* [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version.
@ 2015-12-09 8:37 Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 01/18] stmmac: share reset function between dwmac100 and dwmac1000 Giuseppe Cavallaro
` (17 more replies)
0 siblings, 18 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
This is a subset of patches to rework the driver
in order to improve its performances and make it
more robust under stress conditions.
All patches have been ported on STi mainstream kernel branch and
tested on ARM STiH4xx platforms and newer ones.
This series also updates the driver version and prepares it
to include the new development for GMAC4 generation.
This will be on top of these patch train.
In detail, these patches are:
o to rework and improve the internal DMA bus settings
Fine tuning is mandatory on some platforms for both
performance and stability issues.
o to rework and optimize the descriptor management.
This will help a lot on performance side and preparing
the inclusion on the GMAC4.
o to add a set of optimizations for both xmit and rx functions.
These will help a lot on performance side and making the driver
more robust in case of low memory conditions and under some
stress test, performed for example on IP-STB.
Below some throughput figures obtained on some boxes before and after
the patches.
nuttcp (mbps) iperf (Mbps)
------------------------------------------------------------------
tcp udp tcp udp
tx rx tx rx tx rx tx rx
------------------------------------------
old 680 800 480 506 760 800 600 700
new 830 880 540 630 840 880 700 800
Fabrice Gasnier (3):
stmmac: merge get_rx_owner into rx_status routine.
stmmac: optimize tx clean function
stmmac: fix phy init when attached to a phy
Giuseppe Cavallaro (15):
stmmac: share reset function between dwmac100 and dwmac1000
stmmac: rework DMA bus setting and introduce new platform AXI
structure
stmmac: change descriptor layout
stmmac: remove modulo in stmmac_xmit()
stmmac: add length field to dma data
stmmac: add last_segment field to dma data
stmmac: add is_jumbo field to dma data
stmmac: optimize tx desc management
stmmac: set dirty index out of the loop
stmmac: first frame prep at the end of xmit routine
stmmac: perf, remove modulo in stmmac_rx()
stmmac: do not poll phy handler when attach a switch
stmmac: do not perform zero-copy for rx frames
stmmac: tune rx copy via threshold.
stmmac: update version to Oct_2015
Documentation/devicetree/bindings/net/stmmac.txt | 45 ++-
drivers/net/ethernet/stmicro/stmmac/chain_mode.c | 35 ++-
drivers/net/ethernet/stmicro/stmmac/common.h | 34 ++-
drivers/net/ethernet/stmicro/stmmac/descs.h | 330 +++++++++-----------
drivers/net/ethernet/stmicro/stmmac/descs_com.h | 77 ++---
drivers/net/ethernet/stmicro/stmmac/dwmac100.h | 1 -
drivers/net/ethernet/stmicro/stmmac/dwmac1000.h | 3 +-
.../net/ethernet/stmicro/stmmac/dwmac1000_dma.c | 111 ++++---
drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c | 22 +-
drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h | 39 +++-
drivers/net/ethernet/stmicro/stmmac/dwmac_lib.c | 21 ++
drivers/net/ethernet/stmicro/stmmac/enh_desc.c | 226 ++++++++------
drivers/net/ethernet/stmicro/stmmac/norm_desc.c | 150 +++++----
drivers/net/ethernet/stmicro/stmmac/ring_mode.c | 32 ++-
drivers/net/ethernet/stmicro/stmmac/stmmac.h | 6 +-
.../net/ethernet/stmicro/stmmac/stmmac_ethtool.c | 2 +-
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 344 ++++++++++++++------
drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c | 4 +-
.../net/ethernet/stmicro/stmmac/stmmac_platform.c | 42 +++-
include/linux/stmmac.h | 17 +-
20 files changed, 928 insertions(+), 613 deletions(-)
--
1.7.4.4
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 01/18] stmmac: share reset function between dwmac100 and dwmac1000
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 02/18] stmmac: rework DMA bus setting and introduce new platform AXI structure Giuseppe Cavallaro
` (16 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
This patch is to share the same reset procedure between dwmac100 and
dwmac1000 chips.
This will also help on enhancing the driver and support new chips.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/common.h | 3 +-
drivers/net/ethernet/stmicro/stmmac/dwmac100.h | 1 -
drivers/net/ethernet/stmicro/stmmac/dwmac1000.h | 1 -
.../net/ethernet/stmicro/stmmac/dwmac1000_dma.c | 22 +++----------------
drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c | 22 ++-----------------
drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h | 5 ++++
drivers/net/ethernet/stmicro/stmmac/dwmac_lib.c | 21 +++++++++++++++++++
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 14 ++++++++++--
8 files changed, 46 insertions(+), 43 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/common.h b/drivers/net/ethernet/stmicro/stmmac/common.h
index f4518bc..b4f4950 100644
--- a/drivers/net/ethernet/stmicro/stmmac/common.h
+++ b/drivers/net/ethernet/stmicro/stmmac/common.h
@@ -376,7 +376,8 @@ extern const struct stmmac_desc_ops ndesc_ops;
/* Specific DMA helpers */
struct stmmac_dma_ops {
/* DMA core initialization */
- int (*init) (void __iomem *ioaddr, int pbl, int fb, int mb,
+ int (*reset)(void __iomem *ioaddr);
+ void (*init)(void __iomem *ioaddr, int pbl, int fb, int mb,
int burst_len, u32 dma_tx, u32 dma_rx, int atds);
/* Dump DMA registers */
void (*dump_regs) (void __iomem *ioaddr);
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac100.h b/drivers/net/ethernet/stmicro/stmmac/dwmac100.h
index 2ec6aea..1657acf 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac100.h
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac100.h
@@ -95,7 +95,6 @@
#define DMA_BUS_MODE_DSL_MASK 0x0000007c /* Descriptor Skip Length */
#define DMA_BUS_MODE_DSL_SHIFT 2 /* (in DWORDS) */
#define DMA_BUS_MODE_BAR_BUS 0x00000002 /* Bar-Bus Arbitration */
-#define DMA_BUS_MODE_SFT_RESET 0x00000001 /* Software Reset */
#define DMA_BUS_MODE_DEFAULT 0x00000000
/* DMA Control register defines */
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h b/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
index 8831a05..9d36ae7 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
@@ -221,7 +221,6 @@ enum inter_frame_gap {
/*--- DMA BLOCK defines ---*/
/* DMA Bus Mode register defines */
-#define DMA_BUS_MODE_SFT_RESET 0x00000001 /* Software Reset */
#define DMA_BUS_MODE_DA 0x00000002 /* Arbitration scheme */
#define DMA_BUS_MODE_DSL_MASK 0x0000007c /* Descriptor Skip Length */
#define DMA_BUS_MODE_DSL_SHIFT 2 /* (in DWORDS) */
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c b/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
index 0e8937c..5f0aea5 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
@@ -30,23 +30,10 @@
#include "dwmac1000.h"
#include "dwmac_dma.h"
-static int dwmac1000_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
- int burst_len, u32 dma_tx, u32 dma_rx, int atds)
+static void dwmac1000_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
+ int burst_len, u32 dma_tx, u32 dma_rx, int atds)
{
- u32 value = readl(ioaddr + DMA_BUS_MODE);
- int limit;
-
- /* DMA SW reset */
- value |= DMA_BUS_MODE_SFT_RESET;
- writel(value, ioaddr + DMA_BUS_MODE);
- limit = 10;
- while (limit--) {
- if (!(readl(ioaddr + DMA_BUS_MODE) & DMA_BUS_MODE_SFT_RESET))
- break;
- mdelay(10);
- }
- if (limit < 0)
- return -EBUSY;
+ u32 value;
/*
* Set the DMA PBL (Programmable Burst Length) mode
@@ -102,8 +89,6 @@ static int dwmac1000_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
*/
writel(dma_tx, ioaddr + DMA_TX_BASE_ADDR);
writel(dma_rx, ioaddr + DMA_RCV_BASE_ADDR);
-
- return 0;
}
static u32 dwmac1000_configure_fc(u32 csr6, int rxfifosz)
@@ -205,6 +190,7 @@ static void dwmac1000_rx_watchdog(void __iomem *ioaddr, u32 riwt)
}
const struct stmmac_dma_ops dwmac1000_dma_ops = {
+ .reset = dwmac_dma_reset,
.init = dwmac1000_dma_init,
.dump_regs = dwmac1000_dump_dma_regs,
.dma_mode = dwmac1000_dma_operation_mode,
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c b/drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c
index 9d0971c..c40582a 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c
@@ -32,24 +32,9 @@
#include "dwmac100.h"
#include "dwmac_dma.h"
-static int dwmac100_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
- int burst_len, u32 dma_tx, u32 dma_rx, int atds)
+static void dwmac100_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
+ int burst_len, u32 dma_tx, u32 dma_rx, int atds)
{
- u32 value = readl(ioaddr + DMA_BUS_MODE);
- int limit;
-
- /* DMA SW reset */
- value |= DMA_BUS_MODE_SFT_RESET;
- writel(value, ioaddr + DMA_BUS_MODE);
- limit = 10;
- while (limit--) {
- if (!(readl(ioaddr + DMA_BUS_MODE) & DMA_BUS_MODE_SFT_RESET))
- break;
- mdelay(10);
- }
- if (limit < 0)
- return -EBUSY;
-
/* Enable Application Access by writing to DMA CSR0 */
writel(DMA_BUS_MODE_DEFAULT | (pbl << DMA_BUS_MODE_PBL_SHIFT),
ioaddr + DMA_BUS_MODE);
@@ -62,8 +47,6 @@ static int dwmac100_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
*/
writel(dma_tx, ioaddr + DMA_TX_BASE_ADDR);
writel(dma_rx, ioaddr + DMA_RCV_BASE_ADDR);
-
- return 0;
}
/* Store and Forward capability is not used at all.
@@ -131,6 +114,7 @@ static void dwmac100_dma_diagnostic_fr(void *data, struct stmmac_extra_stats *x,
}
const struct stmmac_dma_ops dwmac100_dma_ops = {
+ .reset = dwmac_dma_reset,
.init = dwmac100_dma_init,
.dump_regs = dwmac100_dump_dma_regs,
.dma_mode = dwmac100_dma_operation_mode,
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h b/drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h
index def266d..13ca90e 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h
@@ -35,6 +35,10 @@
#define DMA_CONTROL 0x00001018 /* Ctrl (Operational Mode) */
#define DMA_INTR_ENA 0x0000101c /* Interrupt Enable */
#define DMA_MISSED_FRAME_CTR 0x00001020 /* Missed Frame Counter */
+
+/* SW Reset */
+#define DMA_BUS_MODE_SFT_RESET 0x00000001 /* Software Reset */
+
/* Rx watchdog register */
#define DMA_RX_WATCHDOG 0x00001024
/* AXI Bus Mode */
@@ -112,5 +116,6 @@ void dwmac_dma_stop_tx(void __iomem *ioaddr);
void dwmac_dma_start_rx(void __iomem *ioaddr);
void dwmac_dma_stop_rx(void __iomem *ioaddr);
int dwmac_dma_interrupt(void __iomem *ioaddr, struct stmmac_extra_stats *x);
+int dwmac_dma_reset(void __iomem *ioaddr);
#endif /* __DWMAC_DMA_H__ */
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac_lib.c b/drivers/net/ethernet/stmicro/stmmac/dwmac_lib.c
index 484e3cf..84e3e84 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac_lib.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac_lib.c
@@ -26,6 +26,27 @@
#define GMAC_HI_REG_AE 0x80000000
+int dwmac_dma_reset(void __iomem *ioaddr)
+{
+ u32 value = readl(ioaddr + DMA_BUS_MODE);
+ int limit;
+
+ /* DMA SW reset */
+ value |= DMA_BUS_MODE_SFT_RESET;
+ writel(value, ioaddr + DMA_BUS_MODE);
+ limit = 10;
+ while (limit--) {
+ if (!(readl(ioaddr + DMA_BUS_MODE) & DMA_BUS_MODE_SFT_RESET))
+ break;
+ mdelay(10);
+ }
+
+ if (limit < 0)
+ return -EBUSY;
+
+ return 0;
+}
+
/* CSR1 enables the transmit DMA to check for new descriptor */
void dwmac_enable_dma_transmission(void __iomem *ioaddr)
{
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 3c6549a..8b75c4b 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1641,6 +1641,7 @@ static int stmmac_init_dma_engine(struct stmmac_priv *priv)
int pbl = DEFAULT_DMA_PBL, fixed_burst = 0, burst_len = 0;
int mixed_burst = 0;
int atds = 0;
+ int ret = 0;
if (priv->plat->dma_cfg) {
pbl = priv->plat->dma_cfg->pbl;
@@ -1652,9 +1653,16 @@ static int stmmac_init_dma_engine(struct stmmac_priv *priv)
if (priv->extend_desc && (priv->mode == STMMAC_RING_MODE))
atds = 1;
- return priv->hw->dma->init(priv->ioaddr, pbl, fixed_burst, mixed_burst,
- burst_len, priv->dma_tx_phy,
- priv->dma_rx_phy, atds);
+ ret = priv->hw->dma->reset(priv->ioaddr);
+ if (ret) {
+ dev_err(priv->device, "Failed to reset the dma\n");
+ return ret;
+ }
+
+ priv->hw->dma->init(priv->ioaddr, pbl, fixed_burst, mixed_burst,
+ burst_len, priv->dma_tx_phy,
+ priv->dma_rx_phy, atds);
+ return ret;
}
/**
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 02/18] stmmac: rework DMA bus setting and introduce new platform AXI structure
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 01/18] stmmac: share reset function between dwmac100 and dwmac1000 Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 03/18] stmmac: change descriptor layout Giuseppe Cavallaro
` (15 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
This patch restructures the DMA bus settings and this is done
by introducing a new platform structure used for programming
the AXI Bus Mode Register inside the DMA module.
This structure can be populated from device-tree as documented in the
binding txt file.
After initializing the DMA, the AXI register can be optionally tuned
for platform drivers based.
This patch also reworks some parameters to make coherent the DMA
configuration now that AXI register is introduced.
For example, the burst_len is managed by using the mentioned axi
support above; so the snps,burst-len parameter has been removed.
It makes sense to provide the AAL parameter from DT to Address-Aligned
Beats inside the Register0 and review the PBL settings when initialize
the engine.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
Documentation/devicetree/bindings/net/stmmac.txt | 45 +++++++---
drivers/net/ethernet/stmicro/stmmac/common.h | 5 +-
drivers/net/ethernet/stmicro/stmmac/dwmac1000.h | 2 +-
.../net/ethernet/stmicro/stmmac/dwmac1000_dma.c | 99 +++++++++++++------
drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c | 2 +-
drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h | 34 +++++++-
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 12 ++-
drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c | 4 +-
.../net/ethernet/stmicro/stmmac/stmmac_platform.c | 42 ++++++++-
include/linux/stmmac.h | 17 +++-
10 files changed, 204 insertions(+), 58 deletions(-)
diff --git a/Documentation/devicetree/bindings/net/stmmac.txt b/Documentation/devicetree/bindings/net/stmmac.txt
index f34fc3c..36720d0 100644
--- a/Documentation/devicetree/bindings/net/stmmac.txt
+++ b/Documentation/devicetree/bindings/net/stmmac.txt
@@ -17,18 +17,6 @@ Required properties:
The 1st cell is reset pre-delay in micro seconds.
The 2nd cell is reset pulse in micro seconds.
The 3rd cell is reset post-delay in micro seconds.
-- snps,pbl Programmable Burst Length
-- snps,fixed-burst Program the DMA to use the fixed burst mode
-- snps,mixed-burst Program the DMA to use the mixed burst mode
-- snps,force_thresh_dma_mode Force DMA to use the threshold mode for
- both tx and rx
-- snps,force_sf_dma_mode Force DMA to use the Store and Forward
- mode for both tx and rx. This flag is
- ignored if force_thresh_dma_mode is set.
-- snps,multicast-filter-bins: Number of multicast filter hash bins
- supported by this device instance
-- snps,perfect-filter-entries: Number of perfect filter entries supported
- by this device instance
Optional properties:
- resets: Should contain a phandle to the STMMAC reset signal, if any
@@ -44,12 +32,42 @@ Optional properties:
available this clock is used for programming the Timestamp Addend Register.
If not passed then the system clock will be used and this is fine on some
platforms.
-- snps,burst_len: The AXI burst lenth value of the AXI BUS MODE register.
- tx-fifo-depth: See ethernet.txt file in the same directory
- rx-fifo-depth: See ethernet.txt file in the same directory
+- snps,pbl Programmable Burst Length
+- snps,aal Address-Aligned Beats
+- snps,fixed-burst Program the DMA to use the fixed burst mode
+- snps,mixed-burst Program the DMA to use the mixed burst mode
+- snps,force_thresh_dma_mode Force DMA to use the threshold mode for
+ both tx and rx
+- snps,force_sf_dma_mode Force DMA to use the Store and Forward
+ mode for both tx and rx. This flag is
+ ignored if force_thresh_dma_mode is set.
+- snps,multicast-filter-bins: Number of multicast filter hash bins
+ supported by this device instance
+- snps,perfect-filter-entries: Number of perfect filter entries supported
+ by this device instance
+- AXI BUS Mode parameters: below the list of all the parameters to program the
+ AXI register inside the DMA module:
+ - snps,lpi_en: enable Low Power Interface
+ - snps,xit_frm: unlock on WoL
+ - snps,wr_osr_lmt: max write oustanding req. limit
+ - snps,rd_osr_lmt: max read oustanding req. limit
+ - snps,kbbe: do not cross 1KiB boundary.
+ - snps,axi_all: align address
+ - snps,blen: this is a vector of supported burst length.
+ - snps,fb: fixed-burst
+ - snps,mb: mixed-burst
+ - snps,rb: rebuild INCRx Burst
Examples:
+ stmmac_axi_setup: stmmac-axi-config {
+ snps,wr_osr_lmt = <0xf>;
+ snps,rd_osr_lmt = <0xf>;
+ snps,blen = <256 128 64 32 0 0 0>;
+ };
+
gmac0: ethernet@e0800000 {
compatible = "st,spear600-gmac";
reg = <0xe0800000 0x8000>;
@@ -65,4 +83,5 @@ Examples:
tx-fifo-depth = <16384>;
clocks = <&clock>;
clock-names = "stmmaceth";
+ snps,axi-config = <&stmmac_axi_setup>;
};
diff --git a/drivers/net/ethernet/stmicro/stmmac/common.h b/drivers/net/ethernet/stmicro/stmmac/common.h
index b4f4950..cdc3806 100644
--- a/drivers/net/ethernet/stmicro/stmmac/common.h
+++ b/drivers/net/ethernet/stmicro/stmmac/common.h
@@ -27,6 +27,7 @@
#include <linux/etherdevice.h>
#include <linux/netdevice.h>
+#include <linux/stmmac.h>
#include <linux/phy.h>
#include <linux/module.h>
#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
@@ -378,7 +379,9 @@ struct stmmac_dma_ops {
/* DMA core initialization */
int (*reset)(void __iomem *ioaddr);
void (*init)(void __iomem *ioaddr, int pbl, int fb, int mb,
- int burst_len, u32 dma_tx, u32 dma_rx, int atds);
+ int aal, u32 dma_tx, u32 dma_rx, int atds);
+ /* Configure the AXI Bus Mode Register */
+ void (*axi)(void __iomem *ioaddr, struct stmmac_axi *axi);
/* Dump DMA registers */
void (*dump_regs) (void __iomem *ioaddr);
/* Set tx/rx threshold in the csr6 register
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h b/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
index 9d36ae7..b0593a4 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac1000.h
@@ -240,7 +240,7 @@ enum rx_tx_priority_ratio {
#define DMA_BUS_MODE_RPBL_MASK 0x003e0000 /* Rx-Programmable Burst Len */
#define DMA_BUS_MODE_RPBL_SHIFT 17
#define DMA_BUS_MODE_USP 0x00800000
-#define DMA_BUS_MODE_PBL 0x01000000
+#define DMA_BUS_MODE_MAXPBL 0x01000000
#define DMA_BUS_MODE_AAL 0x02000000
/* DMA CRS Control and Status Register Mapping */
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c b/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
index 5f0aea5..da32d60 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac1000_dma.c
@@ -30,24 +30,76 @@
#include "dwmac1000.h"
#include "dwmac_dma.h"
+static void dwmac1000_dma_axi(void __iomem *ioaddr, struct stmmac_axi *axi)
+{
+ u32 value = readl(ioaddr + DMA_AXI_BUS_MODE);
+ int i;
+
+ pr_info("dwmac1000: Master AXI performs %s burst length\n",
+ !(value & DMA_AXI_UNDEF) ? "fixed" : "any");
+
+ if (axi->axi_lpi_en)
+ value |= DMA_AXI_EN_LPI;
+ if (axi->axi_xit_frm)
+ value |= DMA_AXI_LPI_XIT_FRM;
+
+ value |= (axi->axi_wr_osr_lmt & DMA_AXI_WR_OSR_LMT_MASK) <<
+ DMA_AXI_WR_OSR_LMT_SHIFT;
+
+ value |= (axi->axi_rd_osr_lmt & DMA_AXI_RD_OSR_LMT_MASK) <<
+ DMA_AXI_RD_OSR_LMT_SHIFT;
+
+ /* Depending on the UNDEF bit the Master AXI will perform any burst
+ * length according to the BLEN programmed (by default all BLEN are
+ * set).
