* [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface
@ 2021-12-01 22:03 Stefan Mätje
2021-12-01 22:03 ` [PATCH 1/4] MAINTAINERS: add Stefan Mätje as maintainer for the esd electronics GmbH CAN drivers Stefan Mätje
` (7 more replies)
0 siblings, 8 replies; 15+ messages in thread
From: Stefan Mätje @ 2021-12-01 22:03 UTC (permalink / raw)
To: linux-can, Marc Kleine-Budde, Wolfgang Grandegger; +Cc: netdev, linux-kernel
The purpose of this patch is to introduce a new CAN driver to support
the esd GmbH 402 family of CAN interface boards. The hardware design
is based on a CAN controller implemented in a FPGA attached to a
PCIe link.
More information on these boards can be found following the links
included in the commit message.
This patch supports all boards but will operate the CAN-FD capable
boards only in Classic-CAN mode. The CAN-FD support will be added
when the initial patch has stabilized.
The patch is reuses the previous work of my former colleague:
Link: https://lore.kernel.org/linux-can/1426592308-23817-1-git-send-email-thomas.koerper@esd.eu/
*Note*: scripts/checkpatch.pl still emits the following warnings:
- esd_402_pci-core.c:270: Possible unnecessary 'out of memory' message
This error message is there to tell the user that the DMA allocation
failed and not an allocation for normal kernel memory.
- esdacc.h:255: The irq_cnt pointer is still declared volatile and
this has a reason and is explained in detail in the header
referencing the exception noted in volatile-considered-harmful.rst.
The patch is based on the linux-can-next testing branch.
Changes in v6:
- Fixed the statistic handling of RX overrun errors and increase
net_device_stats::rx_errors instead of net_device_stats::rx_dropped.
- Added a patch to not increase rx statistics when generating a CAN
rx error message frame as suggested on the linux-can list.
- Added a patch to not not increase rx_bytes statistics for RTR frames
as suggested on the linux-can list.
The last two patches change the statistics handling from the previous
style used in other drivers to the newly suggested one.
Changes in v5:
- Added the initialization for netdev::dev_port as it is implemented
for another CAN driver. See
https://lore.kernel.org/linux-can/20211026180553.1953189-1-mailhol.vincent@wanadoo.fr/
Changes in v4:
- Fixed the build failure on ARCH=arm64 that was found by the Intel
kernel test robot. See
https://lore.kernel.org/linux-can/202109120608.7ZbQXkRh-lkp@intel.com/
Removed error monitoring code that used GCC's built-in compiler
functions for atomic access (__sync_* functions). GCC versions
after 9 (tested with "gcc-10 (Ubuntu 10.3.0-1ubuntu1~20.04)")
don't implement the intrinsic atomic as in-line code but call
"__aarch64_ldadd4_acq_rel" on arm64. This GCC support function
is not exported by the kernel and therefore the module build
post-processing fails.
Removed that code because the error monitoring never showed a
problem during the development this year.
Changes in v3:
- Rework the bus-off restart logic in acc_set_mode() and
handle_core_msg_errstatechange() to call netif_wake_queue() from the
error active event.
- Changed pci402_init_card() to allocate a variable sized array of
struct acc_core using devm_kcalloc() instead of using a fixed size
array in struct pci402_card.
- Changed handle_core_msg_txabort() to release aborted TX frames in
TX FIFO order.
- Fixed the acc_close() function to abort all pending TX request in
esdACC controller.
- Fixed counting of transmit aborts in handle_core_msg_txabort().
It is now done like in can_flush_echo_skb().
- Fixed handle_core_msg_buserr() to create error frames including the
CAN RX and TX error counters that were missing.
- Fixed acc_set_bittiming() neither to touch LOM mode setting of
esdACC controller nor to enter or leave RESET mode.
The esdACC controller is going active on the CAN bus in acc_open()
and is going inactive (RESET mode) again in acc_close().
- Rely on the automatic release of memory fetched by devm_kzalloc().
But still use devm_irq_free() explicitely to make sure that the
interrupt handler is disconnected at that point.
This avoids a possible crash in non-MSI mode due to the IRQ
triggered by another device on the same PCI IRQ line.
- Changed to use DMA map API instead of pci_*_consistent compatibility
wrappers.
- Fixed stale email references and updated copyright information.
- Removed any traces of future CAN-FD support.
Changes in v2:
- Avoid warning triggered by -Wshift-count-overflow on architectures
with 32-bit dma_addr_t.
- Fixed Makefile not to build the kernel module always. Doing this
renamed esd402_pci.c to esd_402_pci-core.c as recommended by Marc.
Stefan Mätje (4):
MAINTAINERS: add Stefan Mätje as maintainer for the esd electronics
GmbH CAN drivers
can: esd: add support for esd GmbH PCIe/402 CAN interface family
can: esd_402_pci: do not increase rx statistics when generating a CAN
rx error message frame
can: esd_402_pci: do not increase rx_bytes statistics for RTR frames
MAINTAINERS | 8 +
drivers/net/can/Kconfig | 1 +
drivers/net/can/Makefile | 1 +
drivers/net/can/esd/Kconfig | 12 +
drivers/net/can/esd/Makefile | 7 +
drivers/net/can/esd/esd_402_pci-core.c | 502 ++++++++++++++++
drivers/net/can/esd/esdacc.c | 772 +++++++++++++++++++++++++
drivers/net/can/esd/esdacc.h | 380 ++++++++++++
8 files changed, 1683 insertions(+)
create mode 100644 drivers/net/can/esd/Kconfig
create mode 100644 drivers/net/can/esd/Makefile
create mode 100644 drivers/net/can/esd/esd_402_pci-core.c
create mode 100644 drivers/net/can/esd/esdacc.c
create mode 100644 drivers/net/can/esd/esdacc.h
base-commit: f659c5c7de7982018bb961cf1f9960e60f526bdf
--
2.25.1
^ permalink raw reply [flat|nested] 15+ messages in thread* [PATCH 1/4] MAINTAINERS: add Stefan Mätje as maintainer for the esd electronics GmbH CAN drivers 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje @ 2021-12-01 22:03 ` Stefan Mätje 2021-12-01 22:03 ` [PATCH 2/4] can: esd: add support for esd GmbH PCIe/402 CAN interface family Stefan Mätje ` (6 subsequent siblings) 7 siblings, 0 replies; 15+ messages in thread From: Stefan Mätje @ 2021-12-01 22:03 UTC (permalink / raw) To: linux-can, Marc Kleine-Budde, Wolfgang Grandegger; +Cc: netdev, linux-kernel Adding myself (Stefan Mätje) as a maintainer for the esd_usb2.c the driver for the CAN-USB/2 and CAN-USB/Micro. Also for the upcoming driver for the PCIe/402 interface card family. Signed-off-by: Stefan Mätje <stefan.maetje@esd.eu> --- MAINTAINERS | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/MAINTAINERS b/MAINTAINERS index 975086c5345d..1cd23dcbdd8e 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -6980,6 +6980,14 @@ S: Maintained F: include/linux/errseq.h F: lib/errseq.c +ESD CAN NETWORK DRIVERS +M: Stefan Mätje <stefan.maetje@esd.eu> +R: socketcan@esd.eu +L: linux-can@vger.kernel.org +S: Maintained +F: drivers/net/can/esd/ +F: drivers/net/can/usb/esd_usb2.c + ET131X NETWORK DRIVER M: Mark Einon <mark.einon@gmail.com> S: Odd Fixes -- 2.25.1 ^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH 2/4] can: esd: add support for esd GmbH PCIe/402 CAN interface family 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje 2021-12-01 22:03 ` [PATCH 1/4] MAINTAINERS: add Stefan Mätje as maintainer for the esd electronics GmbH CAN drivers Stefan Mätje @ 2021-12-01 22:03 ` Stefan Mätje 2022-02-01 17:25 ` [PATCH v6 " Marc Kleine-Budde 2021-12-01 22:03 ` [PATCH 3/4] can: esd_402_pci: do not increase rx statistics when generating a CAN rx error message frame Stefan Mätje ` (5 subsequent siblings) 7 siblings, 1 reply; 15+ messages in thread From: Stefan Mätje @ 2021-12-01 22:03 UTC (permalink / raw) To: linux-can, Marc Kleine-Budde, Wolfgang Grandegger; +Cc: netdev, linux-kernel This patch adds support for the PCI based PCIe/402 CAN interface family from esd GmbH that is available with various form factors (https://esd.eu/technologie/can/can-und-can-fd-interfaces). All boards utilize a FPGA based CAN controller solution developed by esd (esdACC). For more information on the esdACC see https://esd.eu/technologie/can/esd-advanced-can-controller-esdacc. This driver detects all available CAN interface board variants of the family but atm. operates the CAN-FD capable devices in Classic-CAN mode only! A later patch will introduce the CAN-FD functionality in this driver. Co-developed-by: Thomas Körper <thomas.koerper@esd.eu> Signed-off-by: Thomas Körper <thomas.koerper@esd.eu> Signed-off-by: Stefan Mätje <stefan.maetje@esd.eu> --- drivers/net/can/Kconfig | 1 + drivers/net/can/Makefile | 1 + drivers/net/can/esd/Kconfig | 12 + drivers/net/can/esd/Makefile | 7 + drivers/net/can/esd/esd_402_pci-core.c | 502 ++++++++++++++++ drivers/net/can/esd/esdacc.c | 777 +++++++++++++++++++++++++ drivers/net/can/esd/esdacc.h | 380 ++++++++++++ 7 files changed, 1680 insertions(+) create mode 100644 drivers/net/can/esd/Kconfig create mode 100644 drivers/net/can/esd/Makefile create mode 100644 drivers/net/can/esd/esd_402_pci-core.c create mode 100644 drivers/net/can/esd/esdacc.c create mode 100644 drivers/net/can/esd/esdacc.h diff --git a/drivers/net/can/Kconfig b/drivers/net/can/Kconfig index fff259247d52..47cfb6ae0772 100644 --- a/drivers/net/can/Kconfig +++ b/drivers/net/can/Kconfig @@ -170,6 +170,7 @@ config PCH_CAN source "drivers/net/can/c_can/Kconfig" source "drivers/net/can/cc770/Kconfig" +source "drivers/net/can/esd/Kconfig" source "drivers/net/can/ifi_canfd/Kconfig" source "drivers/net/can/m_can/Kconfig" source "drivers/net/can/mscan/Kconfig" diff --git a/drivers/net/can/Makefile b/drivers/net/can/Makefile index a2b4463d8480..015b6fc110d1 100644 --- a/drivers/net/can/Makefile +++ b/drivers/net/can/Makefile @@ -8,6 +8,7 @@ obj-$(CONFIG_CAN_VXCAN) += vxcan.o obj-$(CONFIG_CAN_SLCAN) += slcan.o obj-y += dev/ +obj-y += esd/ obj-y += rcar/ obj-y += spi/ obj-y += usb/ diff --git a/drivers/net/can/esd/Kconfig b/drivers/net/can/esd/Kconfig new file mode 100644 index 000000000000..54bfc366634c --- /dev/null +++ b/drivers/net/can/esd/Kconfig @@ -0,0 +1,12 @@ +# SPDX-License-Identifier: GPL-2.0-only +config CAN_ESD_402_PCI + tristate "esd electronics gmbh CAN-PCI(e)/402 family" + depends on PCI && HAS_DMA + help + Support for C402 card family from esd electronics gmbh. + This card family is based on the ESDACC CAN controller and + available in several form factors: PCI, PCIe, PCIe Mini, + M.2 PCIe, CPCIserial, PMC, XMC (see https://esd.eu/en) + + This driver can also be built as a module. In this case the + module will be called esd_402_pci. diff --git a/drivers/net/can/esd/Makefile b/drivers/net/can/esd/Makefile new file mode 100644 index 000000000000..5dd2d470c286 --- /dev/null +++ b/drivers/net/can/esd/Makefile @@ -0,0 +1,7 @@ +# SPDX-License-Identifier: GPL-2.0-only +# +# Makefile for esd gmbh ESDACC controller driver +# +esd_402_pci-objs := esdacc.o esd_402_pci-core.o + +obj-$(CONFIG_CAN_ESD_402_PCI) += esd_402_pci.o diff --git a/drivers/net/can/esd/esd_402_pci-core.c b/drivers/net/can/esd/esd_402_pci-core.c new file mode 100644 index 000000000000..80d816a78859 --- /dev/null +++ b/drivers/net/can/esd/esd_402_pci-core.c @@ -0,0 +1,502 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh + */ + +#include <linux/kernel.h> +#include <linux/module.h> +#include <linux/interrupt.h> +#include <linux/pci.h> +#include <linux/io.h> +#include <linux/delay.h> +#include <linux/netdevice.h> +#include <linux/dma-mapping.h> +#include <linux/can.h> +#include <linux/can/dev.h> +#include <linux/can/netlink.h> + +#include "esdacc.h" + +#define DRV_NAME "esd_402_pci" + +#define ESD_PCI_DEVICE_ID_PCIE402 0x0402 + +#define PCI402_FPGA_VER_MIN 0x003d +#define PCI402_MAX_CORES 6 +#define PCI402_BAR 0 +#define PCI402_IO_OV_OFFS 0 +#define PCI402_IO_PCIEP_OFFS 0x10000 +#define PCI402_IO_LEN_TOTAL 0x20000 +#define PCI402_IO_LEN_CORE 0x2000 +#define PCI402_PCICFG_MSICAP_CSR 0x52 +#define PCI402_PCICFG_MSICAP_ADDR 0x54 +#define PCI402_PCICFG_MSICAP_DATA 0x5c + +#define PCI402_DMA_MASK DMA_BIT_MASK(32) +#define PCI402_DMA_SIZE ALIGN(0x10000, PAGE_SIZE) + +#define PCI402_PCIEP_OF_INT_ENABLE 0x0050 +#define PCI402_PCIEP_OF_BM_ADDR_LO 0x1000 +#define PCI402_PCIEP_OF_BM_ADDR_HI 0x1004 +#define PCI402_PCIEP_OF_MSI_ADDR_LO 0x1008 +#define PCI402_PCIEP_OF_MSI_ADDR_HI 0x100c + +/* The BTR register capabilities described by the can_bittiming_const structures + * below are valid since ESDACC version 0x0032. + */ + +/* Used if the ESDACC FPGA is built as CAN-Classic version. */ +static const struct can_bittiming_const pci402_bittiming_const = { + .name = "esd_402", + .tseg1_min = 1, + .tseg1_max = 16, + .tseg2_min = 1, + .tseg2_max = 8, + .sjw_max = 4, + .brp_min = 1, + .brp_max = 512, + .brp_inc = 1, +}; + +/* Used if the ESDACC FPGA is built as CAN-FD version. */ +static const struct can_bittiming_const pci402_bittiming_const_canfd = { + .name = "esd_402fd", + .tseg1_min = 1, + .tseg1_max = 256, + .tseg2_min = 1, + .tseg2_max = 128, + .sjw_max = 128, + .brp_min = 1, + .brp_max = 256, + .brp_inc = 1, +}; + +static const struct net_device_ops pci402_acc_netdev_ops = { + .ndo_open = acc_open, + .ndo_stop = acc_close, + .ndo_start_xmit = acc_start_xmit, + .ndo_change_mtu = can_change_mtu +}; + +struct pci402_card { + /* Actually mapped io space, all other iomem derived from this */ + void __iomem *addr; + void __iomem *addr_pciep; + + void *dma_buf; + dma_addr_t dma_hnd; + + struct acc_ov ov; + struct acc_core *cores; + + bool msi_enabled; +}; + +static irqreturn_t pci402_interrupt(int irq, void *dev_id) +{ + struct pci_dev *pdev = dev_id; + struct pci402_card *card = pci_get_drvdata(pdev); + irqreturn_t irq_status; + + irq_status = acc_card_interrupt(&card->ov, card->cores); + + return irq_status; +} + +static int pci402_set_msiconfig(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + u32 addr_lo_offs = 0; + u32 addr_lo = 0; + u32 addr_hi = 0; + u32 data = 0; + u16 csr = 0; + int err; + + err = pci_read_config_word(pdev, PCI402_PCICFG_MSICAP_CSR, &csr); + if (err) + goto failed; + + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_ADDR, &addr_lo); + if (err) + goto failed; + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_ADDR + 4, + &addr_hi); + if (err) + goto failed; + + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_DATA, &data); + if (err) + goto failed; + + addr_lo_offs = addr_lo & 0x0000ffff; + addr_lo &= 0xffff0000; + + if (addr_hi) + addr_lo |= 1; /* Enable 64-Bit addressing in address space */ + + if (!(csr & 0x0001)) { /* Enable bit */ + err = -EINVAL; + goto failed; + } + + iowrite32(addr_lo, card->addr_pciep + PCI402_PCIEP_OF_MSI_ADDR_LO); + iowrite32(addr_hi, card->addr_pciep + PCI402_PCIEP_OF_MSI_ADDR_HI); + acc_ov_write32(&card->ov, ACC_OV_OF_MSI_ADDRESSOFFSET, addr_lo_offs); + acc_ov_write32(&card->ov, ACC_OV_OF_MSI_DATA, data); + + return 0; + +failed: + pci_warn(pdev, "Error while setting MSI configuration:\n" + "CSR: 0x%.4x, addr: 0x%.8x%.8x, data: 0x%.8x\n", + csr, addr_hi, addr_lo, data); + + return err; +} + +static int pci402_init_card(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + + card->ov.addr = card->addr + PCI402_IO_OV_OFFS; + card->addr_pciep = card->addr + PCI402_IO_PCIEP_OFFS; + + acc_reset_fpga(&card->ov); + acc_init_ov(&card->ov, &pdev->dev); + + if (card->ov.version < PCI402_FPGA_VER_MIN) { + pci_err(pdev, + "ESDACC version (0x%.4x) outdated, please update\n", + card->ov.version); + return -EINVAL; + } + + if (card->ov.active_cores > PCI402_MAX_CORES) { + pci_warn(pdev, + "Card has more active cores than supported by driver, %u core(s) will be ignored\n", + card->ov.active_cores - PCI402_MAX_CORES); + card->ov.active_cores = PCI402_MAX_CORES; + } + card->cores = devm_kcalloc(&pdev->dev, card->ov.active_cores, + sizeof(struct acc_core), GFP_KERNEL); + if (!card->cores) + return -ENOMEM; + + if (card->ov.features & ACC_OV_REG_FEAT_MASK_CANFD) { + pci_warn(pdev, + "ESDACC with CAN-FD feature detected. This driver doesn't support CAN-FD yet.\n"); + } + +#ifdef __LITTLE_ENDIAN + /* So card converts all busmastered data to LE for us: */ + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, + ACC_OV_REG_MODE_MASK_ENDIAN_LITTLE); +#endif + + return 0; +} + +static int pci402_init_interrupt(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + int err; + + err = pci_enable_msi(pdev); + if (!err) { + err = pci402_set_msiconfig(pdev); + if (!err) { + card->msi_enabled = true; + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, + ACC_OV_REG_MODE_MASK_MSI_ENABLE); + pci_info(pdev, "MSI enabled\n"); + } + } + + err = devm_request_irq(&pdev->dev, pdev->irq, pci402_interrupt, + IRQF_SHARED, dev_name(&pdev->dev), pdev); + if (err) + goto failure_msidis; + + iowrite32(1, card->addr_pciep + PCI402_PCIEP_OF_INT_ENABLE); + + return 0; + +failure_msidis: + if (card->msi_enabled) { + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, + ACC_OV_REG_MODE_MASK_MSI_ENABLE); + pci_disable_msi(pdev); + card->msi_enabled = false; + } + + return err; +} + +static void pci402_finish_interrupt(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_INT_ENABLE); + devm_free_irq(&pdev->dev, pdev->irq, pdev); + + if (card->msi_enabled) { + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, + ACC_OV_REG_MODE_MASK_MSI_ENABLE); + pci_disable_msi(pdev); + card->msi_enabled = false; + } +} + +static int pci402_init_dma(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + int err; + + err = dma_set_coherent_mask(&pdev->dev, PCI402_DMA_MASK); + if (err) { + pci_err(pdev, "DMA set mask failed!\n"); + return err; + } + + /* The ESDACC DMA engine needs the DMA buffer aligned to a 64k + * boundary. The DMA API guarantees to align the returned buffer to the + * smallest PAGE_SIZE order which is greater than or equal to the + * requested size. With PCI402_DMA_SIZE == 64kB this suffices here. + */ + card->dma_buf = dma_alloc_coherent(&pdev->dev, PCI402_DMA_SIZE, + &card->dma_hnd, GFP_ATOMIC); + if (!card->dma_buf) { + pci_err(pdev, "DMA alloc failed!\n"); + return -ENOMEM; + } + + acc_init_bm_ptr(&card->ov, card->cores, card->dma_buf); + + iowrite32((u32)card->dma_hnd, + card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_LO); + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_HI); + + pci_set_master(pdev); + + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, + ACC_OV_REG_MODE_MASK_BM_ENABLE); + + return 0; +} + +static void pci402_finish_dma(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + int i; + + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, + ACC_OV_REG_MODE_MASK_BM_ENABLE); + + pci_clear_master(pdev); + + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_LO); + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_HI); + + card->ov.bmfifo.messages = NULL; + card->ov.bmfifo.irq_cnt = NULL; + for (i = 0; i < card->ov.active_cores; i++) { + struct acc_core *core = &card->cores[i]; + + core->bmfifo.messages = NULL; + core->bmfifo.irq_cnt = NULL; + } + + dma_free_coherent(&pdev->dev, PCI402_DMA_SIZE, card->dma_buf, + card->dma_hnd); + card->dma_buf = NULL; +} + +static int pci402_init_cores(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + int err; + int i; + + for (i = 0; i < card->ov.active_cores; i++) { + struct acc_core *core = &card->cores[i]; + struct acc_net_priv *priv; + struct net_device *netdev; + u32 fifo_config; + + core->addr = card->ov.addr + (i + 1) * PCI402_IO_LEN_CORE; + + fifo_config = acc_read32(core, ACC_CORE_OF_TXFIFO_CONFIG); + core->tx_fifo_size = (u8)(fifo_config >> 24); + if (core->tx_fifo_size <= 1) { + pci_err(pdev, "Invalid tx_fifo_size!\n"); + err = -EINVAL; + goto failure; + } + + netdev = alloc_candev(sizeof(*priv), core->tx_fifo_size); + if (!netdev) { + err = -ENOMEM; + goto failure; + } + core->netdev = netdev; + + netdev->flags |= IFF_ECHO; + netdev->dev_port = i; + netdev->netdev_ops = &pci402_acc_netdev_ops; + SET_NETDEV_DEV(netdev, &pdev->dev); + + priv = netdev_priv(netdev); + priv->can.state = CAN_STATE_STOPPED; + priv->can.ctrlmode_supported = CAN_CTRLMODE_LISTENONLY | + CAN_CTRLMODE_BERR_REPORTING | + CAN_CTRLMODE_CC_LEN8_DLC | + CAN_CTRLMODE_LOOPBACK; + priv->can.clock.freq = card->ov.core_frequency; + priv->can.bittiming_const = + (card->ov.features & ACC_OV_REG_FEAT_MASK_CANFD) ? + &pci402_bittiming_const_canfd : + &pci402_bittiming_const; + priv->can.do_set_bittiming = acc_set_bittiming; + priv->can.do_set_mode = acc_set_mode; + priv->can.do_get_berr_counter = acc_get_berr_counter; + + priv->core = core; + priv->ov = &card->ov; + + err = register_candev(netdev); + if (err) { + free_candev(core->netdev); + core->netdev = NULL; + goto failure; + } + + netdev_info(netdev, "registered\n"); + } + + return 0; + +failure: + for (i--; i >= 0; i--) { + struct acc_core *core = &card->cores[i]; + + netdev_info(core->netdev, "unregistering...\n"); + unregister_candev(core->netdev); + + free_candev(core->netdev); + core->netdev = NULL; + } + + return err; +} + +static void pci402_finish_cores(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + int i; + + for (i = 0; i < card->ov.active_cores; i++) { + struct acc_core *core = &card->cores[i]; + + netdev_info(core->netdev, "unregister\n"); + unregister_candev(core->netdev); + + free_candev(core->netdev); + core->netdev = NULL; + } +} + +static int pci402_probe(struct pci_dev *pdev, const struct pci_device_id *ent) +{ + struct pci402_card *card = NULL; + int err; + + err = pci_enable_device(pdev); + if (err) + return err; + + card = devm_kzalloc(&pdev->dev, sizeof(*card), GFP_KERNEL); + if (!card) + goto failure_disable_pci; + + pci_set_drvdata(pdev, card); + + err = pci_request_regions(pdev, pci_name(pdev)); + if (err) + goto failure_disable_pci; + + card->addr = pci_iomap(pdev, PCI402_BAR, PCI402_IO_LEN_TOTAL); + if (!card->addr) { + err = -ENOMEM; + goto failure_release_regions; + } + + err = pci402_init_card(pdev); + if (err) + goto failure_unmap; + + err = pci402_init_dma(pdev); + if (err) + goto failure_unmap; + + err = pci402_init_interrupt(pdev); + if (err) + goto failure_finish_dma; + + err = pci402_init_cores(pdev); + if (err) + goto failure_finish_interrupt; + + return 0; + +failure_finish_interrupt: + pci402_finish_interrupt(pdev); + +failure_finish_dma: + pci402_finish_dma(pdev); + +failure_unmap: + pci_iounmap(pdev, card->addr); + +failure_release_regions: + pci_release_regions(pdev); + +failure_disable_pci: + pci_disable_device(pdev); + + return err; +} + +static void pci402_remove(struct pci_dev *pdev) +{ + struct pci402_card *card = pci_get_drvdata(pdev); + + pci402_finish_interrupt(pdev); + pci402_finish_cores(pdev); + pci402_finish_dma(pdev); + pci_iounmap(pdev, card->addr); + pci_release_regions(pdev); + pci_disable_device(pdev); +} + +static const struct pci_device_id pci402_tbl[] = { + { PCI_VENDOR_ID_ESDGMBH, ESD_PCI_DEVICE_ID_PCIE402, + PCI_VENDOR_ID_ESDGMBH, PCI_ANY_ID, + 0U, 0U, 0UL }, + { 0, } +}; +MODULE_DEVICE_TABLE(pci, pci402_tbl); + +static struct pci_driver pci402_driver = { + .name = DRV_NAME, + .id_table = pci402_tbl, + .probe = pci402_probe, + .remove = pci402_remove, +}; + +module_pci_driver(pci402_driver); + +MODULE_DESCRIPTION("Socket-CAN driver for esd CAN 402 card family with esdACC core on PCIe"); +MODULE_AUTHOR("Thomas Körper <socketcan@esd.eu>"); +MODULE_AUTHOR("Stefan Mätje <stefan.maetje@esd.eu>"); +MODULE_LICENSE("GPL"); diff --git a/drivers/net/can/esd/esdacc.c b/drivers/net/can/esd/esdacc.c new file mode 100644 index 000000000000..13f7397dfc4e --- /dev/null +++ b/drivers/net/can/esd/esdacc.c @@ -0,0 +1,777 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh + */ + +#include <linux/ktime.h> +#include <linux/gcd.h> +#include <linux/io.h> +#include <linux/delay.h> +#include "esdacc.h" + +/* ecc value of esdACC equals SJA1000's ECC register */ +#define ACC_ECC_SEG 0x1f +#define ACC_ECC_DIR 0x20 +#define ACC_ECC_BIT 0x00 +#define ACC_ECC_FORM 0x40 +#define ACC_ECC_STUFF 0x80 +#define ACC_ECC_MASK 0xc0 + +#define ACC_BM_IRQ_UNMASK_ALL 0x55555555U +#define ACC_BM_IRQ_MASK_ALL 0xaaaaaaaaU +#define ACC_BM_IRQ_MASK 0x2U +#define ACC_BM_IRQ_UNMASK 0x1U +#define ACC_BM_LENFLAG_TX 0x20 + +#define ACC_REG_STATUS_IDX_STATUS_DOS 16 +#define ACC_REG_STATUS_IDX_STATUS_ES 17 +#define ACC_REG_STATUS_IDX_STATUS_EP 18 +#define ACC_REG_STATUS_IDX_STATUS_BS 19 +#define ACC_REG_STATUS_IDX_STATUS_RBS 20 +#define ACC_REG_STATUS_IDX_STATUS_RS 21 +#define ACC_REG_STATUS_MASK_STATUS_DOS BIT(ACC_REG_STATUS_IDX_STATUS_DOS) +#define ACC_REG_STATUS_MASK_STATUS_ES BIT(ACC_REG_STATUS_IDX_STATUS_ES) +#define ACC_REG_STATUS_MASK_STATUS_EP BIT(ACC_REG_STATUS_IDX_STATUS_EP) +#define ACC_REG_STATUS_MASK_STATUS_BS BIT(ACC_REG_STATUS_IDX_STATUS_BS) +#define ACC_REG_STATUS_MASK_STATUS_RBS BIT(ACC_REG_STATUS_IDX_STATUS_RBS) +#define ACC_REG_STATUS_MASK_STATUS_RS BIT(ACC_REG_STATUS_IDX_STATUS_RS) + +static void acc_resetmode_enter(struct acc_core *core) +{ + int i; + + acc_set_bits(core, ACC_CORE_OF_CTRL_MODE, + ACC_REG_CONTROL_MASK_MODE_RESETMODE); + + for (i = 0; i < 10; i++) { + if (acc_resetmode_entered(core)) + return; + + udelay(5); + } + + netdev_warn(core->netdev, "Entering reset mode timed out\n"); +} + +static void acc_resetmode_leave(struct acc_core *core) +{ + int i; + + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, + ACC_REG_CONTROL_MASK_MODE_RESETMODE); + + for (i = 0; i < 10; i++) { + if (!acc_resetmode_entered(core)) + return; + + udelay(5); + } + + netdev_warn(core->netdev, "Leaving reset mode timed out\n"); +} + +static void acc_txq_put(struct acc_core *core, u32 acc_id, u8 acc_dlc, + const void *data) +{ + acc_write32_noswap(core, ACC_CORE_OF_TXFIFO_DATA_1, + *((const u32 *)(data + 4))); + acc_write32_noswap(core, ACC_CORE_OF_TXFIFO_DATA_0, + *((const u32 *)data)); + acc_write32(core, ACC_CORE_OF_TXFIFO_DLC, acc_dlc); + /* CAN id must be written at last. This write starts TX. */ + acc_write32(core, ACC_CORE_OF_TXFIFO_ID, acc_id); +} + +/* Prepare conversion factor from ESDACC time stamp ticks to ns + * + * The conversion factor ts2ns from time stamp counts to ns is basically + * ts2ns = NSEC_PER_SEC / timestamp_frequency + * + * To avoid an overflow, the ts2ns fraction is truncated with its gcd and + * only the truncated numerator and denominator are used further. + */ +static void acc_init_ov_ts2ns(struct acc_ov *ov) +{ + u32 ts2ns_gcd = (u32)gcd(NSEC_PER_SEC, ov->timestamp_frequency); + + ov->ts2ns_numerator = (u32)NSEC_PER_SEC / ts2ns_gcd; + ov->ts2ns_denominator = ov->timestamp_frequency / ts2ns_gcd; +} + +static ktime_t acc_ts2ktime(struct acc_ov *ov, u64 ts) +{ + u64 ns; + + ts = ts * ov->ts2ns_numerator; + ns = div_u64(ts, ov->ts2ns_denominator); + + return ns_to_ktime(ns); +} + +void acc_init_ov(struct acc_ov *ov, struct device *dev) +{ + u32 temp; + /* For the justification of this see comment on struct acc_bmmsg* + * in esdacc.h. + */ + BUILD_BUG_ON(sizeof(struct acc_bmmsg) != ACC_CORE_DMAMSG_SIZE); + + temp = acc_ov_read32(ov, ACC_OV_OF_VERSION); + ov->version = (u16)temp; + ov->features = (u16)(temp >> 16); + + temp = acc_ov_read32(ov, ACC_OV_OF_INFO); + ov->total_cores = (u8)temp; + ov->active_cores = (u8)(temp >> 8); + + ov->core_frequency = acc_ov_read32(ov, ACC_OV_OF_CANCORE_FREQ); + ov->timestamp_frequency = acc_ov_read32(ov, ACC_OV_OF_TS_FREQ_LO); + acc_init_ov_ts2ns(ov); + + /* Depending on ESDACC feature NEW_PSC enable the new prescaler + * or adjust core_frequency according to the implicit division by 2. + */ + if (ov->features & ACC_OV_REG_FEAT_MASK_NEW_PSC) { + acc_ov_set_bits(ov, ACC_OV_OF_MODE, + ACC_OV_REG_MODE_MASK_NEW_PSC_ENABLE); + } else { + ov->core_frequency /= 2; + } + + dev_info(dev, + "ESDACC v%u, freq: %u/%u, feat/strap: 0x%x/0x%x, cores: %u/%u\n", + ov->version, ov->core_frequency, ov->timestamp_frequency, + ov->features, acc_ov_read32(ov, ACC_OV_OF_INFO) >> 16, + ov->active_cores, ov->total_cores); + dev_dbg(dev, "ESDACC ts2ns: numerator %u, denominator %u\n", + ov->ts2ns_numerator, ov->ts2ns_denominator); +} + +void acc_init_bm_ptr(struct acc_ov *ov, struct acc_core *cores, const void *mem) +{ + unsigned int u; + + /* DMA buffer layout as follows where N is the number of CAN cores + * implemented in the FPGA, i.e. N = ov->total_cores + * + * Layout Section size + * +-----------------------+ + * | FIFO Card/Overview | ACC_CORE_DMABUF_SIZE + * | | + * +-----------------------+ + * | FIFO Core0 | ACC_CORE_DMABUF_SIZE + * | | + * +-----------------------+ + * | ... | ... + * | | + * +-----------------------+ + * | FIFO CoreN | ACC_CORE_DMABUF_SIZE + * | | + * +-----------------------+ + * | irq_cnt Card/Overview | sizeof(u32) + * +-----------------------+ + * | irq_cnt Core0 | sizeof(u32) + * +-----------------------+ + * | ... | ... + * +-----------------------+ + * | irq_cnt CoreN | sizeof(u32) + * +-----------------------+ + */ + ov->bmfifo.messages = mem; + ov->bmfifo.irq_cnt = mem + (ov->total_cores + 1U) * ACC_CORE_DMABUF_SIZE; + + for (u = 0U; u < ov->active_cores; u++) { + struct acc_core *core = &cores[u]; + + core->bmfifo.messages = mem + (u + 1U) * ACC_CORE_DMABUF_SIZE; + core->bmfifo.irq_cnt = ov->bmfifo.irq_cnt + (u + 1U); + } +} + +int acc_open(struct net_device *netdev) +{ + struct acc_net_priv *priv = netdev_priv(netdev); + struct acc_core *core = priv->core; + u32 ctrl_mode; + int err; + + /* Retry to enter RESET mode if out of sync. */ + if (priv->can.state != CAN_STATE_STOPPED) { + netdev_warn(netdev, "Entered %s() with bad can.state: %s\n", + __func__, can_get_state_str(priv->can.state)); + acc_resetmode_enter(core); + if (acc_resetmode_entered(core)) + priv->can.state = CAN_STATE_STOPPED; + } + + err = open_candev(netdev); + if (err) + return err; + + ctrl_mode = ACC_REG_CONTROL_MASK_IE_RXTX | + ACC_REG_CONTROL_MASK_IE_TXERROR | + ACC_REG_CONTROL_MASK_IE_ERRWARN | + ACC_REG_CONTROL_MASK_IE_OVERRUN | + ACC_REG_CONTROL_MASK_IE_ERRPASS; + + if (priv->can.ctrlmode & CAN_CTRLMODE_BERR_REPORTING) + ctrl_mode |= ACC_REG_CONTROL_MASK_IE_BUSERR; + + if (priv->can.ctrlmode & CAN_CTRLMODE_LISTENONLY) + ctrl_mode |= ACC_REG_CONTROL_MASK_MODE_LOM; + + acc_set_bits(core, ACC_CORE_OF_CTRL_MODE, ctrl_mode); + + acc_resetmode_leave(core); + if (!acc_resetmode_entered(core)) + priv->can.state = CAN_STATE_ERROR_ACTIVE; + /* Resync TX FIFO indices to HW state after (re-)start. */ + { + u32 tx_fifo_status = acc_read32(core, ACC_CORE_OF_TXFIFO_STATUS); + + core->tx_fifo_head = tx_fifo_status & 0xff; + core->tx_fifo_tail = (tx_fifo_status >> 8) & 0xff; + } + netif_start_queue(netdev); + return 0; +} + +int acc_close(struct net_device *netdev) +{ + struct acc_net_priv *priv = netdev_priv(netdev); + struct acc_core *core = priv->core; + + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, + ACC_REG_CONTROL_MASK_IE_RXTX | + ACC_REG_CONTROL_MASK_IE_TXERROR | + ACC_REG_CONTROL_MASK_IE_ERRWARN | + ACC_REG_CONTROL_MASK_IE_OVERRUN | + ACC_REG_CONTROL_MASK_IE_ERRPASS | + ACC_REG_CONTROL_MASK_IE_BUSERR); + + netif_stop_queue(netdev); + acc_resetmode_enter(core); + if (acc_resetmode_entered(core)) + priv->can.state = CAN_STATE_STOPPED; + + /* Mark pending TX requests to be aborted after controller restart. */ + acc_write32(core, ACC_CORE_OF_TX_ABORT_MASK, 0xffff); + + /* ACC_REG_CONTROL_MASK_MODE_LOM is only accessible in RESET mode */ + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, + ACC_REG_CONTROL_MASK_MODE_LOM); + + close_candev(netdev); + return 0; +} + +netdev_tx_t acc_start_xmit(struct sk_buff *skb, struct net_device *netdev) +{ + struct acc_net_priv *priv = netdev_priv(netdev); + struct acc_core *core = priv->core; + struct can_frame *cf = (struct can_frame *)skb->data; + u8 tx_fifo_head = core->tx_fifo_head; + int fifo_usage; + u32 acc_id; + u8 acc_dlc; + + /* Access core->tx_fifo_tail only once because it may be changed + * from the interrupt level. + */ + fifo_usage = tx_fifo_head - core->tx_fifo_tail; + if (fifo_usage < 0) + fifo_usage += core->tx_fifo_size; + + if (fifo_usage >= core->tx_fifo_size - 1) { + netdev_err(core->netdev, + "BUG: TX ring full when queue awake!\n"); + netif_stop_queue(netdev); + return NETDEV_TX_BUSY; + } + + if (fifo_usage == core->tx_fifo_size - 2) + netif_stop_queue(netdev); + + acc_dlc = can_get_cc_dlc(cf, priv->can.ctrlmode); + if (cf->can_id & CAN_RTR_FLAG) + acc_dlc |= ACC_CAN_RTR_FLAG; + + if (cf->can_id & CAN_EFF_FLAG) { + acc_id = cf->can_id & CAN_EFF_MASK; + acc_id |= ACC_CAN_EFF_FLAG; + } else { + acc_id = cf->can_id & CAN_SFF_MASK; + } + + can_put_echo_skb(skb, netdev, core->tx_fifo_head, 0); + + tx_fifo_head++; + if (tx_fifo_head >= core->tx_fifo_size) + tx_fifo_head = 0U; + core->tx_fifo_head = tx_fifo_head; + + acc_txq_put(core, acc_id, acc_dlc, cf->data); + + return NETDEV_TX_OK; +} + +int acc_get_berr_counter(const struct net_device *netdev, + struct can_berr_counter *bec) +{ + struct acc_net_priv *priv = netdev_priv(netdev); + u32 core_status = acc_read32(priv->core, ACC_CORE_OF_STATUS); + + bec->txerr = (core_status >> 8) & 0xff; + bec->rxerr = core_status & 0xff; + + return 0; +} + +int acc_set_mode(struct net_device *netdev, enum can_mode mode) +{ + struct acc_net_priv *priv = netdev_priv(netdev); + + switch (mode) { + case CAN_MODE_START: + /* Paranoid FIFO index check. */ + { + const u32 tx_fifo_status = + acc_read32(priv->core, ACC_CORE_OF_TXFIFO_STATUS); + const u8 hw_fifo_head = (u8)tx_fifo_status; + + if (hw_fifo_head != priv->core->tx_fifo_head || + hw_fifo_head != priv->core->tx_fifo_tail) { + netdev_warn(netdev, + "TX FIFO mismatch: T %2u H %2u; TFHW %#08x\n", + priv->core->tx_fifo_tail, + priv->core->tx_fifo_head, + tx_fifo_status); + } + } + acc_resetmode_leave(priv->core); + /* To leave the bus-off state the esdACC controller begins + * here a grace period where it counts 128 "idle conditions" (each + * of 11 consecutive recessive bits) on the bus as required + * by the CAN spec. + * + * During this time the TX FIFO may still contain already + * aborted "zombie" frames that are only drained from the FIFO + * at the end of the grace period. + * + * To not to interfere with this drain process we don't + * call netif_wake_queue() here. When the controller reaches + * the error-active state again, it informs us about that + * with an acc_bmmsg_errstatechange message. Then + * netif_wake_queue() is called from + * handle_core_msg_errstatechange() instead. + */ + break; + + default: + return -EOPNOTSUPP; + } + + return 0; +} + +int acc_set_bittiming(struct net_device *netdev) +{ + struct acc_net_priv *priv = netdev_priv(netdev); + const struct can_bittiming *bt = &priv->can.bittiming; + u32 brp = bt->brp - 1; + u32 btr; + + if (priv->ov->features & ACC_OV_REG_FEAT_MASK_CANFD) { + u32 fbtr = 0; + + netdev_dbg(priv->core->netdev, + "bit timing: brp %u, prop %u, ph1 %u ph2 %u, sjw %u\n", + bt->brp, bt->prop_seg, + bt->phase_seg1, bt->phase_seg2, bt->sjw); + + /* BRP: 8 bits @ bits 7..0 */ + brp &= 0xff; + + /* TSEG1: 8 bits @ bits 7..0 */ + btr = (bt->phase_seg1 + bt->prop_seg - 1) & 0xff; + /* TSEG2: 7 bits @ bits 22..16 */ + btr |= ((bt->phase_seg2 - 1) & 0x7f) << 16; + /* SJW: 7 bits @ bits 30..24 */ + btr |= ((bt->sjw - 1) & 0x7f) << 24; + + /* Keep order of accesses to ACC_CORE_OF_BRP and ACC_CORE_OF_BTR. */ + acc_write32(priv->core, ACC_CORE_OF_BRP, brp); + acc_write32(priv->core, ACC_CORE_OF_BTR, btr); + + netdev_info(priv->core->netdev, + "ESDACC: BRP %u, NBTR 0x%08x, DBTR 0x%08x", + brp, btr, fbtr); + } else { + netdev_dbg(priv->core->netdev, + "bit timing: brp %u, prop %u, ph1 %u ph2 %u, sjw %u\n", + bt->brp, bt->prop_seg, + bt->phase_seg1, bt->phase_seg2, bt->sjw); + + /* BRP: 9 bits @ bits 8..0 */ + brp &= 0x1ff; + + /* TSEG1: 4 bits @ bits 3..0 */ + btr = (bt->phase_seg1 + bt->prop_seg - 1) & 0xf; + /* TSEG2: 3 bits @ bits 18..16*/ + btr |= ((bt->phase_seg2 - 1) & 0x7) << 16; + /* SJW: 2 bits @ bits 25..24 */ + btr |= ((bt->sjw - 1) & 0x3) << 24; + + /* Keep order of accesses to ACC_CORE_OF_BRP and ACC_CORE_OF_BTR. */ + acc_write32(priv->core, ACC_CORE_OF_BRP, brp); + acc_write32(priv->core, ACC_CORE_OF_BTR, btr); + + netdev_info(priv->core->netdev, "ESDACC: BRP %u, BTR 0x%08x", + brp, btr); + } + + return 0; +} + +static void handle_core_msg_rxtxdone(struct acc_core *core, + const struct acc_bmmsg_rxtxdone *msg) +{ + struct acc_net_priv *priv = netdev_priv(core->netdev); + struct net_device_stats *stats = &core->netdev->stats; + struct sk_buff *skb; + + if (msg->dlc.rxtx.len & ACC_BM_LENFLAG_TX) { + u8 tx_fifo_tail = core->tx_fifo_tail; + + if (core->tx_fifo_head == tx_fifo_tail) { + netdev_warn(core->netdev, + "TX interrupt, but queue is empty!?\n"); + return; + } + + /* Direct access echo skb to attach HW time stamp. */ + skb = priv->can.echo_skb[tx_fifo_tail]; + if (skb) { + skb_hwtstamps(skb)->hwtstamp = + acc_ts2ktime(priv->ov, msg->ts); + } + + stats->tx_packets++; + stats->tx_bytes += can_get_echo_skb(core->netdev, tx_fifo_tail, + NULL); + + tx_fifo_tail++; + if (tx_fifo_tail >= core->tx_fifo_size) + tx_fifo_tail = 0U; + core->tx_fifo_tail = tx_fifo_tail; + + netif_wake_queue(core->netdev); + + } else { + struct can_frame *cf; + + skb = alloc_can_skb(core->netdev, &cf); + if (!skb) { + stats->rx_dropped++; + return; + } + + cf->can_id = msg->id & CAN_EFF_MASK; + if (msg->id & ACC_CAN_EFF_FLAG) + cf->can_id |= CAN_EFF_FLAG; + + can_frame_set_cc_len(cf, msg->dlc.rx.len & ACC_CAN_DLC_MASK, + priv->can.ctrlmode); + + if (msg->dlc.rx.len & ACC_CAN_RTR_FLAG) + cf->can_id |= CAN_RTR_FLAG; + else + memcpy(cf->data, msg->data, cf->len); + + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); + + stats->rx_packets++; + stats->rx_bytes += cf->len; + netif_rx(skb); + } +} + +static void handle_core_msg_txabort(struct acc_core *core, + const struct acc_bmmsg_txabort *msg) +{ + struct net_device_stats *stats = &core->netdev->stats; + u8 tx_fifo_tail = core->tx_fifo_tail; + u32 abort_mask = msg->abort_mask; /* u32 extend to avoid warnings later */ + + /* The abort_mask shows which frames were aborted in ESDACC's FIFO. */ + while (tx_fifo_tail != core->tx_fifo_head && (abort_mask)) { + const u32 tail_mask = (1U << tx_fifo_tail); + + if (!(abort_mask & tail_mask)) + break; + abort_mask &= ~tail_mask; + + can_free_echo_skb(core->netdev, tx_fifo_tail, NULL); + stats->tx_dropped++; + stats->tx_aborted_errors++; + + tx_fifo_tail++; + if (tx_fifo_tail >= core->tx_fifo_size) + tx_fifo_tail = 0; + } + core->tx_fifo_tail = tx_fifo_tail; + if (abort_mask) + netdev_warn(core->netdev, "Unhandled aborted messages\n"); + + if (!acc_resetmode_entered(core)) + netif_wake_queue(core->netdev); +} + +static void handle_core_msg_overrun(struct acc_core *core, + const struct acc_bmmsg_overrun *msg) +{ + struct acc_net_priv *priv = netdev_priv(core->netdev); + struct net_device_stats *stats = &core->netdev->stats; + struct can_frame *cf; + struct sk_buff *skb; + + skb = alloc_can_err_skb(core->netdev, &cf); + if (!skb) { + stats->rx_dropped++; + return; + } + + /* lost_cnt may be 0 if not supported by ESDACC version */ + if (msg->lost_cnt) { + stats->rx_errors += msg->lost_cnt; + stats->rx_over_errors += msg->lost_cnt; + } else { + stats->rx_errors++; + stats->rx_over_errors++; + } + + cf->can_id |= CAN_ERR_CRTL; + cf->data[1] = CAN_ERR_CRTL_RX_OVERFLOW; + + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); + + stats->rx_packets++; + stats->rx_bytes += cf->len; + netif_rx(skb); +} + +static void handle_core_msg_buserr(struct acc_core *core, + const struct acc_bmmsg_buserr *msg) +{ + struct acc_net_priv *priv = netdev_priv(core->netdev); + struct net_device_stats *stats = &core->netdev->stats; + struct can_frame *cf; + struct sk_buff *skb; + const u32 reg_status = msg->reg_status; + const u8 rxerr = (u8)reg_status; + const u8 txerr = (u8)(reg_status >> 8); + + priv->can.can_stats.bus_error++; + + skb = alloc_can_err_skb(core->netdev, &cf); + if (!skb) { + stats->rx_dropped++; + return; + } + + cf->can_id |= CAN_ERR_PROT | CAN_ERR_BUSERROR; + + /* msg->ecc acts like SJA1000's ECC register */ + switch (msg->ecc & ACC_ECC_MASK) { + case ACC_ECC_BIT: + cf->data[2] |= CAN_ERR_PROT_BIT; + break; + case ACC_ECC_FORM: + cf->data[2] |= CAN_ERR_PROT_FORM; + break; + case ACC_ECC_STUFF: + cf->data[2] |= CAN_ERR_PROT_STUFF; + break; + default: + cf->data[2] |= CAN_ERR_PROT_UNSPEC; + break; + } + + /* Set error location */ + cf->data[3] = msg->ecc & ACC_ECC_SEG; + + /* Error occurred during transmission? */ + if ((msg->ecc & ACC_ECC_DIR) == 0) { + cf->data[2] |= CAN_ERR_PROT_TX; + stats->tx_errors++; + } else { + stats->rx_errors++; + } + /* Insert CAN TX and RX error counters. */ + cf->data[6] = txerr; + cf->data[7] = rxerr; + + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); + + stats->rx_packets++; + stats->rx_bytes += cf->len; + netif_rx(skb); +} + +static void +handle_core_msg_errstatechange(struct acc_core *core, + const struct acc_bmmsg_errstatechange *msg) +{ + struct acc_net_priv *priv = netdev_priv(core->netdev); + struct net_device_stats *stats = &core->netdev->stats; + struct can_frame *cf = NULL; + struct sk_buff *skb; + const u32 reg_status = msg->reg_status; + const u8 rxerr = (u8)reg_status; + const u8 txerr = (u8)(reg_status >> 8); + enum can_state new_state; + + if (reg_status & ACC_REG_STATUS_MASK_STATUS_BS) { + new_state = CAN_STATE_BUS_OFF; + } else if (reg_status & ACC_REG_STATUS_MASK_STATUS_EP) { + new_state = CAN_STATE_ERROR_PASSIVE; + } else if (reg_status & ACC_REG_STATUS_MASK_STATUS_ES) { + new_state = CAN_STATE_ERROR_WARNING; + } else { + new_state = CAN_STATE_ERROR_ACTIVE; + if (priv->can.state == CAN_STATE_BUS_OFF) { + /* See comment in acc_set_mode() for CAN_MODE_START */ + netif_wake_queue(core->netdev); + } + } + + skb = alloc_can_err_skb(core->netdev, &cf); + + if (new_state != priv->can.state) { + enum can_state tx_state, rx_state; + + tx_state = (txerr >= rxerr) ? + new_state : CAN_STATE_ERROR_ACTIVE; + rx_state = (rxerr >= txerr) ? + new_state : CAN_STATE_ERROR_ACTIVE; + + /* Always call can_change_state() to update the state + * even if alloc_can_err_skb() may have failed. + * can_change_state() can cope with a NULL cf pointer. + */ + can_change_state(core->netdev, cf, tx_state, rx_state); + } + + if (skb) { + cf->data[6] = txerr; + cf->data[7] = rxerr; + + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); + + stats->rx_packets++; + stats->rx_bytes += cf->len; + netif_rx(skb); + } else { + stats->rx_dropped++; + } + + if (new_state == CAN_STATE_BUS_OFF) { + acc_write32(core, ACC_CORE_OF_TX_ABORT_MASK, 0xffff); + can_bus_off(core->netdev); + } +} + +static void handle_core_interrupt(struct acc_core *core) +{ + u32 msg_fifo_head = core->bmfifo.local_irq_cnt & 0xff; + + while (core->bmfifo.msg_fifo_tail != msg_fifo_head) { + const struct acc_bmmsg *msg = + &core->bmfifo.messages[core->bmfifo.msg_fifo_tail]; + + switch (msg->u.msg_id) { + case BM_MSG_ID_RXTXDONE: + handle_core_msg_rxtxdone(core, &msg->u.rxtxdone); + break; + + case BM_MSG_ID_TXABORT: + handle_core_msg_txabort(core, &msg->u.txabort); + break; + + case BM_MSG_ID_OVERRUN: + handle_core_msg_overrun(core, &msg->u.overrun); + break; + + case BM_MSG_ID_BUSERR: + handle_core_msg_buserr(core, &msg->u.buserr); + break; + + case BM_MSG_ID_ERRPASSIVE: + case BM_MSG_ID_ERRWARN: + handle_core_msg_errstatechange(core, + &msg->u.errstatechange); + break; + + default: + /* Ignore all other BM messages (like the CAN-FD messages) */ + break; + } + + core->bmfifo.msg_fifo_tail = + (core->bmfifo.msg_fifo_tail + 1) & 0xff; + } +} + +irqreturn_t acc_card_interrupt(struct acc_ov *ov, struct acc_core *cores) +{ + u32 irqmask; + int i; + + /* First we look for whom interrupts are pending, card/overview + * or any of the cores. Two bits in irqmask are used for each; + * set to ACC_BM_IRQ_MASK then: + */ + irqmask = 0; + if (*ov->bmfifo.irq_cnt != ov->bmfifo.local_irq_cnt) { + irqmask |= ACC_BM_IRQ_MASK; + ov->bmfifo.local_irq_cnt = *ov->bmfifo.irq_cnt; + } + + for (i = 0; i < ov->active_cores; i++) { + struct acc_core *core = &cores[i]; + + if (*core->bmfifo.irq_cnt != core->bmfifo.local_irq_cnt) { + irqmask |= (ACC_BM_IRQ_MASK << (2 * (i + 1))); + core->bmfifo.local_irq_cnt = *core->bmfifo.irq_cnt; + } + } + + if (!irqmask) + return IRQ_NONE; + + /* At second we tell the card we're working on them by writing irqmask, + * call handle_{ov|core}_interrupt and then acknowledge the + * interrupts by writing irq_cnt: + */ + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_MASK, irqmask); + + if (irqmask & ACC_BM_IRQ_MASK) { + /* handle_ov_interrupt(); - no use yet. */ + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_COUNTER, + ov->bmfifo.local_irq_cnt); + } + + for (i = 0; i < ov->active_cores; i++) { + struct acc_core *core = &cores[i]; + + if (irqmask & (ACC_BM_IRQ_MASK << (2 * (i + 1)))) { + handle_core_interrupt(core); + acc_write32(core, ACC_OV_OF_BM_IRQ_COUNTER, + core->bmfifo.local_irq_cnt); + } + } + + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_MASK, ACC_BM_IRQ_UNMASK_ALL); + + return IRQ_HANDLED; +} diff --git a/drivers/net/can/esd/esdacc.h b/drivers/net/can/esd/esdacc.h new file mode 100644 index 000000000000..3e865ececb3e --- /dev/null +++ b/drivers/net/can/esd/esdacc.h @@ -0,0 +1,380 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh + */ + +#include <linux/kernel.h> +#include <linux/netdevice.h> +#include <linux/can/dev.h> + +#define ACC_CAN_EFF_FLAG 0x20000000 +#define ACC_CAN_RTR_FLAG 0x10 +#define ACC_CAN_DLC_MASK 0x0f + +#define ACC_OV_OF_PROBE 0x0000 +#define ACC_OV_OF_VERSION 0x0004 +#define ACC_OV_OF_INFO 0x0008 +#define ACC_OV_OF_CANCORE_FREQ 0x000c +#define ACC_OV_OF_TS_FREQ_LO 0x0010 +#define ACC_OV_OF_TS_FREQ_HI 0x0014 +#define ACC_OV_OF_IRQ_STATUS_CORES 0x0018 +#define ACC_OV_OF_TS_CURR_LO 0x001c +#define ACC_OV_OF_TS_CURR_HI 0x0020 +#define ACC_OV_OF_IRQ_STATUS 0x0028 +#define ACC_OV_OF_MODE 0x002c +#define ACC_OV_OF_BM_IRQ_COUNTER 0x0070 +#define ACC_OV_OF_BM_IRQ_MASK 0x0074 +#define ACC_OV_OF_MSI_DATA 0x0080 +#define ACC_OV_OF_MSI_ADDRESSOFFSET 0x0084 + +/* Feature flags are contained in the upper 16 bit of the version + * register at ACC_OV_OF_VERSION but only used with these masks after + * extraction into an extra variable => (xx - 16). + */ +#define ACC_OV_REG_FEAT_IDX_CANFD (27 - 16) +#define ACC_OV_REG_FEAT_IDX_NEW_PSC (28 - 16) +#define ACC_OV_REG_FEAT_MASK_CANFD BIT(ACC_OV_REG_FEAT_IDX_CANFD) +#define ACC_OV_REG_FEAT_MASK_NEW_PSC BIT(ACC_OV_REG_FEAT_IDX_NEW_PSC) + +#define ACC_OV_REG_MODE_MASK_ENDIAN_LITTLE 0x00000001 +#define ACC_OV_REG_MODE_MASK_BM_ENABLE 0x00000002 +#define ACC_OV_REG_MODE_MASK_MODE_LED 0x00000004 +#define ACC_OV_REG_MODE_MASK_TIMER 0x00000070 +#define ACC_OV_REG_MODE_MASK_TIMER_ENABLE 0x00000010 +#define ACC_OV_REG_MODE_MASK_TIMER_ONE_SHOT 0x00000020 +#define ACC_OV_REG_MODE_MASK_TIMER_ABSOLUTE 0x00000040 +#define ACC_OV_REG_MODE_MASK_TS_SRC 0x00000180 +#define ACC_OV_REG_MODE_MASK_I2C_ENABLE 0x00000800 +#define ACC_OV_REG_MODE_MASK_MSI_ENABLE 0x00004000 +#define ACC_OV_REG_MODE_MASK_NEW_PSC_ENABLE 0x00008000 +#define ACC_OV_REG_MODE_MASK_FPGA_RESET 0x80000000 + +#define ACC_CORE_OF_CTRL_MODE 0x0000 +#define ACC_CORE_OF_STATUS_IRQ 0x0008 +#define ACC_CORE_OF_BRP 0x000c +#define ACC_CORE_OF_BTR 0x0010 +#define ACC_CORE_OF_FBTR 0x0014 +#define ACC_CORE_OF_STATUS 0x0030 +#define ACC_CORE_OF_TXFIFO_CONFIG 0x0048 +#define ACC_CORE_OF_TXFIFO_STATUS 0x004c +#define ACC_CORE_OF_TX_STATUS_IRQ 0x0050 +#define ACC_CORE_OF_TX_ABORT_MASK 0x0054 +#define ACC_CORE_OF_BM_IRQ_COUNTER 0x0070 +#define ACC_CORE_OF_TXFIFO_ID 0x00c0 +#define ACC_CORE_OF_TXFIFO_DLC 0x00c4 +#define ACC_CORE_OF_TXFIFO_DATA_0 0x00c8 +#define ACC_CORE_OF_TXFIFO_DATA_1 0x00cc + +#define ACC_REG_CONTROL_IDX_MODE_RESETMODE 0 +#define ACC_REG_CONTROL_IDX_MODE_LOM 1 +#define ACC_REG_CONTROL_IDX_MODE_STM 2 +#define ACC_REG_CONTROL_IDX_MODE_TRANSEN 5 +#define ACC_REG_CONTROL_IDX_MODE_TS 6 +#define ACC_REG_CONTROL_IDX_MODE_SCHEDULE 7 +#define ACC_REG_CONTROL_MASK_MODE_RESETMODE \ + BIT(ACC_REG_CONTROL_IDX_MODE_RESETMODE) +#define ACC_REG_CONTROL_MASK_MODE_LOM \ + BIT(ACC_REG_CONTROL_IDX_MODE_LOM) +#define ACC_REG_CONTROL_MASK_MODE_STM \ + BIT(ACC_REG_CONTROL_IDX_MODE_STM) +#define ACC_REG_CONTROL_MASK_MODE_TRANSEN \ + BIT(ACC_REG_CONTROL_IDX_MODE_TRANSEN) +#define ACC_REG_CONTROL_MASK_MODE_TS \ + BIT(ACC_REG_CONTROL_IDX_MODE_TS) +#define ACC_REG_CONTROL_MASK_MODE_SCHEDULE \ + BIT(ACC_REG_CONTROL_IDX_MODE_SCHEDULE) + +#define ACC_REG_CONTROL_IDX_IE_RXTX 8 +#define ACC_REG_CONTROL_IDX_IE_TXERROR 9 +#define ACC_REG_CONTROL_IDX_IE_ERRWARN 10 +#define ACC_REG_CONTROL_IDX_IE_OVERRUN 11 +#define ACC_REG_CONTROL_IDX_IE_TSI 12 +#define ACC_REG_CONTROL_IDX_IE_ERRPASS 13 +#define ACC_REG_CONTROL_IDX_IE_BUSERR 15 +#define ACC_REG_CONTROL_MASK_IE_RXTX BIT(ACC_REG_CONTROL_IDX_IE_RXTX) +#define ACC_REG_CONTROL_MASK_IE_TXERROR BIT(ACC_REG_CONTROL_IDX_IE_TXERROR) +#define ACC_REG_CONTROL_MASK_IE_ERRWARN BIT(ACC_REG_CONTROL_IDX_IE_ERRWARN) +#define ACC_REG_CONTROL_MASK_IE_OVERRUN BIT(ACC_REG_CONTROL_IDX_IE_OVERRUN) +#define ACC_REG_CONTROL_MASK_IE_TSI BIT(ACC_REG_CONTROL_IDX_IE_TSI) +#define ACC_REG_CONTROL_MASK_IE_ERRPASS BIT(ACC_REG_CONTROL_IDX_IE_ERRPASS) +#define ACC_REG_CONTROL_MASK_IE_BUSERR BIT(ACC_REG_CONTROL_IDX_IE_BUSERR) + +/* 256 BM_MSGs of 32 byte size */ +#define ACC_CORE_DMAMSG_SIZE 32U +#define ACC_CORE_DMABUF_SIZE (256U * ACC_CORE_DMAMSG_SIZE) + +enum acc_bmmsg_id { + BM_MSG_ID_RXTXDONE = 0x01, + BM_MSG_ID_TXABORT = 0x02, + BM_MSG_ID_OVERRUN = 0x03, + BM_MSG_ID_BUSERR = 0x04, + BM_MSG_ID_ERRPASSIVE = 0x05, + BM_MSG_ID_ERRWARN = 0x06, + BM_MSG_ID_TIMESLICE = 0x07, + BM_MSG_ID_HWTIMER = 0x08, + BM_MSG_ID_HOTPLUG = 0x09, +}; + +/* The struct acc_bmmsg* structure declarations that follow here provide + * access to the ring buffer of bus master messages maintained by the FPGA + * bus master engine. All bus master messages have the same size of + * ACC_CORE_DMAMSG_SIZE and a minimum alignment of ACC_CORE_DMAMSG_SIZE in + * memory. + * + * All structure members are natural aligned. Therefore we should not need + * a __packed attribute. All struct acc_bmmsg* declarations have at least + * reserved* members to fill the structure to the full ACC_CORE_DMAMSG_SIZE. + * + * A failure of this property due padding will be detected at compile time + * by BUILD_BUG_ON(sizeof(struct acc_bmmsg) != ACC_CORE_DMAMSG_SIZE) + */ + +struct acc_bmmsg_rxtxdone { + u8 msg_id; + u8 txfifo_level; + u8 reserved1[2]; + u8 txtsfifo_level; + u8 reserved2[3]; + u32 id; + union { + struct { + u8 len; + u8 reserved0; + u8 bits; + u8 state; + } rxtx; + struct { + u8 len; + u8 msg_lost; + u8 bits; + u8 state; + } rx; + struct { + u8 len; + u8 txfifo_idx; + u8 bits; + u8 state; + } tx; + } dlc; + u8 data[8]; + /* Time stamps in struct acc_ov::timestamp_frequency ticks. */ + u64 ts; +}; + +struct acc_bmmsg_txabort { + u8 msg_id; + u8 txfifo_level; + u16 abort_mask; + u8 txtsfifo_level; + u8 reserved2[1]; + u16 abort_mask_txts; + u64 ts; + u32 reserved3[4]; +}; + +struct acc_bmmsg_overrun { + u8 msg_id; + u8 txfifo_level; + u8 lost_cnt; + u8 reserved1; + u8 txtsfifo_level; + u8 reserved2[3]; + u64 ts; + u32 reserved3[4]; +}; + +struct acc_bmmsg_buserr { + u8 msg_id; + u8 txfifo_level; + u8 ecc; + u8 reserved1; + u8 txtsfifo_level; + u8 reserved2[3]; + u64 ts; + u32 reg_status; + u32 reg_btr; + u32 reserved3[2]; +}; + +struct acc_bmmsg_errstatechange { + u8 msg_id; + u8 txfifo_level; + u8 reserved1[2]; + u8 txtsfifo_level; + u8 reserved2[3]; + u64 ts; + u32 reg_status; + u32 reserved3[3]; +}; + +struct acc_bmmsg_timeslice { + u8 msg_id; + u8 txfifo_level; + u8 reserved1[2]; + u8 txtsfifo_level; + u8 reserved2[3]; + u64 ts; + u32 reserved3[4]; +}; + +struct acc_bmmsg_hwtimer { + u8 msg_id; + u8 reserved1[3]; + u32 reserved2[1]; + u64 timer; + u32 reserved3[4]; +}; + +struct acc_bmmsg_hotplug { + u8 msg_id; + u8 reserved1[3]; + u32 reserved2[7]; +}; + +struct acc_bmmsg { + union { + u8 msg_id; + struct acc_bmmsg_rxtxdone rxtxdone; + struct acc_bmmsg_txabort txabort; + struct acc_bmmsg_overrun overrun; + struct acc_bmmsg_buserr buserr; + struct acc_bmmsg_errstatechange errstatechange; + struct acc_bmmsg_timeslice timeslice; + struct acc_bmmsg_hwtimer hwtimer; + } u; +}; + +/* Regarding Documentation/process/volatile-considered-harmful.rst the + * forth exception applies to the "irq_cnt" member of the structure + * below. The u32 variable "irq_cnt" points to is updated by the ESDACC + * FPGA via DMA. + */ +struct acc_bmfifo { + const struct acc_bmmsg *messages; + /* Bits 0..7: bm_fifo head index */ + volatile const u32 *irq_cnt; + u32 local_irq_cnt; + u32 msg_fifo_tail; +}; + +struct acc_core { + void __iomem *addr; + struct net_device *netdev; + struct acc_bmfifo bmfifo; + u8 tx_fifo_size; + u8 tx_fifo_head; + u8 tx_fifo_tail; +}; + +struct acc_ov { + void __iomem *addr; + struct acc_bmfifo bmfifo; + u32 timestamp_frequency; + u32 ts2ns_numerator; + u32 ts2ns_denominator; + u32 core_frequency; + u16 version; + u16 features; + u8 total_cores; + u8 active_cores; +}; + +struct acc_net_priv { + struct can_priv can; /* must be the first member! */ + struct acc_core *core; + struct acc_ov *ov; +}; + +static inline u32 acc_read32(struct acc_core *core, unsigned short offs) +{ + return ioread32be(core->addr + offs); +} + +static inline void acc_write32(struct acc_core *core, + unsigned short offs, u32 v) +{ + iowrite32be(v, core->addr + offs); +} + +static inline void acc_write32_noswap(struct acc_core *core, + unsigned short offs, u32 v) +{ + iowrite32(v, core->addr + offs); +} + +static inline void acc_set_bits(struct acc_core *core, + unsigned short offs, u32 mask) +{ + u32 v = acc_read32(core, offs); + + v |= mask; + acc_write32(core, offs, v); +} + +static inline void acc_clear_bits(struct acc_core *core, + unsigned short offs, u32 mask) +{ + u32 v = acc_read32(core, offs); + + v &= ~mask; + acc_write32(core, offs, v); +} + +static inline int acc_resetmode_entered(struct acc_core *core) +{ + u32 ctrl = acc_read32(core, ACC_CORE_OF_CTRL_MODE); + + return (ctrl & ACC_REG_CONTROL_MASK_MODE_RESETMODE) != 0; +} + +static inline u32 acc_ov_read32(struct acc_ov *ov, unsigned short offs) +{ + return ioread32be(ov->addr + offs); +} + +static inline void acc_ov_write32(struct acc_ov *ov, + unsigned short offs, u32 v) +{ + iowrite32be(v, ov->addr + offs); +} + +static inline void acc_ov_set_bits(struct acc_ov *ov, + unsigned short offs, u32 b) +{ + u32 v = acc_ov_read32(ov, offs); + + v |= b; + acc_ov_write32(ov, offs, v); +} + +static inline void acc_ov_clear_bits(struct acc_ov *ov, + unsigned short offs, u32 b) +{ + u32 v = acc_ov_read32(ov, offs); + + v &= ~b; + acc_ov_write32(ov, offs, v); +} + +static inline void acc_reset_fpga(struct acc_ov *ov) +{ + acc_ov_write32(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_FPGA_RESET); + + /* Also reset I^2C, to re-detect card addons at every driver start: */ + acc_ov_clear_bits(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_I2C_ENABLE); + mdelay(2); + acc_ov_set_bits(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_I2C_ENABLE); + mdelay(10); +} + +void acc_init_ov(struct acc_ov *ov, struct device *dev); +void acc_init_bm_ptr(struct acc_ov *ov, struct acc_core *cores, + const void *mem); +int acc_open(struct net_device *netdev); +int acc_close(struct net_device *netdev); +netdev_tx_t acc_start_xmit(struct sk_buff *skb, struct net_device *netdev); +int acc_get_berr_counter(const struct net_device *netdev, + struct can_berr_counter *bec); +int acc_set_mode(struct net_device *netdev, enum can_mode mode); +int acc_set_bittiming(struct net_device *netdev); +irqreturn_t acc_card_interrupt(struct acc_ov *ov, struct acc_core *cores); -- 2.25.1 ^ permalink raw reply related [flat|nested] 15+ messages in thread
