Baikal-T1 SoC provides a DW DMA controller to perform low-speed peripherals
Mem-to-Dev and Dev-to-Mem transaction. This is also applicable to the DW
APB SSI devices embedded into the SoC. Currently the DMA-based transfers
are supported by the DW APB SPI driver only as a middle layer code for
Intel MID/Elkhart PCI devices. Seeing the same code can be used for normal
platform DMAC device we introduced a set of patches to fix it within this
series.
First of all we need to add the Tx and Rx DMA channels support into the DW
APB SSI binding. Then there are several fixes and cleanups provided as a
initial preparation for the Generic DMA support integration: add Tx/Rx
finish wait methods, clear DMAC register when done or stopped, Fix native
CS being unset, enable interrupts in accordance with DMA xfer mode,
discard static DW DMA slave structures, discard unused void priv pointer
and dma_width member of the dw_spi structure, provide the DMA Tx/Rx burst
length parametrisation and make sure it's optionally set in accordance
with the DMA max-burst capability.
In order to have the DW APB SSI MMIO driver working with DMA we need to
initialize the paddr field with the physical base address of the DW APB SSI
registers space. Then we unpin the Intel MID specific code from the
generic DMA one and placed it into the spi-dw-pci.c driver, which is a
better place for it anyway. After that the naming cleanups are performed
since the code is going to be used for a generic DMAC device. Finally the
Generic DMA initialization can be added to the generic version of the
DW APB SSI IP.
Last but not least we traditionally convert the legacy plain text-based
dt-binding file with yaml-based one and as a cherry on a cake replace
the manually written DebugFS registers read method with a ready-to-use
for the same purpose regset32 DebugFS interface usage.
This patchset is rebased and tested on the spi/for-next (5.7-rc5):
base-commit: fe9fce6b2cf3 ("Merge remote-tracking branch 'spi/for-5.8' into spi-next")
Link: https://lore.kernel.org/linux-spi/[email protected]/
Changelog v2:
- Rebase on top of the spi repository for-next branch.
- Move bindings conversion patch to the tail of the series.
- Move fixes to the head of the series.
- Apply as many changes as possible to be applied the Generic DMA
functionality support is added and the spi-dw-mid is moved to the
spi-dw-dma driver.
- Discard patch "spi: dw: Fix dma_slave_config used partly uninitialized"
since the problem has already been fixed.
- Add new patch "spi: dw: Discard unused void priv pointer".
- Add new patch "spi: dw: Discard dma_width member of the dw_spi structure".
n_bytes member of the DW SPI data can be used instead.
- Build the DMA functionality into the DW APB SSI core if required instead
of creating a separate kernel module.
- Use conditional statement instead of the ternary operator in the ref
clock getter.
Link: https://lore.kernel.org/linux-spi/[email protected]/
Changelog v3:
- Use spi_delay_exec() method to wait for the DMA operation completion.
- Explicitly initialize the dw_dma_slave members on stack.
- Discard the dws->fifo_len utilization in the Tx FIFO DMA threshold
setting from the patch where we just add the default burst length
constants.
- Use min() method to calculate the optimal burst values.
- Add new patch which moves the spi-dw.c source file to spi-dw-core.c in
order to preserve the DW APB SSI core driver name.
- Add commas in the debugfs_reg32 structure initializer and after the last
entry of the dw_spi_dbgfs_regs array.
Link: https://lore.kernel.org/linux-spi/[email protected]
Changelog v4:
- Get back ndelay() method to wait for an SPI transfer completion.
spi_delay_exec() isn't suitable for the atomic context.
Link: https://lore.kernel.org/linux-spi/[email protected]
Changelog v5:
- Refactor the Tx/Rx DMA-based SPI transfers wait methods.
- Add a new patch "spi: dw: Set xfer effective_speed_hz".
- Add a new patch "spi: dw: Return any value retrieved from the
dma_transfer callback" as a preparation patch before implementing
the local DMA, Tx SPI and Rx SPI transfers wait methods.
- Add a new patch "spi: dw: Locally wait for the DMA transactions
completion", which provides a local DMA transaction complete
method
- Create a dedicated patch which adds the Rx-done wait method:
"spi: dw: Add SPI Rx-done wait method to DMA-based transfer".
- Add more detailed description of the problems the Tx/Rx-wait
methods-related patches fix.
- Wait for the SPI Tx and Rx transfers being finished in the
mid_spi_dma_transfer() method executed in the task context.
- Use spi_delay_exec() to wait for the SPI Tx/Rx completion, since now
the driver calls the wait methods in the kernel thread context.
- Use SPI_DELAY_UNIT_SCK spi_delay unit for Tx-wait delay, since SPI
xfer's are now have the effective_speed_hz initialized.
- Rx-wait for a delay correlated with the APB/SSI synchronous clock
rate instead of using the SPI bus clock rate.
Co-developed-by: Georgy Vlasov <[email protected]>
Signed-off-by: Georgy Vlasov <[email protected]>
Co-developed-by: Ramil Zaripov <[email protected]>
Signed-off-by: Ramil Zaripov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Maxim Kaurkin <[email protected]>
Cc: Pavel Parkhomenko <[email protected]>
Cc: Ekaterina Skachko <[email protected]>
Cc: Vadim Vlasov <[email protected]>
Cc: Alexey Kolotnikov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Andy Shevchenko <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Serge Semin (16):
spi: dw: Set xfer effective_speed_hz
spi: dw: Return any value retrieved from the dma_transfer callback
spi: dw: Locally wait for the DMA transactions completion
spi: dw: Add SPI Tx-done wait method to DMA-based transfer
spi: dw: Add SPI Rx-done wait method to DMA-based transfer
spi: dw: Parameterize the DMA Rx/Tx burst length
spi: dw: Use DMA max burst to set the request thresholds
spi: dw: Fix Rx-only DMA transfers
spi: dw: Add core suffix to the DW APB SSI core source file
spi: dw: Move Non-DMA code to the DW PCIe-SPI driver
spi: dw: Remove DW DMA code dependency from DW_DMAC_PCI
spi: dw: Add DW SPI DMA/PCI/MMIO dependency on the DW SPI core
spi: dw: Cleanup generic DW DMA code namings
spi: dw: Add DMA support to the DW SPI MMIO driver
spi: dw: Use regset32 DebugFS method to create regdump file
dt-bindings: spi: Convert DW SPI binding to DT schema
.../bindings/spi/snps,dw-apb-ssi.txt | 44 --
.../bindings/spi/snps,dw-apb-ssi.yaml | 127 +++++
.../devicetree/bindings/spi/spi-dw.txt | 24 -
drivers/spi/Kconfig | 15 +-
drivers/spi/Makefile | 5 +-
drivers/spi/{spi-dw.c => spi-dw-core.c} | 95 ++--
drivers/spi/spi-dw-dma.c | 482 ++++++++++++++++++
drivers/spi/spi-dw-mid.c | 382 --------------
drivers/spi/spi-dw-mmio.c | 4 +
drivers/spi/spi-dw-pci.c | 50 +-
drivers/spi/spi-dw.h | 20 +-
11 files changed, 719 insertions(+), 529 deletions(-)
delete mode 100644 Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.txt
create mode 100644 Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml
delete mode 100644 Documentation/devicetree/bindings/spi/spi-dw.txt
rename drivers/spi/{spi-dw.c => spi-dw-core.c} (82%)
create mode 100644 drivers/spi/spi-dw-dma.c
delete mode 100644 drivers/spi/spi-dw-mid.c
--
2.26.2
It isn't good to have numeric literals in the code especially if there
are multiple of them and they are related. Let's replace the Tx and Rx
burst level literals with the corresponding constants.
Co-developed-by: Georgy Vlasov <[email protected]>
Signed-off-by: Georgy Vlasov <[email protected]>
Co-developed-by: Ramil Zaripov <[email protected]>
Signed-off-by: Ramil Zaripov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
Changelog v3:
- Discard the dws->fifo_len utilization in the Tx FIFO DMA threshold
setting.
---
drivers/spi/spi-dw-mid.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/drivers/spi/spi-dw-mid.c b/drivers/spi/spi-dw-mid.c
index 4345881ebf66..93463bdba0f8 100644
--- a/drivers/spi/spi-dw-mid.c
+++ b/drivers/spi/spi-dw-mid.c
@@ -21,7 +21,9 @@
#define WAIT_RETRIES 5
#define RX_BUSY 0
+#define RX_BURST_LEVEL 16
#define TX_BUSY 1
+#define TX_BURST_LEVEL 16
static bool mid_spi_dma_chan_filter(struct dma_chan *chan, void *param)
{
@@ -227,7 +229,7 @@ static struct dma_async_tx_descriptor *dw_spi_dma_prepare_tx(struct dw_spi *dws,
memset(&txconf, 0, sizeof(txconf));
txconf.direction = DMA_MEM_TO_DEV;
txconf.dst_addr = dws->dma_addr;
- txconf.dst_maxburst = 16;
+ txconf.dst_maxburst = TX_BURST_LEVEL;
txconf.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
txconf.dst_addr_width = convert_dma_width(dws->n_bytes);
txconf.device_fc = false;
@@ -319,7 +321,7 @@ static struct dma_async_tx_descriptor *dw_spi_dma_prepare_rx(struct dw_spi *dws,
memset(&rxconf, 0, sizeof(rxconf));
rxconf.direction = DMA_DEV_TO_MEM;
rxconf.src_addr = dws->dma_addr;
- rxconf.src_maxburst = 16;
+ rxconf.src_maxburst = RX_BURST_LEVEL;
rxconf.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
rxconf.src_addr_width = convert_dma_width(dws->n_bytes);
rxconf.device_fc = false;
@@ -344,8 +346,8 @@ static int mid_spi_dma_setup(struct dw_spi *dws, struct spi_transfer *xfer)
{
u16 imr = 0, dma_ctrl = 0;
- dw_writel(dws, DW_SPI_DMARDLR, 0xf);
- dw_writel(dws, DW_SPI_DMATDLR, 0x10);
+ dw_writel(dws, DW_SPI_DMARDLR, RX_BURST_LEVEL - 1);
+ dw_writel(dws, DW_SPI_DMATDLR, TX_BURST_LEVEL);
if (xfer->tx_buf) {
dma_ctrl |= SPI_DMA_TDMAE;
--
2.26.2
Generic DMA support is going to be part of the DW APB SSI core object.
In order to preserve the kernel loadable module name as spi-dw.ko, let's
add the "-core" suffix to the object with generic DW APB SSI code and
build it into the target spi-dw.ko driver.
Signed-off-by: Serge Semin <[email protected]>
Suggested-by: Andy Shevchenko <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Georgy Vlasov <[email protected]>
Cc: Ramil Zaripov <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
Changelog v3:
- This is a new patch added as a result of the discussion with Andy
Shevchenko.
---
drivers/spi/Makefile | 1 +
drivers/spi/{spi-dw.c => spi-dw-core.c} | 0
2 files changed, 1 insertion(+)
rename drivers/spi/{spi-dw.c => spi-dw-core.c} (100%)
diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile
index 28f601327f8c..70ebc2a62e5f 100644
--- a/drivers/spi/Makefile
+++ b/drivers/spi/Makefile
@@ -36,6 +36,7 @@ obj-$(CONFIG_SPI_COLDFIRE_QSPI) += spi-coldfire-qspi.o
obj-$(CONFIG_SPI_DAVINCI) += spi-davinci.o
obj-$(CONFIG_SPI_DLN2) += spi-dln2.o
obj-$(CONFIG_SPI_DESIGNWARE) += spi-dw.o
+spi-dw-y := spi-dw-core.o
obj-$(CONFIG_SPI_DW_MMIO) += spi-dw-mmio.o
obj-$(CONFIG_SPI_DW_PCI) += spi-dw-midpci.o
spi-dw-midpci-objs := spi-dw-pci.o spi-dw-mid.o
diff --git a/drivers/spi/spi-dw.c b/drivers/spi/spi-dw-core.c
similarity index 100%
rename from drivers/spi/spi-dw.c
rename to drivers/spi/spi-dw-core.c
--
2.26.2
Even if DMA transactions are finished it doesn't mean that the SPI
transfers are also completed. It's specifically concerns the Tx-only
SPI transfers, since there might be data left in the SPI Tx FIFO after
the DMA engine notifies that the Tx DMA procedure is done. In order to
completely fix the problem first the driver has to wait for the DMA
transaction completion, then for the corresponding SPI operations to be
finished. In this commit we implement the former part of the solution.
