch Siach; +Cc: Linus Walleij, Dan Williams, linux-kernel, linux-arm-kernel
On Tue, Aug 17, 2010 at 07:36:12AM +0300, Baruch Siach wrote:
> Hi Sascha,
>
> On Mon, Aug 16, 2010 at 04:15:40PM +0200, Sascha Hauer wrote:
> > On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> > > 2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> > > > The SDMA engine is a scatter/gather DMA engine which is implemented
> > > > as a seperate coprocessor. SDMA needs its own firmware which is
> > > > requested using the standard request_firmware mechanism. The firmware
> > > > has different entry points for each peripheral type, so drivers
> > > > have to pass the peripheral type to the DMA engine which in turn
> > > > picks the correct firmware entry point from a table contained in
> > > > the firmware image itself.
> > >
> > > Quite fun, if the spec for the microcode is open this opens up
> > > for dynamic firmware generation for specific DMA jobs does it
> > > not?
> >
> > Unfortunately the specs are not open, so we are sticked to the binary
> > microcode from Freescale. I'm pretty sure though that the SDMA engine
> > could do at least a device_prep_dma_xor operation.
>
> Chapter 38 in the i.MX25 Reference Manual seems to include almost everything
> there is to know about the SDMA. Isn't this enough for writing custom SDMA
> microcodes?
Unfortunately not, the assembler is missing.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-17 6:47 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-17 6:47 UTC (permalink / raw)
To: linux-arm-kernel
On Tue, Aug 17, 2010 at 07:36:12AM +0300, Baruch Siach wrote:
> Hi Sascha,
>
> On Mon, Aug 16, 2010 at 04:15:40PM +0200, Sascha Hauer wrote:
> > On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> > > 2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> > > > The SDMA engine is a scatter/gather DMA engine which is implemented
> > > > as a seperate coprocessor. SDMA needs its own firmware which is
> > > > requested using the standard request_firmware mechanism. The firmware
> > > > has different entry points for each peripheral type, so drivers
> > > > have to pass the peripheral type to the DMA engine which in turn
> > > > picks the correct firmware entry point from a table contained in
> > > > the firmware image itself.
> > >
> > > Quite fun, if the spec for the microcode is open this opens up
> > > for dynamic firmware generation for specific DMA jobs does it
> > > not?
> >
> > Unfortunately the specs are not open, so we are sticked to the binary
> > microcode from Freescale. I'm pretty sure though that the SDMA engine
> > could do at least a device_prep_dma_xor operation.
>
> Chapter 38 in the i.MX25 Reference Manual seems to include almost everything
> there is to know about the SDMA. Isn't this enough for writing custom SDMA
> microcodes?
Unfortunately not, the assembler is missing.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-17 6:47 ` Sascha Hauer
(?)
@ 2010-08-18 3:49 ` Alexei Babich
2010-08-18 4:41 ` Baruch Siach
-1 siblings, 1 reply; 78+ messages in thread
From: Alexei Babich @ 2010-08-18 3:49 UTC (permalink / raw)
To: linux-arm-kernel
> Unfortunately not, the assembler is missing.
The instruction set, if I'm not mistaken, is similar to the M-core
--
Regards,
Alexei Babich, circuit design engineer, Rezonans plc., Chelyabinsk, Russia
http://www.rez.ru
Jabber ID: impatt at jabber.ru
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-18 3:49 ` Alexei Babich
@ 2010-08-18 4:41 ` Baruch Siach
0 siblings, 0 replies; 78+ messages in thread
From: Baruch Siach @ 2010-08-18 4:41 UTC (permalink / raw)
To: linux-arm-kernel
Hi Alexei,
On Wed, Aug 18, 2010 at 09:49:11AM +0600, Alexei Babich wrote:
> > Unfortunately not, the assembler is missing.
> The instruction set, if I'm not mistaken, is similar to the M-core
Thanks for the tip.
The document with the necessary M-core assembly instructions details, the
MCORERC/AD, is not publicly available, as far as I can see.
baruch
--
~. .~ Tk Open Systems
=}------------------------------------------------ooO--U--Ooo------------{=
- baruch at tkos.co.il - tel: +972.2.679.5364, http://www.tkos.co.il -
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-17 6:47 ` Sascha Hauer
@ 2010-08-18 11:17 ` Philippe Rétornaz
-1 siblings, 0 replies; 78+ messages in thread
From: Philippe Rétornaz @ 2010-08-18 11:17 UTC (permalink / raw)
To: linux-arm-kernel
Cc: Sascha Hauer, Baruch Siach, Linus Walleij, Dan Williams, linux-kernel
Le mardi, 17 août 2010 08.47:34, Sascha Hauer a écrit :
> On Tue, Aug 17, 2010 at 07:36:12AM +0300, Baruch Siach wrote:
> > Hi Sascha,
> >
> > On Mon, Aug 16, 2010 at 04:15:40PM +0200, Sascha Hauer wrote:
> > > On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> > > > 2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> > > > > The SDMA engine is a scatter/gather DMA engine which is implemented
> > > > > as a seperate coprocessor. SDMA needs its own firmware which is
> > > > > requested using the standard request_firmware mechanism. The
> > > > > firmware has different entry points for each peripheral type, so
> > > > > drivers have to pass the peripheral type to the DMA engine which in
> > > > > turn picks the correct firmware entry point from a table contained
> > > > > in the firmware image itself.
> > > >
> > > > Quite fun, if the spec for the microcode is open this opens up
> > > > for dynamic firmware generation for specific DMA jobs does it
> > > > not?
> > >
> > > Unfortunately the specs are not open, so we are sticked to the binary
> > > microcode from Freescale. I'm pretty sure though that the SDMA engine
> > > could do at least a device_prep_dma_xor operation.
> >
> > Chapter 38 in the i.MX25 Reference Manual seems to include almost
> > everything there is to know about the SDMA. Isn't this enough for writing
> > custom SDMA microcodes?
>
> Unfortunately not, the assembler is missing.
>
I think the instruction list and encoding is available in the imx31 reference
manual at section 40.19.
BTW, I tested the non-dmaengine version of your patch on imx31, works great !
Regards,
Philippe
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-18 11:17 ` Philippe Rétornaz
0 siblings, 0 replies; 78+ messages in thread
From: Philippe Rétornaz @ 2010-08-18 11:17 UTC (permalink / raw)
To: linux-arm-kernel
Le mardi, 17 ao?t 2010 08.47:34, Sascha Hauer a ?crit :
> On Tue, Aug 17, 2010 at 07:36:12AM +0300, Baruch Siach wrote:
> > Hi Sascha,
> >
> > On Mon, Aug 16, 2010 at 04:15:40PM +0200, Sascha Hauer wrote:
> > > On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> > > > 2010/8/16 Sascha Hauer <s.hauer@pengutronix.de>:
> > > > > The SDMA engine is a scatter/gather DMA engine which is implemented
> > > > > as a seperate coprocessor. SDMA needs its own firmware which is
> > > > > requested using the standard request_firmware mechanism. The
> > > > > firmware has different entry points for each peripheral type, so
> > > > > drivers have to pass the peripheral type to the DMA engine which in
> > > > > turn picks the correct firmware entry point from a table contained
> > > > > in the firmware image itself.
> > > >
> > > > Quite fun, if the spec for the microcode is open this opens up
> > > > for dynamic firmware generation for specific DMA jobs does it
> > > > not?
> > >
> > > Unfortunately the specs are not open, so we are sticked to the binary
> > > microcode from Freescale. I'm pretty sure though that the SDMA engine
> > > could do at least a device_prep_dma_xor operation.
> >
> > Chapter 38 in the i.MX25 Reference Manual seems to include almost
> > everything there is to know about the SDMA. Isn't this enough for writing
> > custom SDMA microcodes?
>
> Unfortunately not, the assembler is missing.
>
I think the instruction list and encoding is available in the imx31 reference
manual at section 40.19.
BTW, I tested the non-dmaengine version of your patch on imx31, works great !
Regards,
Philippe
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-23 7:17 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-23 7:17 UTC (permalink / raw)
To: linux-kernel; +Cc: Linus Walleij, Dan Williams, linux-arm-kernel
On Mon, Aug 16, 2010 at 01:07:49PM +0200, Sascha Hauer wrote:
> Currently dmaengine users have to explicitely dereference function
> pointers in struct dma_device. For the convenience of drivers and
> to be more flexible when changing the dmaengine later add static
> inline wrapper functions for the dma commands.
>
> This patch is not complete yet. If there's consensus on this patch
> I'll provide an updated patch with the missing functions.
Dan,
Any comment on this one?
Sascha
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
> include/linux/dmaengine.h | 41 +++++++++++++++++++++++++++++++++++++++++
> 1 files changed, 41 insertions(+), 0 deletions(-)
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index 0df7864..635c60b 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -491,6 +491,47 @@ struct dma_device {
> void (*device_issue_pending)(struct dma_chan *chan);
> };
>
> +static inline int dmaengine_device_control(struct dma_chan *chan,
> + enum dma_ctrl_cmd cmd,
> + unsigned long arg)
> +{
> + return chan->device->device_control(chan, cmd, arg);
> +}
> +
> +static inline int dmaengine_slave_config(struct dma_chan *chan,
> + struct dma_slave_config *config)
> +{
> + return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
> + (unsigned long)config);
> +}
> +
> +static inline int dmaengine_terminate_all(struct dma_chan *chan)
> +{
> + return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
> + struct dma_chan *chan, struct scatterlist *sgl,
> + unsigned int sg_len, enum dma_data_direction direction,
> + unsigned long flags)
> +{
> + return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
> + flags);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction)
> +{
> + return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
> + period_len, direction);
> +}
> +
> +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
> +{
> + return desc->tx_submit(desc);
> +}
> +
> static inline bool dmaengine_check_align(u8 align, size_t off1, size_t off2, size_t len)
> {
> size_t mask;
> --
> 1.7.1
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
@ 2010-08-23 7:17 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-23 7:17 UTC (permalink / raw)
To: linux-arm-kernel
On Mon, Aug 16, 2010 at 01:07:49PM +0200, Sascha Hauer wrote:
> Currently dmaengine users have to explicitely dereference function
> pointers in struct dma_device. For the convenience of drivers and
> to be more flexible when changing the dmaengine later add static
> inline wrapper functions for the dma commands.
>
> This patch is not complete yet. If there's consensus on this patch
> I'll provide an updated patch with the missing functions.
Dan,
Any comment on this one?
Sascha
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
> include/linux/dmaengine.h | 41 +++++++++++++++++++++++++++++++++++++++++
> 1 files changed, 41 insertions(+), 0 deletions(-)
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index 0df7864..635c60b 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -491,6 +491,47 @@ struct dma_device {
> void (*device_issue_pending)(struct dma_chan *chan);
> };
>
> +static inline int dmaengine_device_control(struct dma_chan *chan,
> + enum dma_ctrl_cmd cmd,
> + unsigned long arg)
> +{
> + return chan->device->device_control(chan, cmd, arg);
> +}
> +
> +static inline int dmaengine_slave_config(struct dma_chan *chan,
> + struct dma_slave_config *config)
> +{
> + return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
> + (unsigned long)config);
> +}
> +
> +static inline int dmaengine_terminate_all(struct dma_chan *chan)
> +{
> + return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
> + struct dma_chan *chan, struct scatterlist *sgl,
> + unsigned int sg_len, enum dma_data_direction direction,
> + unsigned long flags)
> +{
> + return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
> + flags);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction)
> +{
> + return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
> + period_len, direction);
> +}
> +
> +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
> +{
> + return desc->tx_submit(desc);
> +}
> +
> static inline bool dmaengine_check_align(u8 align, size_t off1, size_t off2, size_t len)
> {
> size_t mask;
> --
> 1.7.1
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-23 12:57 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-23 12:57 UTC (permalink / raw)
To: linux-kernel; +Cc: Linus Walleij, Dan Williams, linux-arm-kernel
This patch adds support for the Freescale i.MX SDMA engine.
The SDMA engine is a scatter/gather DMA engine which is implemented
as a seperate coprocessor. SDMA needs its own firmware which is
requested using the standard request_firmware mechanism. The firmware
has different entry points for each peripheral type, so drivers
have to pass the peripheral type to the DMA engine which in turn
picks the correct firmware entry point from a table contained in
the firmware image itself.
The original Freescale code also supports support for transfering
data to the internal SRAM which needs different entry points to
the firmware. Support for this is currently not implemented. Also,
support for the ASRC (asymmetric sample rate converter) is skipped.
I took a very simple approach to implement dmaengine support. Only
a single descriptor is statically assigned to a each channel. This
means that transfers can't be queued up but only a single transfer
is in progress. This simplifies implementation a lot and is sufficient
for the usual device/memory transfers.
Changes since v1:
- included comments from Linus Walleij
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
arch/arm/mach-mx3/Kconfig | 2 +
arch/arm/plat-mxc/Kconfig | 10 +
arch/arm/plat-mxc/include/mach/dma.h | 64 ++
arch/arm/plat-mxc/include/mach/sdma.h | 17 +
drivers/dma/Kconfig | 8 +
drivers/dma/Makefile | 1 +
drivers/dma/imx-sdma.c | 1395 +++++++++++++++++++++++++++++++
8 files changed, 1499 insertions(+), 6 deletions(-)
create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
create mode 100644 drivers/dma/imx-sdma.c
diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
index 287431c..ac6fd71 100644
--- a/arch/arm/mach-imx/include/mach/dma-v1.h
+++ b/arch/arm/mach-imx/include/mach/dma-v1.h
@@ -27,6 +27,8 @@
#define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
+#include <mach/dma.h>
+
#define IMX_DMA_CHANNELS 16
#define DMA_MODE_READ 0
@@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
void imx_dma_free(int channel);
-enum imx_dma_prio {
- DMA_PRIO_HIGH = 0,
- DMA_PRIO_MEDIUM = 1,
- DMA_PRIO_LOW = 2
-};
-
int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
#endif /* __MACH_DMA_V1_H__ */
diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
index 85beece..301375c 100644
--- a/arch/arm/mach-mx3/Kconfig
+++ b/arch/arm/mach-mx3/Kconfig
@@ -3,12 +3,14 @@ if ARCH_MX3
config ARCH_MX31
select ARCH_HAS_RNGA
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
bool
config ARCH_MX35
bool
select ARCH_MXC_IOMUX_V3
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
comment "MX3 platforms:"
diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
index 0527e65..6741625 100644
--- a/arch/arm/plat-mxc/Kconfig
+++ b/arch/arm/plat-mxc/Kconfig
@@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
config ARCH_MXC_AUDMUX_V2
bool
+config IMX_HAVE_SDMA
+ bool
+
+config IMX_SDMA
+ depends on IMX_HAVE_SDMA
+ tristate "Enable SDMA support"
+ help
+ Include support for the SDMA engine. The SDMA engine needs additional
+ firmware support. SDMA can be compiled as a module to support loading
+ the firmware when a rootfs is present.
endif
diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
new file mode 100644
index 0000000..69d181f
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/dma.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARCH_MXC_DMA_H__
+#define __ASM_ARCH_MXC_DMA_H__
+
+#include <linux/scatterlist.h>
+
+/*
+ * This enumerates peripheral types. Used for SDMA.
+ */
+enum sdma_peripheral_type {
+ IMX_DMATYPE_SSI, /* MCU domain SSI */
+ IMX_DMATYPE_SSI_SP, /* Shared SSI */
+ IMX_DMATYPE_MMC, /* MMC */
+ IMX_DMATYPE_SDHC, /* SDHC */
+ IMX_DMATYPE_UART, /* MCU domain UART */
+ IMX_DMATYPE_UART_SP, /* Shared UART */
+ IMX_DMATYPE_FIRI, /* FIRI */
+ IMX_DMATYPE_CSPI, /* MCU domain CSPI */
+ IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
+ IMX_DMATYPE_SIM, /* SIM */
+ IMX_DMATYPE_ATA, /* ATA */
+ IMX_DMATYPE_CCM, /* CCM */
+ IMX_DMATYPE_EXT, /* External peripheral */
+ IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
+ IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
+ IMX_DMATYPE_DSP, /* DSP */
+ IMX_DMATYPE_MEMORY, /* Memory */
+ IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
+ IMX_DMATYPE_SPDIF, /* SPDIF */
+ IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
+ IMX_DMATYPE_ASRC, /* ASRC */
+ IMX_DMATYPE_ESAI, /* ESAI */
+};
+
+enum imx_dma_prio {
+ DMA_PRIO_HIGH = 0,
+ DMA_PRIO_MEDIUM = 1,
+ DMA_PRIO_LOW = 2
+};
+
+struct imx_dma_data {
+ int dma_request; /* DMA request line */
+ enum sdma_peripheral_type peripheral_type;
+ int priority;
+};
+
+static inline int imx_dma_is_ipu(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "ipu-core");
+}
+
+static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "imx-sdma");
+}
+
+#endif
diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
new file mode 100644
index 0000000..9be1122
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/sdma.h
@@ -0,0 +1,17 @@
+#ifndef __MACH_MXC_SDMA_H__
+#define __MACH_MXC_SDMA_H__
+
+/**
+ * struct sdma_platform_data - platform specific data for SDMA engine
+ *
+ * @sdma_version The version of this SDMA engine
+ * @cpu_name used to generate the firmware name
+ * @to_version CPU Tape out version
+ */
+struct sdma_platform_data {
+ int sdma_version;
+ char *cpu_name;
+ int to_version;
+};
+
+#endif /* __MACH_MXC_SDMA_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9520cf0..ff68307 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -195,6 +195,14 @@ config PCH_DMA
help
Enable support for the Topcliff PCH DMA engine.
+config IMX_SDMA
+ tristate "i.MX SDMA support"
+ depends on ARCH_MXC
+ select DMA_ENGINE
+ help
+ Support the i.MX SDMA engine. This engine is integrated into
+ Freescale i.MX25/31/35/51 chips.
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 72bd703..14d7a1b 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
obj-$(CONFIG_PCH_DMA) += pch_dma.o
+obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
new file mode 100644
index 0000000..c447fc0
--- /dev/null
+++ b/drivers/dma/imx-sdma.c
@@ -0,0 +1,1395 @@
+/*
+ * drivers/dma/imx-sdma.c
+ *
+ * This file contains a driver for the Freescale Smart DMA engine
+ *
+ * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
+ *
+ * Based on code from Freescale:
+ *
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * The code contained herein is licensed under the GNU General Public
+ * License. You may obtain a copy of the GNU General Public License
+ * Version 2 or later at the following locations:
+ *
+ * http://www.opensource.org/licenses/gpl-license.html
+ * http://www.gnu.org/copyleft/gpl.html
+ */
+
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/dmaengine.h>
+
+#include <asm/irq.h>
+#include <mach/sdma.h>
+#include <mach/dma.h>
+#include <mach/hardware.h>
+
+/* SDMA registers */
+#define SDMA_H_C0PTR (sdma->regs + 0x000)
+#define SDMA_H_INTR (sdma->regs + 0x004)
+#define SDMA_H_STATSTOP (sdma->regs + 0x008)
+#define SDMA_H_START (sdma->regs + 0x00c)
+#define SDMA_H_EVTOVR (sdma->regs + 0x010)
+#define SDMA_H_DSPOVR (sdma->regs + 0x014)
+#define SDMA_H_HOSTOVR (sdma->regs + 0x018)
+#define SDMA_H_EVTPEND (sdma->regs + 0x01c)
+#define SDMA_H_DSPENBL (sdma->regs + 0x020)
+#define SDMA_H_RESET (sdma->regs + 0x024)
+#define SDMA_H_EVTERR (sdma->regs + 0x028)
+#define SDMA_H_INTRMSK (sdma->regs + 0x02c)
+#define SDMA_H_PSW (sdma->regs + 0x030)
+#define SDMA_H_EVTERRDBG (sdma->regs + 0x034)
+#define SDMA_H_CONFIG (sdma->regs + 0x038)
+#define SDMA_ONCE_ENB (sdma->regs + 0x040)
+#define SDMA_ONCE_DATA (sdma->regs + 0x044)
+#define SDMA_ONCE_INSTR (sdma->regs + 0x048)
+#define SDMA_ONCE_STAT (sdma->regs + 0x04c)
+#define SDMA_ONCE_CMD (sdma->regs + 0x050)
+#define SDMA_EVT_MIRROR (sdma->regs + 0x054)
+#define SDMA_ILLINSTADDR (sdma->regs + 0x058)
+#define SDMA_CHN0ADDR (sdma->regs + 0x05c)
+#define SDMA_ONCE_RTB (sdma->regs + 0x060)
+#define SDMA_XTRIG_CONF1 (sdma->regs + 0x070)
+#define SDMA_XTRIG_CONF2 (sdma->regs + 0x074)
+#define SDMA_CHNENBL_0 (sdma->regs + (sdma->version == 2 ? 0x200 : 0x80))
+#define SDMA_CHNPRI_0 (sdma->regs + 0x100)
+
+/*
+ * Buffer descriptor status values.
+ */
+#define BD_DONE 0x01
+#define BD_WRAP 0x02
+#define BD_CONT 0x04
+#define BD_INTR 0x08
+#define BD_RROR 0x10
+#define BD_LAST 0x20
+#define BD_EXTD 0x80
+
+/*
+ * Data Node descriptor status values.
+ */
+#define DND_END_OF_FRAME 0x80
+#define DND_END_OF_XFER 0x40
+#define DND_DONE 0x20
+#define DND_UNUSED 0x01
+
+/*
+ * IPCV2 descriptor status values.
+ */
+#define BD_IPCV2_END_OF_FRAME 0x40
+
+#define IPCV2_MAX_NODES 50
+/*
+ * Error bit set in the CCB status field by the SDMA,
+ * in setbd routine, in case of a transfer error
+ */
+#define DATA_ERROR 0x10000000
+
+/*
+ * Buffer descriptor commands.
+ */
+#define C0_ADDR 0x01
+#define C0_LOAD 0x02
+#define C0_DUMP 0x03
+#define C0_SETCTX 0x07
+#define C0_GETCTX 0x03
+#define C0_SETDM 0x01
+#define C0_SETPM 0x04
+#define C0_GETDM 0x02
+#define C0_GETPM 0x08
+/*
+ * Change endianness indicator in the BD command field
+ */
+#define CHANGE_ENDIANNESS 0x80
+
+/*
+ * Mode/Count of data node descriptors - IPCv2
+ */
+struct sdma_mode_count {
+ u32 count : 16; /* size of the buffer pointed by this BD */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 command : 8; /* command mostlky used for channel 0 */
+};
+
+/*
+ * Buffer descriptor
+ */
+struct sdma_buffer_descriptor {
+ struct sdma_mode_count mode;
+ dma_addr_t buffer_addr; /* address of the buffer described */
+ dma_addr_t ext_buffer_addr; /* extended buffer address */
+} __attribute__ ((packed));
+
+/*
+ * Channel control Block
+ */
+struct sdma_channel_control {
+ dma_addr_t current_bd_ptr; /* current buffer descriptor processed */
+ dma_addr_t base_bd_ptr; /* first element of buffer descriptor array */
+ u32 unused;
+ u32 unused1;
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_state_registers - SDMA context for a channel
+ *
+ * @pc: program counter
+ * @t: test bit: status of arithmetic & test instruction
+ * @rpc: return program counter
+ * @sf: source fault while loading data
+ * @spc: loop start program counter
+ * @df: destination fault while storing data
+ * @epc: loop end program counter
+ * @lm: loop mode
+ */
+struct sdma_state_registers {
+ u32 pc :14;
+ u32 unused1: 1;
+ u32 t : 1;
+ u32 rpc :14;
+ u32 unused0: 1;
+ u32 sf : 1;
+ u32 spc :14;
+ u32 unused2: 1;
+ u32 df : 1;
+ u32 epc :14;
+ u32 lm : 2;
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_context_data - sdma context specific to a channel
+ *
+ * @channel_state: channel state bits
+ * @gReg: general registers
+ * @mda: burst dma destination address register
+ * @msa: burst dma source address register
+ * @ms: burst dma status register
+ * @md: burst dma data register
+ * @pda: peripheral dma destination address register
+ * @psa: peripheral dma source address register
+ * @ps: peripheral dma status register
+ * @pd: peripheral dma data register
+ * @ca: CRC polynomial register
+ * @cs: CRC accumulator register
+ * @dda: dedicated core destination address register
+ * @dsa: dedicated core source address register
+ * @ds: dedicated core status register
+ * @dd: dedicated core data register
+ */
+struct sdma_context_data {
+ struct sdma_state_registers channel_state;
+ u32 gReg[8];
+ u32 mda;
+ u32 msa;
+ u32 ms;
+ u32 md;
+ u32 pda;
+ u32 psa;
+ u32 ps;
+ u32 pd;
+ u32 ca;
+ u32 cs;
+ u32 dda;
+ u32 dsa;
+ u32 ds;
+ u32 dd;
+ u32 scratch0;
+ u32 scratch1;
+ u32 scratch2;
+ u32 scratch3;
+ u32 scratch4;
+ u32 scratch5;
+ u32 scratch6;
+ u32 scratch7;
+} __attribute__ ((packed));
+
+#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
+
+struct sdma_engine;
+
+/**
+ * struct sdma_channel - housekeeping for a SDMA channel
+ *
+ * @sdma pointer to the SDMA engine for this channel
+ * @channel the channel number, matches dmaengine chan_id
+ * @direction transfer type. Needed for setting SDMA script
+ * @peripheral_type Peripheral type. Needed for setting SDMA script
+ * @event_id aka dma request line
+ * @event_id2 for channels that use 2 events
+ * @word_size peripheral access size
+ * @buf_tail ID of the buffer that was processed
+ * @done channel completion
+ * @num_bd max NUM_BD. number of descriptors currently handling
+ */
+struct sdma_channel {
+ struct sdma_engine *sdma;
+ unsigned int channel;
+ enum dma_data_direction direction;
+ enum sdma_peripheral_type peripheral_type;
+ unsigned int event_id;
+ unsigned int event_id2;
+ enum dma_slave_buswidth word_size;
+ unsigned int buf_tail;
+ struct completion done;
+ unsigned int num_bd;
+ struct sdma_buffer_descriptor *bd;
+ dma_addr_t bd_phys;
+ unsigned int pc_from_device, pc_to_device;
+ unsigned long flags;
+ dma_addr_t per_address;
+ u32 event_mask1, event_mask2;
+ u32 watermark_level;
+ u32 shp_addr, per_addr;
+ struct dma_chan chan;
+ spinlock_t lock;
+ struct dma_async_tx_descriptor desc;
+ dma_cookie_t last_completed;
+ enum dma_status status;
+};
+
+#define IMX_DMA_SG_LOOP (1 << 0)
+
+#define MAX_DMA_CHANNELS 32
+#define MXC_SDMA_DEFAULT_PRIORITY 1
+#define MXC_SDMA_MIN_PRIORITY 1
+#define MXC_SDMA_MAX_PRIORITY 7
+
+/*
+ * This enumerates transfer types
+ */
+enum {
+ emi_2_per = 0, /* EMI memory to peripheral */
+ emi_2_int, /* EMI memory to internal RAM */
+ emi_2_emi, /* EMI memory to EMI memory */
+ emi_2_dsp, /* EMI memory to DSP memory */
+ per_2_int, /* Peripheral to internal RAM */
+ per_2_emi, /* Peripheral to internal EMI memory */
+ per_2_dsp, /* Peripheral to DSP memory */
+ per_2_per, /* Peripheral to Peripheral */
+ int_2_per, /* Internal RAM to peripheral */
+ int_2_int, /* Internal RAM to Internal RAM */
+ int_2_emi, /* Internal RAM to EMI memory */
+ int_2_dsp, /* Internal RAM to DSP memory */
+ dsp_2_per, /* DSP memory to peripheral */
+ dsp_2_int, /* DSP memory to internal RAM */
+ dsp_2_emi, /* DSP memory to EMI memory */
+ dsp_2_dsp, /* DSP memory to DSP memory */
+ emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
+ dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
+ dvfs_pll, /* DVFS script with PLL change */
+ dvfs_pdr /* DVFS script without PLL change */
+} sdma_transfer_type;
+
+/**
+ * struct sdma_script_start_addrs - SDMA script start pointers
+ *
+ * start addresses of the different functions in the physical
+ * address space of the SDMA engine.
+ */
+struct sdma_script_start_addrs {
+ u32 ap_2_ap_addr;
+ u32 ap_2_bp_addr;
+ u32 ap_2_ap_fixed_addr;
+ u32 bp_2_ap_addr;
+ u32 loopback_on_dsp_side_addr;
+ u32 mcu_interrupt_only_addr;
+ u32 firi_2_per_addr;
+ u32 firi_2_mcu_addr;
+ u32 per_2_firi_addr;
+ u32 mcu_2_firi_addr;
+ u32 uart_2_per_addr;
+ u32 uart_2_mcu_addr;
+ u32 per_2_app_addr;
+ u32 mcu_2_app_addr;
+ u32 per_2_per_addr;
+ u32 uartsh_2_per_addr;
+ u32 uartsh_2_mcu_addr;
+ u32 per_2_shp_addr;
+ u32 mcu_2_shp_addr;
+ u32 ata_2_mcu_addr;
+ u32 mcu_2_ata_addr;
+ u32 app_2_per_addr;
+ u32 app_2_mcu_addr;
+ u32 shp_2_per_addr;
+ u32 shp_2_mcu_addr;
+ u32 mshc_2_mcu_addr;
+ u32 mcu_2_mshc_addr;
+ u32 spdif_2_mcu_addr;
+ u32 mcu_2_spdif_addr;
+ u32 asrc_2_mcu_addr;
+ u32 ext_mem_2_ipu_addr;
+ u32 descrambler_addr;
+ u32 dptc_dvfs_addr;
+ u32 utra_addr;
+ u32 ram_code_start_addr;
+};
+
+#define SDMA_FIRMWARE_MAGIC 0x414d4453
+
+/**
+ * struct sdma_firmware_header - Layout of the firmware image
+ *
+ * @magic "SDMA"
+ * @version_major increased whenever layout of struct sdma_script_start_addrs
+ * changes.
+ * @version_minor firmware minor version (for binary compatible changes)
+ * @script_addrs_start offset of struct sdma_script_start_addrs in this image
+ * @num_script_addrs Number of script addresses in this image
+ * @ram_code_start offset of SDMA ram image in this firmware image
+ * @ram_code_size size of SDMA ram image
+ */
+struct sdma_firmware_header {
+ u32 magic;
+ u32 version_major;
+ u32 version_minor;
+ u32 script_addrs_start;
+ u32 num_script_addrs;
+ u32 ram_code_start;
+ u32 ram_code_size;
+};
+
+struct sdma_engine {
+ struct device *dev;
+ struct sdma_channel channel[MAX_DMA_CHANNELS];
+ struct sdma_channel_control *channel_control;
+ void __iomem *regs;
+ unsigned int version;
+ unsigned int num_events;
+ struct sdma_context_data *context;
+ dma_addr_t context_phys;
+ struct dma_device dma_device;
+ struct clk *clk;
+};
+
+#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
+#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
+#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
+#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
+
+static int sdma_config_ownership(struct sdma_channel *sdmac,
+ bool event_override, bool mcu_verride, bool dsp_override)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 evt, mcu, dsp;
+
+ if (event_override && mcu_verride && dsp_override)
+ return -EINVAL;
+
+ evt = readl(SDMA_H_EVTOVR);
+ mcu = readl(SDMA_H_HOSTOVR);
+ dsp = readl(SDMA_H_DSPOVR);
+
+ if (dsp_override)
+ dsp &= ~(1 << channel);
+ else
+ dsp |= (1 << channel);
+
+ if (event_override)
+ evt &= ~(1 << channel);
+ else
+ evt |= (1 << channel);
+
+ if (mcu_verride)
+ mcu &= ~(1 << channel);
+ else
+ mcu |= (1 << channel);
+
+ writel(evt, SDMA_H_EVTOVR);
+ writel(mcu, SDMA_H_HOSTOVR);
+ writel(dsp, SDMA_H_DSPOVR);
+
+ return 0;
+}
+
+/*
+ * sdma_run_channel - run a channel and wait till it's done
+ */
+static int sdma_run_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret;
+
+ init_completion(&sdmac->done);
+
+ writel(1 << channel, SDMA_H_START);
+
+ ret = wait_for_completion_timeout(&sdmac->done, HZ);
+
+ return ret ? 0 : -ETIMEDOUT;
+}
+
+static int sdma_load_script(struct sdma_engine *sdma, void *buf, int size,
+ u32 address)
+{
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ void *buf_virt;
+ dma_addr_t buf_phys;
+ int ret;
+
+ buf_virt = dma_alloc_coherent(NULL,
+ size,
+ &buf_phys, GFP_KERNEL);
+ if (!buf_virt)
+ return -ENOMEM;
+
+ bd0->mode.command = C0_SETPM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = size / 2;
+ bd0->buffer_addr = buf_phys;
+ bd0->ext_buffer_addr = address;
+
+ memcpy(buf_virt, buf, size);
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ dma_free_coherent(NULL, size, buf_virt, buf_phys);
+
+ return ret;
+}
+
+static void sdma_event_enable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 val;
+
+ val = readl(SDMA_CHNENBL_0 + event * 4);
+ val |= (1 << channel);
+ writel(val, SDMA_CHNENBL_0 + event * 4);
+}
+
+static void sdma_event_disable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 val;
+
+ val = readl(SDMA_CHNENBL_0 + event * 4);
+ val &= ~(1 << channel);
+ writel(val, SDMA_CHNENBL_0 + event * 4);
+}
+
+static void sdma_handle_channel_loop(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+
+ /*
+ * loop mode. Iterate over descriptors, re-setup them and
+ * call callback function.
+ */
+ while (1) {
+ bd = &sdmac->bd[sdmac->buf_tail];
+
+ if (bd->mode.status & BD_DONE)
+ break;
+
+ if (bd->mode.status & BD_RROR)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ bd->mode.status |= BD_DONE;
+ sdmac->buf_tail++;
+ sdmac->buf_tail %= sdmac->num_bd;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ }
+}
+
+static void mxc_sdma_handle_channel_normal(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+ int i, error = 0;
+
+ /*
+ * non loop mode. Iterate over all descriptors, collect
+ * errors and call callback function
+ */
+ for (i = 0; i < sdmac->num_bd; i++) {
+ bd = &sdmac->bd[i];
+
+ if (bd->mode.status & (BD_DONE | BD_RROR))
+ error = -EIO;
+ }
+
+ if (error)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ sdmac->last_completed = sdmac->desc.cookie;
+}
+
+static void mxc_sdma_handle_channel(struct sdma_channel *sdmac)
+{
+ complete(&sdmac->done);
+
+ /* not interested in channel 0 interrupts */
+ if (sdmac->channel == 0)
+ return;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP)
+ sdma_handle_channel_loop(sdmac);
+ else
+ mxc_sdma_handle_channel_normal(sdmac);
+}
+
+static irqreturn_t sdma_int_handler(int irq, void *dev_id)
+{
+ struct sdma_engine *sdma = dev_id;
+ u32 stat;
+
+ stat = readl(SDMA_H_INTR);
+ writel(stat, SDMA_H_INTR);
+
+ while (stat) {
+ int channel = fls(stat) - 1;
+ struct sdma_channel *sdmac = &sdma->channel[channel];
+
+ mxc_sdma_handle_channel(sdmac);
+
+ stat &= ~(1 << channel);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * Stores the start address of the SDMA scripts
+ */
+static struct sdma_script_start_addrs __sdma_script_addrs;
+static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
+
+/*
+ * sets the pc of SDMA script according to the peripheral type
+ */
+static void sdma_get_pc(struct sdma_channel *sdmac,
+ enum sdma_peripheral_type peripheral_type)
+{
+ int per_2_emi = 0, emi_2_per = 0;
+ /*
+ * These are needed once we start to support transfers between
+ * two peripherals or memory-to-memory transfers
+ */
+ int per_2_per = 0, emi_2_emi = 0;
+
+ sdmac->pc_from_device = 0;
+ sdmac->pc_to_device = 0;
+
+ switch (peripheral_type) {
+ case IMX_DMATYPE_MEMORY:
+ emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
+ break;
+ case IMX_DMATYPE_DSP:
+ emi_2_per = sdma_script_addrs->bp_2_ap_addr;
+ per_2_emi = sdma_script_addrs->ap_2_bp_addr;
+ break;
+ case IMX_DMATYPE_FIRI:
+ per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
+ break;
+ case IMX_DMATYPE_UART:
+ per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_UART_SP:
+ per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ATA:
+ per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
+ break;
+ case IMX_DMATYPE_CSPI:
+ case IMX_DMATYPE_EXT:
+ case IMX_DMATYPE_SSI:
+ per_2_emi = sdma_script_addrs->app_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_SSI_SP:
+ case IMX_DMATYPE_MMC:
+ case IMX_DMATYPE_SDHC:
+ case IMX_DMATYPE_CSPI_SP:
+ case IMX_DMATYPE_ESAI:
+ case IMX_DMATYPE_MSHC_SP:
+ per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ASRC:
+ per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
+ per_2_per = sdma_script_addrs->per_2_per_addr;
+ break;
+ case IMX_DMATYPE_MSHC:
+ per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
+ break;
+ case IMX_DMATYPE_CCM:
+ per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
+ break;
+ case IMX_DMATYPE_SPDIF:
+ per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
+ break;
+ case IMX_DMATYPE_IPU_MEMORY:
+ emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
+ break;
+ default:
+ break;
+ }
+
+ sdmac->pc_from_device = per_2_emi;
+ sdmac->pc_to_device = emi_2_per;
+}
+
+static int sdma_load_context(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int load_address;
+ struct sdma_context_data *context = sdma->context;
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ int ret;
+
+ if (sdmac->direction == DMA_FROM_DEVICE) {
+ load_address = sdmac->pc_from_device;
+ } else {
+ load_address = sdmac->pc_to_device;
+ }
+
+ if (load_address < 0)
+ return load_address;
+
+ dev_dbg(sdma->dev, "load_address = %d\n", load_address);
+ dev_dbg(sdma->dev, "wml = 0x%08x\n", sdmac->watermark_level);
+ dev_dbg(sdma->dev, "shp_addr = 0x%08x\n", sdmac->shp_addr);
+ dev_dbg(sdma->dev, "per_addr = 0x%08x\n", sdmac->per_addr);
+ dev_dbg(sdma->dev, "event_mask1 = 0x%08x\n", sdmac->event_mask1);
+ dev_dbg(sdma->dev, "event_mask2 = 0x%08x\n", sdmac->event_mask2);
+
+ memset(context, 0, sizeof(*context));
+ context->channel_state.pc = load_address;
+
+ /* Send by context the event mask,base address for peripheral
+ * and watermark level
+ */
+ context->gReg[0] = sdmac->event_mask2;
+ context->gReg[1] = sdmac->event_mask1;
+ context->gReg[2] = sdmac->per_addr;
+ context->gReg[6] = sdmac->shp_addr;
+ context->gReg[7] = sdmac->watermark_level;
+
+ bd0->mode.command = C0_SETDM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = sizeof(*context) / 4;
+ bd0->buffer_addr = sdma->context_phys;
+ bd0->ext_buffer_addr = 2048 + (sizeof(*context) / 4) * channel;
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ return ret;
+}
+
+static void sdma_disable_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ writel(1 << channel, SDMA_H_STATSTOP);
+ sdmac->status = DMA_ERROR;
+}
+
+static int sdma_config_channel(struct sdma_channel *sdmac)
+{
+ int ret;
+
+ sdma_disable_channel(sdmac);
+
+ sdmac->event_mask1 = 0;
+ sdmac->event_mask2 = 0;
+ sdmac->shp_addr = 0;
+ sdmac->per_addr = 0;
+
+ if (sdmac->event_id) {
+ if (sdmac->event_id > 32)
+ return -EINVAL;
+ sdma_event_enable(sdmac, sdmac->event_id);
+ }
+
+ switch (sdmac->peripheral_type) {
+ case IMX_DMATYPE_DSP:
+ sdma_config_ownership(sdmac, false, true, true);
+ break;
+ case IMX_DMATYPE_MEMORY:
+ sdma_config_ownership(sdmac, false, true, false);
+ break;
+ default:
+ sdma_config_ownership(sdmac, true, true, false);
+ break;
+ }
+
+ sdma_get_pc(sdmac, sdmac->peripheral_type);
+
+ if ((sdmac->peripheral_type != IMX_DMATYPE_MEMORY) &&
+ (sdmac->peripheral_type != IMX_DMATYPE_DSP)) {
+ /* Handle multiple event channels differently */
+ if (sdmac->event_id2) {
+ sdmac->event_mask2 = 1 << (sdmac->event_id2 % 32);
+ if (sdmac->event_id2 > 31)
+ sdmac->watermark_level |= 1 << 31;
+ sdmac->event_mask1 = 1 << (sdmac->event_id % 32);
+ if (sdmac->event_id > 31)
+ sdmac->watermark_level |= 1 << 30;
+ } else {
+ sdmac->event_mask1 = 1 << sdmac->event_id;
+ sdmac->event_mask2 = 1 << (sdmac->event_id - 32);
+ }
+ /* Watermark Level */
+ sdmac->watermark_level |= sdmac->watermark_level;
+ /* Address */
+ sdmac->shp_addr = sdmac->per_address;
+ } else {
+ sdmac->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
+ }
+
+ ret = sdma_load_context(sdmac);
+
+ return ret;
+}
+
+static int sdma_set_channel_priority(struct sdma_channel *sdmac,
+ unsigned int priority)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ if (priority < MXC_SDMA_MIN_PRIORITY
+ || priority > MXC_SDMA_MAX_PRIORITY) {
+ return -EINVAL;
+ }
+
+ writel(priority, SDMA_CHNPRI_0 + 4 * channel);
+
+ return 0;
+}
+
+static int sdma_request_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret = -EBUSY;
+
+ sdmac->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdmac->bd_phys, GFP_KERNEL);
+ if (!sdmac->bd) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memset(sdmac->bd, 0, PAGE_SIZE);
+
+ sdma->channel_control[channel].base_bd_ptr = sdmac->bd_phys;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ clk_enable(sdma->clk);
+
+ sdma_set_channel_priority(sdmac, MXC_SDMA_DEFAULT_PRIORITY);
+
+ init_completion(&sdmac->done);
+
+ sdmac->buf_tail = 0;
+
+ return 0;
+out:
+
+ return ret;
+}
+
+static void sdma_enable_channel(struct sdma_engine *sdma, int channel)
+{
+ writel(1 << channel, SDMA_H_START);
+}
+
+static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
+{
+ dma_cookie_t cookie = sdma->chan.cookie;
+
+ if (++cookie < 0)
+ cookie = 1;
+
+ sdma->chan.cookie = cookie;
+ sdma->desc.cookie = cookie;
+
+ return cookie;
+}
+
+static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sdma_channel, chan);
+}
+
+static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(tx->chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ dma_cookie_t cookie;
+
+ spin_lock_irq(&sdmac->lock);
+
+ cookie = sdma_assign_cookie(sdmac);
+
+ sdma_enable_channel(sdma, tx->chan->chan_id);
+
+ spin_unlock_irq(&sdmac->lock);
+
+ return cookie;
+}
+
+static int sdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct imx_dma_data *data = chan->private;
+ int prio, ret;
+
+ /* No need to execute this for internal channel 0 */
+ if (chan->chan_id == 0)
+ return 0;
+
+ if (!data)
+ return -EINVAL;
+
+ switch (data->priority) {
+ case DMA_PRIO_HIGH:
+ prio = 3;
+ break;
+ case DMA_PRIO_MEDIUM:
+ prio = 2;
+ break;
+ case DMA_PRIO_LOW:
+ default:
+ prio = 1;
+ break;
+ }
+
+ sdmac->peripheral_type = data->peripheral_type;
+ sdmac->event_id = data->dma_request;
+ ret = sdma_set_channel_priority(sdmac, prio);
+ if (ret)
+ return ret;
+
+ ret = sdma_request_channel(sdmac);
+ if (ret)
+ return ret;
+
+ dma_async_tx_descriptor_init(&sdmac->desc, chan);
+ sdmac->desc.tx_submit = sdma_tx_submit;
+ /* txd.flags will be overwritten in prep funcs */
+ sdmac->desc.flags = DMA_CTRL_ACK;
+
+ return 0;
+}
+
+static void sdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+
+ sdma_disable_channel(sdmac);
+
+ if (sdmac->event_id)
+ sdma_event_disable(sdmac, sdmac->event_id);
+ if (sdmac->event_id2)
+ sdma_event_disable(sdmac, sdmac->event_id2);
+
+ sdmac->event_id = 0;
+ sdmac->event_id2 = 0;
+
+ sdma_set_channel_priority(sdmac, 0);
+
+ dma_free_coherent(NULL, PAGE_SIZE, sdmac->bd, sdmac->bd_phys);
+
+ clk_disable(sdma->clk);
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int ret, i, count;
+ int channel = chan->chan_id;
+ struct scatterlist *sg;
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags = 0;
+
+ dev_dbg(sdma->dev, "setting up %d entries for channel %d.\n",
+ sg_len, channel);
+
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (sg_len > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, sg_len, NUM_BD);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = sgl->dma_address;
+
+ count = sg->length;
+
+ if (count > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
+ channel, count, 0xffff);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ bd->mode.count = count;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES) {
+ ret = -EINVAL;
+ goto err_out;
+ }
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP) {
+ param |= BD_INTR;
+ if (i + 1 == sg_len)
+ param |= BD_WRAP;
+ }
+
+ if (i + 1 == sg_len)
+ param |= BD_INTR;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, count, sg->dma_address,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+ }
+
+ sdmac->num_bd = sg_len;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int num_periods = buf_len / period_len;
+ int channel = chan->chan_id;
+ int ret, i = 0, buf = 0;
+
+ dev_dbg(sdma->dev, "%s channel: %d\n", __func__, channel);
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags |= IMX_DMA_SG_LOOP;
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (num_periods > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, num_periods, NUM_BD);
+ goto err_out;
+ }
+
+ if (period_len > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum period size exceeded: %d > %d\n",
+ channel, period_len, 0xffff);
+ goto err_out;
+ }
+
+ while (buf < buf_len) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = dma_addr;
+
+ bd->mode.count = period_len;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES)
+ goto err_out;
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
+ if (i + 1 == num_periods)
+ param |= BD_WRAP;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, period_len, dma_addr,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+
+ dma_addr += period_len;
+ buf += period_len;
+
+ i++;
+ }
+
+ sdmac->num_bd = num_periods;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ sdmac->status = DMA_ERROR;
+ return NULL;
+}
+
+static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct dma_slave_config *dmaengine_cfg = (void *)arg;
+
+ switch (cmd) {
+ case DMA_TERMINATE_ALL:
+ sdma_disable_channel(sdmac);
+ return 0;
+ case DMA_SLAVE_CONFIG:
+ if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
+ sdmac->per_address = dmaengine_cfg->src_addr;
+ sdmac->watermark_level = dmaengine_cfg->src_maxburst;
+ sdmac->word_size = dmaengine_cfg->src_addr_width;
+ } else {
+ sdmac->per_address = dmaengine_cfg->dst_addr;
+ sdmac->watermark_level = dmaengine_cfg->dst_maxburst;
+ sdmac->word_size = dmaengine_cfg->dst_addr_width;
+ }
+ return sdma_config_channel(sdmac);
+ default:
+ return -ENOSYS;
+ }
+
+ return -EINVAL;
+}
+
+static enum dma_status sdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ dma_cookie_t last_used;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+
+ ret = dma_async_is_complete(cookie, sdmac->last_completed, last_used);
+ dma_set_tx_state(txstate, sdmac->last_completed, last_used, 0);
+
+ return ret;
+}
+
+static void sdma_issue_pending(struct dma_chan *chan)
+{
+ /*
+ * Nothing to do. We only have a single descriptor
+ */
+}
+
+static int __init sdma_init(struct sdma_engine *sdma,
+ void *ram_code, int ram_code_size)
+{
+ int i, ret;
+ dma_addr_t ccb_phys;
+
+ switch (sdma->version) {
+ case 1:
+ sdma->num_events = 32;
+ break;
+ case 2:
+ sdma->num_events = 48;
+ break;
+ default:
+ dev_err(sdma->dev, "Unknown version %d. aborting\n", sdma->version);
+ return -ENODEV;
+ }
+
+ clk_enable(sdma->clk);
+
+ /* Be sure SDMA has not started yet */
+ writel(0, SDMA_H_C0PTR);
+
+ sdma->channel_control = dma_alloc_coherent(NULL,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
+ sizeof(struct sdma_context_data),
+ &ccb_phys, GFP_KERNEL);
+
+ if (!sdma->channel_control) {
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ sdma->context = (void *)sdma->channel_control +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+ sdma->context_phys = ccb_phys +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+
+ /* Zero-out the CCB structures array just allocated */
+ memset(sdma->channel_control, 0,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
+
+ /* disable all channels */
+ for (i = 0; i < sdma->num_events; i++)
+ writel(0, SDMA_CHNENBL_0 + i * 4);
+
+ /* All channels have priority 0 */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++)
+ writel(0, SDMA_CHNPRI_0 + i * 4);
+
+ ret = sdma_request_channel(&sdma->channel[0]);
+ if (ret)
+ goto err_dma_alloc;
+
+ sdma_config_ownership(&sdma->channel[0], false, true, false);
+
+ /* Set Command Channel (Channel Zero) */
+ writel(0x4050, SDMA_CHN0ADDR);
+
+ /* Set bits of CONFIG register but with static context switching */
+ /* FIXME: Check whether to set ACR bit depending on clock ratios */
+ writel(0, SDMA_H_CONFIG);
+
+ writel(ccb_phys, SDMA_H_C0PTR);
+
+ /* download the RAM image for SDMA */
+ sdma_load_script(sdma, ram_code,
+ ram_code_size,
+ sdma_script_addrs->ram_code_start_addr);
+
+ /* Set bits of CONFIG register with given context switching mode */
+ writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
+
+ /* Initializes channel's priorities */
+ sdma_set_channel_priority(&sdma->channel[0], 7);
+
+ clk_disable(sdma->clk);
+
+ return 0;
+
+err_dma_alloc:
+ clk_disable(sdma->clk);
+ dev_err(sdma->dev, "initialisation failed with %d\n", ret);
+ return ret;
+}
+
+static int __devinit sdma_probe(struct platform_device *pdev)
+{
+ int ret;
+ const struct firmware *fw;
+ const struct sdma_firmware_header *header;
+ const struct sdma_script_start_addrs *addr;
+ int irq;
+ unsigned short *ram_code;
+ struct resource *iores;
+ struct sdma_platform_data *pdata = pdev->dev.platform_data;
+ char *fwname;
+ int i;
+ dma_cap_mask_t mask;
+ struct sdma_engine *sdma;
+
+ sdma = kzalloc(sizeof(*sdma), GFP_KERNEL);
+ if (!sdma)
+ return -ENOMEM;
+
+ sdma->dev = &pdev->dev;
+
+ iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ irq = platform_get_irq(pdev, 0);
+ if (!iores || irq < 0 || !pdata) {
+ ret = -EINVAL;
+ goto err_irq;
+ }
+
+ sdma->clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(sdma->clk)) {
+ ret = PTR_ERR(sdma->clk);
+ goto err_clk;
+ }
+
+ sdma->regs = ioremap(iores->start, resource_size(iores));
+ if (!sdma->regs) {
+ ret = -ENOMEM;
+ goto err_ioremap;
+ }
+
+ ret = request_irq(irq, sdma_int_handler, 0, "sdma", sdma);
+ if (ret)
+ goto err_request_irq;
+
+ fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin",
+ pdata->cpu_name, pdata->to_version);
+ if (!fwname) {
+ ret = -ENOMEM;
+ goto err_cputype;
+ }
+
+ ret = request_firmware(&fw, fwname, &pdev->dev);
+ if (ret) {
+ dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
+ fwname, ret);
+ kfree(fwname);
+ goto err_cputype;
+ }
+ kfree(fwname);
+
+ if (fw->size < sizeof(*header))
+ goto err_firmware;
+
+ header = (struct sdma_firmware_header *)fw->data;
+
+ if (header->magic != SDMA_FIRMWARE_MAGIC)
+ goto err_firmware;
+ if (header->ram_code_start + header->ram_code_size > fw->size)
+ goto err_firmware;
+
+ addr = (void *)header + header->script_addrs_start;
+ ram_code = (void *)header + header->ram_code_start;
+ memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
+
+ sdma->version = pdata->sdma_version;
+
+ INIT_LIST_HEAD(&sdma->dma_device.channels);
+ /* Initialize channel parameters */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++) {
+ struct sdma_channel *sdmac = &sdma->channel[i];
+
+ sdmac->sdma = sdma;
+ spin_lock_init(&sdmac->lock);
+
+ dma_cap_set(DMA_SLAVE, sdma->dma_device.cap_mask);
+ dma_cap_set(DMA_CYCLIC, sdma->dma_device.cap_mask);
+
+ sdmac->chan.device = &sdma->dma_device;
+ sdmac->chan.chan_id = i;
+ sdmac->channel = i;
+
+ /* Add the channel to the DMAC list */
+ list_add_tail(&sdmac->chan.device_node, &sdma->dma_device.channels);
+ }
+
+ ret = sdma_init(sdma, ram_code, header->ram_code_size);
+ if (ret)
+ goto err_firmware;
+
+ sdma->dma_device.dev = &pdev->dev;
+
+ sdma->dma_device.device_alloc_chan_resources = sdma_alloc_chan_resources;
+ sdma->dma_device.device_free_chan_resources = sdma_free_chan_resources;
+ sdma->dma_device.device_tx_status = sdma_tx_status;
+ sdma->dma_device.device_prep_slave_sg = sdma_prep_slave_sg;
+ sdma->dma_device.device_prep_dma_cyclic = sdma_prep_dma_cyclic;
+ sdma->dma_device.device_control = sdma_control;
+ sdma->dma_device.device_issue_pending = sdma_issue_pending;
+
+ ret = dma_async_device_register(&sdma->dma_device);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register\n");
+ goto err_firmware;
+ }
+
+ dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
+ header->version_major,
+ header->version_minor);
+
+ /* request channel 0. This is an internal control channel
+ * to the SDMA engine and not available to clients.