* Re: [PATCH v6 2/4] can: esd: add support for esd GmbH PCIe/402 CAN interface family 2021-12-01 22:03 ` [PATCH 2/4] can: esd: add support for esd GmbH PCIe/402 CAN interface family Stefan Mätje @ 2022-02-01 17:25 ` Marc Kleine-Budde 2022-09-30 22:15 ` Stefan Mätje 0 siblings, 1 reply; 15+ messages in thread From: Marc Kleine-Budde @ 2022-02-01 17:25 UTC (permalink / raw) To: Stefan Mätje; +Cc: linux-can, Wolfgang Grandegger, netdev, linux-kernel [-- Attachment #1: Type: text/plain, Size: 59897 bytes --] On 01.12.2021 23:03:26, Stefan Mätje wrote: > This patch adds support for the PCI based PCIe/402 CAN interface family > from esd GmbH that is available with various form factors > (https://esd.eu/technologie/can/can-und-can-fd-interfaces). > > All boards utilize a FPGA based CAN controller solution developed > by esd (esdACC). For more information on the esdACC see > https://esd.eu/technologie/can/esd-advanced-can-controller-esdacc. > > This driver detects all available CAN interface board variants of > the family but atm. operates the CAN-FD capable devices in > Classic-CAN mode only! A later patch will introduce the CAN-FD > functionality in this driver. > > Co-developed-by: Thomas Körper <thomas.koerper@esd.eu> > Signed-off-by: Thomas Körper <thomas.koerper@esd.eu> > Signed-off-by: Stefan Mätje <stefan.maetje@esd.eu> > --- > drivers/net/can/Kconfig | 1 + > drivers/net/can/Makefile | 1 + > drivers/net/can/esd/Kconfig | 12 + > drivers/net/can/esd/Makefile | 7 + > drivers/net/can/esd/esd_402_pci-core.c | 502 ++++++++++++++++ > drivers/net/can/esd/esdacc.c | 777 +++++++++++++++++++++++++ > drivers/net/can/esd/esdacc.h | 380 ++++++++++++ > 7 files changed, 1680 insertions(+) > create mode 100644 drivers/net/can/esd/Kconfig > create mode 100644 drivers/net/can/esd/Makefile > create mode 100644 drivers/net/can/esd/esd_402_pci-core.c > create mode 100644 drivers/net/can/esd/esdacc.c > create mode 100644 drivers/net/can/esd/esdacc.h > > diff --git a/drivers/net/can/Kconfig b/drivers/net/can/Kconfig > index fff259247d52..47cfb6ae0772 100644 > --- a/drivers/net/can/Kconfig > +++ b/drivers/net/can/Kconfig > @@ -170,6 +170,7 @@ config PCH_CAN > > source "drivers/net/can/c_can/Kconfig" > source "drivers/net/can/cc770/Kconfig" > +source "drivers/net/can/esd/Kconfig" > source "drivers/net/can/ifi_canfd/Kconfig" > source "drivers/net/can/m_can/Kconfig" > source "drivers/net/can/mscan/Kconfig" > diff --git a/drivers/net/can/Makefile b/drivers/net/can/Makefile > index a2b4463d8480..015b6fc110d1 100644 > --- a/drivers/net/can/Makefile > +++ b/drivers/net/can/Makefile > @@ -8,6 +8,7 @@ obj-$(CONFIG_CAN_VXCAN) += vxcan.o > obj-$(CONFIG_CAN_SLCAN) += slcan.o > > obj-y += dev/ > +obj-y += esd/ > obj-y += rcar/ > obj-y += spi/ > obj-y += usb/ > diff --git a/drivers/net/can/esd/Kconfig b/drivers/net/can/esd/Kconfig > new file mode 100644 > index 000000000000..54bfc366634c > --- /dev/null > +++ b/drivers/net/can/esd/Kconfig > @@ -0,0 +1,12 @@ > +# SPDX-License-Identifier: GPL-2.0-only > +config CAN_ESD_402_PCI > + tristate "esd electronics gmbh CAN-PCI(e)/402 family" > + depends on PCI && HAS_DMA > + help > + Support for C402 card family from esd electronics gmbh. > + This card family is based on the ESDACC CAN controller and > + available in several form factors: PCI, PCIe, PCIe Mini, > + M.2 PCIe, CPCIserial, PMC, XMC (see https://esd.eu/en) > + > + This driver can also be built as a module. In this case the > + module will be called esd_402_pci. > diff --git a/drivers/net/can/esd/Makefile b/drivers/net/can/esd/Makefile > new file mode 100644 > index 000000000000..5dd2d470c286 > --- /dev/null > +++ b/drivers/net/can/esd/Makefile > @@ -0,0 +1,7 @@ > +# SPDX-License-Identifier: GPL-2.0-only > +# > +# Makefile for esd gmbh ESDACC controller driver > +# > +esd_402_pci-objs := esdacc.o esd_402_pci-core.o > + > +obj-$(CONFIG_CAN_ESD_402_PCI) += esd_402_pci.o > diff --git a/drivers/net/can/esd/esd_402_pci-core.c b/drivers/net/can/esd/esd_402_pci-core.c > new file mode 100644 > index 000000000000..80d816a78859 > --- /dev/null > +++ b/drivers/net/can/esd/esd_402_pci-core.c > @@ -0,0 +1,502 @@ > +// SPDX-License-Identifier: GPL-2.0-only > +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh > + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh > + */ > + > +#include <linux/kernel.h> > +#include <linux/module.h> > +#include <linux/interrupt.h> > +#include <linux/pci.h> > +#include <linux/io.h> > +#include <linux/delay.h> > +#include <linux/netdevice.h> > +#include <linux/dma-mapping.h> > +#include <linux/can.h> > +#include <linux/can/dev.h> > +#include <linux/can/netlink.h> please sort alphabetically > + > +#include "esdacc.h" > + > +#define DRV_NAME "esd_402_pci" > + > +#define ESD_PCI_DEVICE_ID_PCIE402 0x0402 > + > +#define PCI402_FPGA_VER_MIN 0x003d > +#define PCI402_MAX_CORES 6 > +#define PCI402_BAR 0 > +#define PCI402_IO_OV_OFFS 0 > +#define PCI402_IO_PCIEP_OFFS 0x10000 > +#define PCI402_IO_LEN_TOTAL 0x20000 > +#define PCI402_IO_LEN_CORE 0x2000 > +#define PCI402_PCICFG_MSICAP_CSR 0x52 > +#define PCI402_PCICFG_MSICAP_ADDR 0x54 > +#define PCI402_PCICFG_MSICAP_DATA 0x5c > + > +#define PCI402_DMA_MASK DMA_BIT_MASK(32) > +#define PCI402_DMA_SIZE ALIGN(0x10000, PAGE_SIZE) > + > +#define PCI402_PCIEP_OF_INT_ENABLE 0x0050 > +#define PCI402_PCIEP_OF_BM_ADDR_LO 0x1000 > +#define PCI402_PCIEP_OF_BM_ADDR_HI 0x1004 > +#define PCI402_PCIEP_OF_MSI_ADDR_LO 0x1008 > +#define PCI402_PCIEP_OF_MSI_ADDR_HI 0x100c nitpick: please use single space for indention. Sooner or later new stuff is added and the original alignment doesn't fit anymore. > + > +/* The BTR register capabilities described by the can_bittiming_const structures > + * below are valid since ESDACC version 0x0032. > + */ > + > +/* Used if the ESDACC FPGA is built as CAN-Classic version. */ > +static const struct can_bittiming_const pci402_bittiming_const = { > + .name = "esd_402", > + .tseg1_min = 1, > + .tseg1_max = 16, > + .tseg2_min = 1, > + .tseg2_max = 8, > + .sjw_max = 4, > + .brp_min = 1, > + .brp_max = 512, > + .brp_inc = 1, > +}; > + > +/* Used if the ESDACC FPGA is built as CAN-FD version. */ > +static const struct can_bittiming_const pci402_bittiming_const_canfd = { > + .name = "esd_402fd", > + .tseg1_min = 1, > + .tseg1_max = 256, > + .tseg2_min = 1, > + .tseg2_max = 128, > + .sjw_max = 128, > + .brp_min = 1, > + .brp_max = 256, > + .brp_inc = 1, > +}; > + > +static const struct net_device_ops pci402_acc_netdev_ops = { > + .ndo_open = acc_open, > + .ndo_stop = acc_close, > + .ndo_start_xmit = acc_start_xmit, > + .ndo_change_mtu = can_change_mtu > +}; > + > +struct pci402_card { > + /* Actually mapped io space, all other iomem derived from this */ > + void __iomem *addr; > + void __iomem *addr_pciep; > + > + void *dma_buf; > + dma_addr_t dma_hnd; > + > + struct acc_ov ov; > + struct acc_core *cores; > + > + bool msi_enabled; > +}; > + > +static irqreturn_t pci402_interrupt(int irq, void *dev_id) > +{ > + struct pci_dev *pdev = dev_id; > + struct pci402_card *card = pci_get_drvdata(pdev); > + irqreturn_t irq_status; > + > + irq_status = acc_card_interrupt(&card->ov, card->cores); > + > + return irq_status; > +} > + > +static int pci402_set_msiconfig(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + u32 addr_lo_offs = 0; > + u32 addr_lo = 0; > + u32 addr_hi = 0; > + u32 data = 0; > + u16 csr = 0; > + int err; > + > + err = pci_read_config_word(pdev, PCI402_PCICFG_MSICAP_CSR, &csr); > + if (err) > + goto failed; > + > + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_ADDR, &addr_lo); > + if (err) > + goto failed; > + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_ADDR + 4, > + &addr_hi); > + if (err) > + goto failed; > + > + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_DATA, &data); > + if (err) > + goto failed; > + > + addr_lo_offs = addr_lo & 0x0000ffff; > + addr_lo &= 0xffff0000; > + > + if (addr_hi) > + addr_lo |= 1; /* Enable 64-Bit addressing in address space */ > + > + if (!(csr & 0x0001)) { /* Enable bit */ > + err = -EINVAL; > + goto failed; > + } > + > + iowrite32(addr_lo, card->addr_pciep + PCI402_PCIEP_OF_MSI_ADDR_LO); > + iowrite32(addr_hi, card->addr_pciep + PCI402_PCIEP_OF_MSI_ADDR_HI); > + acc_ov_write32(&card->ov, ACC_OV_OF_MSI_ADDRESSOFFSET, addr_lo_offs); > + acc_ov_write32(&card->ov, ACC_OV_OF_MSI_DATA, data); > + > + return 0; > + > +failed: > + pci_warn(pdev, "Error while setting MSI configuration:\n" > + "CSR: 0x%.4x, addr: 0x%.8x%.8x, data: 0x%.8x\n", > + csr, addr_hi, addr_lo, data); > + > + return err; > +} > + > +static int pci402_init_card(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + > + card->ov.addr = card->addr + PCI402_IO_OV_OFFS; > + card->addr_pciep = card->addr + PCI402_IO_PCIEP_OFFS; > + > + acc_reset_fpga(&card->ov); > + acc_init_ov(&card->ov, &pdev->dev); > + > + if (card->ov.version < PCI402_FPGA_VER_MIN) { > + pci_err(pdev, > + "ESDACC version (0x%.4x) outdated, please update\n", > + card->ov.version); > + return -EINVAL; > + } > + > + if (card->ov.active_cores > PCI402_MAX_CORES) { > + pci_warn(pdev, > + "Card has more active cores than supported by driver, %u core(s) will be ignored\n", > + card->ov.active_cores - PCI402_MAX_CORES); > + card->ov.active_cores = PCI402_MAX_CORES; Where is that limitation in this driver? PCI402_MAX_CORES is only used here. > + } > + card->cores = devm_kcalloc(&pdev->dev, card->ov.active_cores, > + sizeof(struct acc_core), GFP_KERNEL); > + if (!card->cores) > + return -ENOMEM; > + > + if (card->ov.features & ACC_OV_REG_FEAT_MASK_CANFD) { > + pci_warn(pdev, > + "ESDACC with CAN-FD feature detected. This driver doesn't support CAN-FD yet.\n"); > + } > + > +#ifdef __LITTLE_ENDIAN > + /* So card converts all busmastered data to LE for us: */ > + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, > + ACC_OV_REG_MODE_MASK_ENDIAN_LITTLE); > +#endif > + > + return 0; > +} > + > +static int pci402_init_interrupt(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + int err; > + > + err = pci_enable_msi(pdev); > + if (!err) { > + err = pci402_set_msiconfig(pdev); > + if (!err) { > + card->msi_enabled = true; > + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, > + ACC_OV_REG_MODE_MASK_MSI_ENABLE); > + pci_info(pdev, "MSI enabled\n"); > + } > + } > + > + err = devm_request_irq(&pdev->dev, pdev->irq, pci402_interrupt, > + IRQF_SHARED, dev_name(&pdev->dev), pdev); > + if (err) > + goto failure_msidis; > + > + iowrite32(1, card->addr_pciep + PCI402_PCIEP_OF_INT_ENABLE); > + > + return 0; > + > +failure_msidis: > + if (card->msi_enabled) { > + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, > + ACC_OV_REG_MODE_MASK_MSI_ENABLE); > + pci_disable_msi(pdev); > + card->msi_enabled = false; > + } > + > + return err; > +} > + > +static void pci402_finish_interrupt(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + > + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_INT_ENABLE); > + devm_free_irq(&pdev->dev, pdev->irq, pdev); A devm manged interrupt gets automatically freed, it makes no sense if you free it manually here. I think it's best to use a non devm IRQ instead. > + > + if (card->msi_enabled) { > + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, > + ACC_OV_REG_MODE_MASK_MSI_ENABLE); > + pci_disable_msi(pdev); > + card->msi_enabled = false; > + } > +} > + > +static int pci402_init_dma(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + int err; > + > + err = dma_set_coherent_mask(&pdev->dev, PCI402_DMA_MASK); > + if (err) { > + pci_err(pdev, "DMA set mask failed!\n"); > + return err; > + } > + > + /* The ESDACC DMA engine needs the DMA buffer aligned to a 64k > + * boundary. The DMA API guarantees to align the returned buffer to the > + * smallest PAGE_SIZE order which is greater than or equal to the > + * requested size. With PCI402_DMA_SIZE == 64kB this suffices here. > + */ > + card->dma_buf = dma_alloc_coherent(&pdev->dev, PCI402_DMA_SIZE, > + &card->dma_hnd, GFP_ATOMIC); Why do you use ATOMIC here? > + if (!card->dma_buf) { > + pci_err(pdev, "DMA alloc failed!\n"); > + return -ENOMEM; > + } > + > + acc_init_bm_ptr(&card->ov, card->cores, card->dma_buf); > + > + iowrite32((u32)card->dma_hnd, > + card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_LO); cast not needed > + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_HI); > + > + pci_set_master(pdev); > + > + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, > + ACC_OV_REG_MODE_MASK_BM_ENABLE); > + > + return 0; > +} > + > +static void pci402_finish_dma(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + int i; > + > + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, > + ACC_OV_REG_MODE_MASK_BM_ENABLE); > + > + pci_clear_master(pdev); > + > + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_LO); > + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_HI); > + > + card->ov.bmfifo.messages = NULL; > + card->ov.bmfifo.irq_cnt = NULL; > + for (i = 0; i < card->ov.active_cores; i++) { > + struct acc_core *core = &card->cores[i]; > + > + core->bmfifo.messages = NULL; > + core->bmfifo.irq_cnt = NULL; > + } > + > + dma_free_coherent(&pdev->dev, PCI402_DMA_SIZE, card->dma_buf, > + card->dma_hnd); > + card->dma_buf = NULL; > +} > + > +static int pci402_init_cores(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + int err; > + int i; > + > + for (i = 0; i < card->ov.active_cores; i++) { > + struct acc_core *core = &card->cores[i]; > + struct acc_net_priv *priv; > + struct net_device *netdev; > + u32 fifo_config; > + > + core->addr = card->ov.addr + (i + 1) * PCI402_IO_LEN_CORE; > + > + fifo_config = acc_read32(core, ACC_CORE_OF_TXFIFO_CONFIG); > + core->tx_fifo_size = (u8)(fifo_config >> 24); cast not needed. > + if (core->tx_fifo_size <= 1) { > + pci_err(pdev, "Invalid tx_fifo_size!\n"); > + err = -EINVAL; > + goto failure; > + } > + > + netdev = alloc_candev(sizeof(*priv), core->tx_fifo_size); > + if (!netdev) { > + err = -ENOMEM; > + goto failure; > + } > + core->netdev = netdev; > + > + netdev->flags |= IFF_ECHO; > + netdev->dev_port = i; > + netdev->netdev_ops = &pci402_acc_netdev_ops; > + SET_NETDEV_DEV(netdev, &pdev->dev); > + > + priv = netdev_priv(netdev); > + priv->can.state = CAN_STATE_STOPPED; The state is automatically set by alloc_candev(). > + priv->can.ctrlmode_supported = CAN_CTRLMODE_LISTENONLY | > + CAN_CTRLMODE_BERR_REPORTING | > + CAN_CTRLMODE_CC_LEN8_DLC | > + CAN_CTRLMODE_LOOPBACK; Please sort these by occurrence in can/netlink.h > + priv->can.clock.freq = card->ov.core_frequency; > + priv->can.bittiming_const = > + (card->ov.features & ACC_OV_REG_FEAT_MASK_CANFD) ? > + &pci402_bittiming_const_canfd : > + &pci402_bittiming_const; nitpick: Please use if (card->ov.features & ACC_OV_REG_FEAT_MASK_CANFD)...else here > + priv->can.do_set_bittiming = acc_set_bittiming; Please call it directly from the open() callback. > + priv->can.do_set_mode = acc_set_mode; > + priv->can.do_get_berr_counter = acc_get_berr_counter; > + > + priv->core = core; > + priv->ov = &card->ov; > + > + err = register_candev(netdev); > + if (err) { > + free_candev(core->netdev); > + core->netdev = NULL; > + goto failure; > + } > + > + netdev_info(netdev, "registered\n"); > + } > + > + return 0; > + > +failure: > + for (i--; i >= 0; i--) { > + struct acc_core *core = &card->cores[i]; > + > + netdev_info(core->netdev, "unregistering...\n"); > + unregister_candev(core->netdev); > + > + free_candev(core->netdev); > + core->netdev = NULL; > + } > + > + return err; > +} > + > +static void pci402_finish_cores(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + int i; > + > + for (i = 0; i < card->ov.active_cores; i++) { > + struct acc_core *core = &card->cores[i]; > + > + netdev_info(core->netdev, "unregister\n"); > + unregister_candev(core->netdev); > + > + free_candev(core->netdev); > + core->netdev = NULL; > + } > +} > + > +static int pci402_probe(struct pci_dev *pdev, const struct pci_device_id *ent) > +{ > + struct pci402_card *card = NULL; > + int err; > + > + err = pci_enable_device(pdev); > + if (err) > + return err; > + > + card = devm_kzalloc(&pdev->dev, sizeof(*card), GFP_KERNEL); > + if (!card) > + goto failure_disable_pci; > + > + pci_set_drvdata(pdev, card); > + > + err = pci_request_regions(pdev, pci_name(pdev)); > + if (err) > + goto failure_disable_pci; > + > + card->addr = pci_iomap(pdev, PCI402_BAR, PCI402_IO_LEN_TOTAL); > + if (!card->addr) { > + err = -ENOMEM; > + goto failure_release_regions; > + } > + > + err = pci402_init_card(pdev); > + if (err) > + goto failure_unmap; > + > + err = pci402_init_dma(pdev); > + if (err) > + goto failure_unmap; > + > + err = pci402_init_interrupt(pdev); > + if (err) > + goto failure_finish_dma; > + > + err = pci402_init_cores(pdev); > + if (err) > + goto failure_finish_interrupt; > + > + return 0; > + > +failure_finish_interrupt: > + pci402_finish_interrupt(pdev); > + > +failure_finish_dma: > + pci402_finish_dma(pdev); > + > +failure_unmap: > + pci_iounmap(pdev, card->addr); > + > +failure_release_regions: > + pci_release_regions(pdev); > + > +failure_disable_pci: > + pci_disable_device(pdev); > + > + return err; > +} > + > +static void pci402_remove(struct pci_dev *pdev) > +{ > + struct pci402_card *card = pci_get_drvdata(pdev); > + > + pci402_finish_interrupt(pdev); > + pci402_finish_cores(pdev); > + pci402_finish_dma(pdev); > + pci_iounmap(pdev, card->addr); > + pci_release_regions(pdev); > + pci_disable_device(pdev); > +} > + > +static const struct pci_device_id pci402_tbl[] = { > + { PCI_VENDOR_ID_ESDGMBH, ESD_PCI_DEVICE_ID_PCIE402, > + PCI_VENDOR_ID_ESDGMBH, PCI_ANY_ID, > + 0U, 0U, 0UL }, > + { 0, } > +}; > +MODULE_DEVICE_TABLE(pci, pci402_tbl); > + > +static struct pci_driver pci402_driver = { > + .name = DRV_NAME, > + .id_table = pci402_tbl, > + .probe = pci402_probe, > + .remove = pci402_remove, > +}; > + > +module_pci_driver(pci402_driver); > + > +MODULE_DESCRIPTION("Socket-CAN driver for esd CAN 402 card family with esdACC core on PCIe"); > +MODULE_AUTHOR("Thomas Körper <socketcan@esd.eu>"); > +MODULE_AUTHOR("Stefan Mätje <stefan.maetje@esd.eu>"); > +MODULE_LICENSE("GPL"); > diff --git a/drivers/net/can/esd/esdacc.c b/drivers/net/can/esd/esdacc.c > new file mode 100644 > index 000000000000..13f7397dfc4e > --- /dev/null > +++ b/drivers/net/can/esd/esdacc.c > @@ -0,0 +1,777 @@ > +// SPDX-License-Identifier: GPL-2.0-only > +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh > + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh > + */ > + > +#include <linux/ktime.h> > +#include <linux/gcd.h> > +#include <linux/io.h> > +#include <linux/delay.h> please sort alphabetically > +#include "esdacc.h" > + > +/* ecc value of esdACC equals SJA1000's ECC register */ > +#define ACC_ECC_SEG 0x1f > +#define ACC_ECC_DIR 0x20 > +#define ACC_ECC_BIT 0x00 > +#define ACC_ECC_FORM 0x40 > +#define ACC_ECC_STUFF 0x80 > +#define ACC_ECC_MASK 0xc0 > + > +#define ACC_BM_IRQ_UNMASK_ALL 0x55555555U > +#define ACC_BM_IRQ_MASK_ALL 0xaaaaaaaaU > +#define ACC_BM_IRQ_MASK 0x2U > +#define ACC_BM_IRQ_UNMASK 0x1U > +#define ACC_BM_LENFLAG_TX 0x20 > + > +#define ACC_REG_STATUS_IDX_STATUS_DOS 16 > +#define ACC_REG_STATUS_IDX_STATUS_ES 17 > +#define ACC_REG_STATUS_IDX_STATUS_EP 18 > +#define ACC_REG_STATUS_IDX_STATUS_BS 19 > +#define ACC_REG_STATUS_IDX_STATUS_RBS 20 > +#define ACC_REG_STATUS_IDX_STATUS_RS 21 > +#define ACC_REG_STATUS_MASK_STATUS_DOS BIT(ACC_REG_STATUS_IDX_STATUS_DOS) > +#define ACC_REG_STATUS_MASK_STATUS_ES BIT(ACC_REG_STATUS_IDX_STATUS_ES) > +#define ACC_REG_STATUS_MASK_STATUS_EP BIT(ACC_REG_STATUS_IDX_STATUS_EP) > +#define ACC_REG_STATUS_MASK_STATUS_BS BIT(ACC_REG_STATUS_IDX_STATUS_BS) > +#define ACC_REG_STATUS_MASK_STATUS_RBS BIT(ACC_REG_STATUS_IDX_STATUS_RBS) > +#define ACC_REG_STATUS_MASK_STATUS_RS BIT(ACC_REG_STATUS_IDX_STATUS_RS) > + > +static void acc_resetmode_enter(struct acc_core *core) > +{ > + int i; > + > + acc_set_bits(core, ACC_CORE_OF_CTRL_MODE, > + ACC_REG_CONTROL_MASK_MODE_RESETMODE); > + > + for (i = 0; i < 10; i++) { > + if (acc_resetmode_entered(core)) > + return; > + > + udelay(5); > + } > + > + netdev_warn(core->netdev, "Entering reset mode timed out\n"); > +} > + > +static void acc_resetmode_leave(struct acc_core *core) > +{ > + int i; > + > + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, > + ACC_REG_CONTROL_MASK_MODE_RESETMODE); > + > + for (i = 0; i < 10; i++) { > + if (!acc_resetmode_entered(core)) > + return; > + > + udelay(5); > + } > + > + netdev_warn(core->netdev, "Leaving reset mode timed out\n"); > +} > + > +static void acc_txq_put(struct acc_core *core, u32 acc_id, u8 acc_dlc, > + const void *data) > +{ > + acc_write32_noswap(core, ACC_CORE_OF_TXFIFO_DATA_1, > + *((const u32 *)(data + 4))); > + acc_write32_noswap(core, ACC_CORE_OF_TXFIFO_DATA_0, > + *((const u32 *)data)); > + acc_write32(core, ACC_CORE_OF_TXFIFO_DLC, acc_dlc); > + /* CAN id must be written at last. This write starts TX. */ > + acc_write32(core, ACC_CORE_OF_TXFIFO_ID, acc_id); > +} > + > +/* Prepare conversion factor from ESDACC time stamp ticks to ns > + * > + * The conversion factor ts2ns from time stamp counts to ns is basically > + * ts2ns = NSEC_PER_SEC / timestamp_frequency > + * > + * To avoid an overflow, the ts2ns fraction is truncated with its gcd and > + * only the truncated numerator and denominator are used further. > + */ > +static void acc_init_ov_ts2ns(struct acc_ov *ov) > +{ > + u32 ts2ns_gcd = (u32)gcd(NSEC_PER_SEC, ov->timestamp_frequency); > + > + ov->ts2ns_numerator = (u32)NSEC_PER_SEC / ts2ns_gcd; > + ov->ts2ns_denominator = ov->timestamp_frequency / ts2ns_gcd; > +} Please don't craft your own time conversion functions. please use cyclecounter/timercounter instead. Have a look at: drivers/net/can/spi/mcp251xfd/mcp251xfd-timestamp.c > + > +static ktime_t acc_ts2ktime(struct acc_ov *ov, u64 ts) > +{ > + u64 ns; > + > + ts = ts * ov->ts2ns_numerator; > + ns = div_u64(ts, ov->ts2ns_denominator); > + > + return ns_to_ktime(ns); > +} > + > +void acc_init_ov(struct acc_ov *ov, struct device *dev) > +{ > + u32 temp; > + /* For the justification of this see comment on struct acc_bmmsg* > + * in esdacc.h. > + */ > + BUILD_BUG_ON(sizeof(struct acc_bmmsg) != ACC_CORE_DMAMSG_SIZE); > + > + temp = acc_ov_read32(ov, ACC_OV_OF_VERSION); > + ov->version = (u16)temp; > + ov->features = (u16)(temp >> 16); casts not needed > + > + temp = acc_ov_read32(ov, ACC_OV_OF_INFO); > + ov->total_cores = (u8)temp; > + ov->active_cores = (u8)(temp >> 8); casts not needed > + > + ov->core_frequency = acc_ov_read32(ov, ACC_OV_OF_CANCORE_FREQ); > + ov->timestamp_frequency = acc_ov_read32(ov, ACC_OV_OF_TS_FREQ_LO); > + acc_init_ov_ts2ns(ov); > + > + /* Depending on ESDACC feature NEW_PSC enable the new prescaler > + * or adjust core_frequency according to the implicit division by 2. > + */ > + if (ov->features & ACC_OV_REG_FEAT_MASK_NEW_PSC) { > + acc_ov_set_bits(ov, ACC_OV_OF_MODE, > + ACC_OV_REG_MODE_MASK_NEW_PSC_ENABLE); > + } else { > + ov->core_frequency /= 2; > + } > + > + dev_info(dev, > + "ESDACC v%u, freq: %u/%u, feat/strap: 0x%x/0x%x, cores: %u/%u\n", > + ov->version, ov->core_frequency, ov->timestamp_frequency, > + ov->features, acc_ov_read32(ov, ACC_OV_OF_INFO) >> 16, > + ov->active_cores, ov->total_cores); > + dev_dbg(dev, "ESDACC ts2ns: numerator %u, denominator %u\n", > + ov->ts2ns_numerator, ov->ts2ns_denominator); > +} > + > +void