TCFQ to XSPI migration for NXP DSPI driver
Subsystems (2)
Vladimir Oltean
maintainercover TCFQ to XSPI migration for NXP DSPI driver
From: Vladimir Oltean <[email protected]>
This series aims to remove the most inefficient transfer method from the NXP DSPI driver.
TCFQ (Transfer Complete Flag) mode works by transferring one word, waiting for its TX confirmation interrupt (or polling on the equivalent status bit), sending the next word, etc, until the buffer is complete.
The issue with this mode is that it's fundamentally incompatible with any sort of batching such as writing to a FIFO. But actually, due to previous patchset ("Compatible string consolidation for NXP DSPI driver"):
https://patchwork.kernel.org/cover/11414593/
all existing users of TCFQ mode today already support a more advanced feature set, in the form of XSPI (extended SPI). XSPI brings 2 extra features:
- Word sizes up to 32 bits. This is sub-utilized today, and acceleration of smaller-than-32 bpw values is provided.
- "Command cycling", basically the ability to write multiple words in a row and receiving an interrupt only after the completion of the last one. This is what enables us to make use of the full FIFO depth of this controller.
Series was tested on the NXP LS1021A-TSN and LS1043A-RDB boards, both functionally as well as from a performance standpoint.
The command used to benchmark the increased throughput was:
spidev_test --device /dev/spidev1.0 --bpw 8 --size 256 --cpha --iter 10000000 --speed 20000000
where spidev1.0 is a dummy spidev node, using a chip select that no peripheral responds to.
On LS1021A, which has a 4-entry-deep FIFO and a less powerful CPU, the performance increase brought by this patchset is from 2700 kbps to 5800 kbps.
On LS1043A, which has a 16-entry-deep FIFO and a more powerful CPU, the performance increases from 4100 kbps to 13700 kbps.
On average, SPI software timestamping is not adversely affected by the extra batching, due to the extra patches.
There is one extra patch which clarifies why the TCFQ users were not converted to the "other" mode in this driver that makes use of the FIFO, which would be EOQ mode.
My request to the many people on CC (known users and/or contributors) is to give this series a test to ensure there are no regressions, and for the Coldfire maintainers to clarify whether the EOQ limitation is acceptable for them in the long run.
Vladimir Oltean (12):
spi: spi-fsl-dspi: Simplify bytes_per_word gymnastics
spi: spi-fsl-dspi: Remove unused chip->void_write_data
spi: spi-fsl-dspi: Don't mask off undefined bits
spi: spi-fsl-dspi: Add comments around dspi_pop_tx and dspi_push_rx
functions
spi: spi-fsl-dspi: Rename fifo_{read,write} and {tx,cmd}_fifo_write
spi: spi-fsl-dspi: Implement .max_message_size method for EOQ mode
spi: Do spi_take_timestamp_pre for as many times as necessary
spi: spi-fsl-dspi: Convert TCFQ users to XSPI FIFO mode
spi: spi-fsl-dspi: Accelerate transfers using larger word size if
possible
spi: spi-fsl-dspi: Optimize dspi_setup_accel for lowest interrupt
count
spi: spi-fsl-dspi: Use EOQ for last word in buffer even for XSPI mode
spi: spi-fsl-dspi: Take software timestamp in dspi_fifo_write
drivers/spi/spi-fsl-dspi.c | 421 ++++++++++++++++++++++++------------- drivers/spi/spi.c | 19 +- include/linux/spi/spi.h | 3 +- 3 files changed, 288 insertions(+), 155 deletions(-)
Signature · 1 line
2.17.1
Vladimir Oltean
maintainerPATCH 1/12 spi: spi-fsl-dspi: Simplify bytes_per_word gymnastics
From: Vladimir Oltean <[email protected]>
Reduce the if-then-else-if-then-else sequence to:
- a simple division in the case of bytes_per_word calculation
- a memcpy command with a variable size. The semantics of larger-than-8 xfer->bits_per_word is that those words are to be interpreted and transmitted in CPU native endianness.
--- drivers/spi/spi-fsl-dspi.c | 21 +++------------------ 1 file changed, 3 insertions(+), 18 deletions(-)
drivers/spi/spi-fsl-dspi.c +3 -18
@@ -252,12 +252,7 @@ static u32 dspi_pop_tx(struct fsl_dspi *dspi)u32 txdata =0 ;if (dspi->tx) { -if (dspi->bytes_per_word ==1 ) - txdata = *(u8 *)dspi->tx; -else if (dspi->bytes_per_word ==2 ) - txdata = *(u16 *)dspi->tx; -else /* dspi->bytes_per_word == 4 */ - txdata = *(u32 *)dspi->tx; +memcpy (&txdata, dspi->tx, dspi->bytes_per_word); dspi->tx += dspi->bytes_per_word; } dspi->len -= dspi->bytes_per_word;
@@ -284,12 +279,7 @@ static void dspi_push_rx(struct fsl_dspi *dspi, u32 rxdata)/* Mask off undefined bits */ rxdata &= (1 << dspi->bits_per_word) -1 ; -if (dspi->bytes_per_word ==1 ) - *(u8 *)dspi->rx = rxdata; -else if (dspi->bytes_per_word ==2 ) - *(u16 *)dspi->rx = rxdata; -else /* dspi->bytes_per_word == 4 */ - *(u32 *)dspi->rx = rxdata; +memcpy (dspi->rx, &rxdata, dspi->bytes_per_word); dspi->rx += dspi->bytes_per_word; }
@@ -814,12 +804,7 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr, dspi->progress =0 ;/* Validated transfer specific frame size (defaults applied) */ dspi->bits_per_word = transfer->bits_per_word; -if (transfer->bits_per_word <=8 ) - dspi->bytes_per_word =1 ; -else if (transfer->bits_per_word <=16 ) - dspi->bytes_per_word =2 ; -else - dspi->bytes_per_word =4 ; + dspi->bytes_per_word =DIV_ROUND_UP (dspi->bits_per_word,8 ); regmap_update_bits(dspi->regmap, SPI_MCR, SPI_MCR_CLR_TXF | SPI_MCR_CLR_RXF,
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Simplify bytes_per_word gymnastics" to the spi tree
The patch
spi: spi-fsl-dspi: Simplify bytes_per_word gymnastics
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
53fadb4d90c762b560a9d0983bb5894129057ea1Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:33 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Simplify bytes_per_word gymnastics
Reduce the if-then-else-if-then-else sequence to:
- a simple division in the case of bytes_per_word calculation
- a memcpy command with a variable size. The semantics of larger-than-8 xfer->bits_per_word is that those words are to be interpreted and transmitted in CPU native endianness.
--- drivers/spi/spi-fsl-dspi.c | 21 +++------------------ 1 file changed, 3 insertions(+), 18 deletions(-)
drivers/spi/spi-fsl-dspi.c +3 -18
@@ -252,12 +252,7 @@ static u32 dspi_pop_tx(struct fsl_dspi *dspi)u32 txdata =0 ;if (dspi->tx) { -if (dspi->bytes_per_word ==1 ) - txdata = *(u8 *)dspi->tx; -else if (dspi->bytes_per_word ==2 ) - txdata = *(u16 *)dspi->tx; -else /* dspi->bytes_per_word == 4 */ - txdata = *(u32 *)dspi->tx; +memcpy (&txdata, dspi->tx, dspi->bytes_per_word); dspi->tx += dspi->bytes_per_word; } dspi->len -= dspi->bytes_per_word;
@@ -284,12 +279,7 @@ static void dspi_push_rx(struct fsl_dspi *dspi, u32 rxdata)/* Mask off undefined bits */ rxdata &= (1 << dspi->bits_per_word) -1 ; -if (dspi->bytes_per_word ==1 ) - *(u8 *)dspi->rx = rxdata; -else if (dspi->bytes_per_word ==2 ) - *(u16 *)dspi->rx = rxdata; -else /* dspi->bytes_per_word == 4 */ - *(u32 *)dspi->rx = rxdata; +memcpy (dspi->rx, &rxdata, dspi->bytes_per_word); dspi->rx += dspi->bytes_per_word; }
@@ -814,12 +804,7 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr, dspi->progress =0 ;/* Validated transfer specific frame size (defaults applied) */ dspi->bits_per_word = transfer->bits_per_word; -if (transfer->bits_per_word <=8 ) - dspi->bytes_per_word =1 ; -else if (transfer->bits_per_word <=16 ) - dspi->bytes_per_word =2 ; -else - dspi->bytes_per_word =4 ; + dspi->bytes_per_word =DIV_ROUND_UP (dspi->bits_per_word,8 ); regmap_update_bits(dspi->regmap, SPI_MCR, SPI_MCR_CLR_TXF | SPI_MCR_CLR_RXF,
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 2/12 spi: spi-fsl-dspi: Remove unused chip->void_write_data
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
This variable has been present since the initial submission of the driver, and held, for some reason, the value of zero, to be sent on the wire in the case there wasn't any TX buffer for the current transfer.
Since quite a while now, however, it isn't doing anything at all.
--- drivers/spi/spi-fsl-dspi.c | 6 ------ 1 file changed, 6 deletions(-)
drivers/spi/spi-fsl-dspi.c +0 -6
@@ -110,7 +110,6 @@struct chip_data {u32 ctar_val; -u16 void_write_data; };enum dspi_trans_mode {
@@ -235,7 +234,6 @@ struct fsl_dspi {const void *tx;void *rx;void *rx_end; -u16 void_write_data;u16 tx_cmd;u8 bits_per_word;u8 bytes_per_word;
@@ -795,8 +793,6 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr, dspi->tx_cmd |=SPI_PUSHR_CMD_CONT ; } - dspi->void_write_data = dspi->cur_chip->void_write_data; - dspi->tx = transfer->tx_buf; dspi->rx = transfer->rx_buf; dspi->rx_end = dspi->rx + transfer->len;
@@ -897,8 +893,6 @@ static int dspi_setup(struct spi_device *spi) sck_cs_delay = pdata->sck_cs_delay; } - chip->void_write_data =0 ; - clkrate =clk_get_rate (dspi->clk);hz_to_spi_baud (&pbr, &br, spi->max_speed_hz, clkrate);
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Remove unused chip->void_write_data" to the spi tree
The patch
spi: spi-fsl-dspi: Remove unused chip->void_write_data
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
6d6af5796e5d9a88ae83c9c753023bba61deb18bMon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:34 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Remove unused chip->void_write_data
This variable has been present since the initial submission of the driver, and held, for some reason, the value of zero, to be sent on the wire in the case there wasn't any TX buffer for the current transfer.
Since quite a while now, however, it isn't doing anything at all.
--- drivers/spi/spi-fsl-dspi.c | 6 ------ 1 file changed, 6 deletions(-)
drivers/spi/spi-fsl-dspi.c +0 -6
@@ -110,7 +110,6 @@struct chip_data {u32 ctar_val; -u16 void_write_data; };enum dspi_trans_mode {
@@ -235,7 +234,6 @@ struct fsl_dspi {const void *tx;void *rx;void *rx_end; -u16 void_write_data;u16 tx_cmd;u8 bits_per_word;u8 bytes_per_word;
@@ -795,8 +793,6 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr, dspi->tx_cmd |=SPI_PUSHR_CMD_CONT ; } - dspi->void_write_data = dspi->cur_chip->void_write_data; - dspi->tx = transfer->tx_buf; dspi->rx = transfer->rx_buf; dspi->rx_end = dspi->rx + transfer->len;
@@ -897,8 +893,6 @@ static int dspi_setup(struct spi_device *spi) sck_cs_delay = pdata->sck_cs_delay; } - chip->void_write_data =0 ; - clkrate =clk_get_rate (dspi->clk);hz_to_spi_baud (&pbr, &br, spi->max_speed_hz, clkrate);
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 3/12 spi: spi-fsl-dspi: Don't mask off undefined bits
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
This is a useless operation, and if the driver needs to do that, there's something deeply wrong going on.
--- drivers/spi/spi-fsl-dspi.c | 3 --- 1 file changed, 3 deletions(-)
drivers/spi/spi-fsl-dspi.c +0 -3
@@ -274,9 +274,6 @@ static void dspi_push_rx(struct fsl_dspi *dspi, u32 rxdata)if (!dspi->rx)return ; -/* Mask off undefined bits */ - rxdata &= (1 << dspi->bits_per_word) -1 ; -memcpy (dspi->rx, &rxdata, dspi->bytes_per_word); dspi->rx += dspi->bytes_per_word; }
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Don't mask off undefined bits" to the spi tree
The patch
spi: spi-fsl-dspi: Don't mask off undefined bits
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
5542bd797190d5d77f1ad3a6df9628f26d117b31Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:35 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Don't mask off undefined bits
This is a useless operation, and if the driver needs to do that, there's something deeply wrong going on.
--- drivers/spi/spi-fsl-dspi.c | 3 --- 1 file changed, 3 deletions(-)
drivers/spi/spi-fsl-dspi.c +0 -3
@@ -274,9 +274,6 @@ static void dspi_push_rx(struct fsl_dspi *dspi, u32 rxdata)if (!dspi->rx)return ; -/* Mask off undefined bits */ - rxdata &= (1 << dspi->bits_per_word) -1 ; -memcpy (dspi->rx, &rxdata, dspi->bytes_per_word); dspi->rx += dspi->bytes_per_word; }
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 4/12 spi: spi-fsl-dspi: Add comments around dspi_pop_tx and dspi_push_rx functions
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
Their names are confusing, since dspi_pop_tx prepares a word to be written to the PUSHR register, and dspi_push_rx gets a word from the POPR register.
--- drivers/spi/spi-fsl-dspi.c | 6 ++++++ 1 file changed, 6 insertions(+)
drivers/spi/spi-fsl-dspi.c +6 -0
@@ -245,6 +245,10 @@ struct fsl_dspi {struct fsl_dspi_dma *dma; }; +/* +* Pop one word from the TX buffer for pushing into the +* PUSHR register (TX FIFO) +*/ static u32 dspi_pop_tx (struct fsl_dspi *dspi) {u32 txdata =0 ;
@@ -257,6 +261,7 @@ static u32 dspi_pop_tx(struct fsl_dspi *dspi)return txdata; } +/* Prepare one TX FIFO entry (txdata plus cmd) */ static u32 dspi_pop_tx_pushr (struct fsl_dspi *dspi) {u16 cmd = dspi->tx_cmd, data =dspi_pop_tx (dspi);
@@ -269,6 +274,7 @@ static u32 dspi_pop_tx_pushr(struct fsl_dspi *dspi)return cmd <<16 | data; } +/* Push one word to the RX buffer from the POPR register (RX FIFO) */ static void dspi_push_rx (struct fsl_dspi *dspi,u32 rxdata) {if (!dspi->rx)
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Add comments around dspi_pop_tx and dspi_push_rx functions" to the spi tree
The patch
spi: spi-fsl-dspi: Add comments around dspi_pop_tx and dspi_push_rx functions
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
8f8303ee05857e1b3084c467dde8bb31c58464bdMon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:36 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Add comments around dspi_pop_tx and
dspi_push_rx functions
Their names are confusing, since dspi_pop_tx prepares a word to be written to the PUSHR register, and dspi_push_rx gets a word from the POPR register.