+ */
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_request_channel(mask, NULL, NULL);
+
+ release_firmware(fw);
+
+ return 0;
+
+err_firmware:
+ release_firmware(fw);
+err_cputype:
+ free_irq(irq, sdma);
+err_request_irq:
+ iounmap(sdma->regs);
+err_ioremap:
+ clk_put(sdma->clk);
+err_clk:
+err_irq:
+ kfree(sdma);
+ return 0;
+}
+
+static int __devexit sdma_remove(struct platform_device *pdev)
+{
+ return -EBUSY;
+}
+
+static struct platform_driver sdma_driver = {
+ .driver = {
+ .name = "imx-sdma",
+ },
+ .probe = sdma_probe,
+ .remove = __devexit_p(sdma_remove),
+};
+
+static int __init sdma_module_init(void)
+{
+ return platform_driver_register(&sdma_driver);
+}
+subsys_initcall(sdma_module_init);
+
+MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("i.MX SDMA driver");
+MODULE_LICENSE("GPL");
--
1.7.1
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-23 12:57 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-23 12:57 UTC (permalink / raw)
To: linux-arm-kernel
This patch adds support for the Freescale i.MX SDMA engine.
The SDMA engine is a scatter/gather DMA engine which is implemented
as a seperate coprocessor. SDMA needs its own firmware which is
requested using the standard request_firmware mechanism. The firmware
has different entry points for each peripheral type, so drivers
have to pass the peripheral type to the DMA engine which in turn
picks the correct firmware entry point from a table contained in
the firmware image itself.
The original Freescale code also supports support for transfering
data to the internal SRAM which needs different entry points to
the firmware. Support for this is currently not implemented. Also,
support for the ASRC (asymmetric sample rate converter) is skipped.
I took a very simple approach to implement dmaengine support. Only
a single descriptor is statically assigned to a each channel. This
means that transfers can't be queued up but only a single transfer
is in progress. This simplifies implementation a lot and is sufficient
for the usual device/memory transfers.
Changes since v1:
- included comments from Linus Walleij
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
arch/arm/mach-mx3/Kconfig | 2 +
arch/arm/plat-mxc/Kconfig | 10 +
arch/arm/plat-mxc/include/mach/dma.h | 64 ++
arch/arm/plat-mxc/include/mach/sdma.h | 17 +
drivers/dma/Kconfig | 8 +
drivers/dma/Makefile | 1 +
drivers/dma/imx-sdma.c | 1395 +++++++++++++++++++++++++++++++
8 files changed, 1499 insertions(+), 6 deletions(-)
create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
create mode 100644 drivers/dma/imx-sdma.c
diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
index 287431c..ac6fd71 100644
--- a/arch/arm/mach-imx/include/mach/dma-v1.h
+++ b/arch/arm/mach-imx/include/mach/dma-v1.h
@@ -27,6 +27,8 @@
#define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
+#include <mach/dma.h>
+
#define IMX_DMA_CHANNELS 16
#define DMA_MODE_READ 0
@@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
void imx_dma_free(int channel);
-enum imx_dma_prio {
- DMA_PRIO_HIGH = 0,
- DMA_PRIO_MEDIUM = 1,
- DMA_PRIO_LOW = 2
-};
-
int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
#endif /* __MACH_DMA_V1_H__ */
diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
index 85beece..301375c 100644
--- a/arch/arm/mach-mx3/Kconfig
+++ b/arch/arm/mach-mx3/Kconfig
@@ -3,12 +3,14 @@ if ARCH_MX3
config ARCH_MX31
select ARCH_HAS_RNGA
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
bool
config ARCH_MX35
bool
select ARCH_MXC_IOMUX_V3
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
comment "MX3 platforms:"
diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
index 0527e65..6741625 100644
--- a/arch/arm/plat-mxc/Kconfig
+++ b/arch/arm/plat-mxc/Kconfig
@@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
config ARCH_MXC_AUDMUX_V2
bool
+config IMX_HAVE_SDMA
+ bool
+
+config IMX_SDMA
+ depends on IMX_HAVE_SDMA
+ tristate "Enable SDMA support"
+ help
+ Include support for the SDMA engine. The SDMA engine needs additional
+ firmware support. SDMA can be compiled as a module to support loading
+ the firmware when a rootfs is present.
endif
diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
new file mode 100644
index 0000000..69d181f
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/dma.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARCH_MXC_DMA_H__
+#define __ASM_ARCH_MXC_DMA_H__
+
+#include <linux/scatterlist.h>
+
+/*
+ * This enumerates peripheral types. Used for SDMA.
+ */
+enum sdma_peripheral_type {
+ IMX_DMATYPE_SSI, /* MCU domain SSI */
+ IMX_DMATYPE_SSI_SP, /* Shared SSI */
+ IMX_DMATYPE_MMC, /* MMC */
+ IMX_DMATYPE_SDHC, /* SDHC */
+ IMX_DMATYPE_UART, /* MCU domain UART */
+ IMX_DMATYPE_UART_SP, /* Shared UART */
+ IMX_DMATYPE_FIRI, /* FIRI */
+ IMX_DMATYPE_CSPI, /* MCU domain CSPI */
+ IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
+ IMX_DMATYPE_SIM, /* SIM */
+ IMX_DMATYPE_ATA, /* ATA */
+ IMX_DMATYPE_CCM, /* CCM */
+ IMX_DMATYPE_EXT, /* External peripheral */
+ IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
+ IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
+ IMX_DMATYPE_DSP, /* DSP */
+ IMX_DMATYPE_MEMORY, /* Memory */
+ IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
+ IMX_DMATYPE_SPDIF, /* SPDIF */
+ IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
+ IMX_DMATYPE_ASRC, /* ASRC */
+ IMX_DMATYPE_ESAI, /* ESAI */
+};
+
+enum imx_dma_prio {
+ DMA_PRIO_HIGH = 0,
+ DMA_PRIO_MEDIUM = 1,
+ DMA_PRIO_LOW = 2
+};
+
+struct imx_dma_data {
+ int dma_request; /* DMA request line */
+ enum sdma_peripheral_type peripheral_type;
+ int priority;
+};
+
+static inline int imx_dma_is_ipu(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "ipu-core");
+}
+
+static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "imx-sdma");
+}
+
+#endif
diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
new file mode 100644
index 0000000..9be1122
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/sdma.h
@@ -0,0 +1,17 @@
+#ifndef __MACH_MXC_SDMA_H__
+#define __MACH_MXC_SDMA_H__
+
+/**
+ * struct sdma_platform_data - platform specific data for SDMA engine
+ *
+ * @sdma_version The version of this SDMA engine
+ * @cpu_name used to generate the firmware name
+ * @to_version CPU Tape out version
+ */
+struct sdma_platform_data {
+ int sdma_version;
+ char *cpu_name;
+ int to_version;
+};
+
+#endif /* __MACH_MXC_SDMA_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9520cf0..ff68307 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -195,6 +195,14 @@ config PCH_DMA
help
Enable support for the Topcliff PCH DMA engine.
+config IMX_SDMA
+ tristate "i.MX SDMA support"
+ depends on ARCH_MXC
+ select DMA_ENGINE
+ help
+ Support the i.MX SDMA engine. This engine is integrated into
+ Freescale i.MX25/31/35/51 chips.
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 72bd703..14d7a1b 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
obj-$(CONFIG_PCH_DMA) += pch_dma.o
+obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
new file mode 100644
index 0000000..c447fc0
--- /dev/null
+++ b/drivers/dma/imx-sdma.c
@@ -0,0 +1,1395 @@
+/*
+ * drivers/dma/imx-sdma.c
+ *
+ * This file contains a driver for the Freescale Smart DMA engine
+ *
+ * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
+ *
+ * Based on code from Freescale:
+ *
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * The code contained herein is licensed under the GNU General Public
+ * License. You may obtain a copy of the GNU General Public License
+ * Version 2 or later at the following locations:
+ *
+ * http://www.opensource.org/licenses/gpl-license.html
+ * http://www.gnu.org/copyleft/gpl.html
+ */
+
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/dmaengine.h>
+
+#include <asm/irq.h>
+#include <mach/sdma.h>
+#include <mach/dma.h>
+#include <mach/hardware.h>
+
+/* SDMA registers */
+#define SDMA_H_C0PTR (sdma->regs + 0x000)
+#define SDMA_H_INTR (sdma->regs + 0x004)
+#define SDMA_H_STATSTOP (sdma->regs + 0x008)
+#define SDMA_H_START (sdma->regs + 0x00c)
+#define SDMA_H_EVTOVR (sdma->regs + 0x010)
+#define SDMA_H_DSPOVR (sdma->regs + 0x014)
+#define SDMA_H_HOSTOVR (sdma->regs + 0x018)
+#define SDMA_H_EVTPEND (sdma->regs + 0x01c)
+#define SDMA_H_DSPENBL (sdma->regs + 0x020)
+#define SDMA_H_RESET (sdma->regs + 0x024)
+#define SDMA_H_EVTERR (sdma->regs + 0x028)
+#define SDMA_H_INTRMSK (sdma->regs + 0x02c)
+#define SDMA_H_PSW (sdma->regs + 0x030)
+#define SDMA_H_EVTERRDBG (sdma->regs + 0x034)
+#define SDMA_H_CONFIG (sdma->regs + 0x038)
+#define SDMA_ONCE_ENB (sdma->regs + 0x040)
+#define SDMA_ONCE_DATA (sdma->regs + 0x044)
+#define SDMA_ONCE_INSTR (sdma->regs + 0x048)
+#define SDMA_ONCE_STAT (sdma->regs + 0x04c)
+#define SDMA_ONCE_CMD (sdma->regs + 0x050)
+#define SDMA_EVT_MIRROR (sdma->regs + 0x054)
+#define SDMA_ILLINSTADDR (sdma->regs + 0x058)
+#define SDMA_CHN0ADDR (sdma->regs + 0x05c)
+#define SDMA_ONCE_RTB (sdma->regs + 0x060)
+#define SDMA_XTRIG_CONF1 (sdma->regs + 0x070)
+#define SDMA_XTRIG_CONF2 (sdma->regs + 0x074)
+#define SDMA_CHNENBL_0 (sdma->regs + (sdma->version == 2 ? 0x200 : 0x80))
+#define SDMA_CHNPRI_0 (sdma->regs + 0x100)
+
+/*
+ * Buffer descriptor status values.
+ */
+#define BD_DONE 0x01
+#define BD_WRAP 0x02
+#define BD_CONT 0x04
+#define BD_INTR 0x08
+#define BD_RROR 0x10
+#define BD_LAST 0x20
+#define BD_EXTD 0x80
+
+/*
+ * Data Node descriptor status values.
+ */
+#define DND_END_OF_FRAME 0x80
+#define DND_END_OF_XFER 0x40
+#define DND_DONE 0x20
+#define DND_UNUSED 0x01
+
+/*
+ * IPCV2 descriptor status values.
+ */
+#define BD_IPCV2_END_OF_FRAME 0x40
+
+#define IPCV2_MAX_NODES 50
+/*
+ * Error bit set in the CCB status field by the SDMA,
+ * in setbd routine, in case of a transfer error
+ */
+#define DATA_ERROR 0x10000000
+
+/*
+ * Buffer descriptor commands.
+ */
+#define C0_ADDR 0x01
+#define C0_LOAD 0x02
+#define C0_DUMP 0x03
+#define C0_SETCTX 0x07
+#define C0_GETCTX 0x03
+#define C0_SETDM 0x01
+#define C0_SETPM 0x04
+#define C0_GETDM 0x02
+#define C0_GETPM 0x08
+/*
+ * Change endianness indicator in the BD command field
+ */
+#define CHANGE_ENDIANNESS 0x80
+
+/*
+ * Mode/Count of data node descriptors - IPCv2
+ */
+struct sdma_mode_count {
+ u32 count : 16; /* size of the buffer pointed by this BD */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 command : 8; /* command mostlky used for channel 0 */
+};
+
+/*
+ * Buffer descriptor
+ */
+struct sdma_buffer_descriptor {
+ struct sdma_mode_count mode;
+ dma_addr_t buffer_addr; /* address of the buffer described */
+ dma_addr_t ext_buffer_addr; /* extended buffer address */
+} __attribute__ ((packed));
+
+/*
+ * Channel control Block
+ */
+struct sdma_channel_control {
+ dma_addr_t current_bd_ptr; /* current buffer descriptor processed */
+ dma_addr_t base_bd_ptr; /* first element of buffer descriptor array */
+ u32 unused;
+ u32 unused1;
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_state_registers - SDMA context for a channel
+ *
+ * @pc: program counter
+ * @t: test bit: status of arithmetic & test instruction
+ * @rpc: return program counter
+ * @sf: source fault while loading data
+ * @spc: loop start program counter
+ * @df: destination fault while storing data
+ * @epc: loop end program counter
+ * @lm: loop mode
+ */
+struct sdma_state_registers {
+ u32 pc :14;
+ u32 unused1: 1;
+ u32 t : 1;
+ u32 rpc :14;
+ u32 unused0: 1;
+ u32 sf : 1;
+ u32 spc :14;
+ u32 unused2: 1;
+ u32 df : 1;
+ u32 epc :14;
+ u32 lm : 2;
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_context_data - sdma context specific to a channel
+ *
+ * @channel_state: channel state bits
+ * @gReg: general registers
+ * @mda: burst dma destination address register
+ * @msa: burst dma source address register
+ * @ms: burst dma status register
+ * @md: burst dma data register
+ * @pda: peripheral dma destination address register
+ * @psa: peripheral dma source address register
+ * @ps: peripheral dma status register
+ * @pd: peripheral dma data register
+ * @ca: CRC polynomial register
+ * @cs: CRC accumulator register
+ * @dda: dedicated core destination address register
+ * @dsa: dedicated core source address register
+ * @ds: dedicated core status register
+ * @dd: dedicated core data register
+ */
+struct sdma_context_data {
+ struct sdma_state_registers channel_state;
+ u32 gReg[8];
+ u32 mda;
+ u32 msa;
+ u32 ms;
+ u32 md;
+ u32 pda;
+ u32 psa;
+ u32 ps;
+ u32 pd;
+ u32 ca;
+ u32 cs;
+ u32 dda;
+ u32 dsa;
+ u32 ds;
+ u32 dd;
+ u32 scratch0;
+ u32 scratch1;
+ u32 scratch2;
+ u32 scratch3;
+ u32 scratch4;
+ u32 scratch5;
+ u32 scratch6;
+ u32 scratch7;
+} __attribute__ ((packed));
+
+#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
+
+struct sdma_engine;
+
+/**
+ * struct sdma_channel - housekeeping for a SDMA channel
+ *
+ * @sdma pointer to the SDMA engine for this channel
+ * @channel the channel number, matches dmaengine chan_id
+ * @direction transfer type. Needed for setting SDMA script
+ * @peripheral_type Peripheral type. Needed for setting SDMA script
+ * @event_id aka dma request line
+ * @event_id2 for channels that use 2 events
+ * @word_size peripheral access size
+ * @buf_tail ID of the buffer that was processed
+ * @done channel completion
+ * @num_bd max NUM_BD. number of descriptors currently handling
+ */
+struct sdma_channel {
+ struct sdma_engine *sdma;
+ unsigned int channel;
+ enum dma_data_direction direction;
+ enum sdma_peripheral_type peripheral_type;
+ unsigned int event_id;
+ unsigned int event_id2;
+ enum dma_slave_buswidth word_size;
+ unsigned int buf_tail;
+ struct completion done;
+ unsigned int num_bd;
+ struct sdma_buffer_descriptor *bd;
+ dma_addr_t bd_phys;
+ unsigned int pc_from_device, pc_to_device;
+ unsigned long flags;
+ dma_addr_t per_address;
+ u32 event_mask1, event_mask2;
+ u32 watermark_level;
+ u32 shp_addr, per_addr;
+ struct dma_chan chan;
+ spinlock_t lock;
+ struct dma_async_tx_descriptor desc;
+ dma_cookie_t last_completed;
+ enum dma_status status;
+};
+
+#define IMX_DMA_SG_LOOP (1 << 0)
+
+#define MAX_DMA_CHANNELS 32
+#define MXC_SDMA_DEFAULT_PRIORITY 1
+#define MXC_SDMA_MIN_PRIORITY 1
+#define MXC_SDMA_MAX_PRIORITY 7
+
+/*
+ * This enumerates transfer types
+ */
+enum {
+ emi_2_per = 0, /* EMI memory to peripheral */
+ emi_2_int, /* EMI memory to internal RAM */
+ emi_2_emi, /* EMI memory to EMI memory */
+ emi_2_dsp, /* EMI memory to DSP memory */
+ per_2_int, /* Peripheral to internal RAM */
+ per_2_emi, /* Peripheral to internal EMI memory */
+ per_2_dsp, /* Peripheral to DSP memory */
+ per_2_per, /* Peripheral to Peripheral */
+ int_2_per, /* Internal RAM to peripheral */
+ int_2_int, /* Internal RAM to Internal RAM */
+ int_2_emi, /* Internal RAM to EMI memory */
+ int_2_dsp, /* Internal RAM to DSP memory */
+ dsp_2_per, /* DSP memory to peripheral */
+ dsp_2_int, /* DSP memory to internal RAM */
+ dsp_2_emi, /* DSP memory to EMI memory */
+ dsp_2_dsp, /* DSP memory to DSP memory */
+ emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
+ dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
+ dvfs_pll, /* DVFS script with PLL change */
+ dvfs_pdr /* DVFS script without PLL change */
+} sdma_transfer_type;
+
+/**
+ * struct sdma_script_start_addrs - SDMA script start pointers
+ *
+ * start addresses of the different functions in the physical
+ * address space of the SDMA engine.
+ */
+struct sdma_script_start_addrs {
+ u32 ap_2_ap_addr;
+ u32 ap_2_bp_addr;
+ u32 ap_2_ap_fixed_addr;
+ u32 bp_2_ap_addr;
+ u32 loopback_on_dsp_side_addr;
+ u32 mcu_interrupt_only_addr;
+ u32 firi_2_per_addr;
+ u32 firi_2_mcu_addr;
+ u32 per_2_firi_addr;
+ u32 mcu_2_firi_addr;
+ u32 uart_2_per_addr;
+ u32 uart_2_mcu_addr;
+ u32 per_2_app_addr;
+ u32 mcu_2_app_addr;
+ u32 per_2_per_addr;
+ u32 uartsh_2_per_addr;
+ u32 uartsh_2_mcu_addr;
+ u32 per_2_shp_addr;
+ u32 mcu_2_shp_addr;
+ u32 ata_2_mcu_addr;
+ u32 mcu_2_ata_addr;
+ u32 app_2_per_addr;
+ u32 app_2_mcu_addr;
+ u32 shp_2_per_addr;
+ u32 shp_2_mcu_addr;
+ u32 mshc_2_mcu_addr;
+ u32 mcu_2_mshc_addr;
+ u32 spdif_2_mcu_addr;
+ u32 mcu_2_spdif_addr;
+ u32 asrc_2_mcu_addr;
+ u32 ext_mem_2_ipu_addr;
+ u32 descrambler_addr;
+ u32 dptc_dvfs_addr;
+ u32 utra_addr;
+ u32 ram_code_start_addr;
+};
+
+#define SDMA_FIRMWARE_MAGIC 0x414d4453
+
+/**
+ * struct sdma_firmware_header - Layout of the firmware image
+ *
+ * @magic "SDMA"
+ * @version_major increased whenever layout of struct sdma_script_start_addrs
+ * changes.
+ * @version_minor firmware minor version (for binary compatible changes)
+ * @script_addrs_start offset of struct sdma_script_start_addrs in this image
+ * @num_script_addrs Number of script addresses in this image
+ * @ram_code_start offset of SDMA ram image in this firmware image
+ * @ram_code_size size of SDMA ram image
+ */
+struct sdma_firmware_header {
+ u32 magic;
+ u32 version_major;
+ u32 version_minor;
+ u32 script_addrs_start;
+ u32 num_script_addrs;
+ u32 ram_code_start;
+ u32 ram_code_size;
+};
+
+struct sdma_engine {
+ struct device *dev;
+ struct sdma_channel channel[MAX_DMA_CHANNELS];
+ struct sdma_channel_control *channel_control;
+ void __iomem *regs;
+ unsigned int version;
+ unsigned int num_events;
+ struct sdma_context_data *context;
+ dma_addr_t context_phys;
+ struct dma_device dma_device;
+ struct clk *clk;
+};
+
+#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
+#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
+#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
+#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
+
+static int sdma_config_ownership(struct sdma_channel *sdmac,
+ bool event_override, bool mcu_verride, bool dsp_override)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 evt, mcu, dsp;
+
+ if (event_override && mcu_verride && dsp_override)
+ return -EINVAL;
+
+ evt = readl(SDMA_H_EVTOVR);
+ mcu = readl(SDMA_H_HOSTOVR);
+ dsp = readl(SDMA_H_DSPOVR);
+
+ if (dsp_override)
+ dsp &= ~(1 << channel);
+ else
+ dsp |= (1 << channel);
+
+ if (event_override)
+ evt &= ~(1 << channel);
+ else
+ evt |= (1 << channel);
+
+ if (mcu_verride)
+ mcu &= ~(1 << channel);
+ else
+ mcu |= (1 << channel);
+
+ writel(evt, SDMA_H_EVTOVR);
+ writel(mcu, SDMA_H_HOSTOVR);
+ writel(dsp, SDMA_H_DSPOVR);
+
+ return 0;
+}
+
+/*
+ * sdma_run_channel - run a channel and wait till it's done
+ */
+static int sdma_run_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret;
+
+ init_completion(&sdmac->done);
+
+ writel(1 << channel, SDMA_H_START);
+
+ ret = wait_for_completion_timeout(&sdmac->done, HZ);
+
+ return ret ? 0 : -ETIMEDOUT;
+}
+
+static int sdma_load_script(struct sdma_engine *sdma, void *buf, int size,
+ u32 address)
+{
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ void *buf_virt;
+ dma_addr_t buf_phys;
+ int ret;
+
+ buf_virt = dma_alloc_coherent(NULL,
+ size,
+ &buf_phys, GFP_KERNEL);
+ if (!buf_virt)
+ return -ENOMEM;
+
+ bd0->mode.command = C0_SETPM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = size / 2;
+ bd0->buffer_addr = buf_phys;
+ bd0->ext_buffer_addr = address;
+
+ memcpy(buf_virt, buf, size);
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ dma_free_coherent(NULL, size, buf_virt, buf_phys);
+
+ return ret;
+}
+
+static void sdma_event_enable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 val;
+
+ val = readl(SDMA_CHNENBL_0 + event * 4);
+ val |= (1 << channel);
+ writel(val, SDMA_CHNENBL_0 + event * 4);
+}
+
+static void sdma_event_disable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 val;
+
+ val = readl(SDMA_CHNENBL_0 + event * 4);
+ val &= ~(1 << channel);
+ writel(val, SDMA_CHNENBL_0 + event * 4);
+}
+
+static void sdma_handle_channel_loop(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+
+ /*
+ * loop mode. Iterate over descriptors, re-setup them and
+ * call callback function.
+ */
+ while (1) {
+ bd = &sdmac->bd[sdmac->buf_tail];
+
+ if (bd->mode.status & BD_DONE)
+ break;
+
+ if (bd->mode.status & BD_RROR)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ bd->mode.status |= BD_DONE;
+ sdmac->buf_tail++;
+ sdmac->buf_tail %= sdmac->num_bd;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ }
+}
+
+static void mxc_sdma_handle_channel_normal(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+ int i, error = 0;
+
+ /*
+ * non loop mode. Iterate over all descriptors, collect
+ * errors and call callback function
+ */
+ for (i = 0; i < sdmac->num_bd; i++) {
+ bd = &sdmac->bd[i];
+
+ if (bd->mode.status & (BD_DONE | BD_RROR))
+ error = -EIO;
+ }
+
+ if (error)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ sdmac->last_completed = sdmac->desc.cookie;
+}
+
+static void mxc_sdma_handle_channel(struct sdma_channel *sdmac)
+{
+ complete(&sdmac->done);
+
+ /* not interested in channel 0 interrupts */
+ if (sdmac->channel == 0)
+ return;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP)
+ sdma_handle_channel_loop(sdmac);
+ else
+ mxc_sdma_handle_channel_normal(sdmac);
+}
+
+static irqreturn_t sdma_int_handler(int irq, void *dev_id)
+{
+ struct sdma_engine *sdma = dev_id;
+ u32 stat;
+
+ stat = readl(SDMA_H_INTR);
+ writel(stat, SDMA_H_INTR);
+
+ while (stat) {
+ int channel = fls(stat) - 1;
+ struct sdma_channel *sdmac = &sdma->channel[channel];
+
+ mxc_sdma_handle_channel(sdmac);
+
+ stat &= ~(1 << channel);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * Stores the start address of the SDMA scripts
+ */
+static struct sdma_script_start_addrs __sdma_script_addrs;
+static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
+
+/*
+ * sets the pc of SDMA script according to the peripheral type
+ */
+static void sdma_get_pc(struct sdma_channel *sdmac,
+ enum sdma_peripheral_type peripheral_type)
+{
+ int per_2_emi = 0, emi_2_per = 0;
+ /*
+ * These are needed once we start to support transfers between
+ * two peripherals or memory-to-memory transfers
+ */
+ int per_2_per = 0, emi_2_emi = 0;
+
+ sdmac->pc_from_device = 0;
+ sdmac->pc_to_device = 0;
+
+ switch (peripheral_type) {
+ case IMX_DMATYPE_MEMORY:
+ emi_2_emi = sdma_script_addrs->ap_2_ap_addr;
+ break;
+ case IMX_DMATYPE_DSP:
+ emi_2_per = sdma_script_addrs->bp_2_ap_addr;
+ per_2_emi = sdma_script_addrs->ap_2_bp_addr;
+ break;
+ case IMX_DMATYPE_FIRI:
+ per_2_emi = sdma_script_addrs->firi_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_firi_addr;
+ break;
+ case IMX_DMATYPE_UART:
+ per_2_emi = sdma_script_addrs->uart_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_UART_SP:
+ per_2_emi = sdma_script_addrs->uartsh_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ATA:
+ per_2_emi = sdma_script_addrs->ata_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_ata_addr;
+ break;
+ case IMX_DMATYPE_CSPI:
+ case IMX_DMATYPE_EXT:
+ case IMX_DMATYPE_SSI:
+ per_2_emi = sdma_script_addrs->app_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_SSI_SP:
+ case IMX_DMATYPE_MMC:
+ case IMX_DMATYPE_SDHC:
+ case IMX_DMATYPE_CSPI_SP:
+ case IMX_DMATYPE_ESAI:
+ case IMX_DMATYPE_MSHC_SP:
+ per_2_emi = sdma_script_addrs->shp_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ASRC:
+ per_2_emi = sdma_script_addrs->asrc_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->asrc_2_mcu_addr;
+ per_2_per = sdma_script_addrs->per_2_per_addr;
+ break;
+ case IMX_DMATYPE_MSHC:
+ per_2_emi = sdma_script_addrs->mshc_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_mshc_addr;
+ break;
+ case IMX_DMATYPE_CCM:
+ per_2_emi = sdma_script_addrs->dptc_dvfs_addr;
+ break;
+ case IMX_DMATYPE_SPDIF:
+ per_2_emi = sdma_script_addrs->spdif_2_mcu_addr;
+ emi_2_per = sdma_script_addrs->mcu_2_spdif_addr;
+ break;
+ case IMX_DMATYPE_IPU_MEMORY:
+ emi_2_per = sdma_script_addrs->ext_mem_2_ipu_addr;
+ break;
+ default:
+ break;
+ }
+
+ sdmac->pc_from_device = per_2_emi;
+ sdmac->pc_to_device = emi_2_per;
+}
+
+static int sdma_load_context(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int load_address;
+ struct sdma_context_data *context = sdma->context;
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ int ret;
+
+ if (sdmac->direction == DMA_FROM_DEVICE) {
+ load_address = sdmac->pc_from_device;
+ } else {
+ load_address = sdmac->pc_to_device;
+ }
+
+ if (load_address < 0)
+ return load_address;
+
+ dev_dbg(sdma->dev, "load_address = %d\n", load_address);
+ dev_dbg(sdma->dev, "wml = 0x%08x\n", sdmac->watermark_level);
+ dev_dbg(sdma->dev, "shp_addr = 0x%08x\n", sdmac->shp_addr);
+ dev_dbg(sdma->dev, "per_addr = 0x%08x\n", sdmac->per_addr);
+ dev_dbg(sdma->dev, "event_mask1 = 0x%08x\n", sdmac->event_mask1);
+ dev_dbg(sdma->dev, "event_mask2 = 0x%08x\n", sdmac->event_mask2);
+
+ memset(context, 0, sizeof(*context));
+ context->channel_state.pc = load_address;
+
+ /* Send by context the event mask,base address for peripheral
+ * and watermark level
+ */
+ context->gReg[0] = sdmac->event_mask2;
+ context->gReg[1] = sdmac->event_mask1;
+ context->gReg[2] = sdmac->per_addr;
+ context->gReg[6] = sdmac->shp_addr;
+ context->gReg[7] = sdmac->watermark_level;
+
+ bd0->mode.command = C0_SETDM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = sizeof(*context) / 4;
+ bd0->buffer_addr = sdma->context_phys;
+ bd0->ext_buffer_addr = 2048 + (sizeof(*context) / 4) * channel;
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ return ret;
+}
+
+static void sdma_disable_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ writel(1 << channel, SDMA_H_STATSTOP);
+ sdmac->status = DMA_ERROR;
+}
+
+static int sdma_config_channel(struct sdma_channel *sdmac)
+{
+ int ret;
+
+ sdma_disable_channel(sdmac);
+
+ sdmac->event_mask1 = 0;
+ sdmac->event_mask2 = 0;
+ sdmac->shp_addr = 0;
+ sdmac->per_addr = 0;
+
+ if (sdmac->event_id) {
+ if (sdmac->event_id > 32)
+ return -EINVAL;
+ sdma_event_enable(sdmac, sdmac->event_id);
+ }
+
+ switch (sdmac->peripheral_type) {
+ case IMX_DMATYPE_DSP:
+ sdma_config_ownership(sdmac, false, true, true);
+ break;
+ case IMX_DMATYPE_MEMORY:
+ sdma_config_ownership(sdmac, false, true, false);
+ break;
+ default:
+ sdma_config_ownership(sdmac, true, true, false);
+ break;
+ }
+
+ sdma_get_pc(sdmac, sdmac->peripheral_type);
+
+ if ((sdmac->peripheral_type != IMX_DMATYPE_MEMORY) &&
+ (sdmac->peripheral_type != IMX_DMATYPE_DSP)) {
+ /* Handle multiple event channels differently */
+ if (sdmac->event_id2) {
+ sdmac->event_mask2 = 1 << (sdmac->event_id2 % 32);
+ if (sdmac->event_id2 > 31)
+ sdmac->watermark_level |= 1 << 31;
+ sdmac->event_mask1 = 1 << (sdmac->event_id % 32);
+ if (sdmac->event_id > 31)
+ sdmac->watermark_level |= 1 << 30;
+ } else {
+ sdmac->event_mask1 = 1 << sdmac->event_id;
+ sdmac->event_mask2 = 1 << (sdmac->event_id - 32);
+ }
+ /* Watermark Level */
+ sdmac->watermark_level |= sdmac->watermark_level;
+ /* Address */
+ sdmac->shp_addr = sdmac->per_address;
+ } else {
+ sdmac->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
+ }
+
+ ret = sdma_load_context(sdmac);
+
+ return ret;
+}
+
+static int sdma_set_channel_priority(struct sdma_channel *sdmac,
+ unsigned int priority)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ if (priority < MXC_SDMA_MIN_PRIORITY
+ || priority > MXC_SDMA_MAX_PRIORITY) {
+ return -EINVAL;
+ }
+
+ writel(priority, SDMA_CHNPRI_0 + 4 * channel);
+
+ return 0;
+}
+
+static int sdma_request_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret = -EBUSY;
+
+ sdmac->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdmac->bd_phys, GFP_KERNEL);
+ if (!sdmac->bd) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memset(sdmac->bd, 0, PAGE_SIZE);
+
+ sdma->channel_control[channel].base_bd_ptr = sdmac->bd_phys;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ clk_enable(sdma->clk);
+
+ sdma_set_channel_priority(sdmac, MXC_SDMA_DEFAULT_PRIORITY);
+
+ init_completion(&sdmac->done);
+
+ sdmac->buf_tail = 0;
+
+ return 0;
+out:
+
+ return ret;
+}
+
+static void sdma_enable_channel(struct sdma_engine *sdma, int channel)
+{
+ writel(1 << channel, SDMA_H_START);
+}
+
+static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
+{
+ dma_cookie_t cookie = sdma->chan.cookie;
+
+ if (++cookie < 0)
+ cookie = 1;
+
+ sdma->chan.cookie = cookie;
+ sdma->desc.cookie = cookie;
+
+ return cookie;
+}
+
+static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sdma_channel, chan);
+}
+
+static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(tx->chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ dma_cookie_t cookie;
+
+ spin_lock_irq(&sdmac->lock);
+
+ cookie = sdma_assign_cookie(sdmac);
+
+ sdma_enable_channel(sdma, tx->chan->chan_id);
+
+ spin_unlock_irq(&sdmac->lock);
+
+ return cookie;
+}
+
+static int sdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct imx_dma_data *data = chan->private;
+ int prio, ret;
+
+ /* No need to execute this for internal channel 0 */
+ if (chan->chan_id == 0)
+ return 0;
+
+ if (!data)
+ return -EINVAL;
+
+ switch (data->priority) {
+ case DMA_PRIO_HIGH:
+ prio = 3;
+ break;
+ case DMA_PRIO_MEDIUM:
+ prio = 2;
+ break;
+ case DMA_PRIO_LOW:
+ default:
+ prio = 1;
+ break;
+ }
+
+ sdmac->peripheral_type = data->peripheral_type;
+ sdmac->event_id = data->dma_request;
+ ret = sdma_set_channel_priority(sdmac, prio);
+ if (ret)
+ return ret;
+
+ ret = sdma_request_channel(sdmac);
+ if (ret)
+ return ret;
+
+ dma_async_tx_descriptor_init(&sdmac->desc, chan);
+ sdmac->desc.tx_submit = sdma_tx_submit;
+ /* txd.flags will be overwritten in prep funcs */
+ sdmac->desc.flags = DMA_CTRL_ACK;
+
+ return 0;
+}
+
+static void sdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+
+ sdma_disable_channel(sdmac);
+
+ if (sdmac->event_id)
+ sdma_event_disable(sdmac, sdmac->event_id);
+ if (sdmac->event_id2)
+ sdma_event_disable(sdmac, sdmac->event_id2);
+
+ sdmac->event_id = 0;
+ sdmac->event_id2 = 0;
+
+ sdma_set_channel_priority(sdmac, 0);
+
+ dma_free_coherent(NULL, PAGE_SIZE, sdmac->bd, sdmac->bd_phys);
+
+ clk_disable(sdma->clk);
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int ret, i, count;
+ int channel = chan->chan_id;
+ struct scatterlist *sg;
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags = 0;
+
+ dev_dbg(sdma->dev, "setting up %d entries for channel %d.\n",
+ sg_len, channel);
+
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (sg_len > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, sg_len, NUM_BD);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = sgl->dma_address;
+
+ count = sg->length;
+
+ if (count > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
+ channel, count, 0xffff);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ bd->mode.count = count;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES) {
+ ret = -EINVAL;
+ goto err_out;
+ }
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP) {
+ param |= BD_INTR;
+ if (i + 1 == sg_len)
+ param |= BD_WRAP;
+ }
+
+ if (i + 1 == sg_len)
+ param |= BD_INTR;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, count, sg->dma_address,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+ }
+
+ sdmac->num_bd = sg_len;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int num_periods = buf_len / period_len;
+ int channel = chan->chan_id;
+ int ret, i = 0, buf = 0;
+
+ dev_dbg(sdma->dev, "%s channel: %d\n", __func__, channel);
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags |= IMX_DMA_SG_LOOP;
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (num_periods > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, num_periods, NUM_BD);
+ goto err_out;
+ }
+
+ if (period_len > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum period size exceeded: %d > %d\n",
+ channel, period_len, 0xffff);
+ goto err_out;
+ }
+
+ while (buf < buf_len) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = dma_addr;
+
+ bd->mode.count = period_len;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES)
+ goto err_out;
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
+ if (i + 1 == num_periods)
+ param |= BD_WRAP;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, period_len, dma_addr,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+
+ dma_addr += period_len;
+ buf += period_len;
+
+ i++;
+ }
+
+ sdmac->num_bd = num_periods;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ sdmac->status = DMA_ERROR;
+ return NULL;
+}
+
+static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct dma_slave_config *dmaengine_cfg = (void *)arg;
+
+ switch (cmd) {
+ case DMA_TERMINATE_ALL:
+ sdma_disable_channel(sdmac);
+ return 0;
+ case DMA_SLAVE_CONFIG:
+ if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
+ sdmac->per_address = dmaengine_cfg->src_addr;
+ sdmac->watermark_level = dmaengine_cfg->src_maxburst;
+ sdmac->word_size = dmaengine_cfg->src_addr_width;
+ } else {
+ sdmac->per_address = dmaengine_cfg->dst_addr;
+ sdmac->watermark_level = dmaengine_cfg->dst_maxburst;
+ sdmac->word_size = dmaengine_cfg->dst_addr_width;
+ }
+ return sdma_config_channel(sdmac);
+ default:
+ return -ENOSYS;
+ }
+
+ return -EINVAL;
+}
+
+static enum dma_status sdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ dma_cookie_t last_used;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+
+ ret = dma_async_is_complete(cookie, sdmac->last_completed, last_used);
+ dma_set_tx_state(txstate, sdmac->last_completed, last_used, 0);
+
+ return ret;
+}
+
+static void sdma_issue_pending(struct dma_chan *chan)
+{
+ /*
+ * Nothing to do. We only have a single descriptor
+ */
+}
+
+static int __init sdma_init(struct sdma_engine *sdma,
+ void *ram_code, int ram_code_size)
+{
+ int i, ret;
+ dma_addr_t ccb_phys;
+
+ switch (sdma->version) {
+ case 1:
+ sdma->num_events = 32;
+ break;
+ case 2:
+ sdma->num_events = 48;
+ break;
+ default:
+ dev_err(sdma->dev, "Unknown version %d. aborting\n", sdma->version);
+ return -ENODEV;
+ }
+
+ clk_enable(sdma->clk);
+
+ /* Be sure SDMA has not started yet */
+ writel(0, SDMA_H_C0PTR);
+
+ sdma->channel_control = dma_alloc_coherent(NULL,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
+ sizeof(struct sdma_context_data),
+ &ccb_phys, GFP_KERNEL);
+
+ if (!sdma->channel_control) {
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ sdma->context = (void *)sdma->channel_control +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+ sdma->context_phys = ccb_phys +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+
+ /* Zero-out the CCB structures array just allocated */
+ memset(sdma->channel_control, 0,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
+
+ /* disable all channels */
+ for (i = 0; i < sdma->num_events; i++)
+ writel(0, SDMA_CHNENBL_0 + i * 4);
+
+ /* All channels have priority 0 */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++)
+ writel(0, SDMA_CHNPRI_0 + i * 4);
+
+ ret = sdma_request_channel(&sdma->channel[0]);
+ if (ret)
+ goto err_dma_alloc;
+
+ sdma_config_ownership(&sdma->channel[0], false, true, false);
+
+ /* Set Command Channel (Channel Zero) */
+ writel(0x4050, SDMA_CHN0ADDR);
+
+ /* Set bits of CONFIG register but with static context switching */
+ /* FIXME: Check whether to set ACR bit depending on clock ratios */
+ writel(0, SDMA_H_CONFIG);
+
+ writel(ccb_phys, SDMA_H_C0PTR);
+
+ /* download the RAM image for SDMA */
+ sdma_load_script(sdma, ram_code,
+ ram_code_size,
+ sdma_script_addrs->ram_code_start_addr);
+
+ /* Set bits of CONFIG register with given context switching mode */
+ writel(SDMA_H_CONFIG_CSM, SDMA_H_CONFIG);
+
+ /* Initializes channel's priorities */
+ sdma_set_channel_priority(&sdma->channel[0], 7);
+
+ clk_disable(sdma->clk);
+
+ return 0;
+
+err_dma_alloc:
+ clk_disable(sdma->clk);
+ dev_err(sdma->dev, "initialisation failed with %d\n", ret);
+ return ret;
+}
+
+static int __devinit sdma_probe(struct platform_device *pdev)
+{
+ int ret;
+ const struct firmware *fw;
+ const struct sdma_firmware_header *header;
+ const struct sdma_script_start_addrs *addr;
+ int irq;
+ unsigned short *ram_code;
+ struct resource *iores;
+ struct sdma_platform_data *pdata = pdev->dev.platform_data;
+ char *fwname;
+ int i;
+ dma_cap_mask_t mask;
+ struct sdma_engine *sdma;
+
+ sdma = kzalloc(sizeof(*sdma), GFP_KERNEL);
+ if (!sdma)
+ return -ENOMEM;
+
+ sdma->dev = &pdev->dev;
+
+ iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ irq = platform_get_irq(pdev, 0);
+ if (!iores || irq < 0 || !pdata) {
+ ret = -EINVAL;
+ goto err_irq;
+ }
+
+ sdma->clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(sdma->clk)) {
+ ret = PTR_ERR(sdma->clk);
+ goto err_clk;
+ }
+
+ sdma->regs = ioremap(iores->start, resource_size(iores));
+ if (!sdma->regs) {
+ ret = -ENOMEM;
+ goto err_ioremap;
+ }
+
+ ret = request_irq(irq, sdma_int_handler, 0, "sdma", sdma);
+ if (ret)
+ goto err_request_irq;
+
+ fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin",
+ pdata->cpu_name, pdata->to_version);
+ if (!fwname) {
+ ret = -ENOMEM;
+ goto err_cputype;
+ }
+
+ ret = request_firmware(&fw, fwname, &pdev->dev);
+ if (ret) {
+ dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
+ fwname, ret);
+ kfree(fwname);
+ goto err_cputype;
+ }
+ kfree(fwname);
+
+ if (fw->size < sizeof(*header))
+ goto err_firmware;
+
+ header = (struct sdma_firmware_header *)fw->data;
+
+ if (header->magic != SDMA_FIRMWARE_MAGIC)
+ goto err_firmware;
+ if (header->ram_code_start + header->ram_code_size > fw->size)
+ goto err_firmware;
+
+ addr = (void *)header + header->script_addrs_start;
+ ram_code = (void *)header + header->ram_code_start;
+ memcpy(&__sdma_script_addrs, addr, sizeof(*addr));
+
+ sdma->version = pdata->sdma_version;
+
+ INIT_LIST_HEAD(&sdma->dma_device.channels);
+ /* Initialize channel parameters */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++) {
+ struct sdma_channel *sdmac = &sdma->channel[i];
+
+ sdmac->sdma = sdma;
+ spin_lock_init(&sdmac->lock);
+
+ dma_cap_set(DMA_SLAVE, sdma->dma_device.cap_mask);
+ dma_cap_set(DMA_CYCLIC, sdma->dma_device.cap_mask);
+
+ sdmac->chan.device = &sdma->dma_device;
+ sdmac->chan.chan_id = i;
+ sdmac->channel = i;
+
+ /* Add the channel to the DMAC list */
+ list_add_tail(&sdmac->chan.device_node, &sdma->dma_device.channels);
+ }
+
+ ret = sdma_init(sdma, ram_code, header->ram_code_size);
+ if (ret)
+ goto err_firmware;
+
+ sdma->dma_device.dev = &pdev->dev;
+
+ sdma->dma_device.device_alloc_chan_resources = sdma_alloc_chan_resources;
+ sdma->dma_device.device_free_chan_resources = sdma_free_chan_resources;
+ sdma->dma_device.device_tx_status = sdma_tx_status;
+ sdma->dma_device.device_prep_slave_sg = sdma_prep_slave_sg;
+ sdma->dma_device.device_prep_dma_cyclic = sdma_prep_dma_cyclic;
+ sdma->dma_device.device_control = sdma_control;
+ sdma->dma_device.device_issue_pending = sdma_issue_pending;
+
+ ret = dma_async_device_register(&sdma->dma_device);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register\n");
+ goto err_firmware;
+ }
+
+ dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
+ header->version_major,
+ header->version_minor);
+
+ /* request channel 0. This is an internal control channel
+ * to the SDMA engine and not available to clients.
+ */
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_request_channel(mask, NULL, NULL);
+
+ release_firmware(fw);
+
+ return 0;
+
+err_firmware:
+ release_firmware(fw);
+err_cputype:
+ free_irq(irq, sdma);
+err_request_irq:
+ iounmap(sdma->regs);
+err_ioremap:
+ clk_put(sdma->clk);
+err_clk:
+err_irq:
+ kfree(sdma);
+ return 0;
+}
+
+static int __devexit sdma_remove(struct platform_device *pdev)
+{
+ return -EBUSY;
+}
+
+static struct platform_driver sdma_driver = {
+ .driver = {
+ .name = "imx-sdma",
+ },
+ .probe = sdma_probe,
+ .remove = __devexit_p(sdma_remove),
+};
+
+static int __init sdma_module_init(void)
+{
+ return platform_driver_register(&sdma_driver);
+}
+subsys_initcall(sdma_module_init);
+
+MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("i.MX SDMA driver");
+MODULE_LICENSE("GPL");
--
1.7.1
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply related [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-23 12:57 ` Sascha Hauer
@ 2010-08-23 17:30 ` Linus Walleij
-1 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-23 17:30 UTC (permalink / raw)
To: Sascha Hauer; +Cc: linux-kernel, Dan Williams, linux-arm-kernel
2010/8/23 Sascha Hauer <s.hauer@pengutronix.de>:
> This patch adds support for the Freescale i.MX SDMA engine.
Great progress!
> (...)
> diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> (...)
> +/* SDMA registers */
> +#define SDMA_H_C0PTR (sdma->regs + 0x000)
> +#define SDMA_H_INTR (sdma->regs + 0x004)
> +#define SDMA_H_STATSTOP (sdma->regs + 0x008)
> +#define SDMA_H_START (sdma->regs + 0x00c)
> +#define SDMA_H_EVTOVR (sdma->regs + 0x010)
> +#define SDMA_H_DSPOVR (sdma->regs + 0x014)
> +#define SDMA_H_HOSTOVR (sdma->regs + 0x018)
> +#define SDMA_H_EVTPEND (sdma->regs + 0x01c)
> +#define SDMA_H_DSPENBL (sdma->regs + 0x020)
> +#define SDMA_H_RESET (sdma->regs + 0x024)
> +#define SDMA_H_EVTERR (sdma->regs + 0x028)
> +#define SDMA_H_INTRMSK (sdma->regs + 0x02c)
> +#define SDMA_H_PSW (sdma->regs + 0x030)
> +#define SDMA_H_EVTERRDBG (sdma->regs + 0x034)
> +#define SDMA_H_CONFIG (sdma->regs + 0x038)
> +#define SDMA_ONCE_ENB (sdma->regs + 0x040)
> +#define SDMA_ONCE_DATA (sdma->regs + 0x044)
> +#define SDMA_ONCE_INSTR (sdma->regs + 0x048)
> +#define SDMA_ONCE_STAT (sdma->regs + 0x04c)
> +#define SDMA_ONCE_CMD (sdma->regs + 0x050)
> +#define SDMA_EVT_MIRROR (sdma->regs + 0x054)
> +#define SDMA_ILLINSTADDR (sdma->regs + 0x058)
> +#define SDMA_CHN0ADDR (sdma->regs + 0x05c)
> +#define SDMA_ONCE_RTB (sdma->regs + 0x060)
> +#define SDMA_XTRIG_CONF1 (sdma->regs + 0x070)
> +#define SDMA_XTRIG_CONF2 (sdma->regs + 0x074)
> +#define SDMA_CHNENBL_0 (sdma->regs + (sdma->version == 2 ? 0x200 : 0x80))
> +#define SDMA_CHNPRI_0 (sdma->regs + 0x100)
These macros expand to the local variable "sdma" which must
be present in all functions using them. I don't know what is
considered moste readable, but I would certainly just
#define SDMA_FOO (0x0123)
(...)
u32 foo = readl(sdma->regs + SDMA_FOO);
That is more common I think.
> (...)
> +/*
> + * Channel control Block
Some kerneldoc here describing especially
@unused: padding for register cast
@usused1: padding for register cast
Hm "unused" and "unused1" wrinkles my binary brain,
can you call the first one "unused0" for my perceptions
sake?
> + */
> +struct sdma_channel_control {
> + dma_addr_t current_bd_ptr; /* current buffer descriptor processed */
> + dma_addr_t base_bd_ptr; /* first element of buffer descriptor array */
> + u32 unused;
> + u32 unused1;
> +} __attribute__ ((packed));
> (...)
> +/**
> + * struct sdma_channel - housekeeping for a SDMA channel
> + *
> + * @sdma pointer to the SDMA engine for this channel
> + * @channel the channel number, matches dmaengine chan_id
> + * @direction transfer type. Needed for setting SDMA script
> + * @peripheral_type Peripheral type. Needed for setting SDMA script
> + * @event_id aka dma request line
> + * @event_id2 for channels that use 2 events
> + * @word_size peripheral access size
> + * @buf_tail ID of the buffer that was processed
> + * @done channel completion
> + * @num_bd max NUM_BD. number of descriptors currently handling
> + */
> +struct sdma_channel {
> + struct sdma_engine *sdma;
> + unsigned int channel;
> + enum dma_data_direction direction;
> + enum sdma_peripheral_type peripheral_type;
> + unsigned int event_id;
> + unsigned int event_id2;
id1 und id2 oder
id0 und id1 fierleicht?
> + enum dma_slave_buswidth word_size;
> + unsigned int buf_tail;
> + struct completion done;
> + unsigned int num_bd;
> + struct sdma_buffer_descriptor *bd;
> + dma_addr_t bd_phys;
> + unsigned int pc_from_device, pc_to_device;
> + unsigned long flags;
> + dma_addr_t per_address;
> + u32 event_mask1, event_mask2;
1 and 2, else unnumbered and 1, else unnumbered and 2 X-)
> + u32 watermark_level;
> + u32 shp_addr, per_addr;
> + struct dma_chan chan;
> + spinlock_t lock;
> + struct dma_async_tx_descriptor desc;
> + dma_cookie_t last_completed;
> + enum dma_status status;
> +};
> +
> +#define IMX_DMA_SG_LOOP (1 << 0)
> +
> +#define MAX_DMA_CHANNELS 32
> +#define MXC_SDMA_DEFAULT_PRIORITY 1
> +#define MXC_SDMA_MIN_PRIORITY 1
> +#define MXC_SDMA_MAX_PRIORITY 7
> +
> +/*
> + * This enumerates transfer types
> + */
> +enum {
> + emi_2_per = 0, /* EMI memory to peripheral */
> + emi_2_int, /* EMI memory to internal RAM */
> + emi_2_emi, /* EMI memory to EMI memory */
> + emi_2_dsp, /* EMI memory to DSP memory */
> + per_2_int, /* Peripheral to internal RAM */
> + per_2_emi, /* Peripheral to internal EMI memory */
> + per_2_dsp, /* Peripheral to DSP memory */
> + per_2_per, /* Peripheral to Peripheral */
> + int_2_per, /* Internal RAM to peripheral */
> + int_2_int, /* Internal RAM to Internal RAM */
> + int_2_emi, /* Internal RAM to EMI memory */
> + int_2_dsp, /* Internal RAM to DSP memory */
> + dsp_2_per, /* DSP memory to peripheral */
> + dsp_2_int, /* DSP memory to internal RAM */
> + dsp_2_emi, /* DSP memory to EMI memory */
> + dsp_2_dsp, /* DSP memory to DSP memory */
> + emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
> + dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
> + dvfs_pll, /* DVFS script with PLL change */
> + dvfs_pdr /* DVFS script without PLL change */
> +} sdma_transfer_type;
Picky me, but it's no type, its an enum. I understand that it is
a technical term...
