| Message ID | 20190801204710.27309-1-uwe@kleine-koenig.org (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | 3c0448d507d42605ab6811bcd61c1cd94390d2d8 |
| Headers | show |
| Series | spi: atmel: add tracing to custom .transfer_one_message callback | expand |
On Thu, Aug 01, 2019 at 10:47:10PM +0200, Uwe Kleine-König wrote: > Driver specific implementations for .transfer_one_message need to call > the tracing stuff themself. This is necessary to make spi tracing > actually useful. Have you tested this for modular builds? IIRC it breaks unfortunately.
Hey Mark, On Fri, Aug 02, 2019 at 12:40:11AM +0100, Mark Brown wrote: > On Thu, Aug 01, 2019 at 10:47:10PM +0200, Uwe Kleine-König wrote: > > Driver specific implementations for .transfer_one_message need to call > > the tracing stuff themself. This is necessary to make spi tracing > > actually useful. > > Have you tested this for modular builds? IIRC it breaks unfortunately. Works for me. Is this just a suspicion or do you have a breaking .config? If the latter, the spi-mxs driver probably has the same problem. When I sent the analogous patch for mxs this was indeed a problem, but this was fixed in ca1438dcb34c7fcad63b6ce14ea63a870b92a69b. Best regards Uwe
On Fri, Aug 02, 2019 at 08:20:18AM +0200, Uwe Kleine-König wrote: > On Fri, Aug 02, 2019 at 12:40:11AM +0100, Mark Brown wrote: > > On Thu, Aug 01, 2019 at 10:47:10PM +0200, Uwe Kleine-König wrote: > > Have you tested this for modular builds? IIRC it breaks unfortunately. > Works for me. Is this just a suspicion or do you have a breaking > .config? If the latter, the spi-mxs driver probably has the same > problem. When I sent the analogous patch for mxs this was indeed a > problem, but this was fixed in > ca1438dcb34c7fcad63b6ce14ea63a870b92a69b. It was a problem I remember happening in the past.
diff --git a/drivers/spi/spi-atmel.c b/drivers/spi/spi-atmel.c index f00b367523cd..acf318e7330c 100644 --- a/drivers/spi/spi-atmel.c +++ b/drivers/spi/spi-atmel.c @@ -23,6 +23,7 @@ #include <linux/gpio/consumer.h> #include <linux/pinctrl/consumer.h> #include <linux/pm_runtime.h> +#include <trace/events/spi.h> /* SPI register offsets */ #define SPI_CR 0x0000 @@ -1409,9 +1410,13 @@ static int atmel_spi_transfer_one_message(struct spi_master *master, msg->actual_length = 0; list_for_each_entry(xfer, &msg->transfers, transfer_list) { + trace_spi_transfer_start(msg, xfer); + ret = atmel_spi_one_transfer(master, msg, xfer); if (ret) goto msg_done; + + trace_spi_transfer_stop(msg, xfer); } if (as->use_pdc)
Driver specific implementations for .transfer_one_message need to call the tracing stuff themself. This is necessary to make spi tracing actually useful. Signed-off-by: Uwe Kleine-König <uwe@kleine-koenig.org> --- drivers/spi/spi-atmel.c | 5 +++++ 1 file changed, 5 insertions(+)