acc_init_bm_ptr(struct acc_ov *ov, struct acc_core *cores, const void *mem) > +{ > + unsigned int u; > + > + /* DMA buffer layout as follows where N is the number of CAN cores > + * implemented in the FPGA, i.e. N = ov->total_cores > + * > + * Layout Section size > + * +-----------------------+ > + * | FIFO Card/Overview | ACC_CORE_DMABUF_SIZE > + * | | > + * +-----------------------+ > + * | FIFO Core0 | ACC_CORE_DMABUF_SIZE > + * | | > + * +-----------------------+ > + * | ... | ... > + * | | > + * +-----------------------+ > + * | FIFO CoreN | ACC_CORE_DMABUF_SIZE > + * | | > + * +-----------------------+ > + * | irq_cnt Card/Overview | sizeof(u32) > + * +-----------------------+ > + * | irq_cnt Core0 | sizeof(u32) > + * +-----------------------+ > + * | ... | ... > + * +-----------------------+ > + * | irq_cnt CoreN | sizeof(u32) > + * +-----------------------+ > + */ > + ov->bmfifo.messages = mem; > + ov->bmfifo.irq_cnt = mem + (ov->total_cores + 1U) * ACC_CORE_DMABUF_SIZE; > + > + for (u = 0U; u < ov->active_cores; u++) { > + struct acc_core *core = &cores[u]; > + > + core->bmfifo.messages = mem + (u + 1U) * ACC_CORE_DMABUF_SIZE; > + core->bmfifo.irq_cnt = ov->bmfifo.irq_cnt + (u + 1U); > + } > +} > + > +int acc_open(struct net_device *netdev) > +{ > + struct acc_net_priv *priv = netdev_priv(netdev); > + struct acc_core *core = priv->core; > + u32 ctrl_mode; > + int err; > + > + /* Retry to enter RESET mode if out of sync. */ > + if (priv->can.state != CAN_STATE_STOPPED) { > + netdev_warn(netdev, "Entered %s() with bad can.state: %s\n", > + __func__, can_get_state_str(priv->can.state)); > + acc_resetmode_enter(core); > + if (acc_resetmode_entered(core)) > + priv->can.state = CAN_STATE_STOPPED; > + } What about always doing a full reset during open(). > + > + err = open_candev(netdev); > + if (err) > + return err; > + > + ctrl_mode = ACC_REG_CONTROL_MASK_IE_RXTX | > + ACC_REG_CONTROL_MASK_IE_TXERROR | > + ACC_REG_CONTROL_MASK_IE_ERRWARN | > + ACC_REG_CONTROL_MASK_IE_OVERRUN | > + ACC_REG_CONTROL_MASK_IE_ERRPASS; > + > + if (priv->can.ctrlmode & CAN_CTRLMODE_BERR_REPORTING) > + ctrl_mode |= ACC_REG_CONTROL_MASK_IE_BUSERR; > + > + if (priv->can.ctrlmode & CAN_CTRLMODE_LISTENONLY) > + ctrl_mode |= ACC_REG_CONTROL_MASK_MODE_LOM; > + > + acc_set_bits(core, ACC_CORE_OF_CTRL_MODE, ctrl_mode); > + > + acc_resetmode_leave(core); > + if (!acc_resetmode_entered(core)) > + priv->can.state = CAN_STATE_ERROR_ACTIVE; If the device didn't leave reset mode, then return with an error. > + /* Resync TX FIFO indices to HW state after (re-)start. */ > + { > + u32 tx_fifo_status = acc_read32(core, ACC_CORE_OF_TXFIFO_STATUS); > + > + core->tx_fifo_head = tx_fifo_status & 0xff; > + core->tx_fifo_tail = (tx_fifo_status >> 8) & 0xff; > + } Why this extra level of indention? > + netif_start_queue(netdev); > + return 0; > +} > + > +int acc_close(struct net_device *netdev) > +{ > + struct acc_net_priv *priv = netdev_priv(netdev); > + struct acc_core *core = priv->core; > + > + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, > + ACC_REG_CONTROL_MASK_IE_RXTX | > + ACC_REG_CONTROL_MASK_IE_TXERROR | > + ACC_REG_CONTROL_MASK_IE_ERRWARN | > + ACC_REG_CONTROL_MASK_IE_OVERRUN | > + ACC_REG_CONTROL_MASK_IE_ERRPASS | > + ACC_REG_CONTROL_MASK_IE_BUSERR); > + > + netif_stop_queue(netdev); > + acc_resetmode_enter(core); > + if (acc_resetmode_entered(core)) > + priv->can.state = CAN_STATE_STOPPED; Better mark the device as stopped, and make full reset during open() > + > + /* Mark pending TX requests to be aborted after controller restart. */ > + acc_write32(core, ACC_CORE_OF_TX_ABORT_MASK, 0xffff); > + > + /* ACC_REG_CONTROL_MASK_MODE_LOM is only accessible in RESET mode */ > + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, > + ACC_REG_CONTROL_MASK_MODE_LOM); > + > + close_candev(netdev); > + return 0; > +} > + > +netdev_tx_t acc_start_xmit(struct sk_buff *skb, struct net_device *netdev) > +{ > + struct acc_net_priv *priv = netdev_priv(netdev); > + struct acc_core *core = priv->core; > + struct can_frame *cf = (struct can_frame *)skb->data; > + u8 tx_fifo_head = core->tx_fifo_head; > + int fifo_usage; > + u32 acc_id; > + u8 acc_dlc; Please add a check for can_dropped_invalid_skb(). > + > + /* Access core->tx_fifo_tail only once because it may be changed > + * from the interrupt level. > + */ > + fifo_usage = tx_fifo_head - core->tx_fifo_tail; > + if (fifo_usage < 0) > + fifo_usage += core->tx_fifo_size; > + > + if (fifo_usage >= core->tx_fifo_size - 1) { > + netdev_err(core->netdev, > + "BUG: TX ring full when queue awake!\n"); > + netif_stop_queue(netdev); > + return NETDEV_TX_BUSY; > + } > + > + if (fifo_usage == core->tx_fifo_size - 2) > + netif_stop_queue(netdev); You need proper memory barriers and double checking: See the mcp251xfd driver: | netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, | struct net_device *ndev) | { | struct mcp251xfd_priv *priv = netdev_priv(ndev); | struct mcp251xfd_tx_ring *tx_ring = priv->tx; | struct mcp251xfd_tx_obj *tx_obj; | unsigned int frame_len; | u8 tx_head; | int err; | | if (can_dropped_invalid_skb(ndev, skb)) | return NETDEV_TX_OK; | | if (mcp251xfd_tx_busy(priv, tx_ring)) | return NETDEV_TX_BUSY; The magic happens in mcp251xfd_tx_busy(): | static bool mcp251xfd_tx_busy(const struct mcp251xfd_priv *priv, | struct mcp251xfd_tx_ring *tx_ring) | { | if (mcp251xfd_get_tx_free(tx_ring) > 0) | return false; | | netif_stop_queue(priv->ndev); | | /* Memory barrier before checking tx_free (head and tail) */ | smp_mb(); | | if (mcp251xfd_get_tx_free(tx_ring) == 0) { If you need at least 2 free slots, adjust this check. | netdev_dbg(priv->ndev, | "Stopping tx-queue (tx_head=0x%08x, tx_tail=0x%08x, len=%d).\n", | tx_ring->head, tx_ring->tail, | tx_ring->head - tx_ring->tail); | | return true; | } | | netif_start_queue(priv->ndev); | | return false; | } > + > + acc_dlc = can_get_cc_dlc(cf, priv->can.ctrlmode); > + if (cf->can_id & CAN_RTR_FLAG) > + acc_dlc |= ACC_CAN_RTR_FLAG; > + > + if (cf->can_id & CAN_EFF_FLAG) { > + acc_id = cf->can_id & CAN_EFF_MASK; > + acc_id |= ACC_CAN_EFF_FLAG; > + } else { > + acc_id = cf->can_id & CAN_SFF_MASK; > + } > + > + can_put_echo_skb(skb, netdev, core->tx_fifo_head, 0); > + > + tx_fifo_head++; > + if (tx_fifo_head >= core->tx_fifo_size) > + tx_fifo_head = 0U; > + core->tx_fifo_head = tx_fifo_head; > + > + acc_txq_put(core, acc_id, acc_dlc, cf->data); > + > + return NETDEV_TX_OK; > +} > + > +int acc_get_berr_counter(const struct net_device *netdev, > + struct can_berr_counter *bec) > +{ > + struct acc_net_priv *priv = netdev_priv(netdev); > + u32 core_status = acc_read32(priv->core, ACC_CORE_OF_STATUS); > + > + bec->txerr = (core_status >> 8) & 0xff; > + bec->rxerr = core_status & 0xff; > + > + return 0; > +} > + > +int acc_set_mode(struct net_device *netdev, enum can_mode mode) > +{ > + struct acc_net_priv *priv = netdev_priv(netdev); > + > + switch (mode) { > + case CAN_MODE_START: > + /* Paranoid FIFO index check. */ > + { > + const u32 tx_fifo_status = > + acc_read32(priv->core, ACC_CORE_OF_TXFIFO_STATUS); > + const u8 hw_fifo_head = (u8)tx_fifo_status; cast not needed > + > + if (hw_fifo_head != priv->core->tx_fifo_head || > + hw_fifo_head != priv->core->tx_fifo_tail) { > + netdev_warn(netdev, > + "TX FIFO mismatch: T %2u H %2u; TFHW %#08x\n", > + priv->core->tx_fifo_tail, > + priv->core->tx_fifo_head, > + tx_fifo_status); > + } > + } > + acc_resetmode_leave(priv->core); > + /* To leave the bus-off state the esdACC controller begins > + * here a grace period where it counts 128 "idle conditions" (each > + * of 11 consecutive recessive bits) on the bus as required > + * by the CAN spec. > + * > + * During this time the TX FIFO may still contain already > + * aborted "zombie" frames that are only drained from the FIFO > + * at the end of the grace period. > + * > + * To not to interfere with this drain process we don't > + * call netif_wake_queue() here. When the controller reaches > + * the error-active state again, it informs us about that > + * with an acc_bmmsg_errstatechange message. Then > + * netif_wake_queue() is called from > + * handle_core_msg_errstatechange() instead. > + */ > + break; Due 128 idle condition auto recovery it's best to shut down/reset the controller on BUS_OFF and re-start it here. It turned out to be less complex form the Linux driver point of view and the individual CAN cores feel much more consistent: Bus off means bus off and recovery is controlled by the kernel. > + > + default: > + return -EOPNOTSUPP; > + } > + > + return 0; > +} > + > +int acc_set_bittiming(struct net_device *netdev) > +{ > + struct acc_net_priv *priv = netdev_priv(netdev); > + const struct can_bittiming *bt = &priv->can.bittiming; > + u32 brp = bt->brp - 1; > + u32 btr; > + > + if (priv->ov->features & ACC_OV_REG_FEAT_MASK_CANFD) { > + u32 fbtr = 0; > + > + netdev_dbg(priv->core->netdev, > + "bit timing: brp %u, prop %u, ph1 %u ph2 %u, sjw %u\n", > + bt->brp, bt->prop_seg, > + bt->phase_seg1, bt->phase_seg2, bt->sjw); > + > + /* BRP: 8 bits @ bits 7..0 */ > + brp &= 0xff; > + > + /* TSEG1: 8 bits @ bits 7..0 */ > + btr = (bt->phase_seg1 + bt->prop_seg - 1) & 0xff; > + /* TSEG2: 7 bits @ bits 22..16 */ > + btr |= ((bt->phase_seg2 - 1) & 0x7f) << 16; > + /* SJW: 7 bits @ bits 30..24 */ > + btr |= ((bt->sjw - 1) & 0x7f) << 24; Please make use of FIELD_PREP() here. > + > + /* Keep order of accesses to ACC_CORE_OF_BRP and ACC_CORE_OF_BTR. */ > + acc_write32(priv->core, ACC_CORE_OF_BRP, brp); > + acc_write32(priv->core, ACC_CORE_OF_BTR, btr); > + > + netdev_info(priv->core->netdev, > + "ESDACC: BRP %u, NBTR 0x%08x, DBTR 0x%08x", > + brp, btr, fbtr); please make this a netdev_dbg() > + } else { > + netdev_dbg(priv->core->netdev, > + "bit timing: brp %u, prop %u, ph1 %u ph2 %u, sjw %u\n", > + bt->brp, bt->prop_seg, > + bt->phase_seg1, bt->phase_seg2, bt->sjw); > + > + /* BRP: 9 bits @ bits 8..0 */ > + brp &= 0x1ff; > + > + /* TSEG1: 4 bits @ bits 3..0 */ > + btr = (bt->phase_seg1 + bt->prop_seg - 1) & 0xf; > + /* TSEG2: 3 bits @ bits 18..16*/ > + btr |= ((bt->phase_seg2 - 1) & 0x7) << 16; > + /* SJW: 2 bits @ bits 25..24 */ > + btr |= ((bt->sjw - 1) & 0x3) << 24; > + > + /* Keep order of accesses to ACC_CORE_OF_BRP and ACC_CORE_OF_BTR. */ > + acc_write32(priv->core, ACC_CORE_OF_BRP, brp); > + acc_write32(priv->core, ACC_CORE_OF_BTR, btr); > + > + netdev_info(priv->core->netdev, "ESDACC: BRP %u, BTR 0x%08x", > + brp, btr); same here > + } > + > + return 0; > +} > + > +static void handle_core_msg_rxtxdone(struct acc_core *core, > + const struct acc_bmmsg_rxtxdone *msg) > +{ > + struct acc_net_priv *priv = netdev_priv(core->netdev); > + struct net_device_stats *stats = &core->netdev->stats; > + struct sk_buff *skb; > + > + if (msg->dlc.rxtx.len & ACC_BM_LENFLAG_TX) { > + u8 tx_fifo_tail = core->tx_fifo_tail; > + > + if (core->tx_fifo_head == tx_fifo_tail) { > + netdev_warn(core->netdev, > + "TX interrupt, but queue is empty!?\n"); > + return; > + } > + > + /* Direct access echo skb to attach HW time stamp. */ > + skb = priv->can.echo_skb[tx_fifo_tail]; > + if (skb) { > + skb_hwtstamps(skb)->hwtstamp = > + acc_ts2ktime(priv->ov, msg->ts); > + } > + > + stats->tx_packets++; > + stats->tx_bytes += can_get_echo_skb(core->netdev, tx_fifo_tail, > + NULL); > + > + tx_fifo_tail++; > + if (tx_fifo_tail >= core->tx_fifo_size) > + tx_fifo_tail = 0U; > + core->tx_fifo_tail = tx_fifo_tail; > + > + netif_wake_queue(core->netdev); You only need to wake the queue if there is space available, and you need a barrier here, see mcp251xfd_handle_tefif(): | int mcp251xfd_handle_tefif(struct mcp251xfd_priv *priv) | { [...] | if (mcp251xfd_get_tx_free(priv->tx)) { | /* Make sure that anybody stopping the queue after | * this sees the new tx_ring->tail. | */ | smp_mb(); | netif_wake_queue(priv->ndev); | } | | return 0; | | } > + > + } else { > + struct can_frame *cf; > + > + skb = alloc_can_skb(core->netdev, &cf); > + if (!skb) { > + stats->rx_dropped++; > + return; > + } > + > + cf->can_id = msg->id & CAN_EFF_MASK; > + if (msg->id & ACC_CAN_EFF_FLAG) > + cf->can_id |= CAN_EFF_FLAG; > + > + can_frame_set_cc_len(cf, msg->dlc.rx.len & ACC_CAN_DLC_MASK, > + priv->can.ctrlmode); > + > + if (msg->dlc.rx.len & ACC_CAN_RTR_FLAG) > + cf->can_id |= CAN_RTR_FLAG; > + else > + memcpy(cf->data, msg->data, cf->len); > + > + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); > + > + stats->rx_packets++; > + stats->rx_bytes += cf->len; > + netif_rx(skb); Consider moving the TX completion and RX handling into NAPI or make use of rx-offload, to avoid reordering of CAN frames in the RX path. > + } > +} > + > +static void handle_core_msg_txabort(struct acc_core *core, > + const struct acc_bmmsg_txabort *msg) > +{ > + struct net_device_stats *stats = &core->netdev->stats; > + u8 tx_fifo_tail = core->tx_fifo_tail; > + u32 abort_mask = msg->abort_mask; /* u32 extend to avoid warnings later */ > + > + /* The abort_mask shows which frames were aborted in ESDACC's FIFO. */ You can directly iterate over all set bits using for_each_set_bit(). > + while (tx_fifo_tail != core->tx_fifo_head && (abort_mask)) { > + const u32 tail_mask = (1U << tx_fifo_tail); > + > + if (!(abort_mask & tail_mask)) > + break; > + abort_mask &= ~tail_mask; > + > + can_free_echo_skb(core->netdev, tx_fifo_tail, NULL); > + stats->tx_dropped++; > + stats->tx_aborted_errors++; > + > + tx_fifo_tail++; > + if (tx_fifo_tail >= core->tx_fifo_size) > + tx_fifo_tail = 0; > + } > + core->tx_fifo_tail = tx_fifo_tail; > + if (abort_mask) > + netdev_warn(core->netdev, "Unhandled aborted messages\n"); > + > + if (!acc_resetmode_entered(core)) > + netif_wake_queue(core->netdev); > +} > + > +static void handle_core_msg_overrun(struct acc_core *core, > + const struct acc_bmmsg_overrun *msg) > +{ > + struct acc_net_priv *priv = netdev_priv(core->netdev); > + struct net_device_stats *stats = &core->netdev->stats; > + struct can_frame *cf; > + struct sk_buff *skb; > + > + skb = alloc_can_err_skb(core->netdev, &cf); > + if (!skb) { > + stats->rx_dropped++; > + return; > + } Please handle the stats, even if the driver fails to allocate and error skb. > + > + /* lost_cnt may be 0 if not supported by ESDACC version */ > + if (msg->lost_cnt) { > + stats->rx_errors += msg->lost_cnt; > + stats->rx_over_errors += msg->lost_cnt; > + } else { > + stats->rx_errors++; > + stats->rx_over_errors++; > + } > + > + cf->can_id |= CAN_ERR_CRTL; > + cf->data[1] = CAN_ERR_CRTL_RX_OVERFLOW; > + > + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); > + > + stats->rx_packets++; > + stats->rx_bytes += cf->len; > + netif_rx(skb); > +} > + > +static void handle_core_msg_buserr(struct acc_core *core, > + const struct acc_bmmsg_buserr *msg) > +{ > + struct acc_net_priv *priv = netdev_priv(core->netdev); > + struct net_device_stats *stats = &core->netdev->stats; > + struct can_frame *cf; > + struct sk_buff *skb; > + const u32 reg_status = msg->reg_status; > + const u8 rxerr = (u8)reg_status; > + const u8 txerr = (u8)(reg_status >> 8); casts not needed > + > + priv->can.can_stats.bus_error++; > + > + skb = alloc_can_err_skb(core->netdev, &cf); > + if (!skb) { > + stats->rx_dropped++; > + return; > + } Please handle the stats, even if the driver fails to allocate and error skb. > + > + cf->can_id |= CAN_ERR_PROT | CAN_ERR_BUSERROR; > + > + /* msg->ecc acts like SJA1000's ECC register */ > + switch (msg->ecc & ACC_ECC_MASK) { > + case ACC_ECC_BIT: > + cf->data[2] |= CAN_ERR_PROT_BIT; > + break; > + case ACC_ECC_FORM: > + cf->data[2] |= CAN_ERR_PROT_FORM; > + break; > + case ACC_ECC_STUFF: > + cf->data[2] |= CAN_ERR_PROT_STUFF; > + break; > + default: > + cf->data[2] |= CAN_ERR_PROT_UNSPEC; > + break; > + } > + > + /* Set error location */ > + cf->data[3] = msg->ecc & ACC_ECC_SEG; > + > + /* Error occurred during transmission? */ > + if ((msg->ecc & ACC_ECC_DIR) == 0) { > + cf->data[2] |= CAN_ERR_PROT_TX; > + stats->tx_errors++; > + } else { > + stats->rx_errors++; > + } > + /* Insert CAN TX and RX error counters. */ > + cf->data[6] = txerr; > + cf->data[7] = rxerr; > + > + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); > + > + stats->rx_packets++; > + stats->rx_bytes += cf->len; > + netif_rx(skb); > +} > + > +static void > +handle_core_msg_errstatechange(struct acc_core *core, > + const struct acc_bmmsg_errstatechange *msg) > +{ > + struct acc_net_priv *priv = netdev_priv(core->netdev); > + struct net_device_stats *stats = &core->netdev->stats; > + struct can_frame *cf = NULL; > + struct sk_buff *skb; > + const u32 reg_status = msg->reg_status; > + const u8 rxerr = (u8)reg_status; > + const u8 txerr = (u8)(reg_status >> 8); cast not needed > + enum can_state new_state; > + > + if (reg_status & ACC_REG_STATUS_MASK_STATUS_BS) { > + new_state = CAN_STATE_BUS_OFF; > + } else if (reg_status & ACC_REG_STATUS_MASK_STATUS_EP) { > + new_state = CAN_STATE_ERROR_PASSIVE; > + } else if (reg_status & ACC_REG_STATUS_MASK_STATUS_ES) { > + new_state = CAN_STATE_ERROR_WARNING; > + } else { > + new_state = CAN_STATE_ERROR_ACTIVE; > + if (priv->can.state == CAN_STATE_BUS_OFF) { > + /* See comment in acc_set_mode() for CAN_MODE_START */ > + netif_wake_queue(core->netdev); See my other comment. Shut down the chip in case of bus off. > + } > + } > + > + skb = alloc_can_err_skb(core->netdev, &cf); > + > + if (new_state != priv->can.state) { > + enum can_state tx_state, rx_state; > + > + tx_state = (txerr >= rxerr) ? > + new_state : CAN_STATE_ERROR_ACTIVE; > + rx_state = (rxerr >= txerr) ? > + new_state : CAN_STATE_ERROR_ACTIVE; > + > + /* Always call can_change_state() to update the state > + * even if alloc_can_err_skb() may have failed. > + * can_change_state() can cope with a NULL cf pointer. > + */ > + can_change_state(core->netdev, cf, tx_state, rx_state); > + } > + > + if (skb) { > + cf->data[6] = txerr; > + cf->data[7] = rxerr; > + > + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); > + > + stats->rx_packets++; > + stats->rx_bytes += cf->len; > + netif_rx(skb); > + } else { > + stats->rx_dropped++; > + } > + > + if (new_state == CAN_STATE_BUS_OFF) { > + acc_write32(core, ACC_CORE_OF_TX_ABORT_MASK, 0xffff); > + can_bus_off(core->netdev); > + } > +} > + > +static void handle_core_interrupt(struct acc_core *core) > +{ > + u32 msg_fifo_head = core->bmfifo.local_irq_cnt & 0xff; > + > + while (core->bmfifo.msg_fifo_tail != msg_fifo_head) { > + const struct acc_bmmsg *msg = > + &core->bmfifo.messages[core->bmfifo.msg_fifo_tail]; > + > + switch (msg->u.msg_id) { > + case BM_MSG_ID_RXTXDONE: > + handle_core_msg_rxtxdone(core, &msg->u.rxtxdone); > + break; > + > + case BM_MSG_ID_TXABORT: > + handle_core_msg_txabort(core, &msg->u.txabort); > + break; > + > + case BM_MSG_ID_OVERRUN: > + handle_core_msg_overrun(core, &msg->u.overrun); > + break; > + > + case BM_MSG_ID_BUSERR: > + handle_core_msg_buserr(core, &msg->u.buserr); > + break; > + > + case BM_MSG_ID_ERRPASSIVE: > + case BM_MSG_ID_ERRWARN: > + handle_core_msg_errstatechange(core, > + &msg->u.errstatechange); > + break; > + > + default: > + /* Ignore all other BM messages (like the CAN-FD messages) */ > + break; > + } > + > + core->bmfifo.msg_fifo_tail = > + (core->bmfifo.msg_fifo_tail + 1) & 0xff; > + } > +} > + > +irqreturn_t acc_card_interrupt(struct acc_ov *ov, struct acc_core *cores) > +{ > + u32 irqmask; > + int i; single space here > + > + /* First we look for whom interrupts are pending, card/overview > + * or any of the cores. Two bits in irqmask are used for each; > + * set to ACC_BM_IRQ_MASK then: > + */ Please explain a bit more detailed how the interrupt handling works. > + irqmask = 0; > + if (*ov->bmfifo.irq_cnt != ov->bmfifo.local_irq_cnt) { > + irqmask |= ACC_BM_IRQ_MASK; > + ov->bmfifo.local_irq_cnt = *ov->bmfifo.irq_cnt; > + } > + > + for (i = 0; i < ov->active_cores; i++) { > + struct acc_core *core = &cores[i]; > + > + if (*core->bmfifo.irq_cnt != core->bmfifo.local_irq_cnt) { > + irqmask |= (ACC_BM_IRQ_MASK << (2 * (i + 1))); > + core->bmfifo.local_irq_cnt = *core->bmfifo.irq_cnt; > + } > + } > + > + if (!irqmask) > + return IRQ_NONE; > + > + /* At second we tell the card we're working on them by writing irqmask, > + * call handle_{ov|core}_interrupt and then acknowledge the > + * interrupts by writing irq_cnt: > + */ > + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_MASK, irqmask); > + > + if (irqmask & ACC_BM_IRQ_MASK) { > + /* handle_ov_interrupt(); - no use yet. */ > + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_COUNTER, > + ov->bmfifo.local_irq_cnt); > + } > + > + for (i = 0; i < ov->active_cores; i++) { > + struct acc_core *core = &cores[i]; > + > + if (irqmask & (ACC_BM_IRQ_MASK << (2 * (i + 1)))) { > + handle_core_interrupt(core); > + acc_write32(core, ACC_OV_OF_BM_IRQ_COUNTER, > + core->bmfifo.local_irq_cnt); > + } > + } > + > + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_MASK, ACC_BM_IRQ_UNMASK_ALL); > + > + return IRQ_HANDLED; > +} > diff --git a/drivers/net/can/esd/esdacc.h b/drivers/net/can/esd/esdacc.h > new file mode 100644 > index 000000000000..3e865ececb3e > --- /dev/null > +++ b/drivers/net/can/esd/esdacc.h > @@ -0,0 +1,380 @@ > +/* SPDX-License-Identifier: GPL-2.0-only */ > +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh > + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh > + */ > + > +#include <linux/kernel.h> > +#include <linux/netdevice.h> > +#include <linux/can/dev.h> > + > +#define ACC_CAN_EFF_FLAG 0x20000000 > +#define ACC_CAN_RTR_FLAG 0x10 > +#define ACC_CAN_DLC_MASK 0x0f > + > +#define ACC_OV_OF_PROBE 0x0000 > +#define ACC_OV_OF_VERSION 0x0004 > +#define ACC_OV_OF_INFO 0x0008 > +#define ACC_OV_OF_CANCORE_FREQ 0x000c > +#define ACC_OV_OF_TS_FREQ_LO 0x0010 > +#define ACC_OV_OF_TS_FREQ_HI 0x0014 > +#define ACC_OV_OF_IRQ_STATUS_CORES 0x0018 > +#define ACC_OV_OF_TS_CURR_LO 0x001c > +#define ACC_OV_OF_TS_CURR_HI 0x0020 > +#define ACC_OV_OF_IRQ_STATUS 0x0028 > +#define ACC_OV_OF_MODE 0x002c > +#define ACC_OV_OF_BM_IRQ_COUNTER 0x0070 > +#define ACC_OV_OF_BM_IRQ_MASK 0x0074 > +#define ACC_OV_OF_MSI_DATA 0x0080 > +#define ACC_OV_OF_MSI_ADDRESSOFFSET 0x0084 > + > +/* Feature flags are contained in the upper 16 bit of the version > + * register at ACC_OV_OF_VERSION but only used with these masks after > + * extraction into an extra variable => (xx - 16). > + */ > +#define ACC_OV_REG_FEAT_IDX_CANFD (27 - 16) > +#define ACC_OV_REG_FEAT_IDX_NEW_PSC (28 - 16) > +#define ACC_OV_REG_FEAT_MASK_CANFD BIT(ACC_OV_REG_FEAT_IDX_CANFD) > +#define ACC_OV_REG_FEAT_MASK_NEW_PSC BIT(ACC_OV_REG_FEAT_IDX_NEW_PSC) > + > +#define ACC_OV_REG_MODE_MASK_ENDIAN_LITTLE 0x00000001 > +#define ACC_OV_REG_MODE_MASK_BM_ENABLE 0x00000002 > +#define ACC_OV_REG_MODE_MASK_MODE_LED 0x00000004 > +#define ACC_OV_REG_MODE_MASK_TIMER 0x00000070 > +#define ACC_OV_REG_MODE_MASK_TIMER_ENABLE 0x00000010 > +#define ACC_OV_REG_MODE_MASK_TIMER_ONE_SHOT 0x00000020 > +#define ACC_OV_REG_MODE_MASK_TIMER_ABSOLUTE 0x00000040 > +#define ACC_OV_REG_MODE_MASK_TS_SRC 0x00000180 > +#define ACC_OV_REG_MODE_MASK_I2C_ENABLE 0x00000800 > +#define ACC_OV_REG_MODE_MASK_MSI_ENABLE 0x00004000 > +#define ACC_OV_REG_MODE_MASK_NEW_PSC_ENABLE 0x00008000 > +#define ACC_OV_REG_MODE_MASK_FPGA_RESET 0x80000000 > + > +#define ACC_CORE_OF_CTRL_MODE 0x0000 > +#define ACC_CORE_OF_STATUS_IRQ 0x0008 > +#define ACC_CORE_OF_BRP 0x000c > +#define ACC_CORE_OF_BTR 0x0010 > +#define ACC_CORE_OF_FBTR 0x0014 > +#define ACC_CORE_OF_STATUS 0x0030 > +#define ACC_CORE_OF_TXFIFO_CONFIG 0x0048 > +#define ACC_CORE_OF_TXFIFO_STATUS 0x004c > +#define ACC_CORE_OF_TX_STATUS_IRQ 0x0050 > +#define ACC_CORE_OF_TX_ABORT_MASK 0x0054 > +#define ACC_CORE_OF_BM_IRQ_COUNTER 0x0070 > +#define ACC_CORE_OF_TXFIFO_ID 0x00c0 > +#define ACC_CORE_OF_TXFIFO_DLC 0x00c4 > +#define ACC_CORE_OF_TXFIFO_DATA_0 0x00c8 > +#define ACC_CORE_OF_TXFIFO_DATA_1 0x00cc > + > +#define ACC_REG_CONTROL_IDX_MODE_RESETMODE 0 > +#define ACC_REG_CONTROL_IDX_MODE_LOM 1 > +#define ACC_REG_CONTROL_IDX_MODE_STM 2 > +#define ACC_REG_CONTROL_IDX_MODE_TRANSEN 5 > +#define ACC_REG_CONTROL_IDX_MODE_TS 6 > +#define ACC_REG_CONTROL_IDX_MODE_SCHEDULE 7 > +#define ACC_REG_CONTROL_MASK_MODE_RESETMODE \ > + BIT(ACC_REG_CONTROL_IDX_MODE_RESETMODE) > +#define ACC_REG_CONTROL_MASK_MODE_LOM \ > + BIT(ACC_REG_CONTROL_IDX_MODE_LOM) > +#define ACC_REG_CONTROL_MASK_MODE_STM \ > + BIT(ACC_REG_CONTROL_IDX_MODE_STM) > +#define ACC_REG_CONTROL_MASK_MODE_TRANSEN \ > + BIT(ACC_REG_CONTROL_IDX_MODE_TRANSEN) > +#define ACC_REG_CONTROL_MASK_MODE_TS \ > + BIT(ACC_REG_CONTROL_IDX_MODE_TS) > +#define ACC_REG_CONTROL_MASK_MODE_SCHEDULE \ > + BIT(ACC_REG_CONTROL_IDX_MODE_SCHEDULE) > + > +#define ACC_REG_CONTROL_IDX_IE_RXTX 8 > +#define ACC_REG_CONTROL_IDX_IE_TXERROR 9 > +#define ACC_REG_CONTROL_IDX_IE_ERRWARN 10 > +#define ACC_REG_CONTROL_IDX_IE_OVERRUN 11 > +#define ACC_REG_CONTROL_IDX_IE_TSI 12 > +#define ACC_REG_CONTROL_IDX_IE_ERRPASS 13 > +#define ACC_REG_CONTROL_IDX_IE_BUSERR 15 > +#define ACC_REG_CONTROL_MASK_IE_RXTX BIT(ACC_REG_CONTROL_IDX_IE_RXTX) > +#define ACC_REG_CONTROL_MASK_IE_TXERROR BIT(ACC_REG_CONTROL_IDX_IE_TXERROR) > +#define ACC_REG_CONTROL_MASK_IE_ERRWARN BIT(ACC_REG_CONTROL_IDX_IE_ERRWARN) > +#define ACC_REG_CONTROL_MASK_IE_OVERRUN BIT(ACC_REG_CONTROL_IDX_IE_OVERRUN) > +#define ACC_REG_CONTROL_MASK_IE_TSI BIT(ACC_REG_CONTROL_IDX_IE_TSI) > +#define ACC_REG_CONTROL_MASK_IE_ERRPASS BIT(ACC_REG_CONTROL_IDX_IE_ERRPASS) > +#define ACC_REG_CONTROL_MASK_IE_BUSERR BIT(ACC_REG_CONTROL_IDX_IE_BUSERR) > + > +/* 256 BM_MSGs of 32 byte size */ > +#define ACC_CORE_DMAMSG_SIZE 32U > +#define ACC_CORE_DMABUF_SIZE (256U * ACC_CORE_DMAMSG_SIZE) > + > +enum acc_bmmsg_id { > + BM_MSG_ID_RXTXDONE = 0x01, > + BM_MSG_ID_TXABORT = 0x02, > + BM_MSG_ID_OVERRUN = 0x03, > + BM_MSG_ID_BUSERR = 0x04, > + BM_MSG_ID_ERRPASSIVE = 0x05, > + BM_MSG_ID_ERRWARN = 0x06, > + BM_MSG_ID_TIMESLICE = 0x07, > + BM_MSG_ID_HWTIMER = 0x08, > + BM_MSG_ID_HOTPLUG = 0x09, > +}; > + > +/* The struct acc_bmmsg* structure declarations that follow here provide > + * access to the ring buffer of bus master messages maintained by the FPGA > + * bus master engine. All bus master messages have the same size of > + * ACC_CORE_DMAMSG_SIZE and a minimum alignment of ACC_CORE_DMAMSG_SIZE in > + * memory. > + * > + * All structure members are natural aligned. Therefore we