+ */
+ for (i = 0; i < AXI_BLEN; i++) {
+ switch (axi->axi_blen[i]) {
+ case 256:
+ value |= DMA_AXI_BLEN256;
+ break;
+ case 128:
+ value |= DMA_AXI_BLEN128;
+ break;
+ case 64:
+ value |= DMA_AXI_BLEN64;
+ break;
+ case 32:
+ value |= DMA_AXI_BLEN32;
+ break;
+ case 16:
+ value |= DMA_AXI_BLEN16;
+ break;
+ case 8:
+ value |= DMA_AXI_BLEN8;
+ break;
+ case 4:
+ value |= DMA_AXI_BLEN4;
+ break;
+ }
+ }
+
+ writel(value, ioaddr + DMA_AXI_BUS_MODE);
+}
+
static void dwmac1000_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
- int burst_len, u32 dma_tx, u32 dma_rx, int atds)
+ int aal, u32 dma_tx, u32 dma_rx, int atds)
{
- u32 value;
+ u32 value = readl(ioaddr + DMA_BUS_MODE);
/*
- * Set the DMA PBL (Programmable Burst Length) mode
- * Before stmmac core 3.50 this mode bit was 4xPBL, and
+ * Set the DMA PBL (Programmable Burst Length) mode.
+ *
+ * Note: before stmmac core 3.50 this mode bit was 4xPBL, and
* post 3.5 mode bit acts as 8*PBL.
- * For core rev < 3.5, when the core is set for 4xPBL mode, the
- * DMA transfers the data in 4, 8, 16, 32, 64 & 128 beats
- * depending on pbl value.
- * For core rev > 3.5, when the core is set for 8xPBL mode, the
- * DMA transfers the data in 8, 16, 32, 64, 128 & 256 beats
- * depending on pbl value.
+ *
+ * This configuration doesn't take care about the Separate PBL
+ * so only the bits: 13-8 are programmed with the PBL passed from the
+ * platform.
*/
- value = DMA_BUS_MODE_PBL | ((pbl << DMA_BUS_MODE_PBL_SHIFT) |
- (pbl << DMA_BUS_MODE_RPBL_SHIFT));
+ value |= DMA_BUS_MODE_MAXPBL;
+ value &= ~DMA_BUS_MODE_PBL_MASK;
+ value |= (pbl << DMA_BUS_MODE_PBL_SHIFT);
/* Set the Fixed burst mode */
if (fb)
@@ -60,26 +112,10 @@ static void dwmac1000_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
if (atds)
value |= DMA_BUS_MODE_ATDS;
- writel(value, ioaddr + DMA_BUS_MODE);
+ if (aal)
+ value |= DMA_BUS_MODE_AAL;
- /* In case of GMAC AXI configuration, program the DMA_AXI_BUS_MODE
- * for supported bursts.
- *
- * Note: This is applicable only for revision GMACv3.61a. For
- * older version this register is reserved and shall have no
- * effect.
- *
- * Note:
- * For Fixed Burst Mode: if we directly write 0xFF to this
- * register using the configurations pass from platform code,
- * this would ensure that all bursts supported by core are set
- * and those which are not supported would remain ineffective.
- *
- * For Non Fixed Burst Mode: provide the maximum value of the
- * burst length. Any burst equal or below the provided burst
- * length would be allowed to perform.
- */
- writel(burst_len, ioaddr + DMA_AXI_BUS_MODE);
+ writel(value, ioaddr + DMA_BUS_MODE);
/* Mask interrupts by writing to CSR7 */
writel(DMA_INTR_DEFAULT_MASK, ioaddr + DMA_INTR_ENA);
@@ -192,6 +228,7 @@ static void dwmac1000_rx_watchdog(void __iomem *ioaddr, u32 riwt)
const struct stmmac_dma_ops dwmac1000_dma_ops = {
.reset = dwmac_dma_reset,
.init = dwmac1000_dma_init,
+ .axi = dwmac1000_dma_axi,
.dump_regs = dwmac1000_dump_dma_regs,
.dma_mode = dwmac1000_dma_operation_mode,
.enable_dma_transmission = dwmac_enable_dma_transmission,
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c b/drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c
index c40582a..61f54c9 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac100_dma.c
@@ -33,7 +33,7 @@
#include "dwmac_dma.h"
static void dwmac100_dma_init(void __iomem *ioaddr, int pbl, int fb, int mb,
- int burst_len, u32 dma_tx, u32 dma_rx, int atds)
+ int aal, u32 dma_tx, u32 dma_rx, int atds)
{
/* Enable Application Access by writing to DMA CSR0 */
writel(DMA_BUS_MODE_DEFAULT | (pbl << DMA_BUS_MODE_PBL_SHIFT),
diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h b/drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h
index 13ca90e..726d9d9 100644
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac_dma.h
@@ -41,8 +41,40 @@
/* Rx watchdog register */
#define DMA_RX_WATCHDOG 0x00001024
-/* AXI Bus Mode */
+
+/* AXI Master Bus Mode */
#define DMA_AXI_BUS_MODE 0x00001028
+
+#define DMA_AXI_EN_LPI BIT(31)
+#define DMA_AXI_LPI_XIT_FRM BIT(30)
+#define DMA_AXI_WR_OSR_LMT GENMASK(23, 20)
+#define DMA_AXI_WR_OSR_LMT_SHIFT 20
+#define DMA_AXI_WR_OSR_LMT_MASK 0xf
+#define DMA_AXI_RD_OSR_LMT GENMASK(19, 16)
+#define DMA_AXI_RD_OSR_LMT_SHIFT 16
+#define DMA_AXI_RD_OSR_LMT_MASK 0xf
+
+#define DMA_AXI_OSR_MAX 0xf
+#define DMA_AXI_MAX_OSR_LIMIT ((DMA_AXI_OSR_MAX << DMA_AXI_WR_OSR_LMT_SHIFT) | \
+ (DMA_AXI_OSR_MAX << DMA_AXI_RD_OSR_LMT_SHIFT))
+#define DMA_AXI_1KBBE BIT(13)
+#define DMA_AXI_AAL BIT(12)
+#define DMA_AXI_BLEN256 BIT(7)
+#define DMA_AXI_BLEN128 BIT(6)
+#define DMA_AXI_BLEN64 BIT(5)
+#define DMA_AXI_BLEN32 BIT(4)
+#define DMA_AXI_BLEN16 BIT(3)
+#define DMA_AXI_BLEN8 BIT(2)
+#define DMA_AXI_BLEN4 BIT(1)
+#define DMA_BURST_LEN_DEFAULT (DMA_AXI_BLEN256 | DMA_AXI_BLEN128 | \
+ DMA_AXI_BLEN64 | DMA_AXI_BLEN32 | \
+ DMA_AXI_BLEN16 | DMA_AXI_BLEN8 | \
+ DMA_AXI_BLEN4)
+
+#define DMA_AXI_UNDEF BIT(0)
+
+#define DMA_AXI_BURST_LEN_MASK 0x000000FE
+
#define DMA_CUR_TX_BUF_ADDR 0x00001050 /* Current Host Tx Buffer */
#define DMA_CUR_RX_BUF_ADDR 0x00001054 /* Current Host Rx Buffer */
#define DMA_HW_FEATURE 0x00001058 /* HW Feature Register */
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 8b75c4b..2dd0279 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1638,7 +1638,7 @@ static void stmmac_check_ether_addr(struct stmmac_priv *priv)
*/
static int stmmac_init_dma_engine(struct stmmac_priv *priv)
{
- int pbl = DEFAULT_DMA_PBL, fixed_burst = 0, burst_len = 0;
+ int pbl = DEFAULT_DMA_PBL, fixed_burst = 0, aal = 0;
int mixed_burst = 0;
int atds = 0;
int ret = 0;
@@ -1647,7 +1647,7 @@ static int stmmac_init_dma_engine(struct stmmac_priv *priv)
pbl = priv->plat->dma_cfg->pbl;
fixed_burst = priv->plat->dma_cfg->fixed_burst;
mixed_burst = priv->plat->dma_cfg->mixed_burst;
- burst_len = priv->plat->dma_cfg->burst_len;
+ aal = priv->plat->dma_cfg->aal;
}
if (priv->extend_desc && (priv->mode == STMMAC_RING_MODE))
@@ -1660,8 +1660,12 @@ static int stmmac_init_dma_engine(struct stmmac_priv *priv)
}
priv->hw->dma->init(priv->ioaddr, pbl, fixed_burst, mixed_burst,
- burst_len, priv->dma_tx_phy,
- priv->dma_rx_phy, atds);
+ aal, priv->dma_tx_phy, priv->dma_rx_phy, atds);
+
+ if ((priv->synopsys_id >= DWMAC_CORE_3_50) &&
+ (priv->plat->axi && priv->hw->dma->axi))
+ priv->hw->dma->axi(priv->ioaddr, priv->plat->axi);
+
return ret;
}
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
index d71a721..ae43887 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
@@ -81,7 +81,7 @@ static void stmmac_default_data(struct plat_stmmacenet_data *plat)
plat->mdio_bus_data->phy_mask = 0;
plat->dma_cfg->pbl = 32;
- plat->dma_cfg->burst_len = DMA_AXI_BLEN_256;
+ /* TODO: AXI */
/* Set default value for multicast hash bins */
plat->multicast_filter_bins = HASH_TABLE_SIZE;
@@ -115,8 +115,8 @@ static int quark_default_data(struct plat_stmmacenet_data *plat,
plat->mdio_bus_data->phy_mask = 0;
plat->dma_cfg->pbl = 16;
- plat->dma_cfg->burst_len = DMA_AXI_BLEN_256;
plat->dma_cfg->fixed_burst = 1;
+ /* AXI (TODO) */
/* Set default value for multicast hash bins */
plat->multicast_filter_bins = HASH_TABLE_SIZE;
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c
index d02691b..76cc581 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c
@@ -96,6 +96,42 @@ static int dwmac1000_validate_ucast_entries(int ucast_entries)
}
/**
+ * stmmac_axi_setup - parse DT parameters for programming the AXI register
+ * @pdev: platform device
+ * @priv: driver private struct.
+ * Description:
+ * if required, from device-tree the AXI internal register can be tuned
+ * by using platform parameters.
+ */
+static struct stmmac_axi *stmmac_axi_setup(struct platform_device *pdev)
+{
+ struct device_node *np;
+ struct stmmac_axi *axi;
+
+ np = of_parse_phandle(pdev->dev.of_node, "snps,axi-config", 0);
+ if (!np)
+ return NULL;
+
+ axi = kzalloc(sizeof(axi), GFP_KERNEL);
+ if (!axi)
+ return ERR_PTR(-ENOMEM);
+
+ axi->axi_lpi_en = of_property_read_bool(np, "snps,lpi_en");
+ axi->axi_xit_frm = of_property_read_bool(np, "snps,xit_frm");
+ axi->axi_kbbe = of_property_read_bool(np, "snps,axi_kbbe");
+ axi->axi_axi_all = of_property_read_bool(np, "snps,axi_all");
+ axi->axi_fb = of_property_read_bool(np, "snps,axi_fb");
+ axi->axi_mb = of_property_read_bool(np, "snps,axi_mb");
+ axi->axi_rb = of_property_read_bool(np, "snps,axi_rb");
+
+ of_property_read_u32(np, "snps,wr_osr_lmt", &axi->axi_wr_osr_lmt);
+ of_property_read_u32(np, "snps,rd_osr_lmt", &axi->axi_rd_osr_lmt);
+ of_property_read_u32_array(np, "snps,blen", axi->axi_blen, AXI_BLEN);
+
+ return axi;
+}
+
+/**
* stmmac_probe_config_dt - parse device-tree driver parameters
* @pdev: platform_device structure
* @plat: driver data platform structure
@@ -216,13 +252,11 @@ stmmac_probe_config_dt(struct platform_device *pdev, const char **mac)
}
plat->dma_cfg = dma_cfg;
of_property_read_u32(np, "snps,pbl", &dma_cfg->pbl);
+ dma_cfg->aal = of_property_read_bool(np, "snps,aal");
dma_cfg->fixed_burst =
of_property_read_bool(np, "snps,fixed-burst");
dma_cfg->mixed_burst =
of_property_read_bool(np, "snps,mixed-burst");
- of_property_read_u32(np, "snps,burst_len", &dma_cfg->burst_len);
- if (dma_cfg->burst_len < 0 || dma_cfg->burst_len > 256)
- dma_cfg->burst_len = 0;
}
plat->force_thresh_dma_mode = of_property_read_bool(np, "snps,force_thresh_dma_mode");
if (plat->force_thresh_dma_mode) {
@@ -230,6 +264,8 @@ stmmac_probe_config_dt(struct platform_device *pdev, const char **mac)
pr_warn("force_sf_dma_mode is ignored if force_thresh_dma_mode is set.");
}
+ plat->axi = stmmac_axi_setup(pdev);
+
return plat;
}
#else
diff --git a/include/linux/stmmac.h b/include/linux/stmmac.h
index eead8ab..6e53fa8 100644
--- a/include/linux/stmmac.h
+++ b/include/linux/stmmac.h
@@ -90,7 +90,21 @@ struct stmmac_dma_cfg {
int pbl;
int fixed_burst;
int mixed_burst;
- int burst_len;
+ bool aal;
+};
+
+#define AXI_BLEN 7
+struct stmmac_axi {
+ bool axi_lpi_en;
+ bool axi_xit_frm;
+ u32 axi_wr_osr_lmt;
+ u32 axi_rd_osr_lmt;
+ bool axi_kbbe;
+ bool axi_axi_all;
+ u32 axi_blen[AXI_BLEN];
+ bool axi_fb;
+ bool axi_mb;
+ bool axi_rb;
};
struct plat_stmmacenet_data {
@@ -122,5 +136,6 @@ struct plat_stmmacenet_data {
int (*init)(struct platform_device *pdev, void *priv);
void (*exit)(struct platform_device *pdev, void *priv);
void *bsp_priv;
+ struct stmmac_axi *axi;
};
#endif
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 03/18] stmmac: change descriptor layout
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 01/18] stmmac: share reset function between dwmac100 and dwmac1000 Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 02/18] stmmac: rework DMA bus setting and introduce new platform AXI structure Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 04/18] stmmac: remove modulo in stmmac_xmit() Giuseppe Cavallaro
` (14 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
This patch completely changes the descriptor layout to improve
the whole performances due to the single read usage of the
descriptors in critical paths.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/descs.h | 330 ++++++++++-------------
drivers/net/ethernet/stmicro/stmmac/descs_com.h | 77 +++---
drivers/net/ethernet/stmicro/stmmac/enh_desc.c | 190 +++++++------
drivers/net/ethernet/stmicro/stmmac/norm_desc.c | 121 +++++----
4 files changed, 355 insertions(+), 363 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/descs.h b/drivers/net/ethernet/stmicro/stmmac/descs.h
index 799c292..2e4c171 100644
--- a/drivers/net/ethernet/stmicro/stmmac/descs.h
+++ b/drivers/net/ethernet/stmicro/stmmac/descs.h
@@ -1,6 +1,6 @@
/*******************************************************************************
- Header File to describe the DMA descriptors.
- Enhanced descriptors have been in case of DWMAC1000 Cores.
+ Header File to describe the DMA descriptors and related definitions.
+ This is for DWMAC100 and 1000 cores.