Note we can't just move the SPI operations wait procedure to the DMA
completion callbacks, since these callbacks might be executed in the
tasklet context (and they will be in case of the DW DMA). In case of
slow SPI bus it can cause significant system performance drop.
Signed-off-by: Serge Semin <[email protected]>
Cc: Georgy Vlasov <[email protected]>
Cc: Ramil Zaripov <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Andy Shevchenko <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
drivers/spi/spi-dw-mid.c | 44 ++++++++++++++++++++++++++++++++++++----
drivers/spi/spi-dw.h | 2 ++
2 files changed, 42 insertions(+), 4 deletions(-)
diff --git a/drivers/spi/spi-dw-mid.c b/drivers/spi/spi-dw-mid.c
index 7ff1acaa55f8..355b641c4483 100644
--- a/drivers/spi/spi-dw-mid.c
+++ b/drivers/spi/spi-dw-mid.c
@@ -11,9 +11,11 @@
#include "spi-dw.h"
#ifdef CONFIG_SPI_DW_MID_DMA
+#include <linux/completion.h>
#include <linux/dma-mapping.h>
#include <linux/dmaengine.h>
#include <linux/irqreturn.h>
+#include <linux/jiffies.h>
#include <linux/pci.h>
#include <linux/platform_data/dma-dw.h>
@@ -66,6 +68,8 @@ static int mid_spi_dma_init_mfld(struct device *dev, struct dw_spi *dws)
dws->master->dma_rx = dws->rxchan;
dws->master->dma_tx = dws->txchan;
+ init_completion(&dws->dma_completion);
+
return 0;
free_rxchan:
@@ -91,6 +95,8 @@ static int mid_spi_dma_init_generic(struct device *dev, struct dw_spi *dws)
dws->master->dma_rx = dws->rxchan;
dws->master->dma_tx = dws->txchan;
+ init_completion(&dws->dma_completion);
+
return 0;
}
@@ -121,7 +127,7 @@ static irqreturn_t dma_transfer(struct dw_spi *dws)
dev_err(&dws->master->dev, "%s: FIFO overrun/underrun\n", __func__);
dws->master->cur_msg->status = -EIO;
- spi_finalize_current_transfer(dws->master);
+ complete(&dws->dma_completion);
return IRQ_HANDLED;
}
@@ -142,6 +148,29 @@ static enum dma_slave_buswidth convert_dma_width(u8 n_bytes) {
return DMA_SLAVE_BUSWIDTH_UNDEFINED;
}
+static int dw_spi_dma_wait(struct dw_spi *dws, struct spi_transfer *xfer)
+{
+ unsigned long long ms;
+
+ ms = xfer->len * MSEC_PER_SEC * BITS_PER_BYTE;
+ do_div(ms, xfer->effective_speed_hz);
+ ms += ms + 200;
+
+ if (ms > UINT_MAX)
+ ms = UINT_MAX;
+
+ ms = wait_for_completion_timeout(&dws->dma_completion,
+ msecs_to_jiffies(ms));
+
+ if (ms == 0) {
+ dev_err(&dws->master->cur_msg->spi->dev,
+ "DMA transaction timed out\n");
+ return -ETIMEDOUT;
+ }
+
+ return 0;
+}
+
/*
* dws->dma_chan_busy is set before the dma transfer starts, callback for tx
* channel will clear a corresponding bit.
@@ -155,7 +184,7 @@ static void dw_spi_dma_tx_done(void *arg)
return;
dw_writel(dws, DW_SPI_DMACR, 0);
- spi_finalize_current_transfer(dws->master);
+ complete(&dws->dma_completion);
}
static struct dma_async_tx_descriptor *dw_spi_dma_prepare_tx(struct dw_spi *dws,
@@ -204,7 +233,7 @@ static void dw_spi_dma_rx_done(void *arg)
return;
dw_writel(dws, DW_SPI_DMACR, 0);
- spi_finalize_current_transfer(dws->master);
+ complete(&dws->dma_completion);
}
static struct dma_async_tx_descriptor *dw_spi_dma_prepare_rx(struct dw_spi *dws,
@@ -260,6 +289,8 @@ static int mid_spi_dma_setup(struct dw_spi *dws, struct spi_transfer *xfer)
/* Set the interrupt mask */
spi_umask_intr(dws, imr);
+ reinit_completion(&dws->dma_completion);
+
dws->transfer_handler = dma_transfer;
return 0;
@@ -268,6 +299,7 @@ static int mid_spi_dma_setup(struct dw_spi *dws, struct spi_transfer *xfer)
static int mid_spi_dma_transfer(struct dw_spi *dws, struct spi_transfer *xfer)
{
struct dma_async_tx_descriptor *txdesc, *rxdesc;
+ int ret;
/* Prepare the TX dma transfer */
txdesc = dw_spi_dma_prepare_tx(dws, xfer);
@@ -288,7 +320,11 @@ static int mid_spi_dma_transfer(struct dw_spi *dws, struct spi_transfer *xfer)
dma_async_issue_pending(dws->txchan);
}
- return 1;
+ ret = dw_spi_dma_wait(dws, xfer);
+ if (ret)
+ return ret;
+
+ return 0;
}
static void mid_spi_dma_stop(struct dw_spi *dws)
diff --git a/drivers/spi/spi-dw.h b/drivers/spi/spi-dw.h
index 79782e93eb12..9585d0c83a6d 100644
--- a/drivers/spi/spi-dw.h
+++ b/drivers/spi/spi-dw.h
@@ -2,6 +2,7 @@
#ifndef DW_SPI_HEADER_H
#define DW_SPI_HEADER_H
+#include <linux/completion.h>
#include <linux/irqreturn.h>
#include <linux/io.h>
#include <linux/scatterlist.h>
@@ -145,6 +146,7 @@ struct dw_spi {
unsigned long dma_chan_busy;
dma_addr_t dma_addr; /* phy address of the Data register */
const struct dw_spi_dma_ops *dma_ops;
+ struct completion dma_completion;
#ifdef CONFIG_DEBUG_FS
struct dentry *debugfs;
--
2.26.2
Since there is a generic method available to initialize the DW SPI DMA
interface on any DT and ACPI-based platforms, which in general can be
designed with not only DW DMAC but with any DMA engine on board, we can
freely remove the CONFIG_DW_DMAC_PCI config from dependency list of
CONFIG_SPI_DW_DMA. Especially seeing that we don't use anything DW DMAC
specific in the new driver.
Co-developed-by: Georgy Vlasov <[email protected]>
Signed-off-by: Georgy Vlasov <[email protected]>
Co-developed-by: Ramil Zaripov <[email protected]>
Signed-off-by: Ramil Zaripov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
drivers/spi/Kconfig | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig
index 03b061975f70..6a84f3dad35c 100644
--- a/drivers/spi/Kconfig
+++ b/drivers/spi/Kconfig
@@ -228,7 +228,7 @@ config SPI_DESIGNWARE
config SPI_DW_DMA
bool "DMA support for DW SPI controller"
- depends on SPI_DESIGNWARE && DW_DMAC_PCI
+ depends on SPI_DESIGNWARE
config SPI_DW_PCI
tristate "PCI interface driver for DW SPI core"
--
2.26.2
Each channel of DMA controller may have a limited length of burst
transaction (number of IO operations performed at ones in a single
DMA client request). This parameter can be used to setup the most
optimal DMA Tx/Rx data level values. In order to avoid the Tx buffer
overrun we can set the DMA Tx level to be of FIFO depth minus the
maximum burst transactions length. To prevent the Rx buffer underflow
the DMA Rx level should be set to the maximum burst transactions length.
This commit setups the DMA channels and the DW SPI DMA Tx/Rx levels
in accordance with these rules.
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
Changelog v3:
- Use min() method to calculate the optimal burst values.
---
drivers/spi/spi-dw-mid.c | 37 +++++++++++++++++++++++++++++++++----
drivers/spi/spi-dw.h | 2 ++
2 files changed, 35 insertions(+), 4 deletions(-)
diff --git a/drivers/spi/spi-dw-mid.c b/drivers/spi/spi-dw-mid.c
index 93463bdba0f8..ff79b4239d68 100644
--- a/drivers/spi/spi-dw-mid.c
+++ b/drivers/spi/spi-dw-mid.c
@@ -36,6 +36,31 @@ static bool mid_spi_dma_chan_filter(struct dma_chan *chan, void *param)
return true;
}
+static void mid_spi_maxburst_init(struct dw_spi *dws)
+{
+ struct dma_slave_caps caps;
+ u32 max_burst, def_burst;
+ int ret;
+
+ def_burst = dws->fifo_len / 2;
+
+ ret = dma_get_slave_caps(dws->rxchan, &caps);
+ if (!ret && caps.max_burst)
+ max_burst = caps.max_burst;
+ else
+ max_burst = RX_BURST_LEVEL;
+
+ dws->rxburst = min(max_burst, def_burst);
+
+ ret = dma_get_slave_caps(dws->txchan, &caps);
+ if (!ret && caps.max_burst)
+ max_burst = caps.max_burst;
+ else
+ max_burst = TX_BURST_LEVEL;
+
+ dws->txburst = min(max_burst, def_burst);
+}
+
static int mid_spi_dma_init_mfld(struct device *dev, struct dw_spi *dws)
{
struct dw_dma_slave slave = {
@@ -73,6 +98,8 @@ static int mid_spi_dma_init_mfld(struct device *dev, struct dw_spi *dws)
init_completion(&dws->dma_completion);
+ mid_spi_maxburst_init(dws);
+
return 0;
free_rxchan:
@@ -100,6 +127,8 @@ static int mid_spi_dma_init_generic(struct device *dev, struct dw_spi *dws)
init_completion(&dws->dma_completion);
+ mid_spi_maxburst_init(dws);
+
return 0;
}
@@ -229,7 +258,7 @@ static struct dma_async_tx_descriptor *dw_spi_dma_prepare_tx(struct dw_spi *dws,
memset(&txconf, 0, sizeof(txconf));
txconf.direction = DMA_MEM_TO_DEV;
txconf.dst_addr = dws->dma_addr;
- txconf.dst_maxburst = TX_BURST_LEVEL;
+ txconf.dst_maxburst = dws->txburst;
txconf.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
txconf.dst_addr_width = convert_dma_width(dws->n_bytes);
txconf.device_fc = false;
@@ -321,7 +350,7 @@ static struct dma_async_tx_descriptor *dw_spi_dma_prepare_rx(struct dw_spi *dws,
memset(&rxconf, 0, sizeof(rxconf));
rxconf.direction = DMA_DEV_TO_MEM;
rxconf.src_addr = dws->dma_addr;
- rxconf.src_maxburst = RX_BURST_LEVEL;
+ rxconf.src_maxburst = dws->rxburst;
rxconf.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
rxconf.src_addr_width = convert_dma_width(dws->n_bytes);
rxconf.device_fc = false;
@@ -346,8 +375,8 @@ static int mid_spi_dma_setup(struct dw_spi *dws, struct spi_transfer *xfer)
{
u16 imr = 0, dma_ctrl = 0;
- dw_writel(dws, DW_SPI_DMARDLR, RX_BURST_LEVEL - 1);
- dw_writel(dws, DW_SPI_DMATDLR, TX_BURST_LEVEL);
+ dw_writel(dws, DW_SPI_DMARDLR, dws->rxburst - 1);
+ dw_writel(dws, DW_SPI_DMATDLR, dws->fifo_len - dws->txburst);
if (xfer->tx_buf) {
dma_ctrl |= SPI_DMA_TDMAE;
diff --git a/drivers/spi/spi-dw.h b/drivers/spi/spi-dw.h
index 9585d0c83a6d..9247670fcdfb 100644
--- a/drivers/spi/spi-dw.h
+++ b/drivers/spi/spi-dw.h
@@ -142,7 +142,9 @@ struct dw_spi {
/* DMA info */
struct dma_chan *txchan;
+ u32 txburst;
struct dma_chan *rxchan;
+ u32 rxburst;
unsigned long dma_chan_busy;
dma_addr_t dma_addr; /* phy address of the Data register */
const struct dw_spi_dma_ops *dma_ops;
--
2.26.2
Modern device tree bindings are supposed to be created as YAML-files
in accordance with dt-schema. This commit replaces two DW SPI legacy
bare text bindings with YAML file. As before the bindings file states
that the corresponding dts node is supposed to be compatible either
with generic DW APB SSI controller or with Microsemi/Amazon/Renesas/Intel
vendors-specific controllers, to have registers, interrupts and clocks
properties. Though in case of Microsemi version of the controller
there must be two registers resources specified. Properties like
clock-names, reg-io-width, cs-gpio, num-cs, DMA and slave device
sub-nodes are optional.