--- drivers/spi/spi-fsl-dspi.c | 6 ++++++ 1 file changed, 6 insertions(+)
drivers/spi/spi-fsl-dspi.c +6 -0
@@ -245,6 +245,10 @@ struct fsl_dspi {struct fsl_dspi_dma *dma; }; +/* +* Pop one word from the TX buffer for pushing into the +* PUSHR register (TX FIFO) +*/ static u32 dspi_pop_tx (struct fsl_dspi *dspi) {u32 txdata =0 ;
@@ -257,6 +261,7 @@ static u32 dspi_pop_tx(struct fsl_dspi *dspi)return txdata; } +/* Prepare one TX FIFO entry (txdata plus cmd) */ static u32 dspi_pop_tx_pushr (struct fsl_dspi *dspi) {u16 cmd = dspi->tx_cmd, data =dspi_pop_tx (dspi);
@@ -269,6 +274,7 @@ static u32 dspi_pop_tx_pushr(struct fsl_dspi *dspi)return cmd <<16 | data; } +/* Push one word to the RX buffer from the POPR register (RX FIFO) */ static void dspi_push_rx (struct fsl_dspi *dspi,u32 rxdata) {if (!dspi->rx)
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 5/12 spi: spi-fsl-dspi: Rename fifo_{read,write} and {tx,cmd}_fifo_write
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
These function names are very generic and it is easy to get confused. Rename them after the hardware register that they are accessing.
--- drivers/spi/spi-fsl-dspi.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-)
drivers/spi/spi-fsl-dspi.c +11 -11
@@ -601,12 +601,12 @@ static void ns_delay_scale(char *psc, char *sc, int delay_ns, } } -static void fifo_write (struct fsl_dspi *dspi) +static void dspi_pushr_write (struct fsl_dspi *dspi) {regmap_write (dspi->regmap,SPI_PUSHR ,dspi_pop_tx_pushr (dspi)); } -static void cmd_fifo_write (struct fsl_dspi *dspi) +static void dspi_pushr_cmd_write (struct fsl_dspi *dspi) {u16 cmd = dspi->tx_cmd;
@@ -615,7 +615,7 @@ static void cmd_fifo_write(struct fsl_dspi *dspi)regmap_write (dspi->regmap_pushr,PUSHR_CMD , cmd); } -static void tx_fifo_write (struct fsl_dspi *dspi,u16 txdata) +static void dspi_pushr_txdata_write (struct fsl_dspi *dspi,u16 txdata) {regmap_write (dspi->regmap_pushr,PUSHR_TX , txdata); }
@@ -631,18 +631,18 @@ static void dspi_tcfq_write(struct fsl_dspi *dspi)*/ u32 data =dspi_pop_tx (dspi); -cmd_fifo_write (dspi); -tx_fifo_write (dspi, data &0xFFFF ); -tx_fifo_write (dspi, data >>16 ); +dspi_pushr_cmd_write (dspi); +dspi_pushr_txdata_write (dspi, data &0xFFFF ); +dspi_pushr_txdata_write (dspi, data >>16 ); }else {/* Write one entry to both TX FIFO and CMD FIFO * simultaneously. */ -fifo_write (dspi); +dspi_pushr_write (dspi); } } -static u32 fifo_read (struct fsl_dspi *dspi) +static u32 dspi_popr_read (struct fsl_dspi *dspi) {u32 rxdata =0 ;
@@ -652,7 +652,7 @@ static u32 fifo_read(struct fsl_dspi *dspi)static void dspi_tcfq_read (struct fsl_dspi *dspi) { -dspi_push_rx (dspi,fifo_read (dspi)); +dspi_push_rx (dspi,dspi_popr_read (dspi)); }static void dspi_eoq_write (struct fsl_dspi *dspi)
@@ -670,7 +670,7 @@ static void dspi_eoq_write(struct fsl_dspi *dspi)if (fifo_size == (dspi->devtype_data->fifo_size -1 )) dspi->tx_cmd |=SPI_PUSHR_CMD_CTCNT ;/* Write combined TX FIFO and CMD FIFO entry */ -fifo_write (dspi); +dspi_pushr_write (dspi); } }
@@ -680,7 +680,7 @@ static void dspi_eoq_read(struct fsl_dspi *dspi)/* Read one FIFO entry and push to rx buffer */ while ((dspi->rx < dspi->rx_end) && fifo_size--) -dspi_push_rx (dspi,fifo_read (dspi)); +dspi_push_rx (dspi,dspi_popr_read (dspi)); }static int dspi_rxtx (struct fsl_dspi *dspi)
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Rename fifo_{read,write} and {tx,cmd}_fifo_write" to the spi tree
The patch
spi: spi-fsl-dspi: Rename fifo_{read,write} and {tx,cmd}_fifo_write
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
547248fbed23f3cd2f6a5937b44fad60993640c4Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:37 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Rename fifo_{read,write} and
{tx,cmd}_fifo_write
These function names are very generic and it is easy to get confused. Rename them after the hardware register that they are accessing.
--- drivers/spi/spi-fsl-dspi.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-)
drivers/spi/spi-fsl-dspi.c +11 -11
@@ -601,12 +601,12 @@ static void ns_delay_scale(char *psc, char *sc, int delay_ns, } } -static void fifo_write (struct fsl_dspi *dspi) +static void dspi_pushr_write (struct fsl_dspi *dspi) {regmap_write (dspi->regmap,SPI_PUSHR ,dspi_pop_tx_pushr (dspi)); } -static void cmd_fifo_write (struct fsl_dspi *dspi) +static void dspi_pushr_cmd_write (struct fsl_dspi *dspi) {u16 cmd = dspi->tx_cmd;
@@ -615,7 +615,7 @@ static void cmd_fifo_write(struct fsl_dspi *dspi)regmap_write (dspi->regmap_pushr,PUSHR_CMD , cmd); } -static void tx_fifo_write (struct fsl_dspi *dspi,u16 txdata) +static void dspi_pushr_txdata_write (struct fsl_dspi *dspi,u16 txdata) {regmap_write (dspi->regmap_pushr,PUSHR_TX , txdata); }
@@ -631,18 +631,18 @@ static void dspi_tcfq_write(struct fsl_dspi *dspi)*/ u32 data =dspi_pop_tx (dspi); -cmd_fifo_write (dspi); -tx_fifo_write (dspi, data &0xFFFF ); -tx_fifo_write (dspi, data >>16 ); +dspi_pushr_cmd_write (dspi); +dspi_pushr_txdata_write (dspi, data &0xFFFF ); +dspi_pushr_txdata_write (dspi, data >>16 ); }else {/* Write one entry to both TX FIFO and CMD FIFO * simultaneously. */ -fifo_write (dspi); +dspi_pushr_write (dspi); } } -static u32 fifo_read (struct fsl_dspi *dspi) +static u32 dspi_popr_read (struct fsl_dspi *dspi) {u32 rxdata =0 ;
@@ -652,7 +652,7 @@ static u32 fifo_read(struct fsl_dspi *dspi)static void dspi_tcfq_read (struct fsl_dspi *dspi) { -dspi_push_rx (dspi,fifo_read (dspi)); +dspi_push_rx (dspi,dspi_popr_read (dspi)); }static void dspi_eoq_write (struct fsl_dspi *dspi)
@@ -670,7 +670,7 @@ static void dspi_eoq_write(struct fsl_dspi *dspi)if (fifo_size == (dspi->devtype_data->fifo_size -1 )) dspi->tx_cmd |=SPI_PUSHR_CMD_CTCNT ;/* Write combined TX FIFO and CMD FIFO entry */ -fifo_write (dspi); +dspi_pushr_write (dspi); } }
@@ -680,7 +680,7 @@ static void dspi_eoq_read(struct fsl_dspi *dspi)/* Read one FIFO entry and push to rx buffer */ while ((dspi->rx < dspi->rx_end) && fifo_size--) -dspi_push_rx (dspi,fifo_read (dspi)); +dspi_push_rx (dspi,dspi_popr_read (dspi)); }static int dspi_rxtx (struct fsl_dspi *dspi)
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 6/12 spi: spi-fsl-dspi: Implement .max_message_size method for EOQ mode
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
When it gets set, End Of Queue Flag halts the DSPI controller and forces the chip select signal to deassert.
This operating mode is not ideal, but it is used for the DSPI instantiations where there is no other notification from the controller that the data in the FIFO has finished transmission. So in practice, it means that transmitting buffers larger than the FIFO size will yield unpredictable results.
The only controller that operates in EOQ mode is MCF5441X (Coldfire). I would say that the way EOQ is used (and documented in the reference manual, too) on this chip is incorrect, and I would personally migrate it to TCFQ, but that's notably worse in terms of performance (it can only use 1 entry of the 16-deep FIFO) and if this limitation didn't bother any Coldfire DSPI user so far, it's likely that we just need to throw an error for larger buffers to make sure that callers are aware their transfers are getting truncated/split.
--- drivers/spi/spi-fsl-dspi.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)
drivers/spi/spi-fsl-dspi.c +17 -0
@@ -1084,6 +1084,22 @@ static int dspi_slave_abort(struct spi_master *master)return 0 ; } +/* +* EOQ mode will inevitably deassert its PCS signal on last word in a queue +* (hardware limitation), so we need to inform the spi_device that larger +* buffers than the FIFO size are going to have the chip select randomly +* toggling, so it has a chance to adapt its message sizes. +*/ +static size_t dspi_max_message_size (struct spi_device *spi) +{ +struct fsl_dspi *dspi =spi_controller_get_devdata (spi->controller); + +if (dspi->devtype_data->trans_mode ==DSPI_EOQ_MODE ) +return dspi->devtype_data->fifo_size; + +return SIZE_MAX ; +} +static int dspi_probe (struct platform_device *pdev) {struct device_node *np = pdev->dev.of_node;
@@ -1105,6 +1121,7 @@ static int dspi_probe(struct platform_device *pdev) ctlr->setup = dspi_setup; ctlr->transfer_one_message = dspi_transfer_one_message; + ctlr->max_message_size = dspi_max_message_size; ctlr->dev.of_node = pdev->dev.of_node; ctlr->cleanup = dspi_cleanup;
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Implement .max_message_size method for EOQ mode" to the spi tree
The patch
spi: spi-fsl-dspi: Implement .max_message_size method for EOQ mode
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
a3185c38dc6cd664b2b576eb9d3e5d2f49101a10Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:38 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Implement .max_message_size method for EOQ
mode
When it gets set, End Of Queue Flag halts the DSPI controller and forces the chip select signal to deassert.
This operating mode is not ideal, but it is used for the DSPI instantiations where there is no other notification from the controller that the data in the FIFO has finished transmission. So in practice, it means that transmitting buffers larger than the FIFO size will yield unpredictable results.
The only controller that operates in EOQ mode is MCF5441X (Coldfire). I would say that the way EOQ is used (and documented in the reference manual, too) on this chip is incorrect, and I would personally migrate it to TCFQ, but that's notably worse in terms of performance (it can only use 1 entry of the 16-deep FIFO) and if this limitation didn't bother any Coldfire DSPI user so far, it's likely that we just need to throw an error for larger buffers to make sure that callers are aware their transfers are getting truncated/split.
--- drivers/spi/spi-fsl-dspi.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+)
drivers/spi/spi-fsl-dspi.c +17 -0
@@ -1084,6 +1084,22 @@ static int dspi_slave_abort(struct spi_master *master)return 0 ; } +/* +* EOQ mode will inevitably deassert its PCS signal on last word in a queue +* (hardware limitation), so we need to inform the spi_device that larger +* buffers than the FIFO size are going to have the chip select randomly +* toggling, so it has a chance to adapt its message sizes. +*/ +static size_t dspi_max_message_size (struct spi_device *spi) +{ +struct fsl_dspi *dspi =spi_controller_get_devdata (spi->controller); + +if (dspi->devtype_data->trans_mode ==DSPI_EOQ_MODE ) +return dspi->devtype_data->fifo_size; + +return SIZE_MAX ; +} +static int dspi_probe (struct platform_device *pdev) {struct device_node *np = pdev->dev.of_node;
@@ -1105,6 +1121,7 @@ static int dspi_probe(struct platform_device *pdev) ctlr->setup = dspi_setup; ctlr->transfer_one_message = dspi_transfer_one_message; + ctlr->max_message_size = dspi_max_message_size; ctlr->dev.of_node = pdev->dev.of_node; ctlr->cleanup = dspi_cleanup;
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 7/12 spi: Do spi_take_timestamp_pre for as many times as necessary
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
When dealing with a SPI controller driver that is sending more than 1 byte at once (or the entire buffer at once), and the SPI peripheral driver has requested timestamping for a byte in the middle of the buffer, we find that spi_take_timestamp_pre never records a "pre" timestamp.
This happens because the function currently expects to be called with the "progress" argument >= to what the peripheral has requested to be timestamped. But clearly there are cases when that isn't going to fly.
And since we can't change the past when we realize that the opportunity to take a "pre" timestamp has just passed and there isn't going to be another one, the approach taken is to keep recording the "pre" timestamp on each call, overwriting the previously recorded one until the "post" timestamp is also taken.