This program is free software; you can redistribute it and/or modify it
under the terms and conditions of the GNU General Public License,
@@ -24,198 +24,164 @@
#ifndef __DESCS_H__
#define __DESCS_H__
+#include <linux/bitops.h>
+
+/* Normal receive descriptor defines */
+
+/* RDES0 */
+#define RDES0_PAYLOAD_CSUM_ERR BIT(0)
+#define RDES0_CRC_ERROR BIT(1)
+#define RDES0_DRIBBLING BIT(2)
+#define RDES0_MII_ERROR BIT(3)
+#define RDES0_RECEIVE_WATCHDOG BIT(4)
+#define RDES0_FRAME_TYPE BIT(5)
+#define RDES0_COLLISION BIT(6)
+#define RDES0_IPC_CSUM_ERROR BIT(7)
+#define RDES0_LAST_DESCRIPTOR BIT(8)
+#define RDES0_FIRST_DESCRIPTOR BIT(9)
+#define RDES0_VLAN_TAG BIT(10)
+#define RDES0_OVERFLOW_ERROR BIT(11)
+#define RDES0_LENGTH_ERROR BIT(12)
+#define RDES0_SA_FILTER_FAIL BIT(13)
+#define RDES0_DESCRIPTOR_ERROR BIT(14)
+#define RDES0_ERROR_SUMMARY BIT(15)
+#define RDES0_FRAME_LEN_MASK GENMASK(29, 16)
+#define RDES0_FRAME_LEN_SHIFT 16
+#define RDES0_DA_FILTER_FAIL BIT(30)
+#define RDES0_OWN BIT(31)
+ /* RDES1 */
+#define RDES1_BUFFER1_SIZE_MASK GENMASK(10, 0)
+#define RDES1_BUFFER2_SIZE_MASK GENMASK(21, 11)
+#define RDES1_BUFFER2_SIZE_SHIFT 11
+#define RDES1_SECOND_ADDRESS_CHAINED BIT(24)
+#define RDES1_END_RING BIT(25)
+#define RDES1_DISABLE_IC BIT(31)
+
+/* Enhanced receive descriptor defines */
+
+/* RDES0 (similar to normal RDES) */
+#define ERDES0_RX_MAC_ADDR BIT(0)
+
+/* RDES1: completely differ from normal desc definitions */
+#define ERDES1_BUFFER1_SIZE_MASK GENMASK(12, 0)
+#define ERDES1_SECOND_ADDRESS_CHAINED BIT(14)
+#define ERDES1_END_RING BIT(15)
+#define ERDES1_BUFFER2_SIZE_MASK GENMASK(28, 16)
+#define ERDES1_BUFFER2_SIZE_SHIFT 16
+#define ERDES1_DISABLE_IC BIT(31)
+
+/* Normal transmit descriptor defines */
+/* TDES0 */
+#define TDES0_DEFERRED BIT(0)
+#define TDES0_UNDERFLOW_ERROR BIT(1)
+#define TDES0_EXCESSIVE_DEFERRAL BIT(2)
+#define TDES0_COLLISION_COUNT_MASK GENMASK(6, 3)
+#define TDES0_VLAN_FRAME BIT(7)
+#define TDES0_EXCESSIVE_COLLISIONS BIT(8)
+#define TDES0_LATE_COLLISION BIT(9)
+#define TDES0_NO_CARRIER BIT(10)
+#define TDES0_LOSS_CARRIER BIT(11)
+#define TDES0_PAYLOAD_ERROR BIT(12)
+#define TDES0_FRAME_FLUSHED BIT(13)
+#define TDES0_JABBER_TIMEOUT BIT(14)
+#define TDES0_ERROR_SUMMARY BIT(15)
+#define TDES0_IP_HEADER_ERROR BIT(16)
+#define TDES0_TIME_STAMP_STATUS BIT(17)
+#define TDES0_OWN BIT(31)
+/* TDES1 */
+#define TDES1_BUFFER1_SIZE_MASK GENMASK(10, 0)
+#define TDES1_BUFFER2_SIZE_MASK GENMASK(21, 11)
+#define TDES1_BUFFER2_SIZE_SHIFT 11
+#define TDES1_TIME_STAMP_ENABLE BIT(22)
+#define TDES1_DISABLE_PADDING BIT(23)
+#define TDES1_SECOND_ADDRESS_CHAINED BIT(24)
+#define TDES1_END_RING BIT(25)
+#define TDES1_CRC_DISABLE BIT(26)
+#define TDES1_CHECKSUM_INSERTION_MASK GENMASK(28, 27)
+#define TDES1_CHECKSUM_INSERTION_SHIFT 27
+#define TDES1_FIRST_SEGMENT BIT(29)
+#define TDES1_LAST_SEGMENT BIT(30)
+#define TDES1_INTERRUPT BIT(31)
+
+/* Enhanced transmit descriptor defines */
+/* TDES0 */
+#define ETDES0_DEFERRED BIT(0)
+#define ETDES0_UNDERFLOW_ERROR BIT(1)
+#define ETDES0_EXCESSIVE_DEFERRAL BIT(2)
+#define ETDES0_COLLISION_COUNT_MASK GENMASK(6, 3)
+#define ETDES0_VLAN_FRAME BIT(7)
+#define ETDES0_EXCESSIVE_COLLISIONS BIT(8)
+#define ETDES0_LATE_COLLISION BIT(9)
+#define ETDES0_NO_CARRIER BIT(10)
+#define ETDES0_LOSS_CARRIER BIT(11)
+#define ETDES0_PAYLOAD_ERROR BIT(12)
+#define ETDES0_FRAME_FLUSHED BIT(13)
+#define ETDES0_JABBER_TIMEOUT BIT(14)
+#define ETDES0_ERROR_SUMMARY BIT(15)
+#define ETDES0_IP_HEADER_ERROR BIT(16)
+#define ETDES0_TIME_STAMP_STATUS BIT(17)
+#define ETDES0_SECOND_ADDRESS_CHAINED BIT(20)
+#define ETDES0_END_RING BIT(21)
+#define ETDES0_CHECKSUM_INSERTION_MASK GENMASK(23, 22)
+#define ETDES0_CHECKSUM_INSERTION_SHIFT 22
+#define ETDES0_TIME_STAMP_ENABLE BIT(25)
+#define ETDES0_DISABLE_PADDING BIT(26)
+#define ETDES0_CRC_DISABLE BIT(27)
+#define ETDES0_FIRST_SEGMENT BIT(28)
+#define ETDES0_LAST_SEGMENT BIT(29)
+#define ETDES0_INTERRUPT BIT(30)
+#define ETDES0_OWN BIT(31)
+/* TDES1 */
+#define ETDES1_BUFFER1_SIZE_MASK GENMASK(12, 0)
+#define ETDES1_BUFFER2_SIZE_MASK GENMASK(28, 16)
+#define ETDES1_BUFFER2_SIZE_SHIFT 16
+
+/* Extended Receive descriptor definitions */
+#define ERDES4_IP_PAYLOAD_TYPE_MASK GENMASK(2, 6)
+#define ERDES4_IP_HDR_ERR BIT(3)
+#define ERDES4_IP_PAYLOAD_ERR BIT(4)
+#define ERDES4_IP_CSUM_BYPASSED BIT(5)
+#define ERDES4_IPV4_PKT_RCVD BIT(6)
+#define ERDES4_IPV6_PKT_RCVD BIT(7)
+#define ERDES4_MSG_TYPE_MASK GENMASK(11, 8)
+#define ERDES4_PTP_FRAME_TYPE BIT(12)
+#define ERDES4_PTP_VER BIT(13)
+#define ERDES4_TIMESTAMP_DROPPED BIT(14)
+#define ERDES4_AV_PKT_RCVD BIT(16)
+#define ERDES4_AV_TAGGED_PKT_RCVD BIT(17)
+#define ERDES4_VLAN_TAG_PRI_VAL_MASK GENMASK(20, 18)
+#define ERDES4_L3_FILTER_MATCH BIT(24)
+#define ERDES4_L4_FILTER_MATCH BIT(25)
+#define ERDES4_L3_L4_FILT_NO_MATCH_MASK GENMASK(27, 26)
+
+/* Extended RDES4 message type definitions */
+#define RDES_EXT_NO_PTP 0
+#define RDES_EXT_SYNC 1
+#define RDES_EXT_FOLLOW_UP 2
+#define RDES_EXT_DELAY_REQ 3
+#define RDES_EXT_DELAY_RESP 4
+#define RDES_EXT_PDELAY_REQ 5
+#define RDES_EXT_PDELAY_RESP 6
+#define RDES_EXT_PDELAY_FOLLOW_UP 7
+
/* Basic descriptor structure for normal and alternate descriptors */
struct dma_desc {
- /* Receive descriptor */
- union {
- struct {
- /* RDES0 */
- u32 payload_csum_error:1;
- u32 crc_error:1;
- u32 dribbling:1;
- u32 mii_error:1;
- u32 receive_watchdog:1;
- u32 frame_type:1;
- u32 collision:1;
- u32 ipc_csum_error:1;
- u32 last_descriptor:1;
- u32 first_descriptor:1;
- u32 vlan_tag:1;
- u32 overflow_error:1;
- u32 length_error:1;
- u32 sa_filter_fail:1;
- u32 descriptor_error:1;
- u32 error_summary:1;
- u32 frame_length:14;
- u32 da_filter_fail:1;
- u32 own:1;
- /* RDES1 */
- u32 buffer1_size:11;
- u32 buffer2_size:11;
- u32 reserved1:2;
- u32 second_address_chained:1;
- u32 end_ring:1;
- u32 reserved2:5;
- u32 disable_ic:1;
-
- } rx;
- struct {
- /* RDES0 */
- u32 rx_mac_addr:1;
- u32 crc_error:1;
- u32 dribbling:1;
- u32 error_gmii:1;
- u32 receive_watchdog:1;
- u32 frame_type:1;
- u32 late_collision:1;
- u32 ipc_csum_error:1;
- u32 last_descriptor:1;
- u32 first_descriptor:1;
- u32 vlan_tag:1;
- u32 overflow_error:1;
- u32 length_error:1;
- u32 sa_filter_fail:1;
- u32 descriptor_error:1;
- u32 error_summary:1;
- u32 frame_length:14;
- u32 da_filter_fail:1;
- u32 own:1;
- /* RDES1 */
- u32 buffer1_size:13;
- u32 reserved1:1;
- u32 second_address_chained:1;
- u32 end_ring:1;
- u32 buffer2_size:13;
- u32 reserved2:2;
- u32 disable_ic:1;
- } erx; /* -- enhanced -- */
-
- /* Transmit descriptor */
- struct {
- /* TDES0 */
- u32 deferred:1;
- u32 underflow_error:1;
- u32 excessive_deferral:1;
- u32 collision_count:4;
- u32 vlan_frame:1;
- u32 excessive_collisions:1;
- u32 late_collision:1;
- u32 no_carrier:1;
- u32 loss_carrier:1;
- u32 payload_error:1;
- u32 frame_flushed:1;
- u32 jabber_timeout:1;
- u32 error_summary:1;
- u32 ip_header_error:1;
- u32 time_stamp_status:1;
- u32 reserved1:13;
- u32 own:1;
- /* TDES1 */
- u32 buffer1_size:11;
- u32 buffer2_size:11;
- u32 time_stamp_enable:1;
- u32 disable_padding:1;
- u32 second_address_chained:1;
- u32 end_ring:1;
- u32 crc_disable:1;
- u32 checksum_insertion:2;
- u32 first_segment:1;
- u32 last_segment:1;
- u32 interrupt:1;
- } tx;
- struct {
- /* TDES0 */
- u32 deferred:1;
- u32 underflow_error:1;
- u32 excessive_deferral:1;
- u32 collision_count:4;
- u32 vlan_frame:1;
- u32 excessive_collisions:1;
- u32 late_collision:1;
- u32 no_carrier:1;
- u32 loss_carrier:1;
- u32 payload_error:1;
- u32 frame_flushed:1;
- u32 jabber_timeout:1;
- u32 error_summary:1;
- u32 ip_header_error:1;
- u32 time_stamp_status:1;
- u32 reserved1:2;
- u32 second_address_chained:1;
- u32 end_ring:1;
- u32 checksum_insertion:2;
- u32 reserved2:1;
- u32 time_stamp_enable:1;
- u32 disable_padding:1;
- u32 crc_disable:1;
- u32 first_segment:1;
- u32 last_segment:1;
- u32 interrupt:1;
- u32 own:1;
- /* TDES1 */
- u32 buffer1_size:13;
- u32 reserved3:3;
- u32 buffer2_size:13;
- u32 reserved4:3;
- } etx; /* -- enhanced -- */
-
- u64 all_flags;
- } des01;
+ unsigned int des0;
+ unsigned int des1;
unsigned int des2;
unsigned int des3;
};
-/* Extended descriptor structure (supported by new SYNP GMAC generations) */
+/* Extended descriptor structure (e.g. >= databook 3.50a) */
struct dma_extended_desc {
- struct dma_desc basic;
- union {
- struct {
- u32 ip_payload_type:3;
- u32 ip_hdr_err:1;
- u32 ip_payload_err:1;
- u32 ip_csum_bypassed:1;
- u32 ipv4_pkt_rcvd:1;
- u32 ipv6_pkt_rcvd:1;
- u32 msg_type:4;
- u32 ptp_frame_type:1;
- u32 ptp_ver:1;
- u32 timestamp_dropped:1;
- u32 reserved:1;
- u32 av_pkt_rcvd:1;
- u32 av_tagged_pkt_rcvd:1;
- u32 vlan_tag_priority_val:3;
- u32 reserved3:3;
- u32 l3_filter_match:1;
- u32 l4_filter_match:1;
- u32 l3_l4_filter_no_match:2;
- u32 reserved4:4;
- } erx;
- struct {
- u32 reserved;
- } etx;
- } des4;
+ struct dma_desc basic; /* Basic descriptors */
+ unsigned int des4; /* Extended Status */
unsigned int des5; /* Reserved */
unsigned int des6; /* Tx/Rx Timestamp Low */
unsigned int des7; /* Tx/Rx Timestamp High */
};
/* Transmit checksum insertion control */
-enum tdes_csum_insertion {
- cic_disabled = 0, /* Checksum Insertion Control */
- cic_only_ip = 1, /* Only IP header */
- /* IP header but pseudoheader is not calculated */
- cic_no_pseudoheader = 2,
- cic_full = 3, /* IP header and pseudoheader */
-};
-
-/* Extended RDES4 definitions */
-#define RDES_EXT_NO_PTP 0
-#define RDES_EXT_SYNC 0x1
-#define RDES_EXT_FOLLOW_UP 0x2
-#define RDES_EXT_DELAY_REQ 0x3
-#define RDES_EXT_DELAY_RESP 0x4
-#define RDES_EXT_PDELAY_REQ 0x5
-#define RDES_EXT_PDELAY_RESP 0x6
-#define RDES_EXT_PDELAY_FOLLOW_UP 0x7
+#define TX_CIC_FULL 3 /* Include IP header and pseudoheader */
#endif /* __DESCS_H__ */
diff --git a/drivers/net/ethernet/stmicro/stmmac/descs_com.h b/drivers/net/ethernet/stmicro/stmmac/descs_com.h
index 6f2cc78..7635a46 100644
--- a/drivers/net/ethernet/stmicro/stmmac/descs_com.h
+++ b/drivers/net/ethernet/stmicro/stmmac/descs_com.h
@@ -35,100 +35,91 @@
/* Enhanced descriptors */
static inline void ehn_desc_rx_set_on_ring(struct dma_desc *p, int end)
{
- p->des01.erx.buffer2_size = BUF_SIZE_8KiB - 1;
- if (end)
- p->des01.erx.end_ring = 1;
-}
+ p->des1 |= ((BUF_SIZE_8KiB - 1) << ERDES1_BUFFER2_SIZE_SHIFT)
+ & ERDES1_BUFFER2_SIZE_MASK;
-static inline void ehn_desc_tx_set_on_ring(struct dma_desc *p, int end)
-{
if (end)
- p->des01.etx.end_ring = 1;
+ p->des1 |= ERDES1_END_RING;
}
-static inline void enh_desc_end_tx_desc_on_ring(struct dma_desc *p, int ter)
+static inline void enh_desc_end_tx_desc_on_ring(struct dma_desc *p, int end)
{
- p->des01.etx.end_ring = ter;
+ if (end)
+ p->des0 |= ETDES0_END_RING;
+ else
+ p->des0 &= ~ETDES0_END_RING;
}
static inline void enh_set_tx_desc_len_on_ring(struct dma_desc *p, int len)
{
if (unlikely(len > BUF_SIZE_4KiB)) {
- p->des01.etx.buffer1_size = BUF_SIZE_4KiB;
- p->des01.etx.buffer2_size = len - BUF_SIZE_4KiB;
+ p->des1 |= (((len - BUF_SIZE_4KiB) << ETDES1_BUFFER2_SIZE_SHIFT)
+ & ETDES1_BUFFER2_SIZE_MASK) | (BUF_SIZE_4KiB
+ & ETDES1_BUFFER1_SIZE_MASK);
} else
- p->des01.etx.buffer1_size = len;
+ p->des1 |= (len & ETDES1_BUFFER1_SIZE_MASK);
}
/* Normal descriptors */
static inline void ndesc_rx_set_on_ring(struct dma_desc *p, int end)
{
- p->des01.rx.buffer2_size = BUF_SIZE_2KiB - 1;
- if (end)
- p->des01.rx.end_ring = 1;
-}
+ p->des1 |= ((BUF_SIZE_2KiB - 1) << RDES1_BUFFER2_SIZE_SHIFT)
+ & RDES1_BUFFER2_SIZE_MASK;
-static inline void ndesc_tx_set_on_ring(struct dma_desc *p, int end)
-{
if (end)
- p->des01.tx.end_ring = 1;
+ p->des1 |= RDES1_END_RING;
}
-static inline void ndesc_end_tx_desc_on_ring(struct dma_desc *p, int ter)
+static inline void ndesc_end_tx_desc_on_ring(struct dma_desc *p, int end)
{
- p->des01.tx.end_ring = ter;
+ if (end)
+ p->des1 |= TDES1_END_RING;
+ else
+ p->des1 &= ~TDES1_END_RING;
}
static inline void norm_set_tx_desc_len_on_ring(struct dma_desc *p, int len)
{
if (unlikely(len > BUF_SIZE_2KiB)) {
- p->des01.etx.buffer1_size = BUF_SIZE_2KiB - 1;
- p->des01.etx.buffer2_size = len - p->des01.etx.buffer1_size;
+ unsigned int buffer1 = (BUF_SIZE_2KiB - 1)
+ & TDES1_BUFFER1_SIZE_MASK;
+ p->des1 |= ((((len - buffer1) << TDES1_BUFFER2_SIZE_SHIFT)
+ & TDES1_BUFFER2_SIZE_MASK) | buffer1);
} else
- p->des01.tx.buffer1_size = len;
+ p->des1 |= (len & TDES1_BUFFER1_SIZE_MASK);
}
/* Specific functions used for Chain mode */
/* Enhanced descriptors */
-static inline void ehn_desc_rx_set_on_chain(struct dma_desc *p, int end)
-{
- p->des01.erx.second_address_chained = 1;
-}
-
-static inline void ehn_desc_tx_set_on_chain(struct dma_desc *p, int end)
+static inline void ehn_desc_rx_set_on_chain(struct dma_desc *p)
{
- p->des01.etx.second_address_chained = 1;
+ p->des1 |= ERDES1_SECOND_ADDRESS_CHAINED;
}
-static inline void enh_desc_end_tx_desc_on_chain(struct dma_desc *p, int ter)
+static inline void enh_desc_end_tx_desc_on_chain(struct dma_desc *p)
{
- p->des01.etx.second_address_chained = 1;
+ p->des0 |= ETDES0_SECOND_ADDRESS_CHAINED;
}
static inline void enh_set_tx_desc_len_on_chain(struct dma_desc *p, int len)
{
- p->des01.etx.buffer1_size = len;
+ p->des1 |= (len & ETDES1_BUFFER1_SIZE_MASK);
}
/* Normal descriptors */
static inline void ndesc_rx_set_on_chain(struct dma_desc *p, int end)
{
- p->des01.rx.second_address_chained = 1;
-}
-
-static inline void ndesc_tx_set_on_chain(struct dma_desc *p, int ring_size)
-{
- p->des01.tx.second_address_chained = 1;
+ p->des1 |= RDES1_SECOND_ADDRESS_CHAINED;
}
-static inline void ndesc_end_tx_desc_on_chain(struct dma_desc *p, int ter)
+static inline void ndesc_tx_set_on_chain(struct dma_desc *p)
{
- p->des01.tx.second_address_chained = 1;
+ p->des1 |= TDES1_SECOND_ADDRESS_CHAINED;
}
static inline void norm_set_tx_desc_len_on_chain(struct dma_desc *p, int len)
{
- p->des01.tx.buffer1_size = len;
+ p->des1 |= len & TDES1_BUFFER1_SIZE_MASK;
}
#endif /* __DESC_COM_H__ */
diff --git a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
index 7d94444..716b807 100644
--- a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
@@ -1,7 +1,7 @@
/*******************************************************************************
This contains the functions to handle the enhanced descriptors.