should not need > + * a __packed attribute. All struct acc_bmmsg* declarations have at least > + * reserved* members to fill the structure to the full ACC_CORE_DMAMSG_SIZE. > + * > + * A failure of this property due padding will be detected at compile time > + * by BUILD_BUG_ON(sizeof(struct acc_bmmsg) != ACC_CORE_DMAMSG_SIZE) > + */ You can use static_assert() directly unter the definition of the struct. > + > +struct acc_bmmsg_rxtxdone { > + u8 msg_id; > + u8 txfifo_level; > + u8 reserved1[2]; > + u8 txtsfifo_level; > + u8 reserved2[3]; > + u32 id; > + union { > + struct { > + u8 len; > + u8 reserved0; > + u8 bits; > + u8 state; > + } rxtx; > + struct { > + u8 len; > + u8 msg_lost; > + u8 bits; > + u8 state; > + } rx; > + struct { > + u8 len; > + u8 txfifo_idx; > + u8 bits; > + u8 state; > + } tx; > + } dlc; > + u8 data[8]; > + /* Time stamps in struct acc_ov::timestamp_frequency ticks. */ > + u64 ts; > +}; > + > +struct acc_bmmsg_txabort { > + u8 msg_id; > + u8 txfifo_level; > + u16 abort_mask; > + u8 txtsfifo_level; > + u8 reserved2[1]; > + u16 abort_mask_txts; > + u64 ts; > + u32 reserved3[4]; > +}; > + > +struct acc_bmmsg_overrun { > + u8 msg_id; > + u8 txfifo_level; > + u8 lost_cnt; > + u8 reserved1; > + u8 txtsfifo_level; > + u8 reserved2[3]; > + u64 ts; > + u32 reserved3[4]; > +}; > + > +struct acc_bmmsg_buserr { > + u8 msg_id; > + u8 txfifo_level; > + u8 ecc; > + u8 reserved1; > + u8 txtsfifo_level; > + u8 reserved2[3]; > + u64 ts; > + u32 reg_status; > + u32 reg_btr; > + u32 reserved3[2]; > +}; > + > +struct acc_bmmsg_errstatechange { > + u8 msg_id; > + u8 txfifo_level; > + u8 reserved1[2]; > + u8 txtsfifo_level; > + u8 reserved2[3]; > + u64 ts; > + u32 reg_status; > + u32 reserved3[3]; > +}; > + > +struct acc_bmmsg_timeslice { > + u8 msg_id; > + u8 txfifo_level; > + u8 reserved1[2]; > + u8 txtsfifo_level; > + u8 reserved2[3]; > + u64 ts; > + u32 reserved3[4]; > +}; > + > +struct acc_bmmsg_hwtimer { > + u8 msg_id; > + u8 reserved1[3]; > + u32 reserved2[1]; > + u64 timer; > + u32 reserved3[4]; > +}; > + > +struct acc_bmmsg_hotplug { > + u8 msg_id; > + u8 reserved1[3]; > + u32 reserved2[7]; > +}; > + > +struct acc_bmmsg { > + union { > + u8 msg_id; > + struct acc_bmmsg_rxtxdone rxtxdone; > + struct acc_bmmsg_txabort txabort; > + struct acc_bmmsg_overrun overrun; > + struct acc_bmmsg_buserr buserr; > + struct acc_bmmsg_errstatechange errstatechange; > + struct acc_bmmsg_timeslice timeslice; > + struct acc_bmmsg_hwtimer hwtimer; > + } u; > +}; Why don't you use the union directly instead of putting it into a struct? > + > +/* Regarding Documentation/process/volatile-considered-harmful.rst the > + * forth exception applies to the "irq_cnt" member of the structure > + * below. The u32 variable "irq_cnt" points to is updated by the ESDACC > + * FPGA via DMA. > + */ > +struct acc_bmfifo { > + const struct acc_bmmsg *messages; > + /* Bits 0..7: bm_fifo head index */ > + volatile const u32 *irq_cnt; > + u32 local_irq_cnt; > + u32 msg_fifo_tail; > +}; > + > +struct acc_core { > + void __iomem *addr; > + struct net_device *netdev; > + struct acc_bmfifo bmfifo; > + u8 tx_fifo_size; > + u8 tx_fifo_head; > + u8 tx_fifo_tail; Is the tx_fifo_size a power of two? If so it's usually easier to use a unsigned int for the head and tail pointers and mask them to their real values. head and tail pointers only ever increase and eventually roll over. As both are unsigned there's no problem if head has rolled over, but the tail doesn't. When you read tail pointer from HW you round_down() to core->tx_fifo_size, add the tail value from HW and add core->tx_fifo_size if it's lower than the old value. This is the only time where you have to take care of rollover. > +}; > + > +struct acc_ov { > + void __iomem *addr; > + struct acc_bmfifo bmfifo; > + u32 timestamp_frequency; > + u32 ts2ns_numerator; > + u32 ts2ns_denominator; > + u32 core_frequency; > + u16 version; > + u16 features; > + u8 total_cores; > + u8 active_cores; > +}; > + > +struct acc_net_priv { > + struct can_priv can; /* must be the first member! */ > + struct acc_core *core; > + struct acc_ov *ov; > +}; > + > +static inline u32 acc_read32(struct acc_core *core, unsigned short offs) > +{ > + return ioread32be(core->addr + offs); > +} > + > +static inline void acc_write32(struct acc_core *core, > + unsigned short offs, u32 v) > +{ > + iowrite32be(v, core->addr + offs); > +} > + > +static inline void acc_write32_noswap(struct acc_core *core, > + unsigned short offs, u32 v) > +{ > + iowrite32(v, core->addr + offs); > +} > + > +static inline void acc_sset_bits(struct acc_core *core, > + unsigned short offs, u32 mask) > +{ > + u32 v = acc_read32(core, offs); > + > + v |= mask; > + acc_write32(core, offs, v); > +} > + > +static inline void acc_clear_bits(struct acc_core *core, > + unsigned short offs, u32 mask) > +{ > + u32 v = acc_read32(core, offs); > + > + v &= ~mask; > + acc_write32(core, offs, v); > +} > + > +static inline int acc_resetmode_entered(struct acc_core *core) > +{ > + u32 ctrl = acc_read32(core, ACC_CORE_OF_CTRL_MODE); > + > + return (ctrl & ACC_REG_CONTROL_MASK_MODE_RESETMODE) != 0; > +} > + > +static inline u32 acc_ov_read32(struct acc_ov *ov, unsigned short offs) > +{ > + return ioread32be(ov->addr + offs); > +} > + > +static inline void acc_ov_write32(struct acc_ov *ov, > + unsigned short offs, u32 v) > +{ > + iowrite32be(v, ov->addr + offs); > +} > + > +static inline void acc_ov_set_bits(struct acc_ov *ov, > + unsigned short offs, u32 b) > +{ > + u32 v = acc_ov_read32(ov, offs); > + > + v |= b; > + acc_ov_write32(ov, offs, v); > +} > + > +static inline void acc_ov_clear_bits(struct acc_ov *ov, > + unsigned short offs, u32 b) > +{ > + u32 v = acc_ov_read32(ov, offs); > + > + v &= ~b; > + acc_ov_write32(ov, offs, v); > +} > + > +static inline void acc_reset_fpga(struct acc_ov *ov) > +{ > + acc_ov_write32(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_FPGA_RESET); > + > + /* Also reset I^2C, to re-detect card addons at every driver start: */ > + acc_ov_clear_bits(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_I2C_ENABLE); > + mdelay(2); > + acc_ov_set_bits(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_I2C_ENABLE); > + mdelay(10); > +} > + > +void acc_init_ov(struct acc_ov *ov, struct device *dev); > +void acc_init_bm_ptr(struct acc_ov *ov, struct acc_core *cores, > + const void *mem); > +int acc_open(struct net_device *netdev); > +int acc_close(struct net_device *netdev); > +netdev_tx_t acc_start_xmit(struct sk_buff *skb, struct net_device *netdev); > +int acc_get_berr_counter(const struct net_device *netdev, > + struct can_berr_counter *bec); > +int acc_set_mode(struct net_device *netdev, enum can_mode mode); > +int acc_set_bittiming(struct net_device *netdev); > +irqreturn_t acc_card_interrupt(struct acc_ov *ov, struct acc_core *cores); > -- > 2.25.1 > > regards, Marc -- Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 488 bytes --] ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v6 2/4] can: esd: add support for esd GmbH PCIe/402 CAN interface family 2022-02-01 17:25 ` [PATCH v6 " Marc Kleine-Budde @ 2022-09-30 22:15 ` Stefan Mätje 2022-10-05 16:04 ` Marc Kleine-Budde 0 siblings, 1 reply; 15+ messages in thread From: Stefan Mätje @ 2022-09-30 22:15 UTC (permalink / raw) To: mkl; +Cc: netdev, linux-kernel, linux-can, wg Hi Marc, I would like to resume the efforts to bring this driver in the Linux kernel after being kicked off track in February by another project. I did a lot of the changes on the driver you recommended, but some stuff is not yet clear to me. Please see my comments in-line of the email below. My local developement is at the moment rebased to linux-can-next:master on 7b584fbb36362340a2d9cfe459e447619eecebea. Should I send a V7 of the patch (rebased to another commit)? How should I proceed? You have commented on many type casts that they would not be needed. But all of them had been introduced by me due to warnings of the compiler in the style of "warning: conversion from ‘u32’ {aka ‘unsigned int’} to ‘u8’ {aka ‘unsigned char’} may change value [-Wconversion]". These are triggered by building the driver with "W=3" as recommended in kernel documentation. Should these warnings generally be ignored and the casts be removed then? Best regards, Stefan Am Dienstag, den 01.02.2022, 18:25 +0100 schrieb Marc Kleine-Budde: > On 01.12.2021 23:03:26, Stefan Mätje wrote: > > This patch adds support for the PCI based PCIe/402 CAN interface family > > from esd GmbH that is available with various form factors > > (https://esd.eu/technologie/can/can-und-can-fd-interfaces). > > > > All boards utilize a FPGA based CAN controller solution developed > > by esd (esdACC). For more information on the esdACC see > > https://esd.eu/technologie/can/esd-advanced-can-controller-esdacc. > > > > This driver detects all available CAN interface board variants of > > the family but atm. operates the CAN-FD capable devices in > > Classic-CAN mode only! A later patch will introduce the CAN-FD > > functionality in this driver. > > > > Co-developed-by: Thomas Körper <thomas.koerper@esd.eu> > > Signed-off-by: Thomas Körper <thomas.koerper@esd.eu> > > Signed-off-by: Stefan Mätje <stefan.maetje@esd.eu> > > --- > > drivers/net/can/Kconfig | 1 + > > drivers/net/can/Makefile | 1 + > > drivers/net/can/esd/Kconfig | 12 + > > drivers/net/can/esd/Makefile | 7 + > > drivers/net/can/esd/esd_402_pci-core.c | 502 ++++++++++++++++ > > drivers/net/can/esd/esdacc.c | 777 +++++++++++++++++++++++++ > > drivers/net/can/esd/esdacc.h | 380 ++++++++++++ > > 7 files changed, 1680 insertions(+) > > create mode 100644 drivers/net/can/esd/Kconfig > > create mode 100644 drivers/net/can/esd/Makefile > > create mode 100644 drivers/net/can/esd/esd_402_pci-core.c > > create mode 100644 drivers/net/can/esd/esdacc.c > > create mode 100644 drivers/net/can/esd/esdacc.h > > > > diff --git a/drivers/net/can/Kconfig b/drivers/net/can/Kconfig > > index fff259247d52..47cfb6ae0772 100644 > > --- a/drivers/net/can/Kconfig > > +++ b/drivers/net/can/Kconfig > > @@ -170,6 +170,7 @@ config PCH_CAN > > > > source "drivers/net/can/c_can/Kconfig" > > source "drivers/net/can/cc770/Kconfig" > > +source "drivers/net/can/esd/Kconfig" > > source "drivers/net/can/ifi_canfd/Kconfig" > > source "drivers/net/can/m_can/Kconfig" > > source "drivers/net/can/mscan/Kconfig" > > diff --git a/drivers/net/can/Makefile b/drivers/net/can/Makefile > > index a2b4463d8480..015b6fc110d1 100644 > > --- a/drivers/net/can/Makefile > > +++ b/drivers/net/can/Makefile > > @@ -8,6 +8,7 @@ obj-$(CONFIG_CAN_VXCAN) += vxcan.o > > obj-$(CONFIG_CAN_SLCAN) += slcan.o > > > > obj-y += dev/ > > +obj-y += esd/ > > obj-y += rcar/ > > obj-y += spi/ > > obj-y += usb/ > > diff --git a/drivers/net/can/esd/Kconfig b/drivers/net/can/esd/Kconfig > > new file mode 100644 > > index 000000000000..54bfc366634c > > --- /dev/null > > +++ b/drivers/net/can/esd/Kconfig > > @@ -0,0 +1,12 @@ > > +# SPDX-License-Identifier: GPL-2.0-only > > +config CAN_ESD_402_PCI > > + tristate "esd electronics gmbh CAN-PCI(e)/402 family" > > + depends on PCI && HAS_DMA > > + help > > + Support for C402 card family from esd electronics gmbh. > > + This card family is based on the ESDACC CAN controller and > > + available in several form factors: PCI, PCIe, PCIe Mini, > > + M.2 PCIe, CPCIserial, PMC, XMC (see https://esd.eu/en) > > + > > + This driver can also be built as a module. In this case the > > + module will be called esd_402_pci. > > diff --git a/drivers/net/can/esd/Makefile b/drivers/net/can/esd/Makefile > > new file mode 100644 > > index 000000000000..5dd2d470c286 > > --- /dev/null > > +++ b/drivers/net/can/esd/Makefile > > @@ -0,0 +1,7 @@ > > +# SPDX-License-Identifier: GPL-2.0-only > > +# > > +# Makefile for esd gmbh ESDACC controller driver > > +# > > +esd_402_pci-objs := esdacc.o esd_402_pci-core.o > > + > > +obj-$(CONFIG_CAN_ESD_402_PCI) += esd_402_pci.o > > diff --git a/drivers/net/can/esd/esd_402_pci-core.c b/drivers/net/can/esd/esd_402_pci-core.c > > new file mode 100644 > > index 000000000000..80d816a78859 > > --- /dev/null > > +++ b/drivers/net/can/esd/esd_402_pci-core.c > > @@ -0,0 +1,502 @@ > > +// SPDX-License-Identifier: GPL-2.0-only > > +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh > > + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh > > + */ > > + > > +#include <linux/kernel.h> > > +#include <linux/module.h> > > +#include <linux/interrupt.h> > > +#include <linux/pci.h> > > +#include <linux/io.h> > > +#include <linux/delay.h> > > +#include <linux/netdevice.h> > > +#include <linux/dma-mapping.h> > > +#include <linux/can.h> > > +#include <linux/can/dev.h> > > +#include <linux/can/netlink.h> > > please sort alphabetically done > > > + > > +#include "esdacc.h" > > + > > +#define DRV_NAME "esd_402_pci" There has been some effort on the mailing list to replace hardcoded driver names with KBUILD_MODNAME. But then this showed up as debateable (see https://lore.kernel.org/linux-can/YyLkrj5v2EiBIXk+@surfacebook/#r) How should I proceed with this DRV_NAME (i.e. keep it use KBUILD_MODNAME) which is only used to set struct pci_driver::name? > > + > > +#define ESD_PCI_DEVICE_ID_PCIE402 0x0402 > > + > > +#define PCI402_FPGA_VER_MIN 0x003d > > +#define PCI402_MAX_CORES 6 > > +#define PCI402_BAR 0 > > +#define PCI402_IO_OV_OFFS 0 > > +#define PCI402_IO_PCIEP_OFFS 0x10000 > > +#define PCI402_IO_LEN_TOTAL 0x20000 > > +#define PCI402_IO_LEN_CORE 0x2000 > > +#define PCI402_PCICFG_MSICAP_CSR 0x52 > > +#define PCI402_PCICFG_MSICAP_ADDR 0x54 > > +#define PCI402_PCICFG_MSICAP_DATA 0x5c > > + > > +#define PCI402_DMA_MASK DMA_BIT_MASK(32) > > +#define PCI402_DMA_SIZE ALIGN(0x10000, PAGE_SIZE) > > + > > +#define PCI402_PCIEP_OF_INT_ENABLE 0x0050 > > +#define PCI402_PCIEP_OF_BM_ADDR_LO 0x1000 > > +#define PCI402_PCIEP_OF_BM_ADDR_HI 0x1004 > > +#define PCI402_PCIEP_OF_MSI_ADDR_LO 0x1008 > > +#define PCI402_PCIEP_OF_MSI_ADDR_HI 0x100c > > nitpick: please use single space for indention. Sooner or later new > stuff is added and the original alignment doesn't fit anymore. Changed to single space indentation. > > + > > +/* The BTR register capabilities described by the can_bittiming_const structures > > + * below are valid since ESDACC version 0x0032. > > + */ > > + > > +/* Used if the ESDACC FPGA is built as CAN-Classic version. */ > > +static const struct can_bittiming_const pci402_bittiming_const = { > > + .name = "esd_402", > > + .tseg1_min = 1, > > + .tseg1_max = 16, > > + .tseg2_min = 1, > > + .tseg2_max = 8, > > + .sjw_max = 4, > > + .brp_min = 1, > > + .brp_max = 512, > > + .brp_inc = 1, > > +}; > > + > > +/* Used if the ESDACC FPGA is built as CAN-FD version. */ > > +static const struct can_bittiming_const pci402_bittiming_const_canfd = { > > + .name = "esd_402fd", > > + .tseg1_min = 1, > > + .tseg1_max = 256, > > + .tseg2_min = 1, > > + .tseg2_max = 128, > > + .sjw_max = 128, > > + .brp_min = 1, > > + .brp_max = 256, > > + .brp_inc = 1, > > +}; > > + > > +static const struct net_device_ops pci402_acc_netdev_ops = { > > + .ndo_open = acc_open, > > + .ndo_stop = acc_close, > > + .ndo_start_xmit = acc_start_xmit, > > + .ndo_change_mtu = can_change_mtu > > +}; > > + > > +struct pci402_card { > > + /* Actually mapped io space, all other iomem derived from this */ > > + void __iomem *addr; > > + void __iomem *addr_pciep; > > + > > + void *dma_buf; > > + dma_addr_t dma_hnd; > > + > > + struct acc_ov ov; > > + struct acc_core *cores; > > + > > + bool msi_enabled; > > +}; > > + > > +static irqreturn_t pci402_interrupt(int irq, void *dev_id) > > +{ > > + struct pci_dev *pdev = dev_id; > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + irqreturn_t irq_status; > > + > > + irq_status = acc_card_interrupt(&card->ov, card->cores); > > + > > + return irq_status; > > +} > > + > > +static int pci402_set_msiconfig(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + u32 addr_lo_offs = 0; > > + u32 addr_lo = 0; > > + u32 addr_hi = 0; > > + u32 data = 0; > > + u16 csr = 0; > > + int err; > > + > > + err = pci_read_config_word(pdev, PCI402_PCICFG_MSICAP_CSR, &csr); > > + if (err) > > + goto failed; > > + > > + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_ADDR, &addr_lo); > > + if (err) > > + goto failed; > > + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_ADDR + 4, > > + &addr_hi); > > + if (err) > > + goto failed; > > + > > + err = pci_read_config_dword(pdev, PCI402_PCICFG_MSICAP_DATA, &data); > > + if (err) > > + goto failed; > > + > > + addr_lo_offs = addr_lo & 0x0000ffff; > > + addr_lo &= 0xffff0000; > > + > > + if (addr_hi) > > + addr_lo |= 1; /* Enable 64-Bit addressing in address space */ > > + > > + if (!(csr & 0x0001)) { /* Enable bit */ > > + err = -EINVAL; > > + goto failed; > > + } > > + > > + iowrite32(addr_lo, card->addr_pciep + PCI402_PCIEP_OF_MSI_ADDR_LO); > > + iowrite32(addr_hi, card->addr_pciep + PCI402_PCIEP_OF_MSI_ADDR_HI); > > + acc_ov_write32(&card->ov, ACC_OV_OF_MSI_ADDRESSOFFSET, addr_lo_offs); > > + acc_ov_write32(&card->ov, ACC_OV_OF_MSI_DATA, data); > > + > > + return 0; > > + > > +failed: > > + pci_warn(pdev, "Error while setting MSI configuration:\n" > > + "CSR: 0x%.4x, addr: 0x%.8x%.8x, data: 0x%.8x\n", > > + csr, addr_hi, addr_lo, data); > > + > > + return err; > > +} Reworked pci402_set_msiconfig() to use defines from pci.h to access MSI capability registers. > > + > > +static int pci402_init_card(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + > > + card->ov.addr = card->addr + PCI402_IO_OV_OFFS; > > + card->addr_pciep = card->addr + PCI402_IO_PCIEP_OFFS; > > + > > + acc_reset_fpga(&card->ov); > > + acc_init_ov(&card->ov, &pdev->dev); > > + > > + if (card->ov.version < PCI402_FPGA_VER_MIN) { > > + pci_err(pdev, > > + "ESDACC version (0x%.4x) outdated, please update\n", > > + card->ov.version); > > + return -EINVAL; > > + } > > + > > + if (card->ov.active_cores > PCI402_MAX_CORES) { > > + pci_warn(pdev, > > + "Card has more active cores than supported by driver, %u core(s) will be ignored\n", > > + card->ov.active_cores - PCI402_MAX_CORES); > > + card->ov.active_cores = PCI402_MAX_CORES; > > Where is that limitation in this driver? PCI402_MAX_CORES is only used here. Most of the structures are allocated dynamically except the size of the DMA area which is fixed. I decided now to abort the initialization of the driver with ov.active_cores > PCI402_MAX_CORES because it can't work properly with so many cores because it would overflow the allocated DMA memory area. PCI402_MAX_CORES is derived from a hard FPGA limit of the current implementation. > > + } > > + card->cores = devm_kcalloc(&pdev->dev, card->ov.active_cores, > > + sizeof(struct acc_core), GFP_KERNEL); > > + if (!card->cores) > > + return -ENOMEM; > > + > > + if (card->ov.features & ACC_OV_REG_FEAT_MASK_CANFD) { > > + pci_warn(pdev, > > + "ESDACC with CAN-FD feature detected. This driver doesn't support CAN-FD yet.\n"); > > + } > > + > > +#ifdef __LITTLE_ENDIAN > > + /* So card converts all busmastered data to LE for us: */ > > + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, > > + ACC_OV_REG_MODE_MASK_ENDIAN_LITTLE); > > +#endif > > + > > + return 0; > > +} > > + > > +static int pci402_init_interrupt(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + int err; > > + > > + err = pci_enable_msi(pdev); > > + if (!err) { > > + err = pci402_set_msiconfig(pdev); > > + if (!err) { > > + card->msi_enabled = true; > > + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, > > + ACC_OV_REG_MODE_MASK_MSI_ENABLE); > > + pci_info(pdev, "MSI enabled\n"); > > + } > > + } > > + > > + err = devm_request_irq(&pdev->dev, pdev->irq, pci402_interrupt, > > + IRQF_SHARED, dev_name(&pdev->dev), pdev); > > + if (err) > > + goto failure_msidis; > > + > > + iowrite32(1, card->addr_pciep + PCI402_PCIEP_OF_INT_ENABLE); > > + > > + return 0; > > + > > +failure_msidis: > > + if (card->msi_enabled) { > > + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, > > + ACC_OV_REG_MODE_MASK_MSI_ENABLE); > > + pci_disable_msi(pdev); > > + card->msi_enabled = false; > > + } > > + > > + return err; > > +} > > + > > +static void pci402_finish_interrupt(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + > > + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_INT_ENABLE); > > + devm_free_irq(&pdev->dev, pdev->irq, pdev); > > A devm manged interrupt gets automatically freed, it makes no sense if > you free it manually here. I think it's best to use a non devm IRQ > instead. When I switched the memory allocation to devm_* allocations I also switched the IRQ request to a devm_* request to not to mix devm_* resource allocation with non-devm_* resource allocation. Should I again use request_irq() / free_irq() pair even in the face of this imbalance? > > + > > + if (card->msi_enabled) { > > + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, > > + ACC_OV_REG_MODE_MASK_MSI_ENABLE); > > + pci_disable_msi(pdev); > > + card->msi_enabled = false; > > + } > > +} > > + > > +static int pci402_init_dma(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + int err; > > + > > + err = dma_set_coherent_mask(&pdev->dev, PCI402_DMA_MASK); > > + if (err) { > > + pci_err(pdev, "DMA set mask failed!\n"); > > + return err; > > + } > > + > > + /* The ESDACC DMA engine needs the DMA buffer aligned to a 64k > > + * boundary. The DMA API guarantees to align the returned buffer to the > > + * smallest PAGE_SIZE order which is greater than or equal to the > > + * requested size. With PCI402_DMA_SIZE == 64kB this suffices here. > > + */ > > + card->dma_buf = dma_alloc_coherent(&pdev->dev, PCI402_DMA_SIZE, > > + &card->dma_hnd, GFP_ATOMIC); > > Why do you use ATOMIC here? Originally a pci_alloc_consistent() was used, which in turn uses GFP_ATOMIC. When pci_alloc_consistent() was replaced GFP_ATOMIC was inherited without deeper thought. Since I believe we are able to sleep during the PCI probe routine when the allocation happens GFP_KERNEL should be ok too. Changed it to GFP_KERNEL. > > + if (!card->dma_buf) { > > + pci_err(pdev, "DMA alloc failed!\n"); > > + return -ENOMEM; > > + } > > + > > + acc_init_bm_ptr(&card->ov, card->cores, card->dma_buf); > > + > > + iowrite32((u32)card->dma_hnd, > > + card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_LO); > > cast not needed See comment at top of email. > > > + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_HI); > > + > > + pci_set_master(pdev); > > + > > + acc_ov_set_bits(&card->ov, ACC_OV_OF_MODE, > > + ACC_OV_REG_MODE_MASK_BM_ENABLE); > > + > > + return 0; > > +} > > + > > +static void pci402_finish_dma(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + int i; > > + > > + acc_ov_clear_bits(&card->ov, ACC_OV_OF_MODE, > > + ACC_OV_REG_MODE_MASK_BM_ENABLE); > > + > > + pci_clear_master(pdev); > > + > > + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_LO); > > + iowrite32(0, card->addr_pciep + PCI402_PCIEP_OF_BM_ADDR_HI); > > + > > + card->ov.bmfifo.messages = NULL; > > + card->ov.bmfifo.irq_cnt = NULL; > > + for (i = 0; i < card->ov.active_cores; i++) { > > + struct acc_core *core = &card->cores[i]; > > + > > + core->bmfifo.messages = NULL; > > + core->bmfifo.irq_cnt = NULL; > > + } > > + > > + dma_free_coherent(&pdev->dev, PCI402_DMA_SIZE, card->dma_buf, > > + card->dma_hnd); > > + card->dma_buf = NULL; > > +} > > + > > +static int pci402_init_cores(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + int err; > > + int i; > > + > > + for (i = 0; i < card->ov.active_cores; i++) { > > + struct acc_core *core = &card->cores[i]; > > + struct acc_net_priv *priv; > > + struct net_device *netdev; > > + u32 fifo_config; > > + > > + core->addr = card->ov.addr + (i + 1) * PCI402_IO_LEN_CORE; > > + > > + fifo_config = acc_read32(core, ACC_CORE_OF_TXFIFO_CONFIG); > > + core->tx_fifo_size = (u8)(fifo_config >> 24); > > cast not needed. > See comment at top of email. > > > + if (core->tx_fifo_size <= 1) { > > + pci_err(pdev, "Invalid tx_fifo_size!\n"); > > + err = -EINVAL; > > + goto failure; > > + } > > + > > + netdev = alloc_candev(sizeof(*priv), core->tx_fifo_size); > > + if (!netdev) { > > + err = -ENOMEM; > > + goto failure; > > + } > > + core->netdev = netdev; > > + > > + netdev->flags |= IFF_ECHO; > > + netdev->dev_port = i; > > + netdev->netdev_ops = &pci402_acc_netdev_ops; > > + SET_NETDEV_DEV(netdev, &pdev->dev); > > + > > + priv = netdev_priv(netdev); > > + priv->can.state = CAN_STATE_STOPPED; > > The state is automatically set by alloc_candev(). Removed this assignment. > > + priv->can.ctrlmode_supported = CAN_CTRLMODE_LISTENONLY | > > + CAN_CTRLMODE_BERR_REPORTING | > > + CAN_CTRLMODE_CC_LEN8_DLC | > > + CAN_CTRLMODE_LOOPBACK; > > Please sort these by occurrence in can/netlink.h Sorted. > > + priv->can.clock.freq = card->ov.core_frequency; > > + priv->can.bittiming_const = > > + (card->ov.features & ACC_OV_REG_FEAT_MASK_CANFD) ? > > + &pci402_bittiming_const_canfd : > > + &pci402_bittiming_const; > > nitpick: > Please use if (card->ov.features & ACC_OV_REG_FEAT_MASK_CANFD)...else here Removed the ternary operator and replaced by if else clause. > > + priv->can.do_set_bittiming = acc_set_bittiming; > > Please call it directly from the open() callback. Unchanged till now. Need to revisit later... > > + priv->can.do_set_mode = acc_set_mode; > > + priv->can.do_get_berr_counter = acc_get_berr_counter; > > + > > + priv->core = core; > > + priv->ov = &card->ov; > > + > > + err = register_candev(netdev); > > + if (err) { > > + free_candev(core->netdev); > > + core->netdev = NULL; > > + goto failure; > > + } > > + > > + netdev_info(netdev, "registered\n"); > > + } > > + > > + return 0; > > + > > +failure: > > + for (i--; i >= 0; i--) { > > + struct acc_core *core = &card->cores[i]; > > + > > + netdev_info(core->netdev, "unregistering...