--- drivers/spi/spi.c | 19 +++++++------------ include/linux/spi/spi.h | 3 +-- 2 files changed, 8 insertions(+), 14 deletions(-)
drivers/spi/spi.c +7 -12
@@ -1515,17 +1515,15 @@ void spi_take_timestamp_pre(struct spi_controller *ctlr,if (!xfer->ptp_sts)return ; -if (xfer->timestamped_pre) +if (xfer->timestamped)return ; -if (progress < xfer->ptp_sts_word_pre) +if (progress > xfer->ptp_sts_word_pre)return ;/* Capture the resolution of the timestamp */ xfer->ptp_sts_word_pre = progress; - xfer->timestamped_pre = true; -if (irqs_off) {local_irq_save (ctlr->irq_flags);preempt_disable ();
@@ -1554,7 +1552,7 @@ void spi_take_timestamp_post(struct spi_controller *ctlr,if (!xfer->ptp_sts)return ; -if (xfer->timestamped_post) +if (xfer->timestamped)return ;if (progress < xfer->ptp_sts_word_post)
@@ -1570,7 +1568,7 @@ void spi_take_timestamp_post(struct spi_controller *ctlr,/* Capture the resolution of the timestamp */ xfer->ptp_sts_word_post = progress; - xfer->timestamped_post = true; + xfer->timestamped = true; }EXPORT_SYMBOL_GPL (spi_take_timestamp_post);
@@ -1675,12 +1673,9 @@ void spi_finalize_current_message(struct spi_controller *ctlr) } } -if (unlikely (ctlr->ptp_sts_supported)) { -list_for_each_entry (xfer, &mesg->transfers, transfer_list) { -WARN_ON_ONCE (xfer->ptp_sts && !xfer->timestamped_pre); -WARN_ON_ONCE (xfer->ptp_sts && !xfer->timestamped_post); - } - } +if (unlikely (ctlr->ptp_sts_supported)) +list_for_each_entry (xfer, &mesg->transfers, transfer_list) +WARN_ON_ONCE (xfer->ptp_sts && !xfer->timestamped);spi_unmap_msg (ctlr, mesg);
include/linux/spi/spi.h +1 -2
@@ -933,8 +933,7 @@ struct spi_transfer {struct ptp_system_timestamp *ptp_sts; -bool timestamped_pre; -bool timestamped_post; +bool timestamped;struct list_head transfer_list; };
Signature · 1 line
2.17.1
Mark Brown
maintainerRe: [PATCH 07/12] spi: Do spi_take_timestamp_pre for as many times as necessary
Quoted 7 lines · Vladimir Oltean
From: Vladimir Oltean <[email protected]>
When dealing with a SPI controller driver that is sending more than 1 byte at once (or the entire buffer at once), and the SPI peripheral driver has requested timestamping for a byte in the middle of the buffer, we find that
spi_take_timestamp_prenever records a "pre" timestamp.
This is a fix and so should have been at the start of the series to make sure there aren't any dependencies.
Vladimir Oltean
maintainerRe: [PATCH 07/12] spi: Do spi_take_timestamp_pre for as many times as necessary
Hi Mark,
Quoted 4 lines · Mark Brown
Quoted 2 lines · Vladimir Oltean
From: Vladimir Oltean <[email protected]>
Quoted 8 lines
Quoted 5 lines
When dealing with a SPI controller driver that is sending more than 1 byte at once (or the entire buffer at once), and the SPI peripheral driver has requested timestamping for a byte in the middle of the buffer, we find that
spi_take_timestamp_prenever records a "pre" timestamp.This is a fix and so should have been at the start of the series to make sure there aren't any dependencies.
My reasoning for not submitting it as a fix is:
- The only driver that uses the functionality so far - spi-fsl-dspi - has worked thus far even with the limitation that only byte-by-byte transfers were supported properly.
- I removed the limitation before actually changing the operating mode of spi-fsl-dspi. Therefore the limitation is effectively never seen.
- New SPI drivers that would want to make use of software timestamping would do so through your SPI for-next branch anyway, where the limitation would be, again, fixed.
Thanks,
-Vladimir
Mark Brown
maintainerRe: [PATCH 07/12] spi: Do spi_take_timestamp_pre for as many times as necessary
Quoted 5 lines · Vladimir Oltean
Quoted 2 lines · Mark Brown
This is a fix and so should have been at the start of the series to make sure there aren't any dependencies.
Quoted 9 lines
My reasoning for not submitting it as a fix is:
- The only driver that uses the functionality so far - spi-fsl-dspi - has worked thus far even with the limitation that only byte-by-byte transfers were supported properly.
- I removed the limitation before actually changing the operating mode of spi-fsl-dspi. Therefore the limitation is effectively never seen.
- New SPI drivers that would want to make use of software timestamping would do so through your SPI for-next branch anyway, where the limitation would be, again, fixed.
That's mostly all true but it's still better to pull fixes like this (or the patch limiting the size) forwards and not have to think if it's safe to not apply them as a fix, it's less effort all round.
Vladimir Oltean
maintainerRe: [PATCH 07/12] spi: Do spi_take_timestamp_pre for as many times as necessary
Quoted 17 lines · Mark Brown
Quoted 5 lines · Vladimir Oltean
Quoted 2 lines · Mark Brown
This is a fix and so should have been at the start of the series to make sure there aren't any dependencies.
Quoted 9 lines
My reasoning for not submitting it as a fix is:
- The only driver that uses the functionality so far - spi-fsl-dspi - has worked thus far even with the limitation that only byte-by-byte transfers were supported properly.
- I removed the limitation before actually changing the operating mode of spi-fsl-dspi. Therefore the limitation is effectively never seen.
- New SPI drivers that would want to make use of software timestamping would do so through your SPI for-next branch anyway, where the limitation would be, again, fixed.
Quoted 3 lines
That's mostly all true but it's still better to pull fixes like this (or the patch limiting the size) forwards and not have to think if it's safe to not apply them as a fix, it's less effort all round.
So do you want me to do something about it now?
Mark Brown
maintainerRe: [PATCH 07/12] spi: Do spi_take_timestamp_pre for as many times as necessary
Quoted 7 lines · Vladimir Oltean
Quoted 3 lines · Mark Brown
That's mostly all true but it's still better to pull fixes like this (or the patch limiting the size) forwards and not have to think if it's safe to not apply them as a fix, it's less effort all round.
So do you want me to do something about it now?
No, it's fine for now but please bear this in mind in future.
Mark Brown
maintainerApplied "spi: Do spi_take_timestamp_pre for as many times as necessary" to the spi tree
In reply to Vladimir OlteanThe patch
spi: Do spi_take_timestamp_pre for as many times as necessary
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
6a726824aaa3adaaf3bcfca3b471408e225f33d6Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:39 +0200
Subject: [PATCH] spi: Do spi_take_timestamp_pre for as many times as necessary
When dealing with a SPI controller driver that is sending more than 1 byte at once (or the entire buffer at once), and the SPI peripheral driver has requested timestamping for a byte in the middle of the buffer, we find that spi_take_timestamp_pre never records a "pre" timestamp.
This happens because the function currently expects to be called with the "progress" argument >= to what the peripheral has requested to be timestamped. But clearly there are cases when that isn't going to fly.
And since we can't change the past when we realize that the opportunity to take a "pre" timestamp has just passed and there isn't going to be another one, the approach taken is to keep recording the "pre" timestamp on each call, overwriting the previously recorded one until the "post" timestamp is also taken.
--- drivers/spi/spi.c | 19 +++++++------------ include/linux/spi/spi.h | 3 +-- 2 files changed, 8 insertions(+), 14 deletions(-)
drivers/spi/spi.c +7 -12
@@ -1515,17 +1515,15 @@ void spi_take_timestamp_pre(struct spi_controller *ctlr,if (!xfer->ptp_sts)return ; -if (xfer->timestamped_pre) +if (xfer->timestamped)return ; -if (progress < xfer->ptp_sts_word_pre) +if (progress > xfer->ptp_sts_word_pre)return ;/* Capture the resolution of the timestamp */ xfer->ptp_sts_word_pre = progress; - xfer->timestamped_pre = true; -if (irqs_off) {local_irq_save (ctlr->irq_flags);preempt_disable ();
@@ -1554,7 +1552,7 @@ void spi_take_timestamp_post(struct spi_controller *ctlr,if (!xfer->ptp_sts)return ; -if (xfer->timestamped_post) +if (xfer->timestamped)return ;if (progress < xfer->ptp_sts_word_post)
@@ -1570,7 +1568,7 @@ void spi_take_timestamp_post(struct spi_controller *ctlr,/* Capture the resolution of the timestamp */ xfer->ptp_sts_word_post = progress; - xfer->timestamped_post = true; + xfer->timestamped = true; }EXPORT_SYMBOL_GPL (spi_take_timestamp_post);
@@ -1675,12 +1673,9 @@ void spi_finalize_current_message(struct spi_controller *ctlr) } } -if (unlikely (ctlr->ptp_sts_supported)) { -list_for_each_entry (xfer, &mesg->transfers, transfer_list) { -WARN_ON_ONCE (xfer->ptp_sts && !xfer->timestamped_pre); -WARN_ON_ONCE (xfer->ptp_sts && !xfer->timestamped_post); - } - } +if (unlikely (ctlr->ptp_sts_supported)) +list_for_each_entry (xfer, &mesg->transfers, transfer_list) +WARN_ON_ONCE (xfer->ptp_sts && !xfer->timestamped);spi_unmap_msg (ctlr, mesg);
include/linux/spi/spi.h +1 -2
@@ -933,8 +933,7 @@ struct spi_transfer {struct ptp_system_timestamp *ptp_sts; -bool timestamped_pre; -bool timestamped_post; +bool timestamped;struct list_head transfer_list; };
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 8/12 spi: spi-fsl-dspi: Convert TCFQ users to XSPI FIFO mode
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
The Transfer Complete Flag (TCF) interrupt gets raised after each write to the TX FIFO (PUSHR) which means that it is not possible to devise a transfer procedure that makes full utilization of the FIFO depth (4 entries on most controllers, 16 entries on some).
On the other hand, XSPI mode has a feature called "command cycling", which allows a single TX command to be run for a pre-specified number of TX words. When the command cycle ends, the Command Transfer Complete Flag bit asserts and raises an interrupt. The advantage in this mode is that the TX FIFO can be better utilized (more words can be batched at once).
Other changes brought by this patch:
- The dspi->rx_end variable has been removed, since now the dspi_fifo_write function sets up dspi->words_in_flight, so dspi_fifo_read knows how much to read without overrunning the RX buffer.
- Stop using poll mode unconditionally for TCFQ mode, since XSPI mode is a little less efficient than that, and so, poll mode doesn't bring as many improvements for XSPI.
- Stop relying on the hardware transfer counter (SPI_TCR_GET_TCNT) and instead increment the message->actual_length based on the newly introduced dspi->words_in_flight variable.
- The CTARE register is now written in the hotpath instead of just at transfer init time, since it contains the DTCP field (transfer preload - the counter indicating how many txdata words will follow), which is a dynamic value.
Due to the fact that the Chip Select toggling setting is part of the command written to the TX FIFO, the ending word of each buffer needs to be sent via its own TX command, so that we have a chance to emit a 1-word command with deasserted PCS.
--- drivers/spi/spi-fsl-dspi.c | 189 ++++++++++++++++++++----------------- 1 file changed, 101 insertions(+), 88 deletions(-)
drivers/spi/spi-fsl-dspi.c +101 -88
@@ -72,6 +72,7 @@#define SPI_RSER 0x30#define SPI_RSER_TCFQE BIT(31)#define SPI_RSER_EOQFE BIT(28) +#define SPI_RSER_CMDTCFE BIT(23)#define SPI_PUSHR 0x34#define SPI_PUSHR_CMD_CONT BIT(15)
@@ -114,14 +115,13 @@ struct chip_data {enum dspi_trans_mode {DSPI_EOQ_MODE =0 , -DSPI_TCFQ_MODE , +DSPI_XSPI_MODE ,DSPI_DMA_MODE , };struct fsl_dspi_devtype_data {enum dspi_trans_mode trans_mode;u8 max_clock_factor; -bool xspi_mode;int fifo_size;int dma_bufsize; };
@@ -147,37 +147,32 @@ static const struct fsl_dspi_devtype_data devtype_data[] = { }, [LS1021A] = { /* Has A-011218 DMA erratum */ - .trans_mode = DSPI_TCFQ_MODE, + .trans_mode = DSPI_XSPI_MODE, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 4, }, [LS1012A] = { /* Has A-011218 DMA erratum */ - .trans_mode = DSPI_TCFQ_MODE, + .trans_mode = DSPI_XSPI_MODE, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 16, }, [LS1043A] = { /* Has A-011218 DMA erratum */ - .trans_mode = DSPI_TCFQ_MODE, + .trans_mode = DSPI_XSPI_MODE, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 16, }, [LS1046A] = { /* Has A-011218 DMA erratum */ - .trans_mode = DSPI_TCFQ_MODE, + .trans_mode = DSPI_XSPI_MODE, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 16, }, [LS2080A] = { .trans_mode = DSPI_DMA_MODE, .dma_bufsize = 8, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 4, }, [LS2085A] = {