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Rob Herring <[email protected]>
Cc: Georgy Vlasov <[email protected]>
Cc: Ramil Zaripov <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Andy Shevchenko <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: [email protected]
---
.../bindings/spi/snps,dw-apb-ssi.txt | 44 ------
.../bindings/spi/snps,dw-apb-ssi.yaml | 127 ++++++++++++++++++
.../devicetree/bindings/spi/spi-dw.txt | 24 ----
3 files changed, 127 insertions(+), 68 deletions(-)
delete mode 100644 Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.txt
create mode 100644 Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml
delete mode 100644 Documentation/devicetree/bindings/spi/spi-dw.txt
diff --git a/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.txt b/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.txt
deleted file mode 100644
index 020e3168ee41..000000000000
--- a/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.txt
+++ /dev/null
@@ -1,44 +0,0 @@
-Synopsys DesignWare AMBA 2.0 Synchronous Serial Interface.
-
-Required properties:
-- compatible : "snps,dw-apb-ssi" or "mscc,<soc>-spi", where soc is "ocelot" or
- "jaguar2", or "amazon,alpine-dw-apb-ssi", or "snps,dwc-ssi-1.01a" or
- "intel,keembay-ssi"
-- reg : The register base for the controller. For "mscc,<soc>-spi", a second
- register set is required (named ICPU_CFG:SPI_MST)
-- interrupts : One interrupt, used by the controller.
-- #address-cells : <1>, as required by generic SPI binding.
-- #size-cells : <0>, also as required by generic SPI binding.
-- clocks : phandles for the clocks, see the description of clock-names below.
- The phandle for the "ssi_clk" is required. The phandle for the "pclk" clock
- is optional. If a single clock is specified but no clock-name, it is the
- "ssi_clk" clock. If both clocks are listed, the "ssi_clk" must be first.
-
-Optional properties:
-- clock-names : Contains the names of the clocks:
- "ssi_clk", for the core clock used to generate the external SPI clock.
- "pclk", the interface clock, required for register access. If a clock domain
- used to enable this clock then it should be named "pclk_clkdomain".
-- cs-gpios : Specifies the gpio pins to be used for chipselects.
-- num-cs : The number of chipselects. If omitted, this will default to 4.
-- reg-io-width : The I/O register width (in bytes) implemented by this
- device. Supported values are 2 or 4 (the default).
-- dmas : Phandle + identifiers of Tx and Rx DMA channels.
-- dma-names : Contains the names of the DMA channels. Must be "tx" and "rx".
-
-Child nodes as per the generic SPI binding.
-
-Example:
-
- spi@fff00000 {
- compatible = "snps,dw-apb-ssi";
- reg = <0xfff00000 0x1000>;
- interrupts = <0 154 4>;
- #address-cells = <1>;
- #size-cells = <0>;
- clocks = <&spi_m_clk>;
- num-cs = <2>;
- cs-gpios = <&gpio0 13 0>,
- <&gpio0 14 0>;
- };
-
diff --git a/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml b/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml
new file mode 100644
index 000000000000..1fcab6415136
--- /dev/null
+++ b/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml
@@ -0,0 +1,127 @@
+# SPDX-License-Identifier: GPL-2.0-only
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/spi/snps,dw-apb-ssi.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Synopsys DesignWare AMBA 2.0 Synchronous Serial Interface
+
+maintainers:
+ - Mark Brown <[email protected]>
+
+allOf:
+ - $ref: "spi-controller.yaml#"
+ - if:
+ properties:
+ compatible:
+ contains:
+ enum:
+ - mscc,ocelot-spi
+ - mscc,jaguar2-spi
+ then:
+ properties:
+ reg:
+ minItems: 2
+
+properties:
+ compatible:
+ oneOf:
+ - description: Generic DW SPI Controller
+ enum:
+ - snps,dw-apb-ssi
+ - snps,dwc-ssi-1.01a
+ - description: Microsemi Ocelot/Jaguar2 SoC SPI Controller
+ items:
+ - enum:
+ - mscc,ocelot-spi
+ - mscc,jaguar2-spi
+ - const: snps,dw-apb-ssi
+ - description: Amazon Alpine SPI Controller
+ const: amazon,alpine-dw-apb-ssi
+ - description: Renesas RZ/N1 SPI Controller
+ items:
+ - const: renesas,rzn1-spi
+ - const: snps,dw-apb-ssi
+ - description: Intel Keem Bay SPI Controller
+ const: intel,keembay-ssi
+
+ reg:
+ minItems: 1
+ items:
+ - description: DW APB SSI controller memory mapped registers
+ - description: SPI MST region map
+
+ interrupts:
+ maxItems: 1
+
+ clocks:
+ minItems: 1
+ items:
+ - description: SPI Controller reference clock source
+ - description: APB interface clock source
+
+ clock-names:
+ minItems: 1
+ items:
+ - const: ssi_clk
+ - const: pclk
+
+ reg-io-width:
+ $ref: /schemas/types.yaml#/definitions/uint32
+ description: I/O register width (in bytes) implemented by this device
+ default: 4
+ enum: [ 2, 4 ]
+
+ num-cs:
+ default: 4
+ minimum: 1
+ maximum: 4
+
+ dmas:
+ items:
+ - description: TX DMA Channel
+ - description: RX DMA Channel
+
+ dma-names:
+ items:
+ - const: tx
+ - const: rx
+
+patternProperties:
+ "^.*@[0-9a-f]+$":
+ type: object
+ properties:
+ reg:
+ minimum: 0
+ maximum: 3
+
+ spi-rx-bus-width:
+ const: 1
+
+ spi-tx-bus-width:
+ const: 1
+
+unevaluatedProperties: false
+
+required:
+ - compatible
+ - reg
+ - "#address-cells"
+ - "#size-cells"
+ - interrupts
+ - clocks
+
+examples:
+ - |
+ spi@fff00000 {
+ compatible = "snps,dw-apb-ssi";
+ reg = <0xfff00000 0x1000>;
+ #address-cells = <1>;
+ #size-cells = <0>;
+ interrupts = <0 154 4>;
+ clocks = <&spi_m_clk>;
+ num-cs = <2>;
+ cs-gpios = <&gpio0 13 0>,
+ <&gpio0 14 0>;
+ };
+...
diff --git a/Documentation/devicetree/bindings/spi/spi-dw.txt b/Documentation/devicetree/bindings/spi/spi-dw.txt
deleted file mode 100644
index 7b63ed601990..000000000000
--- a/Documentation/devicetree/bindings/spi/spi-dw.txt
+++ /dev/null
@@ -1,24 +0,0 @@
-Synopsys DesignWare SPI master
-
-Required properties:
-- compatible: should be "snps,designware-spi"
-- #address-cells: see spi-bus.txt
-- #size-cells: see spi-bus.txt
-- reg: address and length of the spi master registers
-- interrupts: should contain one interrupt
-- clocks: spi clock phandle
-- num-cs: see spi-bus.txt
-
-Optional properties:
-- cs-gpios: see spi-bus.txt
-
-Example:
-
-spi: spi@4020a000 {
- compatible = "snps,designware-spi";
- interrupts = <11 1>;
- reg = <0x4020a000 0x1000>;
- clocks = <&pclk>;
- num-cs = <2>;
- cs-gpios = <&banka 0 0>;
-};
--
2.26.2
DebugFS kernel interface provides a dedicated method to create the
registers dump file. Use it instead of creating a generic DebugFS
file with manually written read callback function.
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Georgy Vlasov <[email protected]>
Cc: Ramil Zaripov <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
Changelog v3:
- Add commas in the debugfs_reg32 structure initializer and after the last
item of the array dw_spi_dbgfs_regs.
---
drivers/spi/spi-dw-core.c | 86 ++++++++++++---------------------------
drivers/spi/spi-dw.h | 2 +
2 files changed, 28 insertions(+), 60 deletions(-)
diff --git a/drivers/spi/spi-dw-core.c b/drivers/spi/spi-dw-core.c
index 4d1849699a12..323c66c5db50 100644
--- a/drivers/spi/spi-dw-core.c
+++ b/drivers/spi/spi-dw-core.c
@@ -29,66 +29,29 @@ struct chip_data {
};
#ifdef CONFIG_DEBUG_FS
-#define SPI_REGS_BUFSIZE 1024
-static ssize_t dw_spi_show_regs(struct file *file, char __user *user_buf,
- size_t count, loff_t *ppos)
-{
- struct dw_spi *dws = file->private_data;
- char *buf;
- u32 len = 0;
- ssize_t ret;
-
- buf = kzalloc(SPI_REGS_BUFSIZE, GFP_KERNEL);
- if (!buf)
- return 0;
-
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "%s registers:\n", dev_name(&dws->master->dev));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "=================================\n");
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "CTRLR0: \t0x%08x\n", dw_readl(dws, DW_SPI_CTRLR0));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "CTRLR1: \t0x%08x\n", dw_readl(dws, DW_SPI_CTRLR1));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "SSIENR: \t0x%08x\n", dw_readl(dws, DW_SPI_SSIENR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "SER: \t\t0x%08x\n", dw_readl(dws, DW_SPI_SER));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "BAUDR: \t\t0x%08x\n", dw_readl(dws, DW_SPI_BAUDR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "TXFTLR: \t0x%08x\n", dw_readl(dws, DW_SPI_TXFTLR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "RXFTLR: \t0x%08x\n", dw_readl(dws, DW_SPI_RXFTLR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "TXFLR: \t\t0x%08x\n", dw_readl(dws, DW_SPI_TXFLR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "RXFLR: \t\t0x%08x\n", dw_readl(dws, DW_SPI_RXFLR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "SR: \t\t0x%08x\n", dw_readl(dws, DW_SPI_SR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "IMR: \t\t0x%08x\n", dw_readl(dws, DW_SPI_IMR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "ISR: \t\t0x%08x\n", dw_readl(dws, DW_SPI_ISR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "DMACR: \t\t0x%08x\n", dw_readl(dws, DW_SPI_DMACR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "DMATDLR: \t0x%08x\n", dw_readl(dws, DW_SPI_DMATDLR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "DMARDLR: \t0x%08x\n", dw_readl(dws, DW_SPI_DMARDLR));
- len += scnprintf(buf + len, SPI_REGS_BUFSIZE - len,
- "=================================\n");
-
- ret = simple_read_from_buffer(user_buf, count, ppos, buf, len);
- kfree(buf);
- return ret;
+
+#define DW_SPI_DBGFS_REG(_name, _off) \
+{ \
+ .name = _name, \
+ .offset = _off, \
}
-static const struct file_operations dw_spi_regs_ops = {
- .owner = THIS_MODULE,
- .open = simple_open,
- .read = dw_spi_show_regs,
- .llseek = default_llseek,
+static const struct debugfs_reg32 dw_spi_dbgfs_regs[] = {
+ DW_SPI_DBGFS_REG("CTRLR0", DW_SPI_CTRLR0),
+ DW_SPI_DBGFS_REG("CTRLR1", DW_SPI_CTRLR1),
+ DW_SPI_DBGFS_REG("SSIENR", DW_SPI_SSIENR),
+ DW_SPI_DBGFS_REG("SER", DW_SPI_SER),
+ DW_SPI_DBGFS_REG("BAUDR", DW_SPI_BAUDR),
+ DW_SPI_DBGFS_REG("TXFTLR", DW_SPI_TXFTLR),
+ DW_SPI_DBGFS_REG("RXFTLR", DW_SPI_RXFTLR),
+ DW_SPI_DBGFS_REG("TXFLR", DW_SPI_TXFLR),
+ DW_SPI_DBGFS_REG("RXFLR", DW_SPI_RXFLR),
+ DW_SPI_DBGFS_REG("SR", DW_SPI_SR),
+ DW_SPI_DBGFS_REG("IMR", DW_SPI_IMR),
+ DW_SPI_DBGFS_REG("ISR", DW_SPI_ISR),
+ DW_SPI_DBGFS_REG("DMACR", DW_SPI_DMACR),
+ DW_SPI_DBGFS_REG("DMATDLR", DW_SPI_DMATDLR),
+ DW_SPI_DBGFS_REG("DMARDLR", DW_SPI_DMARDLR),
};
static int dw_spi_debugfs_init(struct dw_spi *dws)
@@ -100,8 +63,11 @@ static int dw_spi_debugfs_init(struct dw_spi *dws)
if (!dws->debugfs)
return -ENOMEM;
- debugfs_create_file("registers", S_IFREG | S_IRUGO,
- dws->debugfs, (void *)dws, &dw_spi_regs_ops);
+ dws->regset.regs = dw_spi_dbgfs_regs;
+ dws->regset.nregs = ARRAY_SIZE(dw_spi_dbgfs_regs);
+ dws->regset.base = dws->regs;
+ debugfs_create_regset32("registers", 0400, dws->debugfs, &dws->regset);
+
return 0;
}
diff --git a/drivers/spi/spi-dw.h b/drivers/spi/spi-dw.h
index 0b2cd7994513..151ba316619e 100644
--- a/drivers/spi/spi-dw.h
+++ b/drivers/spi/spi-dw.h
@@ -3,6 +3,7 @@
#define DW_SPI_HEADER_H
#include <linux/completion.h>
+#include <linux/debugfs.h>
#include <linux/irqreturn.h>
#include <linux/io.h>
#include <linux/scatterlist.h>
@@ -152,6 +153,7 @@ struct dw_spi {
#ifdef CONFIG_DEBUG_FS
struct dentry *debugfs;
+ struct debugfs_regset32 regset;
#endif
};
--
2.26.2
Since from now the former Intel MID platform layer is used as a generic
DW SPI DMA module, let's alter the internal methods naming to be
DMA-related instead of having the "mid_" prefix.