What about just calling is sdma_transfer? Short and nice.
Or sdma_transfer_line?
> (...)
> +/*
> + * Stores the start address of the SDMA scripts
> + */
> +static struct sdma_script_start_addrs __sdma_script_addrs;
> +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
What's the rationale behind prefixing that variable with __?
The same name for struct and variable is perfectly viable.
Apart from these smallies (and it's all minor stuff) it's nice and clean so:
Reviewed-by: Linus Walleij <linus.walleij@stericsson.com>
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-23 17:30 ` Linus Walleij
0 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-23 17:30 UTC (permalink / raw)
To: linux-arm-kernel
2010/8/23 Sascha Hauer <s.hauer@pengutronix.de>:
> This patch adds support for the Freescale i.MX SDMA engine.
Great progress!
> (...)
> diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> (...)
> +/* SDMA registers */
> +#define SDMA_H_C0PTR ? ? ? ? ? (sdma->regs + 0x000)
> +#define SDMA_H_INTR ? ? ? ? ? ?(sdma->regs + 0x004)
> +#define SDMA_H_STATSTOP ? ? ? ? ? ? ? ?(sdma->regs + 0x008)
> +#define SDMA_H_START ? ? ? ? ? (sdma->regs + 0x00c)
> +#define SDMA_H_EVTOVR ? ? ? ? ?(sdma->regs + 0x010)
> +#define SDMA_H_DSPOVR ? ? ? ? ?(sdma->regs + 0x014)
> +#define SDMA_H_HOSTOVR ? ? ? ? (sdma->regs + 0x018)
> +#define SDMA_H_EVTPEND ? ? ? ? (sdma->regs + 0x01c)
> +#define SDMA_H_DSPENBL ? ? ? ? (sdma->regs + 0x020)
> +#define SDMA_H_RESET ? ? ? ? ? (sdma->regs + 0x024)
> +#define SDMA_H_EVTERR ? ? ? ? ?(sdma->regs + 0x028)
> +#define SDMA_H_INTRMSK ? ? ? ? (sdma->regs + 0x02c)
> +#define SDMA_H_PSW ? ? ? ? ? ? (sdma->regs + 0x030)
> +#define SDMA_H_EVTERRDBG ? ? ? (sdma->regs + 0x034)
> +#define SDMA_H_CONFIG ? ? ? ? ?(sdma->regs + 0x038)
> +#define SDMA_ONCE_ENB ? ? ? ? ?(sdma->regs + 0x040)
> +#define SDMA_ONCE_DATA ? ? ? ? (sdma->regs + 0x044)
> +#define SDMA_ONCE_INSTR ? ? ? ? ? ? ? ?(sdma->regs + 0x048)
> +#define SDMA_ONCE_STAT ? ? ? ? (sdma->regs + 0x04c)
> +#define SDMA_ONCE_CMD ? ? ? ? ?(sdma->regs + 0x050)
> +#define SDMA_EVT_MIRROR ? ? ? ? ? ? ? ?(sdma->regs + 0x054)
> +#define SDMA_ILLINSTADDR ? ? ? (sdma->regs + 0x058)
> +#define SDMA_CHN0ADDR ? ? ? ? ?(sdma->regs + 0x05c)
> +#define SDMA_ONCE_RTB ? ? ? ? ?(sdma->regs + 0x060)
> +#define SDMA_XTRIG_CONF1 ? ? ? (sdma->regs + 0x070)
> +#define SDMA_XTRIG_CONF2 ? ? ? (sdma->regs + 0x074)
> +#define SDMA_CHNENBL_0 ? ? ? ? (sdma->regs + (sdma->version == 2 ? 0x200 : 0x80))
> +#define SDMA_CHNPRI_0 ? ? ? ? ?(sdma->regs + 0x100)
These macros expand to the local variable "sdma" which must
be present in all functions using them. I don't know what is
considered moste readable, but I would certainly just
#define SDMA_FOO (0x0123)
(...)
u32 foo = readl(sdma->regs + SDMA_FOO);
That is more common I think.
> (...)
> +/*
> + * Channel control Block
Some kerneldoc here describing especially
@unused: padding for register cast
@usused1: padding for register cast
Hm "unused" and "unused1" wrinkles my binary brain,
can you call the first one "unused0" for my perceptions
sake?
> + */
> +struct sdma_channel_control {
> + ? ? ? dma_addr_t current_bd_ptr; /* current buffer descriptor processed */
> + ? ? ? dma_addr_t base_bd_ptr; ? ?/* first element of buffer descriptor array */
> + ? ? ? u32 unused;
> + ? ? ? u32 unused1;
> +} __attribute__ ((packed));
> (...)
> +/**
> + * struct sdma_channel - housekeeping for a SDMA channel
> + *
> + * @sdma ? ? ? ? ? ? ? pointer to the SDMA engine for this channel
> + * @channel ? ? ? ? ? ?the channel number, matches dmaengine chan_id
> + * @direction ? ? ? ? ?transfer type. Needed for setting SDMA script
> + * @peripheral_type ? ?Peripheral type. Needed for setting SDMA script
> + * @event_id ? ? ? ? ? aka dma request line
> + * @event_id2 ? ? ? ? ?for channels that use 2 events
> + * @word_size ? ? ? ? ?peripheral access size
> + * @buf_tail ? ? ? ? ? ID of the buffer that was processed
> + * @done ? ? ? ? ? ? ? channel completion
> + * @num_bd ? ? ? ? ? ? max NUM_BD. number of descriptors currently handling
> + */
> +struct sdma_channel {
> + ? ? ? struct sdma_engine ? ? ? ? ? ? ?*sdma;
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?channel;
> + ? ? ? enum dma_data_direction ? ? ? ? direction;
> + ? ? ? enum sdma_peripheral_type ? ? ? peripheral_type;
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?event_id;
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?event_id2;
id1 und id2 oder
id0 und id1 fierleicht?
> + ? ? ? enum dma_slave_buswidth ? ? ? ? word_size;
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?buf_tail;
> + ? ? ? struct completion ? ? ? ? ? ? ? done;
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?num_bd;
> + ? ? ? struct sdma_buffer_descriptor ? *bd;
> + ? ? ? dma_addr_t ? ? ? ? ? ? ? ? ? ? ?bd_phys;
> + ? ? ? unsigned int ? ? ? ? ? ? ? ? ? ?pc_from_device, pc_to_device;
> + ? ? ? unsigned long ? ? ? ? ? ? ? ? ? flags;
> + ? ? ? dma_addr_t ? ? ? ? ? ? ? ? ? ? ?per_address;
> + ? ? ? u32 ? ? ? ? ? ? ? ? ? ? ? ? ? ? event_mask1, event_mask2;
1 and 2, else unnumbered and 1, else unnumbered and 2 X-)
> + ? ? ? u32 ? ? ? ? ? ? ? ? ? ? ? ? ? ? watermark_level;
> + ? ? ? u32 ? ? ? ? ? ? ? ? ? ? ? ? ? ? shp_addr, per_addr;
> + ? ? ? struct dma_chan ? ? ? ? ? ? ? ? chan;
> + ? ? ? spinlock_t ? ? ? ? ? ? ? ? ? ? ?lock;
> + ? ? ? struct dma_async_tx_descriptor ?desc;
> + ? ? ? dma_cookie_t ? ? ? ? ? ? ? ? ? ?last_completed;
> + ? ? ? enum dma_status ? ? ? ? ? ? ? ? status;
> +};
> +
> +#define IMX_DMA_SG_LOOP ? ? ? ? ? ? ? ?(1 << 0)
> +
> +#define MAX_DMA_CHANNELS 32
> +#define MXC_SDMA_DEFAULT_PRIORITY 1
> +#define MXC_SDMA_MIN_PRIORITY 1
> +#define MXC_SDMA_MAX_PRIORITY 7
> +
> +/*
> + * This enumerates transfer types
> + */
> +enum {
> + ? ? ? emi_2_per = 0, ? ? ? ? ?/* EMI memory to peripheral */
> + ? ? ? emi_2_int, ? ? ? ? ? ? ?/* EMI memory to internal RAM */
> + ? ? ? emi_2_emi, ? ? ? ? ? ? ?/* EMI memory to EMI memory */
> + ? ? ? emi_2_dsp, ? ? ? ? ? ? ?/* EMI memory to DSP memory */
> + ? ? ? per_2_int, ? ? ? ? ? ? ?/* Peripheral to internal RAM */
> + ? ? ? per_2_emi, ? ? ? ? ? ? ?/* Peripheral to internal EMI memory */
> + ? ? ? per_2_dsp, ? ? ? ? ? ? ?/* Peripheral to DSP memory */
> + ? ? ? per_2_per, ? ? ? ? ? ? ?/* Peripheral to Peripheral */
> + ? ? ? int_2_per, ? ? ? ? ? ? ?/* Internal RAM to peripheral */
> + ? ? ? int_2_int, ? ? ? ? ? ? ?/* Internal RAM to Internal RAM */
> + ? ? ? int_2_emi, ? ? ? ? ? ? ?/* Internal RAM to EMI memory */
> + ? ? ? int_2_dsp, ? ? ? ? ? ? ?/* Internal RAM to DSP memory */
> + ? ? ? dsp_2_per, ? ? ? ? ? ? ?/* DSP memory to peripheral */
> + ? ? ? dsp_2_int, ? ? ? ? ? ? ?/* DSP memory to internal RAM */
> + ? ? ? dsp_2_emi, ? ? ? ? ? ? ?/* DSP memory to EMI memory */
> + ? ? ? dsp_2_dsp, ? ? ? ? ? ? ?/* DSP memory to DSP memory */
> + ? ? ? emi_2_dsp_loop, ? ? ? ? /* EMI memory to DSP memory loopback */
> + ? ? ? dsp_2_emi_loop, ? ? ? ? /* DSP memory to EMI memory loopback */
> + ? ? ? dvfs_pll, ? ? ? ? ? ? ? /* DVFS script with PLL change ? ? ? */
> + ? ? ? dvfs_pdr ? ? ? ? ? ? ? ?/* DVFS script without PLL change ? ?*/
> +} sdma_transfer_type;
Picky me, but it's no type, its an enum. I understand that it is
a technical term...
What about just calling is sdma_transfer? Short and nice.
Or sdma_transfer_line?
> (...)
> +/*
> + * Stores the start address of the SDMA scripts
> + */
> +static struct sdma_script_start_addrs __sdma_script_addrs;
> +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
What's the rationale behind prefixing that variable with __?
The same name for struct and variable is perfectly viable.
Apart from these smallies (and it's all minor stuff) it's nice and clean so:
Reviewed-by: Linus Walleij <linus.walleij@stericsson.com>
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-23 12:57 ` Sascha Hauer
@ 2010-08-23 17:48 ` Uwe Kleine-König
-1 siblings, 0 replies; 78+ messages in thread
From: Uwe Kleine-König @ 2010-08-23 17:48 UTC (permalink / raw)
To: Sascha Hauer; +Cc: linux-kernel, Linus Walleij, Dan Williams, linux-arm-kernel
Hallo Sascha,
On Mon, Aug 23, 2010 at 02:57:04PM +0200, Sascha Hauer wrote:
>
> This patch adds support for the Freescale i.MX SDMA engine.
>
> The SDMA engine is a scatter/gather DMA engine which is implemented
> as a seperate coprocessor. SDMA needs its own firmware which is
> requested using the standard request_firmware mechanism. The firmware
> has different entry points for each peripheral type, so drivers
> have to pass the peripheral type to the DMA engine which in turn
> picks the correct firmware entry point from a table contained in
> the firmware image itself.
> The original Freescale code also supports support for transfering
> data to the internal SRAM which needs different entry points to
> the firmware. Support for this is currently not implemented. Also,
> support for the ASRC (asymmetric sample rate converter) is skipped.
>
> I took a very simple approach to implement dmaengine support. Only
> a single descriptor is statically assigned to a each channel. This
> means that transfers can't be queued up but only a single transfer
> is in progress. This simplifies implementation a lot and is sufficient
> for the usual device/memory transfers.
>
> Changes since v1:
>
> - included comments from Linus Walleij
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
> arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
> arch/arm/mach-mx3/Kconfig | 2 +
> arch/arm/plat-mxc/Kconfig | 10 +
> arch/arm/plat-mxc/include/mach/dma.h | 64 ++
> arch/arm/plat-mxc/include/mach/sdma.h | 17 +
> drivers/dma/Kconfig | 8 +
> drivers/dma/Makefile | 1 +
> drivers/dma/imx-sdma.c | 1395 +++++++++++++++++++++++++++++++
> 8 files changed, 1499 insertions(+), 6 deletions(-)
> create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
> create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
> create mode 100644 drivers/dma/imx-sdma.c
>
> diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
> index 287431c..ac6fd71 100644
> --- a/arch/arm/mach-imx/include/mach/dma-v1.h
> +++ b/arch/arm/mach-imx/include/mach/dma-v1.h
> @@ -27,6 +27,8 @@
>
> #define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
>
> +#include <mach/dma.h>
> +
> #define IMX_DMA_CHANNELS 16
>
> #define DMA_MODE_READ 0
> @@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
>
> void imx_dma_free(int channel);
>
> -enum imx_dma_prio {
> - DMA_PRIO_HIGH = 0,
> - DMA_PRIO_MEDIUM = 1,
> - DMA_PRIO_LOW = 2
> -};
> -
> int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
>
> #endif /* __MACH_DMA_V1_H__ */
> diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
> index 85beece..301375c 100644
> --- a/arch/arm/mach-mx3/Kconfig
> +++ b/arch/arm/mach-mx3/Kconfig
> @@ -3,12 +3,14 @@ if ARCH_MX3
> config ARCH_MX31
> select ARCH_HAS_RNGA
> select ARCH_MXC_AUDMUX_V2
> + select IMX_HAVE_SDMA
> bool
>
> config ARCH_MX35
> bool
> select ARCH_MXC_IOMUX_V3
> select ARCH_MXC_AUDMUX_V2
> + select IMX_HAVE_SDMA
>
> comment "MX3 platforms:"
>
> diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
> index 0527e65..6741625 100644
> --- a/arch/arm/plat-mxc/Kconfig
> +++ b/arch/arm/plat-mxc/Kconfig
> @@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
> config ARCH_MXC_AUDMUX_V2
> bool
>
> +config IMX_HAVE_SDMA
> + bool
> +
> +config IMX_SDMA
> + depends on IMX_HAVE_SDMA
> + tristate "Enable SDMA support"
> + help
> + Include support for the SDMA engine. The SDMA engine needs additional
> + firmware support. SDMA can be compiled as a module to support loading
> + the firmware when a rootfs is present.
> endif
> diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
> new file mode 100644
> index 0000000..69d181f
> --- /dev/null
> +++ b/arch/arm/plat-mxc/include/mach/dma.h
> @@ -0,0 +1,64 @@
> +/*
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 as
> + * published by the Free Software Foundation.
> + */
> +
> +#ifndef __ASM_ARCH_MXC_DMA_H__
> +#define __ASM_ARCH_MXC_DMA_H__
__MACH_DMA_H__ please
> +
> +#include <linux/scatterlist.h>
> +
> +/*
> + * This enumerates peripheral types. Used for SDMA.
> + */
> +enum sdma_peripheral_type {
> + IMX_DMATYPE_SSI, /* MCU domain SSI */
> + IMX_DMATYPE_SSI_SP, /* Shared SSI */
> + IMX_DMATYPE_MMC, /* MMC */
> + IMX_DMATYPE_SDHC, /* SDHC */
> + IMX_DMATYPE_UART, /* MCU domain UART */
> + IMX_DMATYPE_UART_SP, /* Shared UART */
> + IMX_DMATYPE_FIRI, /* FIRI */
> + IMX_DMATYPE_CSPI, /* MCU domain CSPI */
> + IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
> + IMX_DMATYPE_SIM, /* SIM */
> + IMX_DMATYPE_ATA, /* ATA */
> + IMX_DMATYPE_CCM, /* CCM */
> + IMX_DMATYPE_EXT, /* External peripheral */
> + IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
> + IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
> + IMX_DMATYPE_DSP, /* DSP */
> + IMX_DMATYPE_MEMORY, /* Memory */
> + IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
> + IMX_DMATYPE_SPDIF, /* SPDIF */
> + IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
> + IMX_DMATYPE_ASRC, /* ASRC */
> + IMX_DMATYPE_ESAI, /* ESAI */
> +};
> +
> +enum imx_dma_prio {
> + DMA_PRIO_HIGH = 0,
> + DMA_PRIO_MEDIUM = 1,
> + DMA_PRIO_LOW = 2
> +};
> +
> +struct imx_dma_data {
> + int dma_request; /* DMA request line */
> + enum sdma_peripheral_type peripheral_type;
> + int priority;
> +};
> +
> +static inline int imx_dma_is_ipu(struct dma_chan *chan)
> +{
> + return !strcmp(dev_name(chan->device->dev), "ipu-core");
> +}
> +
> +static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
> +{
> + return !strcmp(dev_name(chan->device->dev), "imx-sdma");
> +}
> +
> +#endif
> diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
> new file mode 100644
> index 0000000..9be1122
> --- /dev/null
> +++ b/arch/arm/plat-mxc/include/mach/sdma.h
> @@ -0,0 +1,17 @@
> +#ifndef __MACH_MXC_SDMA_H__
> +#define __MACH_MXC_SDMA_H__
__MACH_SDMA_H__
> +
> +/**
> + * struct sdma_platform_data - platform specific data for SDMA engine
> + *
> + * @sdma_version The version of this SDMA engine
> + * @cpu_name used to generate the firmware name
> + * @to_version CPU Tape out version
> + */
> +struct sdma_platform_data {
> + int sdma_version;
> + char *cpu_name;
> + int to_version;
> +};
> +
> +#endif /* __MACH_MXC_SDMA_H__ */
> diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
> index 9520cf0..ff68307 100644
> --- a/drivers/dma/Kconfig
> +++ b/drivers/dma/Kconfig
> @@ -195,6 +195,14 @@ config PCH_DMA
> help
> Enable support for the Topcliff PCH DMA engine.
>
> +config IMX_SDMA
> + tristate "i.MX SDMA support"
> + depends on ARCH_MXC
> + select DMA_ENGINE
> + help
> + Support the i.MX SDMA engine. This engine is integrated into
> + Freescale i.MX25/31/35/51 chips.
> +
> config DMA_ENGINE
> bool
>
> diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
> index 72bd703..14d7a1b 100644
> --- a/drivers/dma/Makefile
> +++ b/drivers/dma/Makefile
> @@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
> obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
> obj-$(CONFIG_PL330_DMA) += pl330.o
> obj-$(CONFIG_PCH_DMA) += pch_dma.o
> +obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
> diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> new file mode 100644
> index 0000000..c447fc0
> --- /dev/null
> +++ b/drivers/dma/imx-sdma.c
> @@ -0,0 +1,1395 @@
> +/*
> + * drivers/dma/imx-sdma.c
> + *
> + * This file contains a driver for the Freescale Smart DMA engine
> + *
> + * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
> + *
> + * Based on code from Freescale:
> + *
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * The code contained herein is licensed under the GNU General Public
> + * License. You may obtain a copy of the GNU General Public License
> + * Version 2 or later at the following locations:
> + *
> + * http://www.opensource.org/licenses/gpl-license.html
> + * http://www.gnu.org/copyleft/gpl.html
> + */
> +
> +#include <linux/init.h>
> +#include <linux/types.h>
> +#include <linux/mm.h>
> +#include <linux/interrupt.h>
> +#include <linux/clk.h>
> +#include <linux/wait.h>
> +#include <linux/sched.h>
> +#include <linux/semaphore.h>
> +#include <linux/spinlock.h>
> +#include <linux/device.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/firmware.h>
> +#include <linux/slab.h>
> +#include <linux/platform_device.h>
> +#include <linux/dmaengine.h>
> +
> +#include <asm/irq.h>
> +#include <mach/sdma.h>
> +#include <mach/dma.h>
> +#include <mach/hardware.h>
> +
> +/* SDMA registers */
> +#define SDMA_H_C0PTR (sdma->regs + 0x000)
> +#define SDMA_H_INTR (sdma->regs + 0x004)
> +#define SDMA_H_STATSTOP (sdma->regs + 0x008)
> +#define SDMA_H_START (sdma->regs + 0x00c)
> +#define SDMA_H_EVTOVR (sdma->regs + 0x010)
> +#define SDMA_H_DSPOVR (sdma->regs + 0x014)
> +#define SDMA_H_HOSTOVR (sdma->regs + 0x018)
> +#define SDMA_H_EVTPEND (sdma->regs + 0x01c)
> +#define SDMA_H_DSPENBL (sdma->regs + 0x020)
> +#define SDMA_H_RESET (sdma->regs + 0x024)
> +#define SDMA_H_EVTERR (sdma->regs + 0x028)
> +#define SDMA_H_INTRMSK (sdma->regs + 0x02c)
> +#define SDMA_H_PSW (sdma->regs + 0x030)
> +#define SDMA_H_EVTERRDBG (sdma->regs + 0x034)
> +#define SDMA_H_CONFIG (sdma->regs + 0x038)
> +#define SDMA_ONCE_ENB (sdma->regs + 0x040)
> +#define SDMA_ONCE_DATA (sdma->regs + 0x044)
> +#define SDMA_ONCE_INSTR (sdma->regs + 0x048)
> +#define SDMA_ONCE_STAT (sdma->regs + 0x04c)
> +#define SDMA_ONCE_CMD (sdma->regs + 0x050)
> +#define SDMA_EVT_MIRROR (sdma->regs + 0x054)
> +#define SDMA_ILLINSTADDR (sdma->regs + 0x058)
> +#define SDMA_CHN0ADDR (sdma->regs + 0x05c)
> +#define SDMA_ONCE_RTB (sdma->regs + 0x060)
> +#define SDMA_XTRIG_CONF1 (sdma->regs + 0x070)
> +#define SDMA_XTRIG_CONF2 (sdma->regs + 0x074)
> +#define SDMA_CHNENBL_0 (sdma->regs + (sdma->version == 2 ? 0x200 : 0x80))
> +#define SDMA_CHNPRI_0 (sdma->regs + 0x100)
I'd prefer having an accessor function like
u32 sdma_read(struct whatever *sdma, unsigned offset)
{
return __raw_readl(sdma->regs + offset);
}
...
> +
> +/*
> + * Buffer descriptor status values.
> + */
> +#define BD_DONE 0x01
> +#define BD_WRAP 0x02
> +#define BD_CONT 0x04
> +#define BD_INTR 0x08
> +#define BD_RROR 0x10
> +#define BD_LAST 0x20
> +#define BD_EXTD 0x80
> +
> +/*
> + * Data Node descriptor status values.
> + */
> +#define DND_END_OF_FRAME 0x80
> +#define DND_END_OF_XFER 0x40
> +#define DND_DONE 0x20
> +#define DND_UNUSED 0x01
> +
> +/*
> + * IPCV2 descriptor status values.
> + */
> +#define BD_IPCV2_END_OF_FRAME 0x40
> +
> +#define IPCV2_MAX_NODES 50
> +/*
> + * Error bit set in the CCB status field by the SDMA,
> + * in setbd routine, in case of a transfer error
> + */
> +#define DATA_ERROR 0x10000000
> +
> +/*
> + * Buffer descriptor commands.
> + */
> +#define C0_ADDR 0x01
> +#define C0_LOAD 0x02
> +#define C0_DUMP 0x03
> +#define C0_SETCTX 0x07
> +#define C0_GETCTX 0x03
> +#define C0_SETDM 0x01
> +#define C0_SETPM 0x04
> +#define C0_GETDM 0x02
> +#define C0_GETPM 0x08
> +/*
> + * Change endianness indicator in the BD command field
> + */
> +#define CHANGE_ENDIANNESS 0x80
> +
> +/*
> + * Mode/Count of data node descriptors - IPCv2
> + */
> +struct sdma_mode_count {
> + u32 count : 16; /* size of the buffer pointed by this BD */
> + u32 status : 8; /* E,R,I,C,W,D status bits stored here */
> + u32 command : 8; /* command mostlky used for channel 0 */
s/mostlky/mostly/
> +};
> +
> +/*
> + * Buffer descriptor
> + */
> +struct sdma_buffer_descriptor {
> + struct sdma_mode_count mode;
> + dma_addr_t buffer_addr; /* address of the buffer described */
> + dma_addr_t ext_buffer_addr; /* extended buffer address */
> +} __attribute__ ((packed));
> +
> +/*
> + * Channel control Block
> + */
> +struct sdma_channel_control {
> + dma_addr_t current_bd_ptr; /* current buffer descriptor processed */
> + dma_addr_t base_bd_ptr; /* first element of buffer descriptor array */
> + u32 unused;
> + u32 unused1;
> +} __attribute__ ((packed));
> +
> +/**
> + * struct sdma_state_registers - SDMA context for a channel
> + *
> + * @pc: program counter
> + * @t: test bit: status of arithmetic & test instruction
> + * @rpc: return program counter
> + * @sf: source fault while loading data
> + * @spc: loop start program counter
> + * @df: destination fault while storing data
> + * @epc: loop end program counter
> + * @lm: loop mode
> + */
> +struct sdma_state_registers {
> + u32 pc :14;
> + u32 unused1: 1;
> + u32 t : 1;
> + u32 rpc :14;
> + u32 unused0: 1;
> + u32 sf : 1;
> + u32 spc :14;
> + u32 unused2: 1;
> + u32 df : 1;
> + u32 epc :14;
> + u32 lm : 2;
> +} __attribute__ ((packed));
I'm not sure what CodingStyle recommends, but I'd not add a space
between __attribute__ and (.
> +
> +/**
> + * struct sdma_context_data - sdma context specific to a channel
> + *
> + * @channel_state: channel state bits
> + * @gReg: general registers
> + * @mda: burst dma destination address register
> + * @msa: burst dma source address register
> + * @ms: burst dma status register
> + * @md: burst dma data register
> + * @pda: peripheral dma destination address register
> + * @psa: peripheral dma source address register
> + * @ps: peripheral dma status register
> + * @pd: peripheral dma data register
> + * @ca: CRC polynomial register
> + * @cs: CRC accumulator register
> + * @dda: dedicated core destination address register
> + * @dsa: dedicated core source address register
> + * @ds: dedicated core status register
> + * @dd: dedicated core data register
> + */
> +struct sdma_context_data {
> + struct sdma_state_registers channel_state;
> + u32 gReg[8];
> + u32 mda;
> + u32 msa;
> + u32 ms;
> + u32 md;
> + u32 pda;
> + u32 psa;
> + u32 ps;
> + u32 pd;
> + u32 ca;
> + u32 cs;
> + u32 dda;
> + u32 dsa;
> + u32 ds;
> + u32 dd;
> + u32 scratch0;
> + u32 scratch1;
> + u32 scratch2;
> + u32 scratch3;
> + u32 scratch4;
> + u32 scratch5;
> + u32 scratch6;
> + u32 scratch7;
> +} __attribute__ ((packed));
s/ / /?
> +
> +#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
> +
> +struct sdma_engine;
> +
> +/**
> + * struct sdma_channel - housekeeping for a SDMA channel
> + *
> + * @sdma pointer to the SDMA engine for this channel
> + * @channel the channel number, matches dmaengine chan_id
> + * @direction transfer type. Needed for setting SDMA script
> + * @peripheral_type Peripheral type. Needed for setting SDMA script
> + * @event_id aka dma request line
> + * @event_id2 for channels that use 2 events
> + * @word_size peripheral access size
> + * @buf_tail ID of the buffer that was processed
> + * @done channel completion
> + * @num_bd max NUM_BD. number of descriptors currently handling
> + */
> +struct sdma_channel {
> + struct sdma_engine *sdma;
> + unsigned int channel;
> + enum dma_data_direction direction;
> + enum sdma_peripheral_type peripheral_type;
> + unsigned int event_id;
> + unsigned int event_id2;
> + enum dma_slave_buswidth word_size;
> + unsigned int buf_tail;
> + struct completion done;
> + unsigned int num_bd;
> + struct sdma_buffer_descriptor *bd;
> + dma_addr_t bd_phys;
> + unsigned int pc_from_device, pc_to_device;
> + unsigned long flags;
> + dma_addr_t per_address;
> + u32 event_mask1, event_mask2;
> + u32 watermark_level;
> + u32 shp_addr, per_addr;
> + struct dma_chan chan;
> + spinlock_t lock;
> + struct dma_async_tx_descriptor desc;
> + dma_cookie_t last_completed;
> + enum dma_status status;
> +};
> +
> +#define IMX_DMA_SG_LOOP (1 << 0)
> +
> +#define MAX_DMA_CHANNELS 32
> +#define MXC_SDMA_DEFAULT_PRIORITY 1
> +#define MXC_SDMA_MIN_PRIORITY 1
> +#define MXC_SDMA_MAX_PRIORITY 7
> +
> +/*
> + * This enumerates transfer types
> + */
> +enum {
> + emi_2_per = 0, /* EMI memory to peripheral */
> + emi_2_int, /* EMI memory to internal RAM */
> + emi_2_emi, /* EMI memory to EMI memory */
> + emi_2_dsp, /* EMI memory to DSP memory */
> + per_2_int, /* Peripheral to internal RAM */
s/int/iram/ maybe?
> + per_2_emi, /* Peripheral to internal EMI memory */
> + per_2_dsp, /* Peripheral to DSP memory */
> + per_2_per, /* Peripheral to Peripheral */
> + int_2_per, /* Internal RAM to peripheral */
> + int_2_int, /* Internal RAM to Internal RAM */
> + int_2_emi, /* Internal RAM to EMI memory */
> + int_2_dsp, /* Internal RAM to DSP memory */
> + dsp_2_per, /* DSP memory to peripheral */
> + dsp_2_int, /* DSP memory to internal RAM */
> + dsp_2_emi, /* DSP memory to EMI memory */
> + dsp_2_dsp, /* DSP memory to DSP memory */
> + emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
> + dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
> + dvfs_pll, /* DVFS script with PLL change */
> + dvfs_pdr /* DVFS script without PLL change */
> +} sdma_transfer_type;
> +
> +/**
> + * struct sdma_script_start_addrs - SDMA script start pointers
> + *
> + * start addresses of the different functions in the physical
> + * address space of the SDMA engine.
> + */
> +struct sdma_script_start_addrs {
> + u32 ap_2_ap_addr;
> + u32 ap_2_bp_addr;
> + u32 ap_2_ap_fixed_addr;
> + u32 bp_2_ap_addr;
> + u32 loopback_on_dsp_side_addr;
> + u32 mcu_interrupt_only_addr;
> + u32 firi_2_per_addr;
> + u32 firi_2_mcu_addr;
> + u32 per_2_firi_addr;
> + u32 mcu_2_firi_addr;
> + u32 uart_2_per_addr;
> + u32 uart_2_mcu_addr;
> + u32 per_2_app_addr;
> + u32 mcu_2_app_addr;
> + u32 per_2_per_addr;
> + u32 uartsh_2_per_addr;
> + u32 uartsh_2_mcu_addr;
> + u32 per_2_shp_addr;
> + u32 mcu_2_shp_addr;
> + u32 ata_2_mcu_addr;
> + u32 mcu_2_ata_addr;
> + u32 app_2_per_addr;
> + u32 app_2_mcu_addr;
> + u32 shp_2_per_addr;
> + u32 shp_2_mcu_addr;
> + u32 mshc_2_mcu_addr;
> + u32 mcu_2_mshc_addr;
> + u32 spdif_2_mcu_addr;
> + u32 mcu_2_spdif_addr;
> + u32 asrc_2_mcu_addr;
> + u32 ext_mem_2_ipu_addr;
> + u32 descrambler_addr;
> + u32 dptc_dvfs_addr;
> + u32 utra_addr;
> + u32 ram_code_start_addr;
> +};
You didn't comment my suggestion to use an array here. Something like
struct {
const char *name;
u32 addr;
} start_addr[];
> +
> +#define SDMA_FIRMWARE_MAGIC 0x414d4453
> +
> +/**
> + * struct sdma_firmware_header - Layout of the firmware image
> + *
> + * @magic "SDMA"
> + * @version_major increased whenever layout of struct sdma_script_start_addrs
> + * changes.
> + * @version_minor firmware minor version (for binary compatible changes)
> + * @script_addrs_start offset of struct sdma_script_start_addrs in this image
> + * @num_script_addrs Number of script addresses in this image
> + * @ram_code_start offset of SDMA ram image in this firmware image
> + * @ram_code_size size of SDMA ram image
> + */
> +struct sdma_firmware_header {
> + u32 magic;
> + u32 version_major;
> + u32 version_minor;
> + u32 script_addrs_start;
> + u32 num_script_addrs;
> + u32 ram_code_start;
> + u32 ram_code_size;
> +};
> +
> +struct sdma_engine {
> + struct device *dev;
> + struct sdma_channel channel[MAX_DMA_CHANNELS];
> + struct sdma_channel_control *channel_control;
> + void __iomem *regs;
> + unsigned int version;
> + unsigned int num_events;
> + struct sdma_context_data *context;
> + dma_addr_t context_phys;
> + struct dma_device dma_device;
> + struct clk *clk;
> +};
> +
> +#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
> +#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
> +#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
> +#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
> +
> +static int sdma_config_ownership(struct sdma_channel *sdmac,
> + bool event_override, bool mcu_verride, bool dsp_override)
> +{
> + struct sdma_engine *sdma = sdmac->sdma;
> + int channel = sdmac->channel;
> + u32 evt, mcu, dsp;
> +
> + if (event_override && mcu_verride && dsp_override)
> + return -EINVAL;
> +
> + evt = readl(SDMA_H_EVTOVR);
> + mcu = readl(SDMA_H_HOSTOVR);
> + dsp = readl(SDMA_H_DSPOVR);
__raw_readl?
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | http://www.pengutronix.de/ |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-23 17:48 ` Uwe Kleine-König
0 siblings, 0 replies; 78+ messages in thread
From: Uwe Kleine-König @ 2010-08-23 17:48 UTC (permalink / raw)
To: linux-arm-kernel
Hallo Sascha,
On Mon, Aug 23, 2010 at 02:57:04PM +0200, Sascha Hauer wrote:
>
> This patch adds support for the Freescale i.MX SDMA engine.
>
> The SDMA engine is a scatter/gather DMA engine which is implemented
> as a seperate coprocessor. SDMA needs its own firmware which is
> requested using the standard request_firmware mechanism. The firmware
> has different entry points for each peripheral type, so drivers
> have to pass the peripheral type to the DMA engine which in turn
> picks the correct firmware entry point from a table contained in
> the firmware image itself.
> The original Freescale code also supports support for transfering
> data to the internal SRAM which needs different entry points to
> the firmware. Support for this is currently not implemented. Also,
> support for the ASRC (asymmetric sample rate converter) is skipped.
>
> I took a very simple approach to implement dmaengine support. Only
> a single descriptor is statically assigned to a each channel. This
> means that transfers can't be queued up but only a single transfer
> is in progress. This simplifies implementation a lot and is sufficient
> for the usual device/memory transfers.
>
> Changes since v1:
>
> - included comments from Linus Walleij
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
> arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
> arch/arm/mach-mx3/Kconfig | 2 +
> arch/arm/plat-mxc/Kconfig | 10 +
> arch/arm/plat-mxc/include/mach/dma.h | 64 ++
> arch/arm/plat-mxc/include/mach/sdma.h | 17 +
> drivers/dma/Kconfig | 8 +
> drivers/dma/Makefile | 1 +
> drivers/dma/imx-sdma.c | 1395 +++++++++++++++++++++++++++++++
> 8 files changed, 1499 insertions(+), 6 deletions(-)
> create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
> create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
> create mode 100644 drivers/dma/imx-sdma.c
>
> diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
> index 287431c..ac6fd71 100644
> --- a/arch/arm/mach-imx/include/mach/dma-v1.h
> +++ b/arch/arm/mach-imx/include/mach/dma-v1.h
> @@ -27,6 +27,8 @@
>
> #define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
>
> +#include <mach/dma.h>
> +
> #define IMX_DMA_CHANNELS 16
>
> #define DMA_MODE_READ 0
> @@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
>
> void imx_dma_free(int channel);
>
> -enum imx_dma_prio {
> - DMA_PRIO_HIGH = 0,
> - DMA_PRIO_MEDIUM = 1,
> - DMA_PRIO_LOW = 2
> -};
> -
> int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
>
> #endif /* __MACH_DMA_V1_H__ */
> diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
> index 85beece..301375c 100644
> --- a/arch/arm/mach-mx3/Kconfig
> +++ b/arch/arm/mach-mx3/Kconfig
> @@ -3,12 +3,14 @@ if ARCH_MX3
> config ARCH_MX31
> select ARCH_HAS_RNGA
> select ARCH_MXC_AUDMUX_V2
> + select IMX_HAVE_SDMA
> bool
>
> config ARCH_MX35
> bool
> select ARCH_MXC_IOMUX_V3
> select ARCH_MXC_AUDMUX_V2
> + select IMX_HAVE_SDMA
>
> comment "MX3 platforms:"
>
> diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
> index 0527e65..6741625 100644
> --- a/arch/arm/plat-mxc/Kconfig
> +++ b/arch/arm/plat-mxc/Kconfig
> @@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
> config ARCH_MXC_AUDMUX_V2
> bool
>
> +config IMX_HAVE_SDMA
> + bool
> +
> +config IMX_SDMA
> + depends on IMX_HAVE_SDMA
> + tristate "Enable SDMA support"
> + help
> + Include support for the SDMA engine. The SDMA engine needs additional
> + firmware support. SDMA can be compiled as a module to support loading
> + the firmware when a rootfs is present.
> endif
> diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
> new file mode 100644
> index 0000000..69d181f
> --- /dev/null
> +++ b/arch/arm/plat-mxc/include/mach/dma.h
> @@ -0,0 +1,64 @@
> +/*
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 as
> + * published by the Free Software Foundation.
> + */
> +
> +#ifndef __ASM_ARCH_MXC_DMA_H__
> +#define __ASM_ARCH_MXC_DMA_H__
__MACH_DMA_H__ please
> +
> +#include <linux/scatterlist.h>
> +
> +/*
> + * This enumerates peripheral types. Used for SDMA.
> + */
> +enum sdma_peripheral_type {
> + IMX_DMATYPE_SSI, /* MCU domain SSI */
> + IMX_DMATYPE_SSI_SP, /* Shared SSI */
> + IMX_DMATYPE_MMC, /* MMC */
> + IMX_DMATYPE_SDHC, /* SDHC */
> + IMX_DMATYPE_UART, /* MCU domain UART */
> + IMX_DMATYPE_UART_SP, /* Shared UART */
> + IMX_DMATYPE_FIRI, /* FIRI */
> + IMX_DMATYPE_CSPI, /* MCU domain CSPI */
> + IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
> + IMX_DMATYPE_SIM, /* SIM */
> + IMX_DMATYPE_ATA, /* ATA */
> + IMX_DMATYPE_CCM, /* CCM */
> + IMX_DMATYPE_EXT, /* External peripheral */
> + IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
> + IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
> + IMX_DMATYPE_DSP, /* DSP */
> + IMX_DMATYPE_MEMORY, /* Memory */
> + IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
> + IMX_DMATYPE_SPDIF, /* SPDIF */
> + IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
> + IMX_DMATYPE_ASRC, /* ASRC */
> + IMX_DMATYPE_ESAI, /* ESAI */
> +};
> +
> +enum imx_dma_prio {
> + DMA_PRIO_HIGH = 0,
> + DMA_PRIO_MEDIUM = 1,
> + DMA_PRIO_LOW = 2
> +};
> +
> +struct imx_dma_data {
> + int dma_request; /* DMA request line */
> + enum sdma_peripheral_type peripheral_type;
> + int priority;
> +};
> +
> +static inline int imx_dma_is_ipu(struct dma_chan *chan)
> +{
> + return !strcmp(dev_name(chan->device->dev), "ipu-core");
> +}
> +
> +static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
> +{
> + return !strcmp(dev_name(chan->device->dev), "imx-sdma");
> +}
> +
> +#endif
> diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
> new file mode 100644
> index 0000000..9be1122
> --- /dev/null
> +++ b/arch/arm/plat-mxc/include/mach/sdma.h
> @@ -0,0 +1,17 @@
> +#ifndef __MACH_MXC_SDMA_H__
> +#define __MACH_MXC_SDMA_H__
__MACH_SDMA_H__
> +
> +/**
> + * struct sdma_platform_data - platform specific data for SDMA engine
> + *
> + * @sdma_version The version of this SDMA engine
> + * @cpu_name used to generate the firmware name
> + * @to_version CPU Tape out version
> + */
> +struct sdma_platform_data {
> + int sdma_version;
> + char *cpu_name;
> + int to_version;
> +};
> +
> +#endif /* __MACH_MXC_SDMA_H__ */
> diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
> index 9520cf0..ff68307 100644
> --- a/drivers/dma/Kconfig
> +++ b/drivers/dma/Kconfig
> @@ -195,6 +195,14 @@ config PCH_DMA
> help
> Enable support for the Topcliff PCH DMA engine.
>
> +config IMX_SDMA
> + tristate "i.MX SDMA support"
> + depends on ARCH_MXC
> + select DMA_ENGINE
> + help
> + Support the i.MX SDMA engine. This engine is integrated into
> + Freescale i.MX25/31/35/51 chips.
> +
> config DMA_ENGINE
> bool
>
> diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
> index 72bd703..14d7a1b 100644
> --- a/drivers/dma/Makefile
> +++ b/drivers/dma/Makefile
> @@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
> obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
> obj-$(CONFIG_PL330_DMA) += pl330.o
> obj-$(CONFIG_PCH_DMA) += pch_dma.o
> +obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
> diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> new file mode 100644
> index 0000000..c447fc0
> --- /dev/null
> +++ b/drivers/dma/imx-sdma.c
> @@ -0,0 +1,1395 @@
> +/*
> + * drivers/dma/imx-sdma.c
> + *
> + * This file contains a driver for the Freescale Smart DMA engine
> + *
> + * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
> + *
> + * Based on code from Freescale:
> + *
> + * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
> + *
> + * The code contained herein is licensed under the GNU General Public
> + * License. You may obtain a copy of the GNU General Public License
> + * Version 2 or later at the following locations:
> + *
> + * http://www.opensource.org/licenses/gpl-license.html
> + * http://www.gnu.org/copyleft/gpl.html
> + */
> +
> +#include <linux/init.h>
> +#include <linux/types.h>
> +#include <linux/mm.h>
> +#include <linux/interrupt.h>
> +#include <linux/clk.h>
> +#include <linux/wait.h>
> +#include <linux/sched.h>
> +#include <linux/semaphore.h>
> +#include <linux/spinlock.h>
> +#include <linux/device.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/firmware.h>
> +#include <linux/slab.h>
> +#include <linux/platform_device.h>
> +#include <linux/dmaengine.h>
> +
> +#include <asm/irq.h>
> +#include <mach/sdma.h>
> +#include <mach/dma.h>
> +#include <mach/hardware.h>
> +
> +/* SDMA registers */
> +#define SDMA_H_C0PTR (sdma->regs + 0x000)
> +#define SDMA_H_INTR (sdma->regs + 0x004)
> +#define SDMA_H_STATSTOP (sdma->regs + 0x008)
> +#define SDMA_H_START (sdma->regs + 0x00c)
> +#define SDMA_H_EVTOVR (sdma->regs + 0x010)
> +#define SDMA_H_DSPOVR (sdma->regs + 0x014)
> +#define SDMA_H_HOSTOVR (sdma->regs + 0x018)
> +#define SDMA_H_EVTPEND (sdma->regs + 0x01c)
> +#define SDMA_H_DSPENBL (sdma->regs + 0x020)
> +#define SDMA_H_RESET (sdma->regs + 0x024)
> +#define SDMA_H_EVTERR (sdma->regs + 0x028)
> +#define SDMA_H_INTRMSK (sdma->regs + 0x02c)
> +#define SDMA_H_PSW (sdma->regs + 0x030)
> +#define SDMA_H_EVTERRDBG (sdma->regs + 0x034)
> +#define SDMA_H_CONFIG (sdma->regs + 0x038)
> +#define SDMA_ONCE_ENB (sdma->regs + 0x040)
> +#define SDMA_ONCE_DATA (sdma->regs + 0x044)
> +#define SDMA_ONCE_INSTR (sdma->regs + 0x048)
> +#define SDMA_ONCE_STAT (sdma->regs + 0x04c)
> +#define SDMA_ONCE_CMD (sdma->regs + 0x050)
> +#define SDMA_EVT_MIRROR (sdma->regs + 0x054)
> +#define SDMA_ILLINSTADDR (sdma->regs + 0x058)
> +#define SDMA_CHN0ADDR (sdma->regs + 0x05c)
> +#define SDMA_ONCE_RTB (sdma->regs + 0x060)
> +#define SDMA_XTRIG_CONF1 (sdma->regs + 0x070)
> +#define SDMA_XTRIG_CONF2 (sdma->regs + 0x074)
> +#define SDMA_CHNENBL_0 (sdma->regs + (sdma->version == 2 ? 0x200 : 0x80))
> +#define SDMA_CHNPRI_0 (sdma->regs + 0x100)
I'd prefer having an accessor function like
u32 sdma_read(struct whatever *sdma, unsigned offset)
{
return __raw_readl(sdma->regs + offset);
}
...
> +
> +/*
> + * Buffer descriptor status values.
> + */
> +#define BD_DONE 0x01
> +#define BD_WRAP 0x02
> +#define BD_CONT 0x04
> +#define BD_INTR 0x08
> +#define BD_RROR 0x10
> +#define BD_LAST 0x20
> +#define BD_EXTD 0x80
> +
> +/*
> + * Data Node descriptor status values.
> + */
> +#define DND_END_OF_FRAME 0x80
> +#define DND_END_OF_XFER 0x40
> +#define DND_DONE 0x20
> +#define DND_UNUSED 0x01
> +
> +/*
> + * IPCV2 descriptor status values.
> + */
> +#define BD_IPCV2_END_OF_FRAME 0x40
> +
> +#define IPCV2_MAX_NODES 50
> +/*
> + * Error bit set in the CCB status field by the SDMA,
> + * in setbd routine, in case of a transfer error
> + */
> +#define DATA_ERROR 0x10000000
> +
> +/*
> + * Buffer descriptor commands.
> + */
> +#define C0_ADDR 0x01
> +#define C0_LOAD 0x02
> +#define C0_DUMP 0x03
> +#define C0_SETCTX 0x07
> +#define C0_GETCTX 0x03
> +#define C0_SETDM 0x01
> +#define C0_SETPM 0x04
> +#define C0_GETDM 0x02
> +#define C0_GETPM 0x08
> +/*
> + * Change endianness indicator in the BD command field
> + */
> +#define CHANGE_ENDIANNESS 0x80
> +
> +/*
> + * Mode/Count of data node descriptors - IPCv2
> + */
> +struct sdma_mode_count {
> + u32 count : 16; /* size of the buffer pointed by this BD */
> + u32 status : 8; /* E,R,I,C,W,D status bits stored here */
> + u32 command : 8; /* command mostlky used for channel 0 */
s/mostlky/mostly/
> +};
> +
> +/*
> + * Buffer descriptor
> + */
> +struct sdma_buffer_descriptor {
> + struct sdma_mode_count mode;
> + dma_addr_t buffer_addr; /* address of the buffer described */
> + dma_addr_t ext_buffer_addr; /* extended buffer address */
> +} __attribute__ ((packed));
> +
> +/*
> + * Channel control Block
> + */
> +struct sdma_channel_control {
> + dma_addr_t current_bd_ptr; /* current buffer descriptor processed */
> + dma_addr_t base_bd_ptr; /* first element of buffer descriptor array */
> + u32 unused;
> + u32 unused1;
> +} __attribute__ ((packed));
> +
> +/**
> + * struct sdma_state_registers - SDMA context for a channel
> + *
> + * @pc: program counter
> + * @t: test bit: status of arithmetic & test instruction
> + * @rpc: return program counter
> + * @sf: source fault while loading data
> + * @spc: loop start program counter
> + * @df: destination fault while storing data
> + * @epc: loop end program counter
> + * @lm: loop mode
> + */
> +struct sdma_state_registers {
> + u32 pc :14;
> + u32 unused1: 1;
> + u32 t : 1;
> + u32 rpc :14;
> + u32 unused0: 1;
> + u32 sf : 1;
> + u32 spc :14;
> + u32 unused2: 1;
> + u32 df : 1;
> + u32 epc :14;
> + u32 lm : 2;
> +} __attribute__ ((packed));
I'm not sure what CodingStyle recommends, but I'd not add a space
between __attribute__ and (.
> +
> +/**
> + * struct sdma_context_data - sdma context specific to a channel
> + *
> + * @channel_state: channel state bits
> + * @gReg: general registers
> + * @mda: burst dma destination address register
> + * @msa: burst dma source address register
> + * @ms: burst dma status register
> + * @md: burst dma data register
> + * @pda: peripheral dma destination address register
> + * @psa: peripheral dma source address register
> + * @ps: peripheral dma status register
> + * @pd: peripheral dma data register
> + * @ca: CRC polynomial register
> + * @cs: CRC accumulator register
> + * @dda: dedicated core destination address register
> + * @dsa: dedicated core source address register
> + * @ds: dedicated core status register
> + * @dd: dedicated core data register
> + */
> +struct sdma_context_data {
> + struct sdma_state_registers channel_state;
> + u32 gReg[8];
> + u32 mda;
> + u32 msa;
> + u32 ms;
> + u32 md;
> + u32 pda;
> + u32 psa;
> + u32 ps;
> + u32 pd;
> + u32 ca;
> + u32 cs;
> + u32 dda;
> + u32 dsa;
> + u32 ds;
> + u32 dd;
> + u32 scratch0;
> + u32 scratch1;
> + u32 scratch2;
> + u32 scratch3;
> + u32 scratch4;
> + u32 scratch5;
> + u32 scratch6;
> + u32 scratch7;
> +} __attribute__ ((packed));
s/ / /?
> +
> +#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
> +
> +struct sdma_engine;
> +
> +/**
> + * struct sdma_channel - housekeeping for a SDMA channel
> + *
> + * @sdma pointer to the SDMA engine for this channel
> + * @channel the channel number, matches dmaengine chan_id
> + * @direction transfer type. Needed for setting SDMA script
> + * @peripheral_type Peripheral type. Needed for setting SDMA script
> + * @event_id aka dma request line
> + * @event_id2 for channels that use 2 events
> + * @word_size peripheral access size
> + * @buf_tail ID of the buffer that was processed
> + * @done channel completion
> + * @num_bd max NUM_BD. number of descriptors currently handling
> + */
> +struct sdma_channel {
> + struct sdma_engine *sdma;
> + unsigned int channel;
> + enum dma_data_direction direction;
> + enum sdma_peripheral_type peripheral_type;
> + unsigned int event_id;
> + unsigned int event_id2;
> + enum dma_slave_buswidth word_size;
> + unsigned int buf_tail;
> + struct completion done;
> + unsigned int num_bd;
> + struct sdma_buffer_descriptor *bd;
> + dma_addr_t bd_phys;
> + unsigned int pc_from_device, pc_to_device;
> + unsigned long flags;
> + dma_addr_t per_address;
> + u32 event_mask1, event_mask2;
> + u32 watermark_level;
> + u32 shp_addr, per_addr;
> + struct dma_chan chan;
> + spinlock_t lock;
> + struct dma_async_tx_descriptor desc;
> + dma_cookie_t last_completed;
> + enum dma_status status;
> +};
> +
> +#define IMX_DMA_SG_LOOP (1 << 0)
> +
> +#define MAX_DMA_CHANNELS 32
> +#define MXC_SDMA_DEFAULT_PRIORITY 1
> +#define MXC_SDMA_MIN_PRIORITY 1
> +#define MXC_SDMA_MAX_PRIORITY 7
> +
> +/*
> + * This enumerates transfer types
> + */
> +enum {
> + emi_2_per = 0, /* EMI memory to peripheral */
> + emi_2_int, /* EMI memory to internal RAM */
> + emi_2_emi, /* EMI memory to EMI memory */
> + emi_2_dsp, /* EMI memory to DSP memory */
> + per_2_int, /* Peripheral to internal RAM */
s/int/iram/ maybe?