@@ -190,7 +185,6 @@ static const struct fsl_dspi_devtype_data devtype_data[] = { .trans_mode = DSPI_DMA_MODE, .dma_bufsize = 8, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 4, }, [MCF5441X] = {
@@ -233,7 +227,6 @@ struct fsl_dspi {size_t len;const void *tx;void *rx; -void *rx_end;u16 tx_cmd;u8 bits_per_word;u8 bytes_per_word;
@@ -243,6 +236,8 @@ struct fsl_dspi {u32 waitflags;struct fsl_dspi_dma *dma; + +int words_in_flight; };/*
@@ -610,7 +605,17 @@ static void dspi_pushr_cmd_write(struct fsl_dspi *dspi) {u16 cmd = dspi->tx_cmd; -if (dspi->len >0 ) +/* +* The only time when the PCS doesn't need continuation after this word +* is when it's last. We need to look ahead, because we actually call +* dspi_pop_tx (the function that decrements dspi->len) _after_ +* dspi_pushr_cmd_write with XSPI mode. As for how much in advance? One +* word is enough. If there's more to transmit than that, +* dspi_xspi_write will know to split the FIFO writes in 2, and +* generate a new PUSHR command with the final word that will have PCS +* deasserted (not continued) here. +*/ +if (dspi->len > dspi->bytes_per_word) cmd |=SPI_PUSHR_CMD_CONT ;regmap_write (dspi->regmap_pushr,PUSHR_CMD , cmd); }
@@ -620,93 +625,115 @@ static void dspi_pushr_txdata_write(struct fsl_dspi *dspi, u16 txdata)regmap_write (dspi->regmap_pushr,PUSHR_TX , txdata); } -static void dspi_tcfq_write (struct fsl_dspi *dspi) +static void dspi_xspi_write (struct fsl_dspi *dspi,int cnt) { -/* Clear transfer count */ - dspi->tx_cmd |=SPI_PUSHR_CMD_CTCNT ; +regmap_write (dspi->regmap,SPI_CTARE (0 ), +SPI_FRAME_EBITS (dspi->bits_per_word) | +SPI_CTARE_DTCP (cnt)); -if (dspi->devtype_data->xspi_mode && dspi->bits_per_word >16 ) { -/* Write the CMD FIFO entry first, and then the two -* corresponding TX FIFO entries. -*/ +/* +* Write the CMD FIFO entry first, and then the two +* corresponding TX FIFO entries (or one...). +*/ +dspi_pushr_cmd_write (dspi); + +/* Fill TX FIFO with as many transfers as possible */ +while (cnt--) {u32 data =dspi_pop_tx (dspi); -dspi_pushr_cmd_write (dspi);dspi_pushr_txdata_write (dspi, data &0xFFFF ); -dspi_pushr_txdata_write (dspi, data >>16 ); - }else { -/* Write one entry to both TX FIFO and CMD FIFO -* simultaneously. -*/ -dspi_pushr_write (dspi); +if (dspi->bits_per_word >16 ) +dspi_pushr_txdata_write (dspi, data >>16 ); } } -static u32 dspi_popr_read (struct fsl_dspi *dspi) +static void dspi_xspi_fifo_write (struct fsl_dspi *dspi) { -u32 rxdata =0 ; +int num_fifo_entries = dspi->devtype_data->fifo_size; +int bytes_in_flight; -regmap_read (dspi->regmap,SPI_POPR , &rxdata); -return rxdata; -} +/* In XSPI mode each 32-bit word occupies 2 TX FIFO entries */ +if (dspi->bits_per_word >16 ) + num_fifo_entries /=2 ; -static void dspi_tcfq_read (struct fsl_dspi *dspi) -{ -dspi_push_rx (dspi,dspi_popr_read (dspi)); + dspi->words_in_flight = dspi->len / dspi->bytes_per_word; + +if (dspi->words_in_flight > num_fifo_entries) + dspi->words_in_flight = num_fifo_entries; + + bytes_in_flight = dspi->words_in_flight * dspi->bytes_per_word; + +/* +* If the PCS needs to de-assert (i.e. we're at the end of the buffer +* and cs_change does not want the PCS to stay on), then we need a new +* PUSHR command, since this one (for the body of the buffer) +* necessarily has the CONT bit set. +* So send one word less during this go, to force a split and a command +* with a single word next time, when CONT will be unset. +*/ +if (bytes_in_flight == dspi->len && dspi->words_in_flight >1 && + !(dspi->tx_cmd &SPI_PUSHR_CMD_CONT )) + dspi->words_in_flight--; + +dspi_xspi_write (dspi, dspi->words_in_flight); } -static void dspi_eoq_write (struct fsl_dspi *dspi) +static void dspi_eoq_fifo_write (struct fsl_dspi *dspi) { -int fifo_size = dspi->devtype_data->fifo_size; +int num_fifo_entries = dspi->devtype_data->fifo_size;u16 xfer_cmd = dspi->tx_cmd; + dspi->words_in_flight = num_fifo_entries; +/* Fill TX FIFO with as many transfers as possible */ -while (dspi->len && fifo_size--) { +while (dspi->len && num_fifo_entries--) { dspi->tx_cmd = xfer_cmd;/* Request EOQF for last transfer in FIFO */ -if (dspi->len == dspi->bytes_per_word || fifo_size ==0 ) +if (dspi->len == dspi->bytes_per_word || num_fifo_entries ==0 ) dspi->tx_cmd |=SPI_PUSHR_CMD_EOQ ; -/* Clear transfer count for first transfer in FIFO */ -if (fifo_size == (dspi->devtype_data->fifo_size -1 )) - dspi->tx_cmd |=SPI_PUSHR_CMD_CTCNT ;/* Write combined TX FIFO and CMD FIFO entry */ dspi_pushr_write (dspi); } } -static void dspi_eoq_read (struct fsl_dspi *dspi) +static u32 dspi_popr_read (struct fsl_dspi *dspi) { -int fifo_size = dspi->devtype_data->fifo_size; +u32 rxdata =0 ; +regmap_read (dspi->regmap,SPI_POPR , &rxdata); +return rxdata; +} + +static void dspi_fifo_read (struct fsl_dspi *dspi) +{/* Read one FIFO entry and push to rx buffer */ -while ((dspi->rx < dspi->rx_end) && fifo_size--) +while (dspi->words_in_flight--)dspi_push_rx (dspi,dspi_popr_read (dspi)); } +static void dspi_fifo_write (struct fsl_dspi *dspi) +{ +if (dspi->devtype_data->trans_mode ==DSPI_EOQ_MODE ) +dspi_eoq_fifo_write (dspi); +else +dspi_xspi_fifo_write (dspi); +} +static int dspi_rxtx (struct fsl_dspi *dspi) { +struct spi_transfer *xfer = dspi->cur_transfer;struct spi_message *msg = dspi->cur_msg; -enum dspi_trans_mode trans_mode; -u16 spi_tcnt; -u32 spi_tcr; +int bytes_sent; + +/* Update total number of bytes that were transferred */ + bytes_sent = dspi->words_in_flight * dspi->bytes_per_word; + msg->actual_length += bytes_sent; + dspi->progress += bytes_sent /DIV_ROUND_UP (xfer->bits_per_word,8 );spi_take_timestamp_post (dspi->ctlr, dspi->cur_transfer, dspi->progress, !dspi->irq); -/* Get transfer counter (in number of SPI transfers). It was -* reset to 0 when transfer(s) were started. -*/ -regmap_read (dspi->regmap,SPI_TCR , &spi_tcr); - spi_tcnt =SPI_TCR_GET_TCNT (spi_tcr); -/* Update total number of bytes that were transferred */ - msg->actual_length += spi_tcnt * dspi->bytes_per_word; - dspi->progress += spi_tcnt; - - trans_mode = dspi->devtype_data->trans_mode; -if (trans_mode ==DSPI_EOQ_MODE ) -dspi_eoq_read (dspi); -else if (trans_mode ==DSPI_TCFQ_MODE ) -dspi_tcfq_read (dspi); +dspi_fifo_read (dspi);if (!dspi->len)/* Success! */
@@ -715,10 +742,7 @@ static int dspi_rxtx(struct fsl_dspi *dspi)spi_take_timestamp_pre (dspi->ctlr, dspi->cur_transfer, dspi->progress, !dspi->irq); -if (trans_mode ==DSPI_EOQ_MODE ) -dspi_eoq_write (dspi); -else if (trans_mode ==DSPI_TCFQ_MODE ) -dspi_tcfq_write (dspi); +dspi_fifo_write (dspi);return -EINPROGRESS ; }
@@ -732,7 +756,7 @@ static int dspi_poll(struct fsl_dspi *dspi)regmap_read (dspi->regmap,SPI_SR , &spi_sr);regmap_write (dspi->regmap,SPI_SR , spi_sr); -if (spi_sr & (SPI_SR_EOQF |SPI_SR_TCFQF )) +if (spi_sr & (SPI_SR_EOQF |SPI_SR_CMDTCF ))break ; }while (--tries);
@@ -750,7 +774,7 @@ static irqreturn_t dspi_interrupt(int irq, void *dev_id)regmap_read (dspi->regmap,SPI_SR , &spi_sr);regmap_write (dspi->regmap,SPI_SR , spi_sr); -if (!(spi_sr &SPI_SR_EOQF )) +if (!(spi_sr & (SPI_SR_EOQF |SPI_SR_CMDTCF )))return IRQ_NONE ;if (dspi_rxtx (dspi) ==0 ) {
@@ -798,7 +822,6 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr, dspi->tx = transfer->tx_buf; dspi->rx = transfer->rx_buf; - dspi->rx_end = dspi->rx + transfer->len; dspi->len = transfer->len; dspi->progress =0 ;/* Validated transfer specific frame size (defaults applied) */
@@ -811,10 +834,6 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr,regmap_write (dspi->regmap,SPI_CTAR (0 ), dspi->cur_chip->ctar_val |SPI_FRAME_BITS (transfer->bits_per_word)); -if (dspi->devtype_data->xspi_mode) -regmap_write (dspi->regmap,SPI_CTARE (0 ), -SPI_FRAME_EBITS (transfer->bits_per_word) | -SPI_CTARE_DTCP (1 ));spi_take_timestamp_pre (dspi->ctlr, dspi->cur_transfer, dspi->progress, !dspi->irq);
@@ -823,11 +842,11 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr,switch (trans_mode) {case DSPI_EOQ_MODE :regmap_write (dspi->regmap,SPI_RSER ,SPI_RSER_EOQFE ); -dspi_eoq_write (dspi); +dspi_fifo_write (dspi);break ; -case DSPI_TCFQ_MODE : -regmap_write (dspi->regmap,SPI_RSER ,SPI_RSER_TCFQE ); -dspi_tcfq_write (dspi); +case DSPI_XSPI_MODE : +regmap_write (dspi->regmap,SPI_RSER ,SPI_RSER_CMDTCFE ); +dspi_fifo_write (dspi);break ;case DSPI_DMA_MODE : regmap_write(dspi->regmap, SPI_RSER,
@@ -1053,16 +1072,13 @@ static void dspi_init(struct fsl_dspi *dspi) {unsigned int mcr =SPI_MCR_PCSIS ; -if (dspi->devtype_data->xspi_mode) +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) mcr |=SPI_MCR_XSPI ;if (!spi_controller_is_slave (dspi->ctlr)) mcr |=SPI_MCR_MASTER ;regmap_write (dspi->regmap,SPI_MCR , mcr);regmap_write (dspi->regmap,SPI_SR ,SPI_SR_CLEAR ); -if (dspi->devtype_data->xspi_mode) -regmap_write (dspi->regmap,SPI_CTARE (0 ), -SPI_CTARE_FMSZE (0 ) |SPI_CTARE_DTCP (1 )); }static int dspi_slave_abort (struct spi_master *master)
@@ -1162,7 +1178,7 @@ static int dspi_probe(struct platform_device *pdev) } } -if (dspi->devtype_data->xspi_mode) +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) ctlr->bits_per_word_mask =SPI_BPW_RANGE_MASK (4 ,32 );else ctlr->bits_per_word_mask =SPI_BPW_RANGE_MASK (4 ,16 );
@@ -1174,7 +1190,7 @@ static int dspi_probe(struct platform_device *pdev) goto out_ctlr_put; } -if (dspi->devtype_data->xspi_mode) +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) regmap_config = &dspi_xspi_regmap_config[0 ];else regmap_config = &dspi_regmap_config;
@@ -1186,7 +1202,7 @@ static int dspi_probe(struct platform_device *pdev) goto out_ctlr_put; } -if (dspi->devtype_data->xspi_mode) { +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) { dspi->regmap_pushr =devm_regmap_init_mmio ( &pdev->dev, base +SPI_PUSHR , &dspi_xspi_regmap_config[1 ]);
@@ -1211,9 +1227,6 @@ static int dspi_probe(struct platform_device *pdev)dspi_init (dspi); -if (dspi->devtype_data->trans_mode ==DSPI_TCFQ_MODE ) - goto poll_mode; - dspi->irq =platform_get_irq (pdev,0 );if (dspi->irq <=0 ) { dev_info(&pdev->dev,
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Convert TCFQ users to XSPI FIFO mode" to the spi tree
The patch
spi: spi-fsl-dspi: Convert TCFQ users to XSPI FIFO mode
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
d59c90a2400ffc4d7127a7653f99da1c4a8fa762Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:40 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Convert TCFQ users to XSPI FIFO mode
The Transfer Complete Flag (TCF) interrupt gets raised after each write to the TX FIFO (PUSHR) which means that it is not possible to devise a transfer procedure that makes full utilization of the FIFO depth (4 entries on most controllers, 16 entries on some).
On the other hand, XSPI mode has a feature called "command cycling", which allows a single TX command to be run for a pre-specified number of TX words. When the command cycle ends, the Command Transfer Complete Flag bit asserts and raises an interrupt. The advantage in this mode is that the TX FIFO can be better utilized (more words can be batched at once).
Other changes brought by this patch:
- The dspi->rx_end variable has been removed, since now the dspi_fifo_write function sets up dspi->words_in_flight, so dspi_fifo_read knows how much to read without overrunning the RX buffer.
- Stop using poll mode unconditionally for TCFQ mode, since XSPI mode is a little less efficient than that, and so, poll mode doesn't bring as many improvements for XSPI.
- Stop relying on the hardware transfer counter (SPI_TCR_GET_TCNT) and instead increment the message->actual_length based on the newly introduced dspi->words_in_flight variable.
- The CTARE register is now written in the hotpath instead of just at transfer init time, since it contains the DTCP field (transfer preload - the counter indicating how many txdata words will follow), which is a dynamic value.
Due to the fact that the Chip Select toggling setting is part of the command written to the TX FIFO, the ending word of each buffer needs to be sent via its own TX command, so that we have a chance to emit a 1-word command with deasserted PCS.