Co-developed-by: Georgy Vlasov <[email protected]>
Signed-off-by: Georgy Vlasov <[email protected]>
Co-developed-by: Ramil Zaripov <[email protected]>
Signed-off-by: Ramil Zaripov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
Changelog v2:
- Leave the DMA setup method suffixes to be mfld and generic.
---
drivers/spi/spi-dw-dma.c | 85 ++++++++++++++++++++--------------------
drivers/spi/spi-dw-pci.c | 4 +-
drivers/spi/spi-dw.h | 8 ++--
3 files changed, 49 insertions(+), 48 deletions(-)
diff --git a/drivers/spi/spi-dw-dma.c b/drivers/spi/spi-dw-dma.c
index 30bd9800f2df..69b7051e5323 100644
--- a/drivers/spi/spi-dw-dma.c
+++ b/drivers/spi/spi-dw-dma.c
@@ -23,7 +23,7 @@
#define TX_BUSY 1
#define TX_BURST_LEVEL 16
-static bool mid_spi_dma_chan_filter(struct dma_chan *chan, void *param)
+static bool dw_spi_dma_chan_filter(struct dma_chan *chan, void *param)
{
struct dw_dma_slave *s = param;
@@ -34,7 +34,7 @@ static bool mid_spi_dma_chan_filter(struct dma_chan *chan, void *param)
return true;
}
-static void mid_spi_maxburst_init(struct dw_spi *dws)
+static void dw_spi_dma_maxburst_init(struct dw_spi *dws)
{
struct dma_slave_caps caps;
u32 max_burst, def_burst;
@@ -59,7 +59,7 @@ static void mid_spi_maxburst_init(struct dw_spi *dws)
dws->txburst = min(max_burst, def_burst);
}
-static int mid_spi_dma_init_mfld(struct device *dev, struct dw_spi *dws)
+static int dw_spi_dma_init_mfld(struct device *dev, struct dw_spi *dws)
{
struct dw_dma_slave slave = {
.src_id = 0,
@@ -81,13 +81,13 @@ static int mid_spi_dma_init_mfld(struct device *dev, struct dw_spi *dws)
/* 1. Init rx channel */
slave.dma_dev = &dma_dev->dev;
- dws->rxchan = dma_request_channel(mask, mid_spi_dma_chan_filter, &slave);
+ dws->rxchan = dma_request_channel(mask, dw_spi_dma_chan_filter, &slave);
if (!dws->rxchan)
goto err_exit;
/* 2. Init tx channel */
slave.dst_id = 1;
- dws->txchan = dma_request_channel(mask, mid_spi_dma_chan_filter, &slave);
+ dws->txchan = dma_request_channel(mask, dw_spi_dma_chan_filter, &slave);
if (!dws->txchan)
goto free_rxchan;
@@ -96,7 +96,7 @@ static int mid_spi_dma_init_mfld(struct device *dev, struct dw_spi *dws)
init_completion(&dws->dma_completion);
- mid_spi_maxburst_init(dws);
+ dw_spi_dma_maxburst_init(dws);
return 0;
@@ -107,7 +107,7 @@ static int mid_spi_dma_init_mfld(struct device *dev, struct dw_spi *dws)
return -EBUSY;
}
-static int mid_spi_dma_init_generic(struct device *dev, struct dw_spi *dws)
+static int dw_spi_dma_init_generic(struct device *dev, struct dw_spi *dws)
{
dws->rxchan = dma_request_slave_channel(dev, "rx");
if (!dws->rxchan)
@@ -125,12 +125,12 @@ static int mid_spi_dma_init_generic(struct device *dev, struct dw_spi *dws)
init_completion(&dws->dma_completion);
- mid_spi_maxburst_init(dws);
+ dw_spi_dma_maxburst_init(dws);
return 0;
}
-static void mid_spi_dma_exit(struct dw_spi *dws)
+static void dw_spi_dma_exit(struct dw_spi *dws)
{
if (dws->txchan) {
dmaengine_terminate_sync(dws->txchan);
@@ -145,7 +145,7 @@ static void mid_spi_dma_exit(struct dw_spi *dws)
dw_writel(dws, DW_SPI_DMACR, 0);
}
-static irqreturn_t dma_transfer(struct dw_spi *dws)
+static irqreturn_t dw_spi_dma_transfer_handler(struct dw_spi *dws)
{
u16 irq_status = dw_readl(dws, DW_SPI_ISR);
@@ -161,15 +161,16 @@ static irqreturn_t dma_transfer(struct dw_spi *dws)
return IRQ_HANDLED;
}
-static bool mid_spi_can_dma(struct spi_controller *master,
- struct spi_device *spi, struct spi_transfer *xfer)
+static bool dw_spi_can_dma(struct spi_controller *master,
+ struct spi_device *spi, struct spi_transfer *xfer)
{
struct dw_spi *dws = spi_controller_get_devdata(master);
return xfer->len > dws->fifo_len;
}
-static enum dma_slave_buswidth convert_dma_width(u8 n_bytes) {
+static enum dma_slave_buswidth dw_spi_dma_convert_width(u8 n_bytes)
+{
if (n_bytes == 1)
return DMA_SLAVE_BUSWIDTH_1_BYTE;
else if (n_bytes == 2)
@@ -244,8 +245,8 @@ static void dw_spi_dma_tx_done(void *arg)
complete(&dws->dma_completion);
}
-static struct dma_async_tx_descriptor *dw_spi_dma_prepare_tx(struct dw_spi *dws,
- struct spi_transfer *xfer)
+static struct dma_async_tx_descriptor *
+dw_spi_dma_prepare_tx(struct dw_spi *dws, struct spi_transfer *xfer)
{
struct dma_slave_config txconf;
struct dma_async_tx_descriptor *txdesc;
@@ -258,7 +259,7 @@ static struct dma_async_tx_descriptor *dw_spi_dma_prepare_tx(struct dw_spi *dws,
txconf.dst_addr = dws->dma_addr;
txconf.dst_maxburst = dws->txburst;
txconf.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
- txconf.dst_addr_width = convert_dma_width(dws->n_bytes);
+ txconf.dst_addr_width = dw_spi_dma_convert_width(dws->n_bytes);
txconf.device_fc = false;
dmaengine_slave_config(dws->txchan, &txconf);
@@ -350,7 +351,7 @@ static struct dma_async_tx_descriptor *dw_spi_dma_prepare_rx(struct dw_spi *dws,
rxconf.src_addr = dws->dma_addr;
rxconf.src_maxburst = dws->rxburst;
rxconf.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
- rxconf.src_addr_width = convert_dma_width(dws->n_bytes);
+ rxconf.src_addr_width = dw_spi_dma_convert_width(dws->n_bytes);
rxconf.device_fc = false;
dmaengine_slave_config(dws->rxchan, &rxconf);
@@ -369,7 +370,7 @@ static struct dma_async_tx_descriptor *dw_spi_dma_prepare_rx(struct dw_spi *dws,
return rxdesc;
}
-static int mid_spi_dma_setup(struct dw_spi *dws, struct spi_transfer *xfer)
+static int dw_spi_dma_setup(struct dw_spi *dws, struct spi_transfer *xfer)
{
u16 imr = 0, dma_ctrl = 0;
@@ -391,12 +392,12 @@ static int mid_spi_dma_setup(struct dw_spi *dws, struct spi_transfer *xfer)
reinit_completion(&dws->dma_completion);
- dws->transfer_handler = dma_transfer;
+ dws->transfer_handler = dw_spi_dma_transfer_handler;
return 0;
}
-static int mid_spi_dma_transfer(struct dw_spi *dws, struct spi_transfer *xfer)
+static int dw_spi_dma_transfer(struct dw_spi *dws, struct spi_transfer *xfer)
{
struct dma_async_tx_descriptor *txdesc, *rxdesc;
int ret;
@@ -436,7 +437,7 @@ static int mid_spi_dma_transfer(struct dw_spi *dws, struct spi_transfer *xfer)
return ret;
}
-static void mid_spi_dma_stop(struct dw_spi *dws)
+static void dw_spi_dma_stop(struct dw_spi *dws)
{
if (test_bit(TX_BUSY, &dws->dma_chan_busy)) {
dmaengine_terminate_sync(dws->txchan);
@@ -450,32 +451,32 @@ static void mid_spi_dma_stop(struct dw_spi *dws)
dw_writel(dws, DW_SPI_DMACR, 0);
}
-static const struct dw_spi_dma_ops mfld_dma_ops = {
- .dma_init = mid_spi_dma_init_mfld,
- .dma_exit = mid_spi_dma_exit,
- .dma_setup = mid_spi_dma_setup,
- .can_dma = mid_spi_can_dma,
- .dma_transfer = mid_spi_dma_transfer,
- .dma_stop = mid_spi_dma_stop,
+static const struct dw_spi_dma_ops dw_spi_dma_mfld_ops = {
+ .dma_init = dw_spi_dma_init_mfld,
+ .dma_exit = dw_spi_dma_exit,
+ .dma_setup = dw_spi_dma_setup,
+ .can_dma = dw_spi_can_dma,
+ .dma_transfer = dw_spi_dma_transfer,
+ .dma_stop = dw_spi_dma_stop,
};
-void dw_spi_mid_setup_dma_mfld(struct dw_spi *dws)
+void dw_spi_dma_setup_mfld(struct dw_spi *dws)
{
- dws->dma_ops = &mfld_dma_ops;
+ dws->dma_ops = &dw_spi_dma_mfld_ops;
}
-EXPORT_SYMBOL_GPL(dw_spi_mid_setup_dma_mfld);
-
-static const struct dw_spi_dma_ops generic_dma_ops = {
- .dma_init = mid_spi_dma_init_generic,
- .dma_exit = mid_spi_dma_exit,
- .dma_setup = mid_spi_dma_setup,
- .can_dma = mid_spi_can_dma,
- .dma_transfer = mid_spi_dma_transfer,
- .dma_stop = mid_spi_dma_stop,
+EXPORT_SYMBOL_GPL(dw_spi_dma_setup_mfld);
+
+static const struct dw_spi_dma_ops dw_spi_dma_generic_ops = {
+ .dma_init = dw_spi_dma_init_generic,
+ .dma_exit = dw_spi_dma_exit,
+ .dma_setup = dw_spi_dma_setup,
+ .can_dma = dw_spi_can_dma,
+ .dma_transfer = dw_spi_dma_transfer,
+ .dma_stop = dw_spi_dma_stop,
};
-void dw_spi_mid_setup_dma_generic(struct dw_spi *dws)
+void dw_spi_dma_setup_generic(struct dw_spi *dws)
{
- dws->dma_ops = &generic_dma_ops;
+ dws->dma_ops = &dw_spi_dma_generic_ops;
}
-EXPORT_SYMBOL_GPL(dw_spi_mid_setup_dma_generic);
+EXPORT_SYMBOL_GPL(dw_spi_dma_setup_generic);
diff --git a/drivers/spi/spi-dw-pci.c b/drivers/spi/spi-dw-pci.c
index c13707b8493e..2ea73809ca34 100644
--- a/drivers/spi/spi-dw-pci.c
+++ b/drivers/spi/spi-dw-pci.c
@@ -51,7 +51,7 @@ static int spi_mid_init(struct dw_spi *dws)
/* Register hook to configure CTRLR0 */
dws->update_cr0 = dw_spi_update_cr0;
- dw_spi_mid_setup_dma_mfld(dws);
+ dw_spi_dma_setup_mfld(dws);