- Copyright (C) 2007-2009 STMicroelectronics Ltd
+ Copyright (C) 2007-2014 STMicroelectronics Ltd
This program is free software; you can redistribute it and/or modify it
under the terms and conditions of the GNU General Public License,
@@ -29,44 +29,44 @@
static int enh_desc_get_tx_status(void *data, struct stmmac_extra_stats *x,
struct dma_desc *p, void __iomem *ioaddr)
{
- int ret = 0;
struct net_device_stats *stats = (struct net_device_stats *)data;
+ unsigned int tdes0 = p->des0;
+ int ret = 0;
- if (unlikely(p->des01.etx.error_summary)) {
- if (unlikely(p->des01.etx.jabber_timeout))
+ if (unlikely(tdes0 & ETDES0_ERROR_SUMMARY)) {
+ if (unlikely(tdes0 & ETDES0_JABBER_TIMEOUT))
x->tx_jabber++;
- if (unlikely(p->des01.etx.frame_flushed)) {
+ if (unlikely(tdes0 & ETDES0_FRAME_FLUSHED)) {
x->tx_frame_flushed++;
dwmac_dma_flush_tx_fifo(ioaddr);
}
- if (unlikely(p->des01.etx.loss_carrier)) {
+ if (unlikely(tdes0 & ETDES0_LOSS_CARRIER)) {
x->tx_losscarrier++;
stats->tx_carrier_errors++;
}
- if (unlikely(p->des01.etx.no_carrier)) {
+ if (unlikely(tdes0 & ETDES0_NO_CARRIER)) {
x->tx_carrier++;
stats->tx_carrier_errors++;
}
- if (unlikely(p->des01.etx.late_collision))
- stats->collisions += p->des01.etx.collision_count;
-
- if (unlikely(p->des01.etx.excessive_collisions))
- stats->collisions += p->des01.etx.collision_count;
+ if (unlikely((tdes0 & ETDES0_LATE_COLLISION) ||
+ (tdes0 & ETDES0_EXCESSIVE_COLLISIONS)))
+ stats->collisions +=
+ (tdes0 & ETDES0_COLLISION_COUNT_MASK) >> 3;
- if (unlikely(p->des01.etx.excessive_deferral))
+ if (unlikely(tdes0 & ETDES0_EXCESSIVE_DEFERRAL))
x->tx_deferred++;
- if (unlikely(p->des01.etx.underflow_error)) {
+ if (unlikely(tdes0 & ETDES0_UNDERFLOW_ERROR)) {
dwmac_dma_flush_tx_fifo(ioaddr);
x->tx_underflow++;
}
- if (unlikely(p->des01.etx.ip_header_error))
+ if (unlikely(tdes0 & ETDES0_IP_HEADER_ERROR))
x->tx_ip_header_error++;
- if (unlikely(p->des01.etx.payload_error)) {
+ if (unlikely(tdes0 & ETDES0_PAYLOAD_ERROR)) {
x->tx_payload_error++;
dwmac_dma_flush_tx_fifo(ioaddr);
}
@@ -74,11 +74,11 @@ static int enh_desc_get_tx_status(void *data, struct stmmac_extra_stats *x,
ret = -1;
}
- if (unlikely(p->des01.etx.deferred))
+ if (unlikely(tdes0 & ETDES0_DEFERRED))
x->tx_deferred++;
#ifdef STMMAC_VLAN_TAG_USED
- if (p->des01.etx.vlan_frame)
+ if (tdes0 & ETDES0_VLAN_FRAME)
x->tx_vlan++;
#endif
@@ -87,7 +87,7 @@ static int enh_desc_get_tx_status(void *data, struct stmmac_extra_stats *x,
static int enh_desc_get_tx_len(struct dma_desc *p)
{
- return p->des01.etx.buffer1_size;
+ return (p->des1 & ETDES1_BUFFER1_SIZE_MASK);
}
static int enh_desc_coe_rdes0(int ipc_err, int type, int payload_err)
@@ -126,50 +126,55 @@ static int enh_desc_coe_rdes0(int ipc_err, int type, int payload_err)
static void enh_desc_get_ext_status(void *data, struct stmmac_extra_stats *x,
struct dma_extended_desc *p)
{
- if (unlikely(p->basic.des01.erx.rx_mac_addr)) {
- if (p->des4.erx.ip_hdr_err)
+ unsigned int rdes0 = p->basic.des0;
+ unsigned int rdes4 = p->des4;
+
+ if (unlikely(rdes0 & ERDES0_RX_MAC_ADDR)) {
+ int message_type = (rdes4 & ERDES4_MSG_TYPE_MASK) >> 8;
+
+ if (rdes4 & ERDES4_IP_HDR_ERR)
x->ip_hdr_err++;
- if (p->des4.erx.ip_payload_err)
+ if (rdes4 & ERDES4_IP_PAYLOAD_ERR)
x->ip_payload_err++;
- if (p->des4.erx.ip_csum_bypassed)
+ if (rdes4 & ERDES4_IP_CSUM_BYPASSED)
x->ip_csum_bypassed++;
- if (p->des4.erx.ipv4_pkt_rcvd)
+ if (rdes4 & ERDES4_IPV4_PKT_RCVD)
x->ipv4_pkt_rcvd++;
- if (p->des4.erx.ipv6_pkt_rcvd)
+ if (rdes4 & ERDES4_IPV6_PKT_RCVD)
x->ipv6_pkt_rcvd++;
- if (p->des4.erx.msg_type == RDES_EXT_SYNC)
+ if (message_type == RDES_EXT_SYNC)
x->rx_msg_type_sync++;
- else if (p->des4.erx.msg_type == RDES_EXT_FOLLOW_UP)
+ else if (message_type == RDES_EXT_FOLLOW_UP)
x->rx_msg_type_follow_up++;
- else if (p->des4.erx.msg_type == RDES_EXT_DELAY_REQ)
+ else if (message_type == RDES_EXT_DELAY_REQ)
x->rx_msg_type_delay_req++;
- else if (p->des4.erx.msg_type == RDES_EXT_DELAY_RESP)
+ else if (message_type == RDES_EXT_DELAY_RESP)
x->rx_msg_type_delay_resp++;
- else if (p->des4.erx.msg_type == RDES_EXT_PDELAY_REQ)
+ else if (message_type == RDES_EXT_PDELAY_REQ)
x->rx_msg_type_pdelay_req++;
- else if (p->des4.erx.msg_type == RDES_EXT_PDELAY_RESP)
+ else if (message_type == RDES_EXT_PDELAY_RESP)
x->rx_msg_type_pdelay_resp++;
- else if (p->des4.erx.msg_type == RDES_EXT_PDELAY_FOLLOW_UP)
+ else if (message_type == RDES_EXT_PDELAY_FOLLOW_UP)
x->rx_msg_type_pdelay_follow_up++;
else
x->rx_msg_type_ext_no_ptp++;
- if (p->des4.erx.ptp_frame_type)
+ if (rdes4 & ERDES4_PTP_FRAME_TYPE)
x->ptp_frame_type++;
- if (p->des4.erx.ptp_ver)
+ if (rdes4 & ERDES4_PTP_VER)
x->ptp_ver++;
- if (p->des4.erx.timestamp_dropped)
+ if (rdes4 & ERDES4_TIMESTAMP_DROPPED)
x->timestamp_dropped++;
- if (p->des4.erx.av_pkt_rcvd)
+ if (rdes4 & ERDES4_AV_PKT_RCVD)
x->av_pkt_rcvd++;
- if (p->des4.erx.av_tagged_pkt_rcvd)
+ if (rdes4 & ERDES4_AV_TAGGED_PKT_RCVD)
x->av_tagged_pkt_rcvd++;
- if (p->des4.erx.vlan_tag_priority_val)
+ if ((rdes4 & ERDES4_VLAN_TAG_PRI_VAL_MASK) >> 18)
x->vlan_tag_priority_val++;
- if (p->des4.erx.l3_filter_match)
+ if (rdes4 & ERDES4_L3_FILTER_MATCH)
x->l3_filter_match++;
- if (p->des4.erx.l4_filter_match)
+ if (rdes4 & ERDES4_L4_FILTER_MATCH)
x->l4_filter_match++;
- if (p->des4.erx.l3_l4_filter_no_match)
+ if ((rdes4 & ERDES4_L3_L4_FILT_NO_MATCH_MASK) >> 26)
x->l3_l4_filter_no_match++;
}
}
@@ -177,30 +182,30 @@ static void enh_desc_get_ext_status(void *data, struct stmmac_extra_stats *x,
static int enh_desc_get_rx_status(void *data, struct stmmac_extra_stats *x,
struct dma_desc *p)
{
- int ret = good_frame;
struct net_device_stats *stats = (struct net_device_stats *)data;
+ unsigned int rdes0 = p->des0;
+ int ret = good_frame;
- if (unlikely(p->des01.erx.error_summary)) {
- if (unlikely(p->des01.erx.descriptor_error)) {
+ if (unlikely(rdes0 & RDES0_ERROR_SUMMARY)) {
+ if (unlikely(rdes0 & RDES0_DESCRIPTOR_ERROR)) {
x->rx_desc++;
stats->rx_length_errors++;
}
- if (unlikely(p->des01.erx.overflow_error))
+ if (unlikely(rdes0 & RDES0_OVERFLOW_ERROR))
x->rx_gmac_overflow++;
- if (unlikely(p->des01.erx.ipc_csum_error))
+ if (unlikely(rdes0 & RDES0_IPC_CSUM_ERROR))
pr_err("\tIPC Csum Error/Giant frame\n");
- if (unlikely(p->des01.erx.late_collision)) {
+ if (unlikely(rdes0 & RDES0_COLLISION))
stats->collisions++;
- }
- if (unlikely(p->des01.erx.receive_watchdog))
+ if (unlikely(rdes0 & RDES0_RECEIVE_WATCHDOG))
x->rx_watchdog++;
- if (unlikely(p->des01.erx.error_gmii))
+ if (unlikely(rdes0 & RDES0_MII_ERROR)) /* GMII */
x->rx_mii++;
- if (unlikely(p->des01.erx.crc_error)) {
+ if (unlikely(rdes0 & RDES0_CRC_ERROR)) {
x->rx_crc++;
stats->rx_crc_errors++;
}
@@ -211,26 +216,27 @@ static int enh_desc_get_rx_status(void *data, struct stmmac_extra_stats *x,
* It doesn't match with the information reported into the databook.
* At any rate, we need to understand if the CSUM hw computation is ok
* and report this info to the upper layers. */
- ret = enh_desc_coe_rdes0(p->des01.erx.ipc_csum_error,
- p->des01.erx.frame_type, p->des01.erx.rx_mac_addr);
+ ret = enh_desc_coe_rdes0(!!(rdes0 & RDES0_IPC_CSUM_ERROR),
+ !!(rdes0 & RDES0_FRAME_TYPE),
+ !!(rdes0 & ERDES0_RX_MAC_ADDR));
- if (unlikely(p->des01.erx.dribbling))
+ if (unlikely(rdes0 & RDES0_DRIBBLING))
x->dribbling_bit++;
- if (unlikely(p->des01.erx.sa_filter_fail)) {
+ if (unlikely(rdes0 & RDES0_SA_FILTER_FAIL)) {
x->sa_rx_filter_fail++;
ret = discard_frame;
}
- if (unlikely(p->des01.erx.da_filter_fail)) {
+ if (unlikely(rdes0 & RDES0_DA_FILTER_FAIL)) {
x->da_rx_filter_fail++;
ret = discard_frame;
}
- if (unlikely(p->des01.erx.length_error)) {
+ if (unlikely(rdes0 & RDES0_LENGTH_ERROR)) {
x->rx_length++;
ret = discard_frame;
}
#ifdef STMMAC_VLAN_TAG_USED
- if (p->des01.erx.vlan_tag)
+ if (rdes0 & RDES0_VLAN_TAG)
x->rx_vlan++;
#endif
@@ -240,60 +246,59 @@ static int enh_desc_get_rx_status(void *data, struct stmmac_extra_stats *x,
static void enh_desc_init_rx_desc(struct dma_desc *p, int disable_rx_ic,
int mode, int end)
{
- p->des01.all_flags = 0;
- p->des01.erx.own = 1;
- p->des01.erx.buffer1_size = BUF_SIZE_8KiB - 1;
+ p->des0 |= RDES0_OWN;
+ p->des1 |= ((BUF_SIZE_8KiB - 1) & ERDES1_BUFFER1_SIZE_MASK);
if (mode == STMMAC_CHAIN_MODE)
- ehn_desc_rx_set_on_chain(p, end);
+ ehn_desc_rx_set_on_chain(p);
else
ehn_desc_rx_set_on_ring(p, end);
if (disable_rx_ic)
- p->des01.erx.disable_ic = 1;
+ p->des1 |= ERDES1_DISABLE_IC;
}
static void enh_desc_init_tx_desc(struct dma_desc *p, int mode, int end)
{
- p->des01.all_flags = 0;
+ p->des0 &= ~ETDES0_OWN;
if (mode == STMMAC_CHAIN_MODE)
- ehn_desc_tx_set_on_chain(p, end);
+ enh_desc_end_tx_desc_on_chain(p);
else
- ehn_desc_tx_set_on_ring(p, end);
+ enh_desc_end_tx_desc_on_ring(p, end);
}
static int enh_desc_get_tx_owner(struct dma_desc *p)
{
- return p->des01.etx.own;
+ return (p->des0 & ETDES0_OWN) >> 31;
}
static int enh_desc_get_rx_owner(struct dma_desc *p)
{
- return p->des01.erx.own;
+ return (p->des0 & RDES0_OWN) >> 31;
}
static void enh_desc_set_tx_owner(struct dma_desc *p)
{
- p->des01.etx.own = 1;
+ p->des0 |= ETDES0_OWN;
}
static void enh_desc_set_rx_owner(struct dma_desc *p)
{
- p->des01.erx.own = 1;
+ p->des0 |= RDES0_OWN;
}
static int enh_desc_get_tx_ls(struct dma_desc *p)
{
- return p->des01.etx.last_segment;
+ return (p->des0 & ETDES0_LAST_SEGMENT) >> 29;
}
static void enh_desc_release_tx_desc(struct dma_desc *p, int mode)
{
- int ter = p->des01.etx.end_ring;
+ int ter = (p->des0 & ETDES0_END_RING) >> 21;
memset(p, 0, offsetof(struct dma_desc, des2));
if (mode == STMMAC_CHAIN_MODE)
- enh_desc_end_tx_desc_on_chain(p, ter);
+ enh_desc_end_tx_desc_on_chain(p);
else
enh_desc_end_tx_desc_on_ring(p, ter);
}
@@ -301,49 +306,60 @@ static void enh_desc_release_tx_desc(struct dma_desc *p, int mode)
static void enh_desc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
int csum_flag, int mode)
{
- p->des01.etx.first_segment = is_fs;
+ unsigned int tdes0 = p->des0;
+
+ if (is_fs)
+ tdes0 |= ETDES0_FIRST_SEGMENT;
+ else
+ tdes0 &= ~ETDES0_FIRST_SEGMENT;
+
+ if (likely(csum_flag))
+ tdes0 |= (TX_CIC_FULL << ETDES0_CHECKSUM_INSERTION_SHIFT);
+ else
+ tdes0 &= ~(TX_CIC_FULL << ETDES0_CHECKSUM_INSERTION_SHIFT);
+
+ p->des0 = tdes0;
if (mode == STMMAC_CHAIN_MODE)
enh_set_tx_desc_len_on_chain(p, len);
else
enh_set_tx_desc_len_on_ring(p, len);
-
- if (likely(csum_flag))
- p->des01.etx.checksum_insertion = cic_full;
}
static void enh_desc_clear_tx_ic(struct dma_desc *p)
{
- p->des01.etx.interrupt = 0;
+ p->des0 &= ~ETDES0_INTERRUPT;
}
static void enh_desc_close_tx_desc(struct dma_desc *p)
{
- p->des01.etx.last_segment = 1;
- p->des01.etx.interrupt = 1;
+ p->des0 |= ETDES0_LAST_SEGMENT | ETDES0_INTERRUPT;
}
static int enh_desc_get_rx_frame_len(struct dma_desc *p, int rx_coe_type)
{
+ unsigned int csum = 0;
/* The type-1 checksum offload engines append the checksum at
* the end of frame and the two bytes of checksum are added in
* the length.
* Adjust for that in the framelen for type-1 checksum offload
- * engines. */
+ * engines.
+ */
if (rx_coe_type == STMMAC_RX_COE_TYPE1)
- return p->des01.erx.frame_length - 2;
- else
- return p->des01.erx.frame_length;
+ csum = 2;
+
+ return (((p->des0 & RDES0_FRAME_LEN_MASK) >> RDES0_FRAME_LEN_SHIFT) -
+ csum);
}
static void enh_desc_enable_tx_timestamp(struct dma_desc *p)
{
- p->des01.etx.time_stamp_enable = 1;
+ p->des0 |= ETDES0_TIME_STAMP_ENABLE;
}
static int enh_desc_get_tx_timestamp_status(struct dma_desc *p)
{
- return p->des01.etx.time_stamp_status;
+ return (p->des0 & ETDES0_TIME_STAMP_STATUS) >> 17;
}
static u64 enh_desc_get_timestamp(void *desc, u32 ats)
@@ -368,7 +384,7 @@ static int enh_desc_get_rx_timestamp_status(void *desc, u32 ats)
{
if (ats) {
struct dma_extended_desc *p = (struct dma_extended_desc *)desc;
- return p->basic.des01.erx.ipc_csum_error;
+ return (p->basic.des0 & RDES0_IPC_CSUM_ERROR) >> 7;
} else {
struct dma_desc *p = (struct dma_desc *)desc;
if ((p->des2 == 0xffffffff) && (p->des3 == 0xffffffff))
diff --git a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
index 48c3456..460c573 100644
--- a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
@@ -29,33 +29,38 @@
static int ndesc_get_tx_status(void *data, struct stmmac_extra_stats *x,
struct dma_desc *p, void __iomem *ioaddr)
{
- int ret = 0;
struct net_device_stats *stats = (struct net_device_stats *)data;
+ unsigned int tdes0 = p->des0;
+ int ret = 0;
- if (unlikely(p->des01.tx.error_summary)) {
- if (unlikely(p->des01.tx.underflow_error)) {
+ if (unlikely(tdes0 & TDES0_ERROR_SUMMARY)) {
+ if (unlikely(tdes0 & TDES0_UNDERFLOW_ERROR)) {
x->tx_underflow++;
stats->tx_fifo_errors++;
}
- if (unlikely(p->des01.tx.no_carrier)) {
+ if (unlikely(tdes0 & TDES0_NO_CARRIER)) {
x->tx_carrier++;
stats->tx_carrier_errors++;
}
- if (unlikely(p->des01.tx.loss_carrier)) {
+ if (unlikely(tdes0 & TDES0_LOSS_CARRIER)) {
x->tx_losscarrier++;
stats->tx_carrier_errors++;
}
- if (unlikely((p->des01.tx.excessive_deferral) ||
- (p->des01.tx.excessive_collisions) ||
- (p->des01.tx.late_collision)))
- stats->collisions += p->des01.tx.collision_count;
+ if (unlikely((tdes0 & TDES0_EXCESSIVE_DEFERRAL) ||
+ (tdes0 & TDES0_EXCESSIVE_COLLISIONS) ||
+ (tdes0 & TDES0_LATE_COLLISION))) {
+ unsigned int collisions;
+
+ collisions = (tdes0 & TDES0_COLLISION_COUNT_MASK) >> 3;
+ stats->collisions += collisions;
+ }
ret = -1;
}
- if (p->des01.etx.vlan_frame)
+ if (tdes0 & TDES0_VLAN_FRAME)
x->tx_vlan++;
- if (unlikely(p->des01.tx.deferred))
+ if (unlikely(tdes0 & TDES0_DEFERRED))
x->tx_deferred++;
return ret;
@@ -63,7 +68,7 @@ static int ndesc_get_tx_status(void *data, struct stmmac_extra_stats *x,
static int ndesc_get_tx_len(struct dma_desc *p)
{
- return p->des01.tx.buffer1_size;
+ return (p->des1 & RDES1_BUFFER1_SIZE_MASK);
}
/* This function verifies if each incoming frame has some errors
@@ -74,47 +79,48 @@ static int ndesc_get_rx_status(void *data, struct stmmac_extra_stats *x,
struct dma_desc *p)
{
int ret = good_frame;
+ unsigned int rdes0 = p->des0;
struct net_device_stats *stats = (struct net_device_stats *)data;
- if (unlikely(p->des01.rx.last_descriptor == 0)) {
+ if (unlikely(!(rdes0 & RDES0_LAST_DESCRIPTOR))) {
pr_warn("%s: Oversized frame spanned multiple buffers\n",
__func__);
stats->rx_length_errors++;
return discard_frame;
}
- if (unlikely(p->des01.rx.error_summary)) {
- if (unlikely(p->des01.rx.descriptor_error))
+ if (unlikely(rdes0 & RDES0_ERROR_SUMMARY)) {
+ if (unlikely(rdes0 & RDES0_DESCRIPTOR_ERROR))
x->rx_desc++;
- if (unlikely(p->des01.rx.sa_filter_fail))
+ if (unlikely(rdes0 & RDES0_SA_FILTER_FAIL))
x->sa_filter_fail++;
- if (unlikely(p->des01.rx.overflow_error))
+ if (unlikely(rdes0 & RDES0_OVERFLOW_ERROR))
x->overflow_error++;
- if (unlikely(p->des01.rx.ipc_csum_error))
+ if (unlikely(rdes0 & RDES0_IPC_CSUM_ERROR))
x->ipc_csum_error++;
- if (unlikely(p->des01.rx.collision)) {
+ if (unlikely(rdes0 & RDES0_COLLISION)) {
x->rx_collision++;
stats->collisions++;
}
- if (unlikely(p->des01.rx.crc_error)) {
+ if (unlikely(rdes0 & RDES0_CRC_ERROR)) {
x->rx_crc++;
stats->rx_crc_errors++;
}
ret = discard_frame;
}
- if (unlikely(p->des01.rx.dribbling))
+ if (unlikely(rdes0 & RDES0_DRIBBLING))
x->dribbling_bit++;
- if (unlikely(p->des01.rx.length_error)) {
+ if (unlikely(rdes0 & RDES0_LENGTH_ERROR)) {
x->rx_length++;
ret = discard_frame;
}
- if (unlikely(p->des01.rx.mii_error)) {
+ if (unlikely(rdes0 & RDES0_MII_ERROR)) {
x->rx_mii++;
ret = discard_frame;
}
#ifdef STMMAC_VLAN_TAG_USED
- if (p->des01.rx.vlan_tag)
+ if (rdes0 & RDES0_VLAN_TAG)
x->vlan_tag++;
#endif
return ret;
@@ -123,9 +129,8 @@ static int ndesc_get_rx_status(void *data, struct stmmac_extra_stats *x,
static void ndesc_init_rx_desc(struct dma_desc *p, int disable_rx_ic, int mode,
int end)
{
- p->des01.all_flags = 0;
- p->des01.rx.own = 1;
- p->des01.rx.buffer1_size = BUF_SIZE_2KiB - 1;
+ p->des0 |= RDES0_OWN;
+ p->des1 |= (BUF_SIZE_2KiB - 1) & RDES1_BUFFER1_SIZE_MASK;
if (mode == STMMAC_CHAIN_MODE)
ndesc_rx_set_on_chain(p, end);
@@ -133,50 +138,50 @@ static void ndesc_init_rx_desc(struct dma_desc *p, int disable_rx_ic, int mode,
ndesc_rx_set_on_ring(p, end);
if (disable_rx_ic)
- p->des01.rx.disable_ic = 1;
+ p->des1 |= RDES1_DISABLE_IC;
}
static void ndesc_init_tx_desc(struct dma_desc *p, int mode, int end)
{
- p->des01.all_flags = 0;
+ p->des0 &= ~TDES0_OWN;
if (mode == STMMAC_CHAIN_MODE)
- ndesc_tx_set_on_chain(p, end);
+ ndesc_tx_set_on_chain(p);
else
- ndesc_tx_set_on_ring(p, end);
+ ndesc_end_tx_desc_on_ring(p, end);
}
static int ndesc_get_tx_owner(struct dma_desc *p)
{
- return p->des01.tx.own;
+ return (p->des0 & TDES0_OWN) >> 31;
}
static int ndesc_get_rx_owner(struct dma_desc *p)
{
- return p->des01.rx.own;
+ return (p->des0 & RDES0_OWN) >> 31;
}
static void ndesc_set_tx_owner(struct dma_desc *p)
{
- p->des01.tx.own = 1;
+ p->des0 |= TDES0_OWN;
}
static void ndesc_set_rx_owner(struct dma_desc *p)
{
- p->des01.rx.own = 1;
+ p->des0 |= RDES0_OWN;
}
static int ndesc_get_tx_ls(struct dma_desc *p)
{
- return p->des01.tx.last_segment;
+ return (p->des1 & TDES1_LAST_SEGMENT) >> 30;
}
static void ndesc_release_tx_desc(struct dma_desc *p, int mode)
{
- int ter = p->des01.tx.end_ring;
+ int ter = (p->des1 & TDES1_END_RING) >> 25;
memset(p, 0, offsetof(struct dma_desc, des2));
if (mode == STMMAC_CHAIN_MODE)
- ndesc_end_tx_desc_on_chain(p, ter);
+ ndesc_tx_set_on_chain(p);
else
ndesc_end_tx_desc_on_ring(p, ter);
}
@@ -184,48 +189,62 @@ static void ndesc_release_tx_desc(struct dma_desc *p, int mode)
static void ndesc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
int csum_flag, int mode)
{
- p->des01.tx.first_segment = is_fs;
+ unsigned int tdes1 = p->des1;
+
+ if (is_fs)
+ tdes1 |= TDES1_FIRST_SEGMENT;
+ else
+ tdes1 &= ~TDES1_FIRST_SEGMENT;
+
+ if (likely(csum_flag))
+ tdes1 |= (TX_CIC_FULL) << TDES1_CHECKSUM_INSERTION_SHIFT;
+ else
+ tdes1 &= ~(TX_CIC_FULL << TDES1_CHECKSUM_INSERTION_SHIFT);
+
+ p->des1 = tdes1;
+
if (mode == STMMAC_CHAIN_MODE)
norm_set_tx_desc_len_on_chain(p, len);
else
norm_set_tx_desc_len_on_ring(p, len);
-
- if (likely(csum_flag))
- p->des01.tx.checksum_insertion = cic_full;
}
static void ndesc_clear_tx_ic(struct dma_desc *p)
{
- p->des01.tx.interrupt = 0;
+ p->des1 &= ~TDES1_INTERRUPT;
}
static void ndesc_close_tx_desc(struct dma_desc *p)
{
- p->des01.tx.last_segment = 1;
- p->des01.tx.interrupt = 1;
+ p->des1 |= TDES1_LAST_SEGMENT | TDES1_INTERRUPT;
}
static int ndesc_get_rx_frame_len(struct dma_desc *p, int rx_coe_type)
{
+ unsigned int csum = 0;
+
/* The type-1 checksum offload engines append the checksum at
* the end of frame and the two bytes of checksum are added in
* the length.