--- drivers/spi/spi-fsl-dspi.c | 189 ++++++++++++++++++++----------------- 1 file changed, 101 insertions(+), 88 deletions(-)
drivers/spi/spi-fsl-dspi.c +101 -88
@@ -72,6 +72,7 @@#define SPI_RSER 0x30#define SPI_RSER_TCFQE BIT(31)#define SPI_RSER_EOQFE BIT(28) +#define SPI_RSER_CMDTCFE BIT(23)#define SPI_PUSHR 0x34#define SPI_PUSHR_CMD_CONT BIT(15)
@@ -114,14 +115,13 @@ struct chip_data {enum dspi_trans_mode {DSPI_EOQ_MODE =0 , -DSPI_TCFQ_MODE , +DSPI_XSPI_MODE ,DSPI_DMA_MODE , };struct fsl_dspi_devtype_data {enum dspi_trans_mode trans_mode;u8 max_clock_factor; -bool xspi_mode;int fifo_size;int dma_bufsize; };
@@ -147,37 +147,32 @@ static const struct fsl_dspi_devtype_data devtype_data[] = { }, [LS1021A] = { /* Has A-011218 DMA erratum */ - .trans_mode = DSPI_TCFQ_MODE, + .trans_mode = DSPI_XSPI_MODE, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 4, }, [LS1012A] = { /* Has A-011218 DMA erratum */ - .trans_mode = DSPI_TCFQ_MODE, + .trans_mode = DSPI_XSPI_MODE, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 16, }, [LS1043A] = { /* Has A-011218 DMA erratum */ - .trans_mode = DSPI_TCFQ_MODE, + .trans_mode = DSPI_XSPI_MODE, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 16, }, [LS1046A] = { /* Has A-011218 DMA erratum */ - .trans_mode = DSPI_TCFQ_MODE, + .trans_mode = DSPI_XSPI_MODE, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 16, }, [LS2080A] = { .trans_mode = DSPI_DMA_MODE, .dma_bufsize = 8, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 4, }, [LS2085A] = {
@@ -190,7 +185,6 @@ static const struct fsl_dspi_devtype_data devtype_data[] = { .trans_mode = DSPI_DMA_MODE, .dma_bufsize = 8, .max_clock_factor = 8, - .xspi_mode = true, .fifo_size = 4, }, [MCF5441X] = {
@@ -233,7 +227,6 @@ struct fsl_dspi {size_t len;const void *tx;void *rx; -void *rx_end;u16 tx_cmd;u8 bits_per_word;u8 bytes_per_word;
@@ -243,6 +236,8 @@ struct fsl_dspi {u32 waitflags;struct fsl_dspi_dma *dma; + +int words_in_flight; };/*
@@ -610,7 +605,17 @@ static void dspi_pushr_cmd_write(struct fsl_dspi *dspi) {u16 cmd = dspi->tx_cmd; -if (dspi->len >0 ) +/* +* The only time when the PCS doesn't need continuation after this word +* is when it's last. We need to look ahead, because we actually call +* dspi_pop_tx (the function that decrements dspi->len) _after_ +* dspi_pushr_cmd_write with XSPI mode. As for how much in advance? One +* word is enough. If there's more to transmit than that, +* dspi_xspi_write will know to split the FIFO writes in 2, and +* generate a new PUSHR command with the final word that will have PCS +* deasserted (not continued) here. +*/ +if (dspi->len > dspi->bytes_per_word) cmd |=SPI_PUSHR_CMD_CONT ;regmap_write (dspi->regmap_pushr,PUSHR_CMD , cmd); }
@@ -620,93 +625,115 @@ static void dspi_pushr_txdata_write(struct fsl_dspi *dspi, u16 txdata)regmap_write (dspi->regmap_pushr,PUSHR_TX , txdata); } -static void dspi_tcfq_write (struct fsl_dspi *dspi) +static void dspi_xspi_write (struct fsl_dspi *dspi,int cnt) { -/* Clear transfer count */ - dspi->tx_cmd |=SPI_PUSHR_CMD_CTCNT ; +regmap_write (dspi->regmap,SPI_CTARE (0 ), +SPI_FRAME_EBITS (dspi->bits_per_word) | +SPI_CTARE_DTCP (cnt)); -if (dspi->devtype_data->xspi_mode && dspi->bits_per_word >16 ) { -/* Write the CMD FIFO entry first, and then the two -* corresponding TX FIFO entries. -*/ +/* +* Write the CMD FIFO entry first, and then the two +* corresponding TX FIFO entries (or one...). +*/ +dspi_pushr_cmd_write (dspi); + +/* Fill TX FIFO with as many transfers as possible */ +while (cnt--) {u32 data =dspi_pop_tx (dspi); -dspi_pushr_cmd_write (dspi);dspi_pushr_txdata_write (dspi, data &0xFFFF ); -dspi_pushr_txdata_write (dspi, data >>16 ); - }else { -/* Write one entry to both TX FIFO and CMD FIFO -* simultaneously. -*/ -dspi_pushr_write (dspi); +if (dspi->bits_per_word >16 ) +dspi_pushr_txdata_write (dspi, data >>16 ); } } -static u32 dspi_popr_read (struct fsl_dspi *dspi) +static void dspi_xspi_fifo_write (struct fsl_dspi *dspi) { -u32 rxdata =0 ; +int num_fifo_entries = dspi->devtype_data->fifo_size; +int bytes_in_flight; -regmap_read (dspi->regmap,SPI_POPR , &rxdata); -return rxdata; -} +/* In XSPI mode each 32-bit word occupies 2 TX FIFO entries */ +if (dspi->bits_per_word >16 ) + num_fifo_entries /=2 ; -static void dspi_tcfq_read (struct fsl_dspi *dspi) -{ -dspi_push_rx (dspi,dspi_popr_read (dspi)); + dspi->words_in_flight = dspi->len / dspi->bytes_per_word; + +if (dspi->words_in_flight > num_fifo_entries) + dspi->words_in_flight = num_fifo_entries; + + bytes_in_flight = dspi->words_in_flight * dspi->bytes_per_word; + +/* +* If the PCS needs to de-assert (i.e. we're at the end of the buffer +* and cs_change does not want the PCS to stay on), then we need a new +* PUSHR command, since this one (for the body of the buffer) +* necessarily has the CONT bit set. +* So send one word less during this go, to force a split and a command +* with a single word next time, when CONT will be unset. +*/ +if (bytes_in_flight == dspi->len && dspi->words_in_flight >1 && + !(dspi->tx_cmd &SPI_PUSHR_CMD_CONT )) + dspi->words_in_flight--; + +dspi_xspi_write (dspi, dspi->words_in_flight); } -static void dspi_eoq_write (struct fsl_dspi *dspi) +static void dspi_eoq_fifo_write (struct fsl_dspi *dspi) { -int fifo_size = dspi->devtype_data->fifo_size; +int num_fifo_entries = dspi->devtype_data->fifo_size;u16 xfer_cmd = dspi->tx_cmd; + dspi->words_in_flight = num_fifo_entries; +/* Fill TX FIFO with as many transfers as possible */ -while (dspi->len && fifo_size--) { +while (dspi->len && num_fifo_entries--) { dspi->tx_cmd = xfer_cmd;/* Request EOQF for last transfer in FIFO */ -if (dspi->len == dspi->bytes_per_word || fifo_size ==0 ) +if (dspi->len == dspi->bytes_per_word || num_fifo_entries ==0 ) dspi->tx_cmd |=SPI_PUSHR_CMD_EOQ ; -/* Clear transfer count for first transfer in FIFO */ -if (fifo_size == (dspi->devtype_data->fifo_size -1 )) - dspi->tx_cmd |=SPI_PUSHR_CMD_CTCNT ;/* Write combined TX FIFO and CMD FIFO entry */ dspi_pushr_write (dspi); } } -static void dspi_eoq_read (struct fsl_dspi *dspi) +static u32 dspi_popr_read (struct fsl_dspi *dspi) { -int fifo_size = dspi->devtype_data->fifo_size; +u32 rxdata =0 ; +regmap_read (dspi->regmap,SPI_POPR , &rxdata); +return rxdata; +} + +static void dspi_fifo_read (struct fsl_dspi *dspi) +{/* Read one FIFO entry and push to rx buffer */ -while ((dspi->rx < dspi->rx_end) && fifo_size--) +while (dspi->words_in_flight--)dspi_push_rx (dspi,dspi_popr_read (dspi)); } +static void dspi_fifo_write (struct fsl_dspi *dspi) +{ +if (dspi->devtype_data->trans_mode ==DSPI_EOQ_MODE ) +dspi_eoq_fifo_write (dspi); +else +dspi_xspi_fifo_write (dspi); +} +static int dspi_rxtx (struct fsl_dspi *dspi) { +struct spi_transfer *xfer = dspi->cur_transfer;struct spi_message *msg = dspi->cur_msg; -enum dspi_trans_mode trans_mode; -u16 spi_tcnt; -u32 spi_tcr; +int bytes_sent; + +/* Update total number of bytes that were transferred */ + bytes_sent = dspi->words_in_flight * dspi->bytes_per_word; + msg->actual_length += bytes_sent; + dspi->progress += bytes_sent /DIV_ROUND_UP (xfer->bits_per_word,8 );spi_take_timestamp_post (dspi->ctlr, dspi->cur_transfer, dspi->progress, !dspi->irq); -/* Get transfer counter (in number of SPI transfers). It was -* reset to 0 when transfer(s) were started. -*/ -regmap_read (dspi->regmap,SPI_TCR , &spi_tcr); - spi_tcnt =SPI_TCR_GET_TCNT (spi_tcr); -/* Update total number of bytes that were transferred */ - msg->actual_length += spi_tcnt * dspi->bytes_per_word; - dspi->progress += spi_tcnt; - - trans_mode = dspi->devtype_data->trans_mode; -if (trans_mode ==DSPI_EOQ_MODE ) -dspi_eoq_read (dspi); -else if (trans_mode ==DSPI_TCFQ_MODE ) -dspi_tcfq_read (dspi); +dspi_fifo_read (dspi);if (!dspi->len)/* Success! */
@@ -715,10 +742,7 @@ static int dspi_rxtx(struct fsl_dspi *dspi)spi_take_timestamp_pre (dspi->ctlr, dspi->cur_transfer, dspi->progress, !dspi->irq); -if (trans_mode ==DSPI_EOQ_MODE ) -dspi_eoq_write (dspi); -else if (trans_mode ==DSPI_TCFQ_MODE ) -dspi_tcfq_write (dspi); +dspi_fifo_write (dspi);return -EINPROGRESS ; }
@@ -732,7 +756,7 @@ static int dspi_poll(struct fsl_dspi *dspi)regmap_read (dspi->regmap,SPI_SR , &spi_sr);regmap_write (dspi->regmap,SPI_SR , spi_sr); -if (spi_sr & (SPI_SR_EOQF |SPI_SR_TCFQF )) +if (spi_sr & (SPI_SR_EOQF |SPI_SR_CMDTCF ))break ; }while (--tries);
@@ -750,7 +774,7 @@ static irqreturn_t dspi_interrupt(int irq, void *dev_id)regmap_read (dspi->regmap,SPI_SR , &spi_sr);regmap_write (dspi->regmap,SPI_SR , spi_sr); -if (!(spi_sr &SPI_SR_EOQF )) +if (!(spi_sr & (SPI_SR_EOQF |SPI_SR_CMDTCF )))return IRQ_NONE ;if (dspi_rxtx (dspi) ==0 ) {
@@ -798,7 +822,6 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr, dspi->tx = transfer->tx_buf; dspi->rx = transfer->rx_buf; - dspi->rx_end = dspi->rx + transfer->len; dspi->len = transfer->len; dspi->progress =0 ;/* Validated transfer specific frame size (defaults applied) */
@@ -811,10 +834,6 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr,regmap_write (dspi->regmap,SPI_CTAR (0 ), dspi->cur_chip->ctar_val |SPI_FRAME_BITS (transfer->bits_per_word)); -if (dspi->devtype_data->xspi_mode) -regmap_write (dspi->regmap,SPI_CTARE (0 ), -SPI_FRAME_EBITS (transfer->bits_per_word) | -SPI_CTARE_DTCP (1 ));spi_take_timestamp_pre (dspi->ctlr, dspi->cur_transfer, dspi->progress, !dspi->irq);
@@ -823,11 +842,11 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr,switch (trans_mode) {case DSPI_EOQ_MODE :regmap_write (dspi->regmap,SPI_RSER ,SPI_RSER_EOQFE ); -dspi_eoq_write (dspi); +dspi_fifo_write (dspi);break ; -case DSPI_TCFQ_MODE : -regmap_write (dspi->regmap,SPI_RSER ,SPI_RSER_TCFQE ); -dspi_tcfq_write (dspi); +case DSPI_XSPI_MODE : +regmap_write (dspi->regmap,SPI_RSER ,SPI_RSER_CMDTCFE ); +dspi_fifo_write (dspi);break ;case DSPI_DMA_MODE : regmap_write(dspi->regmap, SPI_RSER,
@@ -1053,16 +1072,13 @@ static void dspi_init(struct fsl_dspi *dspi) {unsigned int mcr =SPI_MCR_PCSIS ; -if (dspi->devtype_data->xspi_mode) +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) mcr |=SPI_MCR_XSPI ;if (!spi_controller_is_slave (dspi->ctlr)) mcr |=SPI_MCR_MASTER ;regmap_write (dspi->regmap,SPI_MCR , mcr);regmap_write (dspi->regmap,SPI_SR ,SPI_SR_CLEAR ); -if (dspi->devtype_data->xspi_mode) -regmap_write (dspi->regmap,SPI_CTARE (0 ), -SPI_CTARE_FMSZE (0 ) |SPI_CTARE_DTCP (1 )); }static int dspi_slave_abort (struct spi_master *master)
@@ -1162,7 +1178,7 @@ static int dspi_probe(struct platform_device *pdev) } } -if (dspi->devtype_data->xspi_mode) +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) ctlr->bits_per_word_mask =SPI_BPW_RANGE_MASK (4 ,32 );else ctlr->bits_per_word_mask =SPI_BPW_RANGE_MASK (4 ,16 );
@@ -1174,7 +1190,7 @@ static int dspi_probe(struct platform_device *pdev) goto out_ctlr_put; } -if (dspi->devtype_data->xspi_mode) +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) regmap_config = &dspi_xspi_regmap_config[0 ];else regmap_config = &dspi_regmap_config;
@@ -1186,7 +1202,7 @@ static int dspi_probe(struct platform_device *pdev) goto out_ctlr_put; } -if (dspi->devtype_data->xspi_mode) { +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) { dspi->regmap_pushr =devm_regmap_init_mmio ( &pdev->dev, base +SPI_PUSHR , &dspi_xspi_regmap_config[1 ]);
@@ -1211,9 +1227,6 @@ static int dspi_probe(struct platform_device *pdev)dspi_init (dspi); -if (dspi->devtype_data->trans_mode ==DSPI_TCFQ_MODE ) - goto poll_mode; - dspi->irq =platform_get_irq (pdev,0 );if (dspi->irq <=0 ) { dev_info(&pdev->dev,
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 9/12 spi: spi-fsl-dspi: Accelerate transfers using larger word size if possible
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
This patch adds logic in the driver to transmit SPI buffers that use bits_per_word=8 with a higher bits_per_word count (multiple of 8).
Currently the following (most common) modes are implemented:
- 8 bits_per_word on 32-bit capable controllers
- 8 bits_per_word on 16-bit capable controllers
- 16 bits_per_word on 32-bit capable controllers
Transfers which are not accelerated are transferred with a hardware bits_per_word value equal to the one of the SPI transfer.
The difference from just extending bits_per_word=32 at the spi_device driver level is that endianness is different - the SPI core wants to treat bits_per_word=32 buffers as arrays of u32 (i.e. words in host CPU endianness). So to preserve endianness when clumping 8x4 bits into 32-bit words, one must perform conversion between CPU and standard (big) endianness.
All appearances (both on the wire as well as in the buffers presented to the peripheral driver) are preserved, just that accesses to the PUSHR and POPR registers are now more efficient, since the same number of reads/writes can now carry more data (2x more data on TX, 4x more data on RX).