\n"); > > + unregister_candev(core->netdev); > > + > > + free_candev(core->netdev); > > + core->netdev = NULL; > > + } > > + > > + return err; > > +} > > + > > +static void pci402_finish_cores(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + int i; > > + > > + for (i = 0; i < card->ov.active_cores; i++) { > > + struct acc_core *core = &card->cores[i]; > > + > > + netdev_info(core->netdev, "unregister\n"); > > + unregister_candev(core->netdev); > > + > > + free_candev(core->netdev); > > + core->netdev = NULL; > > + } > > +} > > + > > +static int pci402_probe(struct pci_dev *pdev, const struct pci_device_id *ent) > > +{ > > + struct pci402_card *card = NULL; > > + int err; > > + > > + err = pci_enable_device(pdev); > > + if (err) > > + return err; > > + > > + card = devm_kzalloc(&pdev->dev, sizeof(*card), GFP_KERNEL); > > + if (!card) > > + goto failure_disable_pci; > > + > > + pci_set_drvdata(pdev, card); > > + > > + err = pci_request_regions(pdev, pci_name(pdev)); > > + if (err) > > + goto failure_disable_pci; > > + > > + card->addr = pci_iomap(pdev, PCI402_BAR, PCI402_IO_LEN_TOTAL); > > + if (!card->addr) { > > + err = -ENOMEM; > > + goto failure_release_regions; > > + } > > + > > + err = pci402_init_card(pdev); > > + if (err) > > + goto failure_unmap; > > + > > + err = pci402_init_dma(pdev); > > + if (err) > > + goto failure_unmap; > > + > > + err = pci402_init_interrupt(pdev); > > + if (err) > > + goto failure_finish_dma; > > + > > + err = pci402_init_cores(pdev); > > + if (err) > > + goto failure_finish_interrupt; > > + > > + return 0; > > + > > +failure_finish_interrupt: > > + pci402_finish_interrupt(pdev); > > + > > +failure_finish_dma: > > + pci402_finish_dma(pdev); > > + > > +failure_unmap: > > + pci_iounmap(pdev, card->addr); > > + > > +failure_release_regions: > > + pci_release_regions(pdev); > > + > > +failure_disable_pci: > > + pci_disable_device(pdev); > > + > > + return err; > > +} > > + > > +static void pci402_remove(struct pci_dev *pdev) > > +{ > > + struct pci402_card *card = pci_get_drvdata(pdev); > > + > > + pci402_finish_interrupt(pdev); > > + pci402_finish_cores(pdev); > > + pci402_finish_dma(pdev); > > + pci_iounmap(pdev, card->addr); > > + pci_release_regions(pdev); > > + pci_disable_device(pdev); > > +} > > + > > +static const struct pci_device_id pci402_tbl[] = { > > + { PCI_VENDOR_ID_ESDGMBH, ESD_PCI_DEVICE_ID_PCIE402, > > + PCI_VENDOR_ID_ESDGMBH, PCI_ANY_ID, > > + 0U, 0U, 0UL }, > > + { 0, } > > +}; > > +MODULE_DEVICE_TABLE(pci, pci402_tbl); > > + > > +static struct pci_driver pci402_driver = { > > + .name = DRV_NAME, > > + .id_table = pci402_tbl, > > + .probe = pci402_probe, > > + .remove = pci402_remove, > > +}; > > + > > +module_pci_driver(pci402_driver); > > + > > +MODULE_DESCRIPTION("Socket-CAN driver for esd CAN 402 card family with esdACC core on PCIe"); > > +MODULE_AUTHOR("Thomas Körper <socketcan@esd.eu>"); > > +MODULE_AUTHOR("Stefan Mätje <stefan.maetje@esd.eu>"); > > +MODULE_LICENSE("GPL"); > > diff --git a/drivers/net/can/esd/esdacc.c b/drivers/net/can/esd/esdacc.c > > new file mode 100644 > > index 000000000000..13f7397dfc4e > > --- /dev/null > > +++ b/drivers/net/can/esd/esdacc.c > > @@ -0,0 +1,777 @@ > > +// SPDX-License-Identifier: GPL-2.0-only > > +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh > > + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh > > + */ > > + > > +#include <linux/ktime.h> > > +#include <linux/gcd.h> > > +#include <linux/io.h> > > +#include <linux/delay.h> > > please sort alphabetically Done > > +#include "esdacc.h" > > + > > +/* ecc value of esdACC equals SJA1000's ECC register */ > > +#define ACC_ECC_SEG 0x1f > > +#define ACC_ECC_DIR 0x20 > > +#define ACC_ECC_BIT 0x00 > > +#define ACC_ECC_FORM 0x40 > > +#define ACC_ECC_STUFF 0x80 > > +#define ACC_ECC_MASK 0xc0 > > + > > +#define ACC_BM_IRQ_UNMASK_ALL 0x55555555U > > +#define ACC_BM_IRQ_MASK_ALL 0xaaaaaaaaU > > +#define ACC_BM_IRQ_MASK 0x2U > > +#define ACC_BM_IRQ_UNMASK 0x1U > > +#define ACC_BM_LENFLAG_TX 0x20 > > + > > +#define ACC_REG_STATUS_IDX_STATUS_DOS 16 > > +#define ACC_REG_STATUS_IDX_STATUS_ES 17 > > +#define ACC_REG_STATUS_IDX_STATUS_EP 18 > > +#define ACC_REG_STATUS_IDX_STATUS_BS 19 > > +#define ACC_REG_STATUS_IDX_STATUS_RBS 20 > > +#define ACC_REG_STATUS_IDX_STATUS_RS 21 > > +#define ACC_REG_STATUS_MASK_STATUS_DOS BIT(ACC_REG_STATUS_IDX_STATUS_DOS) > > +#define ACC_REG_STATUS_MASK_STATUS_ES BIT(ACC_REG_STATUS_IDX_STATUS_ES) > > +#define ACC_REG_STATUS_MASK_STATUS_EP BIT(ACC_REG_STATUS_IDX_STATUS_EP) > > +#define ACC_REG_STATUS_MASK_STATUS_BS BIT(ACC_REG_STATUS_IDX_STATUS_BS) > > +#define ACC_REG_STATUS_MASK_STATUS_RBS BIT(ACC_REG_STATUS_IDX_STATUS_RBS) > > +#define ACC_REG_STATUS_MASK_STATUS_RS BIT(ACC_REG_STATUS_IDX_STATUS_RS) Should these defines also be reindented using only a single space? > > + > > +static void acc_resetmode_enter(struct acc_core *core) > > +{ > > + int i; > > + > > + acc_set_bits(core, ACC_CORE_OF_CTRL_MODE, > > + ACC_REG_CONTROL_MASK_MODE_RESETMODE); > > + > > + for (i = 0; i < 10; i++) { > > + if (acc_resetmode_entered(core)) > > + return; > > + > > + udelay(5); > > + } > > + > > + netdev_warn(core->netdev, "Entering reset mode timed out\n"); > > +} > > + > > +static void acc_resetmode_leave(struct acc_core *core) > > +{ > > + int i; > > + > > + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, > > + ACC_REG_CONTROL_MASK_MODE_RESETMODE); > > + > > + for (i = 0; i < 10; i++) { > > + if (!acc_resetmode_entered(core)) > > + return; > > + > > + udelay(5); > > + } > > + > > + netdev_warn(core->netdev, "Leaving reset mode timed out\n"); > > +} After consultation with the FPGA people the acc_resetmode_enter() and acc_resetmode_leave() will be reworked. > > + > > +static void acc_txq_put(struct acc_core *core, u32 acc_id, u8 acc_dlc, > > + const void *data) > > +{ > > + acc_write32_noswap(core, ACC_CORE_OF_TXFIFO_DATA_1, > > + *((const u32 *)(data + 4))); > > + acc_write32_noswap(core, ACC_CORE_OF_TXFIFO_DATA_0, > > + *((const u32 *)data)); > > + acc_write32(core, ACC_CORE_OF_TXFIFO_DLC, acc_dlc); > > + /* CAN id must be written at last. This write starts TX. */ > > + acc_write32(core, ACC_CORE_OF_TXFIFO_ID, acc_id); > > +} > > + > > +/* Prepare conversion factor from ESDACC time stamp ticks to ns > > + * > > + * The conversion factor ts2ns from time stamp counts to ns is basically > > + * ts2ns = NSEC_PER_SEC / timestamp_frequency > > + * > > + * To avoid an overflow, the ts2ns fraction is truncated with its gcd and > > + * only the truncated numerator and denominator are used further. > > + */ > > +static void acc_init_ov_ts2ns(struct acc_ov *ov) > > +{ > > + u32 ts2ns_gcd = (u32)gcd(NSEC_PER_SEC, ov->timestamp_frequency); > > + > > + ov->ts2ns_numerator = (u32)NSEC_PER_SEC / ts2ns_gcd; > > + ov->ts2ns_denominator = ov->timestamp_frequency / ts2ns_gcd; > > +} > > Please don't craft your own time conversion functions. please use > cyclecounter/timercounter instead. > > Have a look at: > drivers/net/can/spi/mcp251xfd/mcp251xfd-timestamp.c > > > + > > +static ktime_t acc_ts2ktime(struct acc_ov *ov, u64 ts) > > +{ > > + u64 ns; > > + > > + ts = ts * ov->ts2ns_numerator; > > + ns = div_u64(ts, ov->ts2ns_denominator); > > + > > + return ns_to_ktime(ns); > > +} The time conversion is done here. The esdACC provides a full 64-bit timestamp with 80MHz resolution. This timestamp will wrap after(!) the kernel time ktime_t. To convert from 80MHz a multiplication with 12.5 is needed which is written here overly complicated to cope with other possible timestamp frequencies. After all only 40MHz and 80MHz are likely because these are the recommended clock frequencies for CAN-FD capable controllers. So this could be simplified to the code below. #define ESDACC_TS_FACTOR 25 #define ESDACC_TS_80MHZ_SHIFT 1 #define ESDACC_TS_40MHZ_SHIFT 0 static void acc_init_ov_ts2ns(struct acc_ov *ov, u32 timestamp_frequency) { if (80 * MHZ == timestamp_frequency) ov->ts2ns_shift = ESDACC_TS_80MHZ_SHIFT; else if (40 * MHZ == timestamp_frequency) ov->ts2ns_shift = ESDACC_TS_40MHZ_SHIFT; } static ktime_t acc_ts2ktime(struct acc_ov *ov, u64 ts) { u64 ns; ts = ts * ESDACC_TS_FACTOR; ns = ts >> ov->ts2ns_shift; return ns_to_ktime(ns); } > > + > > +void acc_init_ov(struct acc_ov *ov, struct device *dev) > > +{ > > + u32 temp; > > + /* For the justification of this see comment on struct acc_bmmsg* > > + * in esdacc.h. > > + */ > > + BUILD_BUG_ON(sizeof(struct acc_bmmsg) != ACC_CORE_DMAMSG_SIZE); > > + > > + temp = acc_ov_read32(ov, ACC_OV_OF_VERSION); > > + ov->version = (u16)temp; > > + ov->features = (u16)(temp >> 16); > > casts not needed See comment at top of email. > > + > > + temp = acc_ov_read32(ov, ACC_OV_OF_INFO); > > + ov->total_cores = (u8)temp; > > + ov->active_cores = (u8)(temp >> 8); > > casts not needed See comment at top of email. > > + > > + ov->core_frequency = acc_ov_read32(ov, ACC_OV_OF_CANCORE_FREQ); > > + ov->timestamp_frequency = acc_ov_read32(ov, ACC_OV_OF_TS_FREQ_LO); > > + acc_init_ov_ts2ns(ov); > > + > > + /* Depending on ESDACC feature NEW_PSC enable the new prescaler > > + * or adjust core_frequency according to the implicit division by 2. > > + */ > > + if (ov->features & ACC_OV_REG_FEAT_MASK_NEW_PSC) { > > + acc_ov_set_bits(ov, ACC_OV_OF_MODE, > > + ACC_OV_REG_MODE_MASK_NEW_PSC_ENABLE); > > + } else { > > + ov->core_frequency /= 2; > > + } > > + > > + dev_info(dev, > > + "ESDACC v%u, freq: %u/%u, feat/strap: 0x%x/0x%x, cores: %u/%u\n", > > + ov->version, ov->core_frequency, ov->timestamp_frequency, > > + ov->features, acc_ov_read32(ov, ACC_OV_OF_INFO) >> 16, > > + ov->active_cores, ov->total_cores); > > + dev_dbg(dev, "ESDACC ts2ns: numerator %u, denominator %u\n", > > + ov->ts2ns_numerator, ov->ts2ns_denominator); > > +} > > + > > +void acc_init_bm_ptr(struct acc_ov *ov, struct acc_core *cores, const void *mem) > > +{ > > + unsigned int u; > > + > > + /* DMA buffer layout as follows where N is the number of CAN cores > > + * implemented in the FPGA, i.e. N = ov->total_cores > > + * > > + * Layout Section size > > + * +-----------------------+ > > + * | FIFO Card/Overview | ACC_CORE_DMABUF_SIZE > > + * | | > > + * +-----------------------+ > > + * | FIFO Core0 | ACC_CORE_DMABUF_SIZE > > + * | | > > + * +-----------------------+ > > + * | ... | ... > > + * | | > > + * +-----------------------+ > > + * | FIFO CoreN | ACC_CORE_DMABUF_SIZE > > + * | | > > + * +-----------------------+ > > + * | irq_cnt Card/Overview | sizeof(u32) > > + * +-----------------------+ > > + * | irq_cnt Core0 | sizeof(u32) > > + * +-----------------------+ > > + * | ... | ... > > + * +-----------------------+ > > + * | irq_cnt CoreN | sizeof(u32) > > + * +-----------------------+ > > + */ > > + ov->bmfifo.messages = mem; > > + ov->bmfifo.irq_cnt = mem + (ov->total_cores + 1U) * ACC_CORE_DMABUF_SIZE; > > + > > + for (u = 0U; u < ov->active_cores; u++) { > > + struct acc_core *core = &cores[u]; > > + > > + core->bmfifo.messages = mem + (u + 1U) * ACC_CORE_DMABUF_SIZE; > > + core->bmfifo.irq_cnt = ov->bmfifo.irq_cnt + (u + 1U); > > + } > > +} > > + > > +int acc_open(struct net_device *netdev) > > +{ > > + struct acc_net_priv *priv = netdev_priv(netdev); > > + struct acc_core *core = priv->core; > > + u32 ctrl_mode; > > + int err; > > + > > + /* Retry to enter RESET mode if out of sync. */ > > + if (priv->can.state != CAN_STATE_STOPPED) { > > + netdev_warn(netdev, "Entered %s() with bad can.state: %s\n", > > + __func__, can_get_state_str(priv->can.state)); > > + acc_resetmode_enter(core); > > + if (acc_resetmode_entered(core)) > > + priv->can.state = CAN_STATE_STOPPED; > > + } > > What about always doing a full reset during open(). It's not quite clear to me what you mean with a "full reset" during open. The acc_resetmode_enter() switches the FPGA into a state the is equivalent to the "reset mode of a SJA1000" i. e. the CAN controller is off bus and configurable. But due to the rework of the acc_resetmode_enter() function this code will be changed indeed. > > > + > > + err = open_candev(netdev); > > + if (err) > > + return err; > > + > > + ctrl_mode = ACC_REG_CONTROL_MASK_IE_RXTX | > > + ACC_REG_CONTROL_MASK_IE_TXERROR | > > + ACC_REG_CONTROL_MASK_IE_ERRWARN | > > + ACC_REG_CONTROL_MASK_IE_OVERRUN | > > + ACC_REG_CONTROL_MASK_IE_ERRPASS; > > + > > + if (priv->can.ctrlmode & CAN_CTRLMODE_BERR_REPORTING) > > + ctrl_mode |= ACC_REG_CONTROL_MASK_IE_BUSERR; > > + > > + if (priv->can.ctrlmode & CAN_CTRLMODE_LISTENONLY) > > + ctrl_mode |= ACC_REG_CONTROL_MASK_MODE_LOM; > > + > > + acc_set_bits(core, ACC_CORE_OF_CTRL_MODE, ctrl_mode); > > + > > + acc_resetmode_leave(core); > > + if (!acc_resetmode_entered(core)) > > + priv->can.state = CAN_STATE_ERROR_ACTIVE; > > If the device didn't leave reset mode, then return with an error. > > > + /* Resync TX FIFO indices to HW state after (re-)start. */ > > + { > > + u32 tx_fifo_status = acc_read32(core, ACC_CORE_OF_TXFIFO_STATUS); > > + > > + core->tx_fifo_head = tx_fifo_status & 0xff; > > + core->tx_fifo_tail = (tx_fifo_status >> 8) & 0xff; > > + } > > Why this extra level of indention? This was only done to introduce the variable tx_fifo_status late in the function. But the variable declaration is now at the top of the function. > > + netif_start_queue(netdev); > > + return 0; > > +} > > + > > +int acc_close(struct net_device *netdev) > > +{ > > + struct acc_net_priv *priv = netdev_priv(netdev); > > + struct acc_core *core = priv->core; > > + > > + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, > > + ACC_REG_CONTROL_MASK_IE_RXTX | > > + ACC_REG_CONTROL_MASK_IE_TXERROR | > > + ACC_REG_CONTROL_MASK_IE_ERRWARN | > > + ACC_REG_CONTROL_MASK_IE_OVERRUN | > > + ACC_REG_CONTROL_MASK_IE_ERRPASS | > > + ACC_REG_CONTROL_MASK_IE_BUSERR); > > + > > + netif_stop_queue(netdev); > > + acc_resetmode_enter(core); > > + if (acc_resetmode_entered(core)) > > + priv->can.state = CAN_STATE_STOPPED; > > Better mark the device as stopped, and make full reset during open() What do you mean with full reset? > > > + > > + /* Mark pending TX requests to be aborted after controller restart. */ > > + acc_write32(core, ACC_CORE_OF_TX_ABORT_MASK, 0xffff); > > + > > + /* ACC_REG_CONTROL_MASK_MODE_LOM is only accessible in RESET mode */ > > + acc_clear_bits(core, ACC_CORE_OF_CTRL_MODE, > > + ACC_REG_CONTROL_MASK_MODE_LOM); > > + > > + close_candev(netdev); > > + return 0; > > +} > > + > > +netdev_tx_t acc_start_xmit(struct sk_buff *skb, struct net_device *netdev) > > +{ > > + struct acc_net_priv *priv = netdev_priv(netdev); > > + struct acc_core *core = priv->core; > > + struct can_frame *cf = (struct can_frame *)skb->data; > > + u8 tx_fifo_head = core->tx_fifo_head; > > + int fifo_usage; > > + u32 acc_id; > > + u8 acc_dlc; > > Please add a check for can_dropped_invalid_skb(). Added like in other drivers. > > + > > + /* Access core->tx_fifo_tail only once because it may be changed > > + * from the interrupt level. > > + */ > > + fifo_usage = tx_fifo_head - core->tx_fifo_tail; > > + if (fifo_usage < 0) > > + fifo_usage += core->tx_fifo_size; > > + > > + if (fifo_usage >= core->tx_fifo_size - 1) { > > + netdev_err(core->netdev, > > + "BUG: TX ring full when queue awake!\n"); > > + netif_stop_queue(netdev); > > + return NETDEV_TX_BUSY; > > + } > > + > > + if (fifo_usage == core->tx_fifo_size - 2) > > + netif_stop_queue(netdev); I need to revisit this later. > You need proper memory barriers and double checking: See the mcp251xfd > driver: > > > netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, > > struct net_device *ndev) > > { > > struct mcp251xfd_priv *priv = netdev_priv(ndev); > > struct mcp251xfd_tx_ring *tx_ring = priv->tx; > > struct mcp251xfd_tx_obj *tx_obj; > > unsigned int frame_len; > > u8 tx_head; > > int err; > > > > if (can_dropped_invalid_skb(ndev, skb)) > > return NETDEV_TX_OK; > > > > if (mcp251xfd_tx_busy(priv, tx_ring)) > > return NETDEV_TX_BUSY; > > The magic happens in mcp251xfd_tx_busy(): > > > static bool mcp251xfd_tx_busy(const struct mcp251xfd_priv *priv, > > struct mcp251xfd_tx_ring *tx_ring) > > { > > if (mcp251xfd_get_tx_free(tx_ring) > 0) > > return false; > > > > netif_stop_queue(priv->ndev); > > > > /* Memory barrier before checking tx_free (head and tail) */ > > smp_mb(); > > > > if (mcp251xfd_get_tx_free(tx_ring) == 0) { > > If you need at least 2 free slots, adjust this check. > > > netdev_dbg(priv->ndev, > > "Stopping tx-queue (tx_head=0x%08x, tx_tail=0x%08x, len=%d).\n", > > tx_ring->head, tx_ring->tail, > > tx_ring->head - tx_ring->tail); > > > > return true; > > } > > > > netif_start_queue(priv->ndev); > > > > return false; > > } > > + > > + acc_dlc = can_get_cc_dlc(cf, priv->can.ctrlmode); > > + if (cf->can_id & CAN_RTR_FLAG) > > + acc_dlc |= ACC_CAN_RTR_FLAG; > > + > > + if (cf->can_id & CAN_EFF_FLAG) { > > + acc_id = cf->can_id & CAN_EFF_MASK; > > + acc_id |= ACC_CAN_EFF_FLAG; > > + } else { > > + acc_id = cf->can_id & CAN_SFF_MASK; > > + } > > + > > + can_put_echo_skb(skb, netdev, core->tx_fifo_head, 0); > > + > > + tx_fifo_head++; > > + if (tx_fifo_head >= core->tx_fifo_size) > > + tx_fifo_head = 0U; > > + core->tx_fifo_head = tx_fifo_head; > > + > > + acc_txq_put(core, acc_id, acc_dlc, cf->data); > > + > > + return NETDEV_TX_OK; > > +} > > + > > +int acc_get_berr_counter(const struct net_device *netdev, > > + struct can_berr_counter *bec) > > +{ > > + struct acc_net_priv *priv = netdev_priv(netdev); > > + u32 core_status = acc_read32(priv->core, ACC_CORE_OF_STATUS); > > + > > + bec->txerr = (core_status >> 8) & 0xff; > > + bec->rxerr = core_status & 0xff; > > + > > + return 0; > > +} > > + > > +int acc_set_mode(struct net_device *netdev, enum can_mode mode) > > +{ > > + struct acc_net_priv *priv = netdev_priv(netdev); > > + > > + switch (mode) { > > + case CAN_MODE_START: > > + /* Paranoid FIFO index check. */ > > + { > > + const u32 tx_fifo_status = > > + acc_read32(priv->core, ACC_CORE_OF_TXFIFO_STATUS); > > + const u8 hw_fifo_head = (u8)tx_fifo_status; > > cast not needed See comment at top of email. > > + > > + if (hw_fifo_head != priv->core->tx_fifo_head || > > + hw_fifo_head != priv->core->tx_fifo_tail) { > > + netdev_warn(netdev, > > + "TX FIFO mismatch: T %2u H %2u; TFHW %#08x\n", > > + priv->core->tx_fifo_tail, > > + priv->core->tx_fifo_head, > > + tx_fifo_status); > > + } > > + } > > + acc_resetmode_leave(priv->core); > > + /* To leave the bus-off state the esdACC controller begins > > + * here a grace period where it counts 128 "idle conditions" (each > > + * of 11 consecutive recessive bits) on the bus as required > > + * by the CAN spec. > > + * > > + * During this time the TX FIFO may still contain already > > + * aborted "zombie" frames that are only drained from the FIFO > > + * at the end of the grace period. > > + * > > + * To not to interfere with this drain process we don't > > + * call netif_wake_queue() here. When the controller reaches > > + * the error-active state again, it informs us about that > > + * with an acc_bmmsg_errstatechange message. Then > > + * netif_wake_queue() is called from > > + * handle_core_msg_errstatechange() instead. > > + */ > > + break; > > Due 128 idle condition auto recovery it's best to shut down/reset the > controller on BUS_OFF and re-start it here. It turned out to be less > complex form the Linux driver point of view and the individual CAN cores > feel much more consistent: Bus off means bus off and recovery is > controlled by the kernel. > > > + > > + default: > > + return -EOPNOTSUPP; > > + } > > + > > + return 0; > > +} > > + > > +int acc_set_bittiming(struct net_device *netdev) > > +{ > > + struct acc_net_priv *priv = netdev_priv(netdev); > > + const struct can_bittiming *bt = &priv->can.bittiming; > > + u32 brp = bt->brp - 1; > > + u32 btr; > > + > > + if (priv->ov->features & ACC_OV_REG_FEAT_MASK_CANFD) { > > + u32 fbtr = 0; > > + > > + netdev_dbg(priv->core->netdev, > > + "bit timing: brp %u, prop %u, ph1 %u ph2 %u, sjw %u\n", > > + bt->brp, bt->prop_seg, > > + bt->phase_seg1, bt->phase_seg2, bt->sjw); > > + > > + /* BRP: 8 bits @ bits 7..0 */ > > + brp &= 0xff; > > + > > + /* TSEG1: 8 bits @ bits 7..0 */ > > + btr = (bt->phase_seg1 + bt->prop_seg - 1) & 0xff; > > + /* TSEG2: 7 bits @ bits 22..16 */ > > + btr |= ((bt->phase_seg2 - 1) & 0x7f) << 16; > > + /* SJW: 7 bits @ bits 30..24 */ > > + btr |= ((bt->sjw - 1) & 0x7f) << 24; > > Please make use of FIELD_PREP() here. Use now FIELD_PREP() > > + > > + /* Keep order of accesses to ACC_CORE_OF_BRP and ACC_CORE_OF_BTR. */ > > + acc_write32(priv->core, ACC_CORE_OF_BRP, brp); > > + acc_write32(priv->core, ACC_CORE_OF_BTR, btr); > > + > > + netdev_info(priv->core->netdev, > > + "ESDACC: BRP %u, NBTR 0x%08x, DBTR 0x%08x", > > + brp, btr, fbtr); > > please make this a netdev_dbg() Changed it. > > + } else { > > + netdev_dbg(priv->core->netdev, > > + "bit timing: brp %u, prop %u, ph1 %u ph2 %u, sjw %u\n", > > + bt->brp, bt->prop_seg, > > + bt->phase_seg1, bt->phase_seg2, bt->sjw); > > + > > + /* BRP: 9 bits @ bits 8..0 */ > > + brp &= 0x1ff; > > + > > + /* TSEG1: 4 bits @ bits 3..0 */ > > + btr = (bt->phase_seg1 + bt->prop_seg - 1) & 0xf; > > + /* TSEG2: 3 bits @ bits 18..16*/ > > + btr |= ((bt->phase_seg2 - 1) & 0x7) << 16; > > + /* SJW: 2 bits @ bits 25..24 */ > > + btr |= ((bt->sjw - 1) & 0x3) << 24; > > + > > + /* Keep order of accesses to ACC_CORE_OF_BRP and ACC_CORE_OF_BTR. */ > > + acc_write32(priv->core, ACC_CORE_OF_BRP, brp); > > + acc_write32(priv->core, ACC_CORE_OF_BTR, btr); > > + > > + netdev_info(priv->core->netdev, "ESDACC: BRP %u, BTR 0x%08x", > > + brp, btr); > > same here Changed it. > > > + } > > + > > + return 0; > > +} > > + > > +static void handle_core_msg_rxtxdone(struct acc_core *core, > > + const struct acc_bmmsg_rxtxdone *msg) > > +{ > > + struct acc_net_priv *priv = netdev_priv(core->netdev); > > + struct net_device_stats *stats = &core->netdev->stats; > > + struct sk_buff *skb; > > + > > + if (msg->dlc.rxtx.len & ACC_BM_LENFLAG_TX) { > > + u8 tx_fifo_tail = core->tx_fifo_tail; > > + > > + if (core->tx_fifo_head == tx_fifo_tail) { > > + netdev_warn(core->netdev, > > + "TX interrupt, but queue is empty!?\n"); > > + return; > > + } > > + > > + /* Direct access echo skb to attach HW time stamp. */ > > + skb = priv->can.echo_skb[tx_fifo_tail]; > > + if (skb) { > > + skb_hwtstamps(skb)->hwtstamp = > > + acc_ts2ktime(priv->ov, msg->ts); > > + } > > + > > + stats->tx_packets++; > > + stats->tx_bytes += can_get_echo_skb(core->netdev, tx_fifo_tail, > > + NULL); > > + > > + tx_fifo_tail++; > > + if (tx_fifo_tail >= core->tx_fifo_size) > > + tx_fifo_tail = 0U; > > + core->tx_fifo_tail = tx_fifo_tail; > > + > > + netif_wake_queue(core->netdev); I'll revisit this later. > You only need to wake the queue if there is space available, and you > need a barrier here, see mcp251xfd_handle_tefif(): > > > int mcp251xfd_handle_tefif(struct mcp251xfd_priv *priv) > > { > > [...] > > if (mcp251xfd_get_tx_free(priv->tx)) { > > /* Make sure that anybody stopping the queue after > > * this sees the new tx_ring->tail. > > */ > > smp_mb(); > > netif_wake_queue(priv->ndev); > > } > > > > return 0; > > > > } > > + > > + } else { > > + struct can_frame *cf; > > + > > + skb = alloc_can_skb(core->netdev, &cf); > > + if (!skb) { > > + stats->rx_dropped++; > > + return; > > + } > > + > > + cf->can_id = msg->id & CAN_EFF_MASK; > > + if (msg->id & ACC_CAN_EFF_FLAG) > > + cf->can_id |= CAN_EFF_FLAG; > > + > > + can_frame_set_cc_len(cf, msg->dlc.rx.len & ACC_CAN_DLC_MASK, > > + priv->can.ctrlmode); > > + > > + if (msg->dlc.rx.len & ACC_CAN_RTR_FLAG) > > + cf->can_id |= CAN_RTR_FLAG; > > + else > > + memcpy(cf->data, msg->data, cf->len); > > + > > + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); > > + > > + stats->rx_packets++; > > + stats->rx_bytes += cf->len; > > + netif_rx(skb); > > Consider moving the TX completion and RX handling into NAPI or make use > of rx-offload, to avoid reordering of CAN frames in the RX path. > > > + } > > +} > > + > > +static void handle_core_msg_txabort(struct acc_core *core, > > + const struct acc_bmmsg_txabort *msg) > > +{ > > + struct net_device_stats *stats = &core->netdev->stats; > > + u8 tx_fifo_tail = core->tx_fifo_tail; > > + u32 abort_mask = msg->abort_mask; /* u32 extend to avoid warnings later */ > > + > > + /* The abort_mask shows which frames were aborted in ESDACC's FIFO. */ > > You can directly iterate over all set bits using for_each_set_bit(). for_each_set_bit() / for_each_set_bit_from() start at bit 1 or a specified bit but don't wrap around. That would make two loops necessary. Here a single loop that wraps around till the end of the fifo is found or all aborted frames have been found in the right order looks better to me. > > + while (tx_fifo_tail != core->tx_fifo_head && (abort_mask)) { > > + const u32 tail_mask = (1U << tx_fifo_tail); > > + > > + if (!