return 0;
}
@@ -61,7 +61,7 @@ static int spi_generic_init(struct dw_spi *dws)
/* Register hook to configure CTRLR0 */
dws->update_cr0 = dw_spi_update_cr0;
- dw_spi_mid_setup_dma_generic(dws);
+ dw_spi_dma_setup_generic(dws);
return 0;
}
diff --git a/drivers/spi/spi-dw.h b/drivers/spi/spi-dw.h
index 91608cf12636..0b2cd7994513 100644
--- a/drivers/spi/spi-dw.h
+++ b/drivers/spi/spi-dw.h
@@ -259,13 +259,13 @@ extern u32 dw_spi_update_cr0_v1_01a(struct spi_controller *master,
#ifdef CONFIG_SPI_DW_DMA
-extern void dw_spi_mid_setup_dma_mfld(struct dw_spi *dws);
-extern void dw_spi_mid_setup_dma_generic(struct dw_spi *dws);
+extern void dw_spi_dma_setup_mfld(struct dw_spi *dws);
+extern void dw_spi_dma_setup_generic(struct dw_spi *dws);
#else
-static inline void dw_spi_mid_setup_dma_mfld(struct dw_spi *dws) {}
-static inline void dw_spi_mid_setup_dma_generic(struct dw_spi *dws) {}
+static inline void dw_spi_dma_setup_mfld(struct dw_spi *dws) {}
+static inline void dw_spi_dma_setup_generic(struct dw_spi *dws) {}
#endif /* !CONFIG_SPI_DW_DMA */
--
2.26.2
Seeing all of the DW SPI driver components like DW SPI DMA/PCI/MMIO
depend on the DW SPI core code it's better to use the if-endif
conditional kernel config statement to signify that common dependency.
Co-developed-by: Georgy Vlasov <[email protected]>
Signed-off-by: Georgy Vlasov <[email protected]>
Co-developed-by: Ramil Zaripov <[email protected]>
Signed-off-by: Ramil Zaripov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
drivers/spi/Kconfig | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig
index 6a84f3dad35c..3cdf8310d185 100644
--- a/drivers/spi/Kconfig
+++ b/drivers/spi/Kconfig
@@ -226,17 +226,20 @@ config SPI_DESIGNWARE
help
general driver for SPI controller core from DesignWare
+if SPI_DESIGNWARE
+
config SPI_DW_DMA
bool "DMA support for DW SPI controller"
- depends on SPI_DESIGNWARE
config SPI_DW_PCI
tristate "PCI interface driver for DW SPI core"
- depends on SPI_DESIGNWARE && PCI
+ depends on PCI
config SPI_DW_MMIO
tristate "Memory-mapped io interface driver for DW SPI core"
- depends on SPI_DESIGNWARE
+ depends on HAS_IOMEM
+
+endif
config SPI_DLN2
tristate "Diolan DLN-2 USB SPI adapter"
--
2.26.2
DW APB SSI DMA may need to perform the synchronous operations. In that
case the dma_transfer() callback will return 0 as a marker of the SPI
transfer being finished so the SPI core can proceed with the SPI
message trasnfers pumping procedure. This will be needed to fix the
problem when DMA transactions are finished, but there is still data
left in the SPI Tx/Rx buffers. But for now make dma_transfer to
return 1 as the normal dw_spi_transfer_one() method.
Signed-off-by: Serge Semin <[email protected]>
Cc: Georgy Vlasov <[email protected]>
Cc: Ramil Zaripov <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Andy Shevchenko <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
drivers/spi/spi-dw-mid.c | 2 +-
drivers/spi/spi-dw.c | 7 ++-----
2 files changed, 3 insertions(+), 6 deletions(-)
diff --git a/drivers/spi/spi-dw-mid.c b/drivers/spi/spi-dw-mid.c
index b1710132b7b2..7ff1acaa55f8 100644
--- a/drivers/spi/spi-dw-mid.c
+++ b/drivers/spi/spi-dw-mid.c
@@ -288,7 +288,7 @@ static int mid_spi_dma_transfer(struct dw_spi *dws, struct spi_transfer *xfer)
dma_async_issue_pending(dws->txchan);
}
- return 0;
+ return 1;
}
static void mid_spi_dma_stop(struct dw_spi *dws)
diff --git a/drivers/spi/spi-dw.c b/drivers/spi/spi-dw.c
index 050cb2ea0812..6939e003e3e9 100644
--- a/drivers/spi/spi-dw.c
+++ b/drivers/spi/spi-dw.c
@@ -389,11 +389,8 @@ static int dw_spi_transfer_one(struct spi_controller *master,
spi_enable_chip(dws, 1);
- if (dws->dma_mapped) {
- ret = dws->dma_ops->dma_transfer(dws, transfer);
- if (ret < 0)
- return ret;
- }
+ if (dws->dma_mapped)
+ return dws->dma_ops->dma_transfer(dws, transfer);
return 1;
}
--
2.26.2
Tx-only DMA transfers are working perfectly fine since in this case
the code just ignores the Rx FIFO overflow interrupts. But it turns
out the SPI Rx-only transfers are broken since nothing pushing any
data to the shift registers, so the Rx FIFO is left empty and the
SPI core subsystems just returns a timeout error. Since DW DMAC
driver doesn't support something like cyclic write operations of
a single byte to a device register, the only way to support the
Rx-only SPI transfers is to fake it by using a dummy Tx-buffer.
This is what we intend to fix in this commit by setting the
SPI_CONTROLLER_MUST_TX flag for DMA-capable platform.
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Georgy Vlasov <[email protected]>
Cc: Ramil Zaripov <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
drivers/spi/spi-dw.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/spi/spi-dw.c b/drivers/spi/spi-dw.c
index 6939e003e3e9..4d1849699a12 100644
--- a/drivers/spi/spi-dw.c
+++ b/drivers/spi/spi-dw.c
@@ -515,6 +515,7 @@ int dw_spi_add_host(struct device *dev, struct dw_spi *dws)
dev_warn(dev, "DMA init failed\n");
} else {
master->can_dma = dws->dma_ops->can_dma;
+ master->flags |= SPI_CONTROLLER_MUST_TX;
}
}
--
2.26.2
Having any data left in the Rx FIFO after the DMA engine claimed it has
finished all DMA transactions is an abnormal situation, since the DW SPI
controller driver expects to have all the data being fetched and placed
into the SPI Rx buffer at that moment. In case if this has happened we
assume that DMA engine still may be doing the data fetching, thus we give
it sometime to finish. If after a short period of time the data is still
left in the Rx FIFO, the driver will give up waiting and return an error
indicating that the SPI controller/DMA engine must have hung up or failed
at some point of doing their duties.
Fixes: 7063c0d942a1 ("spi/dw_spi: add DMA support")
Co-developed-by: Georgy Vlasov <[email protected]>
Signed-off-by: Georgy Vlasov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Cc: Ramil Zaripov <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Andy Shevchenko <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
Changelog v5:
- Create a dedicated patch which adds the Rx-done wait method.
- Add more detailed description of the problem the patch fixes.
- Wait for the SPI Rx transfer finish in the mid_spi_dma_transfer() method
executed in the task context.
- Use spi_delay_exec() to wait for the SPI Rx completion, since now the
driver does in the kernel thread context.
- Wait for a delay correlated with the APB/SSI synchronous clock rate
instead of using the SPI bus clock rate.
---
drivers/spi/spi-dw-mid.c | 48 +++++++++++++++++++++++++++++++++++++++-
1 file changed, 47 insertions(+), 1 deletion(-)
diff --git a/drivers/spi/spi-dw-mid.c b/drivers/spi/spi-dw-mid.c
index 846e3db91329..4345881ebf66 100644
--- a/drivers/spi/spi-dw-mid.c
+++ b/drivers/spi/spi-dw-mid.c
@@ -248,6 +248,49 @@ static struct dma_async_tx_descriptor *dw_spi_dma_prepare_tx(struct dw_spi *dws,
return txdesc;
}
+static inline bool dw_spi_dma_rx_busy(struct dw_spi *dws)
+{
+ return !!(dw_readl(dws, DW_SPI_SR) & SR_RF_NOT_EMPT);
+}
+
+static int dw_spi_dma_wait_rx_done(struct dw_spi *dws)
+{
+ int retry = WAIT_RETRIES;
+ struct spi_delay delay;
+ unsigned long ns, us;
+ u32 nents;
+
+ /*
+ * It's unlikely that DMA engine is still doing the data fetching, but
+ * if it's let's give it some reasonable time. The timeout calculation
+ * is based on the synchronous APB/SSI reference clock rate, on a
+ * number of data entries left in the Rx FIFO, times a number of clock
+ * periods normally needed for a single APB read/write transaction
+ * without PREADY signal utilized (which is true for the DW APB SSI
+ * controller).
+ */
+ nents = dw_readl(dws, DW_SPI_RXFLR);
+ ns = NSEC_PER_SEC / dws->max_freq * 4 * nents;
+ if (ns <= NSEC_PER_USEC) {
+ delay.unit = SPI_DELAY_UNIT_NSECS;
+ delay.value = ns;
+ } else {
+ us = DIV_ROUND_UP(ns, NSEC_PER_USEC);
+ delay.unit = SPI_DELAY_UNIT_USECS;
+ delay.value = clamp_val(us, 0, USHRT_MAX);
+ }
+
+ while (dw_spi_dma_rx_busy(dws) && retry--)
+ spi_delay_exec(&delay, NULL);
+
+ if (retry < 0) {
+ dev_err(&dws->master->dev, "Rx hanged up\n");
+ return -EIO;
+ }
+
+ return 0;
+}
+
/*
* dws->dma_chan_busy is set before the dma transfer starts, callback for rx
* channel will clear a corresponding bit.