--- drivers/spi/spi-fsl-dspi.c | 160 +++++++++++++++++++++++++++++++------ 1 file changed, 135 insertions(+), 25 deletions(-)
drivers/spi/spi-fsl-dspi.c +135 -25
@@ -228,8 +228,6 @@ struct fsl_dspi {const void *tx;void *rx;u16 tx_cmd; -u8 bits_per_word; -u8 bytes_per_word;const struct fsl_dspi_devtype_data *devtype_data;wait_queue_head_t waitq;
@@ -237,9 +235,70 @@ struct fsl_dspi {struct fsl_dspi_dma *dma; +int oper_word_size; +int oper_bits_per_word; +int words_in_flight; + +void (*host_to_dev)(struct fsl_dspi *dspi,u32 *txdata); +void (*dev_to_host)(struct fsl_dspi *dspi,u32 rxdata); }; +static void dspi_native_host_to_dev (struct fsl_dspi *dspi,u32 *txdata) +{ +memcpy (txdata, dspi->tx, dspi->oper_word_size); + dspi->tx += dspi->oper_word_size; +} + +static void dspi_native_dev_to_host (struct fsl_dspi *dspi,u32 rxdata) +{ +memcpy (dspi->rx, &rxdata, dspi->oper_word_size); + dspi->rx += dspi->oper_word_size; +} + +static void dspi_8on32_host_to_dev (struct fsl_dspi *dspi,u32 *txdata) +{ + *txdata =cpu_to_be32 (*(u32 *)dspi->tx); + dspi->tx +=sizeof (u32); +} + +static void dspi_8on32_dev_to_host (struct fsl_dspi *dspi,u32 rxdata) +{ + *(u32 *)dspi->rx =be32_to_cpu (rxdata); + dspi->rx +=sizeof (u32); +} + +static void dspi_8on16_host_to_dev (struct fsl_dspi *dspi,u32 *txdata) +{ + *txdata =cpu_to_be16 (*(u16 *)dspi->tx); + dspi->tx +=sizeof (u16); +} + +static void dspi_8on16_dev_to_host (struct fsl_dspi *dspi,u32 rxdata) +{ + *(u16 *)dspi->rx =be16_to_cpu (rxdata); + dspi->rx +=sizeof (u16); +} + +static void dspi_16on32_host_to_dev (struct fsl_dspi *dspi,u32 *txdata) +{ +u16 hi = *(u16 *)dspi->tx; +u16 lo = *(u16 *)(dspi->tx +2 ); + + *txdata = (u32 )hi <<16 | lo; + dspi->tx +=sizeof (u32); +} + +static void dspi_16on32_dev_to_host (struct fsl_dspi *dspi,u32 rxdata) +{ +u16 hi = rxdata &0xffff ; +u16 lo = rxdata >>16 ; + + *(u16 *)dspi->rx = lo; + *(u16 *)(dspi->rx +2 ) = hi; + dspi->rx +=sizeof (u32); +} +/* * Pop one word from the TX buffer for pushing into the * PUSHR register (TX FIFO)
@@ -248,11 +307,9 @@ static u32 dspi_pop_tx(struct fsl_dspi *dspi) {u32 txdata =0 ; -if (dspi->tx) { -memcpy (&txdata, dspi->tx, dspi->bytes_per_word); - dspi->tx += dspi->bytes_per_word; - } - dspi->len -= dspi->bytes_per_word; +if (dspi->tx) + dspi->host_to_dev (dspi, &txdata); + dspi->len -= dspi->oper_word_size;return txdata; }
@@ -274,9 +331,7 @@ static void dspi_push_rx(struct fsl_dspi *dspi, u32 rxdata) {if (!dspi->rx)return ; - -memcpy (dspi->rx, &rxdata, dspi->bytes_per_word); - dspi->rx += dspi->bytes_per_word; + dspi->dev_to_host (dspi, rxdata); }static void dspi_tx_dma_callback (void *arg)
@@ -393,8 +448,8 @@ static int dspi_dma_xfer(struct fsl_dspi *dspi) dspi->devtype_data->fifo_size;while (curr_remaining_bytes) {/* Check if current transfer fits the DMA buffer */ - dma->curr_xfer_len = curr_remaining_bytes - / dspi->bytes_per_word; + dma->curr_xfer_len = curr_remaining_bytes / + dspi->oper_word_size;if (dma->curr_xfer_len > bytes_per_buffer) dma->curr_xfer_len = bytes_per_buffer;
@@ -404,8 +459,8 @@ static int dspi_dma_xfer(struct fsl_dspi *dspi) goto exit; }else { -const int len = - dma->curr_xfer_len * dspi->bytes_per_word; +const int len = dma->curr_xfer_len * + dspi->oper_word_size; curr_remaining_bytes -= len; message->actual_length += len;if (curr_remaining_bytes <0 )
@@ -615,7 +670,7 @@ static void dspi_pushr_cmd_write(struct fsl_dspi *dspi)* generate a new PUSHR command with the final word that will have PCS * deasserted (not continued) here. */ -if (dspi->len > dspi->bytes_per_word) +if (dspi->len > dspi->oper_word_size) cmd |=SPI_PUSHR_CMD_CONT ;regmap_write (dspi->regmap_pushr,PUSHR_CMD , cmd); }
@@ -627,8 +682,9 @@ static void dspi_pushr_txdata_write(struct fsl_dspi *dspi, u16 txdata)static void dspi_xspi_write (struct fsl_dspi *dspi,int cnt) { +/* Update CTARE */ regmap_write (dspi->regmap,SPI_CTARE (0 ), -SPI_FRAME_EBITS (dspi->bits_per_word) | +SPI_FRAME_EBITS (dspi->oper_bits_per_word) |SPI_CTARE_DTCP (cnt));/*
@@ -642,7 +698,7 @@ static void dspi_xspi_write(struct fsl_dspi *dspi, int cnt)u32 data =dspi_pop_tx (dspi);dspi_pushr_txdata_write (dspi, data &0xFFFF ); -if (dspi->bits_per_word >16 ) +if (dspi->oper_bits_per_word >16 )dspi_pushr_txdata_write (dspi, data >>16 ); } }
@@ -653,15 +709,20 @@ static void dspi_xspi_fifo_write(struct fsl_dspi *dspi)int bytes_in_flight;/* In XSPI mode each 32-bit word occupies 2 TX FIFO entries */ -if (dspi->bits_per_word >16 ) +if (dspi->oper_word_size ==4 ) num_fifo_entries /=2 ; - dspi->words_in_flight = dspi->len / dspi->bytes_per_word; +/* +* Integer division intentionally trims off odd (or non-multiple of 4) +* numbers of bytes at the end of the buffer, which will be sent next +* time using a smaller oper_word_size. +*/ + dspi->words_in_flight = dspi->len / dspi->oper_word_size;if (dspi->words_in_flight > num_fifo_entries) dspi->words_in_flight = num_fifo_entries; - bytes_in_flight = dspi->words_in_flight * dspi->bytes_per_word; + bytes_in_flight = dspi->words_in_flight * dspi->oper_word_size;/* * If the PCS needs to de-assert (i.e. we're at the end of the buffer
@@ -689,7 +750,7 @@ static void dspi_eoq_fifo_write(struct fsl_dspi *dspi)while (dspi->len && num_fifo_entries--) { dspi->tx_cmd = xfer_cmd;/* Request EOQF for last transfer in FIFO */ -if (dspi->len == dspi->bytes_per_word || num_fifo_entries ==0 ) +if (dspi->len == dspi->oper_word_size || num_fifo_entries ==0 ) dspi->tx_cmd |=SPI_PUSHR_CMD_EOQ ;/* Write combined TX FIFO and CMD FIFO entry */ dspi_pushr_write (dspi);
@@ -711,8 +772,56 @@ static void dspi_fifo_read(struct fsl_dspi *dspi)dspi_push_rx (dspi,dspi_popr_read (dspi)); } +static void dspi_setup_accel (struct fsl_dspi *dspi) +{ +struct spi_transfer *xfer = dspi->cur_transfer; + +/* Start off with maximum supported by hardware */ +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) + dspi->oper_bits_per_word =32 ; +else + dspi->oper_bits_per_word =16 ; + +/* And go down only if the buffer can't be sent with words this big */ +do { +if (dspi->len >=DIV_ROUND_UP (dspi->oper_bits_per_word,8 )) +break ; + + dspi->oper_bits_per_word /=2 ; + }while (dspi->oper_bits_per_word >8 ); + +if (xfer->bits_per_word ==8 && dspi->oper_bits_per_word ==32 ) { + dspi->dev_to_host = dspi_8on32_dev_to_host; + dspi->host_to_dev = dspi_8on32_host_to_dev; + }else if (xfer->bits_per_word ==8 && dspi->oper_bits_per_word ==16 ) { + dspi->dev_to_host = dspi_8on16_dev_to_host; + dspi->host_to_dev = dspi_8on16_host_to_dev; + }else if (xfer->bits_per_word ==16 && dspi->oper_bits_per_word ==32 ) { + dspi->dev_to_host = dspi_16on32_dev_to_host; + dspi->host_to_dev = dspi_16on32_host_to_dev; + }else { +/* No acceleration needed (8<N<=16 on 16, 16<N<=32 on 32) */ + dspi->dev_to_host = dspi_native_dev_to_host; + dspi->host_to_dev = dspi_native_host_to_dev; + dspi->oper_bits_per_word = xfer->bits_per_word; + } + + dspi->oper_word_size =DIV_ROUND_UP (dspi->oper_bits_per_word,8 ); + +/* +* Update CTAR here (code is common for both EOQ and XSPI modes). +* We will update CTARE in the portion specific to XSPI, when we +* also know the preload value (DTCP). +*/ +regmap_write (dspi->regmap,SPI_CTAR (0 ), + dspi->cur_chip->ctar_val | +SPI_FRAME_BITS (dspi->oper_bits_per_word)); +} +static void dspi_fifo_write (struct fsl_dspi *dspi) { +dspi_setup_accel (dspi); +if (dspi->devtype_data->trans_mode ==DSPI_EOQ_MODE )dspi_eoq_fifo_write (dspi);else
@@ -726,7 +835,7 @@ static int dspi_rxtx(struct fsl_dspi *dspi)int bytes_sent;/* Update total number of bytes that were transferred */ - bytes_sent = dspi->words_in_flight * dspi->bytes_per_word; + bytes_sent = dspi->words_in_flight * dspi->oper_word_size; msg->actual_length += bytes_sent; dspi->progress += bytes_sent /DIV_ROUND_UP (xfer->bits_per_word,8 );
@@ -824,13 +933,14 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr, dspi->rx = transfer->rx_buf; dspi->len = transfer->len; dspi->progress =0 ; -/* Validated transfer specific frame size (defaults applied) */ - dspi->bits_per_word = transfer->bits_per_word; - dspi->bytes_per_word =DIV_ROUND_UP (dspi->bits_per_word,8 );regmap_update_bits (dspi->regmap,SPI_MCR ,SPI_MCR_CLR_TXF |SPI_MCR_CLR_RXF ,SPI_MCR_CLR_TXF |SPI_MCR_CLR_RXF ); +/* +* Static CTAR setup for modes that don't dynamically adjust it +* via dspi_setup_accel (aka for DMA) +*/ regmap_write (dspi->regmap,SPI_CTAR (0 ), dspi->cur_chip->ctar_val |SPI_FRAME_BITS (transfer->bits_per_word));
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Accelerate transfers using larger word size if possible" to the spi tree
The patch
spi: spi-fsl-dspi: Accelerate transfers using larger word size if possible
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
6c1c26ecd9a31c24f9ea7dfb174528141dd32361Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:41 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Accelerate transfers using larger word
size if possible
This patch adds logic in the driver to transmit SPI buffers that use bits_per_word=8 with a higher bits_per_word count (multiple of 8).
Currently the following (most common) modes are implemented:
- 8 bits_per_word on 32-bit capable controllers
- 8 bits_per_word on 16-bit capable controllers
- 16 bits_per_word on 32-bit capable controllers
Transfers which are not accelerated are transferred with a hardware bits_per_word value equal to the one of the SPI transfer.
The difference from just extending bits_per_word=32 at the spi_device driver level is that endianness is different - the SPI core wants to treat bits_per_word=32 buffers as arrays of u32 (i.e. words in host CPU endianness). So to preserve endianness when clumping 8x4 bits into 32-bit words, one must perform conversion between CPU and standard (big) endianness.
All appearances (both on the wire as well as in the buffers presented to the peripheral driver) are preserved, just that accesses to the PUSHR and POPR registers are now more efficient, since the same number of reads/writes can now carry more data (2x more data on TX, 4x more data on RX).