> + per_2_emi, /* Peripheral to internal EMI memory */
> + per_2_dsp, /* Peripheral to DSP memory */
> + per_2_per, /* Peripheral to Peripheral */
> + int_2_per, /* Internal RAM to peripheral */
> + int_2_int, /* Internal RAM to Internal RAM */
> + int_2_emi, /* Internal RAM to EMI memory */
> + int_2_dsp, /* Internal RAM to DSP memory */
> + dsp_2_per, /* DSP memory to peripheral */
> + dsp_2_int, /* DSP memory to internal RAM */
> + dsp_2_emi, /* DSP memory to EMI memory */
> + dsp_2_dsp, /* DSP memory to DSP memory */
> + emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
> + dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
> + dvfs_pll, /* DVFS script with PLL change */
> + dvfs_pdr /* DVFS script without PLL change */
> +} sdma_transfer_type;
> +
> +/**
> + * struct sdma_script_start_addrs - SDMA script start pointers
> + *
> + * start addresses of the different functions in the physical
> + * address space of the SDMA engine.
> + */
> +struct sdma_script_start_addrs {
> + u32 ap_2_ap_addr;
> + u32 ap_2_bp_addr;
> + u32 ap_2_ap_fixed_addr;
> + u32 bp_2_ap_addr;
> + u32 loopback_on_dsp_side_addr;
> + u32 mcu_interrupt_only_addr;
> + u32 firi_2_per_addr;
> + u32 firi_2_mcu_addr;
> + u32 per_2_firi_addr;
> + u32 mcu_2_firi_addr;
> + u32 uart_2_per_addr;
> + u32 uart_2_mcu_addr;
> + u32 per_2_app_addr;
> + u32 mcu_2_app_addr;
> + u32 per_2_per_addr;
> + u32 uartsh_2_per_addr;
> + u32 uartsh_2_mcu_addr;
> + u32 per_2_shp_addr;
> + u32 mcu_2_shp_addr;
> + u32 ata_2_mcu_addr;
> + u32 mcu_2_ata_addr;
> + u32 app_2_per_addr;
> + u32 app_2_mcu_addr;
> + u32 shp_2_per_addr;
> + u32 shp_2_mcu_addr;
> + u32 mshc_2_mcu_addr;
> + u32 mcu_2_mshc_addr;
> + u32 spdif_2_mcu_addr;
> + u32 mcu_2_spdif_addr;
> + u32 asrc_2_mcu_addr;
> + u32 ext_mem_2_ipu_addr;
> + u32 descrambler_addr;
> + u32 dptc_dvfs_addr;
> + u32 utra_addr;
> + u32 ram_code_start_addr;
> +};
You didn't comment my suggestion to use an array here. Something like
struct {
const char *name;
u32 addr;
} start_addr[];
> +
> +#define SDMA_FIRMWARE_MAGIC 0x414d4453
> +
> +/**
> + * struct sdma_firmware_header - Layout of the firmware image
> + *
> + * @magic "SDMA"
> + * @version_major increased whenever layout of struct sdma_script_start_addrs
> + * changes.
> + * @version_minor firmware minor version (for binary compatible changes)
> + * @script_addrs_start offset of struct sdma_script_start_addrs in this image
> + * @num_script_addrs Number of script addresses in this image
> + * @ram_code_start offset of SDMA ram image in this firmware image
> + * @ram_code_size size of SDMA ram image
> + */
> +struct sdma_firmware_header {
> + u32 magic;
> + u32 version_major;
> + u32 version_minor;
> + u32 script_addrs_start;
> + u32 num_script_addrs;
> + u32 ram_code_start;
> + u32 ram_code_size;
> +};
> +
> +struct sdma_engine {
> + struct device *dev;
> + struct sdma_channel channel[MAX_DMA_CHANNELS];
> + struct sdma_channel_control *channel_control;
> + void __iomem *regs;
> + unsigned int version;
> + unsigned int num_events;
> + struct sdma_context_data *context;
> + dma_addr_t context_phys;
> + struct dma_device dma_device;
> + struct clk *clk;
> +};
> +
> +#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
> +#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
> +#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
> +#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
> +
> +static int sdma_config_ownership(struct sdma_channel *sdmac,
> + bool event_override, bool mcu_verride, bool dsp_override)
> +{
> + struct sdma_engine *sdma = sdmac->sdma;
> + int channel = sdmac->channel;
> + u32 evt, mcu, dsp;
> +
> + if (event_override && mcu_verride && dsp_override)
> + return -EINVAL;
> +
> + evt = readl(SDMA_H_EVTOVR);
> + mcu = readl(SDMA_H_HOSTOVR);
> + dsp = readl(SDMA_H_DSPOVR);
__raw_readl?
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-K?nig |
Industrial Linux Solutions | http://www.pengutronix.de/ |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-23 17:30 ` Linus Walleij
@ 2010-08-24 6:58 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-24 6:58 UTC (permalink / raw)
To: Linus Walleij; +Cc: linux-kernel, Dan Williams, linux-arm-kernel
On Mon, Aug 23, 2010 at 07:30:34PM +0200, Linus Walleij wrote:
> 2010/8/23 Sascha Hauer <s.hauer@pengutronix.de>:
>
> > This patch adds support for the Freescale i.MX SDMA engine.
>
> Great progress!
>
> > (...)
> > diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> > (...)
> > +/* SDMA registers */
> > +#define SDMA_H_C0PTR (sdma->regs + 0x000)
> > +#define SDMA_H_INTR (sdma->regs + 0x004)
> > +#define SDMA_H_STATSTOP (sdma->regs + 0x008)
> > +#define SDMA_H_START (sdma->regs + 0x00c)
> > +#define SDMA_H_EVTOVR (sdma->regs + 0x010)
> > +#define SDMA_H_DSPOVR (sdma->regs + 0x014)
> > +#define SDMA_H_HOSTOVR (sdma->regs + 0x018)
> > +#define SDMA_H_EVTPEND (sdma->regs + 0x01c)
> > +#define SDMA_H_DSPENBL (sdma->regs + 0x020)
> > +#define SDMA_H_RESET (sdma->regs + 0x024)
> > +#define SDMA_H_EVTERR (sdma->regs + 0x028)
> > +#define SDMA_H_INTRMSK (sdma->regs + 0x02c)
> > +#define SDMA_H_PSW (sdma->regs + 0x030)
> > +#define SDMA_H_EVTERRDBG (sdma->regs + 0x034)
> > +#define SDMA_H_CONFIG (sdma->regs + 0x038)
> > +#define SDMA_ONCE_ENB (sdma->regs + 0x040)
> > +#define SDMA_ONCE_DATA (sdma->regs + 0x044)
> > +#define SDMA_ONCE_INSTR (sdma->regs + 0x048)
> > +#define SDMA_ONCE_STAT (sdma->regs + 0x04c)
> > +#define SDMA_ONCE_CMD (sdma->regs + 0x050)
> > +#define SDMA_EVT_MIRROR (sdma->regs + 0x054)
> > +#define SDMA_ILLINSTADDR (sdma->regs + 0x058)
> > +#define SDMA_CHN0ADDR (sdma->regs + 0x05c)
> > +#define SDMA_ONCE_RTB (sdma->regs + 0x060)
> > +#define SDMA_XTRIG_CONF1 (sdma->regs + 0x070)
> > +#define SDMA_XTRIG_CONF2 (sdma->regs + 0x074)
> > +#define SDMA_CHNENBL_0 (sdma->regs + (sdma->version == 2 ? 0x200 : 0x80))
> > +#define SDMA_CHNPRI_0 (sdma->regs + 0x100)
>
> These macros expand to the local variable "sdma" which must
> be present in all functions using them. I don't know what is
> considered moste readable, but I would certainly just
>
> #define SDMA_FOO (0x0123)
> (...)
> u32 foo = readl(sdma->regs + SDMA_FOO);
>
> That is more common I think.
> > +
> > +/*
> > + * This enumerates transfer types
> > + */
> > +enum {
> > + emi_2_per = 0, /* EMI memory to peripheral */
> > + emi_2_int, /* EMI memory to internal RAM */
> > + emi_2_emi, /* EMI memory to EMI memory */
> > + emi_2_dsp, /* EMI memory to DSP memory */
> > + per_2_int, /* Peripheral to internal RAM */
> > + per_2_emi, /* Peripheral to internal EMI memory */
> > + per_2_dsp, /* Peripheral to DSP memory */
> > + per_2_per, /* Peripheral to Peripheral */
> > + int_2_per, /* Internal RAM to peripheral */
> > + int_2_int, /* Internal RAM to Internal RAM */
> > + int_2_emi, /* Internal RAM to EMI memory */
> > + int_2_dsp, /* Internal RAM to DSP memory */
> > + dsp_2_per, /* DSP memory to peripheral */
> > + dsp_2_int, /* DSP memory to internal RAM */
> > + dsp_2_emi, /* DSP memory to EMI memory */
> > + dsp_2_dsp, /* DSP memory to DSP memory */
> > + emi_2_dsp_loop, /* EMI memory to DSP memory loopback */
> > + dsp_2_emi_loop, /* DSP memory to EMI memory loopback */
> > + dvfs_pll, /* DVFS script with PLL change */
> > + dvfs_pdr /* DVFS script without PLL change */
> > +} sdma_transfer_type;
>
> Picky me, but it's no type, its an enum. I understand that it is
> a technical term...
>
> What about just calling is sdma_transfer? Short and nice.
> Or sdma_transfer_line?
This turned out to be unused anyway, so the simple fix was to remove it.
>
> > (...)
> > +/*
> > + * Stores the start address of the SDMA scripts
> > + */
> > +static struct sdma_script_start_addrs __sdma_script_addrs;
> > +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
>
> What's the rationale behind prefixing that variable with __?
>
> The same name for struct and variable is perfectly viable.
The rationale was to statically allocate a struct
sdma_script_start_addrs and create a pointer to it so that I don't have
to use &__sdma_script_addrs in the code.
I forgot this one while converting the driver to multi instance, so this
is now part of struct sdma_engine.
Fixed the other stuff aswell, I will send an update shortly.
Regards,
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-24 6:58 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-24 6:58 UTC (permalink / raw)
To: linux-arm-kernel
On Mon, Aug 23, 2010 at 07:30:34PM +0200, Linus Walleij wrote:
> 2010/8/23 Sascha Hauer <s.hauer@pengutronix.de>:
>
> > This patch adds support for the Freescale i.MX SDMA engine.
>
> Great progress!
>
> > (...)
> > diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
> > (...)
> > +/* SDMA registers */
> > +#define SDMA_H_C0PTR ? ? ? ? ? (sdma->regs + 0x000)
> > +#define SDMA_H_INTR ? ? ? ? ? ?(sdma->regs + 0x004)
> > +#define SDMA_H_STATSTOP ? ? ? ? ? ? ? ?(sdma->regs + 0x008)
> > +#define SDMA_H_START ? ? ? ? ? (sdma->regs + 0x00c)
> > +#define SDMA_H_EVTOVR ? ? ? ? ?(sdma->regs + 0x010)
> > +#define SDMA_H_DSPOVR ? ? ? ? ?(sdma->regs + 0x014)
> > +#define SDMA_H_HOSTOVR ? ? ? ? (sdma->regs + 0x018)
> > +#define SDMA_H_EVTPEND ? ? ? ? (sdma->regs + 0x01c)
> > +#define SDMA_H_DSPENBL ? ? ? ? (sdma->regs + 0x020)
> > +#define SDMA_H_RESET ? ? ? ? ? (sdma->regs + 0x024)
> > +#define SDMA_H_EVTERR ? ? ? ? ?(sdma->regs + 0x028)
> > +#define SDMA_H_INTRMSK ? ? ? ? (sdma->regs + 0x02c)
> > +#define SDMA_H_PSW ? ? ? ? ? ? (sdma->regs + 0x030)
> > +#define SDMA_H_EVTERRDBG ? ? ? (sdma->regs + 0x034)
> > +#define SDMA_H_CONFIG ? ? ? ? ?(sdma->regs + 0x038)
> > +#define SDMA_ONCE_ENB ? ? ? ? ?(sdma->regs + 0x040)
> > +#define SDMA_ONCE_DATA ? ? ? ? (sdma->regs + 0x044)
> > +#define SDMA_ONCE_INSTR ? ? ? ? ? ? ? ?(sdma->regs + 0x048)
> > +#define SDMA_ONCE_STAT ? ? ? ? (sdma->regs + 0x04c)
> > +#define SDMA_ONCE_CMD ? ? ? ? ?(sdma->regs + 0x050)
> > +#define SDMA_EVT_MIRROR ? ? ? ? ? ? ? ?(sdma->regs + 0x054)
> > +#define SDMA_ILLINSTADDR ? ? ? (sdma->regs + 0x058)
> > +#define SDMA_CHN0ADDR ? ? ? ? ?(sdma->regs + 0x05c)
> > +#define SDMA_ONCE_RTB ? ? ? ? ?(sdma->regs + 0x060)
> > +#define SDMA_XTRIG_CONF1 ? ? ? (sdma->regs + 0x070)
> > +#define SDMA_XTRIG_CONF2 ? ? ? (sdma->regs + 0x074)
> > +#define SDMA_CHNENBL_0 ? ? ? ? (sdma->regs + (sdma->version == 2 ? 0x200 : 0x80))
> > +#define SDMA_CHNPRI_0 ? ? ? ? ?(sdma->regs + 0x100)
>
> These macros expand to the local variable "sdma" which must
> be present in all functions using them. I don't know what is
> considered moste readable, but I would certainly just
>
> #define SDMA_FOO (0x0123)
> (...)
> u32 foo = readl(sdma->regs + SDMA_FOO);
>
> That is more common I think.
> > +
> > +/*
> > + * This enumerates transfer types
> > + */
> > +enum {
> > + ? ? ? emi_2_per = 0, ? ? ? ? ?/* EMI memory to peripheral */
> > + ? ? ? emi_2_int, ? ? ? ? ? ? ?/* EMI memory to internal RAM */
> > + ? ? ? emi_2_emi, ? ? ? ? ? ? ?/* EMI memory to EMI memory */
> > + ? ? ? emi_2_dsp, ? ? ? ? ? ? ?/* EMI memory to DSP memory */
> > + ? ? ? per_2_int, ? ? ? ? ? ? ?/* Peripheral to internal RAM */
> > + ? ? ? per_2_emi, ? ? ? ? ? ? ?/* Peripheral to internal EMI memory */
> > + ? ? ? per_2_dsp, ? ? ? ? ? ? ?/* Peripheral to DSP memory */
> > + ? ? ? per_2_per, ? ? ? ? ? ? ?/* Peripheral to Peripheral */
> > + ? ? ? int_2_per, ? ? ? ? ? ? ?/* Internal RAM to peripheral */
> > + ? ? ? int_2_int, ? ? ? ? ? ? ?/* Internal RAM to Internal RAM */
> > + ? ? ? int_2_emi, ? ? ? ? ? ? ?/* Internal RAM to EMI memory */
> > + ? ? ? int_2_dsp, ? ? ? ? ? ? ?/* Internal RAM to DSP memory */
> > + ? ? ? dsp_2_per, ? ? ? ? ? ? ?/* DSP memory to peripheral */
> > + ? ? ? dsp_2_int, ? ? ? ? ? ? ?/* DSP memory to internal RAM */
> > + ? ? ? dsp_2_emi, ? ? ? ? ? ? ?/* DSP memory to EMI memory */
> > + ? ? ? dsp_2_dsp, ? ? ? ? ? ? ?/* DSP memory to DSP memory */
> > + ? ? ? emi_2_dsp_loop, ? ? ? ? /* EMI memory to DSP memory loopback */
> > + ? ? ? dsp_2_emi_loop, ? ? ? ? /* DSP memory to EMI memory loopback */
> > + ? ? ? dvfs_pll, ? ? ? ? ? ? ? /* DVFS script with PLL change ? ? ? */
> > + ? ? ? dvfs_pdr ? ? ? ? ? ? ? ?/* DVFS script without PLL change ? ?*/
> > +} sdma_transfer_type;
>
> Picky me, but it's no type, its an enum. I understand that it is
> a technical term...
>
> What about just calling is sdma_transfer? Short and nice.
> Or sdma_transfer_line?
This turned out to be unused anyway, so the simple fix was to remove it.
>
> > (...)
> > +/*
> > + * Stores the start address of the SDMA scripts
> > + */
> > +static struct sdma_script_start_addrs __sdma_script_addrs;
> > +static struct sdma_script_start_addrs *sdma_script_addrs = &__sdma_script_addrs;
>
> What's the rationale behind prefixing that variable with __?
>
> The same name for struct and variable is perfectly viable.
The rationale was to statically allocate a struct
sdma_script_start_addrs and create a pointer to it so that I don't have
to use &__sdma_script_addrs in the code.
I forgot this one while converting the driver to multi instance, so this
is now part of struct sdma_engine.
Fixed the other stuff aswell, I will send an update shortly.
Regards,
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v3] dmaengine: Add Freescale i.MX SDMA support
2010-08-16 12:21 ` Linus Walleij
@ 2010-08-24 7:10 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-24 7:10 UTC (permalink / raw)
To: Linus Walleij; +Cc: linux-kernel, Dan Williams, linux-arm-kernel
This patch adds support for the Freescale i.MX SDMA engine.
The SDMA engine is a scatter/gather DMA engine which is implemented
as a seperate coprocessor. SDMA needs its own firmware which is
requested using the standard request_firmware mechanism. The firmware
has different entry points for each peripheral type, so drivers
have to pass the peripheral type to the DMA engine which in turn
picks the correct firmware entry point from a table contained in
the firmware image itself.
The original Freescale code also supports support for transfering
data to the internal SRAM which needs different entry points to
the firmware. Support for this is currently not implemented. Also,
support for the ASRC (asymmetric sample rate converter) is skipped.
I took a very simple approach to implement dmaengine support. Only
a single descriptor is statically assigned to a each channel. This
means that transfers can't be queued up but only a single transfer
is in progress. This simplifies implementation a lot and is sufficient
for the usual device/memory transfers.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Reviewed-by: Linus Walleij <linus.ml.walleij@gmail.com>
---
arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
arch/arm/mach-mx3/Kconfig | 2 +
arch/arm/plat-mxc/Kconfig | 10 +
arch/arm/plat-mxc/include/mach/dma.h | 64 ++
arch/arm/plat-mxc/include/mach/sdma.h | 17 +
drivers/dma/Kconfig | 8 +
drivers/dma/Makefile | 1 +
drivers/dma/imx-sdma.c | 1386 +++++++++++++++++++++++++++++++
8 files changed, 1490 insertions(+), 6 deletions(-)
create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
create mode 100644 drivers/dma/imx-sdma.c
diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
index 287431c..ac6fd71 100644
--- a/arch/arm/mach-imx/include/mach/dma-v1.h
+++ b/arch/arm/mach-imx/include/mach/dma-v1.h
@@ -27,6 +27,8 @@
#define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
+#include <mach/dma.h>
+
#define IMX_DMA_CHANNELS 16
#define DMA_MODE_READ 0
@@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
void imx_dma_free(int channel);
-enum imx_dma_prio {
- DMA_PRIO_HIGH = 0,
- DMA_PRIO_MEDIUM = 1,
- DMA_PRIO_LOW = 2
-};
-
int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
#endif /* __MACH_DMA_V1_H__ */
diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
index 85beece..301375c 100644
--- a/arch/arm/mach-mx3/Kconfig
+++ b/arch/arm/mach-mx3/Kconfig
@@ -3,12 +3,14 @@ if ARCH_MX3
config ARCH_MX31
select ARCH_HAS_RNGA
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
bool
config ARCH_MX35
bool
select ARCH_MXC_IOMUX_V3
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
comment "MX3 platforms:"
diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
index 0527e65..6741625 100644
--- a/arch/arm/plat-mxc/Kconfig
+++ b/arch/arm/plat-mxc/Kconfig
@@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
config ARCH_MXC_AUDMUX_V2
bool
+config IMX_HAVE_SDMA
+ bool
+
+config IMX_SDMA
+ depends on IMX_HAVE_SDMA
+ tristate "Enable SDMA support"
+ help
+ Include support for the SDMA engine. The SDMA engine needs additional
+ firmware support. SDMA can be compiled as a module to support loading
+ the firmware when a rootfs is present.
endif
diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
new file mode 100644
index 0000000..69d181f
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/dma.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARCH_MXC_DMA_H__
+#define __ASM_ARCH_MXC_DMA_H__
+
+#include <linux/scatterlist.h>
+
+/*
+ * This enumerates peripheral types. Used for SDMA.
+ */
+enum sdma_peripheral_type {
+ IMX_DMATYPE_SSI, /* MCU domain SSI */
+ IMX_DMATYPE_SSI_SP, /* Shared SSI */
+ IMX_DMATYPE_MMC, /* MMC */
+ IMX_DMATYPE_SDHC, /* SDHC */
+ IMX_DMATYPE_UART, /* MCU domain UART */
+ IMX_DMATYPE_UART_SP, /* Shared UART */
+ IMX_DMATYPE_FIRI, /* FIRI */
+ IMX_DMATYPE_CSPI, /* MCU domain CSPI */
+ IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
+ IMX_DMATYPE_SIM, /* SIM */
+ IMX_DMATYPE_ATA, /* ATA */
+ IMX_DMATYPE_CCM, /* CCM */
+ IMX_DMATYPE_EXT, /* External peripheral */
+ IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
+ IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
+ IMX_DMATYPE_DSP, /* DSP */
+ IMX_DMATYPE_MEMORY, /* Memory */
+ IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
+ IMX_DMATYPE_SPDIF, /* SPDIF */
+ IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
+ IMX_DMATYPE_ASRC, /* ASRC */
+ IMX_DMATYPE_ESAI, /* ESAI */
+};
+
+enum imx_dma_prio {
+ DMA_PRIO_HIGH = 0,
+ DMA_PRIO_MEDIUM = 1,
+ DMA_PRIO_LOW = 2
+};
+
+struct imx_dma_data {
+ int dma_request; /* DMA request line */
+ enum sdma_peripheral_type peripheral_type;
+ int priority;
+};
+
+static inline int imx_dma_is_ipu(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "ipu-core");
+}
+
+static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "imx-sdma");
+}
+
+#endif
diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
new file mode 100644
index 0000000..9be1122
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/sdma.h
@@ -0,0 +1,17 @@
+#ifndef __MACH_MXC_SDMA_H__
+#define __MACH_MXC_SDMA_H__
+
+/**
+ * struct sdma_platform_data - platform specific data for SDMA engine
+ *
+ * @sdma_version The version of this SDMA engine
+ * @cpu_name used to generate the firmware name
+ * @to_version CPU Tape out version
+ */
+struct sdma_platform_data {
+ int sdma_version;
+ char *cpu_name;
+ int to_version;
+};
+
+#endif /* __MACH_MXC_SDMA_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9520cf0..ff68307 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -195,6 +195,14 @@ config PCH_DMA
help
Enable support for the Topcliff PCH DMA engine.
+config IMX_SDMA
+ tristate "i.MX SDMA support"
+ depends on ARCH_MXC
+ select DMA_ENGINE
+ help
+ Support the i.MX SDMA engine. This engine is integrated into
+ Freescale i.MX25/31/35/51 chips.
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 72bd703..14d7a1b 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
obj-$(CONFIG_PCH_DMA) += pch_dma.o
+obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
new file mode 100644
index 0000000..e7e2d6b
--- /dev/null
+++ b/drivers/dma/imx-sdma.c
@@ -0,0 +1,1386 @@
+/*
+ * drivers/dma/imx-sdma.c
+ *
+ * This file contains a driver for the Freescale Smart DMA engine
+ *
+ * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
+ *
+ * Based on code from Freescale:
+ *
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * The code contained herein is licensed under the GNU General Public
+ * License. You may obtain a copy of the GNU General Public License
+ * Version 2 or later at the following locations:
+ *
+ * http://www.opensource.org/licenses/gpl-license.html
+ * http://www.gnu.org/copyleft/gpl.html
+ */
+
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/dmaengine.h>
+
+#include <asm/irq.h>
+#include <mach/sdma.h>
+#include <mach/dma.h>
+#include <mach/hardware.h>
+
+/* SDMA registers */
+#define SDMA_H_C0PTR 0x000
+#define SDMA_H_INTR 0x004
+#define SDMA_H_STATSTOP 0x008
+#define SDMA_H_START 0x00c
+#define SDMA_H_EVTOVR 0x010
+#define SDMA_H_DSPOVR 0x014
+#define SDMA_H_HOSTOVR 0x018
+#define SDMA_H_EVTPEND 0x01c
+#define SDMA_H_DSPENBL 0x020
+#define SDMA_H_RESET 0x024
+#define SDMA_H_EVTERR 0x028
+#define SDMA_H_INTRMSK 0x02c
+#define SDMA_H_PSW 0x030
+#define SDMA_H_EVTERRDBG 0x034
+#define SDMA_H_CONFIG 0x038
+#define SDMA_ONCE_ENB 0x040
+#define SDMA_ONCE_DATA 0x044
+#define SDMA_ONCE_INSTR 0x048
+#define SDMA_ONCE_STAT 0x04c
+#define SDMA_ONCE_CMD 0x050
+#define SDMA_EVT_MIRROR 0x054
+#define SDMA_ILLINSTADDR 0x058
+#define SDMA_CHN0ADDR 0x05c
+#define SDMA_ONCE_RTB 0x060
+#define SDMA_XTRIG_CONF1 0x070
+#define SDMA_XTRIG_CONF2 0x074
+#define SDMA_CHNENBL0_V2 0x200
+#define SDMA_CHNENBL0_V1 0x080
+#define SDMA_CHNPRI_0 0x100
+
+/*
+ * Buffer descriptor status values.
+ */
+#define BD_DONE 0x01
+#define BD_WRAP 0x02
+#define BD_CONT 0x04
+#define BD_INTR 0x08
+#define BD_RROR 0x10
+#define BD_LAST 0x20
+#define BD_EXTD 0x80
+
+/*
+ * Data Node descriptor status values.
+ */
+#define DND_END_OF_FRAME 0x80
+#define DND_END_OF_XFER 0x40
+#define DND_DONE 0x20
+#define DND_UNUSED 0x01
+
+/*
+ * IPCV2 descriptor status values.
+ */
+#define BD_IPCV2_END_OF_FRAME 0x40
+
+#define IPCV2_MAX_NODES 50
+/*
+ * Error bit set in the CCB status field by the SDMA,
+ * in setbd routine, in case of a transfer error
+ */
+#define DATA_ERROR 0x10000000
+
+/*
+ * Buffer descriptor commands.
+ */
+#define C0_ADDR 0x01
+#define C0_LOAD 0x02
+#define C0_DUMP 0x03
+#define C0_SETCTX 0x07
+#define C0_GETCTX 0x03
+#define C0_SETDM 0x01
+#define C0_SETPM 0x04
+#define C0_GETDM 0x02
+#define C0_GETPM 0x08
+/*
+ * Change endianness indicator in the BD command field
+ */
+#define CHANGE_ENDIANNESS 0x80
+
+/*
+ * Mode/Count of data node descriptors - IPCv2
+ */
+struct sdma_mode_count {
+ u32 count : 16; /* size of the buffer pointed by this BD */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 command : 8; /* command mostlky used for channel 0 */
+};
+
+/*
+ * Buffer descriptor
+ */
+struct sdma_buffer_descriptor {
+ struct sdma_mode_count mode;
+ dma_addr_t buffer_addr; /* address of the buffer described */
+ dma_addr_t ext_buffer_addr; /* extended buffer address */
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_channel_control - Channel control Block
+ *
+ * @current_bd_ptr current buffer descriptor processed
+ * @base_bd_ptr first element of buffer descriptor array
+ * @unused padding. The SDMA engine expects an array of 128 byte
+ * control blocks
+ */
+struct sdma_channel_control {
+ dma_addr_t current_bd_ptr;
+ dma_addr_t base_bd_ptr;
+ u32 unused[2];
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_state_registers - SDMA context for a channel
+ *
+ * @pc: program counter
+ * @t: test bit: status of arithmetic & test instruction
+ * @rpc: return program counter
+ * @sf: source fault while loading data
+ * @spc: loop start program counter
+ * @df: destination fault while storing data
+ * @epc: loop end program counter
+ * @lm: loop mode
+ */
+struct sdma_state_registers {
+ u32 pc :14;
+ u32 unused1: 1;
+ u32 t : 1;
+ u32 rpc :14;
+ u32 unused0: 1;
+ u32 sf : 1;
+ u32 spc :14;
+ u32 unused2: 1;
+ u32 df : 1;
+ u32 epc :14;
+ u32 lm : 2;
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_context_data - sdma context specific to a channel
+ *
+ * @channel_state: channel state bits
+ * @gReg: general registers
+ * @mda: burst dma destination address register
+ * @msa: burst dma source address register
+ * @ms: burst dma status register
+ * @md: burst dma data register
+ * @pda: peripheral dma destination address register
+ * @psa: peripheral dma source address register
+ * @ps: peripheral dma status register
+ * @pd: peripheral dma data register
+ * @ca: CRC polynomial register
+ * @cs: CRC accumulator register
+ * @dda: dedicated core destination address register
+ * @dsa: dedicated core source address register
+ * @ds: dedicated core status register
+ * @dd: dedicated core data register
+ */
+struct sdma_context_data {
+ struct sdma_state_registers channel_state;
+ u32 gReg[8];
+ u32 mda;
+ u32 msa;
+ u32 ms;
+ u32 md;
+ u32 pda;
+ u32 psa;
+ u32 ps;
+ u32 pd;
+ u32 ca;
+ u32 cs;
+ u32 dda;
+ u32 dsa;
+ u32 ds;
+ u32 dd;
+ u32 scratch0;
+ u32 scratch1;
+ u32 scratch2;
+ u32 scratch3;
+ u32 scratch4;
+ u32 scratch5;
+ u32 scratch6;
+ u32 scratch7;
+} __attribute__ ((packed));
+
+#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
+
+struct sdma_engine;
+
+/**
+ * struct sdma_channel - housekeeping for a SDMA channel
+ *
+ * @sdma pointer to the SDMA engine for this channel
+ * @channel the channel number, matches dmaengine chan_id
+ * @direction transfer type. Needed for setting SDMA script
+ * @peripheral_type Peripheral type. Needed for setting SDMA script
+ * @event_id0 aka dma request line
+ * @event_id1 for channels that use 2 events
+ * @word_size peripheral access size
+ * @buf_tail ID of the buffer that was processed
+ * @done channel completion
+ * @num_bd max NUM_BD. number of descriptors currently handling
+ */
+struct sdma_channel {
+ struct sdma_engine *sdma;
+ unsigned int channel;
+ enum dma_data_direction direction;
+ enum sdma_peripheral_type peripheral_type;
+ unsigned int event_id0;
+ unsigned int event_id1;
+ enum dma_slave_buswidth word_size;
+ unsigned int buf_tail;
+ struct completion done;
+ unsigned int num_bd;
+ struct sdma_buffer_descriptor *bd;
+ dma_addr_t bd_phys;
+ unsigned int pc_from_device, pc_to_device;
+ unsigned long flags;
+ dma_addr_t per_address;
+ u32 event_mask0, event_mask1;
+ u32 watermark_level;
+ u32 shp_addr, per_addr;
+ struct dma_chan chan;
+ spinlock_t lock;
+ struct dma_async_tx_descriptor desc;
+ dma_cookie_t last_completed;
+ enum dma_status status;
+};
+
+#define IMX_DMA_SG_LOOP (1 << 0)
+
+#define MAX_DMA_CHANNELS 32
+#define MXC_SDMA_DEFAULT_PRIORITY 1
+#define MXC_SDMA_MIN_PRIORITY 1
+#define MXC_SDMA_MAX_PRIORITY 7
+
+/**
+ * struct sdma_script_start_addrs - SDMA script start pointers
+ *
+ * start addresses of the different functions in the physical
+ * address space of the SDMA engine.
+ */
+struct sdma_script_start_addrs {
+ u32 ap_2_ap_addr;
+ u32 ap_2_bp_addr;
+ u32 ap_2_ap_fixed_addr;
+ u32 bp_2_ap_addr;
+ u32 loopback_on_dsp_side_addr;
+ u32 mcu_interrupt_only_addr;
+ u32 firi_2_per_addr;
+ u32 firi_2_mcu_addr;
+ u32 per_2_firi_addr;
+ u32 mcu_2_firi_addr;
+ u32 uart_2_per_addr;
+ u32 uart_2_mcu_addr;
+ u32 per_2_app_addr;
+ u32 mcu_2_app_addr;
+ u32 per_2_per_addr;
+ u32 uartsh_2_per_addr;
+ u32 uartsh_2_mcu_addr;
+ u32 per_2_shp_addr;
+ u32 mcu_2_shp_addr;
+ u32 ata_2_mcu_addr;
+ u32 mcu_2_ata_addr;
+ u32 app_2_per_addr;
+ u32 app_2_mcu_addr;
+ u32 shp_2_per_addr;
+ u32 shp_2_mcu_addr;
+ u32 mshc_2_mcu_addr;
+ u32 mcu_2_mshc_addr;
+ u32 spdif_2_mcu_addr;
+ u32 mcu_2_spdif_addr;
+ u32 asrc_2_mcu_addr;
+ u32 ext_mem_2_ipu_addr;
+ u32 descrambler_addr;
+ u32 dptc_dvfs_addr;
+ u32 utra_addr;
+ u32 ram_code_start_addr;
+};
+
+#define SDMA_FIRMWARE_MAGIC 0x414d4453
+
+/**
+ * struct sdma_firmware_header - Layout of the firmware image
+ *
+ * @magic "SDMA"
+ * @version_major increased whenever layout of struct sdma_script_start_addrs
+ * changes.
+ * @version_minor firmware minor version (for binary compatible changes)
+ * @script_addrs_start offset of struct sdma_script_start_addrs in this image
+ * @num_script_addrs Number of script addresses in this image
+ * @ram_code_start offset of SDMA ram image in this firmware image
+ * @ram_code_size size of SDMA ram image
+ * @script_addrs Stores the start address of the SDMA scripts
+ * (in SDMA memory space)
+ */
+struct sdma_firmware_header {
+ u32 magic;
+ u32 version_major;
+ u32 version_minor;
+ u32 script_addrs_start;
+ u32 num_script_addrs;
+ u32 ram_code_start;
+ u32 ram_code_size;
+};
+
+struct sdma_engine {
+ struct device *dev;
+ struct sdma_channel channel[MAX_DMA_CHANNELS];
+ struct sdma_channel_control *channel_control;
+ void __iomem *regs;
+ unsigned int version;
+ unsigned int num_events;
+ struct sdma_context_data *context;
+ dma_addr_t context_phys;
+ struct dma_device dma_device;
+ struct clk *clk;
+ struct sdma_script_start_addrs *script_addrs;
+};
+
+#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
+#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
+#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
+#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
+
+static inline u32 chnenbl_ofs(struct sdma_engine *sdma, unsigned int event)
+{
+ u32 chnenbl0 = (sdma->version == 2 ? SDMA_CHNENBL0_V2 : SDMA_CHNENBL0_V1);
+
+ return chnenbl0 + event * 4;
+}
+
+static int sdma_config_ownership(struct sdma_channel *sdmac,
+ bool event_override, bool mcu_verride, bool dsp_override)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 evt, mcu, dsp;
+
+ if (event_override && mcu_verride && dsp_override)
+ return -EINVAL;
+
+ evt = readl(sdma->regs + SDMA_H_EVTOVR);
+ mcu = readl(sdma->regs + SDMA_H_HOSTOVR);
+ dsp = readl(sdma->regs + SDMA_H_DSPOVR);
+
+ if (dsp_override)
+ dsp &= ~(1 << channel);
+ else
+ dsp |= (1 << channel);
+
+ if (event_override)
+ evt &= ~(1 << channel);
+ else
+ evt |= (1 << channel);
+
+ if (mcu_verride)
+ mcu &= ~(1 << channel);
+ else
+ mcu |= (1 << channel);
+
+ writel(evt, sdma->regs + SDMA_H_EVTOVR);
+ writel(mcu, sdma->regs + SDMA_H_HOSTOVR);
+ writel(dsp, sdma->regs + SDMA_H_DSPOVR);
+
+ return 0;
+}
+
+/*
+ * sdma_run_channel - run a channel and wait till it's done
+ */
+static int sdma_run_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret;
+
+ init_completion(&sdmac->done);
+
+ writel(1 << channel, sdma->regs + SDMA_H_START);
+
+ ret = wait_for_completion_timeout(&sdmac->done, HZ);
+
+ return ret ? 0 : -ETIMEDOUT;
+}
+
+static int sdma_load_script(struct sdma_engine *sdma, void *buf, int size,
+ u32 address)
+{
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ void *buf_virt;
+ dma_addr_t buf_phys;
+ int ret;
+
+ buf_virt = dma_alloc_coherent(NULL,
+ size,
+ &buf_phys, GFP_KERNEL);
+ if (!buf_virt)
+ return -ENOMEM;
+
+ bd0->mode.command = C0_SETPM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = size / 2;
+ bd0->buffer_addr = buf_phys;
+ bd0->ext_buffer_addr = address;
+
+ memcpy(buf_virt, buf, size);
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ dma_free_coherent(NULL, size, buf_virt, buf_phys);
+
+ return ret;
+}
+
+static void sdma_event_enable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 val;
+ u32 chnenbl = chnenbl_ofs(sdma, event);
+
+ val = readl(sdma->regs + chnenbl);
+ val |= (1 << channel);
+ writel(val, sdma->regs + chnenbl);
+}
+
+static void sdma_event_disable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 chnenbl = chnenbl_ofs(sdma, event);
+ u32 val;
+
+ val = readl(sdma->regs + chnenbl);
+ val &= ~(1 << channel);
+ writel(val, sdma->regs + chnenbl);
+}
+
+static void sdma_handle_channel_loop(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+
+ /*
+ * loop mode. Iterate over descriptors, re-setup them and
+ * call callback function.
+ */
+ while (1) {
+ bd = &sdmac->bd[sdmac->buf_tail];
+
+ if (bd->mode.status & BD_DONE)
+ break;
+
+ if (bd->mode.status & BD_RROR)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ bd->mode.status |= BD_DONE;
+ sdmac->buf_tail++;
+ sdmac->buf_tail %= sdmac->num_bd;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ }
+}
+
+static void mxc_sdma_handle_channel_normal(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+ int i, error = 0;
+
+ /*
+ * non loop mode. Iterate over all descriptors, collect
+ * errors and call callback function
+ */
+ for (i = 0; i < sdmac->num_bd; i++) {
+ bd = &sdmac->bd[i];
+
+ if (bd->mode.status & (BD_DONE | BD_RROR))
+ error = -EIO;
+ }
+
+ if (error)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ sdmac->last_completed = sdmac->desc.cookie;
+}
+
+static void mxc_sdma_handle_channel(struct sdma_channel *sdmac)
+{
+ complete(&sdmac->done);
+
+ /* not interested in channel 0 interrupts */
+ if (sdmac->channel == 0)
+ return;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP)
+ sdma_handle_channel_loop(sdmac);
+ else
+ mxc_sdma_handle_channel_normal(sdmac);
+}
+
+static irqreturn_t sdma_int_handler(int irq, void *dev_id)
+{
+ struct sdma_engine *sdma = dev_id;
+ u32 stat;
+
+ stat = readl(sdma->regs + SDMA_H_INTR);
+ writel(stat, sdma->regs + SDMA_H_INTR);
+
+ while (stat) {
+ int channel = fls(stat) - 1;
+ struct sdma_channel *sdmac = &sdma->channel[channel];
+
+ mxc_sdma_handle_channel(sdmac);
+
+ stat &= ~(1 << channel);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * sets the pc of SDMA script according to the peripheral type
+ */
+static void sdma_get_pc(struct sdma_channel *sdmac,
+ enum sdma_peripheral_type peripheral_type)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int per_2_emi = 0, emi_2_per = 0;
+ /*
+ * These are needed once we start to support transfers between
+ * two peripherals or memory-to-memory transfers
+ */
+ int per_2_per = 0, emi_2_emi = 0;
+
+ sdmac->pc_from_device = 0;
+ sdmac->pc_to_device = 0;
+
+ switch (peripheral_type) {
+ case IMX_DMATYPE_MEMORY:
+ emi_2_emi = sdma->script_addrs->ap_2_ap_addr;
+ break;
+ case IMX_DMATYPE_DSP:
+ emi_2_per = sdma->script_addrs->bp_2_ap_addr;
+ per_2_emi = sdma->script_addrs->ap_2_bp_addr;
+ break;
+ case IMX_DMATYPE_FIRI:
+ per_2_emi = sdma->script_addrs->firi_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_firi_addr;
+ break;
+ case IMX_DMATYPE_UART:
+ per_2_emi = sdma->script_addrs->uart_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_UART_SP:
+ per_2_emi = sdma->script_addrs->uartsh_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ATA:
+ per_2_emi = sdma->script_addrs->ata_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_ata_addr;
+ break;
+ case IMX_DMATYPE_CSPI:
+ case IMX_DMATYPE_EXT:
+ case IMX_DMATYPE_SSI:
+ per_2_emi = sdma->script_addrs->app_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_SSI_SP:
+ case IMX_DMATYPE_MMC:
+ case IMX_DMATYPE_SDHC:
+ case IMX_DMATYPE_CSPI_SP:
+ case IMX_DMATYPE_ESAI:
+ case IMX_DMATYPE_MSHC_SP:
+ per_2_emi = sdma->script_addrs->shp_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ASRC:
+ per_2_emi = sdma->script_addrs->asrc_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->asrc_2_mcu_addr;
+ per_2_per = sdma->script_addrs->per_2_per_addr;
+ break;
+ case IMX_DMATYPE_MSHC:
+ per_2_emi = sdma->script_addrs->mshc_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_mshc_addr;
+ break;
+ case IMX_DMATYPE_CCM:
+ per_2_emi = sdma->script_addrs->dptc_dvfs_addr;
+ break;
+ case IMX_DMATYPE_SPDIF:
+ per_2_emi = sdma->script_addrs->spdif_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_spdif_addr;
+ break;
+ case IMX_DMATYPE_IPU_MEMORY:
+ emi_2_per = sdma->script_addrs->ext_mem_2_ipu_addr;
+ break;
+ default:
+ break;
+ }
+
+ sdmac->pc_from_device = per_2_emi;
+ sdmac->pc_to_device = emi_2_per;
+}
+
+static int sdma_load_context(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int load_address;
+ struct sdma_context_data *context = sdma->context;
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ int ret;
+
+ if (sdmac->direction == DMA_FROM_DEVICE) {
+ load_address = sdmac->pc_from_device;
+ } else {
+ load_address = sdmac->pc_to_device;
+ }
+
+ if (load_address < 0)
+ return load_address;
+
+ dev_dbg(sdma->dev, "load_address = %d\n", load_address);
+ dev_dbg(sdma->dev, "wml = 0x%08x\n", sdmac->watermark_level);
+ dev_dbg(sdma->dev, "shp_addr = 0x%08x\n", sdmac->shp_addr);
+ dev_dbg(sdma->dev, "per_addr = 0x%08x\n", sdmac->per_addr);
+ dev_dbg(sdma->dev, "event_mask0 = 0x%08x\n", sdmac->event_mask0);
+ dev_dbg(sdma->dev, "event_mask1 = 0x%08x\n", sdmac->event_mask1);
+
+ memset(context, 0, sizeof(*context));
+ context->channel_state.pc = load_address;
+
+ /* Send by context the event mask,base address for peripheral
+ * and watermark level
+ */
+ context->gReg[0] = sdmac->event_mask1;
+ context->gReg[1] = sdmac->event_mask0;
+ context->gReg[2] = sdmac->per_addr;
+ context->gReg[6] = sdmac->shp_addr;
+ context->gReg[7] = sdmac->watermark_level;
+
+ bd0->mode.command = C0_SETDM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = sizeof(*context) / 4;
+ bd0->buffer_addr = sdma->context_phys;
+ bd0->ext_buffer_addr = 2048 + (sizeof(*context) / 4) * channel;
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ return ret;
+}
+
+static void sdma_disable_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ writel(1 << channel, sdma->regs + SDMA_H_STATSTOP);
+ sdmac->status = DMA_ERROR;
+}
+
+static int sdma_config_channel(struct sdma_channel *sdmac)
+{
+ int ret;
+
+ sdma_disable_channel(sdmac);
+
+ sdmac->event_mask0 = 0;
+ sdmac->event_mask1 = 0;
+ sdmac->shp_addr = 0;
+ sdmac->per_addr = 0;
+
+ if (sdmac->event_id0) {
+ if (sdmac->event_id0 > 32)
+ return -EINVAL;
+ sdma_event_enable(sdmac, sdmac->event_id0);
+ }
+
+ switch (sdmac->peripheral_type) {
+ case IMX_DMATYPE_DSP:
+ sdma_config_ownership(sdmac, false, true, true);
+ break;
+ case IMX_DMATYPE_MEMORY:
+ sdma_config_ownership(sdmac, false, true, false);
+ break;
+ default:
+ sdma_config_ownership(sdmac, true, true, false);
+ break;
+ }
+
+ sdma_get_pc(sdmac, sdmac->peripheral_type);
+
+ if ((sdmac->peripheral_type != IMX_DMATYPE_MEMORY) &&
+ (sdmac->peripheral_type != IMX_DMATYPE_DSP)) {
+ /* Handle multiple event channels differently */
+ if (sdmac->event_id1) {
+ sdmac->event_mask1 = 1 << (sdmac->event_id1 % 32);
+ if (sdmac->event_id1 > 31)
+ sdmac->watermark_level |= 1 << 31;
+ sdmac->event_mask0 = 1 << (sdmac->event_id0 % 32);
+ if (sdmac->event_id0 > 31)
+ sdmac->watermark_level |= 1 << 30;
+ } else {
+ sdmac->event_mask0 = 1 << sdmac->event_id0;
+ sdmac->event_mask1 = 1 << (sdmac->event_id0 - 32);
+ }
+ /* Watermark Level */
+ sdmac->watermark_level |= sdmac->watermark_level;
+ /* Address */
+ sdmac->shp_addr = sdmac->per_address;
+ } else {
+ sdmac->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
+ }
+
+ ret = sdma_load_context(sdmac);
+
+ return ret;
+}
+
+static int sdma_set_channel_priority(struct sdma_channel *sdmac,
+ unsigned int priority)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ if (priority < MXC_SDMA_MIN_PRIORITY
+ || priority > MXC_SDMA_MAX_PRIORITY) {
+ return -EINVAL;
+ }
+
+ writel(priority, sdma->regs + SDMA_CHNPRI_0 + 4 * channel);
+
+ return 0;
+}
+
+static int sdma_request_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret = -EBUSY;
+
+ sdmac->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdmac->bd_phys, GFP_KERNEL);
+ if (!sdmac->bd) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memset(sdmac->bd, 0, PAGE_SIZE);
+
+ sdma->channel_control[channel].base_bd_ptr = sdmac->bd_phys;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ clk_enable(sdma->clk);
+
+ sdma_set_channel_priority(sdmac, MXC_SDMA_DEFAULT_PRIORITY);
+
+ init_completion(&sdmac->done);
+
+ sdmac->buf_tail = 0;
+
+ return 0;
+out:
+
+ return ret;
+}
+
+static void sdma_enable_channel(struct sdma_engine *sdma, int channel)
+{
+ writel(1 << channel, sdma->regs + SDMA_H_START);
+}
+
+static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
+{
+ dma_cookie_t cookie = sdma->chan.cookie;
+
+ if (++cookie < 0)
+ cookie = 1;
+
+ sdma->chan.cookie = cookie;
+ sdma->desc.cookie = cookie;
+
+ return cookie;
+}
+
+static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sdma_channel, chan);
+}
+
+static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(tx->chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ dma_cookie_t cookie;
+
+ spin_lock_irq(&sdmac->lock);
+
+ cookie = sdma_assign_cookie(sdmac);
+
+ sdma_enable_channel(sdma, tx->chan->chan_id);
+
+ spin_unlock_irq(&sdmac->lock);
+
+ return cookie;
+}
+
+static int sdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct imx_dma_data *data = chan->private;
+ int prio, ret;
+
+ /* No need to execute this for internal channel 0 */
+ if (chan->chan_id == 0)
+ return 0;
+
+ if (!data)
+ return -EINVAL;
+
+ switch (data->priority) {
+ case DMA_PRIO_HIGH:
+ prio = 3;
+ break;
+ case DMA_PRIO_MEDIUM:
+ prio = 2;
+ break;
+ case DMA_PRIO_LOW:
+ default:
+ prio = 1;
+ break;
+ }
+
+ sdmac->peripheral_type = data->peripheral_type;
+ sdmac->event_id0 = data->dma_request;
+ ret = sdma_set_channel_priority(sdmac, prio);
+ if (ret)
+ return ret;
+
+ ret = sdma_request_channel(sdmac);
+ if (ret)
+ return ret;
+
+ dma_async_tx_descriptor_init(&sdmac->desc, chan);
+ sdmac->desc.tx_submit = sdma_tx_submit;
+ /* txd.flags will be overwritten in prep funcs */
+ sdmac->desc.flags = DMA_CTRL_ACK;
+
+ return 0;
+}
+
+static void sdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+
+ sdma_disable_channel(sdmac);
+
+ if (sdmac->event_id0)
+ sdma_event_disable(sdmac, sdmac->event_id0);
+ if (sdmac->event_id1)
+ sdma_event_disable(sdmac, sdmac->event_id1);
+
+ sdmac->event_id0 = 0;
+ sdmac->event_id1 = 0;
+
+ sdma_set_channel_priority(sdmac, 0);
+
+ dma_free_coherent(NULL, PAGE_SIZE, sdmac->bd, sdmac->bd_phys);
+
+ clk_disable(sdma->clk);
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int ret, i, count;
+ int channel = chan->chan_id;
+ struct scatterlist *sg;
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags = 0;
+
+ dev_dbg(sdma->dev, "setting up %d entries for channel %d.\n",
+ sg_len, channel);
+
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (sg_len > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, sg_len, NUM_BD);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = sgl->dma_address;
+
+ count = sg->length;
+
+ if (count > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
+ channel, count, 0xffff);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ bd->mode.count = count;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES) {
+ ret = -EINVAL;
+ goto err_out;
+ }
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP) {
+ param |= BD_INTR;
+ if (i + 1 == sg_len)
+ param |= BD_WRAP;
+ }
+
+ if (i + 1 == sg_len)
+ param |= BD_INTR;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, count, sg->dma_address,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+ }
+
+ sdmac->num_bd = sg_len;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int num_periods = buf_len / period_len;
+ int channel = chan->chan_id;
+ int ret, i = 0, buf = 0;
+
+ dev_dbg(sdma->dev, "%s channel: %d\n", __func__, channel);
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags |= IMX_DMA_SG_LOOP;
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (num_periods > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, num_periods, NUM_BD);
+ goto err_out;
+ }
+
+ if (period_len > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum period size exceeded: %d > %d\n",
+ channel, period_len, 0xffff);
+ goto err_out;
+ }
+
+ while (buf < buf_len) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = dma_addr;
+
+ bd->mode.count = period_len;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES)
+ goto err_out;
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
+ if (i + 1 == num_periods)
+ param |= BD_WRAP;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, period_len, dma_addr,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+
+ dma_addr += period_len;
+ buf += period_len;
+
+ i++;
+ }
+
+ sdmac->num_bd = num_periods;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ sdmac->status = DMA_ERROR;
+ return NULL;
+}
+
+static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct dma_slave_config *dmaengine_cfg = (void *)arg;
+
+ switch (cmd) {
+ case DMA_TERMINATE_ALL:
+ sdma_disable_channel(sdmac);
+ return 0;
+ case DMA_SLAVE_CONFIG:
+ if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
+ sdmac->per_address = dmaengine_cfg->src_addr;
+ sdmac->watermark_level = dmaengine_cfg->src_maxburst;
+ sdmac->word_size = dmaengine_cfg->src_addr_width;
+ } else {
+ sdmac->per_address = dmaengine_cfg->dst_addr;
+ sdmac->watermark_level = dmaengine_cfg->dst_maxburst;
+ sdmac->word_size = dmaengine_cfg->dst_addr_width;
+ }
+ return sdma_config_channel(sdmac);
+ default:
+ return -ENOSYS;
+ }
+
+ return -EINVAL;
+}
+
+static enum dma_status sdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ dma_cookie_t last_used;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+
+ ret = dma_async_is_complete(cookie, sdmac->last_completed, last_used);
+ dma_set_tx_state(txstate, sdmac->last_completed, last_used, 0);
+
+ return ret;
+}
+
+static void sdma_issue_pending(struct dma_chan *chan)
+{
+ /*
+ * Nothing to do. We only have a single descriptor
+ */
+}
+
+static int __init sdma_init(struct sdma_engine *sdma,
+ void *ram_code, int ram_code_size)
+{
+ int i, ret;
+ dma_addr_t ccb_phys;
+
+ switch (sdma->version) {
+ case 1:
+ sdma->num_events = 32;
+ break;
+ case 2:
+ sdma->num_events = 48;
+ break;
+ default:
+ dev_err(sdma->dev, "Unknown version %d. aborting\n", sdma->version);
+ return -ENODEV;
+ }
+
+ clk_enable(sdma->clk);
+
+ /* Be sure SDMA has not started yet */
+ writel(0, sdma->regs + SDMA_H_C0PTR);
+
+ sdma->channel_control = dma_alloc_coherent(NULL,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
+ sizeof(struct sdma_context_data),
+ &ccb_phys, GFP_KERNEL);
+
+ if (!sdma->channel_control) {
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ sdma->context = (void *)sdma->channel_control +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+ sdma->context_phys = ccb_phys +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+
+ /* Zero-out the CCB structures array just allocated */
+ memset(sdma->channel_control, 0,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
+
+ /* disable all channels */
+ for (i = 0; i < sdma->num_events; i++)
+ writel(0, sdma->regs + chnenbl_ofs(sdma, i));
+
+ /* All channels have priority 0 */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++)
+ writel(0, sdma->regs + SDMA_CHNPRI_0 + i * 4);
+
+ ret = sdma_request_channel(&sdma->channel[0]);
+ if (ret)
+ goto err_dma_alloc;
+
+ sdma_config_ownership(&sdma->channel[0], false, true, false);
+
+ /* Set Command Channel (Channel Zero) */
+ writel(0x4050, sdma->regs + SDMA_CHN0ADDR);
+
+ /* Set bits of CONFIG register but with static context switching */
+ /* FIXME: Check whether to set ACR bit depending on clock ratios */
+ writel(0, sdma->regs + SDMA_H_CONFIG);
+
+ writel(ccb_phys, sdma->regs + SDMA_H_C0PTR);
+
+ /* download the RAM image for SDMA */
+ sdma_load_script(sdma, ram_code,
+ ram_code_size,
+ sdma->script_addrs->ram_code_start_addr);
+
+ /* Set bits of CONFIG register with given context switching mode */
+ writel(SDMA_H_CONFIG_CSM, sdma->regs + SDMA_H_CONFIG);
+
+ /* Initializes channel's priorities */
+ sdma_set_channel_priority(&sdma->channel[0], 7);
+
+ clk_disable(sdma->clk);
+
+ return 0;
+
+err_dma_alloc:
+ clk_disable(sdma->clk);
+ dev_err(sdma->dev, "initialisation failed with %d\n", ret);
+ return ret;
+}
+
+static int __devinit sdma_probe(struct platform_device *pdev)
+{
+ int ret;
+ const struct firmware *fw;
+ const struct sdma_firmware_header *header;
+ const struct sdma_script_start_addrs *addr;
+ int irq;
+ unsigned short *ram_code;
+ struct resource *iores;
+ struct sdma_platform_data *pdata = pdev->dev.platform_data;
+ char *fwname;
+ int i;
+ dma_cap_mask_t mask;
+ struct sdma_engine *sdma;
+
+ sdma = kzalloc(sizeof(*sdma), GFP_KERNEL);
+ if (!sdma)
+ return -ENOMEM;
+
+ sdma->dev = &pdev->dev;
+
+ iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ irq = platform_get_irq(pdev, 0);
+ if (!iores || irq < 0 || !pdata) {
+ ret = -EINVAL;
+ goto err_irq;
+ }
+
+ sdma->clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(sdma->clk)) {
+ ret = PTR_ERR(sdma->clk);
+ goto err_clk;
+ }
+
+ sdma->regs = ioremap(iores->start, resource_size(iores));
+ if (!sdma->regs) {
+ ret = -ENOMEM;
+ goto err_ioremap;
+ }
+
+ ret = request_irq(irq, sdma_int_handler, 0, "sdma", sdma);
+ if (ret)
+ goto err_request_irq;
+
+ fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin",
+ pdata->cpu_name, pdata->to_version);
+ if (!fwname) {
+ ret = -ENOMEM;
+ goto err_cputype;
+ }
+
+ ret = request_firmware(&fw, fwname, &pdev->dev);
+ if (ret) {
+ dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
+ fwname, ret);
+ kfree(fwname);
+ goto err_cputype;
+ }
+ kfree(fwname);
+
+ if (fw->size < sizeof(*header))
+ goto err_firmware;
+
+ header = (struct sdma_firmware_header *)fw->data;
+
+ if (header->magic != SDMA_FIRMWARE_MAGIC)
+ goto err_firmware;
+ if (header->ram_code_start + header->ram_code_size > fw->size)
+ goto err_firmware;
+
+ addr = (void *)header + header->script_addrs_start;
+ ram_code = (void *)header + header->ram_code_start;
+ sdma->script_addrs = kmalloc(sizeof(*addr), GFP_KERNEL);
+ if (!sdma->script_addrs)
+ goto err_firmware;
+ memcpy(sdma->script_addrs, addr, sizeof(*addr));
+
+ sdma->version = pdata->sdma_version;
+
+ INIT_LIST_HEAD(&sdma->dma_device.channels);
+ /* Initialize channel parameters */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++) {
+ struct sdma_channel *sdmac = &sdma->channel[i];
+
+ sdmac->sdma = sdma;
+ spin_lock_init(&sdmac->lock);
+
+ dma_cap_set(DMA_SLAVE, sdma->dma_device.cap_mask);
+ dma_cap_set(DMA_CYCLIC, sdma->dma_device.cap_mask);
+
+ sdmac->chan.device = &sdma->dma_device;
+ sdmac->chan.chan_id = i;
+ sdmac->channel = i;
+
+ /* Add the channel to the DMAC list */
+ list_add_tail(&sdmac->chan.device_node, &sdma->dma_device.channels);
+ }
+
+ ret = sdma_init(sdma, ram_code, header->ram_code_size);
+ if (ret)
+ goto err_init;
+
+ sdma->dma_device.dev = &pdev->dev;
+
+ sdma->dma_device.device_alloc_chan_resources = sdma_alloc_chan_resources;
+ sdma->dma_device.device_free_chan_resources = sdma_free_chan_resources;
+ sdma->dma_device.device_tx_status = sdma_tx_status;
+ sdma->dma_device.device_prep_slave_sg = sdma_prep_slave_sg;
+ sdma->dma_device.device_prep_dma_cyclic = sdma_prep_dma_cyclic;
+ sdma->dma_device.device_control = sdma_control;
+ sdma->dma_device.device_issue_pending = sdma_issue_pending;
+
+ ret = dma_async_device_register(&sdma->dma_device);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register\n");
+ goto err_init;
+ }
+
+ dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
+ header->version_major,
+ header->version_minor);
+
+ /* request channel 0. This is an internal control channel
+ * to the SDMA engine and not available to clients.
