Re: [PATCH v2 04/15] spi: Replace open coded spi_controller_xfer_timeout()

Il 10/07/23 17:49, Andy Shevchenko ha scritto:

Since the new spi_controller_xfer_timeout() helper appeared, we may replace open coded variant in spi_transfer_wait().

Signed-off-by: Andy Shevchenko andriy.shevchenko@linux.intel.com

drivers/spi/spi.c | 25 ++----------------------- include/linux/spi/spi.h | 6 +++++- 2 files changed, 7 insertions(+), 24 deletions(-)

diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c index 125dea8fae00..c99ee4164f11 100644 --- a/drivers/spi/spi.c +++ b/drivers/spi/spi.c @@ -1342,8 +1342,7 @@ static int spi_transfer_wait(struct spi_controller *ctlr, { struct spi_statistics __percpu *statm = ctlr->pcpu_statistics; struct spi_statistics __percpu *stats = msg->spi->pcpu_statistics;

• u32 speed_hz = xfer->speed_hz;
• unsigned long long ms;

• unsigned long ms;

  if (spi_controller_is_slave(ctlr)) { if (wait_for_completion_interruptible(&ctlr->xfer_completion)) {

@@ -1351,29 +1350,9 @@ static int spi_transfer_wait(struct spi_controller *ctlr, return -EINTR; } } else {

• if (!speed_hz)
  

• 	speed_hz = 100000;
  

• /*
  

•  * For each byte we wait for 8 cycles of the SPI clock.
  

•  * Since speed is defined in Hz and we want milliseconds,
  

•  * use respective multiplier, but before the division,
  

•  * otherwise we may get 0 for short transfers.
  

•  */
  

• ms = 8LL * MSEC_PER_SEC * xfer->len;
  

• do_div(ms, speed_hz);
  

• /*
  

•  * Increase it twice and add 200 ms tolerance, use
  

•  * predefined maximum in case of overflow.
  

•  */
  

• ms += ms + 200;
  

• if (ms > UINT_MAX)
  

• 	ms = UINT_MAX;
  

• ms = spi_controller_xfer_timeout(ctlr, xfer);
  

I agree on using helpers, but the logic is slightly changing here: yes it is unlikely (and also probably useless) to get ms == UINT_MAX, but the helper is limiting the maximum timeout value to 500mS, which may not work for some slow controllers/devices.

This should get validated on more than a few platforms, and I'm not sure that this kind of validation would be "fast" to get... so, probably the best thing to do here is to add a warning in case the timeout exceeds 500mS, print the actual value, keep it like this for a kernel version or two and check reports: that would allow to understand what a safe maximum timeout value could be.

Aside from that, I wouldn't drop those nice comments explaining how/why the timeout is calculated: I know how, but not everyone knows in advance.

Regards, Angelo

ms = wait_for_completion_timeout(&ctlr->xfer_completion,
				 msecs_to_jiffies(ms));

• if (ms == 0) { SPI_STATISTICS_INCREMENT_FIELD(statm, timedout); SPI_STATISTICS_INCREMENT_FIELD(stats, timedout);

diff --git a/include/linux/spi/spi.h b/include/linux/spi/spi.h index 32c94eae8926..0ce1cb18a076 100644 --- a/include/linux/spi/spi.h +++ b/include/linux/spi/spi.h @@ -1270,12 +1270,16 @@ static inline bool spi_is_bpw_supported(struct spi_device *spi, u32 bpw)

• that it would take on a single data line and take twice this amount of time
• with a minimum of 500ms to avoid false positives on loaded systems.

• • Assume speed to be 100 kHz if it's not defined at the time of invocation.
• • Returns: Transfer timeout value in milliseconds.
  */ static inline unsigned int spi_controller_xfer_timeout(struct spi_controller *ctlr, struct spi_transfer *xfer) {

• return max(xfer->len * 8 * 2 / (xfer->speed_hz / 1000), 500U);

• u32 speed_hz = xfer->speed_hz ?: 100000;

• return max(xfer->len * 8 * 2 / (speed_hz / 1000), 500U); }

  /*---------------------------------------------------------------------------*/