(abort_mask & tail_mask)) > > + break; > > + abort_mask &= ~tail_mask; > > + > > + can_free_echo_skb(core->netdev, tx_fifo_tail, NULL); > > + stats->tx_dropped++; > > + stats->tx_aborted_errors++; > > + > > + tx_fifo_tail++; > > + if (tx_fifo_tail >= core->tx_fifo_size) > > + tx_fifo_tail = 0; > > + } > > + core->tx_fifo_tail = tx_fifo_tail; > > + if (abort_mask) > > + netdev_warn(core->netdev, "Unhandled aborted messages\n"); > > + > > + if (!acc_resetmode_entered(core)) > > + netif_wake_queue(core->netdev); > > +} > > + > > +static void handle_core_msg_overrun(struct acc_core *core, > > + const struct acc_bmmsg_overrun *msg) > > +{ > > + struct acc_net_priv *priv = netdev_priv(core->netdev); > > + struct net_device_stats *stats = &core->netdev->stats; > > + struct can_frame *cf; > > + struct sk_buff *skb; > > + > > + skb = alloc_can_err_skb(core->netdev, &cf); > > + if (!skb) { > > + stats->rx_dropped++; > > + return; > > + } > > Please handle the stats, even if the driver fails to allocate and error > skb. Changed as needed. > > + > > + /* lost_cnt may be 0 if not supported by ESDACC version */ > > + if (msg->lost_cnt) { > > + stats->rx_errors += msg->lost_cnt; > > + stats->rx_over_errors += msg->lost_cnt; > > + } else { > > + stats->rx_errors++; > > + stats->rx_over_errors++; > > + } > > + > > + cf->can_id |= CAN_ERR_CRTL; > > + cf->data[1] = CAN_ERR_CRTL_RX_OVERFLOW; > > + > > + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); > > + > > + stats->rx_packets++; > > + stats->rx_bytes += cf->len; > > + netif_rx(skb); > > +} > > + > > +static void handle_core_msg_buserr(struct acc_core *core, > > + const struct acc_bmmsg_buserr *msg) > > +{ > > + struct acc_net_priv *priv = netdev_priv(core->netdev); > > + struct net_device_stats *stats = &core->netdev->stats; > > + struct can_frame *cf; > > + struct sk_buff *skb; > > + const u32 reg_status = msg->reg_status; > > + const u8 rxerr = (u8)reg_status; > > + const u8 txerr = (u8)(reg_status >> 8); > > casts not needed See comment at top of email. > > + > > + priv->can.can_stats.bus_error++; > > + > > + skb = alloc_can_err_skb(core->netdev, &cf); > > + if (!skb) { > > + stats->rx_dropped++; > > + return; > > + } > > Please handle the stats, even if the driver fails to allocate and error > skb. Changed as needed. > > + > > + cf->can_id |= CAN_ERR_PROT | CAN_ERR_BUSERROR; > > + > > + /* msg->ecc acts like SJA1000's ECC register */ > > + switch (msg->ecc & ACC_ECC_MASK) { > > + case ACC_ECC_BIT: > > + cf->data[2] |= CAN_ERR_PROT_BIT; > > + break; > > + case ACC_ECC_FORM: > > + cf->data[2] |= CAN_ERR_PROT_FORM; > > + break; > > + case ACC_ECC_STUFF: > > + cf->data[2] |= CAN_ERR_PROT_STUFF; > > + break; > > + default: > > + cf->data[2] |= CAN_ERR_PROT_UNSPEC; > > + break; > > + } > > + > > + /* Set error location */ > > + cf->data[3] = msg->ecc & ACC_ECC_SEG; > > + > > + /* Error occurred during transmission? */ > > + if ((msg->ecc & ACC_ECC_DIR) == 0) { > > + cf->data[2] |= CAN_ERR_PROT_TX; > > + stats->tx_errors++; > > + } else { > > + stats->rx_errors++; > > + } > > + /* Insert CAN TX and RX error counters. */ > > + cf->data[6] = txerr; > > + cf->data[7] = rxerr; > > + > > + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); > > + > > + stats->rx_packets++; > > + stats->rx_bytes += cf->len; > > + netif_rx(skb); > > +} > > + > > +static void > > +handle_core_msg_errstatechange(struct acc_core *core, > > + const struct acc_bmmsg_errstatechange *msg) > > +{ > > + struct acc_net_priv *priv = netdev_priv(core->netdev); > > + struct net_device_stats *stats = &core->netdev->stats; > > + struct can_frame *cf = NULL; > > + struct sk_buff *skb; > > + const u32 reg_status = msg->reg_status; > > + const u8 rxerr = (u8)reg_status; > > + const u8 txerr = (u8)(reg_status >> 8); > > cast not needed See comment at top of email. > > + enum can_state new_state; > > + > > + if (reg_status & ACC_REG_STATUS_MASK_STATUS_BS) { > > + new_state = CAN_STATE_BUS_OFF; > > + } else if (reg_status & ACC_REG_STATUS_MASK_STATUS_EP) { > > + new_state = CAN_STATE_ERROR_PASSIVE; > > + } else if (reg_status & ACC_REG_STATUS_MASK_STATUS_ES) { > > + new_state = CAN_STATE_ERROR_WARNING; > > + } else { > > + new_state = CAN_STATE_ERROR_ACTIVE; > > + if (priv->can.state == CAN_STATE_BUS_OFF) { > > + /* See comment in acc_set_mode() for CAN_MODE_START */ > > + netif_wake_queue(core->netdev); > > See my other comment. Shut down the chip in case of bus off. I don't get it. The esdACC is shut down, if it is in bus-off state. > > + } > > + } > > + > > + skb = alloc_can_err_skb(core->netdev, &cf); > > + > > + if (new_state != priv->can.state) { > > + enum can_state tx_state, rx_state; > > + > > + tx_state = (txerr >= rxerr) ? > > + new_state : CAN_STATE_ERROR_ACTIVE; > > + rx_state = (rxerr >= txerr) ? > > + new_state : CAN_STATE_ERROR_ACTIVE; > > + > > + /* Always call can_change_state() to update the state > > + * even if alloc_can_err_skb() may have failed. > > + * can_change_state() can cope with a NULL cf pointer. > > + */ > > + can_change_state(core->netdev, cf, tx_state, rx_state); > > + } > > + > > + if (skb) { > > + cf->data[6] = txerr; > > + cf->data[7] = rxerr; > > + > > + skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); > > + > > + stats->rx_packets++; > > + stats->rx_bytes += cf->len; > > + netif_rx(skb); > > + } else { > > + stats->rx_dropped++; > > + } > > + > > + if (new_state == CAN_STATE_BUS_OFF) { > > + acc_write32(core, ACC_CORE_OF_TX_ABORT_MASK, 0xffff); > > + can_bus_off(core->netdev); > > + } > > +} > > + > > +static void handle_core_interrupt(struct acc_core *core) > > +{ > > + u32 msg_fifo_head = core->bmfifo.local_irq_cnt & 0xff; > > + > > + while (core->bmfifo.msg_fifo_tail != msg_fifo_head) { > > + const struct acc_bmmsg *msg = > > + &core->bmfifo.messages[core->bmfifo.msg_fifo_tail]; > > + > > + switch (msg->u.msg_id) { > > + case BM_MSG_ID_RXTXDONE: > > + handle_core_msg_rxtxdone(core, &msg->u.rxtxdone); > > + break; > > + > > + case BM_MSG_ID_TXABORT: > > + handle_core_msg_txabort(core, &msg->u.txabort); > > + break; > > + > > + case BM_MSG_ID_OVERRUN: > > + handle_core_msg_overrun(core, &msg->u.overrun); > > + break; > > + > > + case BM_MSG_ID_BUSERR: > > + handle_core_msg_buserr(core, &msg->u.buserr); > > + break; > > + > > + case BM_MSG_ID_ERRPASSIVE: > > + case BM_MSG_ID_ERRWARN: > > + handle_core_msg_errstatechange(core, > > + &msg->u.errstatechange); > > + break; > > + > > + default: > > + /* Ignore all other BM messages (like the CAN-FD messages) */ > > + break; > > + } > > + > > + core->bmfifo.msg_fifo_tail = > > + (core->bmfifo.msg_fifo_tail + 1) & 0xff; > > + } > > +} > > + /** * acc_card_interrupt() - handle the interrupts of an esdACC FPGA * * @ov: overview module structure * @cores: array of core structures * * This function handles all interrupts pending for the overview module and the * CAN cores of the esdACC FPGA. * * It examines for all cores (the overview module core and the CAN cores) * the bmfifo.irq_cnt and compares it with the previously saved * bmfifo.local_irq_cnt. An IRQ is pending if they differ. The esdACC FPGA * updates the bmfifo.irq_cnt values by DMA. * * The pending interrupts are masked by writing to the IRQ mask register at * ACC_OV_OF_BM_IRQ_MASK. This register has for each core a two bit command * field evaluated as follows: * * Define, bit pattern: meaning * 00: no action * ACC_BM_IRQ_UNMASK, 01: unmask interrupt * ACC_BM_IRQ_MASK, 10: mask interrupt * 11: no action * * For each CAN core with a pending IRQ handle_core_interrupt() handles all * busmaster messages from the message FIFO. The last handled message (FIFO * index) is written to the CAN core to acknowledge its handling. * * Last step is to unmask all interrupts in the FPGA using * ACC_BM_IRQ_UNMASK_ALL. */ > > +irqreturn_t acc_card_interrupt(struct acc_ov *ov, struct acc_core *cores) > > +{ > > + u32 irqmask; > > + int i; > > single space here Changed > > + > > + /* First we look for whom interrupts are pending, card/overview > > + * or any of the cores. Two bits in irqmask are used for each; > > + * set to ACC_BM_IRQ_MASK then: > > + */ > > Please explain a bit more detailed how the interrupt handling works. I added a doxygen like documentation comment to this function (see before). > > + irqmask = 0; > > + if (*ov->bmfifo.irq_cnt != ov->bmfifo.local_irq_cnt) { > > + irqmask |= ACC_BM_IRQ_MASK; > > + ov->bmfifo.local_irq_cnt = *ov->bmfifo.irq_cnt; > > + } > > + > > + for (i = 0; i < ov->active_cores; i++) { > > + struct acc_core *core = &cores[i]; > > + > > + if (*core->bmfifo.irq_cnt != core->bmfifo.local_irq_cnt) { > > + irqmask |= (ACC_BM_IRQ_MASK << (2 * (i + 1))); > > + core->bmfifo.local_irq_cnt = *core->bmfifo.irq_cnt; > > + } > > + } > > + > > + if (!irqmask) > > + return IRQ_NONE; > > + > > + /* At second we tell the card we're working on them by writing irqmask, > > + * call handle_{ov|core}_interrupt and then acknowledge the > > + * interrupts by writing irq_cnt: > > + */ > > + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_MASK, irqmask); > > + > > + if (irqmask & ACC_BM_IRQ_MASK) { > > + /* handle_ov_interrupt(); - no use yet. */ > > + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_COUNTER, > > + ov->bmfifo.local_irq_cnt); > > + } > > + > > + for (i = 0; i < ov->active_cores; i++) { > > + struct acc_core *core = &cores[i]; > > + > > + if (irqmask & (ACC_BM_IRQ_MASK << (2 * (i + 1)))) { > > + handle_core_interrupt(core); > > + acc_write32(core, ACC_OV_OF_BM_IRQ_COUNTER, > > + core->bmfifo.local_irq_cnt); > > + } > > + } > > + > > + acc_ov_write32(ov, ACC_OV_OF_BM_IRQ_MASK, ACC_BM_IRQ_UNMASK_ALL); > > + > > + return IRQ_HANDLED; > > +} > > diff --git a/drivers/net/can/esd/esdacc.h b/drivers/net/can/esd/esdacc.h > > new file mode 100644 > > index 000000000000..3e865ececb3e > > --- /dev/null > > +++ b/drivers/net/can/esd/esdacc.h > > @@ -0,0 +1,380 @@ > > +/* SPDX-License-Identifier: GPL-2.0-only */ > > +/* Copyright (C) 2015 - 2016 Thomas Körper, esd electronic system design gmbh > > + * Copyright (C) 2017 - 2021 Stefan Mätje, esd electronics gmbh > > + */ > > + > > +#include <linux/kernel.h> > > +#include <linux/netdevice.h> > > +#include <linux/can/dev.h> > > + > > +#define ACC_CAN_EFF_FLAG 0x20000000 > > +#define ACC_CAN_RTR_FLAG 0x10 > > +#define ACC_CAN_DLC_MASK 0x0f > > + > > +#define ACC_OV_OF_PROBE 0x0000 > > +#define ACC_OV_OF_VERSION 0x0004 > > +#define ACC_OV_OF_INFO 0x0008 > > +#define ACC_OV_OF_CANCORE_FREQ 0x000c > > +#define ACC_OV_OF_TS_FREQ_LO 0x0010 > > +#define ACC_OV_OF_TS_FREQ_HI 0x0014 > > +#define ACC_OV_OF_IRQ_STATUS_CORES 0x0018 > > +#define ACC_OV_OF_TS_CURR_LO 0x001c > > +#define ACC_OV_OF_TS_CURR_HI 0x0020 > > +#define ACC_OV_OF_IRQ_STATUS 0x0028 > > +#define ACC_OV_OF_MODE 0x002c > > +#define ACC_OV_OF_BM_IRQ_COUNTER 0x0070 > > +#define ACC_OV_OF_BM_IRQ_MASK 0x0074 > > +#define ACC_OV_OF_MSI_DATA 0x0080 > > +#define ACC_OV_OF_MSI_ADDRESSOFFSET 0x0084 > > + > > +/* Feature flags are contained in the upper 16 bit of the version > > + * register at ACC_OV_OF_VERSION but only used with these masks after > > + * extraction into an extra variable => (xx - 16). > > + */ > > +#define ACC_OV_REG_FEAT_IDX_CANFD (27 - 16) > > +#define ACC_OV_REG_FEAT_IDX_NEW_PSC (28 - 16) > > +#define ACC_OV_REG_FEAT_MASK_CANFD BIT(ACC_OV_REG_FEAT_IDX_CANFD) > > +#define ACC_OV_REG_FEAT_MASK_NEW_PSC BIT(ACC_OV_REG_FEAT_IDX_NEW_PSC) > > + > > +#define ACC_OV_REG_MODE_MASK_ENDIAN_LITTLE 0x00000001 > > +#define ACC_OV_REG_MODE_MASK_BM_ENABLE 0x00000002 > > +#define ACC_OV_REG_MODE_MASK_MODE_LED 0x00000004 > > +#define ACC_OV_REG_MODE_MASK_TIMER 0x00000070 > > +#define ACC_OV_REG_MODE_MASK_TIMER_ENABLE 0x00000010 > > +#define ACC_OV_REG_MODE_MASK_TIMER_ONE_SHOT 0x00000020 > > +#define ACC_OV_REG_MODE_MASK_TIMER_ABSOLUTE 0x00000040 > > +#define ACC_OV_REG_MODE_MASK_TS_SRC 0x00000180 > > +#define ACC_OV_REG_MODE_MASK_I2C_ENABLE 0x00000800 > > +#define ACC_OV_REG_MODE_MASK_MSI_ENABLE 0x00004000 > > +#define ACC_OV_REG_MODE_MASK_NEW_PSC_ENABLE 0x00008000 > > +#define ACC_OV_REG_MODE_MASK_FPGA_RESET 0x80000000 > > + > > +#define ACC_CORE_OF_CTRL_MODE 0x0000 > > +#define ACC_CORE_OF_STATUS_IRQ 0x0008 > > +#define ACC_CORE_OF_BRP 0x000c > > +#define ACC_CORE_OF_BTR 0x0010 > > +#define ACC_CORE_OF_FBTR 0x0014 > > +#define ACC_CORE_OF_STATUS 0x0030 > > +#define ACC_CORE_OF_TXFIFO_CONFIG 0x0048 > > +#define ACC_CORE_OF_TXFIFO_STATUS 0x004c > > +#define ACC_CORE_OF_TX_STATUS_IRQ 0x0050 > > +#define ACC_CORE_OF_TX_ABORT_MASK 0x0054 > > +#define ACC_CORE_OF_BM_IRQ_COUNTER 0x0070 > > +#define ACC_CORE_OF_TXFIFO_ID 0x00c0 > > +#define ACC_CORE_OF_TXFIFO_DLC 0x00c4 > > +#define ACC_CORE_OF_TXFIFO_DATA_0 0x00c8 > > +#define ACC_CORE_OF_TXFIFO_DATA_1 0x00cc > > + > > +#define ACC_REG_CONTROL_IDX_MODE_RESETMODE 0 > > +#define ACC_REG_CONTROL_IDX_MODE_LOM 1 > > +#define ACC_REG_CONTROL_IDX_MODE_STM 2 > > +#define ACC_REG_CONTROL_IDX_MODE_TRANSEN 5 > > +#define ACC_REG_CONTROL_IDX_MODE_TS 6 > > +#define ACC_REG_CONTROL_IDX_MODE_SCHEDULE 7 > > +#define ACC_REG_CONTROL_MASK_MODE_RESETMODE \ > > + BIT(ACC_REG_CONTROL_IDX_MODE_RESETMODE) > > +#define ACC_REG_CONTROL_MASK_MODE_LOM \ > > + BIT(ACC_REG_CONTROL_IDX_MODE_LOM) > > +#define ACC_REG_CONTROL_MASK_MODE_STM \ > > + BIT(ACC_REG_CONTROL_IDX_MODE_STM) > > +#define ACC_REG_CONTROL_MASK_MODE_TRANSEN \ > > + BIT(ACC_REG_CONTROL_IDX_MODE_TRANSEN) > > +#define ACC_REG_CONTROL_MASK_MODE_TS \ > > + BIT(ACC_REG_CONTROL_IDX_MODE_TS) > > +#define ACC_REG_CONTROL_MASK_MODE_SCHEDULE \ > > + BIT(ACC_REG_CONTROL_IDX_MODE_SCHEDULE) > > + > > +#define ACC_REG_CONTROL_IDX_IE_RXTX 8 > > +#define ACC_REG_CONTROL_IDX_IE_TXERROR 9 > > +#define ACC_REG_CONTROL_IDX_IE_ERRWARN 10 > > +#define ACC_REG_CONTROL_IDX_IE_OVERRUN 11 > > +#define ACC_REG_CONTROL_IDX_IE_TSI 12 > > +#define ACC_REG_CONTROL_IDX_IE_ERRPASS 13 > > +#define ACC_REG_CONTROL_IDX_IE_BUSERR 15 > > +#define ACC_REG_CONTROL_MASK_IE_RXTX BIT(ACC_REG_CONTROL_IDX_IE_RXTX) > > +#define ACC_REG_CONTROL_MASK_IE_TXERROR BIT(ACC_REG_CONTROL_IDX_IE_TXERROR) > > +#define ACC_REG_CONTROL_MASK_IE_ERRWARN BIT(ACC_REG_CONTROL_IDX_IE_ERRWARN) > > +#define ACC_REG_CONTROL_MASK_IE_OVERRUN BIT(ACC_REG_CONTROL_IDX_IE_OVERRUN) > > +#define ACC_REG_CONTROL_MASK_IE_TSI BIT(ACC_REG_CONTROL_IDX_IE_TSI) > > +#define ACC_REG_CONTROL_MASK_IE_ERRPASS BIT(ACC_REG_CONTROL_IDX_IE_ERRPASS) > > +#define ACC_REG_CONTROL_MASK_IE_BUSERR BIT(ACC_REG_CONTROL_IDX_IE_BUSERR) > > + > > +/* 256 BM_MSGs of 32 byte size */ > > +#define ACC_CORE_DMAMSG_SIZE 32U > > +#define ACC_CORE_DMABUF_SIZE (256U * ACC_CORE_DMAMSG_SIZE) > > + > > +enum acc_bmmsg_id { > > + BM_MSG_ID_RXTXDONE = 0x01, > > + BM_MSG_ID_TXABORT = 0x02, > > + BM_MSG_ID_OVERRUN = 0x03, > > + BM_MSG_ID_BUSERR = 0x04, > > + BM_MSG_ID_ERRPASSIVE = 0x05, > > + BM_MSG_ID_ERRWARN = 0x06, > > + BM_MSG_ID_TIMESLICE = 0x07, > > + BM_MSG_ID_HWTIMER = 0x08, > > + BM_MSG_ID_HOTPLUG = 0x09, > > +}; > > + > > +/* The struct acc_bmmsg* structure declarations that follow here provide > > + * access to the ring buffer of bus master messages maintained by the FPGA > > + * bus master engine. All bus master messages have the same size of > > + * ACC_CORE_DMAMSG_SIZE and a minimum alignment of ACC_CORE_DMAMSG_SIZE in > > + * memory. > > + * > > + * All structure members are natural aligned. Therefore we should not need > > + * a __packed attribute. All struct acc_bmmsg* declarations have at least > > + * reserved* members to fill the structure to the full ACC_CORE_DMAMSG_SIZE. > > + * > > + * A failure of this property due padding will be detected at compile time > > + * by BUILD_BUG_ON(sizeof(struct acc_bmmsg) != ACC_CORE_DMAMSG_SIZE) > > + */ > > You can use static_assert() directly unter the definition of the struct. Have changed it to use static_assert(). > > + > > +struct acc_bmmsg_rxtxdone { > > + u8 msg_id; > > + u8 txfifo_level; > > + u8 reserved1[2]; > > + u8 txtsfifo_level; > > + u8 reserved2[3]; > > + u32 id; > > + union { > > + struct { > > + u8 len; > > + u8 reserved0; > > + u8 bits; > > + u8 state; > > + } rxtx; > > + struct { > > + u8 len; > > + u8 msg_lost; > > + u8 bits; > > + u8 state; > > + } rx; > > + struct { > > + u8 len; > > + u8 txfifo_idx; > > + u8 bits; > > + u8 state; > > + } tx; > > + } dlc; > > + u8 data[8]; > > + /* Time stamps in struct acc_ov::timestamp_frequency ticks. */ > > + u64 ts; > > +}; > > + > > +struct acc_bmmsg_txabort { > > + u8 msg_id; > > + u8 txfifo_level; > > + u16 abort_mask; > > + u8 txtsfifo_level; > > + u8 reserved2[1]; > > + u16 abort_mask_txts; > > + u64 ts; > > + u32 reserved3[4]; > > +}; > > + > > +struct acc_bmmsg_overrun { > > + u8 msg_id; > > + u8 txfifo_level; > > + u8 lost_cnt; > > + u8 reserved1; > > + u8 txtsfifo_level; > > + u8 reserved2[3]; > > + u64 ts; > > + u32 reserved3[4]; > > +}; > > + > > +struct acc_bmmsg_buserr { > > + u8 msg_id; > > + u8 txfifo_level; > > + u8 ecc; > > + u8 reserved1; > > + u8 txtsfifo_level; > > + u8 reserved2[3]; > > + u64 ts; > > + u32 reg_status; > > + u32 reg_btr; > > + u32 reserved3[2]; > > +}; > > + > > +struct acc_bmmsg_errstatechange { > > + u8 msg_id; > > + u8 txfifo_level; > > + u8 reserved1[2]; > > + u8 txtsfifo_level; > > + u8 reserved2[3]; > > + u64 ts; > > + u32 reg_status; > > + u32 reserved3[3]; > > +}; > > + > > +struct acc_bmmsg_timeslice { > > + u8 msg_id; > > + u8 txfifo_level; > > + u8 reserved1[2]; > > + u8 txtsfifo_level; > > + u8 reserved2[3]; > > + u64 ts; > > + u32 reserved3[4]; > > +}; > > + > > +struct acc_bmmsg_hwtimer { > > + u8 msg_id; > > + u8 reserved1[3]; > > + u32 reserved2[1]; > > + u64 timer; > > + u32 reserved3[4]; > > +}; > > + > > +struct acc_bmmsg_hotplug { > > + u8 msg_id; > > + u8 reserved1[3]; > > + u32 reserved2[7]; > > +}; > > + > > +struct acc_bmmsg { > > + union { > > + u8 msg_id; > > + struct acc_bmmsg_rxtxdone rxtxdone; > > + struct acc_bmmsg_txabort txabort; > > + struct acc_bmmsg_overrun overrun; > > + struct acc_bmmsg_buserr buserr; > > + struct acc_bmmsg_errstatechange errstatechange; > > + struct acc_bmmsg_timeslice timeslice; > > + struct acc_bmmsg_hwtimer hwtimer; > > + } u; > > +}; > > Why don't you use the union directly instead of putting it into a struct? Directly use the union is better and changed now. > > + > > +/* Regarding Documentation/process/volatile-considered-harmful.rst the > > + * forth exception applies to the "irq_cnt" member of the structure > > + * below. The u32 variable "irq_cnt" points to is updated by the ESDACC > > + * FPGA via DMA. > > + */ > > +struct acc_bmfifo { > > + const struct acc_bmmsg *messages; > > + /* Bits 0..7: bm_fifo head index */ > > + volatile const u32 *irq_cnt; > > + u32 local_irq_cnt; > > + u32 msg_fifo_tail; > > +}; > > + > > +struct acc_core { > > + void __iomem *addr; > > + struct net_device *netdev; > > + struct acc_bmfifo bmfifo; > > + u8 tx_fifo_size; > > + u8 tx_fifo_head; > > + u8 tx_fifo_tail; > > Is the tx_fifo_size a power of two? If so it's usually easier to use a > unsigned int for the head and tail pointers and mask them to their real > values. head and tail pointers only ever increase and eventually roll > over. As both are unsigned there's no problem if head has rolled over, > but the tail doesn't. > > When you read tail pointer from HW you round_down() to > core->tx_fifo_size, add the tail value from HW and add > core->tx_fifo_size if it's lower than the old value. This is the only > time where you have to take care of rollover. I'll have a look at it. > > +}; > > + > > +struct acc_ov { > > + void __iomem *addr; > > + struct acc_bmfifo bmfifo; > > + u32 timestamp_frequency; > > + u32 ts2ns_numerator; > > + u32 ts2ns_denominator; > > + u32 core_frequency; > > + u16 version; > > + u16 features; > > + u8 total_cores; > > + u8 active_cores; > > +}; > > + > > +struct acc_net_priv { > > + struct can_priv can; /* must be the first member! */ > > + struct acc_core *core; > > + struct acc_ov *ov; > > +}; > > + > > +static inline u32 acc_read32(struct acc_core *core, unsigned short offs) > > +{ > > + return ioread32be(core->addr + offs); > > +} > > + > > +static inline void acc_write32(struct acc_core *core, > > + unsigned short offs, u32 v) > > +{ > > + iowrite32be(v, core->addr + offs); > > +} > > + > > +static inline void acc_write32_noswap(struct acc_core *core, > > + unsigned short offs, u32 v) > > +{ > > + iowrite32(v, core->addr + offs); > > +} > > + > > +static inline void acc_sset_bits(struct acc_core *core, > > + unsigned short offs, u32 mask) > > +{ > > + u32 v = acc_read32(core, offs); > > + > > + v |= mask; > > + acc_write32(core, offs, v); > > +} > > + > > +static inline void acc_clear_bits(struct acc_core *core, > > + unsigned short offs, u32 mask) > > +{ > > + u32 v = acc_read32(core, offs); > > + > > + v &= ~mask; > > + acc_write32(core, offs, v); > > +} > > + > > +static inline int acc_resetmode_entered(struct acc_core *core) > > +{ > > + u32 ctrl = acc_read32(core, ACC_CORE_OF_CTRL_MODE); > > + > > + return (ctrl & ACC_REG_CONTROL_MASK_MODE_RESETMODE) != 0; > > +} > > + > > +static inline u32 acc_ov_read32(struct acc_ov *ov, unsigned short offs) > > +{ > > + return ioread32be(ov->addr + offs); > > +} > > + > > +static inline void acc_ov_write32(struct acc_ov *ov, > > + unsigned short offs, u32 v) > > +{ > > + iowrite32be(v, ov->addr + offs); > > +} > > + > > +static inline void acc_ov_set_bits(struct acc_ov *ov, > > + unsigned short offs, u32 b) > > +{ > > + u32 v = acc_ov_read32(ov, offs); > > + > > + v |= b; > > + acc_ov_write32(ov, offs, v); > > +} > > + > > +static inline void acc_ov_clear_bits(struct acc_ov *ov, > > + unsigned short offs, u32 b) > > +{ > > + u32 v = acc_ov_read32(ov, offs); > > + > > + v &= ~b; > > + acc_ov_write32(ov, offs, v); > > +} > > + > > +static inline void acc_reset_fpga(struct acc_ov *ov) > > +{ > > + acc_ov_write32(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_FPGA_RESET); > > + > > + /* Also reset I^2C, to re-detect card addons at every driver start: */ > > + acc_ov_clear_bits(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_I2C_ENABLE); > > + mdelay(2); > > + acc_ov_set_bits(ov, ACC_OV_OF_MODE, ACC_OV_REG_MODE_MASK_I2C_ENABLE); > > + mdelay(10); > > +} > > + > > +void acc_init_ov(struct acc_ov *ov, struct device *dev); > > +void acc_init_bm_ptr(struct acc_ov *ov, struct acc_core *cores, > > + const void *mem); > > +int acc_open(struct net_device *netdev); > > +int acc_close(struct net_device *netdev); > > +netdev_tx_t acc_start_xmit(struct sk_buff *skb, struct net_device *netdev); > > +int acc_get_berr_counter(const struct net_device *netdev, > > + struct can_berr_counter *bec); > > +int acc_set_mode(struct net_device *netdev, enum can_mode mode); > > +int acc_set_bittiming(struct net_device *netdev); > > +irqreturn_t acc_card_interrupt(struct acc_ov *ov, struct acc_core *cores); > > -- > > 2.25.1 > > > > > > regards, > Marc > ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v6 2/4] can: esd: add support for esd GmbH PCIe/402 CAN interface family 2022-09-30 22:15 ` Stefan Mätje @ 2022-10-05 16:04 ` Marc Kleine-Budde 0 siblings, 0 replies; 15+ messages in thread From: Marc Kleine-Budde @ 2022-10-05 16:04 UTC (permalink / raw) To: Stefan Mätje; +Cc: netdev, linux-kernel, linux-can, wg [-- Attachment #1: Type: text/plain, Size: 1565 bytes --] On 30.09.2022 22:15:59, Stefan Mätje wrote: > I would like to resume the efforts to bring this driver in the Linux > kernel after being kicked off track in February by another project. \o/ > I did a lot of the changes on the driver you recommended, but some > stuff is not yet clear to me. Please see my comments in-line of the > email below. > > My local developement is at the moment rebased to > linux-can-next:master on 7b584fbb36362340a2d9cfe459e447619eecebea. > Should I send a V7 of the patch (rebased to another commit)? How > should I proceed? You can use latest net-next/main as base version. > You have commented on many type casts that they would not be needed. > But all of them had been introduced by me due to warnings of the > compiler in the style of "warning: conversion from ‘u32’ {aka > ‘unsigned int’} to ‘u8’ {aka ‘unsigned char’} may change value > [-Wconversion]". These are triggered by building the driver with "W=3" > as recommended in kernel documentation. Oh? Is there a recommendation for W=3? I can only find a W=1: | https://elixir.bootlin.com/linux/v6.0/source/Documentation/process/maintainer-netdev.rst#L235 > Should these warnings generally be ignored and the casts be removed > then? regards, Marc -- Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 488 bytes --] ^ permalink raw reply [flat|nested] 15+ messages in thread
* [PATCH 3/4] can: esd_402_pci: do not increase rx statistics when generating a CAN rx error message frame 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje 2021-12-01 22:03 ` [PATCH 1/4] MAINTAINERS: add Stefan Mätje as maintainer for the esd electronics GmbH CAN drivers Stefan Mätje 2021-12-01 22:03 ` [PATCH 2/4] can: esd: add support for esd GmbH PCIe/402 CAN interface family Stefan Mätje @ 2021-12-01 22:03 ` Stefan Mätje 2021-12-01 22:03 ` [PATCH 4/4] can: esd_402_pci: do not increase rx_bytes statistics for RTR frames Stefan Mätje ` (4 subsequent siblings) 7 siblings, 0 replies; 15+ messages in thread From: Stefan Mätje @ 2021-12-01 22:03 UTC (permalink / raw) To: linux-can, Marc Kleine-Budde, Wolfgang Grandegger; +Cc: netdev, linux-kernel The CAN error message frames (i.e. error skb) are an interface specific to socket CAN. The payload of the CAN error message frames does not correspond to any actual data sent on the wire. Only an error flag and a delimiter are transmitted when an error occurs (c.f. ISO 11898-1 section 10.4.4.2 "Error flag"). For this reason, it makes no sense to increment the rx_packets and rx_bytes fields of struct net_device_stats because no actual payload were transmitted on the wire. This patch brings the esd_402_pci driver in line with the other CAN drivers which have been changed after a suggestion of Vincent Mailhol. Suggested-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr> Signed-off-by: Stefan Mätje <stefan.maetje@esd.eu> --- drivers/net/can/esd/esdacc.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/drivers/net/can/esd/esdacc.c b/drivers/net/can/esd/esdacc.c index 13f7397dfc4e..f74248662cd6 100644 --- a/drivers/net/can/esd/esdacc.c +++ b/drivers/net/can/esd/esdacc.c @@ -555,8 +555,6 @@ static void handle_core_msg_overrun(struct acc_core *core, skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); - stats->rx_packets++; - stats->rx_bytes += cf->len; netif_rx(skb); } @@ -613,8 +611,6 @@ static void handle_core_msg_buserr(struct acc_core *core, skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); - stats->rx_packets++; - stats->rx_bytes += cf->len; netif_rx(skb); } @@ -668,8 +664,6 @@ handle_core_msg_errstatechange(struct acc_core *core, skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); - stats->rx_packets++; - stats->rx_bytes += cf->len; netif_rx(skb); } else { stats->rx_dropped++; -- 2.25.1 ^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH 4/4] can: esd_402_pci: do not increase rx_bytes statistics for RTR frames 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje ` (2 preceding siblings ...) 2021-12-01 22:03 ` [PATCH 3/4] can: esd_402_pci: do not increase rx statistics when generating a CAN rx error message frame Stefan Mätje @ 2021-12-01 22:03 ` Stefan Mätje 2021-12-01 22:09 ` [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje ` (3 subsequent siblings) 7 siblings, 0 replies; 15+ messages in thread From: Stefan Mätje @ 2021-12-01 22:03 UTC (permalink / raw) To: linux-can, Marc Kleine-Budde, Wolfgang Grandegger; +Cc: netdev, linux-kernel The actual payload length of the CAN Remote Transmission Request (RTR) frames is always 0, i.e. nothing is transmitted on the wire. However, those RTR frames still use the DLC to indicate the length of the requested frame. As such, net_device_stats:rx_bytes should not be increased for the RTR frames. This patch brings the esd_402_pci driver in line with the other CAN drivers which have been changed after a suggestion of Vincent Mailhol. Suggested-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr> Signed-off-by: Stefan Mätje <stefan.maetje@esd.eu> --- drivers/net/can/esd/esdacc.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/drivers/net/can/esd/esdacc.c b/drivers/net/can/esd/esdacc.c index f74248662cd6..99c1e1bdd72d 100644 --- a/drivers/net/can/esd/esdacc.c +++ b/drivers/net/can/esd/esdacc.c @@ -483,15 +483,16 @@ static void handle_core_msg_rxtxdone(struct acc_core *core, can_frame_set_cc_len(cf, msg->dlc.rx.len & ACC_CAN_DLC_MASK, priv->can.ctrlmode); - if (msg->dlc.rx.len & ACC_CAN_RTR_FLAG) + if (msg->dlc.rx.len & ACC_CAN_RTR_FLAG) { cf->can_id |= CAN_RTR_FLAG; - else + } else { memcpy(cf->data, msg->data, cf->len); + stats->rx_bytes += cf->len; + } + stats->rx_packets++; skb_hwtstamps(skb)->hwtstamp = acc_ts2ktime(priv->ov, msg->ts); - stats->rx_packets++; - stats->rx_bytes += cf->len; netif_rx(skb); } } -- 2.25.1 ^ permalink raw reply related [flat|nested] 15+ messages in thread