+ */
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_request_channel(mask, NULL, NULL);
+
+ release_firmware(fw);
+
+ return 0;
+
+err_init:
+ kfree(sdma->script_addrs);
+err_firmware:
+ release_firmware(fw);
+err_cputype:
+ free_irq(irq, sdma);
+err_request_irq:
+ iounmap(sdma->regs);
+err_ioremap:
+ clk_put(sdma->clk);
+err_clk:
+err_irq:
+ kfree(sdma);
+ return 0;
+}
+
+static int __devexit sdma_remove(struct platform_device *pdev)
+{
+ return -EBUSY;
+}
+
+static struct platform_driver sdma_driver = {
+ .driver = {
+ .name = "imx-sdma",
+ },
+ .probe = sdma_probe,
+ .remove = __devexit_p(sdma_remove),
+};
+
+static int __init sdma_module_init(void)
+{
+ return platform_driver_register(&sdma_driver);
+}
+subsys_initcall(sdma_module_init);
+
+MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("i.MX SDMA driver");
+MODULE_LICENSE("GPL");
--
1.7.1
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH 3/3 v3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-24 7:10 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-24 7:10 UTC (permalink / raw)
To: linux-arm-kernel
This patch adds support for the Freescale i.MX SDMA engine.
The SDMA engine is a scatter/gather DMA engine which is implemented
as a seperate coprocessor. SDMA needs its own firmware which is
requested using the standard request_firmware mechanism. The firmware
has different entry points for each peripheral type, so drivers
have to pass the peripheral type to the DMA engine which in turn
picks the correct firmware entry point from a table contained in
the firmware image itself.
The original Freescale code also supports support for transfering
data to the internal SRAM which needs different entry points to
the firmware. Support for this is currently not implemented. Also,
support for the ASRC (asymmetric sample rate converter) is skipped.
I took a very simple approach to implement dmaengine support. Only
a single descriptor is statically assigned to a each channel. This
means that transfers can't be queued up but only a single transfer
is in progress. This simplifies implementation a lot and is sufficient
for the usual device/memory transfers.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Reviewed-by: Linus Walleij <linus.ml.walleij@gmail.com>
---
arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
arch/arm/mach-mx3/Kconfig | 2 +
arch/arm/plat-mxc/Kconfig | 10 +
arch/arm/plat-mxc/include/mach/dma.h | 64 ++
arch/arm/plat-mxc/include/mach/sdma.h | 17 +
drivers/dma/Kconfig | 8 +
drivers/dma/Makefile | 1 +
drivers/dma/imx-sdma.c | 1386 +++++++++++++++++++++++++++++++
8 files changed, 1490 insertions(+), 6 deletions(-)
create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
create mode 100644 drivers/dma/imx-sdma.c
diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
index 287431c..ac6fd71 100644
--- a/arch/arm/mach-imx/include/mach/dma-v1.h
+++ b/arch/arm/mach-imx/include/mach/dma-v1.h
@@ -27,6 +27,8 @@
#define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
+#include <mach/dma.h>
+
#define IMX_DMA_CHANNELS 16
#define DMA_MODE_READ 0
@@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
void imx_dma_free(int channel);
-enum imx_dma_prio {
- DMA_PRIO_HIGH = 0,
- DMA_PRIO_MEDIUM = 1,
- DMA_PRIO_LOW = 2
-};
-
int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
#endif /* __MACH_DMA_V1_H__ */
diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
index 85beece..301375c 100644
--- a/arch/arm/mach-mx3/Kconfig
+++ b/arch/arm/mach-mx3/Kconfig
@@ -3,12 +3,14 @@ if ARCH_MX3
config ARCH_MX31
select ARCH_HAS_RNGA
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
bool
config ARCH_MX35
bool
select ARCH_MXC_IOMUX_V3
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
comment "MX3 platforms:"
diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
index 0527e65..6741625 100644
--- a/arch/arm/plat-mxc/Kconfig
+++ b/arch/arm/plat-mxc/Kconfig
@@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
config ARCH_MXC_AUDMUX_V2
bool
+config IMX_HAVE_SDMA
+ bool
+
+config IMX_SDMA
+ depends on IMX_HAVE_SDMA
+ tristate "Enable SDMA support"
+ help
+ Include support for the SDMA engine. The SDMA engine needs additional
+ firmware support. SDMA can be compiled as a module to support loading
+ the firmware when a rootfs is present.
endif
diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
new file mode 100644
index 0000000..69d181f
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/dma.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARCH_MXC_DMA_H__
+#define __ASM_ARCH_MXC_DMA_H__
+
+#include <linux/scatterlist.h>
+
+/*
+ * This enumerates peripheral types. Used for SDMA.
+ */
+enum sdma_peripheral_type {
+ IMX_DMATYPE_SSI, /* MCU domain SSI */
+ IMX_DMATYPE_SSI_SP, /* Shared SSI */
+ IMX_DMATYPE_MMC, /* MMC */
+ IMX_DMATYPE_SDHC, /* SDHC */
+ IMX_DMATYPE_UART, /* MCU domain UART */
+ IMX_DMATYPE_UART_SP, /* Shared UART */
+ IMX_DMATYPE_FIRI, /* FIRI */
+ IMX_DMATYPE_CSPI, /* MCU domain CSPI */
+ IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
+ IMX_DMATYPE_SIM, /* SIM */
+ IMX_DMATYPE_ATA, /* ATA */
+ IMX_DMATYPE_CCM, /* CCM */
+ IMX_DMATYPE_EXT, /* External peripheral */
+ IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
+ IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
+ IMX_DMATYPE_DSP, /* DSP */
+ IMX_DMATYPE_MEMORY, /* Memory */
+ IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
+ IMX_DMATYPE_SPDIF, /* SPDIF */
+ IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
+ IMX_DMATYPE_ASRC, /* ASRC */
+ IMX_DMATYPE_ESAI, /* ESAI */
+};
+
+enum imx_dma_prio {
+ DMA_PRIO_HIGH = 0,
+ DMA_PRIO_MEDIUM = 1,
+ DMA_PRIO_LOW = 2
+};
+
+struct imx_dma_data {
+ int dma_request; /* DMA request line */
+ enum sdma_peripheral_type peripheral_type;
+ int priority;
+};
+
+static inline int imx_dma_is_ipu(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "ipu-core");
+}
+
+static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "imx-sdma");
+}
+
+#endif
diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
new file mode 100644
index 0000000..9be1122
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/sdma.h
@@ -0,0 +1,17 @@
+#ifndef __MACH_MXC_SDMA_H__
+#define __MACH_MXC_SDMA_H__
+
+/**
+ * struct sdma_platform_data - platform specific data for SDMA engine
+ *
+ * @sdma_version The version of this SDMA engine
+ * @cpu_name used to generate the firmware name
+ * @to_version CPU Tape out version
+ */
+struct sdma_platform_data {
+ int sdma_version;
+ char *cpu_name;
+ int to_version;
+};
+
+#endif /* __MACH_MXC_SDMA_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9520cf0..ff68307 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -195,6 +195,14 @@ config PCH_DMA
help
Enable support for the Topcliff PCH DMA engine.
+config IMX_SDMA
+ tristate "i.MX SDMA support"
+ depends on ARCH_MXC
+ select DMA_ENGINE
+ help
+ Support the i.MX SDMA engine. This engine is integrated into
+ Freescale i.MX25/31/35/51 chips.
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 72bd703..14d7a1b 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
obj-$(CONFIG_PCH_DMA) += pch_dma.o
+obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
new file mode 100644
index 0000000..e7e2d6b
--- /dev/null
+++ b/drivers/dma/imx-sdma.c
@@ -0,0 +1,1386 @@
+/*
+ * drivers/dma/imx-sdma.c
+ *
+ * This file contains a driver for the Freescale Smart DMA engine
+ *
+ * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
+ *
+ * Based on code from Freescale:
+ *
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * The code contained herein is licensed under the GNU General Public
+ * License. You may obtain a copy of the GNU General Public License
+ * Version 2 or later at the following locations:
+ *
+ * http://www.opensource.org/licenses/gpl-license.html
+ * http://www.gnu.org/copyleft/gpl.html
+ */
+
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/dmaengine.h>
+
+#include <asm/irq.h>
+#include <mach/sdma.h>
+#include <mach/dma.h>
+#include <mach/hardware.h>
+
+/* SDMA registers */
+#define SDMA_H_C0PTR 0x000
+#define SDMA_H_INTR 0x004
+#define SDMA_H_STATSTOP 0x008
+#define SDMA_H_START 0x00c
+#define SDMA_H_EVTOVR 0x010
+#define SDMA_H_DSPOVR 0x014
+#define SDMA_H_HOSTOVR 0x018
+#define SDMA_H_EVTPEND 0x01c
+#define SDMA_H_DSPENBL 0x020
+#define SDMA_H_RESET 0x024
+#define SDMA_H_EVTERR 0x028
+#define SDMA_H_INTRMSK 0x02c
+#define SDMA_H_PSW 0x030
+#define SDMA_H_EVTERRDBG 0x034
+#define SDMA_H_CONFIG 0x038
+#define SDMA_ONCE_ENB 0x040
+#define SDMA_ONCE_DATA 0x044
+#define SDMA_ONCE_INSTR 0x048
+#define SDMA_ONCE_STAT 0x04c
+#define SDMA_ONCE_CMD 0x050
+#define SDMA_EVT_MIRROR 0x054
+#define SDMA_ILLINSTADDR 0x058
+#define SDMA_CHN0ADDR 0x05c
+#define SDMA_ONCE_RTB 0x060
+#define SDMA_XTRIG_CONF1 0x070
+#define SDMA_XTRIG_CONF2 0x074
+#define SDMA_CHNENBL0_V2 0x200
+#define SDMA_CHNENBL0_V1 0x080
+#define SDMA_CHNPRI_0 0x100
+
+/*
+ * Buffer descriptor status values.
+ */
+#define BD_DONE 0x01
+#define BD_WRAP 0x02
+#define BD_CONT 0x04
+#define BD_INTR 0x08
+#define BD_RROR 0x10
+#define BD_LAST 0x20
+#define BD_EXTD 0x80
+
+/*
+ * Data Node descriptor status values.
+ */
+#define DND_END_OF_FRAME 0x80
+#define DND_END_OF_XFER 0x40
+#define DND_DONE 0x20
+#define DND_UNUSED 0x01
+
+/*
+ * IPCV2 descriptor status values.
+ */
+#define BD_IPCV2_END_OF_FRAME 0x40
+
+#define IPCV2_MAX_NODES 50
+/*
+ * Error bit set in the CCB status field by the SDMA,
+ * in setbd routine, in case of a transfer error
+ */
+#define DATA_ERROR 0x10000000
+
+/*
+ * Buffer descriptor commands.
+ */
+#define C0_ADDR 0x01
+#define C0_LOAD 0x02
+#define C0_DUMP 0x03
+#define C0_SETCTX 0x07
+#define C0_GETCTX 0x03
+#define C0_SETDM 0x01
+#define C0_SETPM 0x04
+#define C0_GETDM 0x02
+#define C0_GETPM 0x08
+/*
+ * Change endianness indicator in the BD command field
+ */
+#define CHANGE_ENDIANNESS 0x80
+
+/*
+ * Mode/Count of data node descriptors - IPCv2
+ */
+struct sdma_mode_count {
+ u32 count : 16; /* size of the buffer pointed by this BD */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 command : 8; /* command mostlky used for channel 0 */
+};
+
+/*
+ * Buffer descriptor
+ */
+struct sdma_buffer_descriptor {
+ struct sdma_mode_count mode;
+ dma_addr_t buffer_addr; /* address of the buffer described */
+ dma_addr_t ext_buffer_addr; /* extended buffer address */
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_channel_control - Channel control Block
+ *
+ * @current_bd_ptr current buffer descriptor processed
+ * @base_bd_ptr first element of buffer descriptor array
+ * @unused padding. The SDMA engine expects an array of 128 byte
+ * control blocks
+ */
+struct sdma_channel_control {
+ dma_addr_t current_bd_ptr;
+ dma_addr_t base_bd_ptr;
+ u32 unused[2];
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_state_registers - SDMA context for a channel
+ *
+ * @pc: program counter
+ * @t: test bit: status of arithmetic & test instruction
+ * @rpc: return program counter
+ * @sf: source fault while loading data
+ * @spc: loop start program counter
+ * @df: destination fault while storing data
+ * @epc: loop end program counter
+ * @lm: loop mode
+ */
+struct sdma_state_registers {
+ u32 pc :14;
+ u32 unused1: 1;
+ u32 t : 1;
+ u32 rpc :14;
+ u32 unused0: 1;
+ u32 sf : 1;
+ u32 spc :14;
+ u32 unused2: 1;
+ u32 df : 1;
+ u32 epc :14;
+ u32 lm : 2;
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_context_data - sdma context specific to a channel
+ *
+ * @channel_state: channel state bits
+ * @gReg: general registers
+ * @mda: burst dma destination address register
+ * @msa: burst dma source address register
+ * @ms: burst dma status register
+ * @md: burst dma data register
+ * @pda: peripheral dma destination address register
+ * @psa: peripheral dma source address register
+ * @ps: peripheral dma status register
+ * @pd: peripheral dma data register
+ * @ca: CRC polynomial register
+ * @cs: CRC accumulator register
+ * @dda: dedicated core destination address register
+ * @dsa: dedicated core source address register
+ * @ds: dedicated core status register
+ * @dd: dedicated core data register
+ */
+struct sdma_context_data {
+ struct sdma_state_registers channel_state;
+ u32 gReg[8];
+ u32 mda;
+ u32 msa;
+ u32 ms;
+ u32 md;
+ u32 pda;
+ u32 psa;
+ u32 ps;
+ u32 pd;
+ u32 ca;
+ u32 cs;
+ u32 dda;
+ u32 dsa;
+ u32 ds;
+ u32 dd;
+ u32 scratch0;
+ u32 scratch1;
+ u32 scratch2;
+ u32 scratch3;
+ u32 scratch4;
+ u32 scratch5;
+ u32 scratch6;
+ u32 scratch7;
+} __attribute__ ((packed));
+
+#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
+
+struct sdma_engine;
+
+/**
+ * struct sdma_channel - housekeeping for a SDMA channel
+ *
+ * @sdma pointer to the SDMA engine for this channel
+ * @channel the channel number, matches dmaengine chan_id
+ * @direction transfer type. Needed for setting SDMA script
+ * @peripheral_type Peripheral type. Needed for setting SDMA script
+ * @event_id0 aka dma request line
+ * @event_id1 for channels that use 2 events
+ * @word_size peripheral access size
+ * @buf_tail ID of the buffer that was processed
+ * @done channel completion
+ * @num_bd max NUM_BD. number of descriptors currently handling
+ */
+struct sdma_channel {
+ struct sdma_engine *sdma;
+ unsigned int channel;
+ enum dma_data_direction direction;
+ enum sdma_peripheral_type peripheral_type;
+ unsigned int event_id0;
+ unsigned int event_id1;
+ enum dma_slave_buswidth word_size;
+ unsigned int buf_tail;
+ struct completion done;
+ unsigned int num_bd;
+ struct sdma_buffer_descriptor *bd;
+ dma_addr_t bd_phys;
+ unsigned int pc_from_device, pc_to_device;
+ unsigned long flags;
+ dma_addr_t per_address;
+ u32 event_mask0, event_mask1;
+ u32 watermark_level;
+ u32 shp_addr, per_addr;
+ struct dma_chan chan;
+ spinlock_t lock;
+ struct dma_async_tx_descriptor desc;
+ dma_cookie_t last_completed;
+ enum dma_status status;
+};
+
+#define IMX_DMA_SG_LOOP (1 << 0)
+
+#define MAX_DMA_CHANNELS 32
+#define MXC_SDMA_DEFAULT_PRIORITY 1
+#define MXC_SDMA_MIN_PRIORITY 1
+#define MXC_SDMA_MAX_PRIORITY 7
+
+/**
+ * struct sdma_script_start_addrs - SDMA script start pointers
+ *
+ * start addresses of the different functions in the physical
+ * address space of the SDMA engine.
+ */
+struct sdma_script_start_addrs {
+ u32 ap_2_ap_addr;
+ u32 ap_2_bp_addr;
+ u32 ap_2_ap_fixed_addr;
+ u32 bp_2_ap_addr;
+ u32 loopback_on_dsp_side_addr;
+ u32 mcu_interrupt_only_addr;
+ u32 firi_2_per_addr;
+ u32 firi_2_mcu_addr;
+ u32 per_2_firi_addr;
+ u32 mcu_2_firi_addr;
+ u32 uart_2_per_addr;
+ u32 uart_2_mcu_addr;
+ u32 per_2_app_addr;
+ u32 mcu_2_app_addr;
+ u32 per_2_per_addr;
+ u32 uartsh_2_per_addr;
+ u32 uartsh_2_mcu_addr;
+ u32 per_2_shp_addr;
+ u32 mcu_2_shp_addr;
+ u32 ata_2_mcu_addr;
+ u32 mcu_2_ata_addr;
+ u32 app_2_per_addr;
+ u32 app_2_mcu_addr;
+ u32 shp_2_per_addr;
+ u32 shp_2_mcu_addr;
+ u32 mshc_2_mcu_addr;
+ u32 mcu_2_mshc_addr;
+ u32 spdif_2_mcu_addr;
+ u32 mcu_2_spdif_addr;
+ u32 asrc_2_mcu_addr;
+ u32 ext_mem_2_ipu_addr;
+ u32 descrambler_addr;
+ u32 dptc_dvfs_addr;
+ u32 utra_addr;
+ u32 ram_code_start_addr;
+};
+
+#define SDMA_FIRMWARE_MAGIC 0x414d4453
+
+/**
+ * struct sdma_firmware_header - Layout of the firmware image
+ *
+ * @magic "SDMA"
+ * @version_major increased whenever layout of struct sdma_script_start_addrs
+ * changes.
+ * @version_minor firmware minor version (for binary compatible changes)
+ * @script_addrs_start offset of struct sdma_script_start_addrs in this image
+ * @num_script_addrs Number of script addresses in this image
+ * @ram_code_start offset of SDMA ram image in this firmware image
+ * @ram_code_size size of SDMA ram image
+ * @script_addrs Stores the start address of the SDMA scripts
+ * (in SDMA memory space)
+ */
+struct sdma_firmware_header {
+ u32 magic;
+ u32 version_major;
+ u32 version_minor;
+ u32 script_addrs_start;
+ u32 num_script_addrs;
+ u32 ram_code_start;
+ u32 ram_code_size;
+};
+
+struct sdma_engine {
+ struct device *dev;
+ struct sdma_channel channel[MAX_DMA_CHANNELS];
+ struct sdma_channel_control *channel_control;
+ void __iomem *regs;
+ unsigned int version;
+ unsigned int num_events;
+ struct sdma_context_data *context;
+ dma_addr_t context_phys;
+ struct dma_device dma_device;
+ struct clk *clk;
+ struct sdma_script_start_addrs *script_addrs;
+};
+
+#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
+#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
+#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
+#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
+
+static inline u32 chnenbl_ofs(struct sdma_engine *sdma, unsigned int event)
+{
+ u32 chnenbl0 = (sdma->version == 2 ? SDMA_CHNENBL0_V2 : SDMA_CHNENBL0_V1);
+
+ return chnenbl0 + event * 4;
+}
+
+static int sdma_config_ownership(struct sdma_channel *sdmac,
+ bool event_override, bool mcu_verride, bool dsp_override)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 evt, mcu, dsp;
+
+ if (event_override && mcu_verride && dsp_override)
+ return -EINVAL;
+
+ evt = readl(sdma->regs + SDMA_H_EVTOVR);
+ mcu = readl(sdma->regs + SDMA_H_HOSTOVR);
+ dsp = readl(sdma->regs + SDMA_H_DSPOVR);
+
+ if (dsp_override)
+ dsp &= ~(1 << channel);
+ else
+ dsp |= (1 << channel);
+
+ if (event_override)
+ evt &= ~(1 << channel);
+ else
+ evt |= (1 << channel);
+
+ if (mcu_verride)
+ mcu &= ~(1 << channel);
+ else
+ mcu |= (1 << channel);
+
+ writel(evt, sdma->regs + SDMA_H_EVTOVR);
+ writel(mcu, sdma->regs + SDMA_H_HOSTOVR);
+ writel(dsp, sdma->regs + SDMA_H_DSPOVR);
+
+ return 0;
+}
+
+/*
+ * sdma_run_channel - run a channel and wait till it's done
+ */
+static int sdma_run_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret;
+
+ init_completion(&sdmac->done);
+
+ writel(1 << channel, sdma->regs + SDMA_H_START);
+
+ ret = wait_for_completion_timeout(&sdmac->done, HZ);
+
+ return ret ? 0 : -ETIMEDOUT;
+}
+
+static int sdma_load_script(struct sdma_engine *sdma, void *buf, int size,
+ u32 address)
+{
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ void *buf_virt;
+ dma_addr_t buf_phys;
+ int ret;
+
+ buf_virt = dma_alloc_coherent(NULL,
+ size,
+ &buf_phys, GFP_KERNEL);
+ if (!buf_virt)
+ return -ENOMEM;
+
+ bd0->mode.command = C0_SETPM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = size / 2;
+ bd0->buffer_addr = buf_phys;
+ bd0->ext_buffer_addr = address;
+
+ memcpy(buf_virt, buf, size);
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ dma_free_coherent(NULL, size, buf_virt, buf_phys);
+
+ return ret;
+}
+
+static void sdma_event_enable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 val;
+ u32 chnenbl = chnenbl_ofs(sdma, event);
+
+ val = readl(sdma->regs + chnenbl);
+ val |= (1 << channel);
+ writel(val, sdma->regs + chnenbl);
+}
+
+static void sdma_event_disable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 chnenbl = chnenbl_ofs(sdma, event);
+ u32 val;
+
+ val = readl(sdma->regs + chnenbl);
+ val &= ~(1 << channel);
+ writel(val, sdma->regs + chnenbl);
+}
+
+static void sdma_handle_channel_loop(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+
+ /*
+ * loop mode. Iterate over descriptors, re-setup them and
+ * call callback function.
+ */
+ while (1) {
+ bd = &sdmac->bd[sdmac->buf_tail];
+
+ if (bd->mode.status & BD_DONE)
+ break;
+
+ if (bd->mode.status & BD_RROR)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ bd->mode.status |= BD_DONE;
+ sdmac->buf_tail++;
+ sdmac->buf_tail %= sdmac->num_bd;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ }
+}
+
+static void mxc_sdma_handle_channel_normal(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+ int i, error = 0;
+
+ /*
+ * non loop mode. Iterate over all descriptors, collect
+ * errors and call callback function
+ */
+ for (i = 0; i < sdmac->num_bd; i++) {
+ bd = &sdmac->bd[i];
+
+ if (bd->mode.status & (BD_DONE | BD_RROR))
+ error = -EIO;
+ }
+
+ if (error)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ sdmac->last_completed = sdmac->desc.cookie;
+}
+
+static void mxc_sdma_handle_channel(struct sdma_channel *sdmac)
+{
+ complete(&sdmac->done);
+
+ /* not interested in channel 0 interrupts */
+ if (sdmac->channel == 0)
+ return;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP)
+ sdma_handle_channel_loop(sdmac);
+ else
+ mxc_sdma_handle_channel_normal(sdmac);
+}
+
+static irqreturn_t sdma_int_handler(int irq, void *dev_id)
+{
+ struct sdma_engine *sdma = dev_id;
+ u32 stat;
+
+ stat = readl(sdma->regs + SDMA_H_INTR);
+ writel(stat, sdma->regs + SDMA_H_INTR);
+
+ while (stat) {
+ int channel = fls(stat) - 1;
+ struct sdma_channel *sdmac = &sdma->channel[channel];
+
+ mxc_sdma_handle_channel(sdmac);
+
+ stat &= ~(1 << channel);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * sets the pc of SDMA script according to the peripheral type
+ */
+static void sdma_get_pc(struct sdma_channel *sdmac,
+ enum sdma_peripheral_type peripheral_type)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int per_2_emi = 0, emi_2_per = 0;
+ /*
+ * These are needed once we start to support transfers between
+ * two peripherals or memory-to-memory transfers
+ */
+ int per_2_per = 0, emi_2_emi = 0;
+
+ sdmac->pc_from_device = 0;
+ sdmac->pc_to_device = 0;
+
+ switch (peripheral_type) {
+ case IMX_DMATYPE_MEMORY:
+ emi_2_emi = sdma->script_addrs->ap_2_ap_addr;
+ break;
+ case IMX_DMATYPE_DSP:
+ emi_2_per = sdma->script_addrs->bp_2_ap_addr;
+ per_2_emi = sdma->script_addrs->ap_2_bp_addr;
+ break;
+ case IMX_DMATYPE_FIRI:
+ per_2_emi = sdma->script_addrs->firi_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_firi_addr;
+ break;
+ case IMX_DMATYPE_UART:
+ per_2_emi = sdma->script_addrs->uart_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_UART_SP:
+ per_2_emi = sdma->script_addrs->uartsh_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ATA:
+ per_2_emi = sdma->script_addrs->ata_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_ata_addr;
+ break;
+ case IMX_DMATYPE_CSPI:
+ case IMX_DMATYPE_EXT:
+ case IMX_DMATYPE_SSI:
+ per_2_emi = sdma->script_addrs->app_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_SSI_SP:
+ case IMX_DMATYPE_MMC:
+ case IMX_DMATYPE_SDHC:
+ case IMX_DMATYPE_CSPI_SP:
+ case IMX_DMATYPE_ESAI:
+ case IMX_DMATYPE_MSHC_SP:
+ per_2_emi = sdma->script_addrs->shp_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ASRC:
+ per_2_emi = sdma->script_addrs->asrc_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->asrc_2_mcu_addr;
+ per_2_per = sdma->script_addrs->per_2_per_addr;
+ break;
+ case IMX_DMATYPE_MSHC:
+ per_2_emi = sdma->script_addrs->mshc_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_mshc_addr;
+ break;
+ case IMX_DMATYPE_CCM:
+ per_2_emi = sdma->script_addrs->dptc_dvfs_addr;
+ break;
+ case IMX_DMATYPE_SPDIF:
+ per_2_emi = sdma->script_addrs->spdif_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_spdif_addr;
+ break;
+ case IMX_DMATYPE_IPU_MEMORY:
+ emi_2_per = sdma->script_addrs->ext_mem_2_ipu_addr;
+ break;
+ default:
+ break;
+ }
+
+ sdmac->pc_from_device = per_2_emi;
+ sdmac->pc_to_device = emi_2_per;
+}
+
+static int sdma_load_context(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int load_address;
+ struct sdma_context_data *context = sdma->context;
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ int ret;
+
+ if (sdmac->direction == DMA_FROM_DEVICE) {
+ load_address = sdmac->pc_from_device;
+ } else {
+ load_address = sdmac->pc_to_device;
+ }
+
+ if (load_address < 0)
+ return load_address;
+
+ dev_dbg(sdma->dev, "load_address = %d\n", load_address);
+ dev_dbg(sdma->dev, "wml = 0x%08x\n", sdmac->watermark_level);
+ dev_dbg(sdma->dev, "shp_addr = 0x%08x\n", sdmac->shp_addr);
+ dev_dbg(sdma->dev, "per_addr = 0x%08x\n", sdmac->per_addr);
+ dev_dbg(sdma->dev, "event_mask0 = 0x%08x\n", sdmac->event_mask0);
+ dev_dbg(sdma->dev, "event_mask1 = 0x%08x\n", sdmac->event_mask1);
+
+ memset(context, 0, sizeof(*context));
+ context->channel_state.pc = load_address;
+
+ /* Send by context the event mask,base address for peripheral
+ * and watermark level
+ */
+ context->gReg[0] = sdmac->event_mask1;
+ context->gReg[1] = sdmac->event_mask0;
+ context->gReg[2] = sdmac->per_addr;
+ context->gReg[6] = sdmac->shp_addr;
+ context->gReg[7] = sdmac->watermark_level;
+
+ bd0->mode.command = C0_SETDM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = sizeof(*context) / 4;
+ bd0->buffer_addr = sdma->context_phys;
+ bd0->ext_buffer_addr = 2048 + (sizeof(*context) / 4) * channel;
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ return ret;
+}
+
+static void sdma_disable_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ writel(1 << channel, sdma->regs + SDMA_H_STATSTOP);
+ sdmac->status = DMA_ERROR;
+}
+
+static int sdma_config_channel(struct sdma_channel *sdmac)
+{
+ int ret;
+
+ sdma_disable_channel(sdmac);
+
+ sdmac->event_mask0 = 0;
+ sdmac->event_mask1 = 0;
+ sdmac->shp_addr = 0;
+ sdmac->per_addr = 0;
+
+ if (sdmac->event_id0) {
+ if (sdmac->event_id0 > 32)
+ return -EINVAL;
+ sdma_event_enable(sdmac, sdmac->event_id0);
+ }
+
+ switch (sdmac->peripheral_type) {
+ case IMX_DMATYPE_DSP:
+ sdma_config_ownership(sdmac, false, true, true);
+ break;
+ case IMX_DMATYPE_MEMORY:
+ sdma_config_ownership(sdmac, false, true, false);
+ break;
+ default:
+ sdma_config_ownership(sdmac, true, true, false);
+ break;
+ }
+
+ sdma_get_pc(sdmac, sdmac->peripheral_type);
+
+ if ((sdmac->peripheral_type != IMX_DMATYPE_MEMORY) &&
+ (sdmac->peripheral_type != IMX_DMATYPE_DSP)) {
+ /* Handle multiple event channels differently */
+ if (sdmac->event_id1) {
+ sdmac->event_mask1 = 1 << (sdmac->event_id1 % 32);
+ if (sdmac->event_id1 > 31)
+ sdmac->watermark_level |= 1 << 31;
+ sdmac->event_mask0 = 1 << (sdmac->event_id0 % 32);
+ if (sdmac->event_id0 > 31)
+ sdmac->watermark_level |= 1 << 30;
+ } else {
+ sdmac->event_mask0 = 1 << sdmac->event_id0;
+ sdmac->event_mask1 = 1 << (sdmac->event_id0 - 32);
+ }
+ /* Watermark Level */
+ sdmac->watermark_level |= sdmac->watermark_level;
+ /* Address */
+ sdmac->shp_addr = sdmac->per_address;
+ } else {
+ sdmac->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
+ }
+
+ ret = sdma_load_context(sdmac);
+
+ return ret;
+}
+
+static int sdma_set_channel_priority(struct sdma_channel *sdmac,
+ unsigned int priority)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ if (priority < MXC_SDMA_MIN_PRIORITY
+ || priority > MXC_SDMA_MAX_PRIORITY) {
+ return -EINVAL;
+ }
+
+ writel(priority, sdma->regs + SDMA_CHNPRI_0 + 4 * channel);
+
+ return 0;
+}
+
+static int sdma_request_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret = -EBUSY;
+
+ sdmac->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdmac->bd_phys, GFP_KERNEL);
+ if (!sdmac->bd) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memset(sdmac->bd, 0, PAGE_SIZE);
+
+ sdma->channel_control[channel].base_bd_ptr = sdmac->bd_phys;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ clk_enable(sdma->clk);
+
+ sdma_set_channel_priority(sdmac, MXC_SDMA_DEFAULT_PRIORITY);
+
+ init_completion(&sdmac->done);
+
+ sdmac->buf_tail = 0;
+
+ return 0;
+out:
+
+ return ret;
+}
+
+static void sdma_enable_channel(struct sdma_engine *sdma, int channel)
+{
+ writel(1 << channel, sdma->regs + SDMA_H_START);
+}
+
+static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
+{
+ dma_cookie_t cookie = sdma->chan.cookie;
+
+ if (++cookie < 0)
+ cookie = 1;
+
+ sdma->chan.cookie = cookie;
+ sdma->desc.cookie = cookie;
+
+ return cookie;
+}
+
+static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sdma_channel, chan);
+}
+
+static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(tx->chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ dma_cookie_t cookie;
+
+ spin_lock_irq(&sdmac->lock);
+
+ cookie = sdma_assign_cookie(sdmac);
+
+ sdma_enable_channel(sdma, tx->chan->chan_id);
+
+ spin_unlock_irq(&sdmac->lock);
+
+ return cookie;
+}
+
+static int sdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct imx_dma_data *data = chan->private;
+ int prio, ret;
+
+ /* No need to execute this for internal channel 0 */
+ if (chan->chan_id == 0)
+ return 0;
+
+ if (!data)
+ return -EINVAL;
+
+ switch (data->priority) {
+ case DMA_PRIO_HIGH:
+ prio = 3;
+ break;
+ case DMA_PRIO_MEDIUM:
+ prio = 2;
+ break;
+ case DMA_PRIO_LOW:
+ default:
+ prio = 1;
+ break;
+ }
+
+ sdmac->peripheral_type = data->peripheral_type;
+ sdmac->event_id0 = data->dma_request;
+ ret = sdma_set_channel_priority(sdmac, prio);
+ if (ret)
+ return ret;
+
+ ret = sdma_request_channel(sdmac);
+ if (ret)
+ return ret;
+
+ dma_async_tx_descriptor_init(&sdmac->desc, chan);
+ sdmac->desc.tx_submit = sdma_tx_submit;
+ /* txd.flags will be overwritten in prep funcs */
+ sdmac->desc.flags = DMA_CTRL_ACK;
+
+ return 0;
+}
+
+static void sdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+
+ sdma_disable_channel(sdmac);
+
+ if (sdmac->event_id0)
+ sdma_event_disable(sdmac, sdmac->event_id0);
+ if (sdmac->event_id1)
+ sdma_event_disable(sdmac, sdmac->event_id1);
+
+ sdmac->event_id0 = 0;
+ sdmac->event_id1 = 0;
+
+ sdma_set_channel_priority(sdmac, 0);
+
+ dma_free_coherent(NULL, PAGE_SIZE, sdmac->bd, sdmac->bd_phys);
+
+ clk_disable(sdma->clk);
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int ret, i, count;
+ int channel = chan->chan_id;
+ struct scatterlist *sg;
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags = 0;
+
+ dev_dbg(sdma->dev, "setting up %d entries for channel %d.\n",
+ sg_len, channel);
+
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (sg_len > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, sg_len, NUM_BD);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = sgl->dma_address;
+
+ count = sg->length;
+
+ if (count > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
+ channel, count, 0xffff);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ bd->mode.count = count;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES) {
+ ret = -EINVAL;
+ goto err_out;
+ }
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP) {
+ param |= BD_INTR;
+ if (i + 1 == sg_len)
+ param |= BD_WRAP;
+ }
+
+ if (i + 1 == sg_len)
+ param |= BD_INTR;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, count, sg->dma_address,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+ }
+
+ sdmac->num_bd = sg_len;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int num_periods = buf_len / period_len;
+ int channel = chan->chan_id;
+ int ret, i = 0, buf = 0;
+
+ dev_dbg(sdma->dev, "%s channel: %d\n", __func__, channel);
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags |= IMX_DMA_SG_LOOP;
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (num_periods > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, num_periods, NUM_BD);
+ goto err_out;
+ }
+
+ if (period_len > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum period size exceeded: %d > %d\n",
+ channel, period_len, 0xffff);
+ goto err_out;
+ }
+
+ while (buf < buf_len) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = dma_addr;
+
+ bd->mode.count = period_len;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES)
+ goto err_out;
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
+ if (i + 1 == num_periods)
+ param |= BD_WRAP;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, period_len, dma_addr,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+
+ dma_addr += period_len;
+ buf += period_len;
+
+ i++;
+ }
+
+ sdmac->num_bd = num_periods;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ sdmac->status = DMA_ERROR;
+ return NULL;
+}
+
+static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct dma_slave_config *dmaengine_cfg = (void *)arg;
+
+ switch (cmd) {
+ case DMA_TERMINATE_ALL:
+ sdma_disable_channel(sdmac);
+ return 0;
+ case DMA_SLAVE_CONFIG:
+ if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
+ sdmac->per_address = dmaengine_cfg->src_addr;
+ sdmac->watermark_level = dmaengine_cfg->src_maxburst;
+ sdmac->word_size = dmaengine_cfg->src_addr_width;
+ } else {
+ sdmac->per_address = dmaengine_cfg->dst_addr;
+ sdmac->watermark_level = dmaengine_cfg->dst_maxburst;
+ sdmac->word_size = dmaengine_cfg->dst_addr_width;
+ }
+ return sdma_config_channel(sdmac);
+ default:
+ return -ENOSYS;
+ }
+
+ return -EINVAL;
+}
+
+static enum dma_status sdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ dma_cookie_t last_used;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+
+ ret = dma_async_is_complete(cookie, sdmac->last_completed, last_used);
+ dma_set_tx_state(txstate, sdmac->last_completed, last_used, 0);
+
+ return ret;
+}
+
+static void sdma_issue_pending(struct dma_chan *chan)
+{
+ /*
+ * Nothing to do. We only have a single descriptor
+ */
+}
+
+static int __init sdma_init(struct sdma_engine *sdma,
+ void *ram_code, int ram_code_size)
+{
+ int i, ret;
+ dma_addr_t ccb_phys;
+
+ switch (sdma->version) {
+ case 1:
+ sdma->num_events = 32;
+ break;
+ case 2:
+ sdma->num_events = 48;
+ break;
+ default:
+ dev_err(sdma->dev, "Unknown version %d. aborting\n", sdma->version);
+ return -ENODEV;
+ }
+
+ clk_enable(sdma->clk);
+
+ /* Be sure SDMA has not started yet */
+ writel(0, sdma->regs + SDMA_H_C0PTR);
+
+ sdma->channel_control = dma_alloc_coherent(NULL,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
+ sizeof(struct sdma_context_data),
+ &ccb_phys, GFP_KERNEL);
+
+ if (!sdma->channel_control) {
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ sdma->context = (void *)sdma->channel_control +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+ sdma->context_phys = ccb_phys +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+
+ /* Zero-out the CCB structures array just allocated */
+ memset(sdma->channel_control, 0,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
+
+ /* disable all channels */
+ for (i = 0; i < sdma->num_events; i++)
+ writel(0, sdma->regs + chnenbl_ofs(sdma, i));
+
+ /* All channels have priority 0 */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++)
+ writel(0, sdma->regs + SDMA_CHNPRI_0 + i * 4);
+
+ ret = sdma_request_channel(&sdma->channel[0]);
+ if (ret)
+ goto err_dma_alloc;
+
+ sdma_config_ownership(&sdma->channel[0], false, true, false);
+
+ /* Set Command Channel (Channel Zero) */
+ writel(0x4050, sdma->regs + SDMA_CHN0ADDR);
+
+ /* Set bits of CONFIG register but with static context switching */
+ /* FIXME: Check whether to set ACR bit depending on clock ratios */
+ writel(0, sdma->regs + SDMA_H_CONFIG);
+
+ writel(ccb_phys, sdma->regs + SDMA_H_C0PTR);
+
+ /* download the RAM image for SDMA */
+ sdma_load_script(sdma, ram_code,
+ ram_code_size,
+ sdma->script_addrs->ram_code_start_addr);
+
+ /* Set bits of CONFIG register with given context switching mode */
+ writel(SDMA_H_CONFIG_CSM, sdma->regs + SDMA_H_CONFIG);
+
+ /* Initializes channel's priorities */
+ sdma_set_channel_priority(&sdma->channel[0], 7);
+
+ clk_disable(sdma->clk);
+
+ return 0;
+
+err_dma_alloc:
+ clk_disable(sdma->clk);
+ dev_err(sdma->dev, "initialisation failed with %d\n", ret);
+ return ret;
+}
+
+static int __devinit sdma_probe(struct platform_device *pdev)
+{
+ int ret;
+ const struct firmware *fw;
+ const struct sdma_firmware_header *header;
+ const struct sdma_script_start_addrs *addr;
+ int irq;
+ unsigned short *ram_code;
+ struct resource *iores;
+ struct sdma_platform_data *pdata = pdev->dev.platform_data;
+ char *fwname;
+ int i;
+ dma_cap_mask_t mask;
+ struct sdma_engine *sdma;
+
+ sdma = kzalloc(sizeof(*sdma), GFP_KERNEL);
+ if (!sdma)
+ return -ENOMEM;
+
+ sdma->dev = &pdev->dev;
+
+ iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ irq = platform_get_irq(pdev, 0);
+ if (!iores || irq < 0 || !pdata) {
+ ret = -EINVAL;
+ goto err_irq;
+ }
+
+ sdma->clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(sdma->clk)) {
+ ret = PTR_ERR(sdma->clk);
+ goto err_clk;
+ }
+
+ sdma->regs = ioremap(iores->start, resource_size(iores));
+ if (!sdma->regs) {
+ ret = -ENOMEM;
+ goto err_ioremap;
+ }
+
+ ret = request_irq(irq, sdma_int_handler, 0, "sdma", sdma);
+ if (ret)
+ goto err_request_irq;
+
+ fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin",
+ pdata->cpu_name, pdata->to_version);
+ if (!fwname) {
+ ret = -ENOMEM;
+ goto err_cputype;
+ }
+
+ ret = request_firmware(&fw, fwname, &pdev->dev);
+ if (ret) {
+ dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
+ fwname, ret);
+ kfree(fwname);
+ goto err_cputype;
+ }
+ kfree(fwname);
+
+ if (fw->size < sizeof(*header))
+ goto err_firmware;
+
+ header = (struct sdma_firmware_header *)fw->data;
+
+ if (header->magic != SDMA_FIRMWARE_MAGIC)
+ goto err_firmware;
+ if (header->ram_code_start + header->ram_code_size > fw->size)
+ goto err_firmware;
+
+ addr = (void *)header + header->script_addrs_start;
+ ram_code = (void *)header + header->ram_code_start;
+ sdma->script_addrs = kmalloc(sizeof(*addr), GFP_KERNEL);
+ if (!sdma->script_addrs)
+ goto err_firmware;
+ memcpy(sdma->script_addrs, addr, sizeof(*addr));
+
+ sdma->version = pdata->sdma_version;
+
+ INIT_LIST_HEAD(&sdma->dma_device.channels);
+ /* Initialize channel parameters */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++) {
+ struct sdma_channel *sdmac = &sdma->channel[i];
+
+ sdmac->sdma = sdma;
+ spin_lock_init(&sdmac->lock);
+
+ dma_cap_set(DMA_SLAVE, sdma->dma_device.cap_mask);
+ dma_cap_set(DMA_CYCLIC, sdma->dma_device.cap_mask);
+
+ sdmac->chan.device = &sdma->dma_device;
+ sdmac->chan.chan_id = i;
+ sdmac->channel = i;
+
+ /* Add the channel to the DMAC list */
+ list_add_tail(&sdmac->chan.device_node, &sdma->dma_device.channels);
+ }
+
+ ret = sdma_init(sdma, ram_code, header->ram_code_size);
+ if (ret)
+ goto err_init;
+
+ sdma->dma_device.dev = &pdev->dev;
+
+ sdma->dma_device.device_alloc_chan_resources = sdma_alloc_chan_resources;
+ sdma->dma_device.device_free_chan_resources = sdma_free_chan_resources;
+ sdma->dma_device.device_tx_status = sdma_tx_status;
+ sdma->dma_device.device_prep_slave_sg = sdma_prep_slave_sg;
+ sdma->dma_device.device_prep_dma_cyclic = sdma_prep_dma_cyclic;
+ sdma->dma_device.device_control = sdma_control;
+ sdma->dma_device.device_issue_pending = sdma_issue_pending;
+
+ ret = dma_async_device_register(&sdma->dma_device);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register\n");
+ goto err_init;
+ }
+
+ dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
+ header->version_major,
+ header->version_minor);
+
+ /* request channel 0. This is an internal control channel
+ * to the SDMA engine and not available to clients.
+ */
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_request_channel(mask, NULL, NULL);
+
+ release_firmware(fw);
+
+ return 0;
+
+err_init:
+ kfree(sdma->script_addrs);
+err_firmware:
+ release_firmware(fw);
+err_cputype:
+ free_irq(irq, sdma);
+err_request_irq:
+ iounmap(sdma->regs);
+err_ioremap:
+ clk_put(sdma->clk);
+err_clk:
+err_irq:
+ kfree(sdma);
+ return 0;
+}
+
+static int __devexit sdma_remove(struct platform_device *pdev)
+{
+ return -EBUSY;
+}
+
+static struct platform_driver sdma_driver = {
+ .driver = {
+ .name = "imx-sdma",
+ },
+ .probe = sdma_probe,
+ .remove = __devexit_p(sdma_remove),
+};
+
+static int __init sdma_module_init(void)
+{
+ return platform_driver_register(&sdma_driver);
+}
+subsys_initcall(sdma_module_init);
+
+MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("i.MX SDMA driver");
+MODULE_LICENSE("GPL");
--
1.7.1
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply related [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-23 12:57 ` Sascha Hauer
@ 2010-08-24 7:58 ` Lothar Waßmann
-1 siblings, 0 replies; 78+ messages in thread
From: Lothar Waßmann @ 2010-08-24 7:58 UTC (permalink / raw)
To: Sascha Hauer; +Cc: linux-kernel, Linus Walleij, Dan Williams, linux-arm-kernel
Hi,
> This patch adds support for the Freescale i.MX SDMA engine.
>
> The SDMA engine is a scatter/gather DMA engine which is implemented
> as a seperate coprocessor. SDMA needs its own firmware which is
> requested using the standard request_firmware mechanism. The firmware
> has different entry points for each peripheral type, so drivers
> have to pass the peripheral type to the DMA engine which in turn
> picks the correct firmware entry point from a table contained in
> the firmware image itself.
> The original Freescale code also supports support for transfering
> data to the internal SRAM which needs different entry points to
> the firmware. Support for this is currently not implemented. Also,
> support for the ASRC (asymmetric sample rate converter) is skipped.
>
> I took a very simple approach to implement dmaengine support. Only
> a single descriptor is statically assigned to a each channel. This
> means that transfers can't be queued up but only a single transfer
> is in progress. This simplifies implementation a lot and is sufficient
> for the usual device/memory transfers.
>
> Changes since v1:
>
[...]
> +static int __devinit sdma_probe(struct platform_device *pdev)
>
__devinit/__devexit is for hot-pluggable devices. I don't think the
SDMA controller is hot-pluggable, so __init/__exit could be used here.