* Adjust for that in the framelen for type-1 checksum offload
- * engines. */
+ * engines
+ */
if (rx_coe_type == STMMAC_RX_COE_TYPE1)
- return p->des01.rx.frame_length - 2;
- else
- return p->des01.rx.frame_length;
+ csum = 2;
+
+ return (((p->des0 & RDES0_FRAME_LEN_MASK) >> RDES0_FRAME_LEN_SHIFT) -
+ csum);
+
}
static void ndesc_enable_tx_timestamp(struct dma_desc *p)
{
- p->des01.tx.time_stamp_enable = 1;
+ p->des1 |= TDES1_TIME_STAMP_ENABLE;
}
static int ndesc_get_tx_timestamp_status(struct dma_desc *p)
{
- return p->des01.tx.time_stamp_status;
+ return (p->des0 & TDES0_TIME_STAMP_STATUS) >> 17;
}
static u64 ndesc_get_timestamp(void *desc, u32 ats)
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 04/18] stmmac: remove modulo in stmmac_xmit()
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (2 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 03/18] stmmac: change descriptor layout Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 05/18] stmmac: add length field to dma data Giuseppe Cavallaro
` (13 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro, David McKay
The indexes into the ring buffer are always incremented, and
the entry is accessed via doing a modulo to find the "real" index.
I have changed this for 3 reasons:
* It is inefficient, modulo is an expensive operation.
* It is also broken when the counter wraps.
* It makes my head hurt trying to cope with huge numbers
This patch replaces the modulo with a simple if clamp.
Signed-off-by: David McKay <david.mckay@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/chain_mode.c | 15 ++++++----
drivers/net/ethernet/stmicro/stmmac/ring_mode.c | 7 +++-
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 31 +++++++++++++++-----
3 files changed, 37 insertions(+), 16 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
index cf28dab..399a7ed 100644
--- a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
@@ -32,7 +32,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
{
struct stmmac_priv *priv = (struct stmmac_priv *)p;
unsigned int txsize = priv->dma_tx_size;
- unsigned int entry = priv->cur_tx % txsize;
+ unsigned int entry = priv->cur_tx;
struct dma_desc *desc = priv->dma_tx + entry;
unsigned int nopaged_len = skb_headlen(skb);
unsigned int bmax;
@@ -54,7 +54,8 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
while (len != 0) {
priv->tx_skbuff[entry] = NULL;
- entry = (++priv->cur_tx) % txsize;
+ if (++entry >= txsize)
+ entry = 0;
desc = priv->dma_tx + entry;
if (len > bmax) {
@@ -82,6 +83,9 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
len = 0;
}
}
+
+ priv->cur_tx = entry;
+
return entry;
}
@@ -151,10 +155,9 @@ static void stmmac_clean_desc3(void *priv_ptr, struct dma_desc *p)
* 1588-2002 time stamping is enabled, hence reinitialize it
* to keep explicit chaining in the descriptor.
*/
- p->des3 = (unsigned int)(priv->dma_tx_phy +
- (((priv->dirty_tx + 1) %
- priv->dma_tx_size) *
- sizeof(struct dma_desc)));
+ p->des3 = (unsigned int)((priv->dma_tx_phy +
+ priv->dirty_tx + 1) *
+ sizeof(struct dma_desc));
}
const struct stmmac_mode_ops chain_mode_ops = {
diff --git a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
index 5dd50c6..2560855 100644
--- a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
@@ -32,7 +32,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
{
struct stmmac_priv *priv = (struct stmmac_priv *)p;
unsigned int txsize = priv->dma_tx_size;
- unsigned int entry = priv->cur_tx % txsize;
+ unsigned int entry = priv->cur_tx;
struct dma_desc *desc;
unsigned int nopaged_len = skb_headlen(skb);
unsigned int bmax, len;
@@ -62,7 +62,8 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
STMMAC_RING_MODE);
wmb();
priv->tx_skbuff[entry] = NULL;
- entry = (++priv->cur_tx) % txsize;
+ if (++entry >= txsize)
+ entry = 0;
if (priv->extend_desc)
desc = (struct dma_desc *)(priv->dma_etx + entry);
@@ -90,6 +91,8 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
STMMAC_RING_MODE);
}
+ priv->cur_tx = entry;
+
return entry;
}
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 2dd0279..366da1f 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -201,7 +201,14 @@ static void print_pkt(unsigned char *buf, int len)
static inline u32 stmmac_tx_avail(struct stmmac_priv *priv)
{
- return priv->dirty_tx + priv->dma_tx_size - priv->cur_tx - 1;
+ unsigned avail;
+
+ if (priv->dirty_tx > priv->cur_tx)
+ avail = priv->dirty_tx - priv->cur_tx - 1;
+ else
+ avail = priv->dma_tx_size - priv->cur_tx + priv->dirty_tx - 1;
+
+ return avail;
}
/**
@@ -1315,16 +1322,16 @@ static void stmmac_dma_operation_mode(struct stmmac_priv *priv)
*/
static void stmmac_tx_clean(struct stmmac_priv *priv)
{
- unsigned int txsize = priv->dma_tx_size;
unsigned int bytes_compl = 0, pkts_compl = 0;
+ unsigned int txsize = priv->dma_tx_size;
+ unsigned int entry = priv->dirty_tx;
spin_lock(&priv->tx_lock);
priv->xstats.tx_clean++;
- while (priv->dirty_tx != priv->cur_tx) {
+ while (entry != priv->cur_tx) {
int last;
- unsigned int entry = priv->dirty_tx % txsize;
struct sk_buff *skb = priv->tx_skbuff[entry];
struct dma_desc *p;
@@ -1381,7 +1388,9 @@ static void stmmac_tx_clean(struct stmmac_priv *priv)
priv->hw->desc->release_tx_desc(p, priv->mode);
- priv->dirty_tx++;
+ if (++entry >= txsize)
+ entry = 0;
+ priv->dirty_tx = entry;
}
netdev_completed_queue(priv->dev, pkts_compl, bytes_compl);
@@ -1981,7 +1990,8 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
if (priv->tx_path_in_lpi_mode)
stmmac_disable_eee_mode(priv);
- entry = priv->cur_tx % txsize;
+ entry = priv->cur_tx;
+
csum_insertion = (skb->ip_summed == CHECKSUM_PARTIAL);
@@ -2016,7 +2026,9 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
int len = skb_frag_size(frag);
priv->tx_skbuff[entry] = NULL;
- entry = (++priv->cur_tx) % txsize;
+ if (++entry >= txsize)
+ entry = 0;
+
if (priv->extend_desc)
desc = (struct dma_desc *)(priv->dma_etx + entry);
else
@@ -2059,7 +2071,10 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
priv->hw->desc->set_tx_owner(first);
wmb();
- priv->cur_tx++;
+ if (++entry >= txsize)
+ entry = 0;
+
+ priv->cur_tx = entry;
if (netif_msg_pktdata(priv)) {
pr_debug("%s: curr %d dirty=%d entry=%d, first=%p, nfrags=%d",
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 05/18] stmmac: add length field to dma data
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (3 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 04/18] stmmac: remove modulo in stmmac_xmit() Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 06/18] stmmac: add last_segment " Giuseppe Cavallaro
` (12 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro, Fabrice Gasnier
Currently, the code pulls out the length field when
unmapping a buffer directly from the descriptor. This will result
in an uncached read to a dma_alloc_coherent() region. There is no
need to do this, so this patch simply puts the value directly into
a data structure which will hit the cache.
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/chain_mode.c | 3 +++
drivers/net/ethernet/stmicro/stmmac/ring_mode.c | 5 +++++
drivers/net/ethernet/stmicro/stmmac/stmmac.h | 1 +
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 11 +++++++----
4 files changed, 16 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
index 399a7ed..688795e 100644
--- a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
@@ -50,6 +50,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
if (dma_mapping_error(priv->device, desc->des2))
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
+ priv->tx_skbuff_dma[entry].len = bmax;
priv->hw->desc->prepare_tx_desc(desc, 1, bmax, csum, STMMAC_CHAIN_MODE);
while (len != 0) {
@@ -65,6 +66,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
if (dma_mapping_error(priv->device, desc->des2))
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
+ priv->tx_skbuff_dma[entry].len = bmax;
priv->hw->desc->prepare_tx_desc(desc, 0, bmax, csum,
STMMAC_CHAIN_MODE);
priv->hw->desc->set_tx_owner(desc);
@@ -77,6 +79,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
if (dma_mapping_error(priv->device, desc->des2))
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
+ priv->tx_skbuff_dma[entry].len = len;
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum,
STMMAC_CHAIN_MODE);
priv->hw->desc->set_tx_owner(desc);
diff --git a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
index 2560855..3300029 100644
--- a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
@@ -57,6 +57,8 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
+ priv->tx_skbuff_dma[entry].len = bmax;
+
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 1, bmax, csum,
STMMAC_RING_MODE);
@@ -75,6 +77,8 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
if (dma_mapping_error(priv->device, desc->des2))
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
+ priv->tx_skbuff_dma[entry].len = len;
+
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum,
STMMAC_RING_MODE);
@@ -86,6 +90,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
if (dma_mapping_error(priv->device, desc->des2))
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
+ priv->tx_skbuff_dma[entry].len = nopaged_len;
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 1, nopaged_len, csum,
STMMAC_RING_MODE);
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
index 1f3b33a..8517a04 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
@@ -45,6 +45,7 @@ struct stmmac_resources {
struct stmmac_tx_info {
dma_addr_t buf;
bool map_as_page;
+ unsigned len;
};
struct stmmac_priv {
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 366da1f..f9baf5d 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1110,6 +1110,7 @@ static int init_dma_desc_rings(struct net_device *dev, gfp_t flags)
p->des2 = 0;
priv->tx_skbuff_dma[i].buf = 0;
priv->tx_skbuff_dma[i].map_as_page = false;
+ priv->tx_skbuff_dma[i].len = 0;
priv->tx_skbuff[i] = NULL;
}
@@ -1153,12 +1154,12 @@ static void dma_free_tx_skbufs(struct stmmac_priv *priv)
if (priv->tx_skbuff_dma[i].map_as_page)
dma_unmap_page(priv->device,
priv->tx_skbuff_dma[i].buf,
- priv->hw->desc->get_tx_len(p),
+ priv->tx_skbuff_dma[i].len,
DMA_TO_DEVICE);
else
dma_unmap_single(priv->device,
priv->tx_skbuff_dma[i].buf,
- priv->hw->desc->get_tx_len(p),
+ priv->tx_skbuff_dma[i].len,
DMA_TO_DEVICE);
}
@@ -1367,12 +1368,12 @@ static void stmmac_tx_clean(struct stmmac_priv *priv)
if (priv->tx_skbuff_dma[entry].map_as_page)
dma_unmap_page(priv->device,
priv->tx_skbuff_dma[entry].buf,
- priv->hw->desc->get_tx_len(p),
+ priv->tx_skbuff_dma[entry].len,
DMA_TO_DEVICE);
else
dma_unmap_single(priv->device,
priv->tx_skbuff_dma[entry].buf,
- priv->hw->desc->get_tx_len(p),
+ priv->tx_skbuff_dma[entry].len,
DMA_TO_DEVICE);
priv->tx_skbuff_dma[entry].buf = 0;
priv->tx_skbuff_dma[entry].map_as_page = false;
@@ -2012,6 +2013,7 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
if (dma_mapping_error(priv->device, desc->des2))
goto dma_map_err;
priv->tx_skbuff_dma[entry].buf = desc->des2;
+ priv->tx_skbuff_dma[entry].len = nopaged_len;
priv->hw->desc->prepare_tx_desc(desc, 1, nopaged_len,
csum_insertion, priv->mode);
} else {
@@ -2041,6 +2043,7 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].map_as_page = true;
+ priv->tx_skbuff_dma[entry].len = len;
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum_insertion,
priv->mode);
wmb();
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 06/18] stmmac: add last_segment field to dma data
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (4 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 05/18] stmmac: add length field to dma data Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 07/18] stmmac: add is_jumbo " Giuseppe Cavallaro
` (11 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro, Fabrice Gasnier
last_segment field is read twice from dma descriptors in stmmac_clean().
Add last_segment to dma data so that this flag is from priv
structure in cache instead of memory.
It avoids reading twice from memory for each loop in stmmac_clean().
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/chain_mode.c | 3 ++-
drivers/net/ethernet/stmicro/stmmac/stmmac.h | 1 +
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 5 ++++-
3 files changed, 7 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
index 688795e..e985459 100644
--- a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
@@ -152,8 +152,9 @@ static void stmmac_refill_desc3(void *priv_ptr, struct dma_desc *p)
static void stmmac_clean_desc3(void *priv_ptr, struct dma_desc *p)
{
struct stmmac_priv *priv = (struct stmmac_priv *)priv_ptr;
+ unsigned int entry = priv->dirty_tx;
- if (priv->hw->desc->get_tx_ls(p) && !priv->extend_desc)
+ if (priv->tx_skbuff_dma[entry].last_segment && !priv->extend_desc)
/* NOTE: Device will overwrite des3 with timestamp value if
* 1588-2002 time stamping is enabled, hence reinitialize it
* to keep explicit chaining in the descriptor.
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
index 8517a04..f15c493 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
@@ -46,6 +46,7 @@ struct stmmac_tx_info {
dma_addr_t buf;
bool map_as_page;
unsigned len;
+ bool last_segment;
};
struct stmmac_priv {
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index f9baf5d..3136c0b 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1111,6 +1111,7 @@ static int init_dma_desc_rings(struct net_device *dev, gfp_t flags)
priv->tx_skbuff_dma[i].buf = 0;
priv->tx_skbuff_dma[i].map_as_page = false;
priv->tx_skbuff_dma[i].len = 0;
+ priv->tx_skbuff_dma[i].last_segment = false;
priv->tx_skbuff[i] = NULL;
}
@@ -1346,7 +1347,7 @@ static void stmmac_tx_clean(struct stmmac_priv *priv)
break;
/* Verify tx error by looking at the last segment. */
- last = priv->hw->desc->get_tx_ls(p);
+ last = priv->tx_skbuff_dma[entry].last_segment;
if (likely(last)) {
int tx_error =
priv->hw->desc->tx_status(&priv->dev->stats,
@@ -1379,6 +1380,7 @@ static void stmmac_tx_clean(struct stmmac_priv *priv)
priv->tx_skbuff_dma[entry].map_as_page = false;
}
priv->hw->mode->clean_desc3(priv, p);
+ priv->tx_skbuff_dma[entry].last_segment = false;
if (likely(skb != NULL)) {
pkts_compl++;
@@ -2055,6 +2057,7 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
/* Finalize the latest segment. */
priv->hw->desc->close_tx_desc(desc);
+ priv->tx_skbuff_dma[entry].last_segment = true;
wmb();
/* According to the coalesce parameter the IC bit for the latest
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 07/18] stmmac: add is_jumbo field to dma data
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (5 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 06/18] stmmac: add last_segment " Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 08/18] stmmac: merge get_rx_owner into rx_status routine Giuseppe Cavallaro
` (10 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro, Fabrice Gasnier
Optimize tx_clean by avoiding a des3 read in stmmac_clean_desc3().
In ring mode, TX, des3 seems only used when xmit a jumbo frame.
In case of normal descriptors, it may also be used for time
stamping.
Clean it in the above two case, without reading it.
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/chain_mode.c | 3 ++-
drivers/net/ethernet/stmicro/stmmac/ring_mode.c | 11 ++++++++++-
drivers/net/ethernet/stmicro/stmmac/stmmac.h | 1 +
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 1 +
4 files changed, 14 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
index e985459..d6d56ee 100644
--- a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
@@ -154,7 +154,8 @@ static void stmmac_clean_desc3(void *priv_ptr, struct dma_desc *p)
struct stmmac_priv *priv = (struct stmmac_priv *)priv_ptr;
unsigned int entry = priv->dirty_tx;
- if (priv->tx_skbuff_dma[entry].last_segment && !priv->extend_desc)
+ if (priv->tx_skbuff_dma[entry].last_segment && !priv->extend_desc &&
+ priv->hwts_tx_en)
/* NOTE: Device will overwrite des3 with timestamp value if
* 1588-2002 time stamping is enabled, hence reinitialize it
* to keep explicit chaining in the descriptor.
diff --git a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
index 3300029..ffab4d7 100644
--- a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
@@ -58,6 +58,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].len = bmax;
+ priv->tx_skbuff_dma[entry].is_jumbo = true;
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 1, bmax, csum,
@@ -78,6 +79,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].len = len;
+ priv->tx_skbuff_dma[entry].is_jumbo = true;
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum,
@@ -91,6 +93,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].len = nopaged_len;
+ priv->tx_skbuff_dma[entry].is_jumbo = true;
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 1, nopaged_len, csum,
STMMAC_RING_MODE);
@@ -128,7 +131,13 @@ static void stmmac_init_desc3(struct dma_desc *p)
static void stmmac_clean_desc3(void *priv_ptr, struct dma_desc *p)
{
- if (unlikely(p->des3))
+ struct stmmac_priv *priv = (struct stmmac_priv *)priv_ptr;
+ unsigned int entry = priv->dirty_tx;
+
+ /* des3 is only used for jumbo frames tx or time stamping */
+ if (unlikely(priv->tx_skbuff_dma[entry].is_jumbo ||
+ (priv->tx_skbuff_dma[entry].last_segment &&
+ !priv->extend_desc && priv->hwts_tx_en)))
p->des3 = 0;
}
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
index f15c493..b73e8d8 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
@@ -47,6 +47,7 @@ struct stmmac_tx_info {
bool map_as_page;
unsigned len;
bool last_segment;
+ bool is_jumbo;
};
struct stmmac_priv {
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 3136c0b..073a929 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1381,6 +1381,7 @@ static void stmmac_tx_clean(struct stmmac_priv *priv)
}
priv->hw->mode->clean_desc3(priv, p);
priv->tx_skbuff_dma[entry].last_segment = false;
+ priv->tx_skbuff_dma[entry].is_jumbo = false;
if (likely(skb != NULL)) {
pkts_compl++;
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 08/18] stmmac: merge get_rx_owner into rx_status routine.
