[igt-dev] [PATCH i-g-t 3/3] tests/xe_waitfence: Add abstime wait subtest
Zbigniew Kempczyński
zbigniew.kempczynski at intel.com
Fri Jun 2 07:45:50 UTC 2023
On Thu, Jun 01, 2023 at 07:38:43PM +0200, Kamil Konieczny wrote:
> Hi Zbigniew,
>
> On 2023-05-31 at 21:30:43 +0200, Zbigniew Kempczyński wrote:
> > User fence wait supports two methods of passing timeout - relative
> > and absolute. Add absolute with passing timeout as nanoseconds.
> ------------------ ^
> Please clarify somewhat description, maybe insert
> "subtest for testing " ?
>
> >
> > Signed-off-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
> > Cc: Kamil Konieczny <kamil.konieczny at linux.intel.com>
> >
> > ---
> > Requires changes in kernel uapi proposed in:
> > https://patchwork.freedesktop.org/series/118670/
> > ---
> > tests/xe/xe_waitfence.c | 38 ++++++++++++++++++++++++++++++++++----
> > 1 file changed, 34 insertions(+), 4 deletions(-)
> >
> > diff --git a/tests/xe/xe_waitfence.c b/tests/xe/xe_waitfence.c
> > index cdfcacdb47..1440ae4e7e 100644
> > --- a/tests/xe/xe_waitfence.c
> > +++ b/tests/xe/xe_waitfence.c
> > @@ -37,12 +37,23 @@ static void do_bind(int fd, uint32_t vm, uint32_t bo, uint64_t offset,
> > xe_vm_bind(fd, vm, bo, offset, addr, size, sync, 1);
> > }
> >
> > +#define MS_TO_NS(ms) (((long)ms) * 1000000)
> > +enum waittype {
> > + RELATIVE,
> > + ABSTIME,
> > +};
> > +
> > /**
> > * SUBTEST: test
> -------------- ^
>
> > - * Description: Check basic waitfences functionality
> > + * Description: Check basic waitfences functionality with timeout
> > + * as relative timeout in nanoseconds
> > + *
> > + * SUBTEST: test-abstime
> -------------- ^
> This name is a little confusing, consider changing it with
> separate patch, maybe before this one ? From description maybe
> it should be "basic" ?
To be honest I would rename this to 'relative' and 'absolute' subtests.
Any objections about that?
Thank you for the review.
--
Zbigniew
>
> Regards,
> Kamil
>
> > + * Description: Check basic waitfences functionality with timeout
> > + * passed as absolute time in nanoseconds
> > */
> > static void
> > -test(int fd)
> > +test(int fd, enum waittype wt)
> > {
> > uint32_t bo_1;
> > uint32_t bo_2;
> > @@ -51,6 +62,7 @@ test(int fd)
> > uint32_t bo_5;
> > uint32_t bo_6;
> > uint32_t bo_7;
> > + long timeout;
> >
> > uint32_t vm = xe_vm_create(fd, 0, 0);
> > bo_1 = xe_bo_create_flags(fd, vm, 0x40000, MY_FLAG);
> > @@ -67,7 +79,22 @@ test(int fd)
> > do_bind(fd, vm, bo_6, 0, 0xc0040000, 0x1c0000, 6);
> > bo_7 = xe_bo_create_flags(fd, vm, 0x10000, MY_FLAG);
> > do_bind(fd, vm, bo_7, 0, 0xeffff0000, 0x10000, 7);
> > - xe_wait_ufence(fd, &wait_fence, 7, NULL, 2000);
> > +
> > + if (wt == RELATIVE) {
> > + timeout = xe_wait_ufence(fd, &wait_fence, 7, NULL, MS_TO_NS(10));
> > + igt_info("wait type: RELATIVE - timeout: %ld, timeout left: %ld\n",
> > + MS_TO_NS(10), timeout);
> > + } else {
> > + struct timespec ts;
> > + long left;
> > +
> > + clock_gettime(CLOCK_MONOTONIC, &ts);
> > + timeout = ts.tv_sec * 1e9 + ts.tv_nsec + (long) MS_TO_NS(10);
> > + left = xe_wait_ufence_abstime(fd, &wait_fence, 7, NULL, timeout);
> > + igt_info("wait type: ABSTIME - timeout: %ld, left: %ld\n",
> > + timeout, left);
> > + }
> > +
> > xe_vm_unbind_sync(fd, vm, 0, 0x200000, 0x40000);
> > xe_vm_unbind_sync(fd, vm, 0, 0xc0000000, 0x40000);
> > xe_vm_unbind_sync(fd, vm, 0, 0x180000000, 0x40000);
> > @@ -94,7 +121,10 @@ igt_main
> > }
> >
> > igt_subtest("test")
> > - test(fd);
> > + test(fd, RELATIVE);
> > +
> > + igt_subtest("test-abstime")
> > + test(fd, ABSTIME);
> >
> > igt_fixture {
> > xe_device_put(fd);
> > --
> > 2.34.1
> >
More information about the igt-dev
mailing list