> +{
> + int ret;
> + const struct firmware *fw;
> + const struct sdma_firmware_header *header;
> + const struct sdma_script_start_addrs *addr;
> + int irq;
> + unsigned short *ram_code;
> + struct resource *iores;
> + struct sdma_platform_data *pdata = pdev->dev.platform_data;
> + char *fwname;
> + int i;
> + dma_cap_mask_t mask;
> + struct sdma_engine *sdma;
> +
> + sdma = kzalloc(sizeof(*sdma), GFP_KERNEL);
> + if (!sdma)
> + return -ENOMEM;
> +
> + sdma->dev = &pdev->dev;
> +
> + iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + irq = platform_get_irq(pdev, 0);
> + if (!iores || irq < 0 || !pdata) {
> + ret = -EINVAL;
> + goto err_irq;
> + }
> +
> + sdma->clk = clk_get(&pdev->dev, NULL);
> + if (IS_ERR(sdma->clk)) {
> + ret = PTR_ERR(sdma->clk);
> + goto err_clk;
> + }
> +
request_mem_region()?
> + sdma->regs = ioremap(iores->start, resource_size(iores));
> + if (!sdma->regs) {
> + ret = -ENOMEM;
> + goto err_ioremap;
> + }
> +
> +
> +static int __devexit sdma_remove(struct platform_device *pdev)
>
see above.
Lothar Waßmann
--
___________________________________________________________
Ka-Ro electronics GmbH | Pascalstraße 22 | D - 52076 Aachen
Phone: +49 2408 1402-0 | Fax: +49 2408 1402-10
Geschäftsführer: Matthias Kaussen
Handelsregistereintrag: Amtsgericht Aachen, HRB 4996
www.karo-electronics.de | info@karo-electronics.de
___________________________________________________________
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-24 7:58 ` Lothar Waßmann
0 siblings, 0 replies; 78+ messages in thread
From: Lothar Waßmann @ 2010-08-24 7:58 UTC (permalink / raw)
To: linux-arm-kernel
Hi,
> This patch adds support for the Freescale i.MX SDMA engine.
>
> The SDMA engine is a scatter/gather DMA engine which is implemented
> as a seperate coprocessor. SDMA needs its own firmware which is
> requested using the standard request_firmware mechanism. The firmware
> has different entry points for each peripheral type, so drivers
> have to pass the peripheral type to the DMA engine which in turn
> picks the correct firmware entry point from a table contained in
> the firmware image itself.
> The original Freescale code also supports support for transfering
> data to the internal SRAM which needs different entry points to
> the firmware. Support for this is currently not implemented. Also,
> support for the ASRC (asymmetric sample rate converter) is skipped.
>
> I took a very simple approach to implement dmaengine support. Only
> a single descriptor is statically assigned to a each channel. This
> means that transfers can't be queued up but only a single transfer
> is in progress. This simplifies implementation a lot and is sufficient
> for the usual device/memory transfers.
>
> Changes since v1:
>
[...]
> +static int __devinit sdma_probe(struct platform_device *pdev)
>
__devinit/__devexit is for hot-pluggable devices. I don't think the
SDMA controller is hot-pluggable, so __init/__exit could be used here.
> +{
> + int ret;
> + const struct firmware *fw;
> + const struct sdma_firmware_header *header;
> + const struct sdma_script_start_addrs *addr;
> + int irq;
> + unsigned short *ram_code;
> + struct resource *iores;
> + struct sdma_platform_data *pdata = pdev->dev.platform_data;
> + char *fwname;
> + int i;
> + dma_cap_mask_t mask;
> + struct sdma_engine *sdma;
> +
> + sdma = kzalloc(sizeof(*sdma), GFP_KERNEL);
> + if (!sdma)
> + return -ENOMEM;
> +
> + sdma->dev = &pdev->dev;
> +
> + iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> + irq = platform_get_irq(pdev, 0);
> + if (!iores || irq < 0 || !pdata) {
> + ret = -EINVAL;
> + goto err_irq;
> + }
> +
> + sdma->clk = clk_get(&pdev->dev, NULL);
> + if (IS_ERR(sdma->clk)) {
> + ret = PTR_ERR(sdma->clk);
> + goto err_clk;
> + }
> +
request_mem_region()?
> + sdma->regs = ioremap(iores->start, resource_size(iores));
> + if (!sdma->regs) {
> + ret = -ENOMEM;
> + goto err_ioremap;
> + }
> +
> +
> +static int __devexit sdma_remove(struct platform_device *pdev)
>
see above.
Lothar Wa?mann
--
___________________________________________________________
Ka-Ro electronics GmbH | Pascalstra?e 22 | D - 52076 Aachen
Phone: +49 2408 1402-0 | Fax: +49 2408 1402-10
Gesch?ftsf?hrer: Matthias Kaussen
Handelsregistereintrag: Amtsgericht Aachen, HRB 4996
www.karo-electronics.de | info at karo-electronics.de
___________________________________________________________
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-24 7:58 ` Lothar Waßmann
@ 2010-08-24 15:01 ` Linus Walleij
-1 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-24 15:01 UTC (permalink / raw)
To: Lothar Waßmann
Cc: Sascha Hauer, linux-kernel, Dan Williams, linux-arm-kernel
2010/8/24 Lothar Waßmann <LW@karo-electronics.de>:
> [Sacha]
> [...]
>> +static int __devinit sdma_probe(struct platform_device *pdev)
>>
> __devinit/__devexit is for hot-pluggable devices. I don't think the
> SDMA controller is hot-pluggable, so __init/__exit could be used here.
Then the .probe member has to be removed from the
struct platform_driver and platform_driver_probe() used instead
of platform_driver_register().
But it'd work fine I think.
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-24 15:01 ` Linus Walleij
0 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-24 15:01 UTC (permalink / raw)
To: linux-arm-kernel
2010/8/24 Lothar Wa?mann <LW@karo-electronics.de>:
> [Sacha]
> [...]
>> +static int __devinit sdma_probe(struct platform_device *pdev)
>>
> __devinit/__devexit is for hot-pluggable devices. I don't think the
> SDMA controller is hot-pluggable, so __init/__exit could be used here.
Then the .probe member has to be removed from the
struct platform_driver and platform_driver_probe() used instead
of platform_driver_register().
But it'd work fine I think.
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v3] dmaengine: Add Freescale i.MX SDMA support
2010-08-16 11:07 ` Sascha Hauer
@ 2010-08-27 12:22 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-27 12:22 UTC (permalink / raw)
To: linux-kernel; +Cc: Linus Walleij, Dan Williams, linux-arm-kernel
This patch adds support for the Freescale i.MX SDMA engine.
The SDMA engine is a scatter/gather DMA engine which is implemented
as a seperate coprocessor. SDMA needs its own firmware which is
requested using the standard request_firmware mechanism. The firmware
has different entry points for each peripheral type, so drivers
have to pass the peripheral type to the DMA engine which in turn
picks the correct firmware entry point from a table contained in
the firmware image itself.
The original Freescale code also supports support for transfering
data to the internal SRAM which needs different entry points to
the firmware. Support for this is currently not implemented. Also,
support for the ASRC (asymmetric sample rate converter) is skipped.
I took a very simple approach to implement dmaengine support. Only
a single descriptor is statically assigned to a each channel. This
means that transfers can't be queued up but only a single transfer
is in progress. This simplifies implementation a lot and is sufficient
for the usual device/memory transfers.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Reviewed-by: Linus Walleij <linus.ml.walleij@gmail.com>
---
changes since v2:
- use __raw_readl/__raw_writel
- use __init/__exit instead of __devinit/__devexit
- use request_mem_region.
arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
arch/arm/mach-mx3/Kconfig | 2 +
arch/arm/plat-mxc/Kconfig | 10 +
arch/arm/plat-mxc/include/mach/dma.h | 64 ++
arch/arm/plat-mxc/include/mach/sdma.h | 17 +
drivers/dma/Kconfig | 8 +
drivers/dma/Makefile | 1 +
drivers/dma/imx-sdma.c | 1392 +++++++++++++++++++++++++++++++
8 files changed, 1496 insertions(+), 6 deletions(-)
create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
create mode 100644 drivers/dma/imx-sdma.c
diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
index 287431c..ac6fd71 100644
--- a/arch/arm/mach-imx/include/mach/dma-v1.h
+++ b/arch/arm/mach-imx/include/mach/dma-v1.h
@@ -27,6 +27,8 @@
#define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
+#include <mach/dma.h>
+
#define IMX_DMA_CHANNELS 16
#define DMA_MODE_READ 0
@@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
void imx_dma_free(int channel);
-enum imx_dma_prio {
- DMA_PRIO_HIGH = 0,
- DMA_PRIO_MEDIUM = 1,
- DMA_PRIO_LOW = 2
-};
-
int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
#endif /* __MACH_DMA_V1_H__ */
diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
index 85beece..301375c 100644
--- a/arch/arm/mach-mx3/Kconfig
+++ b/arch/arm/mach-mx3/Kconfig
@@ -3,12 +3,14 @@ if ARCH_MX3
config ARCH_MX31
select ARCH_HAS_RNGA
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
bool
config ARCH_MX35
bool
select ARCH_MXC_IOMUX_V3
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
comment "MX3 platforms:"
diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
index 0527e65..6741625 100644
--- a/arch/arm/plat-mxc/Kconfig
+++ b/arch/arm/plat-mxc/Kconfig
@@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
config ARCH_MXC_AUDMUX_V2
bool
+config IMX_HAVE_SDMA
+ bool
+
+config IMX_SDMA
+ depends on IMX_HAVE_SDMA
+ tristate "Enable SDMA support"
+ help
+ Include support for the SDMA engine. The SDMA engine needs additional
+ firmware support. SDMA can be compiled as a module to support loading
+ the firmware when a rootfs is present.
endif
diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
new file mode 100644
index 0000000..69d181f
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/dma.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARCH_MXC_DMA_H__
+#define __ASM_ARCH_MXC_DMA_H__
+
+#include <linux/scatterlist.h>
+
+/*
+ * This enumerates peripheral types. Used for SDMA.
+ */
+enum sdma_peripheral_type {
+ IMX_DMATYPE_SSI, /* MCU domain SSI */
+ IMX_DMATYPE_SSI_SP, /* Shared SSI */
+ IMX_DMATYPE_MMC, /* MMC */
+ IMX_DMATYPE_SDHC, /* SDHC */
+ IMX_DMATYPE_UART, /* MCU domain UART */
+ IMX_DMATYPE_UART_SP, /* Shared UART */
+ IMX_DMATYPE_FIRI, /* FIRI */
+ IMX_DMATYPE_CSPI, /* MCU domain CSPI */
+ IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
+ IMX_DMATYPE_SIM, /* SIM */
+ IMX_DMATYPE_ATA, /* ATA */
+ IMX_DMATYPE_CCM, /* CCM */
+ IMX_DMATYPE_EXT, /* External peripheral */
+ IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
+ IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
+ IMX_DMATYPE_DSP, /* DSP */
+ IMX_DMATYPE_MEMORY, /* Memory */
+ IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
+ IMX_DMATYPE_SPDIF, /* SPDIF */
+ IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
+ IMX_DMATYPE_ASRC, /* ASRC */
+ IMX_DMATYPE_ESAI, /* ESAI */
+};
+
+enum imx_dma_prio {
+ DMA_PRIO_HIGH = 0,
+ DMA_PRIO_MEDIUM = 1,
+ DMA_PRIO_LOW = 2
+};
+
+struct imx_dma_data {
+ int dma_request; /* DMA request line */
+ enum sdma_peripheral_type peripheral_type;
+ int priority;
+};
+
+static inline int imx_dma_is_ipu(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "ipu-core");
+}
+
+static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "imx-sdma");
+}
+
+#endif
diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
new file mode 100644
index 0000000..9be1122
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/sdma.h
@@ -0,0 +1,17 @@
+#ifndef __MACH_MXC_SDMA_H__
+#define __MACH_MXC_SDMA_H__
+
+/**
+ * struct sdma_platform_data - platform specific data for SDMA engine
+ *
+ * @sdma_version The version of this SDMA engine
+ * @cpu_name used to generate the firmware name
+ * @to_version CPU Tape out version
+ */
+struct sdma_platform_data {
+ int sdma_version;
+ char *cpu_name;
+ int to_version;
+};
+
+#endif /* __MACH_MXC_SDMA_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9520cf0..ff68307 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -195,6 +195,14 @@ config PCH_DMA
help
Enable support for the Topcliff PCH DMA engine.
+config IMX_SDMA
+ tristate "i.MX SDMA support"
+ depends on ARCH_MXC
+ select DMA_ENGINE
+ help
+ Support the i.MX SDMA engine. This engine is integrated into
+ Freescale i.MX25/31/35/51 chips.
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 72bd703..14d7a1b 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
obj-$(CONFIG_PCH_DMA) += pch_dma.o
+obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
new file mode 100644
index 0000000..1dd0be1
--- /dev/null
+++ b/drivers/dma/imx-sdma.c
@@ -0,0 +1,1392 @@
+/*
+ * drivers/dma/imx-sdma.c
+ *
+ * This file contains a driver for the Freescale Smart DMA engine
+ *
+ * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
+ *
+ * Based on code from Freescale:
+ *
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * The code contained herein is licensed under the GNU General Public
+ * License. You may obtain a copy of the GNU General Public License
+ * Version 2 or later at the following locations:
+ *
+ * http://www.opensource.org/licenses/gpl-license.html
+ * http://www.gnu.org/copyleft/gpl.html
+ */
+
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/dmaengine.h>
+
+#include <asm/irq.h>
+#include <mach/sdma.h>
+#include <mach/dma.h>
+#include <mach/hardware.h>
+
+/* SDMA registers */
+#define SDMA_H_C0PTR 0x000
+#define SDMA_H_INTR 0x004
+#define SDMA_H_STATSTOP 0x008
+#define SDMA_H_START 0x00c
+#define SDMA_H_EVTOVR 0x010
+#define SDMA_H_DSPOVR 0x014
+#define SDMA_H_HOSTOVR 0x018
+#define SDMA_H_EVTPEND 0x01c
+#define SDMA_H_DSPENBL 0x020
+#define SDMA_H_RESET 0x024
+#define SDMA_H_EVTERR 0x028
+#define SDMA_H_INTRMSK 0x02c
+#define SDMA_H_PSW 0x030
+#define SDMA_H_EVTERRDBG 0x034
+#define SDMA_H_CONFIG 0x038
+#define SDMA_ONCE_ENB 0x040
+#define SDMA_ONCE_DATA 0x044
+#define SDMA_ONCE_INSTR 0x048
+#define SDMA_ONCE_STAT 0x04c
+#define SDMA_ONCE_CMD 0x050
+#define SDMA_EVT_MIRROR 0x054
+#define SDMA_ILLINSTADDR 0x058
+#define SDMA_CHN0ADDR 0x05c
+#define SDMA_ONCE_RTB 0x060
+#define SDMA_XTRIG_CONF1 0x070
+#define SDMA_XTRIG_CONF2 0x074
+#define SDMA_CHNENBL0_V2 0x200
+#define SDMA_CHNENBL0_V1 0x080
+#define SDMA_CHNPRI_0 0x100
+
+/*
+ * Buffer descriptor status values.
+ */
+#define BD_DONE 0x01
+#define BD_WRAP 0x02
+#define BD_CONT 0x04
+#define BD_INTR 0x08
+#define BD_RROR 0x10
+#define BD_LAST 0x20
+#define BD_EXTD 0x80
+
+/*
+ * Data Node descriptor status values.
+ */
+#define DND_END_OF_FRAME 0x80
+#define DND_END_OF_XFER 0x40
+#define DND_DONE 0x20
+#define DND_UNUSED 0x01
+
+/*
+ * IPCV2 descriptor status values.
+ */
+#define BD_IPCV2_END_OF_FRAME 0x40
+
+#define IPCV2_MAX_NODES 50
+/*
+ * Error bit set in the CCB status field by the SDMA,
+ * in setbd routine, in case of a transfer error
+ */
+#define DATA_ERROR 0x10000000
+
+/*
+ * Buffer descriptor commands.
+ */
+#define C0_ADDR 0x01
+#define C0_LOAD 0x02
+#define C0_DUMP 0x03
+#define C0_SETCTX 0x07
+#define C0_GETCTX 0x03
+#define C0_SETDM 0x01
+#define C0_SETPM 0x04
+#define C0_GETDM 0x02
+#define C0_GETPM 0x08
+/*
+ * Change endianness indicator in the BD command field
+ */
+#define CHANGE_ENDIANNESS 0x80
+
+/*
+ * Mode/Count of data node descriptors - IPCv2
+ */
+struct sdma_mode_count {
+ u32 count : 16; /* size of the buffer pointed by this BD */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 command : 8; /* command mostlky used for channel 0 */
+};
+
+/*
+ * Buffer descriptor
+ */
+struct sdma_buffer_descriptor {
+ struct sdma_mode_count mode;
+ dma_addr_t buffer_addr; /* address of the buffer described */
+ dma_addr_t ext_buffer_addr; /* extended buffer address */
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_channel_control - Channel control Block
+ *
+ * @current_bd_ptr current buffer descriptor processed
+ * @base_bd_ptr first element of buffer descriptor array
+ * @unused padding. The SDMA engine expects an array of 128 byte
+ * control blocks
+ */
+struct sdma_channel_control {
+ dma_addr_t current_bd_ptr;
+ dma_addr_t base_bd_ptr;
+ u32 unused[2];
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_state_registers - SDMA context for a channel
+ *
+ * @pc: program counter
+ * @t: test bit: status of arithmetic & test instruction
+ * @rpc: return program counter
+ * @sf: source fault while loading data
+ * @spc: loop start program counter
+ * @df: destination fault while storing data
+ * @epc: loop end program counter
+ * @lm: loop mode
+ */
+struct sdma_state_registers {
+ u32 pc :14;
+ u32 unused1: 1;
+ u32 t : 1;
+ u32 rpc :14;
+ u32 unused0: 1;
+ u32 sf : 1;
+ u32 spc :14;
+ u32 unused2: 1;
+ u32 df : 1;
+ u32 epc :14;
+ u32 lm : 2;
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_context_data - sdma context specific to a channel
+ *
+ * @channel_state: channel state bits
+ * @gReg: general registers
+ * @mda: burst dma destination address register
+ * @msa: burst dma source address register
+ * @ms: burst dma status register
+ * @md: burst dma data register
+ * @pda: peripheral dma destination address register
+ * @psa: peripheral dma source address register
+ * @ps: peripheral dma status register
+ * @pd: peripheral dma data register
+ * @ca: CRC polynomial register
+ * @cs: CRC accumulator register
+ * @dda: dedicated core destination address register
+ * @dsa: dedicated core source address register
+ * @ds: dedicated core status register
+ * @dd: dedicated core data register
+ */
+struct sdma_context_data {
+ struct sdma_state_registers channel_state;
+ u32 gReg[8];
+ u32 mda;
+ u32 msa;
+ u32 ms;
+ u32 md;
+ u32 pda;
+ u32 psa;
+ u32 ps;
+ u32 pd;
+ u32 ca;
+ u32 cs;
+ u32 dda;
+ u32 dsa;
+ u32 ds;
+ u32 dd;
+ u32 scratch0;
+ u32 scratch1;
+ u32 scratch2;
+ u32 scratch3;
+ u32 scratch4;
+ u32 scratch5;
+ u32 scratch6;
+ u32 scratch7;
+} __attribute__ ((packed));
+
+#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
+
+struct sdma_engine;
+
+/**
+ * struct sdma_channel - housekeeping for a SDMA channel
+ *
+ * @sdma pointer to the SDMA engine for this channel
+ * @channel the channel number, matches dmaengine chan_id
+ * @direction transfer type. Needed for setting SDMA script
+ * @peripheral_type Peripheral type. Needed for setting SDMA script
+ * @event_id0 aka dma request line
+ * @event_id1 for channels that use 2 events
+ * @word_size peripheral access size
+ * @buf_tail ID of the buffer that was processed
+ * @done channel completion
+ * @num_bd max NUM_BD. number of descriptors currently handling
+ */
+struct sdma_channel {
+ struct sdma_engine *sdma;
+ unsigned int channel;
+ enum dma_data_direction direction;
+ enum sdma_peripheral_type peripheral_type;
+ unsigned int event_id0;
+ unsigned int event_id1;
+ enum dma_slave_buswidth word_size;
+ unsigned int buf_tail;
+ struct completion done;
+ unsigned int num_bd;
+ struct sdma_buffer_descriptor *bd;
+ dma_addr_t bd_phys;
+ unsigned int pc_from_device, pc_to_device;
+ unsigned long flags;
+ dma_addr_t per_address;
+ u32 event_mask0, event_mask1;
+ u32 watermark_level;
+ u32 shp_addr, per_addr;
+ struct dma_chan chan;
+ spinlock_t lock;
+ struct dma_async_tx_descriptor desc;
+ dma_cookie_t last_completed;
+ enum dma_status status;
+};
+
+#define IMX_DMA_SG_LOOP (1 << 0)
+
+#define MAX_DMA_CHANNELS 32
+#define MXC_SDMA_DEFAULT_PRIORITY 1
+#define MXC_SDMA_MIN_PRIORITY 1
+#define MXC_SDMA_MAX_PRIORITY 7
+
+/**
+ * struct sdma_script_start_addrs - SDMA script start pointers
+ *
+ * start addresses of the different functions in the physical
+ * address space of the SDMA engine.
+ */
+struct sdma_script_start_addrs {
+ u32 ap_2_ap_addr;
+ u32 ap_2_bp_addr;
+ u32 ap_2_ap_fixed_addr;
+ u32 bp_2_ap_addr;
+ u32 loopback_on_dsp_side_addr;
+ u32 mcu_interrupt_only_addr;
+ u32 firi_2_per_addr;
+ u32 firi_2_mcu_addr;
+ u32 per_2_firi_addr;
+ u32 mcu_2_firi_addr;
+ u32 uart_2_per_addr;
+ u32 uart_2_mcu_addr;
+ u32 per_2_app_addr;
+ u32 mcu_2_app_addr;
+ u32 per_2_per_addr;
+ u32 uartsh_2_per_addr;
+ u32 uartsh_2_mcu_addr;
+ u32 per_2_shp_addr;
+ u32 mcu_2_shp_addr;
+ u32 ata_2_mcu_addr;
+ u32 mcu_2_ata_addr;
+ u32 app_2_per_addr;
+ u32 app_2_mcu_addr;
+ u32 shp_2_per_addr;
+ u32 shp_2_mcu_addr;
+ u32 mshc_2_mcu_addr;
+ u32 mcu_2_mshc_addr;
+ u32 spdif_2_mcu_addr;
+ u32 mcu_2_spdif_addr;
+ u32 asrc_2_mcu_addr;
+ u32 ext_mem_2_ipu_addr;
+ u32 descrambler_addr;
+ u32 dptc_dvfs_addr;
+ u32 utra_addr;
+ u32 ram_code_start_addr;
+};
+
+#define SDMA_FIRMWARE_MAGIC 0x414d4453
+
+/**
+ * struct sdma_firmware_header - Layout of the firmware image
+ *
+ * @magic "SDMA"
+ * @version_major increased whenever layout of struct sdma_script_start_addrs
+ * changes.
+ * @version_minor firmware minor version (for binary compatible changes)
+ * @script_addrs_start offset of struct sdma_script_start_addrs in this image
+ * @num_script_addrs Number of script addresses in this image
+ * @ram_code_start offset of SDMA ram image in this firmware image
+ * @ram_code_size size of SDMA ram image
+ * @script_addrs Stores the start address of the SDMA scripts
+ * (in SDMA memory space)
+ */
+struct sdma_firmware_header {
+ u32 magic;
+ u32 version_major;
+ u32 version_minor;
+ u32 script_addrs_start;
+ u32 num_script_addrs;
+ u32 ram_code_start;
+ u32 ram_code_size;
+};
+
+struct sdma_engine {
+ struct device *dev;
+ struct sdma_channel channel[MAX_DMA_CHANNELS];
+ struct sdma_channel_control *channel_control;
+ void __iomem *regs;
+ unsigned int version;
+ unsigned int num_events;
+ struct sdma_context_data *context;
+ dma_addr_t context_phys;
+ struct dma_device dma_device;
+ struct clk *clk;
+ struct sdma_script_start_addrs *script_addrs;
+};
+
+#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
+#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
+#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
+#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
+
+static inline u32 chnenbl_ofs(struct sdma_engine *sdma, unsigned int event)
+{
+ u32 chnenbl0 = (sdma->version == 2 ? SDMA_CHNENBL0_V2 : SDMA_CHNENBL0_V1);
+
+ return chnenbl0 + event * 4;
+}
+
+static int sdma_config_ownership(struct sdma_channel *sdmac,
+ bool event_override, bool mcu_verride, bool dsp_override)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 evt, mcu, dsp;
+
+ if (event_override && mcu_verride && dsp_override)
+ return -EINVAL;
+
+ evt = __raw_readl(sdma->regs + SDMA_H_EVTOVR);
+ mcu = __raw_readl(sdma->regs + SDMA_H_HOSTOVR);
+ dsp = __raw_readl(sdma->regs + SDMA_H_DSPOVR);
+
+ if (dsp_override)
+ dsp &= ~(1 << channel);
+ else
+ dsp |= (1 << channel);
+
+ if (event_override)
+ evt &= ~(1 << channel);
+ else
+ evt |= (1 << channel);
+
+ if (mcu_verride)
+ mcu &= ~(1 << channel);
+ else
+ mcu |= (1 << channel);
+
+ __raw_writel(evt, sdma->regs + SDMA_H_EVTOVR);
+ __raw_writel(mcu, sdma->regs + SDMA_H_HOSTOVR);
+ __raw_writel(dsp, sdma->regs + SDMA_H_DSPOVR);
+
+ return 0;
+}
+
+/*
+ * sdma_run_channel - run a channel and wait till it's done
+ */
+static int sdma_run_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret;
+
+ init_completion(&sdmac->done);
+
+ __raw_writel(1 << channel, sdma->regs + SDMA_H_START);
+
+ ret = wait_for_completion_timeout(&sdmac->done, HZ);
+
+ return ret ? 0 : -ETIMEDOUT;
+}
+
+static int sdma_load_script(struct sdma_engine *sdma, void *buf, int size,
+ u32 address)
+{
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ void *buf_virt;
+ dma_addr_t buf_phys;
+ int ret;
+
+ buf_virt = dma_alloc_coherent(NULL,
+ size,
+ &buf_phys, GFP_KERNEL);
+ if (!buf_virt)
+ return -ENOMEM;
+
+ bd0->mode.command = C0_SETPM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = size / 2;
+ bd0->buffer_addr = buf_phys;
+ bd0->ext_buffer_addr = address;
+
+ memcpy(buf_virt, buf, size);
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ dma_free_coherent(NULL, size, buf_virt, buf_phys);
+
+ return ret;
+}
+
+static void sdma_event_enable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 val;
+ u32 chnenbl = chnenbl_ofs(sdma, event);
+
+ val = __raw_readl(sdma->regs + chnenbl);
+ val |= (1 << channel);
+ __raw_writel(val, sdma->regs + chnenbl);
+}
+
+static void sdma_event_disable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 chnenbl = chnenbl_ofs(sdma, event);
+ u32 val;
+
+ val = __raw_readl(sdma->regs + chnenbl);
+ val &= ~(1 << channel);
+ __raw_writel(val, sdma->regs + chnenbl);
+}
+
+static void sdma_handle_channel_loop(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+
+ /*
+ * loop mode. Iterate over descriptors, re-setup them and
+ * call callback function.
+ */
+ while (1) {
+ bd = &sdmac->bd[sdmac->buf_tail];
+
+ if (bd->mode.status & BD_DONE)
+ break;
+
+ if (bd->mode.status & BD_RROR)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ bd->mode.status |= BD_DONE;
+ sdmac->buf_tail++;
+ sdmac->buf_tail %= sdmac->num_bd;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ }
+}
+
+static void mxc_sdma_handle_channel_normal(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+ int i, error = 0;
+
+ /*
+ * non loop mode. Iterate over all descriptors, collect
+ * errors and call callback function
+ */
+ for (i = 0; i < sdmac->num_bd; i++) {
+ bd = &sdmac->bd[i];
+
+ if (bd->mode.status & (BD_DONE | BD_RROR))
+ error = -EIO;
+ }
+
+ if (error)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ sdmac->last_completed = sdmac->desc.cookie;
+}
+
+static void mxc_sdma_handle_channel(struct sdma_channel *sdmac)
+{
+ complete(&sdmac->done);
+
+ /* not interested in channel 0 interrupts */
+ if (sdmac->channel == 0)
+ return;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP)
+ sdma_handle_channel_loop(sdmac);
+ else
+ mxc_sdma_handle_channel_normal(sdmac);
+}
+
+static irqreturn_t sdma_int_handler(int irq, void *dev_id)
+{
+ struct sdma_engine *sdma = dev_id;
+ u32 stat;
+
+ stat = __raw_readl(sdma->regs + SDMA_H_INTR);
+ __raw_writel(stat, sdma->regs + SDMA_H_INTR);
+
+ while (stat) {
+ int channel = fls(stat) - 1;
+ struct sdma_channel *sdmac = &sdma->channel[channel];
+
+ mxc_sdma_handle_channel(sdmac);
+
+ stat &= ~(1 << channel);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * sets the pc of SDMA script according to the peripheral type
+ */
+static void sdma_get_pc(struct sdma_channel *sdmac,
+ enum sdma_peripheral_type peripheral_type)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int per_2_emi = 0, emi_2_per = 0;
+ /*
+ * These are needed once we start to support transfers between
+ * two peripherals or memory-to-memory transfers
+ */
+ int per_2_per = 0, emi_2_emi = 0;
+
+ sdmac->pc_from_device = 0;
+ sdmac->pc_to_device = 0;
+
+ switch (peripheral_type) {
+ case IMX_DMATYPE_MEMORY:
+ emi_2_emi = sdma->script_addrs->ap_2_ap_addr;
+ break;
+ case IMX_DMATYPE_DSP:
+ emi_2_per = sdma->script_addrs->bp_2_ap_addr;
+ per_2_emi = sdma->script_addrs->ap_2_bp_addr;
+ break;
+ case IMX_DMATYPE_FIRI:
+ per_2_emi = sdma->script_addrs->firi_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_firi_addr;
+ break;
+ case IMX_DMATYPE_UART:
+ per_2_emi = sdma->script_addrs->uart_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_UART_SP:
+ per_2_emi = sdma->script_addrs->uartsh_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ATA:
+ per_2_emi = sdma->script_addrs->ata_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_ata_addr;
+ break;
+ case IMX_DMATYPE_CSPI:
+ case IMX_DMATYPE_EXT:
+ case IMX_DMATYPE_SSI:
+ per_2_emi = sdma->script_addrs->app_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_SSI_SP:
+ case IMX_DMATYPE_MMC:
+ case IMX_DMATYPE_SDHC:
+ case IMX_DMATYPE_CSPI_SP:
+ case IMX_DMATYPE_ESAI:
+ case IMX_DMATYPE_MSHC_SP:
+ per_2_emi = sdma->script_addrs->shp_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ASRC:
+ per_2_emi = sdma->script_addrs->asrc_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->asrc_2_mcu_addr;
+ per_2_per = sdma->script_addrs->per_2_per_addr;
+ break;
+ case IMX_DMATYPE_MSHC:
+ per_2_emi = sdma->script_addrs->mshc_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_mshc_addr;
+ break;
+ case IMX_DMATYPE_CCM:
+ per_2_emi = sdma->script_addrs->dptc_dvfs_addr;
+ break;
+ case IMX_DMATYPE_SPDIF:
+ per_2_emi = sdma->script_addrs->spdif_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_spdif_addr;
+ break;
+ case IMX_DMATYPE_IPU_MEMORY:
+ emi_2_per = sdma->script_addrs->ext_mem_2_ipu_addr;
+ break;
+ default:
+ break;
+ }
+
+ sdmac->pc_from_device = per_2_emi;
+ sdmac->pc_to_device = emi_2_per;
+}
+
+static int sdma_load_context(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int load_address;
+ struct sdma_context_data *context = sdma->context;
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ int ret;
+
+ if (sdmac->direction == DMA_FROM_DEVICE) {
+ load_address = sdmac->pc_from_device;
+ } else {
+ load_address = sdmac->pc_to_device;
+ }
+
+ if (load_address < 0)
+ return load_address;
+
+ dev_dbg(sdma->dev, "load_address = %d\n", load_address);
+ dev_dbg(sdma->dev, "wml = 0x%08x\n", sdmac->watermark_level);
+ dev_dbg(sdma->dev, "shp_addr = 0x%08x\n", sdmac->shp_addr);
+ dev_dbg(sdma->dev, "per_addr = 0x%08x\n", sdmac->per_addr);
+ dev_dbg(sdma->dev, "event_mask0 = 0x%08x\n", sdmac->event_mask0);
+ dev_dbg(sdma->dev, "event_mask1 = 0x%08x\n", sdmac->event_mask1);
+
+ memset(context, 0, sizeof(*context));
+ context->channel_state.pc = load_address;
+
+ /* Send by context the event mask,base address for peripheral
+ * and watermark level
+ */
+ context->gReg[0] = sdmac->event_mask1;
+ context->gReg[1] = sdmac->event_mask0;
+ context->gReg[2] = sdmac->per_addr;
+ context->gReg[6] = sdmac->shp_addr;
+ context->gReg[7] = sdmac->watermark_level;
+
+ bd0->mode.command = C0_SETDM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = sizeof(*context) / 4;
+ bd0->buffer_addr = sdma->context_phys;
+ bd0->ext_buffer_addr = 2048 + (sizeof(*context) / 4) * channel;
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ return ret;
+}
+
+static void sdma_disable_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ __raw_writel(1 << channel, sdma->regs + SDMA_H_STATSTOP);
+ sdmac->status = DMA_ERROR;
+}
+
+static int sdma_config_channel(struct sdma_channel *sdmac)
+{
+ int ret;
+
+ sdma_disable_channel(sdmac);
+
+ sdmac->event_mask0 = 0;
+ sdmac->event_mask1 = 0;
+ sdmac->shp_addr = 0;
+ sdmac->per_addr = 0;
+
+ if (sdmac->event_id0) {
+ if (sdmac->event_id0 > 32)
+ return -EINVAL;
+ sdma_event_enable(sdmac, sdmac->event_id0);
+ }
+
+ switch (sdmac->peripheral_type) {
+ case IMX_DMATYPE_DSP:
+ sdma_config_ownership(sdmac, false, true, true);
+ break;
+ case IMX_DMATYPE_MEMORY:
+ sdma_config_ownership(sdmac, false, true, false);
+ break;
+ default:
+ sdma_config_ownership(sdmac, true, true, false);
+ break;
+ }
+
+ sdma_get_pc(sdmac, sdmac->peripheral_type);
+
+ if ((sdmac->peripheral_type != IMX_DMATYPE_MEMORY) &&
+ (sdmac->peripheral_type != IMX_DMATYPE_DSP)) {
+ /* Handle multiple event channels differently */
+ if (sdmac->event_id1) {
+ sdmac->event_mask1 = 1 << (sdmac->event_id1 % 32);
+ if (sdmac->event_id1 > 31)
+ sdmac->watermark_level |= 1 << 31;
+ sdmac->event_mask0 = 1 << (sdmac->event_id0 % 32);
+ if (sdmac->event_id0 > 31)
+ sdmac->watermark_level |= 1 << 30;
+ } else {
+ sdmac->event_mask0 = 1 << sdmac->event_id0;
+ sdmac->event_mask1 = 1 << (sdmac->event_id0 - 32);
+ }
+ /* Watermark Level */
+ sdmac->watermark_level |= sdmac->watermark_level;
+ /* Address */
+ sdmac->shp_addr = sdmac->per_address;
+ } else {
+ sdmac->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
+ }
+
+ ret = sdma_load_context(sdmac);
+
+ return ret;
+}
+
+static int sdma_set_channel_priority(struct sdma_channel *sdmac,
+ unsigned int priority)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ if (priority < MXC_SDMA_MIN_PRIORITY
+ || priority > MXC_SDMA_MAX_PRIORITY) {
+ return -EINVAL;
+ }
+
+ __raw_writel(priority, sdma->regs + SDMA_CHNPRI_0 + 4 * channel);
+
+ return 0;
+}
+
+static int sdma_request_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret = -EBUSY;
+
+ sdmac->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdmac->bd_phys, GFP_KERNEL);
+ if (!sdmac->bd) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memset(sdmac->bd, 0, PAGE_SIZE);
+
+ sdma->channel_control[channel].base_bd_ptr = sdmac->bd_phys;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ clk_enable(sdma->clk);
+
+ sdma_set_channel_priority(sdmac, MXC_SDMA_DEFAULT_PRIORITY);
+
+ init_completion(&sdmac->done);
+
+ sdmac->buf_tail = 0;
+
+ return 0;
+out:
+
+ return ret;
+}
+
+static void sdma_enable_channel(struct sdma_engine *sdma, int channel)
+{
+ __raw_writel(1 << channel, sdma->regs + SDMA_H_START);
+}
+
+static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
+{
+ dma_cookie_t cookie = sdma->chan.cookie;
+
+ if (++cookie < 0)
+ cookie = 1;
+
+ sdma->chan.cookie = cookie;
+ sdma->desc.cookie = cookie;
+
+ return cookie;
+}
+
+static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sdma_channel, chan);
+}
+
+static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(tx->chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ dma_cookie_t cookie;
+
+ spin_lock_irq(&sdmac->lock);
+
+ cookie = sdma_assign_cookie(sdmac);
+
+ sdma_enable_channel(sdma, tx->chan->chan_id);
+
+ spin_unlock_irq(&sdmac->lock);
+
+ return cookie;
+}
+
+static int sdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct imx_dma_data *data = chan->private;
+ int prio, ret;
+
+ /* No need to execute this for internal channel 0 */
+ if (chan->chan_id == 0)
+ return 0;
+
+ if (!data)
+ return -EINVAL;
+
+ switch (data->priority) {
+ case DMA_PRIO_HIGH:
+ prio = 3;
+ break;
+ case DMA_PRIO_MEDIUM:
+ prio = 2;
+ break;
+ case DMA_PRIO_LOW:
+ default:
+ prio = 1;
+ break;
+ }
+
+ sdmac->peripheral_type = data->peripheral_type;
+ sdmac->event_id0 = data->dma_request;
+ ret = sdma_set_channel_priority(sdmac, prio);
+ if (ret)
+ return ret;
+
+ ret = sdma_request_channel(sdmac);
+ if (ret)
+ return ret;
+
+ dma_async_tx_descriptor_init(&sdmac->desc, chan);
+ sdmac->desc.tx_submit = sdma_tx_submit;
+ /* txd.flags will be overwritten in prep funcs */
+ sdmac->desc.flags = DMA_CTRL_ACK;
+
+ return 0;
+}
+
+static void sdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+
+ sdma_disable_channel(sdmac);
+
+ if (sdmac->event_id0)
+ sdma_event_disable(sdmac, sdmac->event_id0);
+ if (sdmac->event_id1)
+ sdma_event_disable(sdmac, sdmac->event_id1);
+
+ sdmac->event_id0 = 0;
+ sdmac->event_id1 = 0;
+
+ sdma_set_channel_priority(sdmac, 0);
+
+ dma_free_coherent(NULL, PAGE_SIZE, sdmac->bd, sdmac->bd_phys);
+
+ clk_disable(sdma->clk);
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int ret, i, count;
+ int channel = chan->chan_id;
+ struct scatterlist *sg;
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags = 0;
+
+ dev_dbg(sdma->dev, "setting up %d entries for channel %d.\n",
+ sg_len, channel);
+
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (sg_len > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, sg_len, NUM_BD);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = sgl->dma_address;
+
+ count = sg->length;
+
+ if (count > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
+ channel, count, 0xffff);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ bd->mode.count = count;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES) {
+ ret = -EINVAL;
+ goto err_out;
+ }
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP) {
+ param |= BD_INTR;
+ if (i + 1 == sg_len)
+ param |= BD_WRAP;
+ }
+
+ if (i + 1 == sg_len)
+ param |= BD_INTR;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, count, sg->dma_address,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+ }
+
+ sdmac->num_bd = sg_len;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int num_periods = buf_len / period_len;
+ int channel = chan->chan_id;
+ int ret, i = 0, buf = 0;
+
+ dev_dbg(sdma->dev, "%s channel: %d\n", __func__, channel);
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags |= IMX_DMA_SG_LOOP;
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (num_periods > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, num_periods, NUM_BD);
+ goto err_out;
+ }
+
+ if (period_len > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum period size exceeded: %d > %d\n",
+ channel, period_len, 0xffff);
+ goto err_out;
+ }
+
+ while (buf < buf_len) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = dma_addr;
+
+ bd->mode.count = period_len;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES)
+ goto err_out;
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
+ if (i + 1 == num_periods)
+ param |= BD_WRAP;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, period_len, dma_addr,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+
+ dma_addr += period_len;
+ buf += period_len;
+
+ i++;
+ }
+
+ sdmac->num_bd = num_periods;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ sdmac->status = DMA_ERROR;
+ return NULL;
+}
+
+static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct dma_slave_config *dmaengine_cfg = (void *)arg;
+
+ switch (cmd) {
+ case DMA_TERMINATE_ALL:
+ sdma_disable_channel(sdmac);
+ return 0;
+ case DMA_SLAVE_CONFIG:
+ if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
+ sdmac->per_address = dmaengine_cfg->src_addr;
+ sdmac->watermark_level = dmaengine_cfg->src_maxburst;
+ sdmac->word_size = dmaengine_cfg->src_addr_width;
+ } else {
+ sdmac->per_address = dmaengine_cfg->dst_addr;
+ sdmac->watermark_level = dmaengine_cfg->dst_maxburst;
+ sdmac->word_size = dmaengine_cfg->dst_addr_width;
+ }
+ return sdma_config_channel(sdmac);
+ default:
+ return -ENOSYS;
+ }
+
+ return -EINVAL;
+}
+
+static enum dma_status sdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ dma_cookie_t last_used;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+
+ ret = dma_async_is_complete(cookie, sdmac->last_completed, last_used);
+ dma_set_tx_state(txstate, sdmac->last_completed, last_used, 0);
+
+ return ret;
+}
+
+static void sdma_issue_pending(struct dma_chan *chan)
+{
+ /*
+ * Nothing to do. We only have a single descriptor
+ */
+}
+
+static int __init sdma_init(struct sdma_engine *sdma,
+ void *ram_code, int ram_code_size)
+{
+ int i, ret;
+ dma_addr_t ccb_phys;
+
+ switch (sdma->version) {
+ case 1:
+ sdma->num_events = 32;
+ break;
+ case 2:
+ sdma->num_events = 48;
+ break;
+ default:
+ dev_err(sdma->dev, "Unknown version %d. aborting\n", sdma->version);
+ return -ENODEV;
+ }
+
+ clk_enable(sdma->clk);
+
+ /* Be sure SDMA has not started yet */
+ __raw_writel(0, sdma->regs + SDMA_H_C0PTR);
+
+ sdma->channel_control = dma_alloc_coherent(NULL,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
+ sizeof(struct sdma_context_data),
+ &ccb_phys, GFP_KERNEL);
+
+ if (!sdma->channel_control) {
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ sdma->context = (void *)sdma->channel_control +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+ sdma->context_phys = ccb_phys +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+
+ /* Zero-out the CCB structures array just allocated */
+ memset(sdma->channel_control, 0,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
+
+ /* disable all channels */
+ for (i = 0; i < sdma->num_events; i++)
+ __raw_writel(0, sdma->regs + chnenbl_ofs(sdma, i));
+
+ /* All channels have priority 0 */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++)
+ __raw_writel(0, sdma->regs + SDMA_CHNPRI_0 + i * 4);
+
+ ret = sdma_request_channel(&sdma->channel[0]);
+ if (ret)
+ goto err_dma_alloc;
+
+ sdma_config_ownership(&sdma->channel[0], false, true, false);
+
+ /* Set Command Channel (Channel Zero) */
+ __raw_writel(0x4050, sdma->regs + SDMA_CHN0ADDR);
+
+ /* Set bits of CONFIG register but with static context switching */
+ /* FIXME: Check whether to set ACR bit depending on clock ratios */
+ __raw_writel(0, sdma->regs + SDMA_H_CONFIG);
+
+ __raw_writel(ccb_phys, sdma->regs + SDMA_H_C0PTR);
+
+ /* download the RAM image for SDMA */
+ sdma_load_script(sdma, ram_code,
+ ram_code_size,
+ sdma->script_addrs->ram_code_start_addr);
+
+ /* Set bits of CONFIG register with given context switching mode */
+ __raw_writel(SDMA_H_CONFIG_CSM, sdma->regs + SDMA_H_CONFIG);
+
+ /* Initializes channel's priorities */
+ sdma_set_channel_priority(&sdma->channel[0], 7);
+
+ clk_disable(sdma->clk);
+
+ return 0;
+
+err_dma_alloc:
+ clk_disable(sdma->clk);
+ dev_err(sdma->dev, "initialisation failed with %d\n", ret);
+ return ret;
+}
+
+static int __init sdma_probe(struct platform_device *pdev)
+{
+ int ret;
+ const struct firmware *fw;
+ const struct sdma_firmware_header *header;
+ const struct sdma_script_start_addrs *addr;
+ int irq;
+ unsigned short *ram_code;
+ struct resource *iores;
+ struct sdma_platform_data *pdata = pdev->dev.platform_data;
+ char *fwname;
+ int i;
+ dma_cap_mask_t mask;
+ struct sdma_engine *sdma;
+
+ sdma = kzalloc(sizeof(*sdma), GFP_KERNEL);
+ if (!sdma)
+ return -ENOMEM;
+
+ sdma->dev = &pdev->dev;
+
+ iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ irq = platform_get_irq(pdev, 0);
+ if (!iores || irq < 0 || !pdata) {
+ ret = -EINVAL;
+ goto err_irq;
+ }
+
+ if (!request_mem_region(iores->start, resource_size(iores), pdev->name)) {
+ ret = -EBUSY;
+ goto err_request_region;
+ }
+
+ sdma->clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(sdma->clk)) {
+ ret = PTR_ERR(sdma->clk);
+ goto err_clk;
+ }
+
+ sdma->regs = ioremap(iores->start, resource_size(iores));
+ if (!sdma->regs) {
+ ret = -ENOMEM;
+ goto err_ioremap;
+ }
+
+ ret = request_irq(irq, sdma_int_handler, 0, "sdma", sdma);
+ if (ret)
+ goto err_request_irq;
+
+ fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin",
+ pdata->cpu_name, pdata->to_version);
+ if (!fwname) {
+ ret = -ENOMEM;
+ goto err_cputype;
+ }
+
+ ret = request_firmware(&fw, fwname, &pdev->dev);
+ if (ret) {
+ dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
+ fwname, ret);
+ kfree(fwname);
+ goto err_cputype;
+ }
+ kfree(fwname);
+
+ if (fw->size < sizeof(*header))
+ goto err_firmware;
+
+ header = (struct sdma_firmware_header *)fw->data;
+
+ if (header->magic != SDMA_FIRMWARE_MAGIC)
+ goto err_firmware;
+ if (header->ram_code_start + header->ram_code_size > fw->size)
+ goto err_firmware;
+
+ addr = (void *)header + header->script_addrs_start;
+ ram_code = (void *)header + header->ram_code_start;
+ sdma->script_addrs = kmalloc(sizeof(*addr), GFP_KERNEL);
+ if (!sdma->script_addrs)
+ goto err_firmware;
+ memcpy(sdma->script_addrs, addr, sizeof(*addr));
+
+ sdma->version = pdata->sdma_version;
+
+ INIT_LIST_HEAD(&sdma->dma_device.channels);
+ /* Initialize channel parameters */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++) {
+ struct sdma_channel *sdmac = &sdma->channel[i];
+
+ sdmac->sdma = sdma;
+ spin_lock_init(&sdmac->lock);
+
+ dma_cap_set(DMA_SLAVE, sdma->dma_device.cap_mask);
+ dma_cap_set(DMA_CYCLIC, sdma->dma_device.cap_mask);
+
+ sdmac->chan.device = &sdma->dma_device;
+ sdmac->chan.chan_id = i;
+ sdmac->channel = i;
+
+ /* Add the channel to the DMAC list */
+ list_add_tail(&sdmac->chan.device_node, &sdma->dma_device.channels);
+ }
+
+ ret = sdma_init(sdma, ram_code, header->ram_code_size);
+ if (ret)
+ goto err_init;
+
+ sdma->dma_device.dev = &pdev->dev;
+
+ sdma->dma_device.device_alloc_chan_resources = sdma_alloc_chan_resources;
+ sdma->dma_device.device_free_chan_resources = sdma_free_chan_resources;
+ sdma->dma_device.device_tx_status = sdma_tx_status;
+ sdma->dma_device.device_prep_slave_sg = sdma_prep_slave_sg;
+ sdma->dma_device.device_prep_dma_cyclic = sdma_prep_dma_cyclic;
+ sdma->dma_device.device_control = sdma_control;
+ sdma->dma_device.device_issue_pending = sdma_issue_pending;
+
+ ret = dma_async_device_register(&sdma->dma_device);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register\n");
+ goto err_init;
+ }
+
+ dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
+ header->version_major,
+ header->version_minor);
+
+ /* request channel 0. This is an internal control channel
+ * to the SDMA engine and not available to clients.
+ */
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_request_channel(mask, NULL, NULL);
+
+ release_firmware(fw);
+
+ return 0;
+
+err_init:
+ kfree(sdma->script_addrs);
+err_firmware:
+ release_firmware(fw);
+err_cputype:
+ free_irq(irq, sdma);
+err_request_irq:
+ iounmap(sdma->regs);
+err_ioremap:
+ clk_put(sdma->clk);
+err_clk:
+ release_mem_region(iores->start, resource_size(iores));
+err_request_region:
+err_irq:
+ kfree(sdma);
+ return 0;
+}
+
+static int __exit sdma_remove(struct platform_device *pdev)
+{
+ return -EBUSY;
+}
+
+static struct platform_driver sdma_driver = {
+ .driver = {
+ .name = "imx-sdma",
+ },
+ .remove = __exit_p(sdma_remove),
+};
+
+static int __init sdma_module_init(void)
+{
+ return platform_driver_probe(&sdma_driver, sdma_probe);
+}
+subsys_initcall(sdma_module_init);
+
+MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("i.MX SDMA driver");
+MODULE_LICENSE("GPL");
--
1.7.1
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply related [flat|nested] 78+ messages in thread
* [PATCH 3/3 v3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-27 12:22 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-27 12:22 UTC (permalink / raw)
To: linux-arm-kernel
This patch adds support for the Freescale i.MX SDMA engine.
The SDMA engine is a scatter/gather DMA engine which is implemented
as a seperate coprocessor. SDMA needs its own firmware which is
requested using the standard request_firmware mechanism. The firmware
has different entry points for each peripheral type, so drivers
have to pass the peripheral type to the DMA engine which in turn
picks the correct firmware entry point from a table contained in
the firmware image itself.
The original Freescale code also supports support for transfering
data to the internal SRAM which needs different entry points to
the firmware. Support for this is currently not implemented. Also,
support for the ASRC (asymmetric sample rate converter) is skipped.