--- drivers/spi/spi-fsl-dspi.c | 160 +++++++++++++++++++++++++++++++------ 1 file changed, 135 insertions(+), 25 deletions(-)
drivers/spi/spi-fsl-dspi.c +135 -25
@@ -228,8 +228,6 @@ struct fsl_dspi {const void *tx;void *rx;u16 tx_cmd; -u8 bits_per_word; -u8 bytes_per_word;const struct fsl_dspi_devtype_data *devtype_data;wait_queue_head_t waitq;
@@ -237,9 +235,70 @@ struct fsl_dspi {struct fsl_dspi_dma *dma; +int oper_word_size; +int oper_bits_per_word; +int words_in_flight; + +void (*host_to_dev)(struct fsl_dspi *dspi,u32 *txdata); +void (*dev_to_host)(struct fsl_dspi *dspi,u32 rxdata); }; +static void dspi_native_host_to_dev (struct fsl_dspi *dspi,u32 *txdata) +{ +memcpy (txdata, dspi->tx, dspi->oper_word_size); + dspi->tx += dspi->oper_word_size; +} + +static void dspi_native_dev_to_host (struct fsl_dspi *dspi,u32 rxdata) +{ +memcpy (dspi->rx, &rxdata, dspi->oper_word_size); + dspi->rx += dspi->oper_word_size; +} + +static void dspi_8on32_host_to_dev (struct fsl_dspi *dspi,u32 *txdata) +{ + *txdata =cpu_to_be32 (*(u32 *)dspi->tx); + dspi->tx +=sizeof (u32); +} + +static void dspi_8on32_dev_to_host (struct fsl_dspi *dspi,u32 rxdata) +{ + *(u32 *)dspi->rx =be32_to_cpu (rxdata); + dspi->rx +=sizeof (u32); +} + +static void dspi_8on16_host_to_dev (struct fsl_dspi *dspi,u32 *txdata) +{ + *txdata =cpu_to_be16 (*(u16 *)dspi->tx); + dspi->tx +=sizeof (u16); +} + +static void dspi_8on16_dev_to_host (struct fsl_dspi *dspi,u32 rxdata) +{ + *(u16 *)dspi->rx =be16_to_cpu (rxdata); + dspi->rx +=sizeof (u16); +} + +static void dspi_16on32_host_to_dev (struct fsl_dspi *dspi,u32 *txdata) +{ +u16 hi = *(u16 *)dspi->tx; +u16 lo = *(u16 *)(dspi->tx +2 ); + + *txdata = (u32 )hi <<16 | lo; + dspi->tx +=sizeof (u32); +} + +static void dspi_16on32_dev_to_host (struct fsl_dspi *dspi,u32 rxdata) +{ +u16 hi = rxdata &0xffff ; +u16 lo = rxdata >>16 ; + + *(u16 *)dspi->rx = lo; + *(u16 *)(dspi->rx +2 ) = hi; + dspi->rx +=sizeof (u32); +} +/* * Pop one word from the TX buffer for pushing into the * PUSHR register (TX FIFO)
@@ -248,11 +307,9 @@ static u32 dspi_pop_tx(struct fsl_dspi *dspi) {u32 txdata =0 ; -if (dspi->tx) { -memcpy (&txdata, dspi->tx, dspi->bytes_per_word); - dspi->tx += dspi->bytes_per_word; - } - dspi->len -= dspi->bytes_per_word; +if (dspi->tx) + dspi->host_to_dev (dspi, &txdata); + dspi->len -= dspi->oper_word_size;return txdata; }
@@ -274,9 +331,7 @@ static void dspi_push_rx(struct fsl_dspi *dspi, u32 rxdata) {if (!dspi->rx)return ; - -memcpy (dspi->rx, &rxdata, dspi->bytes_per_word); - dspi->rx += dspi->bytes_per_word; + dspi->dev_to_host (dspi, rxdata); }static void dspi_tx_dma_callback (void *arg)
@@ -393,8 +448,8 @@ static int dspi_dma_xfer(struct fsl_dspi *dspi) dspi->devtype_data->fifo_size;while (curr_remaining_bytes) {/* Check if current transfer fits the DMA buffer */ - dma->curr_xfer_len = curr_remaining_bytes - / dspi->bytes_per_word; + dma->curr_xfer_len = curr_remaining_bytes / + dspi->oper_word_size;if (dma->curr_xfer_len > bytes_per_buffer) dma->curr_xfer_len = bytes_per_buffer;
@@ -404,8 +459,8 @@ static int dspi_dma_xfer(struct fsl_dspi *dspi) goto exit; }else { -const int len = - dma->curr_xfer_len * dspi->bytes_per_word; +const int len = dma->curr_xfer_len * + dspi->oper_word_size; curr_remaining_bytes -= len; message->actual_length += len;if (curr_remaining_bytes <0 )
@@ -615,7 +670,7 @@ static void dspi_pushr_cmd_write(struct fsl_dspi *dspi)* generate a new PUSHR command with the final word that will have PCS * deasserted (not continued) here. */ -if (dspi->len > dspi->bytes_per_word) +if (dspi->len > dspi->oper_word_size) cmd |=SPI_PUSHR_CMD_CONT ;regmap_write (dspi->regmap_pushr,PUSHR_CMD , cmd); }
@@ -627,8 +682,9 @@ static void dspi_pushr_txdata_write(struct fsl_dspi *dspi, u16 txdata)static void dspi_xspi_write (struct fsl_dspi *dspi,int cnt) { +/* Update CTARE */ regmap_write (dspi->regmap,SPI_CTARE (0 ), -SPI_FRAME_EBITS (dspi->bits_per_word) | +SPI_FRAME_EBITS (dspi->oper_bits_per_word) |SPI_CTARE_DTCP (cnt));/*
@@ -642,7 +698,7 @@ static void dspi_xspi_write(struct fsl_dspi *dspi, int cnt)u32 data =dspi_pop_tx (dspi);dspi_pushr_txdata_write (dspi, data &0xFFFF ); -if (dspi->bits_per_word >16 ) +if (dspi->oper_bits_per_word >16 )dspi_pushr_txdata_write (dspi, data >>16 ); } }
@@ -653,15 +709,20 @@ static void dspi_xspi_fifo_write(struct fsl_dspi *dspi)int bytes_in_flight;/* In XSPI mode each 32-bit word occupies 2 TX FIFO entries */ -if (dspi->bits_per_word >16 ) +if (dspi->oper_word_size ==4 ) num_fifo_entries /=2 ; - dspi->words_in_flight = dspi->len / dspi->bytes_per_word; +/* +* Integer division intentionally trims off odd (or non-multiple of 4) +* numbers of bytes at the end of the buffer, which will be sent next +* time using a smaller oper_word_size. +*/ + dspi->words_in_flight = dspi->len / dspi->oper_word_size;if (dspi->words_in_flight > num_fifo_entries) dspi->words_in_flight = num_fifo_entries; - bytes_in_flight = dspi->words_in_flight * dspi->bytes_per_word; + bytes_in_flight = dspi->words_in_flight * dspi->oper_word_size;/* * If the PCS needs to de-assert (i.e. we're at the end of the buffer
@@ -689,7 +750,7 @@ static void dspi_eoq_fifo_write(struct fsl_dspi *dspi)while (dspi->len && num_fifo_entries--) { dspi->tx_cmd = xfer_cmd;/* Request EOQF for last transfer in FIFO */ -if (dspi->len == dspi->bytes_per_word || num_fifo_entries ==0 ) +if (dspi->len == dspi->oper_word_size || num_fifo_entries ==0 ) dspi->tx_cmd |=SPI_PUSHR_CMD_EOQ ;/* Write combined TX FIFO and CMD FIFO entry */ dspi_pushr_write (dspi);
@@ -711,8 +772,56 @@ static void dspi_fifo_read(struct fsl_dspi *dspi)dspi_push_rx (dspi,dspi_popr_read (dspi)); } +static void dspi_setup_accel (struct fsl_dspi *dspi) +{ +struct spi_transfer *xfer = dspi->cur_transfer; + +/* Start off with maximum supported by hardware */ +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) + dspi->oper_bits_per_word =32 ; +else + dspi->oper_bits_per_word =16 ; + +/* And go down only if the buffer can't be sent with words this big */ +do { +if (dspi->len >=DIV_ROUND_UP (dspi->oper_bits_per_word,8 )) +break ; + + dspi->oper_bits_per_word /=2 ; + }while (dspi->oper_bits_per_word >8 ); + +if (xfer->bits_per_word ==8 && dspi->oper_bits_per_word ==32 ) { + dspi->dev_to_host = dspi_8on32_dev_to_host; + dspi->host_to_dev = dspi_8on32_host_to_dev; + }else if (xfer->bits_per_word ==8 && dspi->oper_bits_per_word ==16 ) { + dspi->dev_to_host = dspi_8on16_dev_to_host; + dspi->host_to_dev = dspi_8on16_host_to_dev; + }else if (xfer->bits_per_word ==16 && dspi->oper_bits_per_word ==32 ) { + dspi->dev_to_host = dspi_16on32_dev_to_host; + dspi->host_to_dev = dspi_16on32_host_to_dev; + }else { +/* No acceleration needed (8<N<=16 on 16, 16<N<=32 on 32) */ + dspi->dev_to_host = dspi_native_dev_to_host; + dspi->host_to_dev = dspi_native_host_to_dev; + dspi->oper_bits_per_word = xfer->bits_per_word; + } + + dspi->oper_word_size =DIV_ROUND_UP (dspi->oper_bits_per_word,8 ); + +/* +* Update CTAR here (code is common for both EOQ and XSPI modes). +* We will update CTARE in the portion specific to XSPI, when we +* also know the preload value (DTCP). +*/ +regmap_write (dspi->regmap,SPI_CTAR (0 ), + dspi->cur_chip->ctar_val | +SPI_FRAME_BITS (dspi->oper_bits_per_word)); +} +static void dspi_fifo_write (struct fsl_dspi *dspi) { +dspi_setup_accel (dspi); +if (dspi->devtype_data->trans_mode ==DSPI_EOQ_MODE )dspi_eoq_fifo_write (dspi);else
@@ -726,7 +835,7 @@ static int dspi_rxtx(struct fsl_dspi *dspi)int bytes_sent;/* Update total number of bytes that were transferred */ - bytes_sent = dspi->words_in_flight * dspi->bytes_per_word; + bytes_sent = dspi->words_in_flight * dspi->oper_word_size; msg->actual_length += bytes_sent; dspi->progress += bytes_sent /DIV_ROUND_UP (xfer->bits_per_word,8 );
@@ -824,13 +933,14 @@ static int dspi_transfer_one_message(struct spi_controller *ctlr, dspi->rx = transfer->rx_buf; dspi->len = transfer->len; dspi->progress =0 ; -/* Validated transfer specific frame size (defaults applied) */ - dspi->bits_per_word = transfer->bits_per_word; - dspi->bytes_per_word =DIV_ROUND_UP (dspi->bits_per_word,8 );regmap_update_bits (dspi->regmap,SPI_MCR ,SPI_MCR_CLR_TXF |SPI_MCR_CLR_RXF ,SPI_MCR_CLR_TXF |SPI_MCR_CLR_RXF ); +/* +* Static CTAR setup for modes that don't dynamically adjust it +* via dspi_setup_accel (aka for DMA) +*/ regmap_write (dspi->regmap,SPI_CTAR (0 ), dspi->cur_chip->ctar_val |SPI_FRAME_BITS (transfer->bits_per_word));
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 10/12 spi: spi-fsl-dspi: Optimize dspi_setup_accel for lowest interrupt count
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
Currently, a SPI transfer that is not multiple of the highest supported word width (e.g. 4 bytes) will be transmitted as follows (assume a 30-byte buffer transmitted through a 32-bit wide FIFO that is 32 bytes deep):
- First 28 bytes are sent as 7 words of 32 bits each
- Last 2 bytes are sent as 1 word of 16 bits size
But if the dspi_setup_accel function had decided to use a lower oper_bits_per_word value (16 instead of 32), there would have been enough space in the TX FIFO to fit the entire buffer in one go (15 words of 16 bits each).
What we're actually trying to avoid is mixing word sizes within the same run with the TX FIFO, since there is an erratum surrounding this, and invalid data might get transmitted.
So this patch adds special cases for when the remaining length of the buffer can be sent in one go as 8-bit or 16-bit words, otherwise it falls back to the standard logic of sending as many bytes as possible at the highest oper_bits_per_word value possible.
The benefit is that there will be one less CMDFQ/EOQ interrupt to service when the entire buffer is transmitted during a single go, and that will improve the overall latency of the transfer.
--- drivers/spi/spi-fsl-dspi.c | 36 +++++++++++++++++++++++++----------- 1 file changed, 25 insertions(+), 11 deletions(-)
drivers/spi/spi-fsl-dspi.c +25 -11
@@ -775,20 +775,34 @@ static void dspi_fifo_read(struct fsl_dspi *dspi)static void dspi_setup_accel (struct fsl_dspi *dspi) {struct spi_transfer *xfer = dspi->cur_transfer; +bool odd = !!(dspi->len &1 ); -/* Start off with maximum supported by hardware */ -if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) - dspi->oper_bits_per_word =32 ; -else +/* No accel for frames not multiple of 8 bits at the moment */ +if (xfer->bits_per_word %8 ) + goto no_accel; + +if (!odd && dspi->len <= dspi->devtype_data->fifo_size *2 ) { dspi->oper_bits_per_word =16 ; + }else if (odd && dspi->len <= dspi->devtype_data->fifo_size) { + dspi->oper_bits_per_word =8 ; + }else { +/* Start off with maximum supported by hardware */ +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) + dspi->oper_bits_per_word =32 ; +else + dspi->oper_bits_per_word =16 ; -/* And go down only if the buffer can't be sent with words this big */ -do { -if (dspi->len >=DIV_ROUND_UP (dspi->oper_bits_per_word,8 )) -break ; +/* +* And go down only if the buffer can't be sent with +* words this big +*/ +do { +if (dspi->len >=DIV_ROUND_UP (dspi->oper_bits_per_word,8 )) +break ; - dspi->oper_bits_per_word /=2 ; - }while (dspi->oper_bits_per_word >8 ); + dspi->oper_bits_per_word /=2 ; + }while (dspi->oper_bits_per_word >8 ); + }if (xfer->bits_per_word ==8 && dspi->oper_bits_per_word ==32 ) { dspi->dev_to_host = dspi_8on32_dev_to_host;
@@ -800,7 +814,7 @@ static void dspi_setup_accel(struct fsl_dspi *dspi) dspi->dev_to_host = dspi_16on32_dev_to_host; dspi->host_to_dev = dspi_16on32_host_to_dev; }else { -/* No acceleration needed (8<N<=16 on 16, 16<N<=32 on 32) */ +no_accel: dspi->dev_to_host = dspi_native_dev_to_host; dspi->host_to_dev = dspi_native_host_to_dev; dspi->oper_bits_per_word = xfer->bits_per_word;
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Optimize dspi_setup_accel for lowest interrupt count" to the spi tree
The patch
spi: spi-fsl-dspi: Optimize dspi_setup_accel for lowest interrupt count
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
6365504d42d90c68555ee40cdf297a1f187cb4a3Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:42 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Optimize dspi_setup_accel for lowest
interrupt count
Currently, a SPI transfer that is not multiple of the highest supported word width (e.g. 4 bytes) will be transmitted as follows (assume a 30-byte buffer transmitted through a 32-bit wide FIFO that is 32 bytes deep):
- First 28 bytes are sent as 7 words of 32 bits each
- Last 2 bytes are sent as 1 word of 16 bits size
But if the dspi_setup_accel function had decided to use a lower oper_bits_per_word value (16 instead of 32), there would have been enough space in the TX FIFO to fit the entire buffer in one go (15 words of 16 bits each).
What we're actually trying to avoid is mixing word sizes within the same run with the TX FIFO, since there is an erratum surrounding this, and invalid data might get transmitted.
So this patch adds special cases for when the remaining length of the buffer can be sent in one go as 8-bit or 16-bit words, otherwise it falls back to the standard logic of sending as many bytes as possible at the highest oper_bits_per_word value possible.
The benefit is that there will be one less CMDFQ/EOQ interrupt to service when the entire buffer is transmitted during a single go, and that will improve the overall latency of the transfer.
--- drivers/spi/spi-fsl-dspi.c | 36 +++++++++++++++++++++++++----------- 1 file changed, 25 insertions(+), 11 deletions(-)
drivers/spi/spi-fsl-dspi.c +25 -11
@@ -775,20 +775,34 @@ static void dspi_fifo_read(struct fsl_dspi *dspi)static void dspi_setup_accel (struct fsl_dspi *dspi) {struct spi_transfer *xfer = dspi->cur_transfer; +bool odd = !!(dspi->len &1 ); -/* Start off with maximum supported by hardware */ -if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) - dspi->oper_bits_per_word =32 ; -else +/* No accel for frames not multiple of 8 bits at the moment */ +if (xfer->bits_per_word %8 ) + goto no_accel; + +if (!odd && dspi->len <= dspi->devtype_data->fifo_size *2 ) { dspi->oper_bits_per_word =16 ; + }else if (odd && dspi->len <= dspi->devtype_data->fifo_size) { + dspi->oper_bits_per_word =8 ; + }else { +/* Start off with maximum supported by hardware */ +if (dspi->devtype_data->trans_mode ==DSPI_XSPI_MODE ) + dspi->oper_bits_per_word =32 ; +else + dspi->oper_bits_per_word =16 ; -/* And go down only if the buffer can't be sent with words this big */ -do { -if (dspi->len >=DIV_ROUND_UP (dspi->oper_bits_per_word,8 )) -break ; +/* +* And go down only if the buffer can't be sent with +* words this big +*/ +do { +if (dspi->len >=DIV_ROUND_UP (dspi->oper_bits_per_word,8 )) +break ; - dspi->oper_bits_per_word /=2 ; - }while (dspi->oper_bits_per_word >8 ); + dspi->oper_bits_per_word /=2 ; + }while (dspi->oper_bits_per_word >8 ); + }if (xfer->bits_per_word ==8 && dspi->oper_bits_per_word ==32 ) { dspi->dev_to_host = dspi_8on32_dev_to_host;
@@ -800,7 +814,7 @@ static void dspi_setup_accel(struct fsl_dspi *dspi) dspi->dev_to_host = dspi_16on32_dev_to_host; dspi->host_to_dev = dspi_16on32_host_to_dev; }else { -/* No acceleration needed (8<N<=16 on 16, 16<N<=32 on 32) */ +no_accel: dspi->dev_to_host = dspi_native_dev_to_host; dspi->host_to_dev = dspi_native_host_to_dev; dspi->oper_bits_per_word = xfer->bits_per_word;
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 11/12 spi: spi-fsl-dspi: Use EOQ for last word in buffer even for XSPI mode
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
The EOQ mode has a hardware limitation in that it stops the transmission (including the deassertion of the chip select signal) once the host CPU requests end-of-queue for a particular word in the TX FIFO.