@@ -358,7 +401,10 @@ static int mid_spi_dma_transfer(struct dw_spi *dws, struct spi_transfer *xfer)
return ret;
}
- return 0;
+ if (rxdesc && dws->master->cur_msg->status == -EINPROGRESS)
+ ret = dw_spi_dma_wait_rx_done(dws);
+
+ return ret;
}
static void mid_spi_dma_stop(struct dw_spi *dws)
--
2.26.2
Seeing DW APB SSI controller doesn't support setting the exactly
requested SPI bus frequency, but only a rounded frequency determined
by means of the odd-numbered half-worded reference clock divider,
it would be good tune the SPI core up and initialize the current
transfer effective_speed_hz. By doing so the core will be able to
execute the xfer-related delays with better accuracy.
Signed-off-by: Serge Semin <[email protected]>
Cc: Georgy Vlasov <[email protected]>
Cc: Ramil Zaripov <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Andy Shevchenko <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
drivers/spi/spi-dw.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/spi/spi-dw.c b/drivers/spi/spi-dw.c
index 9d6904d30104..050cb2ea0812 100644
--- a/drivers/spi/spi-dw.c
+++ b/drivers/spi/spi-dw.c
@@ -352,6 +352,7 @@ static int dw_spi_transfer_one(struct spi_controller *master,
spi_set_clk(dws, chip->clk_div);
}
+ transfer->effective_speed_hz = dws->max_freq / chip->clk_div;
dws->n_bytes = DIV_ROUND_UP(transfer->bits_per_word, BITS_PER_BYTE);
cr0 = dws->update_cr0(master, spi, transfer);
--
2.26.2
This is a preparation patch before adding the DW DMA support into the
DW SPI MMIO driver. We need to unpin the Non-DMA-specific code from the
intended to be generic DW APB SSI DMA code. This isn't that hard,
since the most part of the spi-dw-mid.c driver in fact implements a
generic DMA interface for the DW SPI controller driver. The only Intel
MID specifics concern getting the max frequency from the MRST Clock
Control Unit and fetching the DMA controller channels from
corresponding PCIe DMA controller. Since first one is related with the
SPI interface configuration we moved it' implementation into the
DW PCIe-SPI driver module. After that former spi-dw-mid.c file
can be just renamed to be the DW SPI DMA module optionally compiled in to
the DW APB SSI core driver.
Co-developed-by: Georgy Vlasov <[email protected]>
Signed-off-by: Georgy Vlasov <[email protected]>
Co-developed-by: Ramil Zaripov <[email protected]>
Signed-off-by: Ramil Zaripov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
Changelog v2:
- Compile the DW SPI DMA module into the DW APB SSI core instead of being
a separate driver.
---
drivers/spi/Kconfig | 8 +--
drivers/spi/Makefile | 4 +-
drivers/spi/{spi-dw-mid.c => spi-dw-dma.c} | 66 +++-------------------
drivers/spi/spi-dw-pci.c | 50 +++++++++++++++-
drivers/spi/spi-dw.h | 14 ++++-
5 files changed, 73 insertions(+), 69 deletions(-)
rename drivers/spi/{spi-dw-mid.c => spi-dw-dma.c} (88%)
diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig
index 741b9140992a..03b061975f70 100644
--- a/drivers/spi/Kconfig
+++ b/drivers/spi/Kconfig
@@ -226,14 +226,14 @@ config SPI_DESIGNWARE
help
general driver for SPI controller core from DesignWare
+config SPI_DW_DMA
+ bool "DMA support for DW SPI controller"
+ depends on SPI_DESIGNWARE && DW_DMAC_PCI
+
config SPI_DW_PCI
tristate "PCI interface driver for DW SPI core"
depends on SPI_DESIGNWARE && PCI
-config SPI_DW_MID_DMA
- bool "DMA support for DW SPI controller on Intel MID platform"
- depends on SPI_DW_PCI && DW_DMAC_PCI
-
config SPI_DW_MMIO
tristate "Memory-mapped io interface driver for DW SPI core"
depends on SPI_DESIGNWARE
diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile
index 70ebc2a62e5f..c4aa80085257 100644
--- a/drivers/spi/Makefile
+++ b/drivers/spi/Makefile
@@ -37,9 +37,9 @@ obj-$(CONFIG_SPI_DAVINCI) += spi-davinci.o
obj-$(CONFIG_SPI_DLN2) += spi-dln2.o
obj-$(CONFIG_SPI_DESIGNWARE) += spi-dw.o
spi-dw-y := spi-dw-core.o
+spi-dw-$(CONFIG_SPI_DW_DMA) += spi-dw-dma.o
obj-$(CONFIG_SPI_DW_MMIO) += spi-dw-mmio.o
-obj-$(CONFIG_SPI_DW_PCI) += spi-dw-midpci.o
-spi-dw-midpci-objs := spi-dw-pci.o spi-dw-mid.o
+obj-$(CONFIG_SPI_DW_PCI) += spi-dw-pci.o
obj-$(CONFIG_SPI_EFM32) += spi-efm32.o
obj-$(CONFIG_SPI_EP93XX) += spi-ep93xx.o
obj-$(CONFIG_SPI_FALCON) += spi-falcon.o
diff --git a/drivers/spi/spi-dw-mid.c b/drivers/spi/spi-dw-dma.c
similarity index 88%
rename from drivers/spi/spi-dw-mid.c
rename to drivers/spi/spi-dw-dma.c
index ff79b4239d68..30bd9800f2df 100644
--- a/drivers/spi/spi-dw-mid.c
+++ b/drivers/spi/spi-dw-dma.c
@@ -1,16 +1,10 @@
// SPDX-License-Identifier: GPL-2.0-only
/*
- * Special handling for DW core on Intel MID platform
+ * Special handling for DW DMA core
*
* Copyright (c) 2009, 2014 Intel Corporation.
*/
-#include <linux/spi/spi.h>
-#include <linux/types.h>
-
-#include "spi-dw.h"
-
-#ifdef CONFIG_SPI_DW_MID_DMA
#include <linux/completion.h>
#include <linux/dma-mapping.h>
#include <linux/dmaengine.h>
@@ -18,6 +12,10 @@
#include <linux/jiffies.h>
#include <linux/pci.h>
#include <linux/platform_data/dma-dw.h>
+#include <linux/spi/spi.h>
+#include <linux/types.h>
+
+#include "spi-dw.h"
#define WAIT_RETRIES 5
#define RX_BUSY 0
@@ -461,10 +459,11 @@ static const struct dw_spi_dma_ops mfld_dma_ops = {
.dma_stop = mid_spi_dma_stop,
};
-static void dw_spi_mid_setup_dma_mfld(struct dw_spi *dws)
+void dw_spi_mid_setup_dma_mfld(struct dw_spi *dws)
{
dws->dma_ops = &mfld_dma_ops;
}
+EXPORT_SYMBOL_GPL(dw_spi_mid_setup_dma_mfld);
static const struct dw_spi_dma_ops generic_dma_ops = {
.dma_init = mid_spi_dma_init_generic,
@@ -475,55 +474,8 @@ static const struct dw_spi_dma_ops generic_dma_ops = {
.dma_stop = mid_spi_dma_stop,
};
-static void dw_spi_mid_setup_dma_generic(struct dw_spi *dws)
+void dw_spi_mid_setup_dma_generic(struct dw_spi *dws)
{
dws->dma_ops = &generic_dma_ops;
}
-#else /* CONFIG_SPI_DW_MID_DMA */
-static inline void dw_spi_mid_setup_dma_mfld(struct dw_spi *dws) {}
-static inline void dw_spi_mid_setup_dma_generic(struct dw_spi *dws) {}
-#endif
-
-/* Some specific info for SPI0 controller on Intel MID */
-
-/* HW info for MRST Clk Control Unit, 32b reg per controller */
-#define MRST_SPI_CLK_BASE 100000000 /* 100m */
-#define MRST_CLK_SPI_REG 0xff11d86c
-#define CLK_SPI_BDIV_OFFSET 0
-#define CLK_SPI_BDIV_MASK 0x00000007
-#define CLK_SPI_CDIV_OFFSET 9
-#define CLK_SPI_CDIV_MASK 0x00000e00
-#define CLK_SPI_DISABLE_OFFSET 8
-
-int dw_spi_mid_init_mfld(struct dw_spi *dws)
-{
- void __iomem *clk_reg;
- u32 clk_cdiv;
-
- clk_reg = ioremap(MRST_CLK_SPI_REG, 16);
- if (!clk_reg)
- return -ENOMEM;
-
- /* Get SPI controller operating freq info */
- clk_cdiv = readl(clk_reg + dws->bus_num * sizeof(u32));
- clk_cdiv &= CLK_SPI_CDIV_MASK;
- clk_cdiv >>= CLK_SPI_CDIV_OFFSET;
- dws->max_freq = MRST_SPI_CLK_BASE / (clk_cdiv + 1);
-
- iounmap(clk_reg);
-
- /* Register hook to configure CTRLR0 */
- dws->update_cr0 = dw_spi_update_cr0;
-
- dw_spi_mid_setup_dma_mfld(dws);
- return 0;
-}
-
-int dw_spi_mid_init_generic(struct dw_spi *dws)
-{
- /* Register hook to configure CTRLR0 */
- dws->update_cr0 = dw_spi_update_cr0;
-
- dw_spi_mid_setup_dma_generic(dws);
- return 0;
-}
+EXPORT_SYMBOL_GPL(dw_spi_mid_setup_dma_generic);
diff --git a/drivers/spi/spi-dw-pci.c b/drivers/spi/spi-dw-pci.c
index dde54a918b5d..c13707b8493e 100644
--- a/drivers/spi/spi-dw-pci.c
+++ b/drivers/spi/spi-dw-pci.c
@@ -15,6 +15,15 @@
#define DRIVER_NAME "dw_spi_pci"
+/* HW info for MRST Clk Control Unit, 32b reg per controller */
+#define MRST_SPI_CLK_BASE 100000000 /* 100m */
+#define MRST_CLK_SPI_REG 0xff11d86c
+#define CLK_SPI_BDIV_OFFSET 0
+#define CLK_SPI_BDIV_MASK 0x00000007
+#define CLK_SPI_CDIV_OFFSET 9
+#define CLK_SPI_CDIV_MASK 0x00000e00
+#define CLK_SPI_DISABLE_OFFSET 8
+
struct spi_pci_desc {
int (*setup)(struct dw_spi *);
u16 num_cs;
@@ -22,20 +31,55 @@ struct spi_pci_desc {
u32 max_freq;
};
+static int spi_mid_init(struct dw_spi *dws)
+{
+ void __iomem *clk_reg;
+ u32 clk_cdiv;
+
+ clk_reg = ioremap(MRST_CLK_SPI_REG, 16);
+ if (!clk_reg)
+ return -ENOMEM;
+
+ /* Get SPI controller operating freq info */
+ clk_cdiv = readl(clk_reg + dws->bus_num * sizeof(u32));
+ clk_cdiv &= CLK_SPI_CDIV_MASK;
+ clk_cdiv >>= CLK_SPI_CDIV_OFFSET;
+ dws->max_freq = MRST_SPI_CLK_BASE / (clk_cdiv + 1);
+
+ iounmap(clk_reg);
+
+ /* Register hook to configure CTRLR0 */
+ dws->update_cr0 = dw_spi_update_cr0;
+
+ dw_spi_mid_setup_dma_mfld(dws);
+
+ return 0;
+}
+
+static int spi_generic_init(struct dw_spi *dws)
+{
+ /* Register hook to configure CTRLR0 */
+ dws->update_cr0 = dw_spi_update_cr0;
+
+ dw_spi_mid_setup_dma_generic(dws);
+
+ return 0;
+}
+
static struct spi_pci_desc spi_pci_mid_desc_1 = {
- .setup = dw_spi_mid_init_mfld,
+ .setup = spi_mid_init,
.num_cs = 5,
.bus_num = 0,
};
static struct spi_pci_desc spi_pci_mid_desc_2 = {
- .setup = dw_spi_mid_init_mfld,
+ .setup = spi_mid_init,
.num_cs = 2,
.bus_num = 1,
};
static struct spi_pci_desc spi_pci_ehl_desc = {
- .setup = dw_spi_mid_init_generic,
+ .setup = spi_generic_init,
.num_cs = 2,
.bus_num = -1,
.max_freq = 100000000,
diff --git a/drivers/spi/spi-dw.h b/drivers/spi/spi-dw.h
index 9247670fcdfb..91608cf12636 100644
--- a/drivers/spi/spi-dw.h
+++ b/drivers/spi/spi-dw.h
@@ -257,8 +257,16 @@ extern u32 dw_spi_update_cr0_v1_01a(struct spi_controller *master,
struct spi_device *spi,
struct spi_transfer *transfer);
-/* platform related setup */
-extern int dw_spi_mid_init_mfld(struct dw_spi *dws);
-extern int dw_spi_mid_init_generic(struct dw_spi *dws);
+#ifdef CONFIG_SPI_DW_DMA
+
+extern void dw_spi_mid_setup_dma_mfld(struct dw_spi *dws);
+extern void dw_spi_mid_setup_dma_generic(struct dw_spi *dws);
+
+#else
+
+static inline void dw_spi_mid_setup_dma_mfld(struct dw_spi *dws) {}
+static inline void dw_spi_mid_setup_dma_generic(struct dw_spi *dws) {}
+
+#endif /* !CONFIG_SPI_DW_DMA */
#endif /* DW_SPI_HEADER_H */
--
2.26.2
Since the common code in the spi-dw-dma.c driver is ready to be used
by the MMIO driver and now provides a method to generically (on any
DT or ACPI-based platforms) retrieve the Tx/Rx DMA channel handlers,
we can use it and a set of the common DW SPI DMA callbacks to enable
DMA at least for generic "snps,dw-apb-ssi" and "snps,dwc-ssi-1.01a"
devices.