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (6 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 07/18] stmmac: add is_jumbo " Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 09/18] stmmac: optimize tx desc management Giuseppe Cavallaro
` (9 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Fabrice Gasnier
From: Fabrice Gasnier <fabrice.gasnier@st.com>
The RDES0 register can be read several times while doing RX of a
packet.
This patch slightly improves RX path performance by reading rdes0
once for two operation: check rx owner, get rx status bits.
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/common.h | 10 +++++-----
drivers/net/ethernet/stmicro/stmmac/enh_desc.c | 9 +++------
drivers/net/ethernet/stmicro/stmmac/norm_desc.c | 9 +++------
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 9 +++++----
4 files changed, 16 insertions(+), 21 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/common.h b/drivers/net/ethernet/stmicro/stmmac/common.h
index cdc3806..26ca46e 100644
--- a/drivers/net/ethernet/stmicro/stmmac/common.h
+++ b/drivers/net/ethernet/stmicro/stmmac/common.h
@@ -234,10 +234,11 @@ struct stmmac_extra_stats {
/* Rx IPC status */
enum rx_frame_status {
- good_frame = 0,
- discard_frame = 1,
- csum_none = 2,
- llc_snap = 4,
+ good_frame = 0x0,
+ discard_frame = 0x1,
+ csum_none = 0x2,
+ llc_snap = 0x4,
+ dma_own = 0x8,
};
enum dma_irq_status {
@@ -352,7 +353,6 @@ struct stmmac_desc_ops {
/* Get the buffer size from the descriptor */
int (*get_tx_len) (struct dma_desc *p);
/* Handle extra events on specific interrupts hw dependent */
- int (*get_rx_owner) (struct dma_desc *p);
void (*set_rx_owner) (struct dma_desc *p);
/* Get the receive frame size */
int (*get_rx_frame_len) (struct dma_desc *p, int rx_coe_type);
diff --git a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
index 716b807..1a2fce9 100644
--- a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
@@ -186,6 +186,9 @@ static int enh_desc_get_rx_status(void *data, struct stmmac_extra_stats *x,
unsigned int rdes0 = p->des0;
int ret = good_frame;
+ if (unlikely(rdes0 & RDES0_OWN))
+ return dma_own;
+
if (unlikely(rdes0 & RDES0_ERROR_SUMMARY)) {
if (unlikely(rdes0 & RDES0_DESCRIPTOR_ERROR)) {
x->rx_desc++;
@@ -272,11 +275,6 @@ static int enh_desc_get_tx_owner(struct dma_desc *p)
return (p->des0 & ETDES0_OWN) >> 31;
}
-static int enh_desc_get_rx_owner(struct dma_desc *p)
-{
- return (p->des0 & RDES0_OWN) >> 31;
-}
-
static void enh_desc_set_tx_owner(struct dma_desc *p)
{
p->des0 |= ETDES0_OWN;
@@ -402,7 +400,6 @@ const struct stmmac_desc_ops enh_desc_ops = {
.init_rx_desc = enh_desc_init_rx_desc,
.init_tx_desc = enh_desc_init_tx_desc,
.get_tx_owner = enh_desc_get_tx_owner,
- .get_rx_owner = enh_desc_get_rx_owner,
.release_tx_desc = enh_desc_release_tx_desc,
.prepare_tx_desc = enh_desc_prepare_tx_desc,
.clear_tx_ic = enh_desc_clear_tx_ic,
diff --git a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
index 460c573..5a91932 100644
--- a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
@@ -82,6 +82,9 @@ static int ndesc_get_rx_status(void *data, struct stmmac_extra_stats *x,
unsigned int rdes0 = p->des0;
struct net_device_stats *stats = (struct net_device_stats *)data;
+ if (unlikely(rdes0 & RDES0_OWN))
+ return dma_own;
+
if (unlikely(!(rdes0 & RDES0_LAST_DESCRIPTOR))) {
pr_warn("%s: Oversized frame spanned multiple buffers\n",
__func__);
@@ -155,11 +158,6 @@ static int ndesc_get_tx_owner(struct dma_desc *p)
return (p->des0 & TDES0_OWN) >> 31;
}
-static int ndesc_get_rx_owner(struct dma_desc *p)
-{
- return (p->des0 & RDES0_OWN) >> 31;
-}
-
static void ndesc_set_tx_owner(struct dma_desc *p)
{
p->des0 |= TDES0_OWN;
@@ -277,7 +275,6 @@ const struct stmmac_desc_ops ndesc_ops = {
.init_rx_desc = ndesc_init_rx_desc,
.init_tx_desc = ndesc_init_tx_desc,
.get_tx_owner = ndesc_get_tx_owner,
- .get_rx_owner = ndesc_get_rx_owner,
.release_tx_desc = ndesc_release_tx_desc,
.prepare_tx_desc = ndesc_prepare_tx_desc,
.clear_tx_ic = ndesc_clear_tx_ic,
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 073a929..19382a1 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -2227,7 +2227,11 @@ static int stmmac_rx(struct stmmac_priv *priv, int limit)
else
p = priv->dma_rx + entry;
- if (priv->hw->desc->get_rx_owner(p))
+ /* read the status of the incoming frame */
+ status = priv->hw->desc->rx_status(&priv->dev->stats,
+ &priv->xstats, p);
+ /* check if managed by the DMA otherwise go ahead */
+ if (unlikely(status & dma_own))
break;
count++;
@@ -2238,9 +2242,6 @@ static int stmmac_rx(struct stmmac_priv *priv, int limit)
else
prefetch(priv->dma_rx + next_entry);
- /* read the status of the incoming frame */
- status = priv->hw->desc->rx_status(&priv->dev->stats,
- &priv->xstats, p);
if ((priv->extend_desc) && (priv->hw->desc->rx_extended_status))
priv->hw->desc->rx_extended_status(&priv->dev->stats,
&priv->xstats,
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 09/18] stmmac: optimize tx desc management
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (7 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 08/18] stmmac: merge get_rx_owner into rx_status routine Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 10/18] stmmac: optimize tx clean function Giuseppe Cavallaro
` (8 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro, Fabrice Gasnier
This patch is to optimize the way to manage the TDES inside the
xmit function. When prepare the frame, some settings (e.g. OWN
bit) can be merged. This has been reworked to improve the tx
performances.
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/chain_mode.c | 13 +++++++----
drivers/net/ethernet/stmicro/stmmac/common.h | 5 +--
drivers/net/ethernet/stmicro/stmmac/enh_desc.c | 22 ++++++++++++++------
drivers/net/ethernet/stmicro/stmmac/norm_desc.c | 15 +++++++------
drivers/net/ethernet/stmicro/stmmac/ring_mode.c | 7 ++---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 16 +++++---------
6 files changed, 42 insertions(+), 36 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
index d6d56ee..c9310ef 100644
--- a/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/chain_mode.c
@@ -51,7 +51,9 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].len = bmax;
- priv->hw->desc->prepare_tx_desc(desc, 1, bmax, csum, STMMAC_CHAIN_MODE);
+ /* do not close the descriptor and do not set own bit */
+ priv->hw->desc->prepare_tx_desc(desc, 1, bmax, csum, STMMAC_CHAIN_MODE,
+ 0, false);
while (len != 0) {
priv->tx_skbuff[entry] = NULL;
@@ -68,8 +70,8 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].len = bmax;
priv->hw->desc->prepare_tx_desc(desc, 0, bmax, csum,
- STMMAC_CHAIN_MODE);
- priv->hw->desc->set_tx_owner(desc);
+ STMMAC_CHAIN_MODE, 1,
+ false);
len -= bmax;
i++;
} else {
@@ -80,9 +82,10 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
return -1;
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].len = len;
+ /* last descriptor can be set now */
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum,
- STMMAC_CHAIN_MODE);
- priv->hw->desc->set_tx_owner(desc);
+ STMMAC_CHAIN_MODE, 1,
+ true);
len = 0;
}
}
diff --git a/drivers/net/ethernet/stmicro/stmmac/common.h b/drivers/net/ethernet/stmicro/stmmac/common.h
index 26ca46e..e58d81c 100644
--- a/drivers/net/ethernet/stmicro/stmmac/common.h
+++ b/drivers/net/ethernet/stmicro/stmmac/common.h
@@ -334,12 +334,11 @@ struct stmmac_desc_ops {
/* Invoked by the xmit function to prepare the tx descriptor */
void (*prepare_tx_desc) (struct dma_desc *p, int is_fs, int len,
- int csum_flag, int mode);
+ int csum_flag, int mode, int tx_own,
+ bool ls_ic);
/* Set/get the owner of the descriptor */
void (*set_tx_owner) (struct dma_desc *p);
int (*get_tx_owner) (struct dma_desc *p);
- /* Invoked by the xmit function to close the tx descriptor */
- void (*close_tx_desc) (struct dma_desc *p);
/* Clean the tx descriptor as soon as the tx irq is received */
void (*release_tx_desc) (struct dma_desc *p, int mode);
/* Clear interrupt on tx frame completion. When this bit is
diff --git a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
index 1a2fce9..1c9a520 100644
--- a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
@@ -302,7 +302,8 @@ static void enh_desc_release_tx_desc(struct dma_desc *p, int mode)
}
static void enh_desc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
- int csum_flag, int mode)
+ int csum_flag, int mode, int tx_own,
+ bool ls_ic)
{
unsigned int tdes0 = p->des0;
@@ -316,6 +317,19 @@ static void enh_desc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
else
tdes0 &= ~(TX_CIC_FULL << ETDES0_CHECKSUM_INSERTION_SHIFT);
+ if (tx_own)
+ tdes0 |= ETDES0_OWN;
+
+ if (is_fs & tx_own)
+ /* When the own bit, for the first frame, has to be set, all
+ * descriptors for the same frame has to be set before, to
+ * avoid race condition.
+ */
+ wmb();
+
+ if (ls_ic)
+ tdes0 |= ETDES0_LAST_SEGMENT | ETDES0_INTERRUPT;
+
p->des0 = tdes0;
if (mode == STMMAC_CHAIN_MODE)
@@ -329,11 +343,6 @@ static void enh_desc_clear_tx_ic(struct dma_desc *p)
p->des0 &= ~ETDES0_INTERRUPT;
}
-static void enh_desc_close_tx_desc(struct dma_desc *p)
-{
- p->des0 |= ETDES0_LAST_SEGMENT | ETDES0_INTERRUPT;
-}
-
static int enh_desc_get_rx_frame_len(struct dma_desc *p, int rx_coe_type)
{
unsigned int csum = 0;
@@ -403,7 +412,6 @@ const struct stmmac_desc_ops enh_desc_ops = {
.release_tx_desc = enh_desc_release_tx_desc,
.prepare_tx_desc = enh_desc_prepare_tx_desc,
.clear_tx_ic = enh_desc_clear_tx_ic,
- .close_tx_desc = enh_desc_close_tx_desc,
.get_tx_ls = enh_desc_get_tx_ls,
.set_tx_owner = enh_desc_set_tx_owner,
.set_rx_owner = enh_desc_set_rx_owner,
diff --git a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
index 5a91932..df2603e 100644
--- a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
@@ -185,7 +185,8 @@ static void ndesc_release_tx_desc(struct dma_desc *p, int mode)
}
static void ndesc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
- int csum_flag, int mode)
+ int csum_flag, int mode, int tx_own,
+ bool ls_ic)
{
unsigned int tdes1 = p->des1;
@@ -199,6 +200,12 @@ static void ndesc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
else
tdes1 &= ~(TX_CIC_FULL << TDES1_CHECKSUM_INSERTION_SHIFT);
+ if (tx_own)
+ tdes1 |= TDES0_OWN;
+
+ if (ls_ic)
+ tdes1 |= TDES1_LAST_SEGMENT | TDES1_INTERRUPT;
+
p->des1 = tdes1;
if (mode == STMMAC_CHAIN_MODE)
@@ -212,11 +219,6 @@ static void ndesc_clear_tx_ic(struct dma_desc *p)
p->des1 &= ~TDES1_INTERRUPT;
}
-static void ndesc_close_tx_desc(struct dma_desc *p)
-{
- p->des1 |= TDES1_LAST_SEGMENT | TDES1_INTERRUPT;
-}
-
static int ndesc_get_rx_frame_len(struct dma_desc *p, int rx_coe_type)
{
unsigned int csum = 0;
@@ -278,7 +280,6 @@ const struct stmmac_desc_ops ndesc_ops = {
.release_tx_desc = ndesc_release_tx_desc,
.prepare_tx_desc = ndesc_prepare_tx_desc,
.clear_tx_ic = ndesc_clear_tx_ic,
- .close_tx_desc = ndesc_close_tx_desc,
.get_tx_ls = ndesc_get_tx_ls,
.set_tx_owner = ndesc_set_tx_owner,
.set_rx_owner = ndesc_set_rx_owner,
diff --git a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
index ffab4d7..4978a6c 100644
--- a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
@@ -62,7 +62,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 1, bmax, csum,
- STMMAC_RING_MODE);
+ STMMAC_RING_MODE, 0, false);
wmb();
priv->tx_skbuff[entry] = NULL;
if (++entry >= txsize)
@@ -83,9 +83,8 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum,
- STMMAC_RING_MODE);
+ STMMAC_RING_MODE, 1, true);
wmb();
- priv->hw->desc->set_tx_owner(desc);
} else {
desc->des2 = dma_map_single(priv->device, skb->data,
nopaged_len, DMA_TO_DEVICE);
@@ -96,7 +95,7 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
priv->tx_skbuff_dma[entry].is_jumbo = true;
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 1, nopaged_len, csum,
- STMMAC_RING_MODE);
+ STMMAC_RING_MODE, 0, true);
}
priv->cur_tx = entry;
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 19382a1..2f2f607 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -2017,8 +2017,10 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
goto dma_map_err;
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].len = nopaged_len;
+ /* do not set the own at this stage */
priv->hw->desc->prepare_tx_desc(desc, 1, nopaged_len,
- csum_insertion, priv->mode);
+ csum_insertion, priv->mode, 0,
+ nfrags == 0);
} else {
desc = first;
entry = priv->hw->mode->jumbo_frm(priv, skb, csum_insertion);
@@ -2029,6 +2031,7 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
for (i = 0; i < nfrags; i++) {
const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
int len = skb_frag_size(frag);
+ bool last_segment = (i == (nfrags - 1));
priv->tx_skbuff[entry] = NULL;
if (++entry >= txsize)
@@ -2048,19 +2051,12 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
priv->tx_skbuff_dma[entry].map_as_page = true;
priv->tx_skbuff_dma[entry].len = len;
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum_insertion,
- priv->mode);
- wmb();
- priv->hw->desc->set_tx_owner(desc);
- wmb();
+ priv->mode, 1, last_segment);
+ priv->tx_skbuff_dma[entry].last_segment = last_segment;
}
priv->tx_skbuff[entry] = skb;
- /* Finalize the latest segment. */
- priv->hw->desc->close_tx_desc(desc);
- priv->tx_skbuff_dma[entry].last_segment = true;
-
- wmb();
/* According to the coalesce parameter the IC bit for the latest
* segment could be reset and the timer re-started to invoke the
* stmmac_tx function. This approach takes care about the fragments.
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 10/18] stmmac: optimize tx clean function
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (8 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 09/18] stmmac: optimize tx desc management Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 11/18] stmmac: set dirty index out of the loop Giuseppe Cavallaro
` (7 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Fabrice Gasnier
From: Fabrice Gasnier <fabrice.gasnier@st.com>
This patch "inline" get_tx_owner and get_ls routines. It Results in a
unique read to tdes0, instead of three, to check TX_OWN and LS bits,
and other status bits.
It helps improve driver TX path by removing two uncached read/writes
inside TX clean loop for enhanced descriptors but not for normal ones
because the des1 must be read in any case.
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/common.h | 8 ++++++
drivers/net/ethernet/stmicro/stmmac/enh_desc.c | 12 +++++++-
drivers/net/ethernet/stmicro/stmmac/norm_desc.c | 13 ++++++++-
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 27 ++++++++++-----------
4 files changed, 42 insertions(+), 18 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/common.h b/drivers/net/ethernet/stmicro/stmmac/common.h
index e58d81c..fb01e94 100644
--- a/drivers/net/ethernet/stmicro/stmmac/common.h
+++ b/drivers/net/ethernet/stmicro/stmmac/common.h
@@ -241,6 +241,14 @@ enum rx_frame_status {
dma_own = 0x8,
};
+/* Tx status */
+enum tx_frame_status {
+ tx_done = 0x0,
+ tx_not_ls = 0x1,
+ tx_err = 0x2,
+ tx_dma_own = 0x4,
+};
+
enum dma_irq_status {
tx_hard_error = 0x1,
tx_hard_error_bump_tc = 0x2,
diff --git a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
index 1c9a520..b672e51 100644
--- a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
@@ -31,7 +31,15 @@ static int enh_desc_get_tx_status(void *data, struct stmmac_extra_stats *x,
{
struct net_device_stats *stats = (struct net_device_stats *)data;
unsigned int tdes0 = p->des0;
- int ret = 0;
+ int ret = tx_done;
+
+ /* Get tx owner first */
+ if (unlikely(tdes0 & ETDES0_OWN))
+ return tx_dma_own;
+
+ /* Verify tx error by looking at the last segment. */
+ if (likely(!(tdes0 & ETDES0_LAST_SEGMENT)))
+ return tx_not_ls;
if (unlikely(tdes0 & ETDES0_ERROR_SUMMARY)) {
if (unlikely(tdes0 & ETDES0_JABBER_TIMEOUT))
@@ -71,7 +79,7 @@ static int enh_desc_get_tx_status(void *data, struct stmmac_extra_stats *x,
dwmac_dma_flush_tx_fifo(ioaddr);
}
- ret = -1;
+ ret = tx_err;
}
if (unlikely(tdes0 & ETDES0_DEFERRED))
diff --git a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
index df2603e..9ccae3c 100644
--- a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
@@ -31,7 +31,16 @@ static int ndesc_get_tx_status(void *data, struct stmmac_extra_stats *x,
{
struct net_device_stats *stats = (struct net_device_stats *)data;
unsigned int tdes0 = p->des0;
- int ret = 0;
+ unsigned int tdes1 = p->des1;
+ int ret = tx_done;
+
+ /* Get tx owner first */
+ if (unlikely(tdes0 & TDES0_OWN))
+ return tx_dma_own;
+
+ /* Verify tx error by looking at the last segment. */
+ if (likely(!(tdes1 & TDES1_LAST_SEGMENT)))
+ return tx_not_ls;
if (unlikely(tdes0 & TDES0_ERROR_SUMMARY)) {
if (unlikely(tdes0 & TDES0_UNDERFLOW_ERROR)) {
@@ -54,7 +63,7 @@ static int ndesc_get_tx_status(void *data, struct stmmac_extra_stats *x,
collisions = (tdes0 & TDES0_COLLISION_COUNT_MASK) >> 3;
stats->collisions += collisions;
}
- ret = -1;
+ ret = tx_err;
}
if (tdes0 & TDES0_VLAN_FRAME)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 2f2f607..e51eee3 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1333,32 +1333,31 @@ static void stmmac_tx_clean(struct stmmac_priv *priv)
priv->xstats.tx_clean++;
while (entry != priv->cur_tx) {
- int last;
struct sk_buff *skb = priv->tx_skbuff[entry];
struct dma_desc *p;
+ int status;
if (priv->extend_desc)
p = (struct dma_desc *)(priv->dma_etx + entry);
else
p = priv->dma_tx + entry;
- /* Check if the descriptor is owned by the DMA. */
- if (priv->hw->desc->get_tx_owner(p))
- break;
-
- /* Verify tx error by looking at the last segment. */
- last = priv->tx_skbuff_dma[entry].last_segment;
- if (likely(last)) {
- int tx_error =
- priv->hw->desc->tx_status(&priv->dev->stats,
+ status = priv->hw->desc->tx_status(&priv->dev->stats,
&priv->xstats, p,
priv->ioaddr);
- if (likely(tx_error == 0)) {
+ /* Check if the descriptor is owned by the DMA */
+ if (unlikely(status & tx_dma_own))
+ break;
+
+ /* Just consider the last segment and ...*/
+ if (likely(!(status & tx_not_ls))) {
+ /* ... verify the status error condition */
+ if (unlikely(status & tx_err)) {
+ priv->dev->stats.tx_errors++;
+ } else {
priv->dev->stats.tx_packets++;
priv->xstats.tx_pkt_n++;
- } else
- priv->dev->stats.tx_errors++;
-
+ }
stmmac_get_tx_hwtstamp(priv, entry, skb);
}
if (netif_msg_tx_done(priv))
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 11/18] stmmac: set dirty index out of the loop
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (9 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 10/18] stmmac: optimize tx clean function Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 12/18] stmmac: first frame prep at the end of xmit routine Giuseppe Cavallaro
` (6 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
The dirty index can be updated out of the loop where all the
tx resources are claimed. This will help on performances too.