* Re: [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje ` (3 preceding siblings ...) 2021-12-01 22:03 ` [PATCH 4/4] can: esd_402_pci: do not increase rx_bytes statistics for RTR frames Stefan Mätje @ 2021-12-01 22:09 ` Stefan Mätje 2022-01-07 13:40 ` Stefan Mätje 2022-01-25 16:25 ` Marc Kleine-Budde ` (2 subsequent siblings) 7 siblings, 1 reply; 15+ messages in thread From: Stefan Mätje @ 2021-12-01 22:09 UTC (permalink / raw) To: linux-can, wg, mkl; +Cc: netdev, linux-kernel Hi, this is really patch v6. Missed to change the email subject. My bad. Best regards, Stefan Mätje Am Mittwoch, den 01.12.2021, 23:03 +0100 schrieb Stefan Mätje: > The purpose of this patch is to introduce a new CAN driver to support > the esd GmbH 402 family of CAN interface boards. The hardware design > is based on a CAN controller implemented in a FPGA attached to a > PCIe link. > > More information on these boards can be found following the links > included in the commit message. > > This patch supports all boards but will operate the CAN-FD capable > boards only in Classic-CAN mode. The CAN-FD support will be added > when the initial patch has stabilized. > > The patch is reuses the previous work of my former colleague: > Link: > https://lore.kernel.org/linux-can/1426592308-23817-1-git-send-email-thomas.koerper@esd.eu/ > > *Note*: scripts/checkpatch.pl still emits the following warnings: > - esd_402_pci-core.c:270: Possible unnecessary 'out of memory' message > This error message is there to tell the user that the DMA allocation > failed and not an allocation for normal kernel memory. > - esdacc.h:255: The irq_cnt pointer is still declared volatile and > this has a reason and is explained in detail in the header > referencing the exception noted in volatile-considered-harmful.rst. > > The patch is based on the linux-can-next testing branch. > > Changes in v6: > - Fixed the statistic handling of RX overrun errors and increase > net_device_stats::rx_errors instead of net_device_stats::rx_dropped. > - Added a patch to not increase rx statistics when generating a CAN > rx error message frame as suggested on the linux-can list. > - Added a patch to not not increase rx_bytes statistics for RTR frames > as suggested on the linux-can list. > > The last two patches change the statistics handling from the previous > style used in other drivers to the newly suggested one. > ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface 2021-12-01 22:09 ` [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje @ 2022-01-07 13:40 ` Stefan Mätje 0 siblings, 0 replies; 15+ messages in thread From: Stefan Mätje @ 2022-01-07 13:40 UTC (permalink / raw) To: linux-can, wg, mkl; +Cc: netdev, linux-kernel Hi, is there any chance to get this patch https://lore.kernel.org/linux-can/20211201220328.3079270-1-stefan.maetje@esd.eu/ in the upcoming 5.17 release? The patch still applies cleanly to linux-can-next "testing" at this tag: https://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next.git/tag/?h=linux-can-next-for-5.17-20220105 If it helps I would provide a patch v7 that is rebased to the mentioned tag. Any comment is appreciated. Best regards, Stefan Mätje Am Mittwoch, den 01.12.2021, 22:09 +0000 schrieb Stefan Mätje: > Hi, > > this is really patch v6. Missed to change the email subject. My bad. > > Best regards, > Stefan Mätje > > Am Mittwoch, den 01.12.2021, 23:03 +0100 schrieb Stefan Mätje: > > The purpose of this patch is to introduce a new CAN driver to support > > the esd GmbH 402 family of CAN interface boards. The hardware design > > is based on a CAN controller implemented in a FPGA attached to a > > PCIe link. > > > > More information on these boards can be found following the links > > included in the commit message. > > > > This patch supports all boards but will operate the CAN-FD capable > > boards only in Classic-CAN mode. The CAN-FD support will be added > > when the initial patch has stabilized. > > > > The patch is reuses the previous work of my former colleague: > > Link: > > https://lore.kernel.org/linux-can/1426592308-23817-1-git-send-email-thomas.koerper@esd.eu/ > > > > *Note*: scripts/checkpatch.pl still emits the following warnings: > > - esd_402_pci-core.c:270: Possible unnecessary 'out of memory' message > > This error message is there to tell the user that the DMA allocation > > failed and not an allocation for normal kernel memory. > > - esdacc.h:255: The irq_cnt pointer is still declared volatile and > > this has a reason and is explained in detail in the header > > referencing the exception noted in volatile-considered-harmful.rst. > > > > The patch is based on the linux-can-next testing branch. > > > > Changes in v6: > > - Fixed the statistic handling of RX overrun errors and increase > > net_device_stats::rx_errors instead of net_device_stats::rx_dropped. > > - Added a patch to not increase rx statistics when generating a CAN > > rx error message frame as suggested on the linux-can list. > > - Added a patch to not not increase rx_bytes statistics for RTR frames > > as suggested on the linux-can list. > > > > The last two patches change the statistics handling from the previous > > style used in other drivers to the newly suggested one. > > ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje ` (4 preceding siblings ...) 2021-12-01 22:09 ` [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje @ 2022-01-25 16:25 ` Marc Kleine-Budde 2022-01-27 19:14 ` Stefan Mätje 2022-02-01 17:27 ` Marc Kleine-Budde 2022-02-02 12:21 ` Marc Kleine-Budde 7 siblings, 1 reply; 15+ messages in thread From: Marc Kleine-Budde @ 2022-01-25 16:25 UTC (permalink / raw) To: Stefan Mätje; +Cc: linux-can, Wolfgang Grandegger, netdev, linux-kernel [-- Attachment #1: Type: text/plain, Size: 1062 bytes --] On 01.12.2021 23:03:24, Stefan Mätje wrote: > *Note*: scripts/checkpatch.pl still emits the following warnings: > - esd_402_pci-core.c:270: Possible unnecessary 'out of memory' message > This error message is there to tell the user that the DMA allocation > failed and not an allocation for normal kernel memory. The kernel takes care of printing a error message in case the DMA mem allocation fails. This is why checkpatch asks you to remove that message. > - esdacc.h:255: The irq_cnt pointer is still declared volatile and > this has a reason and is explained in detail in the header > referencing the exception noted in volatile-considered-harmful.rst. I'll look into this, I'll probably ask you to explain the IRQ demux to me :) regards, Marc -- Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 488 bytes --] ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface 2022-01-25 16:25 ` Marc Kleine-Budde @ 2022-01-27 19:14 ` Stefan Mätje 2022-01-28 8:11 ` Marc Kleine-Budde 0 siblings, 1 reply; 15+ messages in thread From: Stefan Mätje @ 2022-01-27 19:14 UTC (permalink / raw) To: mkl; +Cc: netdev, linux-kernel, linux-can, wg Am Dienstag, den 25.01.2022, 17:25 +0100 schrieb Marc Kleine-Budde: > On 01.12.2021 23:03:24, Stefan Mätje wrote: > > *Note*: scripts/checkpatch.pl still emits the following warnings: > > - esd_402_pci-core.c:270: Possible unnecessary 'out of memory' message > > This error message is there to tell the user that the DMA allocation > > failed and not an allocation for normal kernel memory. > > The kernel takes care of printing a error message in case the DMA mem > allocation fails. This is why checkpatch asks you to remove that message. Hello Marc, I've triggered a kernel failure message for the DMA allocation by exceeding the DMA allocation limit. If you say it is clear from that message that the DMA allocation fails (see included kernel messages below) I'll throw out the extra "DMA alloc failed" message from the esd_402_pci driver. The driver's message is shown for both boards in my computer while the kernel message is printed only once. See the kernel messages below. Best regards, Stefan System Design Phone: +49-511-37298-146 E-Mail: stefan.maetje@esd.eu _______________________________________ esd electronics gmbh Vahrenwalder Str. 207 30165 Hannover www.esd.eu Quality Products – Made in Germany _______________________________________ Register Hannover HRB 51373 - VAT-ID DE 115672832 General Manager: Klaus Detering [ 778.083472] esd_402_pci 0000:01:00.0: ESDACC v71, freq: 80000000/80000000, feat/strap: 0x3c90/0x143f, cores: 4/4 [ 778.083479] acc_init_ov:146: esd_402_pci 0000:01:00.0: ESDACC ts2ns: numerator 25, denominator 2 [ 778.083481] esd_402_pci 0000:01:00.0: ESDACC with CAN-FD feature detected. This driver doesn't support CAN-FD yet. [ 778.083489] ------------[ cut here ]------------ [ 778.083490] WARNING: CPU: 0 PID: 24545 at mm/page_alloc.c:5344 __alloc_pages+0x292/0x330 [ 778.083496] Modules linked in: esd_402_pci(+) tcp_diag inet_diag unix_diag nfsv3 nfs_acl rpcsec_gss_krb5 auth_rpcgss nfsv4 nfs lockd grace fscache netfs binfmt_misc nls_iso8859_1 intel_rapl_msr mei_hdcp intel_rapl_common snd_hda_codec_hdmi x86_pkg_temp_thermal intel_powerclamp snd_hda_codec_realtek snd_hda_codec_generic ledtrig_audio coretemp kvm_intel kvm crct10dif_pclmul ghash_clmulni_intel aesni_intel crypto_simd snd_hda_intel snd_intel_dspcfg cryptd snd_intel_sdw_acpi rapl snd_hda_codec intel_cstate snd_hda_core esd_usb2 input_leds snd_hwdep snd_pcm snd_seq_midi snd_seq_midi_event snd_rawmidi i915 serio_raw at24 efi_pstore video ttm snd_seq drm_kms_helper plx_pci sja1000 snd_seq_device cec snd_timer rc_core can_dev i2c_algo_bit snd fb_sys_fops syscopyarea sysfillrect mei_me sysimgblt soundcore mei mac_hid sch_fq_codel msr parport_pc ppdev lp drm parport sunrpc ip_tables x_tables autofs4 gpio_ich crc32_pclmul psmouse i2c_i801 r8169 realtek ahci i2c_smbus libahci lpc_ich wmi [ 778.083547] [last unloaded: esd_402_pci] [ 778.083549] CPU: 0 PID: 24545 Comm: modprobe Not tainted 5.16.0-rc7-gcc-9+ #2 [ 778.083551] Hardware name: Acer Veriton M2610G/Veriton M2610G, BIOS P01-B1 06/18/2012 [ 778.083553] RIP: 0010:__alloc_pages+0x292/0x330 [ 778.083556] Code: d4 89 94 7e 0f 85 3d ff ff ff e8 e8 ab d2 ff e9 33 ff ff ff e8 bf cf fb ff 48 89 c7 e9 a5 fe ff ff 41 81 e4 00 20 00 00 75 8e <0f> 0b eb 8a a9 00 00 08 00 75 6a 44 89 e1 80 e1 7f a9 00 00 04 00 [ 778.083557] RSP: 0018:ffffb4e84240b918 EFLAGS: 00010246 [ 778.083559] RAX: 0000000000000000 RBX: 0000000000000a24 RCX: 0000000000000000 [ 778.083561] RDX: 0000000000000000 RSI: 000000000000000c RDI: 0000000000000a24 [ 778.083562] RBP: ffffb4e84240b970 R08: 00000000ffffffff R09: 00000000ffffffff [ 778.083563] R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000000 [ 778.083564] R13: 000000000000000c R14: ffff9a08c0e3d0d0 R15: 0000000001000000 [ 778.083565] FS: 00007f7d730df540(0000) GS:ffff9a09f7a00000(0000) knlGS:0000000000000000 [ 778.083567] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 778.083568] CR2: 00007ffe83526b40 CR3: 0000000008128006 CR4: 00000000000606f0 [ 778.083570] Call Trace: [ 778.083571] <TASK> [ 778.083574] __dma_direct_alloc_pages+0x8e/0x120 [ 778.083578] dma_direct_alloc+0x66/0x2a0 [ 778.083580] dma_alloc_attrs+0x3e/0x50 [ 778.083582] pci402_probe+0x211/0x64a [esd_402_pci] [ 778.083586] ? kernfs_link_sibling+0x99/0xe0 [ 778.083590] local_pci_probe+0x4b/0x90 [ 778.083594] ? pci_match_device+0xde/0x130 [ 778.083597] pci_device_probe+0xd8/0x1d0 [ 778.083600] really_probe+0x1d2/0x3d0 [ 778.083604] __driver_probe_device+0x109/0x180 [ 778.083607] driver_probe_device+0x23/0xa0 [ 778.083610] __driver_attach+0xbd/0x160 [ 778.083613] ? __device_attach_driver+0xe0/0xe0 [ 778.083616] bus_for_each_dev+0x7e/0xc0 [ 778.083619] driver_attach+0x1e/0x20 [ 778.083621] bus_add_driver+0x152/0x1f0 [ 778.083624] driver_register+0x74/0xd0 [ 778.083626] ? 0xffffffffc0c2e000 [ 778.083627] __pci_register_driver+0x68/0x70 [ 778.083630] pci402_driver_init+0x23/0x1000 [esd_402_pci] [ 778.083633] do_one_initcall+0x48/0x210 [ 778.083636] ? kmem_cache_alloc_trace+0x32e/0x3f0 [ 778.083639] do_init_module+0x62/0x250 [ 778.083641] load_module+0x261c/0x2890 [ 778.083645] __do_sys_finit_module+0xbf/0x120 [ 778.083646] ? __do_sys_finit_module+0xbf/0x120 [ 778.083649] __x64_sys_finit_module+0x1a/0x20 [ 778.083650] do_syscall_64+0x3b/0xc0 [ 778.083654] entry_SYSCALL_64_after_hwframe+0x44/0xae [ 778.083657] RIP: 0033:0x7f7d7322489d [ 778.083658] Code: 00 c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d c3 f5 0c 00 f7 d8 64 89 01 48 [ 778.083660] RSP: 002b:00007ffe83529b68 EFLAGS: 00000246 ORIG_RAX: 0000000000000139 [ 778.083662] RAX: ffffffffffffffda RBX: 000055db638f40d0 RCX: 00007f7d7322489d [ 778.083663] RDX: 0000000000000000 RSI: 000055db638f4730 RDI: 0000000000000003 [ 778.083664] RBP: 0000000000040000 R08: 0000000000000000 R09: 0000000000000000 [ 778.083665] R10: 0000000000000003 R11: 0000000000000246 R12: 000055db638f4730 [ 778.083666] R13: 0000000000000000 R14: 000055db638f43e0 R15: 000055db638f40d0 [ 778.083669] </TASK> [ 778.083670] ---[ end trace 8ac4b05d87d41ee5 ]--- [ 778.083671] esd_402_pci 0000:01:00.0: DMA alloc failed! [ 778.083710] esd_402_pci: probe of 0000:01:00.0 failed with error -12 [ 778.095803] esd_402_pci 0000:05:00.0: ESDACC v72, freq: 80000000/80000000, feat/strap: 0x3c90/0x3d, cores: 2/4 [ 778.095810] acc_init_ov:146: esd_402_pci 0000:05:00.0: ESDACC ts2ns: numerator 25, denominator 2 [ 778.095812] esd_402_pci 0000:05:00.0: ESDACC with CAN-FD feature detected. This driver doesn't support CAN-FD yet. [ 778.095822] esd_402_pci 0000:05:00.0: DMA alloc failed! [ 778.095882] esd_402_pci: probe of 0000:05:00.0 failed with error -12 ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface 2022-01-27 19:14 ` Stefan Mätje @ 2022-01-28 8:11 ` Marc Kleine-Budde 0 siblings, 0 replies; 15+ messages in thread From: Marc Kleine-Budde @ 2022-01-28 8:11 UTC (permalink / raw) To: Stefan Mätje; +Cc: netdev, linux-kernel, linux-can, wg [-- Attachment #1: Type: text/plain, Size: 6503 bytes --] On 27.01.2022 19:14:04, Stefan Mätje wrote: > > The kernel takes care of printing a error message in case the DMA mem > > allocation fails. This is why checkpatch asks you to remove that message. > > I've triggered a kernel failure message for the DMA allocation by exceeding > the DMA allocation limit. If you say it is clear from that message that the > DMA allocation fails (see included kernel messages below) I'll throw out the > extra "DMA alloc failed" message from the esd_402_pci driver. The driver's > message is shown for both boards in my computer while the kernel message > is printed only once. > > See the kernel messages below. > [ 778.083489] ------------[ cut here ]------------ > [ 778.083490] WARNING: CPU: 0 PID: 24545 at mm/page_alloc.c:5344 __alloc_pages+0x292/0x330 > [ 778.083496] Modules linked in: esd_402_pci(+) tcp_diag inet_diag unix_diag nfsv3 nfs_acl rpcsec_gss_krb5 auth_rpcgss nfsv4 nfs > lockd grace fscache netfs binfmt_misc nls_iso8859_1 intel_rapl_msr mei_hdcp intel_rapl_common snd_hda_codec_hdmi > x86_pkg_temp_thermal intel_powerclamp snd_hda_codec_realtek snd_hda_codec_generic ledtrig_audio coretemp kvm_intel kvm > crct10dif_pclmul ghash_clmulni_intel aesni_intel crypto_simd snd_hda_intel snd_intel_dspcfg cryptd snd_intel_sdw_acpi rapl > snd_hda_codec intel_cstate snd_hda_core esd_usb2 input_leds snd_hwdep snd_pcm snd_seq_midi snd_seq_midi_event snd_rawmidi i915 > serio_raw at24 efi_pstore video ttm snd_seq drm_kms_helper plx_pci sja1000 snd_seq_device cec snd_timer rc_core can_dev i2c_algo_bit > snd fb_sys_fops syscopyarea sysfillrect mei_me sysimgblt soundcore mei mac_hid sch_fq_codel msr parport_pc ppdev lp drm parport > sunrpc ip_tables x_tables autofs4 gpio_ich crc32_pclmul psmouse i2c_i801 r8169 realtek ahci i2c_smbus libahci lpc_ich wmi > [ 778.083547] [last unloaded: esd_402_pci] > [ 778.083549] CPU: 0 PID: 24545 Comm: modprobe Not tainted 5.16.0-rc7-gcc-9+ #2 > [ 778.083551] Hardware name: Acer Veriton M2610G/Veriton M2610G, BIOS P01-B1 06/18/2012 > [ 778.083553] RIP: 0010:__alloc_pages+0x292/0x330 > [ 778.083556] Code: d4 89 94 7e 0f 85 3d ff ff ff e8 e8 ab d2 ff e9 33 ff ff ff e8 bf cf fb ff 48 89 c7 e9 a5 fe ff ff 41 81 e4 00 > 20 00 00 75 8e <0f> 0b eb 8a a9 00 00 08 00 75 6a 44 89 e1 80 e1 7f a9 00 00 04 00 > [ 778.083557] RSP: 0018:ffffb4e84240b918 EFLAGS: 00010246 > [ 778.083559] RAX: 0000000000000000 RBX: 0000000000000a24 RCX: 0000000000000000 > [ 778.083561] RDX: 0000000000000000 RSI: 000000000000000c RDI: 0000000000000a24 > [ 778.083562] RBP: ffffb4e84240b970 R08: 00000000ffffffff R09: 00000000ffffffff > [ 778.083563] R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000000 > [ 778.083564] R13: 000000000000000c R14: ffff9a08c0e3d0d0 R15: 0000000001000000 > [ 778.083565] FS: 00007f7d730df540(0000) GS:ffff9a09f7a00000(0000) knlGS:0000000000000000 > [ 778.083567] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > [ 778.083568] CR2: 00007ffe83526b40 CR3: 0000000008128006 CR4: 00000000000606f0 > [ 778.083570] Call Trace: > [ 778.083571] <TASK> VVVVVVVVVVVVVVVVVVVVVVVV > [ 778.083574] __dma_direct_alloc_pages+0x8e/0x120 > [ 778.083578] dma_direct_alloc+0x66/0x2a0 > [ 778.083580] dma_alloc_attrs+0x3e/0x50 ^^^^^^^^^^^^^^^ > [ 778.083582] pci402_probe+0x211/0x64a [esd_402_pci] > [ 778.083586] ? kernfs_link_sibling+0x99/0xe0 > [ 778.083590] local_pci_probe+0x4b/0x90 > [ 778.083594] ? pci_match_device+0xde/0x130 > [ 778.083597] pci_device_probe+0xd8/0x1d0 > [ 778.083600] really_probe+0x1d2/0x3d0 > [ 778.083604] __driver_probe_device+0x109/0x180 > [ 778.083607] driver_probe_device+0x23/0xa0 > [ 778.083610] __driver_attach+0xbd/0x160 > [ 778.083613] ? __device_attach_driver+0xe0/0xe0 > [ 778.083616] bus_for_each_dev+0x7e/0xc0 > [ 778.083619] driver_attach+0x1e/0x20 > [ 778.083621] bus_add_driver+0x152/0x1f0 > [ 778.083624] driver_register+0x74/0xd0 > [ 778.083626] ? 0xffffffffc0c2e000 > [ 778.083627] __pci_register_driver+0x68/0x70 > [ 778.083630] pci402_driver_init+0x23/0x1000 [esd_402_pci] > [ 778.083633] do_one_initcall+0x48/0x210 > [ 778.083636] ? kmem_cache_alloc_trace+0x32e/0x3f0 > [ 778.083639] do_init_module+0x62/0x250 > [ 778.083641] load_module+0x261c/0x2890 > [ 778.083645] __do_sys_finit_module+0xbf/0x120 > [ 778.083646] ? __do_sys_finit_module+0xbf/0x120 > [ 778.083649] __x64_sys_finit_module+0x1a/0x20 > [ 778.083650] do_syscall_64+0x3b/0xc0 > [ 778.083654] entry_SYSCALL_64_after_hwframe+0x44/0xae > [ 778.083657] RIP: 0033:0x7f7d7322489d > [ 778.083658] Code: 00 c3 66 2e 0f 1f 84 00 00 00 00 00 90 f3 0f 1e fa 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b > 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d c3 f5 0c 00 f7 d8 64 89 01 48 > [ 778.083660] RSP: 002b:00007ffe83529b68 EFLAGS: 00000246 ORIG_RAX: 0000000000000139 > [ 778.083662] RAX: ffffffffffffffda RBX: 000055db638f40d0 RCX: 00007f7d7322489d > [ 778.083663] RDX: 0000000000000000 RSI: 000055db638f4730 RDI: 0000000000000003 > [ 778.083664] RBP: 0000000000040000 R08: 0000000000000000 R09: 0000000000000000 > [ 778.083665] R10: 0000000000000003 R11: 0000000000000246 R12: 000055db638f4730 > [ 778.083666] R13: 0000000000000000 R14: 000055db638f43e0 R15: 000055db638f40d0 > [ 778.083669] </TASK> > [ 778.083670] ---[ end trace 8ac4b05d87d41ee5 ]--- > [ 778.083671] esd_402_pci 0000:01:00.0: DMA alloc failed! > [ 778.083710] esd_402_pci: probe of 0000:01:00.0 failed with error -12 > [ 778.095803] esd_402_pci 0000:05:00.0: ESDACC v72, freq: 80000000/80000000, feat/strap: 0x3c90/0x3d, cores: 2/4 > [ 778.095810] acc_init_ov:146: esd_402_pci 0000:05:00.0: ESDACC ts2ns: numerator 25, denominator 2 > [ 778.095812] esd_402_pci 0000:05:00.0: ESDACC with CAN-FD feature detected. This driver doesn't support CAN-FD yet. > [ 778.095822] esd_402_pci 0000:05:00.0: DMA alloc failed! > [ 778.095882] esd_402_pci: probe of 0000:05:00.0 failed with error -12 Looking at the back trace, at least to a kernel developer, it's clear that this is DMA allocation failure. Marc -- Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 488 bytes --] ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje ` (5 preceding siblings ...) 2022-01-25 16:25 ` Marc Kleine-Budde @ 2022-02-01 17:27 ` Marc Kleine-Budde 2022-02-02 12:21 ` Marc Kleine-Budde 7 siblings, 0 replies; 15+ messages in thread From: Marc Kleine-Budde @ 2022-02-01 17:27 UTC (permalink / raw) To: Stefan Mätje; +Cc: linux-can, Wolfgang Grandegger, netdev, linux-kernel [-- Attachment #1: Type: text/plain, Size: 2199 bytes --] On 01.12.2021 23:03:24, Stefan Mätje wrote: > The purpose of this patch is to introduce a new CAN driver to support > the esd GmbH 402 family of CAN interface boards. The hardware design > is based on a CAN controller implemented in a FPGA attached to a > PCIe link. > > More information on these boards can be found following the links > included in the commit message. > > This patch supports all boards but will operate the CAN-FD capable > boards only in Classic-CAN mode. The CAN-FD support will be added > when the initial patch has stabilized. > > The patch is reuses the previous work of my former colleague: > Link: https://lore.kernel.org/linux-can/1426592308-23817-1-git-send-email-thomas.koerper@esd.eu/ > > *Note*: scripts/checkpatch.pl still emits the following warnings: > - esd_402_pci-core.c:270: Possible unnecessary 'out of memory' message > This error message is there to tell the user that the DMA allocation > failed and not an allocation for normal kernel memory. > - esdacc.h:255: The irq_cnt pointer is still declared volatile and > this has a reason and is explained in detail in the header > referencing the exception noted in volatile-considered-harmful.rst. > > The patch is based on the linux-can-next testing branch. > > Changes in v6: > - Fixed the statistic handling of RX overrun errors and increase > net_device_stats::rx_errors instead of net_device_stats::rx_dropped. > - Added a patch to not increase rx statistics when generating a CAN > rx error message frame as suggested on the linux-can list. > - Added a patch to not not increase rx_bytes statistics for RTR frames > as suggested on the linux-can list. > > The last two patches change the statistics handling from the previous > style used in other drivers to the newly suggested one. Please squash the last 2 patches into patch 2. Marc -- Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 488 bytes --] ^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje ` (6 preceding siblings ...) 2022-02-01 17:27 ` Marc Kleine-Budde @ 2022-02-02 12:21 ` Marc Kleine-Budde 7 siblings, 0 replies; 15+ messages in thread From: Marc Kleine-Budde @ 2022-02-02 12:21 UTC (permalink / raw) To: Stefan Mätje; +Cc: linux-can, Wolfgang Grandegger, netdev, linux-kernel [-- Attachment #1: Type: text/plain, Size: 1574 bytes --] On 01.12.2021 23:03:24, Stefan Mätje wrote: > The purpose of this patch is to introduce a new CAN driver to support > the esd GmbH 402 family of CAN interface boards. The hardware design > is based on a CAN controller implemented in a FPGA attached to a > PCIe link. > > More information on these boards can be found following the links > included in the commit message. > > This patch supports all boards but will operate the CAN-FD capable > boards only in Classic-CAN mode. The CAN-FD support will be added > when the initial patch has stabilized. > > The patch is reuses the previous work of my former colleague: > Link: https://lore.kernel.org/linux-can/1426592308-23817-1-git-send-email-thomas.koerper@esd.eu/ > > *Note*: scripts/checkpatch.pl still emits the following warnings: > - esd_402_pci-core.c:270: Possible unnecessary 'out of memory' message > This error message is there to tell the user that the DMA allocation > failed and not an allocation for normal kernel memory. > - esdacc.h:255: The irq_cnt pointer is still declared volatile and > this has a reason and is explained in detail in the header > referencing the exception noted in volatile-considered-harmful.rst. I think you can use READ_ONCE() instead of the volatile. Marc -- Pengutronix e.K. | Marc Kleine-Budde | Embedded Linux | https://www.pengutronix.de | Vertretung West/Dortmund | Phone: +49-231-2826-924 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | [-- Attachment #2: signature.asc --] [-- Type: application/pgp-signature, Size: 488 bytes --] ^ permalink raw reply [flat|nested] 15+ messages in thread
end of thread, other threads:[~2022-10-05 16:04 UTC | newest] Thread overview: 15+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2021-12-01 22:03 [PATCH 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje 2021-12-01 22:03 ` [PATCH 1/4] MAINTAINERS: add Stefan Mätje as maintainer for the esd electronics GmbH CAN drivers Stefan Mätje 2021-12-01 22:03 ` [PATCH 2/4] can: esd: add support for esd GmbH PCIe/402 CAN interface family Stefan Mätje 2022-02-01 17:25 ` [PATCH v6 " Marc Kleine-Budde 2022-09-30 22:15 ` Stefan Mätje 2022-10-05 16:04 ` Marc Kleine-Budde 2021-12-01 22:03 ` [PATCH 3/4] can: esd_402_pci: do not increase rx statistics when generating a CAN rx error message frame Stefan Mätje 2021-12-01 22:03 ` [PATCH 4/4] can: esd_402_pci: do not increase rx_bytes statistics for RTR frames Stefan Mätje 2021-12-01 22:09 ` [PATCH v6 0/4] can: esd: add support for esd GmbH PCIe/402 CAN interface Stefan Mätje 2022-01-07 13:40 ` Stefan Mätje 2022-01-25 16:25 ` Marc Kleine-Budde 2022-01-27 19:14 ` Stefan Mätje 2022-01-28 8:11 ` Marc Kleine-Budde 2022-02-01 17:27 ` Marc Kleine-Budde 2022-02-02 12:21 ` Marc Kleine-Budde
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).