Co-developed-by: Georgy Vlasov <[email protected]>
Signed-off-by: Georgy Vlasov <[email protected]>
Co-developed-by: Ramil Zaripov <[email protected]>
Signed-off-by: Ramil Zaripov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Reviewed-by: Andy Shevchenko <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Feng Tang <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
---
drivers/spi/spi-dw-mmio.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/spi/spi-dw-mmio.c b/drivers/spi/spi-dw-mmio.c
index 0894b4c09496..e23d0c53a664 100644
--- a/drivers/spi/spi-dw-mmio.c
+++ b/drivers/spi/spi-dw-mmio.c
@@ -149,6 +149,8 @@ static int dw_spi_dw_apb_init(struct platform_device *pdev,
/* Register hook to configure CTRLR0 */
dwsmmio->dws.update_cr0 = dw_spi_update_cr0;
+ dw_spi_dma_setup_generic(&dwsmmio->dws);
+
return 0;
}
@@ -158,6 +160,8 @@ static int dw_spi_dwc_ssi_init(struct platform_device *pdev,
/* Register hook to configure CTRLR0 */
dwsmmio->dws.update_cr0 = dw_spi_update_cr0_v1_01a;
+ dw_spi_dma_setup_generic(&dwsmmio->dws);
+
return 0;
}
--
2.26.2
On Fri, May 29, 2020 at 7:02 AM Serge Semin
<[email protected]> wrote:
>
> Even if DMA transactions are finished it doesn't mean that the SPI
> transfers are also completed. It's specifically concerns the Tx-only
> SPI transfers, since there might be data left in the SPI Tx FIFO after
> the DMA engine notifies that the Tx DMA procedure is done. In order to
> completely fix the problem first the driver has to wait for the DMA
> transaction completion, then for the corresponding SPI operations to be
> finished. In this commit we implement the former part of the solution.
>
> Note we can't just move the SPI operations wait procedure to the DMA
> completion callbacks, since these callbacks might be executed in the
> tasklet context (and they will be in case of the DW DMA). In case of
> slow SPI bus it can cause significant system performance drop.
I read commit message, I read the code. What's going on here since you
repeated xfer_completion (and its wait routine) from SPI core and I'm
wondering what happened to it? Why we are not calling
spi_finalize_current_transfer()?
...
> dws->master->cur_msg->status = -EIO;
> - spi_finalize_current_transfer(dws->master);
> + complete(&dws->dma_completion);
> return IRQ_HANDLED;
...
> +static int dw_spi_dma_wait(struct dw_spi *dws, struct spi_transfer *xfer)
> +{
> + unsigned long long ms;
> +
> + ms = xfer->len * MSEC_PER_SEC * BITS_PER_BYTE;
> + do_div(ms, xfer->effective_speed_hz);
> + ms += ms + 200;
> +
> + if (ms > UINT_MAX)
> + ms = UINT_MAX;
> +
> + ms = wait_for_completion_timeout(&dws->dma_completion,
> + msecs_to_jiffies(ms));
> +
> + if (ms == 0) {
> + dev_err(&dws->master->cur_msg->spi->dev,
> + "DMA transaction timed out\n");
> + return -ETIMEDOUT;
> + }
> +
> + return 0;
> +}
> +
> /*
> * dws->dma_chan_busy is set before the dma transfer starts, callback for tx
> * channel will clear a corresponding bit.
> @@ -155,7 +184,7 @@ static void dw_spi_dma_tx_done(void *arg)
> return;
>
> dw_writel(dws, DW_SPI_DMACR, 0);
> - spi_finalize_current_transfer(dws->master);
> + complete(&dws->dma_completion);
> }
>
> static struct dma_async_tx_descriptor *dw_spi_dma_prepare_tx(struct dw_spi *dws,
> @@ -204,7 +233,7 @@ static void dw_spi_dma_rx_done(void *arg)
> return;
>
> dw_writel(dws, DW_SPI_DMACR, 0);
> - spi_finalize_current_transfer(dws->master);
> + complete(&dws->dma_completion);
> }
--
With Best Regards,
Andy Shevchenko
On Fri, May 29, 2020 at 10:55:32AM +0300, Andy Shevchenko wrote:
> On Fri, May 29, 2020 at 7:02 AM Serge Semin
> <[email protected]> wrote:
> >
> > Even if DMA transactions are finished it doesn't mean that the SPI
> > transfers are also completed. It's specifically concerns the Tx-only
> > SPI transfers, since there might be data left in the SPI Tx FIFO after
> > the DMA engine notifies that the Tx DMA procedure is done. In order to
> > completely fix the problem first the driver has to wait for the DMA
> > transaction completion, then for the corresponding SPI operations to be
> > finished. In this commit we implement the former part of the solution.
> >
> > Note we can't just move the SPI operations wait procedure to the DMA
> > completion callbacks, since these callbacks might be executed in the
> > tasklet context (and they will be in case of the DW DMA). In case of
> > slow SPI bus it can cause significant system performance drop.
>
> I read commit message, I read the code. What's going on here since you
> repeated xfer_completion (and its wait routine) from SPI core and I'm
> wondering what happened to it? Why we are not calling
> spi_finalize_current_transfer()?
We discussed that in v4. You complained about using ndelay() for slow SPI bus,
which may cause too long atomic context execution. We agreed. Since we can't wait
in the tasklet context and using a dedicated kernel thread for waiting would be too
much, Me and Mark agreed, that even if it causes us of the local wait-function
re-implementation the best approach would be not to use the generic
spi_transfer_wait() method, but instead wait for the DMA transactions locally
in the DMA driver and just return 0 from the transfer_one callback indicating
that the SPI transfer is finished and there is no need for SPI core to wait. As
a lot of DMA-based SPI drivers do.
If you don't understand what the commit message says, just say so. I'll
reformulate it.
-Sergey
>
> ...
>
> > dws->master->cur_msg->status = -EIO;
> > - spi_finalize_current_transfer(dws->master);
> > + complete(&dws->dma_completion);
> > return IRQ_HANDLED;
>
> ...
>
> > +static int dw_spi_dma_wait(struct dw_spi *dws, struct spi_transfer *xfer)
> > +{
> > + unsigned long long ms;
> > +
> > + ms = xfer->len * MSEC_PER_SEC * BITS_PER_BYTE;
> > + do_div(ms, xfer->effective_speed_hz);
> > + ms += ms + 200;
> > +
> > + if (ms > UINT_MAX)
> > + ms = UINT_MAX;
> > +
> > + ms = wait_for_completion_timeout(&dws->dma_completion,
> > + msecs_to_jiffies(ms));
> > +
> > + if (ms == 0) {
> > + dev_err(&dws->master->cur_msg->spi->dev,
> > + "DMA transaction timed out\n");
> > + return -ETIMEDOUT;
> > + }
> > +
> > + return 0;
> > +}
> > +
> > /*
> > * dws->dma_chan_busy is set before the dma transfer starts, callback for tx
> > * channel will clear a corresponding bit.
> > @@ -155,7 +184,7 @@ static void dw_spi_dma_tx_done(void *arg)
> > return;
> >
> > dw_writel(dws, DW_SPI_DMACR, 0);
> > - spi_finalize_current_transfer(dws->master);
> > + complete(&dws->dma_completion);
> > }
> >
> > static struct dma_async_tx_descriptor *dw_spi_dma_prepare_tx(struct dw_spi *dws,
> > @@ -204,7 +233,7 @@ static void dw_spi_dma_rx_done(void *arg)
> > return;
> >
> > dw_writel(dws, DW_SPI_DMACR, 0);
> > - spi_finalize_current_transfer(dws->master);
> > + complete(&dws->dma_completion);
> > }
>
>
> --
> With Best Regards,
> Andy Shevchenko
Hello!
On 29.05.2020 6:58, Serge Semin wrote:
> Seeing DW APB SSI controller doesn't support setting the exactly
> requested SPI bus frequency, but only a rounded frequency determined
> by means of the odd-numbered half-worded reference clock divider,
> it would be good tune the SPI core up and initialize the current
^ to?
> transfer effective_speed_hz. By doing so the core will be able to
> execute the xfer-related delays with better accuracy.
>
> Signed-off-by: Serge Semin <[email protected]>
> Cc: Georgy Vlasov <[email protected]>
> Cc: Ramil Zaripov <[email protected]>
> Cc: Alexey Malahov <[email protected]>
> Cc: Thomas Bogendoerfer <[email protected]>
> Cc: Arnd Bergmann <[email protected]>
> Cc: Feng Tang <[email protected]>
> Cc: Andy Shevchenko <[email protected]>
> Cc: Rob Herring <[email protected]>
> Cc: [email protected]
> Cc: [email protected]
[...]
MBR, Sergei
On Fri, May 29, 2020 at 11:12:04AM +0300, Serge Semin wrote:
> On Fri, May 29, 2020 at 10:55:32AM +0300, Andy Shevchenko wrote:
> > On Fri, May 29, 2020 at 7:02 AM Serge Semin
> > <[email protected]> wrote:
> > >
> > > Even if DMA transactions are finished it doesn't mean that the SPI
> > > transfers are also completed. It's specifically concerns the Tx-only
> > > SPI transfers, since there might be data left in the SPI Tx FIFO after
> > > the DMA engine notifies that the Tx DMA procedure is done. In order to
> > > completely fix the problem first the driver has to wait for the DMA
> > > transaction completion, then for the corresponding SPI operations to be
> > > finished. In this commit we implement the former part of the solution.
> > >
> > > Note we can't just move the SPI operations wait procedure to the DMA
> > > completion callbacks, since these callbacks might be executed in the
> > > tasklet context (and they will be in case of the DW DMA). In case of
> > > slow SPI bus it can cause significant system performance drop.
> >
>
> > I read commit message, I read the code. What's going on here since you
> > repeated xfer_completion (and its wait routine) from SPI core and I'm
> > wondering what happened to it? Why we are not calling
> > spi_finalize_current_transfer()?