I took a very simple approach to implement dmaengine support. Only
a single descriptor is statically assigned to a each channel. This
means that transfers can't be queued up but only a single transfer
is in progress. This simplifies implementation a lot and is sufficient
for the usual device/memory transfers.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Reviewed-by: Linus Walleij <linus.ml.walleij@gmail.com>
---
changes since v2:
- use __raw_readl/__raw_writel
- use __init/__exit instead of __devinit/__devexit
- use request_mem_region.
arch/arm/mach-imx/include/mach/dma-v1.h | 8 +-
arch/arm/mach-mx3/Kconfig | 2 +
arch/arm/plat-mxc/Kconfig | 10 +
arch/arm/plat-mxc/include/mach/dma.h | 64 ++
arch/arm/plat-mxc/include/mach/sdma.h | 17 +
drivers/dma/Kconfig | 8 +
drivers/dma/Makefile | 1 +
drivers/dma/imx-sdma.c | 1392 +++++++++++++++++++++++++++++++
8 files changed, 1496 insertions(+), 6 deletions(-)
create mode 100644 arch/arm/plat-mxc/include/mach/dma.h
create mode 100644 arch/arm/plat-mxc/include/mach/sdma.h
create mode 100644 drivers/dma/imx-sdma.c
diff --git a/arch/arm/mach-imx/include/mach/dma-v1.h b/arch/arm/mach-imx/include/mach/dma-v1.h
index 287431c..ac6fd71 100644
--- a/arch/arm/mach-imx/include/mach/dma-v1.h
+++ b/arch/arm/mach-imx/include/mach/dma-v1.h
@@ -27,6 +27,8 @@
#define imx_has_dma_v1() (cpu_is_mx1() || cpu_is_mx21() || cpu_is_mx27())
+#include <mach/dma.h>
+
#define IMX_DMA_CHANNELS 16
#define DMA_MODE_READ 0
@@ -96,12 +98,6 @@ int imx_dma_request(int channel, const char *name);
void imx_dma_free(int channel);
-enum imx_dma_prio {
- DMA_PRIO_HIGH = 0,
- DMA_PRIO_MEDIUM = 1,
- DMA_PRIO_LOW = 2
-};
-
int imx_dma_request_by_prio(const char *name, enum imx_dma_prio prio);
#endif /* __MACH_DMA_V1_H__ */
diff --git a/arch/arm/mach-mx3/Kconfig b/arch/arm/mach-mx3/Kconfig
index 85beece..301375c 100644
--- a/arch/arm/mach-mx3/Kconfig
+++ b/arch/arm/mach-mx3/Kconfig
@@ -3,12 +3,14 @@ if ARCH_MX3
config ARCH_MX31
select ARCH_HAS_RNGA
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
bool
config ARCH_MX35
bool
select ARCH_MXC_IOMUX_V3
select ARCH_MXC_AUDMUX_V2
+ select IMX_HAVE_SDMA
comment "MX3 platforms:"
diff --git a/arch/arm/plat-mxc/Kconfig b/arch/arm/plat-mxc/Kconfig
index 0527e65..6741625 100644
--- a/arch/arm/plat-mxc/Kconfig
+++ b/arch/arm/plat-mxc/Kconfig
@@ -109,4 +109,14 @@ config ARCH_MXC_AUDMUX_V1
config ARCH_MXC_AUDMUX_V2
bool
+config IMX_HAVE_SDMA
+ bool
+
+config IMX_SDMA
+ depends on IMX_HAVE_SDMA
+ tristate "Enable SDMA support"
+ help
+ Include support for the SDMA engine. The SDMA engine needs additional
+ firmware support. SDMA can be compiled as a module to support loading
+ the firmware when a rootfs is present.
endif
diff --git a/arch/arm/plat-mxc/include/mach/dma.h b/arch/arm/plat-mxc/include/mach/dma.h
new file mode 100644
index 0000000..69d181f
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/dma.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __ASM_ARCH_MXC_DMA_H__
+#define __ASM_ARCH_MXC_DMA_H__
+
+#include <linux/scatterlist.h>
+
+/*
+ * This enumerates peripheral types. Used for SDMA.
+ */
+enum sdma_peripheral_type {
+ IMX_DMATYPE_SSI, /* MCU domain SSI */
+ IMX_DMATYPE_SSI_SP, /* Shared SSI */
+ IMX_DMATYPE_MMC, /* MMC */
+ IMX_DMATYPE_SDHC, /* SDHC */
+ IMX_DMATYPE_UART, /* MCU domain UART */
+ IMX_DMATYPE_UART_SP, /* Shared UART */
+ IMX_DMATYPE_FIRI, /* FIRI */
+ IMX_DMATYPE_CSPI, /* MCU domain CSPI */
+ IMX_DMATYPE_CSPI_SP, /* Shared CSPI */
+ IMX_DMATYPE_SIM, /* SIM */
+ IMX_DMATYPE_ATA, /* ATA */
+ IMX_DMATYPE_CCM, /* CCM */
+ IMX_DMATYPE_EXT, /* External peripheral */
+ IMX_DMATYPE_MSHC, /* Memory Stick Host Controller */
+ IMX_DMATYPE_MSHC_SP, /* Shared Memory Stick Host Controller */
+ IMX_DMATYPE_DSP, /* DSP */
+ IMX_DMATYPE_MEMORY, /* Memory */
+ IMX_DMATYPE_FIFO_MEMORY,/* FIFO type Memory */
+ IMX_DMATYPE_SPDIF, /* SPDIF */
+ IMX_DMATYPE_IPU_MEMORY, /* IPU Memory */
+ IMX_DMATYPE_ASRC, /* ASRC */
+ IMX_DMATYPE_ESAI, /* ESAI */
+};
+
+enum imx_dma_prio {
+ DMA_PRIO_HIGH = 0,
+ DMA_PRIO_MEDIUM = 1,
+ DMA_PRIO_LOW = 2
+};
+
+struct imx_dma_data {
+ int dma_request; /* DMA request line */
+ enum sdma_peripheral_type peripheral_type;
+ int priority;
+};
+
+static inline int imx_dma_is_ipu(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "ipu-core");
+}
+
+static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
+{
+ return !strcmp(dev_name(chan->device->dev), "imx-sdma");
+}
+
+#endif
diff --git a/arch/arm/plat-mxc/include/mach/sdma.h b/arch/arm/plat-mxc/include/mach/sdma.h
new file mode 100644
index 0000000..9be1122
--- /dev/null
+++ b/arch/arm/plat-mxc/include/mach/sdma.h
@@ -0,0 +1,17 @@
+#ifndef __MACH_MXC_SDMA_H__
+#define __MACH_MXC_SDMA_H__
+
+/**
+ * struct sdma_platform_data - platform specific data for SDMA engine
+ *
+ * @sdma_version The version of this SDMA engine
+ * @cpu_name used to generate the firmware name
+ * @to_version CPU Tape out version
+ */
+struct sdma_platform_data {
+ int sdma_version;
+ char *cpu_name;
+ int to_version;
+};
+
+#endif /* __MACH_MXC_SDMA_H__ */
diff --git a/drivers/dma/Kconfig b/drivers/dma/Kconfig
index 9520cf0..ff68307 100644
--- a/drivers/dma/Kconfig
+++ b/drivers/dma/Kconfig
@@ -195,6 +195,14 @@ config PCH_DMA
help
Enable support for the Topcliff PCH DMA engine.
+config IMX_SDMA
+ tristate "i.MX SDMA support"
+ depends on ARCH_MXC
+ select DMA_ENGINE
+ help
+ Support the i.MX SDMA engine. This engine is integrated into
+ Freescale i.MX25/31/35/51 chips.
+
config DMA_ENGINE
bool
diff --git a/drivers/dma/Makefile b/drivers/dma/Makefile
index 72bd703..14d7a1b 100644
--- a/drivers/dma/Makefile
+++ b/drivers/dma/Makefile
@@ -25,3 +25,4 @@ obj-$(CONFIG_TIMB_DMA) += timb_dma.o
obj-$(CONFIG_STE_DMA40) += ste_dma40.o ste_dma40_ll.o
obj-$(CONFIG_PL330_DMA) += pl330.o
obj-$(CONFIG_PCH_DMA) += pch_dma.o
+obj-$(CONFIG_IMX_SDMA) += imx-sdma.o
diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c
new file mode 100644
index 0000000..1dd0be1
--- /dev/null
+++ b/drivers/dma/imx-sdma.c
@@ -0,0 +1,1392 @@
+/*
+ * drivers/dma/imx-sdma.c
+ *
+ * This file contains a driver for the Freescale Smart DMA engine
+ *
+ * Copyright 2010 Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>
+ *
+ * Based on code from Freescale:
+ *
+ * Copyright 2004-2009 Freescale Semiconductor, Inc. All Rights Reserved.
+ *
+ * The code contained herein is licensed under the GNU General Public
+ * License. You may obtain a copy of the GNU General Public License
+ * Version 2 or later at the following locations:
+ *
+ * http://www.opensource.org/licenses/gpl-license.html
+ * http://www.gnu.org/copyleft/gpl.html
+ */
+
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/interrupt.h>
+#include <linux/clk.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <linux/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/platform_device.h>
+#include <linux/dmaengine.h>
+
+#include <asm/irq.h>
+#include <mach/sdma.h>
+#include <mach/dma.h>
+#include <mach/hardware.h>
+
+/* SDMA registers */
+#define SDMA_H_C0PTR 0x000
+#define SDMA_H_INTR 0x004
+#define SDMA_H_STATSTOP 0x008
+#define SDMA_H_START 0x00c
+#define SDMA_H_EVTOVR 0x010
+#define SDMA_H_DSPOVR 0x014
+#define SDMA_H_HOSTOVR 0x018
+#define SDMA_H_EVTPEND 0x01c
+#define SDMA_H_DSPENBL 0x020
+#define SDMA_H_RESET 0x024
+#define SDMA_H_EVTERR 0x028
+#define SDMA_H_INTRMSK 0x02c
+#define SDMA_H_PSW 0x030
+#define SDMA_H_EVTERRDBG 0x034
+#define SDMA_H_CONFIG 0x038
+#define SDMA_ONCE_ENB 0x040
+#define SDMA_ONCE_DATA 0x044
+#define SDMA_ONCE_INSTR 0x048
+#define SDMA_ONCE_STAT 0x04c
+#define SDMA_ONCE_CMD 0x050
+#define SDMA_EVT_MIRROR 0x054
+#define SDMA_ILLINSTADDR 0x058
+#define SDMA_CHN0ADDR 0x05c
+#define SDMA_ONCE_RTB 0x060
+#define SDMA_XTRIG_CONF1 0x070
+#define SDMA_XTRIG_CONF2 0x074
+#define SDMA_CHNENBL0_V2 0x200
+#define SDMA_CHNENBL0_V1 0x080
+#define SDMA_CHNPRI_0 0x100
+
+/*
+ * Buffer descriptor status values.
+ */
+#define BD_DONE 0x01
+#define BD_WRAP 0x02
+#define BD_CONT 0x04
+#define BD_INTR 0x08
+#define BD_RROR 0x10
+#define BD_LAST 0x20
+#define BD_EXTD 0x80
+
+/*
+ * Data Node descriptor status values.
+ */
+#define DND_END_OF_FRAME 0x80
+#define DND_END_OF_XFER 0x40
+#define DND_DONE 0x20
+#define DND_UNUSED 0x01
+
+/*
+ * IPCV2 descriptor status values.
+ */
+#define BD_IPCV2_END_OF_FRAME 0x40
+
+#define IPCV2_MAX_NODES 50
+/*
+ * Error bit set in the CCB status field by the SDMA,
+ * in setbd routine, in case of a transfer error
+ */
+#define DATA_ERROR 0x10000000
+
+/*
+ * Buffer descriptor commands.
+ */
+#define C0_ADDR 0x01
+#define C0_LOAD 0x02
+#define C0_DUMP 0x03
+#define C0_SETCTX 0x07
+#define C0_GETCTX 0x03
+#define C0_SETDM 0x01
+#define C0_SETPM 0x04
+#define C0_GETDM 0x02
+#define C0_GETPM 0x08
+/*
+ * Change endianness indicator in the BD command field
+ */
+#define CHANGE_ENDIANNESS 0x80
+
+/*
+ * Mode/Count of data node descriptors - IPCv2
+ */
+struct sdma_mode_count {
+ u32 count : 16; /* size of the buffer pointed by this BD */
+ u32 status : 8; /* E,R,I,C,W,D status bits stored here */
+ u32 command : 8; /* command mostlky used for channel 0 */
+};
+
+/*
+ * Buffer descriptor
+ */
+struct sdma_buffer_descriptor {
+ struct sdma_mode_count mode;
+ dma_addr_t buffer_addr; /* address of the buffer described */
+ dma_addr_t ext_buffer_addr; /* extended buffer address */
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_channel_control - Channel control Block
+ *
+ * @current_bd_ptr current buffer descriptor processed
+ * @base_bd_ptr first element of buffer descriptor array
+ * @unused padding. The SDMA engine expects an array of 128 byte
+ * control blocks
+ */
+struct sdma_channel_control {
+ dma_addr_t current_bd_ptr;
+ dma_addr_t base_bd_ptr;
+ u32 unused[2];
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_state_registers - SDMA context for a channel
+ *
+ * @pc: program counter
+ * @t: test bit: status of arithmetic & test instruction
+ * @rpc: return program counter
+ * @sf: source fault while loading data
+ * @spc: loop start program counter
+ * @df: destination fault while storing data
+ * @epc: loop end program counter
+ * @lm: loop mode
+ */
+struct sdma_state_registers {
+ u32 pc :14;
+ u32 unused1: 1;
+ u32 t : 1;
+ u32 rpc :14;
+ u32 unused0: 1;
+ u32 sf : 1;
+ u32 spc :14;
+ u32 unused2: 1;
+ u32 df : 1;
+ u32 epc :14;
+ u32 lm : 2;
+} __attribute__ ((packed));
+
+/**
+ * struct sdma_context_data - sdma context specific to a channel
+ *
+ * @channel_state: channel state bits
+ * @gReg: general registers
+ * @mda: burst dma destination address register
+ * @msa: burst dma source address register
+ * @ms: burst dma status register
+ * @md: burst dma data register
+ * @pda: peripheral dma destination address register
+ * @psa: peripheral dma source address register
+ * @ps: peripheral dma status register
+ * @pd: peripheral dma data register
+ * @ca: CRC polynomial register
+ * @cs: CRC accumulator register
+ * @dda: dedicated core destination address register
+ * @dsa: dedicated core source address register
+ * @ds: dedicated core status register
+ * @dd: dedicated core data register
+ */
+struct sdma_context_data {
+ struct sdma_state_registers channel_state;
+ u32 gReg[8];
+ u32 mda;
+ u32 msa;
+ u32 ms;
+ u32 md;
+ u32 pda;
+ u32 psa;
+ u32 ps;
+ u32 pd;
+ u32 ca;
+ u32 cs;
+ u32 dda;
+ u32 dsa;
+ u32 ds;
+ u32 dd;
+ u32 scratch0;
+ u32 scratch1;
+ u32 scratch2;
+ u32 scratch3;
+ u32 scratch4;
+ u32 scratch5;
+ u32 scratch6;
+ u32 scratch7;
+} __attribute__ ((packed));
+
+#define NUM_BD (int)(PAGE_SIZE / sizeof(struct sdma_buffer_descriptor))
+
+struct sdma_engine;
+
+/**
+ * struct sdma_channel - housekeeping for a SDMA channel
+ *
+ * @sdma pointer to the SDMA engine for this channel
+ * @channel the channel number, matches dmaengine chan_id
+ * @direction transfer type. Needed for setting SDMA script
+ * @peripheral_type Peripheral type. Needed for setting SDMA script
+ * @event_id0 aka dma request line
+ * @event_id1 for channels that use 2 events
+ * @word_size peripheral access size
+ * @buf_tail ID of the buffer that was processed
+ * @done channel completion
+ * @num_bd max NUM_BD. number of descriptors currently handling
+ */
+struct sdma_channel {
+ struct sdma_engine *sdma;
+ unsigned int channel;
+ enum dma_data_direction direction;
+ enum sdma_peripheral_type peripheral_type;
+ unsigned int event_id0;
+ unsigned int event_id1;
+ enum dma_slave_buswidth word_size;
+ unsigned int buf_tail;
+ struct completion done;
+ unsigned int num_bd;
+ struct sdma_buffer_descriptor *bd;
+ dma_addr_t bd_phys;
+ unsigned int pc_from_device, pc_to_device;
+ unsigned long flags;
+ dma_addr_t per_address;
+ u32 event_mask0, event_mask1;
+ u32 watermark_level;
+ u32 shp_addr, per_addr;
+ struct dma_chan chan;
+ spinlock_t lock;
+ struct dma_async_tx_descriptor desc;
+ dma_cookie_t last_completed;
+ enum dma_status status;
+};
+
+#define IMX_DMA_SG_LOOP (1 << 0)
+
+#define MAX_DMA_CHANNELS 32
+#define MXC_SDMA_DEFAULT_PRIORITY 1
+#define MXC_SDMA_MIN_PRIORITY 1
+#define MXC_SDMA_MAX_PRIORITY 7
+
+/**
+ * struct sdma_script_start_addrs - SDMA script start pointers
+ *
+ * start addresses of the different functions in the physical
+ * address space of the SDMA engine.
+ */
+struct sdma_script_start_addrs {
+ u32 ap_2_ap_addr;
+ u32 ap_2_bp_addr;
+ u32 ap_2_ap_fixed_addr;
+ u32 bp_2_ap_addr;
+ u32 loopback_on_dsp_side_addr;
+ u32 mcu_interrupt_only_addr;
+ u32 firi_2_per_addr;
+ u32 firi_2_mcu_addr;
+ u32 per_2_firi_addr;
+ u32 mcu_2_firi_addr;
+ u32 uart_2_per_addr;
+ u32 uart_2_mcu_addr;
+ u32 per_2_app_addr;
+ u32 mcu_2_app_addr;
+ u32 per_2_per_addr;
+ u32 uartsh_2_per_addr;
+ u32 uartsh_2_mcu_addr;
+ u32 per_2_shp_addr;
+ u32 mcu_2_shp_addr;
+ u32 ata_2_mcu_addr;
+ u32 mcu_2_ata_addr;
+ u32 app_2_per_addr;
+ u32 app_2_mcu_addr;
+ u32 shp_2_per_addr;
+ u32 shp_2_mcu_addr;
+ u32 mshc_2_mcu_addr;
+ u32 mcu_2_mshc_addr;
+ u32 spdif_2_mcu_addr;
+ u32 mcu_2_spdif_addr;
+ u32 asrc_2_mcu_addr;
+ u32 ext_mem_2_ipu_addr;
+ u32 descrambler_addr;
+ u32 dptc_dvfs_addr;
+ u32 utra_addr;
+ u32 ram_code_start_addr;
+};
+
+#define SDMA_FIRMWARE_MAGIC 0x414d4453
+
+/**
+ * struct sdma_firmware_header - Layout of the firmware image
+ *
+ * @magic "SDMA"
+ * @version_major increased whenever layout of struct sdma_script_start_addrs
+ * changes.
+ * @version_minor firmware minor version (for binary compatible changes)
+ * @script_addrs_start offset of struct sdma_script_start_addrs in this image
+ * @num_script_addrs Number of script addresses in this image
+ * @ram_code_start offset of SDMA ram image in this firmware image
+ * @ram_code_size size of SDMA ram image
+ * @script_addrs Stores the start address of the SDMA scripts
+ * (in SDMA memory space)
+ */
+struct sdma_firmware_header {
+ u32 magic;
+ u32 version_major;
+ u32 version_minor;
+ u32 script_addrs_start;
+ u32 num_script_addrs;
+ u32 ram_code_start;
+ u32 ram_code_size;
+};
+
+struct sdma_engine {
+ struct device *dev;
+ struct sdma_channel channel[MAX_DMA_CHANNELS];
+ struct sdma_channel_control *channel_control;
+ void __iomem *regs;
+ unsigned int version;
+ unsigned int num_events;
+ struct sdma_context_data *context;
+ dma_addr_t context_phys;
+ struct dma_device dma_device;
+ struct clk *clk;
+ struct sdma_script_start_addrs *script_addrs;
+};
+
+#define SDMA_H_CONFIG_DSPDMA (1 << 12) /* indicates if the DSPDMA is used */
+#define SDMA_H_CONFIG_RTD_PINS (1 << 11) /* indicates if Real-Time Debug pins are enabled */
+#define SDMA_H_CONFIG_ACR (1 << 4) /* indicates if AHB freq /core freq = 2 or 1 */
+#define SDMA_H_CONFIG_CSM (3) /* indicates which context switch mode is selected*/
+
+static inline u32 chnenbl_ofs(struct sdma_engine *sdma, unsigned int event)
+{
+ u32 chnenbl0 = (sdma->version == 2 ? SDMA_CHNENBL0_V2 : SDMA_CHNENBL0_V1);
+
+ return chnenbl0 + event * 4;
+}
+
+static int sdma_config_ownership(struct sdma_channel *sdmac,
+ bool event_override, bool mcu_verride, bool dsp_override)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 evt, mcu, dsp;
+
+ if (event_override && mcu_verride && dsp_override)
+ return -EINVAL;
+
+ evt = __raw_readl(sdma->regs + SDMA_H_EVTOVR);
+ mcu = __raw_readl(sdma->regs + SDMA_H_HOSTOVR);
+ dsp = __raw_readl(sdma->regs + SDMA_H_DSPOVR);
+
+ if (dsp_override)
+ dsp &= ~(1 << channel);
+ else
+ dsp |= (1 << channel);
+
+ if (event_override)
+ evt &= ~(1 << channel);
+ else
+ evt |= (1 << channel);
+
+ if (mcu_verride)
+ mcu &= ~(1 << channel);
+ else
+ mcu |= (1 << channel);
+
+ __raw_writel(evt, sdma->regs + SDMA_H_EVTOVR);
+ __raw_writel(mcu, sdma->regs + SDMA_H_HOSTOVR);
+ __raw_writel(dsp, sdma->regs + SDMA_H_DSPOVR);
+
+ return 0;
+}
+
+/*
+ * sdma_run_channel - run a channel and wait till it's done
+ */
+static int sdma_run_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret;
+
+ init_completion(&sdmac->done);
+
+ __raw_writel(1 << channel, sdma->regs + SDMA_H_START);
+
+ ret = wait_for_completion_timeout(&sdmac->done, HZ);
+
+ return ret ? 0 : -ETIMEDOUT;
+}
+
+static int sdma_load_script(struct sdma_engine *sdma, void *buf, int size,
+ u32 address)
+{
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ void *buf_virt;
+ dma_addr_t buf_phys;
+ int ret;
+
+ buf_virt = dma_alloc_coherent(NULL,
+ size,
+ &buf_phys, GFP_KERNEL);
+ if (!buf_virt)
+ return -ENOMEM;
+
+ bd0->mode.command = C0_SETPM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = size / 2;
+ bd0->buffer_addr = buf_phys;
+ bd0->ext_buffer_addr = address;
+
+ memcpy(buf_virt, buf, size);
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ dma_free_coherent(NULL, size, buf_virt, buf_phys);
+
+ return ret;
+}
+
+static void sdma_event_enable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 val;
+ u32 chnenbl = chnenbl_ofs(sdma, event);
+
+ val = __raw_readl(sdma->regs + chnenbl);
+ val |= (1 << channel);
+ __raw_writel(val, sdma->regs + chnenbl);
+}
+
+static void sdma_event_disable(struct sdma_channel *sdmac, unsigned int event)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ u32 chnenbl = chnenbl_ofs(sdma, event);
+ u32 val;
+
+ val = __raw_readl(sdma->regs + chnenbl);
+ val &= ~(1 << channel);
+ __raw_writel(val, sdma->regs + chnenbl);
+}
+
+static void sdma_handle_channel_loop(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+
+ /*
+ * loop mode. Iterate over descriptors, re-setup them and
+ * call callback function.
+ */
+ while (1) {
+ bd = &sdmac->bd[sdmac->buf_tail];
+
+ if (bd->mode.status & BD_DONE)
+ break;
+
+ if (bd->mode.status & BD_RROR)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ bd->mode.status |= BD_DONE;
+ sdmac->buf_tail++;
+ sdmac->buf_tail %= sdmac->num_bd;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ }
+}
+
+static void mxc_sdma_handle_channel_normal(struct sdma_channel *sdmac)
+{
+ struct sdma_buffer_descriptor *bd;
+ int i, error = 0;
+
+ /*
+ * non loop mode. Iterate over all descriptors, collect
+ * errors and call callback function
+ */
+ for (i = 0; i < sdmac->num_bd; i++) {
+ bd = &sdmac->bd[i];
+
+ if (bd->mode.status & (BD_DONE | BD_RROR))
+ error = -EIO;
+ }
+
+ if (error)
+ sdmac->status = DMA_ERROR;
+ else
+ sdmac->status = DMA_SUCCESS;
+
+ if (sdmac->desc.callback)
+ sdmac->desc.callback(sdmac->desc.callback_param);
+ sdmac->last_completed = sdmac->desc.cookie;
+}
+
+static void mxc_sdma_handle_channel(struct sdma_channel *sdmac)
+{
+ complete(&sdmac->done);
+
+ /* not interested in channel 0 interrupts */
+ if (sdmac->channel == 0)
+ return;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP)
+ sdma_handle_channel_loop(sdmac);
+ else
+ mxc_sdma_handle_channel_normal(sdmac);
+}
+
+static irqreturn_t sdma_int_handler(int irq, void *dev_id)
+{
+ struct sdma_engine *sdma = dev_id;
+ u32 stat;
+
+ stat = __raw_readl(sdma->regs + SDMA_H_INTR);
+ __raw_writel(stat, sdma->regs + SDMA_H_INTR);
+
+ while (stat) {
+ int channel = fls(stat) - 1;
+ struct sdma_channel *sdmac = &sdma->channel[channel];
+
+ mxc_sdma_handle_channel(sdmac);
+
+ stat &= ~(1 << channel);
+ }
+
+ return IRQ_HANDLED;
+}
+
+/*
+ * sets the pc of SDMA script according to the peripheral type
+ */
+static void sdma_get_pc(struct sdma_channel *sdmac,
+ enum sdma_peripheral_type peripheral_type)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int per_2_emi = 0, emi_2_per = 0;
+ /*
+ * These are needed once we start to support transfers between
+ * two peripherals or memory-to-memory transfers
+ */
+ int per_2_per = 0, emi_2_emi = 0;
+
+ sdmac->pc_from_device = 0;
+ sdmac->pc_to_device = 0;
+
+ switch (peripheral_type) {
+ case IMX_DMATYPE_MEMORY:
+ emi_2_emi = sdma->script_addrs->ap_2_ap_addr;
+ break;
+ case IMX_DMATYPE_DSP:
+ emi_2_per = sdma->script_addrs->bp_2_ap_addr;
+ per_2_emi = sdma->script_addrs->ap_2_bp_addr;
+ break;
+ case IMX_DMATYPE_FIRI:
+ per_2_emi = sdma->script_addrs->firi_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_firi_addr;
+ break;
+ case IMX_DMATYPE_UART:
+ per_2_emi = sdma->script_addrs->uart_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_UART_SP:
+ per_2_emi = sdma->script_addrs->uartsh_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ATA:
+ per_2_emi = sdma->script_addrs->ata_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_ata_addr;
+ break;
+ case IMX_DMATYPE_CSPI:
+ case IMX_DMATYPE_EXT:
+ case IMX_DMATYPE_SSI:
+ per_2_emi = sdma->script_addrs->app_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_app_addr;
+ break;
+ case IMX_DMATYPE_SSI_SP:
+ case IMX_DMATYPE_MMC:
+ case IMX_DMATYPE_SDHC:
+ case IMX_DMATYPE_CSPI_SP:
+ case IMX_DMATYPE_ESAI:
+ case IMX_DMATYPE_MSHC_SP:
+ per_2_emi = sdma->script_addrs->shp_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_shp_addr;
+ break;
+ case IMX_DMATYPE_ASRC:
+ per_2_emi = sdma->script_addrs->asrc_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->asrc_2_mcu_addr;
+ per_2_per = sdma->script_addrs->per_2_per_addr;
+ break;
+ case IMX_DMATYPE_MSHC:
+ per_2_emi = sdma->script_addrs->mshc_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_mshc_addr;
+ break;
+ case IMX_DMATYPE_CCM:
+ per_2_emi = sdma->script_addrs->dptc_dvfs_addr;
+ break;
+ case IMX_DMATYPE_SPDIF:
+ per_2_emi = sdma->script_addrs->spdif_2_mcu_addr;
+ emi_2_per = sdma->script_addrs->mcu_2_spdif_addr;
+ break;
+ case IMX_DMATYPE_IPU_MEMORY:
+ emi_2_per = sdma->script_addrs->ext_mem_2_ipu_addr;
+ break;
+ default:
+ break;
+ }
+
+ sdmac->pc_from_device = per_2_emi;
+ sdmac->pc_to_device = emi_2_per;
+}
+
+static int sdma_load_context(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int load_address;
+ struct sdma_context_data *context = sdma->context;
+ struct sdma_buffer_descriptor *bd0 = sdma->channel[0].bd;
+ int ret;
+
+ if (sdmac->direction == DMA_FROM_DEVICE) {
+ load_address = sdmac->pc_from_device;
+ } else {
+ load_address = sdmac->pc_to_device;
+ }
+
+ if (load_address < 0)
+ return load_address;
+
+ dev_dbg(sdma->dev, "load_address = %d\n", load_address);
+ dev_dbg(sdma->dev, "wml = 0x%08x\n", sdmac->watermark_level);
+ dev_dbg(sdma->dev, "shp_addr = 0x%08x\n", sdmac->shp_addr);
+ dev_dbg(sdma->dev, "per_addr = 0x%08x\n", sdmac->per_addr);
+ dev_dbg(sdma->dev, "event_mask0 = 0x%08x\n", sdmac->event_mask0);
+ dev_dbg(sdma->dev, "event_mask1 = 0x%08x\n", sdmac->event_mask1);
+
+ memset(context, 0, sizeof(*context));
+ context->channel_state.pc = load_address;
+
+ /* Send by context the event mask,base address for peripheral
+ * and watermark level
+ */
+ context->gReg[0] = sdmac->event_mask1;
+ context->gReg[1] = sdmac->event_mask0;
+ context->gReg[2] = sdmac->per_addr;
+ context->gReg[6] = sdmac->shp_addr;
+ context->gReg[7] = sdmac->watermark_level;
+
+ bd0->mode.command = C0_SETDM;
+ bd0->mode.status = BD_DONE | BD_INTR | BD_WRAP | BD_EXTD;
+ bd0->mode.count = sizeof(*context) / 4;
+ bd0->buffer_addr = sdma->context_phys;
+ bd0->ext_buffer_addr = 2048 + (sizeof(*context) / 4) * channel;
+
+ ret = sdma_run_channel(&sdma->channel[0]);
+
+ return ret;
+}
+
+static void sdma_disable_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ __raw_writel(1 << channel, sdma->regs + SDMA_H_STATSTOP);
+ sdmac->status = DMA_ERROR;
+}
+
+static int sdma_config_channel(struct sdma_channel *sdmac)
+{
+ int ret;
+
+ sdma_disable_channel(sdmac);
+
+ sdmac->event_mask0 = 0;
+ sdmac->event_mask1 = 0;
+ sdmac->shp_addr = 0;
+ sdmac->per_addr = 0;
+
+ if (sdmac->event_id0) {
+ if (sdmac->event_id0 > 32)
+ return -EINVAL;
+ sdma_event_enable(sdmac, sdmac->event_id0);
+ }
+
+ switch (sdmac->peripheral_type) {
+ case IMX_DMATYPE_DSP:
+ sdma_config_ownership(sdmac, false, true, true);
+ break;
+ case IMX_DMATYPE_MEMORY:
+ sdma_config_ownership(sdmac, false, true, false);
+ break;
+ default:
+ sdma_config_ownership(sdmac, true, true, false);
+ break;
+ }
+
+ sdma_get_pc(sdmac, sdmac->peripheral_type);
+
+ if ((sdmac->peripheral_type != IMX_DMATYPE_MEMORY) &&
+ (sdmac->peripheral_type != IMX_DMATYPE_DSP)) {
+ /* Handle multiple event channels differently */
+ if (sdmac->event_id1) {
+ sdmac->event_mask1 = 1 << (sdmac->event_id1 % 32);
+ if (sdmac->event_id1 > 31)
+ sdmac->watermark_level |= 1 << 31;
+ sdmac->event_mask0 = 1 << (sdmac->event_id0 % 32);
+ if (sdmac->event_id0 > 31)
+ sdmac->watermark_level |= 1 << 30;
+ } else {
+ sdmac->event_mask0 = 1 << sdmac->event_id0;
+ sdmac->event_mask1 = 1 << (sdmac->event_id0 - 32);
+ }
+ /* Watermark Level */
+ sdmac->watermark_level |= sdmac->watermark_level;
+ /* Address */
+ sdmac->shp_addr = sdmac->per_address;
+ } else {
+ sdmac->watermark_level = 0; /* FIXME: M3_BASE_ADDRESS */
+ }
+
+ ret = sdma_load_context(sdmac);
+
+ return ret;
+}
+
+static int sdma_set_channel_priority(struct sdma_channel *sdmac,
+ unsigned int priority)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+
+ if (priority < MXC_SDMA_MIN_PRIORITY
+ || priority > MXC_SDMA_MAX_PRIORITY) {
+ return -EINVAL;
+ }
+
+ __raw_writel(priority, sdma->regs + SDMA_CHNPRI_0 + 4 * channel);
+
+ return 0;
+}
+
+static int sdma_request_channel(struct sdma_channel *sdmac)
+{
+ struct sdma_engine *sdma = sdmac->sdma;
+ int channel = sdmac->channel;
+ int ret = -EBUSY;
+
+ sdmac->bd = dma_alloc_coherent(NULL, PAGE_SIZE, &sdmac->bd_phys, GFP_KERNEL);
+ if (!sdmac->bd) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ memset(sdmac->bd, 0, PAGE_SIZE);
+
+ sdma->channel_control[channel].base_bd_ptr = sdmac->bd_phys;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ clk_enable(sdma->clk);
+
+ sdma_set_channel_priority(sdmac, MXC_SDMA_DEFAULT_PRIORITY);
+
+ init_completion(&sdmac->done);
+
+ sdmac->buf_tail = 0;
+
+ return 0;
+out:
+
+ return ret;
+}
+
+static void sdma_enable_channel(struct sdma_engine *sdma, int channel)
+{
+ __raw_writel(1 << channel, sdma->regs + SDMA_H_START);
+}
+
+static dma_cookie_t sdma_assign_cookie(struct sdma_channel *sdma)
+{
+ dma_cookie_t cookie = sdma->chan.cookie;
+
+ if (++cookie < 0)
+ cookie = 1;
+
+ sdma->chan.cookie = cookie;
+ sdma->desc.cookie = cookie;
+
+ return cookie;
+}
+
+static struct sdma_channel *to_sdma_chan(struct dma_chan *chan)
+{
+ return container_of(chan, struct sdma_channel, chan);
+}
+
+static dma_cookie_t sdma_tx_submit(struct dma_async_tx_descriptor *tx)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(tx->chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ dma_cookie_t cookie;
+
+ spin_lock_irq(&sdmac->lock);
+
+ cookie = sdma_assign_cookie(sdmac);
+
+ sdma_enable_channel(sdma, tx->chan->chan_id);
+
+ spin_unlock_irq(&sdmac->lock);
+
+ return cookie;
+}
+
+static int sdma_alloc_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct imx_dma_data *data = chan->private;
+ int prio, ret;
+
+ /* No need to execute this for internal channel 0 */
+ if (chan->chan_id == 0)
+ return 0;
+
+ if (!data)
+ return -EINVAL;
+
+ switch (data->priority) {
+ case DMA_PRIO_HIGH:
+ prio = 3;
+ break;
+ case DMA_PRIO_MEDIUM:
+ prio = 2;
+ break;
+ case DMA_PRIO_LOW:
+ default:
+ prio = 1;
+ break;
+ }
+
+ sdmac->peripheral_type = data->peripheral_type;
+ sdmac->event_id0 = data->dma_request;
+ ret = sdma_set_channel_priority(sdmac, prio);
+ if (ret)
+ return ret;
+
+ ret = sdma_request_channel(sdmac);
+ if (ret)
+ return ret;
+
+ dma_async_tx_descriptor_init(&sdmac->desc, chan);
+ sdmac->desc.tx_submit = sdma_tx_submit;
+ /* txd.flags will be overwritten in prep funcs */
+ sdmac->desc.flags = DMA_CTRL_ACK;
+
+ return 0;
+}
+
+static void sdma_free_chan_resources(struct dma_chan *chan)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+
+ sdma_disable_channel(sdmac);
+
+ if (sdmac->event_id0)
+ sdma_event_disable(sdmac, sdmac->event_id0);
+ if (sdmac->event_id1)
+ sdma_event_disable(sdmac, sdmac->event_id1);
+
+ sdmac->event_id0 = 0;
+ sdmac->event_id1 = 0;
+
+ sdma_set_channel_priority(sdmac, 0);
+
+ dma_free_coherent(NULL, PAGE_SIZE, sdmac->bd, sdmac->bd_phys);
+
+ clk_disable(sdma->clk);
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_slave_sg(
+ struct dma_chan *chan, struct scatterlist *sgl,
+ unsigned int sg_len, enum dma_data_direction direction,
+ unsigned long flags)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int ret, i, count;
+ int channel = chan->chan_id;
+ struct scatterlist *sg;
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags = 0;
+
+ dev_dbg(sdma->dev, "setting up %d entries for channel %d.\n",
+ sg_len, channel);
+
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (sg_len > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, sg_len, NUM_BD);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ for_each_sg(sgl, sg, sg_len, i) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = sgl->dma_address;
+
+ count = sg->length;
+
+ if (count > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum bytes for sg entry exceeded: %d > %d\n",
+ channel, count, 0xffff);
+ ret = -EINVAL;
+ goto err_out;
+ }
+
+ bd->mode.count = count;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES) {
+ ret = -EINVAL;
+ goto err_out;
+ }
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT;
+
+ if (sdmac->flags & IMX_DMA_SG_LOOP) {
+ param |= BD_INTR;
+ if (i + 1 == sg_len)
+ param |= BD_WRAP;
+ }
+
+ if (i + 1 == sg_len)
+ param |= BD_INTR;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, count, sg->dma_address,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+ }
+
+ sdmac->num_bd = sg_len;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ return NULL;
+}
+
+static struct dma_async_tx_descriptor *sdma_prep_dma_cyclic(
+ struct dma_chan *chan, dma_addr_t dma_addr, size_t buf_len,
+ size_t period_len, enum dma_data_direction direction)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct sdma_engine *sdma = sdmac->sdma;
+ int num_periods = buf_len / period_len;
+ int channel = chan->chan_id;
+ int ret, i = 0, buf = 0;
+
+ dev_dbg(sdma->dev, "%s channel: %d\n", __func__, channel);
+
+ if (sdmac->status == DMA_IN_PROGRESS)
+ return NULL;
+
+ sdmac->status = DMA_IN_PROGRESS;
+
+ sdmac->flags |= IMX_DMA_SG_LOOP;
+ sdmac->direction = direction;
+ ret = sdma_load_context(sdmac);
+ if (ret)
+ goto err_out;
+
+ if (num_periods > NUM_BD) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum number of sg exceeded: %d > %d\n",
+ channel, num_periods, NUM_BD);
+ goto err_out;
+ }
+
+ if (period_len > 0xffff) {
+ dev_err(sdma->dev, "SDMA channel %d: maximum period size exceeded: %d > %d\n",
+ channel, period_len, 0xffff);
+ goto err_out;
+ }
+
+ while (buf < buf_len) {
+ struct sdma_buffer_descriptor *bd = &sdmac->bd[i];
+ int param;
+
+ bd->buffer_addr = dma_addr;
+
+ bd->mode.count = period_len;
+
+ if (sdmac->word_size > DMA_SLAVE_BUSWIDTH_4_BYTES)
+ goto err_out;
+ if (sdmac->word_size == DMA_SLAVE_BUSWIDTH_4_BYTES)
+ bd->mode.command = 0;
+ else
+ bd->mode.command = sdmac->word_size;
+
+ param = BD_DONE | BD_EXTD | BD_CONT | BD_INTR;
+ if (i + 1 == num_periods)
+ param |= BD_WRAP;
+
+ dev_dbg(sdma->dev, "entry %d: count: %d dma: 0x%08x %s%s\n",
+ i, period_len, dma_addr,
+ param & BD_WRAP ? "wrap" : "",
+ param & BD_INTR ? " intr" : "");
+
+ bd->mode.status = param;
+
+ dma_addr += period_len;
+ buf += period_len;
+
+ i++;
+ }
+
+ sdmac->num_bd = num_periods;
+ sdma->channel_control[channel].current_bd_ptr = sdmac->bd_phys;
+
+ return &sdmac->desc;
+err_out:
+ sdmac->status = DMA_ERROR;
+ return NULL;
+}
+
+static int sdma_control(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
+ unsigned long arg)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ struct dma_slave_config *dmaengine_cfg = (void *)arg;
+
+ switch (cmd) {
+ case DMA_TERMINATE_ALL:
+ sdma_disable_channel(sdmac);
+ return 0;
+ case DMA_SLAVE_CONFIG:
+ if (dmaengine_cfg->direction == DMA_FROM_DEVICE) {
+ sdmac->per_address = dmaengine_cfg->src_addr;
+ sdmac->watermark_level = dmaengine_cfg->src_maxburst;
+ sdmac->word_size = dmaengine_cfg->src_addr_width;
+ } else {
+ sdmac->per_address = dmaengine_cfg->dst_addr;
+ sdmac->watermark_level = dmaengine_cfg->dst_maxburst;
+ sdmac->word_size = dmaengine_cfg->dst_addr_width;
+ }
+ return sdma_config_channel(sdmac);
+ default:
+ return -ENOSYS;
+ }
+
+ return -EINVAL;
+}
+
+static enum dma_status sdma_tx_status(struct dma_chan *chan,
+ dma_cookie_t cookie,
+ struct dma_tx_state *txstate)
+{
+ struct sdma_channel *sdmac = to_sdma_chan(chan);
+ dma_cookie_t last_used;
+ enum dma_status ret;
+
+ last_used = chan->cookie;
+
+ ret = dma_async_is_complete(cookie, sdmac->last_completed, last_used);
+ dma_set_tx_state(txstate, sdmac->last_completed, last_used, 0);
+
+ return ret;
+}
+
+static void sdma_issue_pending(struct dma_chan *chan)
+{
+ /*
+ * Nothing to do. We only have a single descriptor
+ */
+}
+
+static int __init sdma_init(struct sdma_engine *sdma,
+ void *ram_code, int ram_code_size)
+{
+ int i, ret;
+ dma_addr_t ccb_phys;
+
+ switch (sdma->version) {
+ case 1:
+ sdma->num_events = 32;
+ break;
+ case 2:
+ sdma->num_events = 48;
+ break;
+ default:
+ dev_err(sdma->dev, "Unknown version %d. aborting\n", sdma->version);
+ return -ENODEV;
+ }
+
+ clk_enable(sdma->clk);
+
+ /* Be sure SDMA has not started yet */
+ __raw_writel(0, sdma->regs + SDMA_H_C0PTR);
+
+ sdma->channel_control = dma_alloc_coherent(NULL,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control) +
+ sizeof(struct sdma_context_data),
+ &ccb_phys, GFP_KERNEL);
+
+ if (!sdma->channel_control) {
+ ret = -ENOMEM;
+ goto err_dma_alloc;
+ }
+
+ sdma->context = (void *)sdma->channel_control +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+ sdma->context_phys = ccb_phys +
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control);
+
+ /* Zero-out the CCB structures array just allocated */
+ memset(sdma->channel_control, 0,
+ MAX_DMA_CHANNELS * sizeof (struct sdma_channel_control));
+
+ /* disable all channels */
+ for (i = 0; i < sdma->num_events; i++)
+ __raw_writel(0, sdma->regs + chnenbl_ofs(sdma, i));
+
+ /* All channels have priority 0 */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++)
+ __raw_writel(0, sdma->regs + SDMA_CHNPRI_0 + i * 4);
+
+ ret = sdma_request_channel(&sdma->channel[0]);
+ if (ret)
+ goto err_dma_alloc;
+
+ sdma_config_ownership(&sdma->channel[0], false, true, false);
+
+ /* Set Command Channel (Channel Zero) */
+ __raw_writel(0x4050, sdma->regs + SDMA_CHN0ADDR);
+
+ /* Set bits of CONFIG register but with static context switching */
+ /* FIXME: Check whether to set ACR bit depending on clock ratios */
+ __raw_writel(0, sdma->regs + SDMA_H_CONFIG);
+
+ __raw_writel(ccb_phys, sdma->regs + SDMA_H_C0PTR);
+
+ /* download the RAM image for SDMA */
+ sdma_load_script(sdma, ram_code,
+ ram_code_size,
+ sdma->script_addrs->ram_code_start_addr);
+
+ /* Set bits of CONFIG register with given context switching mode */
+ __raw_writel(SDMA_H_CONFIG_CSM, sdma->regs + SDMA_H_CONFIG);
+
+ /* Initializes channel's priorities */
+ sdma_set_channel_priority(&sdma->channel[0], 7);
+
+ clk_disable(sdma->clk);
+
+ return 0;
+
+err_dma_alloc:
+ clk_disable(sdma->clk);
+ dev_err(sdma->dev, "initialisation failed with %d\n", ret);
+ return ret;
+}
+
+static int __init sdma_probe(struct platform_device *pdev)
+{
+ int ret;
+ const struct firmware *fw;
+ const struct sdma_firmware_header *header;
+ const struct sdma_script_start_addrs *addr;
+ int irq;
+ unsigned short *ram_code;
+ struct resource *iores;
+ struct sdma_platform_data *pdata = pdev->dev.platform_data;
+ char *fwname;
+ int i;
+ dma_cap_mask_t mask;
+ struct sdma_engine *sdma;
+
+ sdma = kzalloc(sizeof(*sdma), GFP_KERNEL);
+ if (!sdma)
+ return -ENOMEM;
+
+ sdma->dev = &pdev->dev;
+
+ iores = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ irq = platform_get_irq(pdev, 0);
+ if (!iores || irq < 0 || !pdata) {
+ ret = -EINVAL;
+ goto err_irq;
+ }
+
+ if (!request_mem_region(iores->start, resource_size(iores), pdev->name)) {
+ ret = -EBUSY;
+ goto err_request_region;
+ }
+
+ sdma->clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(sdma->clk)) {
+ ret = PTR_ERR(sdma->clk);
+ goto err_clk;
+ }
+
+ sdma->regs = ioremap(iores->start, resource_size(iores));
+ if (!sdma->regs) {
+ ret = -ENOMEM;
+ goto err_ioremap;
+ }
+
+ ret = request_irq(irq, sdma_int_handler, 0, "sdma", sdma);
+ if (ret)
+ goto err_request_irq;
+
+ fwname = kasprintf(GFP_KERNEL, "sdma-%s-to%d.bin",
+ pdata->cpu_name, pdata->to_version);
+ if (!fwname) {
+ ret = -ENOMEM;
+ goto err_cputype;
+ }
+
+ ret = request_firmware(&fw, fwname, &pdev->dev);
+ if (ret) {
+ dev_err(&pdev->dev, "request firmware \"%s\" failed with %d\n",
+ fwname, ret);
+ kfree(fwname);
+ goto err_cputype;
+ }
+ kfree(fwname);
+
+ if (fw->size < sizeof(*header))
+ goto err_firmware;
+
+ header = (struct sdma_firmware_header *)fw->data;
+
+ if (header->magic != SDMA_FIRMWARE_MAGIC)
+ goto err_firmware;
+ if (header->ram_code_start + header->ram_code_size > fw->size)
+ goto err_firmware;
+
+ addr = (void *)header + header->script_addrs_start;
+ ram_code = (void *)header + header->ram_code_start;
+ sdma->script_addrs = kmalloc(sizeof(*addr), GFP_KERNEL);
+ if (!sdma->script_addrs)
+ goto err_firmware;
+ memcpy(sdma->script_addrs, addr, sizeof(*addr));
+
+ sdma->version = pdata->sdma_version;
+
+ INIT_LIST_HEAD(&sdma->dma_device.channels);
+ /* Initialize channel parameters */
+ for (i = 0; i < MAX_DMA_CHANNELS; i++) {
+ struct sdma_channel *sdmac = &sdma->channel[i];
+
+ sdmac->sdma = sdma;
+ spin_lock_init(&sdmac->lock);
+
+ dma_cap_set(DMA_SLAVE, sdma->dma_device.cap_mask);
+ dma_cap_set(DMA_CYCLIC, sdma->dma_device.cap_mask);
+
+ sdmac->chan.device = &sdma->dma_device;
+ sdmac->chan.chan_id = i;
+ sdmac->channel = i;
+
+ /* Add the channel to the DMAC list */
+ list_add_tail(&sdmac->chan.device_node, &sdma->dma_device.channels);
+ }
+
+ ret = sdma_init(sdma, ram_code, header->ram_code_size);
+ if (ret)
+ goto err_init;
+
+ sdma->dma_device.dev = &pdev->dev;
+
+ sdma->dma_device.device_alloc_chan_resources = sdma_alloc_chan_resources;
+ sdma->dma_device.device_free_chan_resources = sdma_free_chan_resources;
+ sdma->dma_device.device_tx_status = sdma_tx_status;
+ sdma->dma_device.device_prep_slave_sg = sdma_prep_slave_sg;
+ sdma->dma_device.device_prep_dma_cyclic = sdma_prep_dma_cyclic;
+ sdma->dma_device.device_control = sdma_control;
+ sdma->dma_device.device_issue_pending = sdma_issue_pending;
+
+ ret = dma_async_device_register(&sdma->dma_device);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register\n");
+ goto err_init;
+ }
+
+ dev_info(&pdev->dev, "initialized (firmware %d.%d)\n",
+ header->version_major,
+ header->version_minor);
+
+ /* request channel 0. This is an internal control channel
+ * to the SDMA engine and not available to clients.
+ */
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ dma_request_channel(mask, NULL, NULL);
+
+ release_firmware(fw);
+
+ return 0;
+
+err_init:
+ kfree(sdma->script_addrs);
+err_firmware:
+ release_firmware(fw);
+err_cputype:
+ free_irq(irq, sdma);
+err_request_irq:
+ iounmap(sdma->regs);
+err_ioremap:
+ clk_put(sdma->clk);
+err_clk:
+ release_mem_region(iores->start, resource_size(iores));
+err_request_region:
+err_irq:
+ kfree(sdma);
+ return 0;
+}
+
+static int __exit sdma_remove(struct platform_device *pdev)
+{
+ return -EBUSY;
+}
+
+static struct platform_driver sdma_driver = {
+ .driver = {
+ .name = "imx-sdma",
+ },
+ .remove = __exit_p(sdma_remove),
+};
+
+static int __init sdma_module_init(void)
+{
+ return platform_driver_probe(&sdma_driver, sdma_probe);
+}
+subsys_initcall(sdma_module_init);
+
+MODULE_AUTHOR("Sascha Hauer, Pengutronix <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("i.MX SDMA driver");
+MODULE_LICENSE("GPL");
--
1.7.1
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply related [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-23 17:48 ` Uwe Kleine-König
@ 2010-08-28 15:18 ` Linus Walleij
-1 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-28 15:18 UTC (permalink / raw)
To: Uwe Kleine-König
Cc: Sascha Hauer, linux-kernel, Dan Williams, linux-arm-kernel
2010/8/23 Uwe Kleine-König <u.kleine-koenig@pengutronix.de>:
>> + evt = readl(SDMA_H_EVTOVR);
>> + mcu = readl(SDMA_H_HOSTOVR);
>> + dsp = readl(SDMA_H_DSPOVR);
> __raw_readl?
Sorry I never understood this __raw_[read|write][b|w|l] vs.
plain read[b|w|l] proliferation in some drivers and code.
What's the reason for?
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-28 15:18 ` Linus Walleij
0 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-28 15:18 UTC (permalink / raw)
To: linux-arm-kernel
2010/8/23 Uwe Kleine-K?nig <u.kleine-koenig@pengutronix.de>:
>> + ? ? evt = readl(SDMA_H_EVTOVR);
>> + ? ? mcu = readl(SDMA_H_HOSTOVR);
>> + ? ? dsp = readl(SDMA_H_DSPOVR);
> __raw_readl?
Sorry I never understood this __raw_[read|write][b|w|l] vs.
plain read[b|w|l] proliferation in some drivers and code.
What's the reason for?
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-28 15:18 ` Linus Walleij
@ 2010-08-28 15:27 ` Marek Vasut
-1 siblings, 0 replies; 78+ messages in thread
From: Marek Vasut @ 2010-08-28 15:27 UTC (permalink / raw)
To: linux-arm-kernel
Cc: Linus Walleij, Uwe Kleine-König, Sascha Hauer, Dan Williams,
linux-kernel
Dne So 28. srpna 2010 17:18:17 Linus Walleij napsal(a):
> 2010/8/23 Uwe Kleine-König <u.kleine-koenig@pengutronix.de>:
> >> + evt = readl(SDMA_H_EVTOVR);
> >> + mcu = readl(SDMA_H_HOSTOVR);
> >> + dsp = readl(SDMA_H_DSPOVR);
> >
> > __raw_readl?
>
> Sorry I never understood this __raw_[read|write][b|w|l] vs.
> plain read[b|w|l] proliferation in some drivers and code.
>
> What's the reason for?
Hey,
this trick is, if you create the VA<->PA mapping at the kernel start (eg. see
how pxa_map_io() is replacedon some devices for instance), you then use the VA
address you specified and use __raw_{read,write}[b,w,l](). So use
__raw_{read,write}[b,w,l]() on drivers specific for certain device and use
{read,write}[b,w,l]() on ioremap()ed memory areas, aka. in common drivers.
Cheers
>
> Yours,
> Linus Walleij
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-28 15:27 ` Marek Vasut
0 siblings, 0 replies; 78+ messages in thread
From: Marek Vasut @ 2010-08-28 15:27 UTC (permalink / raw)
To: linux-arm-kernel
Dne So 28. srpna 2010 17:18:17 Linus Walleij napsal(a):
> 2010/8/23 Uwe Kleine-K?nig <u.kleine-koenig@pengutronix.de>:
> >> + evt = readl(SDMA_H_EVTOVR);
> >> + mcu = readl(SDMA_H_HOSTOVR);
> >> + dsp = readl(SDMA_H_DSPOVR);
> >
> > __raw_readl?