Also a useless debug printk has been removed from the main loop.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 5 +----
1 files changed, 1 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index e51eee3..38468e9 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1360,9 +1360,6 @@ static void stmmac_tx_clean(struct stmmac_priv *priv)
}
stmmac_get_tx_hwtstamp(priv, entry, skb);
}
- if (netif_msg_tx_done(priv))
- pr_debug("%s: curr %d, dirty %d\n", __func__,
- priv->cur_tx, priv->dirty_tx);
if (likely(priv->tx_skbuff_dma[entry].buf)) {
if (priv->tx_skbuff_dma[entry].map_as_page)
@@ -1393,8 +1390,8 @@ static void stmmac_tx_clean(struct stmmac_priv *priv)
if (++entry >= txsize)
entry = 0;
- priv->dirty_tx = entry;
}
+ priv->dirty_tx = entry;
netdev_completed_queue(priv->dev, pkts_compl, bytes_compl);
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 12/18] stmmac: first frame prep at the end of xmit routine
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (10 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 11/18] stmmac: set dirty index out of the loop Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 13/18] stmmac: perf, remove modulo in stmmac_rx() Giuseppe Cavallaro
` (5 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro, Fabrice Gasnier
This patch is to fill the first descriptor just before granting
the DMA engine so at the end of the xmit.
The patch takes care about the algorithm adopted to mitigate the
interrupts, then it fixes the last segment in case of no fragments.
Moreover, this new implementation does not pass any "ter" field when
prepare the descriptors because this is not necessary.
The patch also details the memory barrier in the xmit.
As final results, this patch guarantees the same performances
but fixing a case if small datagram are sent. In fact, this
kind of test is impacted if no coalesce is done.
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/common.h | 7 +-
drivers/net/ethernet/stmicro/stmmac/enh_desc.c | 25 +++--
drivers/net/ethernet/stmicro/stmmac/norm_desc.c | 24 ++--
drivers/net/ethernet/stmicro/stmmac/ring_mode.c | 2 -
.../net/ethernet/stmicro/stmmac/stmmac_ethtool.c | 2 +-
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 114 +++++++++++--------
6 files changed, 95 insertions(+), 79 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/common.h b/drivers/net/ethernet/stmicro/stmmac/common.h
index fb01e94..3e13b10 100644
--- a/drivers/net/ethernet/stmicro/stmmac/common.h
+++ b/drivers/net/ethernet/stmicro/stmmac/common.h
@@ -96,7 +96,7 @@ struct stmmac_extra_stats {
unsigned long napi_poll;
unsigned long tx_normal_irq_n;
unsigned long tx_clean;
- unsigned long tx_reset_ic_bit;
+ unsigned long tx_set_ic_bit;
unsigned long irq_receive_pmt_irq_n;
/* MMC info */
unsigned long mmc_tx_irq_n;
@@ -342,8 +342,7 @@ struct stmmac_desc_ops {
/* Invoked by the xmit function to prepare the tx descriptor */
void (*prepare_tx_desc) (struct dma_desc *p, int is_fs, int len,
- int csum_flag, int mode, int tx_own,
- bool ls_ic);
+ int csum_flag, int mode, int tx_own, bool ls);
/* Set/get the owner of the descriptor */
void (*set_tx_owner) (struct dma_desc *p);
int (*get_tx_owner) (struct dma_desc *p);
@@ -351,7 +350,7 @@ struct stmmac_desc_ops {
void (*release_tx_desc) (struct dma_desc *p, int mode);
/* Clear interrupt on tx frame completion. When this bit is
* set an interrupt happens as soon as the frame is transmitted */
- void (*clear_tx_ic) (struct dma_desc *p);
+ void (*set_tx_ic)(struct dma_desc *p);
/* Last tx segment reports the transmit status */
int (*get_tx_ls) (struct dma_desc *p);
/* Return the transmit status looking at the TDES1 */
diff --git a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
index b672e51..76e1c15 100644
--- a/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/enh_desc.c
@@ -311,10 +311,15 @@ static void enh_desc_release_tx_desc(struct dma_desc *p, int mode)
static void enh_desc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
int csum_flag, int mode, int tx_own,
- bool ls_ic)
+ bool ls)
{
unsigned int tdes0 = p->des0;
+ if (mode == STMMAC_CHAIN_MODE)
+ enh_set_tx_desc_len_on_chain(p, len);
+ else
+ enh_set_tx_desc_len_on_ring(p, len);
+
if (is_fs)
tdes0 |= ETDES0_FIRST_SEGMENT;
else
@@ -325,6 +330,10 @@ static void enh_desc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
else
tdes0 &= ~(TX_CIC_FULL << ETDES0_CHECKSUM_INSERTION_SHIFT);
+ if (ls)
+ tdes0 |= ETDES0_LAST_SEGMENT;
+
+ /* Finally set the OWN bit. Later the DMA will start! */
if (tx_own)
tdes0 |= ETDES0_OWN;
@@ -335,20 +344,12 @@ static void enh_desc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
*/
wmb();
- if (ls_ic)
- tdes0 |= ETDES0_LAST_SEGMENT | ETDES0_INTERRUPT;
-
p->des0 = tdes0;
-
- if (mode == STMMAC_CHAIN_MODE)
- enh_set_tx_desc_len_on_chain(p, len);
- else
- enh_set_tx_desc_len_on_ring(p, len);
}
-static void enh_desc_clear_tx_ic(struct dma_desc *p)
+static void enh_desc_set_tx_ic(struct dma_desc *p)
{
- p->des0 &= ~ETDES0_INTERRUPT;
+ p->des0 |= ETDES0_INTERRUPT;
}
static int enh_desc_get_rx_frame_len(struct dma_desc *p, int rx_coe_type)
@@ -419,7 +420,7 @@ const struct stmmac_desc_ops enh_desc_ops = {
.get_tx_owner = enh_desc_get_tx_owner,
.release_tx_desc = enh_desc_release_tx_desc,
.prepare_tx_desc = enh_desc_prepare_tx_desc,
- .clear_tx_ic = enh_desc_clear_tx_ic,
+ .set_tx_ic = enh_desc_set_tx_ic,
.get_tx_ls = enh_desc_get_tx_ls,
.set_tx_owner = enh_desc_set_tx_owner,
.set_rx_owner = enh_desc_set_rx_owner,
diff --git a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
index 9ccae3c..f227be6 100644
--- a/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
+++ b/drivers/net/ethernet/stmicro/stmmac/norm_desc.c
@@ -195,10 +195,15 @@ static void ndesc_release_tx_desc(struct dma_desc *p, int mode)
static void ndesc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
int csum_flag, int mode, int tx_own,
- bool ls_ic)
+ bool ls)
{
unsigned int tdes1 = p->des1;
+ if (mode == STMMAC_CHAIN_MODE)
+ norm_set_tx_desc_len_on_chain(p, len);
+ else
+ norm_set_tx_desc_len_on_ring(p, len);
+
if (is_fs)
tdes1 |= TDES1_FIRST_SEGMENT;
else
@@ -209,23 +214,18 @@ static void ndesc_prepare_tx_desc(struct dma_desc *p, int is_fs, int len,
else
tdes1 &= ~(TX_CIC_FULL << TDES1_CHECKSUM_INSERTION_SHIFT);
+ if (ls)
+ tdes1 |= TDES1_LAST_SEGMENT;
+
if (tx_own)
tdes1 |= TDES0_OWN;
- if (ls_ic)
- tdes1 |= TDES1_LAST_SEGMENT | TDES1_INTERRUPT;
-
p->des1 = tdes1;
-
- if (mode == STMMAC_CHAIN_MODE)
- norm_set_tx_desc_len_on_chain(p, len);
- else
- norm_set_tx_desc_len_on_ring(p, len);
}
-static void ndesc_clear_tx_ic(struct dma_desc *p)
+static void ndesc_set_tx_ic(struct dma_desc *p)
{
- p->des1 &= ~TDES1_INTERRUPT;
+ p->des1 |= TDES1_INTERRUPT;
}
static int ndesc_get_rx_frame_len(struct dma_desc *p, int rx_coe_type)
@@ -288,7 +288,7 @@ const struct stmmac_desc_ops ndesc_ops = {
.get_tx_owner = ndesc_get_tx_owner,
.release_tx_desc = ndesc_release_tx_desc,
.prepare_tx_desc = ndesc_prepare_tx_desc,
- .clear_tx_ic = ndesc_clear_tx_ic,
+ .set_tx_ic = ndesc_set_tx_ic,
.get_tx_ls = ndesc_get_tx_ls,
.set_tx_owner = ndesc_set_tx_owner,
.set_rx_owner = ndesc_set_rx_owner,
diff --git a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
index 4978a6c..1e90276 100644
--- a/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
+++ b/drivers/net/ethernet/stmicro/stmmac/ring_mode.c
@@ -63,7 +63,6 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 1, bmax, csum,
STMMAC_RING_MODE, 0, false);
- wmb();
priv->tx_skbuff[entry] = NULL;
if (++entry >= txsize)
entry = 0;
@@ -84,7 +83,6 @@ static int stmmac_jumbo_frm(void *p, struct sk_buff *skb, int csum)
desc->des3 = desc->des2 + BUF_SIZE_4KiB;
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum,
STMMAC_RING_MODE, 1, true);
- wmb();
} else {
desc->des2 = dma_map_single(priv->device, skb->data,
nopaged_len, DMA_TO_DEVICE);
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c
index 4c6486c..c803d4c 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c
@@ -97,7 +97,7 @@ static const struct stmmac_stats stmmac_gstrings_stats[] = {
STMMAC_STAT(napi_poll),
STMMAC_STAT(tx_normal_irq_n),
STMMAC_STAT(tx_clean),
- STMMAC_STAT(tx_reset_ic_bit),
+ STMMAC_STAT(tx_set_ic_bit),
STMMAC_STAT(irq_receive_pmt_irq_n),
/* MMC info */
STMMAC_STAT(mmc_tx_irq_n),
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 38468e9..0a39331 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1967,13 +1967,13 @@ static int stmmac_release(struct net_device *dev)
static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct stmmac_priv *priv = netdev_priv(dev);
+ unsigned int nopaged_len = skb_headlen(skb);
unsigned int txsize = priv->dma_tx_size;
- int entry;
int i, csum_insertion = 0, is_jumbo = 0;
int nfrags = skb_shinfo(skb)->nr_frags;
+ unsigned int entry, first_entry;
struct dma_desc *desc, *first;
- unsigned int nopaged_len = skb_headlen(skb);
- unsigned int enh_desc = priv->plat->enh_desc;
+ unsigned int enh_desc;
spin_lock(&priv->tx_lock);
@@ -1991,34 +1991,25 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
stmmac_disable_eee_mode(priv);
entry = priv->cur_tx;
-
+ first_entry = entry;
csum_insertion = (skb->ip_summed == CHECKSUM_PARTIAL);
- if (priv->extend_desc)
+ if (likely(priv->extend_desc))
desc = (struct dma_desc *)(priv->dma_etx + entry);
else
desc = priv->dma_tx + entry;
first = desc;
+ priv->tx_skbuff[first_entry] = skb;
+
+ enh_desc = priv->plat->enh_desc;
/* To program the descriptors according to the size of the frame */
if (enh_desc)
is_jumbo = priv->hw->mode->is_jumbo_frm(skb->len, enh_desc);
- if (likely(!is_jumbo)) {
- desc->des2 = dma_map_single(priv->device, skb->data,
- nopaged_len, DMA_TO_DEVICE);
- if (dma_mapping_error(priv->device, desc->des2))
- goto dma_map_err;
- priv->tx_skbuff_dma[entry].buf = desc->des2;
- priv->tx_skbuff_dma[entry].len = nopaged_len;
- /* do not set the own at this stage */
- priv->hw->desc->prepare_tx_desc(desc, 1, nopaged_len,
- csum_insertion, priv->mode, 0,
- nfrags == 0);
- } else {
- desc = first;
+ if (unlikely(is_jumbo)) {
entry = priv->hw->mode->jumbo_frm(priv, skb, csum_insertion);
if (unlikely(entry < 0))
goto dma_map_err;
@@ -2029,11 +2020,10 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
int len = skb_frag_size(frag);
bool last_segment = (i == (nfrags - 1));
- priv->tx_skbuff[entry] = NULL;
if (++entry >= txsize)
entry = 0;
- if (priv->extend_desc)
+ if (likely(priv->extend_desc))
desc = (struct dma_desc *)(priv->dma_etx + entry);
else
desc = priv->dma_tx + entry;
@@ -2043,42 +2033,26 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
if (dma_mapping_error(priv->device, desc->des2))
goto dma_map_err; /* should reuse desc w/o issues */
+ priv->tx_skbuff[entry] = NULL;
priv->tx_skbuff_dma[entry].buf = desc->des2;
priv->tx_skbuff_dma[entry].map_as_page = true;
priv->tx_skbuff_dma[entry].len = len;
+ priv->tx_skbuff_dma[entry].last_segment = last_segment;
+
+ /* Prepare the descriptor and set the own bit too */
priv->hw->desc->prepare_tx_desc(desc, 0, len, csum_insertion,
priv->mode, 1, last_segment);
- priv->tx_skbuff_dma[entry].last_segment = last_segment;
}
- priv->tx_skbuff[entry] = skb;
-
- /* According to the coalesce parameter the IC bit for the latest
- * segment could be reset and the timer re-started to invoke the
- * stmmac_tx function. This approach takes care about the fragments.
- */
- priv->tx_count_frames += nfrags + 1;
- if (priv->tx_coal_frames > priv->tx_count_frames) {
- priv->hw->desc->clear_tx_ic(desc);
- priv->xstats.tx_reset_ic_bit++;
- mod_timer(&priv->txtimer,
- STMMAC_COAL_TIMER(priv->tx_coal_timer));
- } else
- priv->tx_count_frames = 0;
-
- /* To avoid raise condition */
- priv->hw->desc->set_tx_owner(first);
- wmb();
-
if (++entry >= txsize)
entry = 0;
priv->cur_tx = entry;
if (netif_msg_pktdata(priv)) {
- pr_debug("%s: curr %d dirty=%d entry=%d, first=%p, nfrags=%d",
- __func__, (priv->cur_tx % txsize),
- (priv->dirty_tx % txsize), entry, first, nfrags);
+ pr_debug("%s: curr=%d dirty=%d f=%d, e=%d, first=%p, nfrags=%d",
+ __func__, priv->cur_tx, priv->dirty_tx, first_entry,
+ entry, first, nfrags);
if (priv->extend_desc)
stmmac_display_ring((void *)priv->dma_etx, txsize, 1);
@@ -2088,6 +2062,7 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
pr_debug(">>> frame to be transmitted: ");
print_pkt(skb->data, skb->len);
}
+
if (unlikely(stmmac_tx_avail(priv) <= (MAX_SKB_FRAGS + 1))) {
if (netif_msg_hw(priv))
pr_debug("%s: stop transmitted packets\n", __func__);
@@ -2096,16 +2071,59 @@ static netdev_tx_t stmmac_xmit(struct sk_buff *skb, struct net_device *dev)
dev->stats.tx_bytes += skb->len;
- if (unlikely((skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
- priv->hwts_tx_en)) {
- /* declare that device is doing timestamping */
- skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
- priv->hw->desc->enable_tx_timestamp(first);
+ /* According to the coalesce parameter the IC bit for the latest
+ * segment is reset and the timer re-started to clean the tx status.
+ * This approach takes care about the fragments: desc is the first
+ * element in case of no SG.
+ */
+ priv->tx_count_frames += nfrags + 1;
+ if (likely(priv->tx_coal_frames > priv->tx_count_frames)) {
+ mod_timer(&priv->txtimer,
+ STMMAC_COAL_TIMER(priv->tx_coal_timer));
+ } else {
+ priv->tx_count_frames = 0;
+ priv->hw->desc->set_tx_ic(desc);
+ priv->xstats.tx_set_ic_bit++;
}
if (!priv->hwts_tx_en)
skb_tx_timestamp(skb);
+ /* Ready to fill the first descriptor and set the OWN bit w/o any
+ * problems because all the descriptors are actually ready to be
+ * passed to the DMA engine.
+ */
+ if (likely(!is_jumbo)) {
+ bool last_segment = (nfrags == 0);
+
+ first->des2 = dma_map_single(priv->device, skb->data,
+ nopaged_len, DMA_TO_DEVICE);
+ if (dma_mapping_error(priv->device, first->des2))
+ goto dma_map_err;
+
+ priv->tx_skbuff_dma[first_entry].buf = first->des2;
+ priv->tx_skbuff_dma[first_entry].len = nopaged_len;
+ priv->tx_skbuff_dma[first_entry].last_segment = last_segment;
+
+ if (unlikely((skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
+ priv->hwts_tx_en)) {
+ /* declare that device is doing timestamping */
+ skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
+ priv->hw->desc->enable_tx_timestamp(first);
+ }
+
+ /* Prepare the first descriptor setting the OWN bit too */
+ priv->hw->desc->prepare_tx_desc(first, 1, nopaged_len,
+ csum_insertion, priv->mode, 1,
+ last_segment);
+
+ /* The own bit must be the latest setting done when prepare the
+ * descriptor and then barrier is needed to make sure that
+ * all is coherent before granting the DMA engine.
+ */
+ smp_wmb();
+ }
+
netdev_sent_queue(dev, skb->len);
priv->hw->dma->enable_dma_transmission(priv->ioaddr);
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 13/18] stmmac: perf, remove modulo in stmmac_rx()
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (11 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 12/18] stmmac: first frame prep at the end of xmit routine Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 17:21 ` David Laight
2015-12-09 8:37 ` [PATCH (net-next.git) 14/18] stmmac: do not poll phy handler when attach a switch Giuseppe Cavallaro
` (4 subsequent siblings)
17 siblings, 1 reply; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro, Fabrice Gasnier
The indexes into the ring buffer are always incremented, and
the entry is accessed via doing a modulo to find the "real" index.
Modulo is an expensive operation.
This patch replaces the modulo with a simple if clamp.
It helps improve stmmac RX path as it's being called inside RX loop.
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 28 ++++++++++++++++++---
1 files changed, 24 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 0a39331..52ac648 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -211,6 +211,18 @@ static inline u32 stmmac_tx_avail(struct stmmac_priv *priv)
return avail;
}
+static inline u32 stmmac_rx_dirty(struct stmmac_priv *priv)
+{
+ unsigned dirty;
+
+ if (priv->dirty_rx <= priv->cur_rx)
+ dirty = priv->cur_rx - priv->dirty_rx;
+ else
+ dirty = priv->dma_rx_size - priv->dirty_rx + priv->cur_rx;
+
+ return dirty;
+}
+
/**
* stmmac_hw_fix_mac_speed - callback for speed selection
* @priv: driver private structure
@@ -2165,9 +2177,10 @@ static inline void stmmac_rx_refill(struct stmmac_priv *priv)
{
unsigned int rxsize = priv->dma_rx_size;
int bfsize = priv->dma_buf_sz;
+ unsigned int entry = priv->dirty_rx;
+ int dirty = stmmac_rx_dirty(priv);
- for (; priv->cur_rx - priv->dirty_rx > 0; priv->dirty_rx++) {
- unsigned int entry = priv->dirty_rx % rxsize;
+ while (dirty-- > 0) {
struct dma_desc *p;
if (priv->extend_desc)
@@ -2203,6 +2216,10 @@ static inline void stmmac_rx_refill(struct stmmac_priv *priv)
wmb();
priv->hw->desc->set_rx_owner(p);
wmb();
+
+ if (unlikely(++priv->dirty_rx >= rxsize))
+ priv->dirty_rx = 0;
+ entry = priv->dirty_rx;
}
}
@@ -2216,7 +2233,7 @@ static inline void stmmac_rx_refill(struct stmmac_priv *priv)
static int stmmac_rx(struct stmmac_priv *priv, int limit)
{
unsigned int rxsize = priv->dma_rx_size;
- unsigned int entry = priv->cur_rx % rxsize;
+ unsigned int entry = priv->cur_rx;
unsigned int next_entry;
unsigned int count = 0;
int coe = priv->hw->rx_csum;
@@ -2246,7 +2263,10 @@ static int stmmac_rx(struct stmmac_priv *priv, int limit)
count++;
- next_entry = (++priv->cur_rx) % rxsize;
+ if (unlikely(++priv->cur_rx >= rxsize))
+ priv->cur_rx = 0;
+ next_entry = priv->cur_rx;
+
if (priv->extend_desc)
prefetch(priv->dma_erx + next_entry);
else
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 14/18] stmmac: do not poll phy handler when attach a switch
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (12 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 13/18] stmmac: perf, remove modulo in stmmac_rx() Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 15/18] stmmac: fix phy init when attached to a phy Giuseppe Cavallaro
` (3 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
This patch avoids to call the stmmac_adjust_link when
the driver is connected to a switch by using the FIXED_PHY
support. Prior this patch the phydev->irq was set as PHY_POLL
so periodically the phy handler was invoked spending useless
time because the link cannot actually change.