And XSPI mode has a limitation in that we need a separate CMD FIFO entry for the last byte in the buffer, where the chip select signal needs to be deasserted. It's not a functional limitation, but it's rather clunky and the fact that we need to halt the pipeline and write a single entry to the TX FIFO whenever a buffer ends brings the throughput down when transmitting small buffers.
So the idea here is to use EOQ's limitation in our favor when using XSPI mode. Stop special-casing that final word in the buffer, and just kill the chip select signal by issuing an EOQ for that last word. Now it can be mixed in with all the other words in the current TX FIFO train.
A small trick here is that we still keep using the XSPI-specific signaling via the CMDTCFQ interrupt in RSER, and not enabling the EOQ interrupt, in order to avoid hardware weirdness (potential races with separate interrupts being raised for CMDTCFQ and EOQ for what is in fact the end of the same transmission). That is just theoretical, but it's good to be cautious, and the EOQ interrupt isn't needed.
--- drivers/spi/spi-fsl-dspi.c | 22 +++++++++++++--------- 1 file changed, 13 insertions(+), 9 deletions(-)
drivers/spi/spi-fsl-dspi.c +13 -9
@@ -656,10 +656,8 @@ static void dspi_pushr_write(struct fsl_dspi *dspi)regmap_write (dspi->regmap,SPI_PUSHR ,dspi_pop_tx_pushr (dspi)); } -static void dspi_pushr_cmd_write (struct fsl_dspi *dspi) +static void dspi_pushr_cmd_write (struct fsl_dspi *dspi,u16 cmd) { -u16 cmd = dspi->tx_cmd; -/* * The only time when the PCS doesn't need continuation after this word * is when it's last. We need to look ahead, because we actually call
@@ -680,8 +678,13 @@ static void dspi_pushr_txdata_write(struct fsl_dspi *dspi, u16 txdata)regmap_write (dspi->regmap_pushr,PUSHR_TX , txdata); } -static void dspi_xspi_write (struct fsl_dspi *dspi,int cnt) +static void dspi_xspi_write (struct fsl_dspi *dspi,int cnt,bool eoq) { +u16 tx_cmd = dspi->tx_cmd; + +if (eoq) + tx_cmd |=SPI_PUSHR_CMD_EOQ ; +/* Update CTARE */ regmap_write(dspi->regmap, SPI_CTARE(0), SPI_FRAME_EBITS(dspi->oper_bits_per_word) |
@@ -691,7 +694,7 @@ static void dspi_xspi_write(struct fsl_dspi *dspi, int cnt)* Write the CMD FIFO entry first, and then the two * corresponding TX FIFO entries (or one...). */ -dspi_pushr_cmd_write (dspi); +dspi_pushr_cmd_write (dspi, tx_cmd);/* Fill TX FIFO with as many transfers as possible */ while (cnt--) {
@@ -707,6 +710,7 @@ static void dspi_xspi_fifo_write(struct fsl_dspi *dspi) {int num_fifo_entries = dspi->devtype_data->fifo_size;int bytes_in_flight; +bool eoq = false;/* In XSPI mode each 32-bit word occupies 2 TX FIFO entries */ if (dspi->oper_word_size ==4 )
@@ -732,11 +736,11 @@ static void dspi_xspi_fifo_write(struct fsl_dspi *dspi)* So send one word less during this go, to force a split and a command * with a single word next time, when CONT will be unset. */ -if (bytes_in_flight == dspi->len && dspi->words_in_flight >1 && - !(dspi->tx_cmd &SPI_PUSHR_CMD_CONT )) - dspi->words_in_flight--; +if (!(dspi->tx_cmd &SPI_PUSHR_CMD_CONT ) && + bytes_in_flight == dspi->len) + eoq = true; -dspi_xspi_write (dspi, dspi->words_in_flight); +dspi_xspi_write (dspi, dspi->words_in_flight, eoq); }static void dspi_eoq_fifo_write (struct fsl_dspi *dspi)
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Use EOQ for last word in buffer even for XSPI mode" to the spi tree
The patch
spi: spi-fsl-dspi: Use EOQ for last word in buffer even for XSPI mode
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
ea93ed4c181bd42d27b49b612d56f4ceb23d1d6cMon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:43 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Use EOQ for last word in buffer even for
XSPI mode
The EOQ mode has a hardware limitation in that it stops the transmission (including the deassertion of the chip select signal) once the host CPU requests end-of-queue for a particular word in the TX FIFO.
And XSPI mode has a limitation in that we need a separate CMD FIFO entry for the last byte in the buffer, where the chip select signal needs to be deasserted. It's not a functional limitation, but it's rather clunky and the fact that we need to halt the pipeline and write a single entry to the TX FIFO whenever a buffer ends brings the throughput down when transmitting small buffers.
So the idea here is to use EOQ's limitation in our favor when using XSPI mode. Stop special-casing that final word in the buffer, and just kill the chip select signal by issuing an EOQ for that last word. Now it can be mixed in with all the other words in the current TX FIFO train.
A small trick here is that we still keep using the XSPI-specific signaling via the CMDTCFQ interrupt in RSER, and not enabling the EOQ interrupt, in order to avoid hardware weirdness (potential races with separate interrupts being raised for CMDTCFQ and EOQ for what is in fact the end of the same transmission). That is just theoretical, but it's good to be cautious, and the EOQ interrupt isn't needed.
--- drivers/spi/spi-fsl-dspi.c | 22 +++++++++++++--------- 1 file changed, 13 insertions(+), 9 deletions(-)
drivers/spi/spi-fsl-dspi.c +13 -9
@@ -656,10 +656,8 @@ static void dspi_pushr_write(struct fsl_dspi *dspi)regmap_write (dspi->regmap,SPI_PUSHR ,dspi_pop_tx_pushr (dspi)); } -static void dspi_pushr_cmd_write (struct fsl_dspi *dspi) +static void dspi_pushr_cmd_write (struct fsl_dspi *dspi,u16 cmd) { -u16 cmd = dspi->tx_cmd; -/* * The only time when the PCS doesn't need continuation after this word * is when it's last. We need to look ahead, because we actually call
@@ -680,8 +678,13 @@ static void dspi_pushr_txdata_write(struct fsl_dspi *dspi, u16 txdata)regmap_write (dspi->regmap_pushr,PUSHR_TX , txdata); } -static void dspi_xspi_write (struct fsl_dspi *dspi,int cnt) +static void dspi_xspi_write (struct fsl_dspi *dspi,int cnt,bool eoq) { +u16 tx_cmd = dspi->tx_cmd; + +if (eoq) + tx_cmd |=SPI_PUSHR_CMD_EOQ ; +/* Update CTARE */ regmap_write(dspi->regmap, SPI_CTARE(0), SPI_FRAME_EBITS(dspi->oper_bits_per_word) |
@@ -691,7 +694,7 @@ static void dspi_xspi_write(struct fsl_dspi *dspi, int cnt)* Write the CMD FIFO entry first, and then the two * corresponding TX FIFO entries (or one...). */ -dspi_pushr_cmd_write (dspi); +dspi_pushr_cmd_write (dspi, tx_cmd);/* Fill TX FIFO with as many transfers as possible */ while (cnt--) {
@@ -707,6 +710,7 @@ static void dspi_xspi_fifo_write(struct fsl_dspi *dspi) {int num_fifo_entries = dspi->devtype_data->fifo_size;int bytes_in_flight; +bool eoq = false;/* In XSPI mode each 32-bit word occupies 2 TX FIFO entries */ if (dspi->oper_word_size ==4 )
@@ -732,11 +736,11 @@ static void dspi_xspi_fifo_write(struct fsl_dspi *dspi)* So send one word less during this go, to force a split and a command * with a single word next time, when CONT will be unset. */ -if (bytes_in_flight == dspi->len && dspi->words_in_flight >1 && - !(dspi->tx_cmd &SPI_PUSHR_CMD_CONT )) - dspi->words_in_flight--; +if (!(dspi->tx_cmd &SPI_PUSHR_CMD_CONT ) && + bytes_in_flight == dspi->len) + eoq = true; -dspi_xspi_write (dspi, dspi->words_in_flight); +dspi_xspi_write (dspi, dspi->words_in_flight, eoq); }static void dspi_eoq_fifo_write (struct fsl_dspi *dspi)
Signature · 1 line
2.20.1
Vladimir Oltean
maintainerPATCH 12/12 spi: spi-fsl-dspi: Take software timestamp in dspi_fifo_write
In reply to Vladimir OlteanFrom: Vladimir Oltean <[email protected]>
Although the SPI system timestamps are supposed to reflect the moment that the peripheral has received a word rather than the moment when the CPU has enqueued that word to the FIFO, in practice it is easier to just record the latter time than the former (with a smaller error).
With the recent migration of TCFQ users from poll back to interrupt mode (this time for XSPI FIFO), it's wiser to keep the interrupt latency outside of the measurement of the PTP system timestamp itself. If there proves to be any constant offset that requires static compensation, that can always be added later. So far that does not appear to be the case at least on the LS1021A-TSN board, where testing shows that the phc2sys offset is able to remain within +/- 200 ns even after 68 hours of testing.
--- drivers/spi/spi-fsl-dspi.c | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-)
drivers/spi/spi-fsl-dspi.c +9 -10
@@ -838,19 +838,18 @@ static void dspi_setup_accel(struct fsl_dspi *dspi)static void dspi_fifo_write (struct fsl_dspi *dspi) { +struct spi_transfer *xfer = dspi->cur_transfer; +struct spi_message *msg = dspi->cur_msg; +int bytes_sent; +dspi_setup_accel (dspi); +spi_take_timestamp_pre (dspi->ctlr, xfer, dspi->progress, !dspi->irq); +if (dspi->devtype_data->trans_mode ==DSPI_EOQ_MODE )dspi_eoq_fifo_write (dspi);else dspi_xspi_fifo_write (dspi); -} - -static int dspi_rxtx (struct fsl_dspi *dspi) -{ -struct spi_transfer *xfer = dspi->cur_transfer; -struct spi_message *msg = dspi->cur_msg; -int bytes_sent;/* Update total number of bytes that were transferred */ bytes_sent = dspi->words_in_flight * dspi->oper_word_size;
@@ -859,16 +858,16 @@ static int dspi_rxtx(struct fsl_dspi *dspi)spi_take_timestamp_post (dspi->ctlr, dspi->cur_transfer, dspi->progress, !dspi->irq); +} +static int dspi_rxtx (struct fsl_dspi *dspi) +{dspi_fifo_read (dspi);if (!dspi->len)/* Success! */ return 0 ; -spi_take_timestamp_pre (dspi->ctlr, dspi->cur_transfer, - dspi->progress, !dspi->irq); -dspi_fifo_write (dspi);return -EINPROGRESS ;
Signature · 1 line
2.17.1
Mark Brown
maintainerApplied "spi: spi-fsl-dspi: Take software timestamp in dspi_fifo_write" to the spi tree
The patch
spi: spi-fsl-dspi: Take software timestamp in dspi_fifo_write
has been applied to the spi tree at
https://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi.git
All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.
You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.
If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.
Please add any relevant lists and maintainers to the CCs when replying to this mail.
Thanks,
Mark
Quoted 1 line
From
e9bac90036d394b01cc7d5297a11d33b8ab92a91Mon Sep 17 00:00:00 2001
From: Vladimir Oltean <[email protected]>
Date: Thu, 5 Mar 2020 00:00:44 +0200
Subject: [PATCH] spi: spi-fsl-dspi: Take software timestamp in dspi_fifo_write
Although the SPI system timestamps are supposed to reflect the moment that the peripheral has received a word rather than the moment when the CPU has enqueued that word to the FIFO, in practice it is easier to just record the latter time than the former (with a smaller error).
With the recent migration of TCFQ users from poll back to interrupt mode (this time for XSPI FIFO), it's wiser to keep the interrupt latency outside of the measurement of the PTP system timestamp itself. If there proves to be any constant offset that requires static compensation, that can always be added later. So far that does not appear to be the case at least on the LS1021A-TSN board, where testing shows that the phc2sys offset is able to remain within +/- 200 ns even after 68 hours of testing.
--- drivers/spi/spi-fsl-dspi.c | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-)
drivers/spi/spi-fsl-dspi.c +9 -10
@@ -838,19 +838,18 @@ static void dspi_setup_accel(struct fsl_dspi *dspi)static void dspi_fifo_write (struct fsl_dspi *dspi) { +struct spi_transfer *xfer = dspi->cur_transfer; +struct spi_message *msg = dspi->cur_msg; +int bytes_sent; +dspi_setup_accel (dspi); +spi_take_timestamp_pre (dspi->ctlr, xfer, dspi->progress, !dspi->irq); +if (dspi->devtype_data->trans_mode ==DSPI_EOQ_MODE )dspi_eoq_fifo_write (dspi);else dspi_xspi_fifo_write (dspi); -} - -static int dspi_rxtx (struct fsl_dspi *dspi) -{ -struct spi_transfer *xfer = dspi->cur_transfer; -struct spi_message *msg = dspi->cur_msg; -int bytes_sent;/* Update total number of bytes that were transferred */ bytes_sent = dspi->words_in_flight * dspi->oper_word_size;
@@ -859,16 +858,16 @@ static int dspi_rxtx(struct fsl_dspi *dspi)spi_take_timestamp_post (dspi->ctlr, dspi->cur_transfer, dspi->progress, !dspi->irq); +} +static int dspi_rxtx (struct fsl_dspi *dspi) +{dspi_fifo_read (dspi);if (!dspi->len)/* Success! */ return 0 ; -spi_take_timestamp_pre (dspi->ctlr, dspi->cur_transfer, - dspi->progress, !dspi->irq); -dspi_fifo_write (dspi);return -EINPROGRESS ;
Signature · 1 line
2.20.1