>
> Sorry I never understood this __raw_[read|write][b|w|l] vs.
> plain read[b|w|l] proliferation in some drivers and code.
>
> What's the reason for?
Hey,
this trick is, if you create the VA<->PA mapping at the kernel start (eg. see
how pxa_map_io() is replacedon some devices for instance), you then use the VA
address you specified and use __raw_{read,write}[b,w,l](). So use
__raw_{read,write}[b,w,l]() on drivers specific for certain device and use
{read,write}[b,w,l]() on ioremap()ed memory areas, aka. in common drivers.
Cheers
>
> Yours,
> Linus Walleij
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-28 15:27 ` Marek Vasut
@ 2010-08-28 16:18 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-28 16:18 UTC (permalink / raw)
To: Marek Vasut
Cc: linux-arm-kernel, Linus Walleij, Dan Williams, linux-kernel,
Uwe Kleine-König
On Sat, Aug 28, 2010 at 05:27:10PM +0200, Marek Vasut wrote:
> Dne So 28. srpna 2010 17:18:17 Linus Walleij napsal(a):
> > 2010/8/23 Uwe Kleine-König <u.kleine-koenig@pengutronix.de>:
> > >> + evt = readl(SDMA_H_EVTOVR);
> > >> + mcu = readl(SDMA_H_HOSTOVR);
> > >> + dsp = readl(SDMA_H_DSPOVR);
> > >
> > > __raw_readl?
> >
> > Sorry I never understood this __raw_[read|write][b|w|l] vs.
> > plain read[b|w|l] proliferation in some drivers and code.
> >
> > What's the reason for?
>
> Hey,
>
> this trick is, if you create the VA<->PA mapping at the kernel start (eg. see
> how pxa_map_io() is replacedon some devices for instance), you then use the VA
> address you specified and use __raw_{read,write}[b,w,l](). So use
> __raw_{read,write}[b,w,l]() on drivers specific for certain device and use
> {read,write}[b,w,l]() on ioremap()ed memory areas, aka. in common drivers.
Nope, this has nothing to do with static mappings vs. ioremap. The
difference is that read[b,w,l] do little endian accesses suitable for
PCI whereas the __raw_* functions do accesses in CPU endianess.
Peripherals integrated into a SoC like the SDMA engine here are
normally accessible in native endianess and thus need the __raw_*
functions. An external network controller (for example a LAN9117) will
probably need the non raw functions. Note that 99% of the arm users use
little endian only and thus cpu_to_le* is a noop, so both types will work
for most people.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-28 16:18 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-28 16:18 UTC (permalink / raw)
To: linux-arm-kernel
On Sat, Aug 28, 2010 at 05:27:10PM +0200, Marek Vasut wrote:
> Dne So 28. srpna 2010 17:18:17 Linus Walleij napsal(a):
> > 2010/8/23 Uwe Kleine-K?nig <u.kleine-koenig@pengutronix.de>:
> > >> + evt = readl(SDMA_H_EVTOVR);
> > >> + mcu = readl(SDMA_H_HOSTOVR);
> > >> + dsp = readl(SDMA_H_DSPOVR);
> > >
> > > __raw_readl?
> >
> > Sorry I never understood this __raw_[read|write][b|w|l] vs.
> > plain read[b|w|l] proliferation in some drivers and code.
> >
> > What's the reason for?
>
> Hey,
>
> this trick is, if you create the VA<->PA mapping at the kernel start (eg. see
> how pxa_map_io() is replacedon some devices for instance), you then use the VA
> address you specified and use __raw_{read,write}[b,w,l](). So use
> __raw_{read,write}[b,w,l]() on drivers specific for certain device and use
> {read,write}[b,w,l]() on ioremap()ed memory areas, aka. in common drivers.
Nope, this has nothing to do with static mappings vs. ioremap. The
difference is that read[b,w,l] do little endian accesses suitable for
PCI whereas the __raw_* functions do accesses in CPU endianess.
Peripherals integrated into a SoC like the SDMA engine here are
normally accessible in native endianess and thus need the __raw_*
functions. An external network controller (for example a LAN9117) will
probably need the non raw functions. Note that 99% of the arm users use
little endian only and thus cpu_to_le* is a noop, so both types will work
for most people.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-28 16:18 ` Sascha Hauer
@ 2010-08-28 16:30 ` Marek Vasut
-1 siblings, 0 replies; 78+ messages in thread
From: Marek Vasut @ 2010-08-28 16:30 UTC (permalink / raw)
To: Sascha Hauer
Cc: linux-arm-kernel, Linus Walleij, Dan Williams, linux-kernel,
Uwe Kleine-König
Dne So 28. srpna 2010 18:18:58 Sascha Hauer napsal(a):
> On Sat, Aug 28, 2010 at 05:27:10PM +0200, Marek Vasut wrote:
> > Dne So 28. srpna 2010 17:18:17 Linus Walleij napsal(a):
> > > 2010/8/23 Uwe Kleine-König <u.kleine-koenig@pengutronix.de>:
> > > >> + evt = readl(SDMA_H_EVTOVR);
> > > >> + mcu = readl(SDMA_H_HOSTOVR);
> > > >> + dsp = readl(SDMA_H_DSPOVR);
> > > >
> > > > __raw_readl?
> > >
> > > Sorry I never understood this __raw_[read|write][b|w|l] vs.
> > > plain read[b|w|l] proliferation in some drivers and code.
> > >
> > > What's the reason for?
> >
> > Hey,
> >
> > this trick is, if you create the VA<->PA mapping at the kernel start (eg.
> > see how pxa_map_io() is replacedon some devices for instance), you then
> > use the VA address you specified and use __raw_{read,write}[b,w,l](). So
> > use __raw_{read,write}[b,w,l]() on drivers specific for certain device
> > and use {read,write}[b,w,l]() on ioremap()ed memory areas, aka. in
> > common drivers.
>
> Nope, this has nothing to do with static mappings vs. ioremap. The
> difference is that read[b,w,l] do little endian accesses suitable for
> PCI whereas the __raw_* functions do accesses in CPU endianess.
> Peripherals integrated into a SoC like the SDMA engine here are
> normally accessible in native endianess and thus need the __raw_*
> functions. An external network controller (for example a LAN9117) will
> probably need the non raw functions. Note that 99% of the arm users use
> little endian only and thus cpu_to_le* is a noop, so both types will work
> for most people.
Sorry, you got me here. Why do we have io{read,write} then btw. ? That's for
x86's io space, right ?
Thanks, cheers and sorry for the confusion
>
> Sascha
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-28 16:30 ` Marek Vasut
0 siblings, 0 replies; 78+ messages in thread
From: Marek Vasut @ 2010-08-28 16:30 UTC (permalink / raw)
To: linux-arm-kernel
Dne So 28. srpna 2010 18:18:58 Sascha Hauer napsal(a):
> On Sat, Aug 28, 2010 at 05:27:10PM +0200, Marek Vasut wrote:
> > Dne So 28. srpna 2010 17:18:17 Linus Walleij napsal(a):
> > > 2010/8/23 Uwe Kleine-K?nig <u.kleine-koenig@pengutronix.de>:
> > > >> + evt = readl(SDMA_H_EVTOVR);
> > > >> + mcu = readl(SDMA_H_HOSTOVR);
> > > >> + dsp = readl(SDMA_H_DSPOVR);
> > > >
> > > > __raw_readl?
> > >
> > > Sorry I never understood this __raw_[read|write][b|w|l] vs.
> > > plain read[b|w|l] proliferation in some drivers and code.
> > >
> > > What's the reason for?
> >
> > Hey,
> >
> > this trick is, if you create the VA<->PA mapping at the kernel start (eg.
> > see how pxa_map_io() is replacedon some devices for instance), you then
> > use the VA address you specified and use __raw_{read,write}[b,w,l](). So
> > use __raw_{read,write}[b,w,l]() on drivers specific for certain device
> > and use {read,write}[b,w,l]() on ioremap()ed memory areas, aka. in
> > common drivers.
>
> Nope, this has nothing to do with static mappings vs. ioremap. The
> difference is that read[b,w,l] do little endian accesses suitable for
> PCI whereas the __raw_* functions do accesses in CPU endianess.
> Peripherals integrated into a SoC like the SDMA engine here are
> normally accessible in native endianess and thus need the __raw_*
> functions. An external network controller (for example a LAN9117) will
> probably need the non raw functions. Note that 99% of the arm users use
> little endian only and thus cpu_to_le* is a noop, so both types will work
> for most people.
Sorry, you got me here. Why do we have io{read,write} then btw. ? That's for
x86's io space, right ?
Thanks, cheers and sorry for the confusion
>
> Sascha
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-28 16:30 ` Marek Vasut
@ 2010-08-28 17:20 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-28 17:20 UTC (permalink / raw)
To: Marek Vasut
Cc: Linus Walleij, Dan Williams, linux-kernel, linux-arm-kernel,
Uwe Kleine-König
On Sat, Aug 28, 2010 at 06:30:05PM +0200, Marek Vasut wrote:
> Dne So 28. srpna 2010 18:18:58 Sascha Hauer napsal(a):
> > On Sat, Aug 28, 2010 at 05:27:10PM +0200, Marek Vasut wrote:
> > > Dne So 28. srpna 2010 17:18:17 Linus Walleij napsal(a):
> > > > 2010/8/23 Uwe Kleine-König <u.kleine-koenig@pengutronix.de>:
> > > > >> + evt = readl(SDMA_H_EVTOVR);
> > > > >> + mcu = readl(SDMA_H_HOSTOVR);
> > > > >> + dsp = readl(SDMA_H_DSPOVR);
> > > > >
> > > > > __raw_readl?
> > > >
> > > > Sorry I never understood this __raw_[read|write][b|w|l] vs.
> > > > plain read[b|w|l] proliferation in some drivers and code.
> > > >
> > > > What's the reason for?
> > >
> > > Hey,
> > >
> > > this trick is, if you create the VA<->PA mapping at the kernel start (eg.
> > > see how pxa_map_io() is replacedon some devices for instance), you then
> > > use the VA address you specified and use __raw_{read,write}[b,w,l](). So
> > > use __raw_{read,write}[b,w,l]() on drivers specific for certain device
> > > and use {read,write}[b,w,l]() on ioremap()ed memory areas, aka. in
> > > common drivers.
> >
> > Nope, this has nothing to do with static mappings vs. ioremap. The
> > difference is that read[b,w,l] do little endian accesses suitable for
> > PCI whereas the __raw_* functions do accesses in CPU endianess.
> > Peripherals integrated into a SoC like the SDMA engine here are
> > normally accessible in native endianess and thus need the __raw_*
> > functions. An external network controller (for example a LAN9117) will
> > probably need the non raw functions. Note that 99% of the arm users use
> > little endian only and thus cpu_to_le* is a noop, so both types will work
> > for most people.
>
> Sorry, you got me here. Why do we have io{read,write} then btw. ? That's for
> x86's io space, right ?
Looking at the implementation in lib/iomap.c it seems that
io{read,write} can do both PIO and MMIO accesses depending on the
address given to them.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-28 17:20 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-28 17:20 UTC (permalink / raw)
To: linux-arm-kernel
On Sat, Aug 28, 2010 at 06:30:05PM +0200, Marek Vasut wrote:
> Dne So 28. srpna 2010 18:18:58 Sascha Hauer napsal(a):
> > On Sat, Aug 28, 2010 at 05:27:10PM +0200, Marek Vasut wrote:
> > > Dne So 28. srpna 2010 17:18:17 Linus Walleij napsal(a):
> > > > 2010/8/23 Uwe Kleine-K?nig <u.kleine-koenig@pengutronix.de>:
> > > > >> + evt = readl(SDMA_H_EVTOVR);
> > > > >> + mcu = readl(SDMA_H_HOSTOVR);
> > > > >> + dsp = readl(SDMA_H_DSPOVR);
> > > > >
> > > > > __raw_readl?
> > > >
> > > > Sorry I never understood this __raw_[read|write][b|w|l] vs.
> > > > plain read[b|w|l] proliferation in some drivers and code.
> > > >
> > > > What's the reason for?
> > >
> > > Hey,
> > >
> > > this trick is, if you create the VA<->PA mapping at the kernel start (eg.
> > > see how pxa_map_io() is replacedon some devices for instance), you then
> > > use the VA address you specified and use __raw_{read,write}[b,w,l](). So
> > > use __raw_{read,write}[b,w,l]() on drivers specific for certain device
> > > and use {read,write}[b,w,l]() on ioremap()ed memory areas, aka. in
> > > common drivers.
> >
> > Nope, this has nothing to do with static mappings vs. ioremap. The
> > difference is that read[b,w,l] do little endian accesses suitable for
> > PCI whereas the __raw_* functions do accesses in CPU endianess.
> > Peripherals integrated into a SoC like the SDMA engine here are
> > normally accessible in native endianess and thus need the __raw_*
> > functions. An external network controller (for example a LAN9117) will
> > probably need the non raw functions. Note that 99% of the arm users use
> > little endian only and thus cpu_to_le* is a noop, so both types will work
> > for most people.
>
> Sorry, you got me here. Why do we have io{read,write} then btw. ? That's for
> x86's io space, right ?
Looking at the implementation in lib/iomap.c it seems that
io{read,write} can do both PIO and MMIO accesses depending on the
address given to them.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-28 16:18 ` Sascha Hauer
@ 2010-08-29 12:35 ` Linus Walleij
-1 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-29 12:35 UTC (permalink / raw)
To: Sascha Hauer
Cc: Marek Vasut, linux-arm-kernel, Dan Williams, linux-kernel,
Uwe Kleine-König
2010/8/28 Sascha Hauer <s.hauer@pengutronix.de>:
> Peripherals integrated into a SoC like the SDMA engine here are
> normally accessible in native endianess and thus need the __raw_*
> functions.
So the SDMA actually switch and twist around the endianness of
its registers if it's synthesized into a bigendian version of the system?
Or hardware-dynamically even depending on the setting of the
endianness bit in the ARM core?
How can that possibly even work...
But I'm still impressed by the statemachines some silicon
engineers come up with so wouldn't surprise me. X-)
I understand the code now atleast. Our hardware is always
synthesized LE so we so handily use [read|write][b|w|l]
everywhere, luckily.
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-29 12:35 ` Linus Walleij
0 siblings, 0 replies; 78+ messages in thread
From: Linus Walleij @ 2010-08-29 12:35 UTC (permalink / raw)
To: linux-arm-kernel
2010/8/28 Sascha Hauer <s.hauer@pengutronix.de>:
> Peripherals integrated into a SoC like the SDMA engine here are
> normally accessible in native endianess and thus need the __raw_*
> functions.
So the SDMA actually switch and twist around the endianness of
its registers if it's synthesized into a bigendian version of the system?
Or hardware-dynamically even depending on the setting of the
endianness bit in the ARM core?
How can that possibly even work...
But I'm still impressed by the statemachines some silicon
engineers come up with so wouldn't surprise me. X-)
I understand the code now atleast. Our hardware is always
synthesized LE so we so handily use [read|write][b|w|l]
everywhere, luckily.
Yours,
Linus Walleij
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v3] dmaengine: Add Freescale i.MX SDMA support
2010-08-27 12:22 ` Sascha Hauer
(?)
@ 2010-08-29 21:46 ` Marc Reilly
-1 siblings, 0 replies; 78+ messages in thread
From: Marc Reilly @ 2010-08-29 21:46 UTC (permalink / raw)
To: linux-arm-kernel
Hi,
Just a typo and a minor niggle..
On Friday, August 27, 2010 10:22:31 pm Sascha Hauer wrote:
> This patch adds support for the Freescale i.MX SDMA engine.
> +static int sdma_config_ownership(struct sdma_channel *sdmac,
> + bool event_override, bool mcu_verride, bool dsp_override)
> +{
> + struct sdma_engine *sdma = sdmac->sdma;
> + int channel = sdmac->channel;
> + u32 evt, mcu, dsp;
> +
> + if (event_override && mcu_verride && dsp_override)
> + return -EINVAL;
> +
> + evt = __raw_readl(sdma->regs + SDMA_H_EVTOVR);
> + mcu = __raw_readl(sdma->regs + SDMA_H_HOSTOVR);
> + dsp = __raw_readl(sdma->regs + SDMA_H_DSPOVR);
> +
> + if (dsp_override)
> + dsp &= ~(1 << channel);
> + else
> + dsp |= (1 << channel);
> +
> + if (event_override)
> + evt &= ~(1 << channel);
> + else
> + evt |= (1 << channel);
> +
> + if (mcu_verride)
> + mcu &= ~(1 << channel);
> + else
> + mcu |= (1 << channel);
> +
> + __raw_writel(evt, sdma->regs + SDMA_H_EVTOVR);
> + __raw_writel(mcu, sdma->regs + SDMA_H_HOSTOVR);
> + __raw_writel(dsp, sdma->regs + SDMA_H_DSPOVR);
> +
> + return 0;
> +}
s/mcu_verride/mcu_override
> +static void sdma_event_enable(struct sdma_channel *sdmac, unsigned int
> event) +{
> + struct sdma_engine *sdma = sdmac->sdma;
> + int channel = sdmac->channel;
> + u32 val;
> + u32 chnenbl = chnenbl_ofs(sdma, event);
> +
> + val = __raw_readl(sdma->regs + chnenbl);
> + val |= (1 << channel);
> + __raw_writel(val, sdma->regs + chnenbl);
> +}
s/chnenbl/chanenbl ?
(or chanenable, etc.)
Everything else was easily readable, but I did a double take when I read
through this.
Cheers,
Marc
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-29 12:35 ` Linus Walleij
@ 2010-08-30 12:55 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-30 12:55 UTC (permalink / raw)
To: Linus Walleij
Cc: Marek Vasut, linux-arm-kernel, Dan Williams, linux-kernel,
Uwe Kleine-König
On Sun, Aug 29, 2010 at 02:35:01PM +0200, Linus Walleij wrote:
> 2010/8/28 Sascha Hauer <s.hauer@pengutronix.de>:
>
> > Peripherals integrated into a SoC like the SDMA engine here are
> > normally accessible in native endianess and thus need the __raw_*
> > functions.
>
> So the SDMA actually switch and twist around the endianness of
> its registers if it's synthesized into a bigendian version of the system?
>
> Or hardware-dynamically even depending on the setting of the
> endianness bit in the ARM core?
I think it's the bus between the ARM core and the periherals which
changes the endianess. I have never tried running an i.MX in big endian
mode, so I can only guess how the system really behaves in BE mode.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-08-30 12:55 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-08-30 12:55 UTC (permalink / raw)
To: linux-arm-kernel
On Sun, Aug 29, 2010 at 02:35:01PM +0200, Linus Walleij wrote:
> 2010/8/28 Sascha Hauer <s.hauer@pengutronix.de>:
>
> > Peripherals integrated into a SoC like the SDMA engine here are
> > normally accessible in native endianess and thus need the __raw_*
> > functions.
>
> So the SDMA actually switch and twist around the endianness of
> its registers if it's synthesized into a bigendian version of the system?
>
> Or hardware-dynamically even depending on the setting of the
> endianness bit in the ARM core?
I think it's the bus between the ARM core and the periherals which
changes the endianess. I have never tried running an i.MX in big endian
mode, so I can only guess how the system really behaves in BE mode.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
2010-08-28 16:30 ` Marek Vasut
@ 2010-09-02 11:20 ` Russell King - ARM Linux
-1 siblings, 0 replies; 78+ messages in thread
From: Russell King - ARM Linux @ 2010-09-02 11:20 UTC (permalink / raw)
To: Marek Vasut
Cc: Sascha Hauer, Linus Walleij, Dan Williams, linux-kernel,
linux-arm-kernel, Uwe Kleine-König
On Sat, Aug 28, 2010 at 06:30:05PM +0200, Marek Vasut wrote:
> Dne So 28. srpna 2010 18:18:58 Sascha Hauer napsal(a):
> > Nope, this has nothing to do with static mappings vs. ioremap. The
> > difference is that read[b,w,l] do little endian accesses suitable for
> > PCI whereas the __raw_* functions do accesses in CPU endianess.
> > Peripherals integrated into a SoC like the SDMA engine here are
> > normally accessible in native endianess and thus need the __raw_*
> > functions. An external network controller (for example a LAN9117) will
> > probably need the non raw functions. Note that 99% of the arm users use
> > little endian only and thus cpu_to_le* is a noop, so both types will work
> > for most people.
>
> Sorry, you got me here. Why do we have io{read,write} then btw. ? That's for
> x86's io space, right ?
io{read,write} are for use with ioremap/ioport_map, and allow drivers to
be written which can access registers via MMIO or the PC IO space.
If your driver doesn't support the PC IO space (iow, doesn't use
ioport_map) there's no point using the io{read,write} APIs.
Also note that there's one very big difference between read[bwl] and
__raw_read[bwl]. The former have a barrier to ensure correct ordering
for drivers doing DMA, the latter do not. So if you use the latter and
you care about data being visible to a DMA agent, you have to ensure
you have proper barriers in place.
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3 v2] dmaengine: Add Freescale i.MX SDMA support
@ 2010-09-02 11:20 ` Russell King - ARM Linux
0 siblings, 0 replies; 78+ messages in thread
From: Russell King - ARM Linux @ 2010-09-02 11:20 UTC (permalink / raw)
To: linux-arm-kernel
On Sat, Aug 28, 2010 at 06:30:05PM +0200, Marek Vasut wrote:
> Dne So 28. srpna 2010 18:18:58 Sascha Hauer napsal(a):
> > Nope, this has nothing to do with static mappings vs. ioremap. The
> > difference is that read[b,w,l] do little endian accesses suitable for
> > PCI whereas the __raw_* functions do accesses in CPU endianess.
> > Peripherals integrated into a SoC like the SDMA engine here are
> > normally accessible in native endianess and thus need the __raw_*
> > functions. An external network controller (for example a LAN9117) will
> > probably need the non raw functions. Note that 99% of the arm users use
> > little endian only and thus cpu_to_le* is a noop, so both types will work
> > for most people.
>
> Sorry, you got me here. Why do we have io{read,write} then btw. ? That's for
> x86's io space, right ?
io{read,write} are for use with ioremap/ioport_map, and allow drivers to
be written which can access registers via MMIO or the PC IO space.
If your driver doesn't support the PC IO space (iow, doesn't use
ioport_map) there's no point using the io{read,write} APIs.
Also note that there's one very big difference between read[bwl] and
__raw_read[bwl]. The former have a barrier to ensure correct ordering
for drivers doing DMA, the latter do not. So if you use the latter and
you care about data being visible to a DMA agent, you have to ensure
you have proper barriers in place.
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
2010-08-16 12:21 ` Linus Walleij
@ 2010-09-02 14:06 ` Russell King - ARM Linux
-1 siblings, 0 replies; 78+ messages in thread
From: Russell King - ARM Linux @ 2010-09-02 14:06 UTC (permalink / raw)
To: Linus Walleij; +Cc: Sascha Hauer, Dan Williams, linux-kernel, linux-arm-kernel
On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> > +/*
> > + * Buffer descriptor
> > + */
> > +struct sdma_buffer_descriptor {
> > + struct sdma_mode_count mode;
> > + u32 buffer_addr; /* address of the buffer described */
> > + u32 ext_buffer_addr; /* extended buffer address */
>
> Shouldn't these be dma_addr_t? OK that's probably u32
> anyway but just to make a marker...
If this is describing hardware, then it makes sense to use uNN rather
than dma_addr_t. dma_addr_t may be subject to change in the future.
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support
@ 2010-09-02 14:06 ` Russell King - ARM Linux
0 siblings, 0 replies; 78+ messages in thread
From: Russell King - ARM Linux @ 2010-09-02 14:06 UTC (permalink / raw)
To: linux-arm-kernel
On Mon, Aug 16, 2010 at 02:21:06PM +0200, Linus Walleij wrote:
> > +/*
> > + * Buffer descriptor
> > + */
> > +struct sdma_buffer_descriptor {
> > + ? ? ? struct sdma_mode_count ?mode;
> > + ? ? ? u32 buffer_addr; ? ?/* address of the buffer described */
> > + ? ? ? u32 ext_buffer_addr; /* extended buffer address */
>
> Shouldn't these be dma_addr_t? OK that's probably u32
> anyway but just to make a marker...
If this is describing hardware, then it makes sense to use uNN rather
than dma_addr_t. dma_addr_t may be subject to change in the future.
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 1/3] dmaengine: add possibility for cyclic transfers
2010-08-16 11:07 ` Sascha Hauer
@ 2010-09-20 13:01 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-09-20 13:01 UTC (permalink / raw)
To: Dan Williams
Cc: Linus Walleij, linux-kernel, linux-arm-kernel, Haavard Skinnemoen
Hi Dan,
Any comment to this patch?
Sascha
On Mon, Aug 16, 2010 at 01:07:48PM +0200, Sascha Hauer wrote:
> Cyclic transfers are useful for audio where a single buffer divided
> in periods has to be transfered endlessly until stopped. After being
> prepared the transfer is started using the dma_async_descriptor->tx_submit
> function. dma_async_descriptor->callback is called after each period.
> The transfer is stopped using the DMA_TERMINATE_ALL callback.
> While being used for cyclic transfers the channel cannot be used
> for other transfer types.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
> ---
> drivers/dma/dmaengine.c | 2 ++
> include/linux/dmaengine.h | 6 +++++-
> 2 files changed, 7 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index 9d31d5e..e5e79ce 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -692,6 +692,8 @@ int dma_async_device_register(struct dma_device *device)
> !device->device_prep_dma_interrupt);
> BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> !device->device_prep_slave_sg);
> + BUG_ON(dma_has_cap(DMA_CYCLIC, device->cap_mask) &&
> + !device->device_prep_dma_cyclic);
> BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> !device->device_control);
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index c61d4ca..0df7864 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -67,10 +67,11 @@ enum dma_transaction_type {
> DMA_PRIVATE,
> DMA_ASYNC_TX,
> DMA_SLAVE,
> + DMA_CYCLIC,
> };
>
> /* last transaction type for creation of the capabilities mask */
> -#define DMA_TX_TYPE_END (DMA_SLAVE + 1)
> +#define DMA_TX_TYPE_END (DMA_CYCLIC + 1)
>
>
> /**
> @@ -478,6 +479,9 @@ struct dma_device {
> struct dma_chan *chan, struct scatterlist *sgl,
> unsigned int sg_len, enum dma_data_direction direction,
> unsigned long flags);
> + struct dma_async_tx_descriptor *(*device_prep_dma_cyclic)(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction);
> int (*device_control)(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> unsigned long arg);
>
> --
> 1.7.1
>
>
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
@ 2010-09-20 13:01 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-09-20 13:01 UTC (permalink / raw)
To: linux-arm-kernel
Hi Dan,
Any comment to this patch?
Sascha
On Mon, Aug 16, 2010 at 01:07:48PM +0200, Sascha Hauer wrote:
> Cyclic transfers are useful for audio where a single buffer divided
> in periods has to be transfered endlessly until stopped. After being
> prepared the transfer is started using the dma_async_descriptor->tx_submit
> function. dma_async_descriptor->callback is called after each period.
> The transfer is stopped using the DMA_TERMINATE_ALL callback.
> While being used for cyclic transfers the channel cannot be used
> for other transfer types.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> Cc: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
> ---
> drivers/dma/dmaengine.c | 2 ++
> include/linux/dmaengine.h | 6 +++++-
> 2 files changed, 7 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index 9d31d5e..e5e79ce 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -692,6 +692,8 @@ int dma_async_device_register(struct dma_device *device)
> !device->device_prep_dma_interrupt);
> BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> !device->device_prep_slave_sg);
> + BUG_ON(dma_has_cap(DMA_CYCLIC, device->cap_mask) &&
> + !device->device_prep_dma_cyclic);
> BUG_ON(dma_has_cap(DMA_SLAVE, device->cap_mask) &&
> !device->device_control);
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index c61d4ca..0df7864 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -67,10 +67,11 @@ enum dma_transaction_type {
> DMA_PRIVATE,
> DMA_ASYNC_TX,
> DMA_SLAVE,
> + DMA_CYCLIC,
> };
>
> /* last transaction type for creation of the capabilities mask */
> -#define DMA_TX_TYPE_END (DMA_SLAVE + 1)
> +#define DMA_TX_TYPE_END (DMA_CYCLIC + 1)
>
>
> /**
> @@ -478,6 +479,9 @@ struct dma_device {
> struct dma_chan *chan, struct scatterlist *sgl,
> unsigned int sg_len, enum dma_data_direction direction,
> unsigned long flags);
> + struct dma_async_tx_descriptor *(*device_prep_dma_cyclic)(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction);
> int (*device_control)(struct dma_chan *chan, enum dma_ctrl_cmd cmd,
> unsigned long arg);
>
> --
> 1.7.1
>
>
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
2010-08-16 11:07 ` Sascha Hauer
(?)
(?)
@ 2010-09-20 13:02 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-09-20 13:02 UTC (permalink / raw)
To: Dan Williams; +Cc: Linus Walleij, Dan Williams, linux-arm-kernel, linux-kernel
Hi Dan,
Any comment?
Sascha
On Mon, Aug 16, 2010 at 01:07:49PM +0200, Sascha Hauer wrote:
> Currently dmaengine users have to explicitely dereference function
> pointers in struct dma_device. For the convenience of drivers and
> to be more flexible when changing the dmaengine later add static
> inline wrapper functions for the dma commands.
>
> This patch is not complete yet. If there's consensus on this patch
> I'll provide an updated patch with the missing functions.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
> include/linux/dmaengine.h | 41 +++++++++++++++++++++++++++++++++++++++++
> 1 files changed, 41 insertions(+), 0 deletions(-)
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index 0df7864..635c60b 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -491,6 +491,47 @@ struct dma_device {
> void (*device_issue_pending)(struct dma_chan *chan);
> };
>
> +static inline int dmaengine_device_control(struct dma_chan *chan,
> + enum dma_ctrl_cmd cmd,
> + unsigned long arg)
> +{
> + return chan->device->device_control(chan, cmd, arg);
> +}
> +
> +static inline int dmaengine_slave_config(struct dma_chan *chan,
> + struct dma_slave_config *config)
> +{
> + return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
> + (unsigned long)config);
> +}
> +
> +static inline int dmaengine_terminate_all(struct dma_chan *chan)
> +{
> + return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
> + struct dma_chan *chan, struct scatterlist *sgl,
> + unsigned int sg_len, enum dma_data_direction direction,
> + unsigned long flags)
> +{
> + return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
> + flags);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction)
> +{
> + return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
> + period_len, direction);
> +}
> +
> +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
> +{
> + return desc->tx_submit(desc);
> +}
> +
> static inline bool dmaengine_check_align(u8 align, size_t off1, size_t off2, size_t len)
> {
> size_t mask;
> --
> 1.7.1
>
>
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 1/3] dmaengine: add possibility for cyclic transfers
2010-09-20 13:01 ` Sascha Hauer
@ 2010-09-23 19:42 ` Dan Williams
-1 siblings, 0 replies; 78+ messages in thread
From: Dan Williams @ 2010-09-23 19:42 UTC (permalink / raw)
To: Sascha Hauer
Cc: Linus Walleij, linux-kernel, linux-arm-kernel, Haavard Skinnemoen
On Mon, Sep 20, 2010 at 6:01 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> Hi Dan,
>
> Any comment to this patch?
Looks good to me especially given the unique implications for
callbacks. Although, I'd like to put a description of the
prep_dma_cyclic semantics in the source code. Can you respin this
with a description added to the kernel-doc for struct dma_device?
--
Dan
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
@ 2010-09-23 19:42 ` Dan Williams
0 siblings, 0 replies; 78+ messages in thread
From: Dan Williams @ 2010-09-23 19:42 UTC (permalink / raw)
To: linux-arm-kernel
On Mon, Sep 20, 2010 at 6:01 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> Hi Dan,
>
> Any comment to this patch?
Looks good to me especially given the unique implications for
callbacks. Although, I'd like to put a description of the
prep_dma_cyclic semantics in the source code. Can you respin this
with a description added to the kernel-doc for struct dma_device?
--
Dan
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
2010-08-16 11:07 ` Sascha Hauer
@ 2010-09-23 19:53 ` Dan Williams
-1 siblings, 0 replies; 78+ messages in thread
From: Dan Williams @ 2010-09-23 19:53 UTC (permalink / raw)
To: Sascha Hauer; +Cc: linux-kernel, Linus Walleij, linux-arm-kernel
On Mon, Aug 16, 2010 at 4:07 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> Currently dmaengine users have to explicitely dereference function
> pointers in struct dma_device. For the convenience of drivers and
> to be more flexible when changing the dmaengine later add static
> inline wrapper functions for the dma commands.
>
> This patch is not complete yet. If there's consensus on this patch
> I'll provide an updated patch with the missing functions.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
> include/linux/dmaengine.h | 41 +++++++++++++++++++++++++++++++++++++++++
> 1 files changed, 41 insertions(+), 0 deletions(-)
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index 0df7864..635c60b 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -491,6 +491,47 @@ struct dma_device {
> void (*device_issue_pending)(struct dma_chan *chan);
> };
>
> +static inline int dmaengine_device_control(struct dma_chan *chan,
> + enum dma_ctrl_cmd cmd,
> + unsigned long arg)
> +{
> + return chan->device->device_control(chan, cmd, arg);
> +}
> +
> +static inline int dmaengine_slave_config(struct dma_chan *chan,
> + struct dma_slave_config *config)
> +{
> + return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
> + (unsigned long)config);
> +}
> +
> +static inline int dmaengine_terminate_all(struct dma_chan *chan)
> +{
> + return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
> + struct dma_chan *chan, struct scatterlist *sgl,
> + unsigned int sg_len, enum dma_data_direction direction,
> + unsigned long flags)
> +{
> + return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
> + flags);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
> + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + size_t period_len, enum dma_data_direction direction)
> +{
> + return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
> + period_len, direction);
> +}
> +
No strong disagreements on the above, the type safety of
dmaengine_slave_config() is nice.
> +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
> +{
> + return desc->tx_submit(desc);
> +}
This one can drop the tx.
--
Dan
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
@ 2010-09-23 19:53 ` Dan Williams
0 siblings, 0 replies; 78+ messages in thread
From: Dan Williams @ 2010-09-23 19:53 UTC (permalink / raw)
To: linux-arm-kernel
On Mon, Aug 16, 2010 at 4:07 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> Currently dmaengine users have to explicitely dereference function
> pointers in struct dma_device. For the convenience of drivers and
> to be more flexible when changing the dmaengine later add static
> inline wrapper functions for the dma commands.
>
> This patch is not complete yet. If there's consensus on this patch
> I'll provide an updated patch with the missing functions.
>
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
> ?include/linux/dmaengine.h | ? 41 +++++++++++++++++++++++++++++++++++++++++
> ?1 files changed, 41 insertions(+), 0 deletions(-)
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index 0df7864..635c60b 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -491,6 +491,47 @@ struct dma_device {
> ? ? ? ?void (*device_issue_pending)(struct dma_chan *chan);
> ?};
>
> +static inline int dmaengine_device_control(struct dma_chan *chan,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?enum dma_ctrl_cmd cmd,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?unsigned long arg)
> +{
> + ? ? ? return chan->device->device_control(chan, cmd, arg);
> +}
> +
> +static inline int dmaengine_slave_config(struct dma_chan *chan,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? struct dma_slave_config *config)
> +{
> + ? ? ? return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
> + ? ? ? ? ? ? ? ? ? ? ? (unsigned long)config);
> +}
> +
> +static inline int dmaengine_terminate_all(struct dma_chan *chan)
> +{
> + ? ? ? return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
> + ? ? ? ? ? ? ? struct dma_chan *chan, struct scatterlist *sgl,
> + ? ? ? ? ? ? ? unsigned int sg_len, enum dma_data_direction direction,
> + ? ? ? ? ? ? ? unsigned long flags)
> +{
> + ? ? ? return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
> + ? ? ? ? ? ? ? ? ? ? ? flags);
> +}
> +
> +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
> + ? ? ? ? ? ? ? struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> + ? ? ? ? ? ? ? size_t period_len, enum dma_data_direction direction)
> +{
> + ? ? ? return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
> + ? ? ? ? ? ? ? ? ? ? ? period_len, direction);
> +}
> +
No strong disagreements on the above, the type safety of
dmaengine_slave_config() is nice.
> +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
> +{
> + ? ? ? return desc->tx_submit(desc);
> +}
This one can drop the tx.
--
Dan
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
2010-09-23 19:53 ` Dan Williams
@ 2010-09-24 7:25 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-09-24 7:25 UTC (permalink / raw)
To: Dan Williams; +Cc: linux-kernel, Linus Walleij, linux-arm-kernel
Hi Dan,
On Thu, Sep 23, 2010 at 12:53:58PM -0700, Dan Williams wrote:
> > diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> > index 0df7864..635c60b 100644
> > --- a/include/linux/dmaengine.h
> > +++ b/include/linux/dmaengine.h
> > @@ -491,6 +491,47 @@ struct dma_device {
> > void (*device_issue_pending)(struct dma_chan *chan);
> > };
> >
> > +static inline int dmaengine_device_control(struct dma_chan *chan,
> > + enum dma_ctrl_cmd cmd,
> > + unsigned long arg)
> > +{
> > + return chan->device->device_control(chan, cmd, arg);
> > +}
> > +
> > +static inline int dmaengine_slave_config(struct dma_chan *chan,
> > + struct dma_slave_config *config)
> > +{
> > + return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
> > + (unsigned long)config);
> > +}
> > +
> > +static inline int dmaengine_terminate_all(struct dma_chan *chan)
> > +{
> > + return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
> > +}
> > +
> > +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
> > + struct dma_chan *chan, struct scatterlist *sgl,
> > + unsigned int sg_len, enum dma_data_direction direction,
> > + unsigned long flags)
> > +{
> > + return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
> > + flags);
> > +}
> > +
> > +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
> > + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> > + size_t period_len, enum dma_data_direction direction)
> > +{
> > + return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
> > + period_len, direction);
> > +}
> > +
>
> No strong disagreements on the above, the type safety of
> dmaengine_slave_config() is nice.
So you have only small disagreements? ;) I can drop the dmaengine_prep_*
functions and only keep the device_control functions if like it better.
>
> > +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
> > +{
> > + return desc->tx_submit(desc);
> > +}
>
> This one can drop the tx.
You mean the function should be named dmaengine_submit?
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
@ 2010-09-24 7:25 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-09-24 7:25 UTC (permalink / raw)
To: linux-arm-kernel
Hi Dan,
On Thu, Sep 23, 2010 at 12:53:58PM -0700, Dan Williams wrote:
> > diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> > index 0df7864..635c60b 100644
> > --- a/include/linux/dmaengine.h
> > +++ b/include/linux/dmaengine.h
> > @@ -491,6 +491,47 @@ struct dma_device {
> > ? ? ? ?void (*device_issue_pending)(struct dma_chan *chan);
> > ?};
> >
> > +static inline int dmaengine_device_control(struct dma_chan *chan,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?enum dma_ctrl_cmd cmd,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?unsigned long arg)
> > +{
> > + ? ? ? return chan->device->device_control(chan, cmd, arg);
> > +}
> > +
> > +static inline int dmaengine_slave_config(struct dma_chan *chan,
> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? struct dma_slave_config *config)
> > +{
> > + ? ? ? return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
> > + ? ? ? ? ? ? ? ? ? ? ? (unsigned long)config);
> > +}
> > +
> > +static inline int dmaengine_terminate_all(struct dma_chan *chan)
> > +{
> > + ? ? ? return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
> > +}
> > +
> > +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
> > + ? ? ? ? ? ? ? struct dma_chan *chan, struct scatterlist *sgl,
> > + ? ? ? ? ? ? ? unsigned int sg_len, enum dma_data_direction direction,
> > + ? ? ? ? ? ? ? unsigned long flags)
> > +{
> > + ? ? ? return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
> > + ? ? ? ? ? ? ? ? ? ? ? flags);
> > +}
> > +
> > +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
> > + ? ? ? ? ? ? ? struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
> > + ? ? ? ? ? ? ? size_t period_len, enum dma_data_direction direction)
> > +{
> > + ? ? ? return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
> > + ? ? ? ? ? ? ? ? ? ? ? period_len, direction);
> > +}
> > +
>
> No strong disagreements on the above, the type safety of
> dmaengine_slave_config() is nice.
So you have only small disagreements? ;) I can drop the dmaengine_prep_*
functions and only keep the device_control functions if like it better.
>
> > +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
> > +{
> > + ? ? ? return desc->tx_submit(desc);
> > +}
>
> This one can drop the tx.
You mean the function should be named dmaengine_submit?
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 1/3] dmaengine: add possibility for cyclic transfers
2010-09-23 19:42 ` Dan Williams
@ 2010-09-24 7:25 ` Sascha Hauer
-1 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-09-24 7:25 UTC (permalink / raw)
To: Dan Williams
Cc: Linus Walleij, linux-kernel, linux-arm-kernel, Haavard Skinnemoen
On Thu, Sep 23, 2010 at 12:42:20PM -0700, Dan Williams wrote:
> On Mon, Sep 20, 2010 at 6:01 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > Hi Dan,
> >
> > Any comment to this patch?
>
> Looks good to me especially given the unique implications for
> callbacks. Although, I'd like to put a description of the
> prep_dma_cyclic semantics in the source code. Can you respin this
> with a description added to the kernel-doc for struct dma_device?
Sure, will do.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 1/3] dmaengine: add possibility for cyclic transfers
@ 2010-09-24 7:25 ` Sascha Hauer
0 siblings, 0 replies; 78+ messages in thread
From: Sascha Hauer @ 2010-09-24 7:25 UTC (permalink / raw)
To: linux-arm-kernel
On Thu, Sep 23, 2010 at 12:42:20PM -0700, Dan Williams wrote:
> On Mon, Sep 20, 2010 at 6:01 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > Hi Dan,
> >
> > Any comment to this patch?
>
> Looks good to me especially given the unique implications for
> callbacks. Although, I'd like to put a description of the
> prep_dma_cyclic semantics in the source code. Can you respin this
> with a description added to the kernel-doc for struct dma_device?
Sure, will do.
Sascha
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
^ permalink raw reply [flat|nested] 78+ messages in thread
* Re: [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
2010-09-24 7:25 ` Sascha Hauer
@ 2010-09-24 15:45 ` Dan Williams
-1 siblings, 0 replies; 78+ messages in thread
From: Dan Williams @ 2010-09-24 15:45 UTC (permalink / raw)
To: Sascha Hauer; +Cc: linux-kernel, Linus Walleij, linux-arm-kernel
On Fri, Sep 24, 2010 at 12:25 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> Hi Dan,
>
> On Thu, Sep 23, 2010 at 12:53:58PM -0700, Dan Williams wrote:
>> > diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
>> > index 0df7864..635c60b 100644
>> > --- a/include/linux/dmaengine.h
>> > +++ b/include/linux/dmaengine.h
>> > @@ -491,6 +491,47 @@ struct dma_device {
>> > void (*device_issue_pending)(struct dma_chan *chan);
>> > };
>> >
>> > +static inline int dmaengine_device_control(struct dma_chan *chan,
>> > + enum dma_ctrl_cmd cmd,
>> > + unsigned long arg)
>> > +{
>> > + return chan->device->device_control(chan, cmd, arg);
>> > +}
>> > +
>> > +static inline int dmaengine_slave_config(struct dma_chan *chan,
>> > + struct dma_slave_config *config)
>> > +{
>> > + return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
>> > + (unsigned long)config);
>> > +}
>> > +
>> > +static inline int dmaengine_terminate_all(struct dma_chan *chan)
>> > +{
>> > + return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
>> > +}
>> > +
>> > +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
>> > + struct dma_chan *chan, struct scatterlist *sgl,
>> > + unsigned int sg_len, enum dma_data_direction direction,
>> > + unsigned long flags)
>> > +{
>> > + return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
>> > + flags);
>> > +}
>> > +
>> > +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
>> > + struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
>> > + size_t period_len, enum dma_data_direction direction)
>> > +{
>> > + return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
>> > + period_len, direction);
>> > +}
>> > +
>>
>> No strong disagreements on the above, the type safety of
>> dmaengine_slave_config() is nice.
>
> So you have only small disagreements? ;) I can drop the dmaengine_prep_*
> functions and only keep the device_control functions if like it better.
>
The prep versions may just be code churn, unless there is another
advantage for having a helper... debug/tracing perhaps?
>>
>> > +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
>> > +{
>> > + return desc->tx_submit(desc);
>> > +}
>>
>> This one can drop the tx.
>
> You mean the function should be named dmaengine_submit?
Yes, the "tx" is often mistaken, understandably, for "transmit".
Linus proposed renaming ->tx_submit() to just ->submit(). This helper
would be a softer way to introduce that rename.
--
Dan
^ permalink raw reply [flat|nested] 78+ messages in thread
* [PATCH 2/3] dmaengine: add wrapper functions for dmaengine
@ 2010-09-24 15:45 ` Dan Williams
0 siblings, 0 replies; 78+ messages in thread
From: Dan Williams @ 2010-09-24 15:45 UTC (permalink / raw)
To: linux-arm-kernel
On Fri, Sep 24, 2010 at 12:25 AM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> Hi Dan,
>
> On Thu, Sep 23, 2010 at 12:53:58PM -0700, Dan Williams wrote:
>> > diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
>> > index 0df7864..635c60b 100644
>> > --- a/include/linux/dmaengine.h
>> > +++ b/include/linux/dmaengine.h
>> > @@ -491,6 +491,47 @@ struct dma_device {
>> > ? ? ? ?void (*device_issue_pending)(struct dma_chan *chan);
>> > ?};
>> >
>> > +static inline int dmaengine_device_control(struct dma_chan *chan,
>> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?enum dma_ctrl_cmd cmd,
>> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?unsigned long arg)
>> > +{
>> > + ? ? ? return chan->device->device_control(chan, cmd, arg);
>> > +}
>> > +
>> > +static inline int dmaengine_slave_config(struct dma_chan *chan,
>> > + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? struct dma_slave_config *config)
>> > +{
>> > + ? ? ? return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
>> > + ? ? ? ? ? ? ? ? ? ? ? (unsigned long)config);
>> > +}
>> > +
>> > +static inline int dmaengine_terminate_all(struct dma_chan *chan)
>> > +{
>> > + ? ? ? return dmaengine_device_control(chan, DMA_TERMINATE_ALL, 0);
>> > +}
>> > +
>> > +static inline struct dma_async_tx_descriptor *dmaengine_prep_slave_sg(
>> > + ? ? ? ? ? ? ? struct dma_chan *chan, struct scatterlist *sgl,
>> > + ? ? ? ? ? ? ? unsigned int sg_len, enum dma_data_direction direction,
>> > + ? ? ? ? ? ? ? unsigned long flags)
>> > +{
>> > + ? ? ? return chan->device->device_prep_slave_sg(chan, sgl, sg_len, direction,
>> > + ? ? ? ? ? ? ? ? ? ? ? flags);
>> > +}
>> > +
>> > +static inline struct dma_async_tx_descriptor *dmaengine_prep_cyclic(
>> > + ? ? ? ? ? ? ? struct dma_chan *chan, dma_addr_t buf_addr, size_t buf_len,
>> > + ? ? ? ? ? ? ? size_t period_len, enum dma_data_direction direction)
>> > +{
>> > + ? ? ? return chan->device->device_prep_dma_cyclic(chan, buf_addr, buf_len,
>> > + ? ? ? ? ? ? ? ? ? ? ? period_len, direction);
>> > +}
>> > +
>>
>> No strong disagreements on the above, the type safety of
>> dmaengine_slave_config() is nice.
>
> So you have only small disagreements? ;) I can drop the dmaengine_prep_*
> functions and only keep the device_control functions if like it better.
>
The prep versions may just be code churn, unless there is another
advantage for having a helper... debug/tracing perhaps?
>>
>> > +static inline int dmaengine_tx_submit(struct dma_async_tx_descriptor *desc)
>> > +{
>> > + ? ? ? return desc->tx_submit(desc);
>> > +}
>>
>> This one can drop the tx.
>
> You mean the function should be named dmaengine_submit?
Yes, the "tx" is often mistaken, understandably, for "transmit".
Linus proposed renaming ->tx_submit() to just ->submit(). This helper
would be a softer way to introduce that rename.
--
Dan
^ permalink raw reply [flat|nested] 78+ messages in thread
end of thread, other threads:[~2010-09-24 15:45 UTC | newest]
Thread overview: 78+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2010-08-16 11:07 [RFC] dmaengine: assorted patches and Freescale SDMA support Sascha Hauer
2010-08-16 11:07 ` Sascha Hauer
2010-08-16 11:07 ` [PATCH 1/3] dmaengine: add possibility for cyclic transfers Sascha Hauer
2010-08-16 11:07 ` Sascha Hauer
2010-08-16 11:56 ` Lothar Waßmann
2010-08-16 11:56 ` Lothar Waßmann
2010-08-16 12:27 ` Linus Walleij
2010-08-16 12:27 ` Linus Walleij
2010-08-16 12:32 ` Sascha Hauer
2010-08-16 12:32 ` Sascha Hauer
2010-08-16 12:22 ` Linus Walleij
2010-08-16 12:22 ` Linus Walleij
2010-09-20 13:01 ` Sascha Hauer
2010-09-20 13:01 ` Sascha Hauer
2010-09-23 19:42 ` Dan Williams
2010-09-23 19:42 ` Dan Williams
2010-09-24 7:25 ` Sascha Hauer
2010-09-24 7:25 ` Sascha Hauer
2010-08-16 11:07 ` [PATCH 2/3] dmaengine: add wrapper functions for dmaengine Sascha Hauer
2010-08-16 11:07 ` Sascha Hauer
2010-08-23 7:17 ` Sascha Hauer
2010-08-23 7:17 ` Sascha Hauer
2010-09-20 13:02 ` Sascha Hauer
2010-09-23 19:53 ` Dan Williams
2010-09-23 19:53 ` Dan Williams
2010-09-24 7:25 ` Sascha Hauer
2010-09-24 7:25 ` Sascha Hauer
2010-09-24 15:45 ` Dan Williams
2010-09-24 15:45 ` Dan Williams
2010-08-16 11:07 ` [PATCH 3/3] dmaengine: Add Freescale i.MX SDMA support Sascha Hauer
2010-08-16 11:07 ` Sascha Hauer
2010-08-16 12:21 ` Linus Walleij
2010-08-16 12:21 ` Linus Walleij
2010-08-16 14:15 ` Sascha Hauer
2010-08-16 14:15 ` Sascha Hauer
2010-08-17 4:36 ` Baruch Siach
2010-08-17 4:36 ` Baruch Siach
2010-08-17 6:47 ` Sascha Hauer
2010-08-17 6:47 ` Sascha Hauer
2010-08-18 3:49 ` Alexei Babich
2010-08-18 4:41 ` Baruch Siach
2010-08-18 11:17 ` Philippe Rétornaz
2010-08-18 11:17 ` Philippe Rétornaz
2010-08-24 7:10 ` [PATCH 3/3 v3] " Sascha Hauer
2010-08-24 7:10 ` Sascha Hauer
2010-09-02 14:06 ` [PATCH 3/3] " Russell King - ARM Linux
2010-09-02 14:06 ` Russell King - ARM Linux
2010-08-23 12:57 ` [PATCH 3/3 v2] " Sascha Hauer
2010-08-23 12:57 ` Sascha Hauer
2010-08-23 17:30 ` Linus Walleij
2010-08-23 17:30 ` Linus Walleij
2010-08-24 6:58 ` Sascha Hauer
2010-08-24 6:58 ` Sascha Hauer
2010-08-23 17:48 ` Uwe Kleine-König
2010-08-23 17:48 ` Uwe Kleine-König
2010-08-28 15:18 ` Linus Walleij
2010-08-28 15:18 ` Linus Walleij
2010-08-28 15:27 ` Marek Vasut
2010-08-28 15:27 ` Marek Vasut
2010-08-28 16:18 ` Sascha Hauer
2010-08-28 16:18 ` Sascha Hauer
2010-08-28 16:30 ` Marek Vasut
2010-08-28 16:30 ` Marek Vasut
2010-08-28 17:20 ` Sascha Hauer
2010-08-28 17:20 ` Sascha Hauer
2010-09-02 11:20 ` Russell King - ARM Linux
2010-09-02 11:20 ` Russell King - ARM Linux
2010-08-29 12:35 ` Linus Walleij
2010-08-29 12:35 ` Linus Walleij
2010-08-30 12:55 ` Sascha Hauer
2010-08-30 12:55 ` Sascha Hauer
2010-08-24 7:58 ` Lothar Waßmann
2010-08-24 7:58 ` Lothar Waßmann
2010-08-24 15:01 ` Linus Walleij
2010-08-24 15:01 ` Linus Walleij
2010-08-27 12:22 ` [PATCH 3/3 v3] " Sascha Hauer
2010-08-27 12:22 ` Sascha Hauer
2010-08-29 21:46 ` Marc Reilly
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.