[PATCH igt 6/9] msm/mapping: Wait for devcore to become available

[Rob Clark]

On Tue, Jul 1, 2025 at 9:20 AM Kamil Konieczny
<kamil.konieczny at linux.intel.com> wrote:
>
> Hi Rob,
> On 2025-06-30 at 11:09:00 -0700, Rob Clark wrote:
> > From: Rob Clark <rob.clark at oss.qualcomm.com>
>
> Is this the same as 1/9 patch?

No, but similar.  Both tests have a similar, but subtly different,
function to read the devcore.  (In the mapping test, we don't have to
read the whole file, but we want to return the first part so the test
can parse it.. in the recovery test we don't need to parse it, but
want to explicitly read the entire devcore.)

BR,
-R

>
> One more nit just spotted, it is useally 'i-g-t' after PATCH:
>
> [PATCH i-g-t 6/9] tests/msm/msm_mapping: Wait for devcore to become available
>
> Regards,
> Kamil
>
> >
> > The devcore could take some time to show up, so add a igt_wait() with
> > timeout so we don't fail the test if the devcore is not immediately
> > available.
> >
> > This addresses a source of flakeyness.
> >
> > Signed-off-by: Rob Clark <rob.clark at oss.qualcomm.com>
> > ---
> >  tests/msm/msm_mapping.c | 16 ++++++++++------
> >  1 file changed, 10 insertions(+), 6 deletions(-)
> >
> > diff --git a/tests/msm/msm_mapping.c b/tests/msm/msm_mapping.c
> > index 846385bb5206..03a9e814c931 100644
> > --- a/tests/msm/msm_mapping.c
> > +++ b/tests/msm/msm_mapping.c
> > @@ -42,14 +42,18 @@
> >   */
> >
> >  static char *
> > -get_and_clear_devcore(void)
> > +get_and_clear_devcore(int timeout_ms)
> >  {
> >       glob_t glob_buf = {0};
> >       char *buf = NULL;
> > -     int ret, fd;
> > +     int fd;
> >
> > -     ret = glob("/sys/class/devcoredump/devcd*/data", GLOB_NOSORT, NULL, &glob_buf);
> > -     if ((ret == GLOB_NOMATCH) || !glob_buf.gl_pathc)
> > +     /* The devcore shows up asynchronously, so it might not be
> > +      * immediately available:
> > +      */
> > +     if (!igt_wait(glob("/sys/class/devcoredump/devcd*/data",
> > +                        GLOB_NOSORT, NULL, &glob_buf) != GLOB_NOMATCH,
> > +                   timeout_ms, 100))
> >               return NULL;
> >
> >       fd = open(glob_buf.gl_pathv[0], O_RDWR);
> > @@ -175,7 +179,7 @@ do_mapping_test(struct msm_pipe *pipe, const char *buffername, bool write)
> >       int fence_fd, ret;
> >
> >       /* Clear any existing devcore's: */
> > -     while ((devcore = get_and_clear_devcore())) {
> > +     while ((devcore = get_and_clear_devcore(0))) {
> >               free(devcore);
> >       }
> >
> > @@ -208,7 +212,7 @@ do_mapping_test(struct msm_pipe *pipe, const char *buffername, bool write)
> >       /* And now we should have gotten a devcore from the iova fault
> >        * triggered by the read or write:
> >        */
> > -     devcore = get_and_clear_devcore();
> > +     devcore = get_and_clear_devcore(1000);
> >       igt_fail_on(!devcore);
> >
> >       /* Make sure the devcore is from iova fault: */
> > --
> > 2.50.0
> >