Note that the stmmac_adjust_link will be called just one
time and this guarantees that the ST glue logic will be
setup according to the mode and speed fixed.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 5 +++++
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 52ac648..cc04943 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -884,6 +884,11 @@ static int stmmac_init_phy(struct net_device *dev)
phy_disconnect(phydev);
return -ENODEV;
}
+
+ /* If attached to a switch, there is no reason to poll phy handler */
+ if (!strcmp(priv->plat->phy_bus_name, "fixed"))
+ phydev->irq = PHY_IGNORE_INTERRUPT;
+
pr_debug("stmmac_init_phy: %s: attached to PHY (UID 0x%x)"
" Link = %d\n", dev->name, phydev->phy_id, phydev->link);
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 15/18] stmmac: fix phy init when attached to a phy
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (13 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 14/18] stmmac: do not poll phy handler when attach a switch Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 16/18] stmmac: do not perform zero-copy for rx frames Giuseppe Cavallaro
` (2 subsequent siblings)
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Fabrice Gasnier, Giuseppe Cavallaro
From: Fabrice Gasnier <fabrice.gasnier@st.com>
phy_bus_name can be NULL when "fixed-link" property isn't used.
Then, since "stmmac: do not poll phy handler when attach a switch",
phy_bus_name ptr needs to be checked before strcmp is called.
Signed-off-by: Fabrice Gasnier <fabrice.gasnier@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 5 +++--
1 files changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index cc04943..6b83213 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -886,8 +886,9 @@ static int stmmac_init_phy(struct net_device *dev)
}
/* If attached to a switch, there is no reason to poll phy handler */
- if (!strcmp(priv->plat->phy_bus_name, "fixed"))
- phydev->irq = PHY_IGNORE_INTERRUPT;
+ if (priv->plat->phy_bus_name)
+ if (!strcmp(priv->plat->phy_bus_name, "fixed"))
+ phydev->irq = PHY_IGNORE_INTERRUPT;
pr_debug("stmmac_init_phy: %s: attached to PHY (UID 0x%x)"
" Link = %d\n", dev->name, phydev->phy_id, phydev->link);
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 16/18] stmmac: do not perform zero-copy for rx frames
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (14 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 15/18] stmmac: fix phy init when attached to a phy Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-12 1:09 ` David Miller
2015-12-09 8:37 ` [PATCH (net-next.git) 17/18] stmmac: tune rx copy via threshold Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 18/18] stmmac: update version to Oct_2015 Giuseppe Cavallaro
17 siblings, 1 reply; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
This patch is to allow this driver to copy tiny frames during the reception
process. This is giving more stability while stressing the driver on STi
embedded systems.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 67 ++++++++++++++++-----
1 files changed, 52 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 6b83213..d68775a 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -98,6 +98,10 @@ static int buf_sz = DEFAULT_BUFSIZE;
module_param(buf_sz, int, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(buf_sz, "DMA buffer size");
+static int minrx = 256;
+module_param(minrx, int, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(minrx, "Copy only tiny-frames");
+
static const u32 default_msg_level = (NETIF_MSG_DRV | NETIF_MSG_PROBE |
NETIF_MSG_LINK | NETIF_MSG_IFUP |
NETIF_MSG_IFDOWN | NETIF_MSG_TIMER);
@@ -147,6 +151,8 @@ static void stmmac_verify_args(void)
pause = PAUSE_TIME;
if (eee_timer < 0)
eee_timer = STMMAC_DEFAULT_LPI_TIMER;
+ if (minrx < 0)
+ minrx = ETH_FRAME_LEN;
}
/**
@@ -2198,8 +2204,7 @@ static inline void stmmac_rx_refill(struct stmmac_priv *priv)
struct sk_buff *skb;
skb = netdev_alloc_skb_ip_align(priv->dev, bfsize);
-
- if (unlikely(skb == NULL))
+ if (unlikely(!skb))
break;
priv->rx_skbuff[entry] = skb;
@@ -2322,23 +2327,52 @@ static int stmmac_rx(struct stmmac_priv *priv, int limit)
pr_debug("\tframe size %d, COE: %d\n",
frame_len, status);
}
- skb = priv->rx_skbuff[entry];
- if (unlikely(!skb)) {
- pr_err("%s: Inconsistent Rx descriptor chain\n",
- priv->dev->name);
- priv->dev->stats.rx_dropped++;
- break;
+
+ if (unlikely(frame_len < minrx)) {
+ skb = netdev_alloc_skb_ip_align(priv->dev,
+ frame_len);
+ if (unlikely(!skb)) {
+ if (net_ratelimit())
+ dev_warn(priv->device,
+ "packet dropped\n");
+ priv->dev->stats.rx_dropped++;
+ break;
+ }
+
+ dma_sync_single_for_cpu(priv->device,
+ priv->rx_skbuff_dma
+ [entry], frame_len,
+ DMA_FROM_DEVICE);
+ skb_copy_to_linear_data(skb,
+ priv->
+ rx_skbuff[entry]->data,
+ frame_len);
+
+ skb_put(skb, frame_len);
+ dma_sync_single_for_device(priv->device,
+ priv->rx_skbuff_dma
+ [entry], frame_len,
+ DMA_FROM_DEVICE);
+ } else {
+ skb = priv->rx_skbuff[entry];
+ if (unlikely(!skb)) {
+ pr_err("%s: Inconsistent Rx chain\n",
+ priv->dev->name);
+ priv->dev->stats.rx_dropped++;
+ break;
+ }
+ prefetch(skb->data - NET_IP_ALIGN);
+ priv->rx_skbuff[entry] = NULL;
+
+ skb_put(skb, frame_len);
+ dma_unmap_single(priv->device,
+ priv->rx_skbuff_dma[entry],
+ priv->dma_buf_sz,
+ DMA_FROM_DEVICE);
}
- prefetch(skb->data - NET_IP_ALIGN);
- priv->rx_skbuff[entry] = NULL;
stmmac_get_rx_hwtstamp(priv, entry, skb);
- skb_put(skb, frame_len);
- dma_unmap_single(priv->device,
- priv->rx_skbuff_dma[entry],
- priv->dma_buf_sz, DMA_FROM_DEVICE);
-
if (netif_msg_pktdata(priv)) {
pr_debug("frame received (%dbytes)", frame_len);
print_pkt(skb->data, frame_len);
@@ -3235,6 +3269,9 @@ static int __init stmmac_cmdline_opt(char *str)
} else if (!strncmp(opt, "chain_mode:", 11)) {
if (kstrtoint(opt + 11, 0, &chain_mode))
goto err;
+ } else if (!strncmp(opt, "minrx:", 6)) {
+ if (kstrtoint(opt + 6, 0, &minrx))
+ goto err;
}
}
return 0;
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 17/18] stmmac: tune rx copy via threshold.
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (15 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 16/18] stmmac: do not perform zero-copy for rx frames Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 18/18] stmmac: update version to Oct_2015 Giuseppe Cavallaro
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
There is a threshold now used to also limit the skb allocation
when use zero-copy. This is to avoid that there are incoherence
in the ring due to a failure on skb allocation under very
aggressive testing and under low memory conditions.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/stmmac.h | 1 +
drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 26 +++++++++++++++++++-
2 files changed, 25 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
index b73e8d8..55180e0 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
@@ -76,6 +76,7 @@ struct stmmac_priv {
unsigned int dirty_rx;
unsigned int dma_rx_size;
unsigned int dma_buf_sz;
+ unsigned int rx_zeroc_thresh;
u32 rx_riwt;
int hwts_rx_en;
dma_addr_t *rx_skbuff_dma;
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index d68775a..74653cf 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -204,6 +204,7 @@ static void print_pkt(unsigned char *buf, int len)
/* minimum number of free TX descriptors required to wake up TX process */
#define STMMAC_TX_THRESH(x) (x->dma_tx_size/4)
+#define STMMAC_RX_THRESH(x) (x->dma_rx_size / 4)
static inline u32 stmmac_tx_avail(struct stmmac_priv *priv)
{
@@ -2179,6 +2180,14 @@ static void stmmac_rx_vlan(struct net_device *dev, struct sk_buff *skb)
}
+static inline int stmmac_rx_threshold_count(struct stmmac_priv *priv)
+{
+ if (priv->rx_zeroc_thresh < STMMAC_RX_THRESH(priv))
+ return 0;
+
+ return 1;
+}
+
/**
* stmmac_rx_refill - refill used skb preallocated buffers
* @priv: driver private structure
@@ -2204,8 +2213,15 @@ static inline void stmmac_rx_refill(struct stmmac_priv *priv)
struct sk_buff *skb;
skb = netdev_alloc_skb_ip_align(priv->dev, bfsize);
- if (unlikely(!skb))
+ if (unlikely(!skb)) {
+ /* so for a while no zero-copy! */
+ priv->rx_zeroc_thresh = STMMAC_RX_THRESH(priv);
+ if (unlikely(net_ratelimit()))
+ dev_err(priv->device,
+ "fail to alloc skb entry %d\n",
+ entry);
break;
+ }
priv->rx_skbuff[entry] = skb;
priv->rx_skbuff_dma[entry] =
@@ -2221,9 +2237,13 @@ static inline void stmmac_rx_refill(struct stmmac_priv *priv)
priv->hw->mode->refill_desc3(priv, p);
+ if (priv->rx_zeroc_thresh > 0)
+ priv->rx_zeroc_thresh--;
+
if (netif_msg_rx_status(priv))
pr_debug("\trefill entry #%d\n", entry);
}
+
wmb();
priv->hw->desc->set_rx_owner(p);
wmb();
@@ -2328,7 +2348,8 @@ static int stmmac_rx(struct stmmac_priv *priv, int limit)
frame_len, status);
}
- if (unlikely(frame_len < minrx)) {
+ if (unlikely((frame_len < minrx) ||
+ stmmac_rx_threshold_count(priv))) {
skb = netdev_alloc_skb_ip_align(priv->dev,
frame_len);
if (unlikely(!skb)) {
@@ -2363,6 +2384,7 @@ static int stmmac_rx(struct stmmac_priv *priv, int limit)
}
prefetch(skb->data - NET_IP_ALIGN);
priv->rx_skbuff[entry] = NULL;
+ priv->rx_zeroc_thresh++;
skb_put(skb, frame_len);
dma_unmap_single(priv->device,
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH (net-next.git) 18/18] stmmac: update version to Oct_2015
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
` (16 preceding siblings ...)
2015-12-09 8:37 ` [PATCH (net-next.git) 17/18] stmmac: tune rx copy via threshold Giuseppe Cavallaro
@ 2015-12-09 8:37 ` Giuseppe Cavallaro
17 siblings, 0 replies; 24+ messages in thread
From: Giuseppe Cavallaro @ 2015-12-09 8:37 UTC (permalink / raw)
To: netdev; +Cc: alexandre.torgue, Giuseppe Cavallaro
This patch just updates the driver to the version fully
tested on STi platforms. This version is Oct_2015.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
---
drivers/net/ethernet/stmicro/stmmac/stmmac.h | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
index 55180e0..99c64a1 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
@@ -24,7 +24,7 @@
#define __STMMAC_H__
#define STMMAC_RESOURCE_NAME "stmmaceth"
-#define DRV_MODULE_VERSION "March_2013"
+#define DRV_MODULE_VERSION "Oct_2015"
#include <linux/clk.h>
#include <linux/stmmac.h>
--
1.7.4.4
^ permalink raw reply related [flat|nested] 24+ messages in thread
* RE: [PATCH (net-next.git) 13/18] stmmac: perf, remove modulo in stmmac_rx()
2015-12-09 8:37 ` [PATCH (net-next.git) 13/18] stmmac: perf, remove modulo in stmmac_rx() Giuseppe Cavallaro
@ 2015-12-09 17:21 ` David Laight
2015-12-10 5:11 ` Giuseppe CAVALLARO
0 siblings, 1 reply; 24+ messages in thread
From: David Laight @ 2015-12-09 17:21 UTC (permalink / raw)
To: 'Giuseppe Cavallaro', netdev; +Cc: alexandre.torgue, Fabrice Gasnier
From: Giuseppe Cavallaro
> Sent: 09 December 2015 08:38
> The indexes into the ring buffer are always incremented, and
> the entry is accessed via doing a modulo to find the "real" index.
> Modulo is an expensive operation.
>
> This patch replaces the modulo with a simple if clamp.
> It helps improve stmmac RX path as it's being called inside RX loop.
Is dma_rx_size always a power of 2 ?
If so you can replace the % by & and remove the conditionals.
If you have changed the read and write values to indexes
then you need to be certain that the 'ring full' and 'ring empty'
cases are correctly distinguished.
David
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH (net-next.git) 13/18] stmmac: perf, remove modulo in stmmac_rx()
2015-12-09 17:21 ` David Laight
@ 2015-12-10 5:11 ` Giuseppe CAVALLARO
2015-12-10 5:51 ` Giuseppe CAVALLARO
0 siblings, 1 reply; 24+ messages in thread
From: Giuseppe CAVALLARO @ 2015-12-10 5:11 UTC (permalink / raw)
To: David Laight, netdev; +Cc: alexandre.torgue, Fabrice Gasnier
On 12/9/2015 6:21 PM, David Laight wrote:
> From: Giuseppe Cavallaro
>> Sent: 09 December 2015 08:38
>> The indexes into the ring buffer are always incremented, and
>> the entry is accessed via doing a modulo to find the "real" index.
>> Modulo is an expensive operation.
>>
>> This patch replaces the modulo with a simple if clamp.
>> It helps improve stmmac RX path as it's being called inside RX loop.
>
> Is dma_rx_size always a power of 2 ?
no
> If so you can replace the % by & and remove the conditionals.
>
> If you have changed the read and write values to indexes
> then you need to be certain that the 'ring full' and 'ring empty'
> cases are correctly distinguished.
this is implicitly managed by dirty and curr index.
thx
peppe
>
> David
>
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH (net-next.git) 13/18] stmmac: perf, remove modulo in stmmac_rx()
2015-12-10 5:11 ` Giuseppe CAVALLARO
@ 2015-12-10 5:51 ` Giuseppe CAVALLARO
0 siblings, 0 replies; 24+ messages in thread
From: Giuseppe CAVALLARO @ 2015-12-10 5:51 UTC (permalink / raw)
To: David Laight, netdev; +Cc: alexandre.torgue, Fabrice Gasnier
On 12/10/2015 6:11 AM, Giuseppe CAVALLARO wrote:
> On 12/9/2015 6:21 PM, David Laight wrote:
>> From: Giuseppe Cavallaro
>>> Sent: 09 December 2015 08:38
>>> The indexes into the ring buffer are always incremented, and
>>> the entry is accessed via doing a modulo to find the "real" index.
>>> Modulo is an expensive operation.
>>>
>>> This patch replaces the modulo with a simple if clamp.
>>> It helps improve stmmac RX path as it's being called inside RX loop.
>>
>> Is dma_rx_size always a power of 2 ?
>
> no
>
>> If so you can replace the % by & and remove the conditionals.
>>
>> If you have changed the read and write values to indexes
>> then you need to be certain that the 'ring full' and 'ring empty'
>> cases are correctly distinguished.
>
> this is implicitly managed by dirty and curr index.
... also the dma will fail with unavailable buffer.
peppe
>
> thx
> peppe
>
>>
>> David
>>
>
>
>
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH (net-next.git) 16/18] stmmac: do not perform zero-copy for rx frames
2015-12-09 8:37 ` [PATCH (net-next.git) 16/18] stmmac: do not perform zero-copy for rx frames Giuseppe Cavallaro
@ 2015-12-12 1:09 ` David Miller
2015-12-16 11:33 ` Giuseppe CAVALLARO
0 siblings, 1 reply; 24+ messages in thread
From: David Miller @ 2015-12-12 1:09 UTC (permalink / raw)
To: peppe.cavallaro; +Cc: netdev, alexandre.torgue
From: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Date: Wed, 9 Dec 2015 09:37:52 +0100
> @@ -98,6 +98,10 @@ static int buf_sz = DEFAULT_BUFSIZE;
> module_param(buf_sz, int, S_IRUGO | S_IWUSR);
> MODULE_PARM_DESC(buf_sz, "DMA buffer size");
>
> +static int minrx = 256;
> +module_param(minrx, int, S_IRUGO | S_IWUSR);
> +MODULE_PARM_DESC(minrx, "Copy only tiny-frames");
> +
When you type module_param() in a network driver, you ought to receive
an electric shock in order to remind you that except in the most extreme
cases module parameters are absolutely not appropriate.
In this case we have an ethtool tunable people can use to control copy
break values like this, so use that instead.
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH (net-next.git) 16/18] stmmac: do not perform zero-copy for rx frames
2015-12-12 1:09 ` David Miller
@ 2015-12-16 11:33 ` Giuseppe CAVALLARO
0 siblings, 0 replies; 24+ messages in thread
From: Giuseppe CAVALLARO @ 2015-12-16 11:33 UTC (permalink / raw)
To: David Miller; +Cc: netdev, alexandre.torgue
On 12/12/2015 2:09 AM, David Miller wrote:
> From: Giuseppe Cavallaro <peppe.cavallaro@st.com>
> Date: Wed, 9 Dec 2015 09:37:52 +0100
>
>> @@ -98,6 +98,10 @@ static int buf_sz = DEFAULT_BUFSIZE;
>> module_param(buf_sz, int, S_IRUGO | S_IWUSR);
>> MODULE_PARM_DESC(buf_sz, "DMA buffer size");
>>
>> +static int minrx = 256;
>> +module_param(minrx, int, S_IRUGO | S_IWUSR);
>> +MODULE_PARM_DESC(minrx, "Copy only tiny-frames");
>> +
>
> When you type module_param() in a network driver, you ought to receive
> an electric shock in order to remind you that except in the most extreme
> cases module parameters are absolutely not appropriate.
>
> In this case we have an ethtool tunable people can use to control copy
> break values like this, so use that instead.
ok, v2 will have this change. Let me know if there is other to fix.
Peppe
^ permalink raw reply [flat|nested] 24+ messages in thread
end of thread, other threads:[~2015-12-16 11:33 UTC | newest]
Thread overview: 24+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-12-09 8:37 [PATCH (net-next.git) 00/18] stmmac: enhance driver performances and update the version Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 01/18] stmmac: share reset function between dwmac100 and dwmac1000 Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 02/18] stmmac: rework DMA bus setting and introduce new platform AXI structure Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 03/18] stmmac: change descriptor layout Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 04/18] stmmac: remove modulo in stmmac_xmit() Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 05/18] stmmac: add length field to dma data Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 06/18] stmmac: add last_segment " Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 07/18] stmmac: add is_jumbo " Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 08/18] stmmac: merge get_rx_owner into rx_status routine Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 09/18] stmmac: optimize tx desc management Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 10/18] stmmac: optimize tx clean function Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 11/18] stmmac: set dirty index out of the loop Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 12/18] stmmac: first frame prep at the end of xmit routine Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 13/18] stmmac: perf, remove modulo in stmmac_rx() Giuseppe Cavallaro
2015-12-09 17:21 ` David Laight
2015-12-10 5:11 ` Giuseppe CAVALLARO
2015-12-10 5:51 ` Giuseppe CAVALLARO
2015-12-09 8:37 ` [PATCH (net-next.git) 14/18] stmmac: do not poll phy handler when attach a switch Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 15/18] stmmac: fix phy init when attached to a phy Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 16/18] stmmac: do not perform zero-copy for rx frames Giuseppe Cavallaro
2015-12-12 1:09 ` David Miller
2015-12-16 11:33 ` Giuseppe CAVALLARO
2015-12-09 8:37 ` [PATCH (net-next.git) 17/18] stmmac: tune rx copy via threshold Giuseppe Cavallaro
2015-12-09 8:37 ` [PATCH (net-next.git) 18/18] stmmac: update version to Oct_2015 Giuseppe Cavallaro
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.