>
> We discussed that in v4. You complained about using ndelay() for slow SPI bus,
> which may cause too long atomic context execution. We agreed. Since we can't wait
> in the tasklet context and using a dedicated kernel thread for waiting would be too
> much, Me and Mark agreed, that
> even if it causes us of the local wait-function
> re-implementation the best approach would be not to use the generic
> spi_transfer_wait() method, but instead wait for the DMA transactions locally
> in the DMA driver and just return 0 from the transfer_one callback indicating
> that the SPI transfer is finished and there is no need for SPI core to wait. As
> a lot of DMA-based SPI drivers do.
The above is missed in the commit message.
> If you don't understand what the commit message says, just say so. I'll
> reformulate it.
See above. A bit of elaboration would be good. Thank you!
--
With Best Regards,
Andy Shevchenko
On Fri, May 29, 2020 at 06:59:03AM +0300, Serge Semin wrote:
> Having any data left in the Rx FIFO after the DMA engine claimed it has
> finished all DMA transactions is an abnormal situation, since the DW SPI
> controller driver expects to have all the data being fetched and placed
> into the SPI Rx buffer at that moment. In case if this has happened we
> assume that DMA engine still may be doing the data fetching, thus we give
> it sometime to finish. If after a short period of time the data is still
> left in the Rx FIFO, the driver will give up waiting and return an error
> indicating that the SPI controller/DMA engine must have hung up or failed
> at some point of doing their duties.
...
> +static int dw_spi_dma_wait_rx_done(struct dw_spi *dws)
> +{
> + int retry = WAIT_RETRIES;
> + struct spi_delay delay;
> + unsigned long ns, us;
> + u32 nents;
> +
> + /*
> + * It's unlikely that DMA engine is still doing the data fetching, but
> + * if it's let's give it some reasonable time. The timeout calculation
> + * is based on the synchronous APB/SSI reference clock rate, on a
> + * number of data entries left in the Rx FIFO, times a number of clock
> + * periods normally needed for a single APB read/write transaction
> + * without PREADY signal utilized (which is true for the DW APB SSI
> + * controller).
> + */
> + nents = dw_readl(dws, DW_SPI_RXFLR);
> + ns = NSEC_PER_SEC / dws->max_freq * 4 * nents;
I think we may slightly increase precision by writing this like
ns = 4 * NSEC_PER_SEC / dws->max_freq * nents;
> + if (ns <= NSEC_PER_USEC) {
> + delay.unit = SPI_DELAY_UNIT_NSECS;
> + delay.value = ns;
> + } else {
> + us = DIV_ROUND_UP(ns, NSEC_PER_USEC);
> + delay.unit = SPI_DELAY_UNIT_USECS;
> + delay.value = clamp_val(us, 0, USHRT_MAX);
> + }
> +
> + while (dw_spi_dma_rx_busy(dws) && retry--)
> + spi_delay_exec(&delay, NULL);
> +
> + if (retry < 0) {
> + dev_err(&dws->master->dev, "Rx hanged up\n");
> + return -EIO;
> + }
> +
> + return 0;
> +}
--
With Best Regards,
Andy Shevchenko
On Fri, May 29, 2020 at 11:49:12AM +0300, Sergei Shtylyov wrote:
> Hello!
>
> On 29.05.2020 6:58, Serge Semin wrote:
>
> > Seeing DW APB SSI controller doesn't support setting the exactly
> > requested SPI bus frequency, but only a rounded frequency determined
> > by means of the odd-numbered half-worded reference clock divider,
> > it would be good tune the SPI core up and initialize the current
> ^ to?
Thanks! I'll fix it in the next patchset version.
-Sergey
>
> > transfer effective_speed_hz. By doing so the core will be able to
> > execute the xfer-related delays with better accuracy.
> >
> > Signed-off-by: Serge Semin <[email protected]>
> > Cc: Georgy Vlasov <[email protected]>
> > Cc: Ramil Zaripov <[email protected]>
> > Cc: Alexey Malahov <[email protected]>
> > Cc: Thomas Bogendoerfer <[email protected]>
> > Cc: Arnd Bergmann <[email protected]>
> > Cc: Feng Tang <[email protected]>
> > Cc: Andy Shevchenko <[email protected]>
> > Cc: Rob Herring <[email protected]>
> > Cc: [email protected]
> > Cc: [email protected]
> [...]
>
> MBR, Sergei
On Fri, May 29, 2020 at 12:26:10PM +0300, Andy Shevchenko wrote:
> On Fri, May 29, 2020 at 11:12:04AM +0300, Serge Semin wrote:
> > On Fri, May 29, 2020 at 10:55:32AM +0300, Andy Shevchenko wrote:
> > > On Fri, May 29, 2020 at 7:02 AM Serge Semin
> > > <[email protected]> wrote:
> > > >
> > > > Even if DMA transactions are finished it doesn't mean that the SPI
> > > > transfers are also completed. It's specifically concerns the Tx-only
> > > > SPI transfers, since there might be data left in the SPI Tx FIFO after
> > > > the DMA engine notifies that the Tx DMA procedure is done. In order to
> > > > completely fix the problem first the driver has to wait for the DMA
> > > > transaction completion, then for the corresponding SPI operations to be
> > > > finished. In this commit we implement the former part of the solution.
> > > >
> > > > Note we can't just move the SPI operations wait procedure to the DMA
> > > > completion callbacks, since these callbacks might be executed in the
> > > > tasklet context (and they will be in case of the DW DMA). In case of
> > > > slow SPI bus it can cause significant system performance drop.
> > >
> >
> > > I read commit message, I read the code. What's going on here since you
> > > repeated xfer_completion (and its wait routine) from SPI core and I'm
> > > wondering what happened to it? Why we are not calling
> > > spi_finalize_current_transfer()?
> >
> > We discussed that in v4. You complained about using ndelay() for slow SPI bus,
> > which may cause too long atomic context execution. We agreed. Since we can't wait
> > in the tasklet context and using a dedicated kernel thread for waiting would be too
> > much, Me and Mark agreed, that
>
> > even if it causes us of the local wait-function
> > re-implementation the best approach would be not to use the generic
> > spi_transfer_wait() method, but instead wait for the DMA transactions locally
> > in the DMA driver and just return 0 from the transfer_one callback indicating
> > that the SPI transfer is finished and there is no need for SPI core to wait. As
> > a lot of DMA-based SPI drivers do.
>
> The above is missed in the commit message.
>
> > If you don't understand what the commit message says, just say so. I'll
> > reformulate it.
>
> See above. A bit of elaboration would be good. Thank you!
Agreed. I'll create a more detailed commit description, which will have the
info you cited.
-Sergey
>
> --
> With Best Regards,
> Andy Shevchenko
>
>
On Fri, May 29, 2020 at 12:46:48PM +0300, Andy Shevchenko wrote:
> On Fri, May 29, 2020 at 06:59:03AM +0300, Serge Semin wrote:
> > Having any data left in the Rx FIFO after the DMA engine claimed it has
> > finished all DMA transactions is an abnormal situation, since the DW SPI
> > controller driver expects to have all the data being fetched and placed
> > into the SPI Rx buffer at that moment. In case if this has happened we
> > assume that DMA engine still may be doing the data fetching, thus we give
> > it sometime to finish. If after a short period of time the data is still
> > left in the Rx FIFO, the driver will give up waiting and return an error
> > indicating that the SPI controller/DMA engine must have hung up or failed
> > at some point of doing their duties.
>
> ...
>
> > +static int dw_spi_dma_wait_rx_done(struct dw_spi *dws)
> > +{
> > + int retry = WAIT_RETRIES;
> > + struct spi_delay delay;
> > + unsigned long ns, us;
> > + u32 nents;
> > +
> > + /*
> > + * It's unlikely that DMA engine is still doing the data fetching, but
> > + * if it's let's give it some reasonable time. The timeout calculation
> > + * is based on the synchronous APB/SSI reference clock rate, on a
> > + * number of data entries left in the Rx FIFO, times a number of clock
> > + * periods normally needed for a single APB read/write transaction
> > + * without PREADY signal utilized (which is true for the DW APB SSI
> > + * controller).
> > + */
> > + nents = dw_readl(dws, DW_SPI_RXFLR);
>
> > + ns = NSEC_PER_SEC / dws->max_freq * 4 * nents;
>
> I think we may slightly increase precision by writing this like
>
> ns = 4 * NSEC_PER_SEC / dws->max_freq * nents;
Good point. Although both 4 and NSEC_PER_SEC are signed. The later is
1000000000L. Formally speaking on x32 systems (4 * 1000 000 000L) equals
to a negative value. Though overflow still won't happen so the result will
be correct. Anyway to be on a safe side it would be better to use an explicit
unsigned literal:
+ ns = 4U * NSEC_PER_SEC / dws->max_freq * nents;
-Sergey
>
>
> > + if (ns <= NSEC_PER_USEC) {
> > + delay.unit = SPI_DELAY_UNIT_NSECS;
> > + delay.value = ns;
> > + } else {
> > + us = DIV_ROUND_UP(ns, NSEC_PER_USEC);
> > + delay.unit = SPI_DELAY_UNIT_USECS;
> > + delay.value = clamp_val(us, 0, USHRT_MAX);
> > + }
> > +
> > + while (dw_spi_dma_rx_busy(dws) && retry--)
> > + spi_delay_exec(&delay, NULL);
> > +
> > + if (retry < 0) {
> > + dev_err(&dws->master->dev, "Rx hanged up\n");
> > + return -EIO;
> > + }
> > +
> > + return 0;
> > +}
>
> --
> With Best Regards,
> Andy Shevchenko
>
>
On Fri, May 29, 2020 at 01:13:28PM +0300, Serge Semin wrote:
> On Fri, May 29, 2020 at 12:46:48PM +0300, Andy Shevchenko wrote:
> > On Fri, May 29, 2020 at 06:59:03AM +0300, Serge Semin wrote:
> > > Having any data left in the Rx FIFO after the DMA engine claimed it has
> > > finished all DMA transactions is an abnormal situation, since the DW SPI
> > > controller driver expects to have all the data being fetched and placed
> > > into the SPI Rx buffer at that moment. In case if this has happened we
> > > assume that DMA engine still may be doing the data fetching, thus we give
> > > it sometime to finish. If after a short period of time the data is still
> > > left in the Rx FIFO, the driver will give up waiting and return an error
> > > indicating that the SPI controller/DMA engine must have hung up or failed
> > > at some point of doing their duties.
> >
> > ...
> >
> > > +static int dw_spi_dma_wait_rx_done(struct dw_spi *dws)
> > > +{
> > > + int retry = WAIT_RETRIES;
> > > + struct spi_delay delay;
> > > + unsigned long ns, us;
> > > + u32 nents;
> > > +
> > > + /*
> > > + * It's unlikely that DMA engine is still doing the data fetching, but
> > > + * if it's let's give it some reasonable time. The timeout calculation
> > > + * is based on the synchronous APB/SSI reference clock rate, on a
> > > + * number of data entries left in the Rx FIFO, times a number of clock
> > > + * periods normally needed for a single APB read/write transaction
> > > + * without PREADY signal utilized (which is true for the DW APB SSI
> > > + * controller).
> > > + */
> > > + nents = dw_readl(dws, DW_SPI_RXFLR);
> >
>
> > > + ns = NSEC_PER_SEC / dws->max_freq * 4 * nents;
> >
> > I think we may slightly increase precision by writing this like
> >
> > ns = 4 * NSEC_PER_SEC / dws->max_freq * nents;
>
> Good point. Although both 4 and NSEC_PER_SEC are signed. The later is
> 1000000000L. Formally speaking on x32 systems (4 * 1000 000 000L) equals
> to a negative value. Though overflow still won't happen so the result will
> be correct. Anyway to be on a safe side it would be better to use an explicit
> unsigned literal:
>
> + ns = 4U * NSEC_PER_SEC / dws->max_freq * nents;
Yes, right.
--
With Best Regards,
Andy Shevchenko