From: Matthew Brost <matthew.brost@intel.com>
To: Kamil Konieczny <kamil.konieczny@linux.intel.com>
Cc: <igt-dev@lists.freedesktop.org>
Subject: Re: [PATCH v2 1/2] tests/xe/xe_vm: Add section to test zero number of VM binds
Date: Tue, 23 Jul 2024 18:41:32 +0000 [thread overview]
Message-ID: <Zp/5XP8eoNJ1CUyw@DUT025-TGLU.fm.intel.com> (raw)
In-Reply-To: <20240723124611.fdk6rvshlgnckrkc@kamilkon-DESK.igk.intel.com>
On Tue, Jul 23, 2024 at 02:46:11PM +0200, Kamil Konieczny wrote:
> Hi Matthew,
> On 2024-07-19 at 12:41:06 -0700, Matthew Brost wrote:
>
> a nit about a subject, you wrote:
>
> [PATCH v2 1/2] tests/xe/xe_vm: Add section to test zero number of VM binds
>
> imho this could be:
>
> [PATCH v2 1/2] tests/intel/xe_vm: Test zero number of VM binds
>
Will fix.
> > Verify the bind queue ordering works wrt out-sync signaling when zero
> > number of binds passed to VM bind IOCTL.
> >
> > v2:
> > - Fix build error
> >
> > Signed-off-by: Matthew Brost <matthew.brost@intel.com>
> > ---
> > lib/xe/xe_ioctl.c | 13 ++++++++++
> > lib/xe/xe_ioctl.h | 2 ++
> > tests/intel/xe_vm.c | 63 +++++++++++++++++++++++++++++++++++++++++++++
> > 3 files changed, 78 insertions(+)
> >
> > diff --git a/lib/xe/xe_ioctl.c b/lib/xe/xe_ioctl.c
> > index ae43ffd15e..27a4a4e70d 100644
> > --- a/lib/xe/xe_ioctl.c
> > +++ b/lib/xe/xe_ioctl.c
> > @@ -154,6 +154,19 @@ void xe_vm_bind_async(int fd, uint32_t vm, uint32_t exec_queue, uint32_t bo,
> > num_syncs, 0, 0);
> > }
> >
>
> You should document each new public lib function.
>
> > +void xe_vm_bind_zero(int fd, uint32_t vm, uint32_t exec_queue,
> > + struct drm_xe_sync *sync, uint32_t num_syncs)
> > +{
> > + struct drm_xe_vm_bind bind = {
> > + .vm_id = vm,
> > + .num_syncs = num_syncs,
> > + .syncs = (uintptr_t)sync,
> > + .exec_queue_id = exec_queue,
> > + };
> > +
> > + igt_assert_eq(igt_ioctl(fd, DRM_IOCTL_XE_VM_BIND, &bind), 0);
> > +}
> > +
>
> Why new lib function? Why not just using __xe_vm_bind?
>
Maybe I'll just move this to xe_vm.c as a static function as I don't any
other test would use this.
Matt
> Regards,
> Kamil
>
> > void xe_vm_bind_async_flags(int fd, uint32_t vm, uint32_t exec_queue, uint32_t bo,
> > uint64_t offset, uint64_t addr, uint64_t size,
> > struct drm_xe_sync *sync, uint32_t num_syncs,
> > diff --git a/lib/xe/xe_ioctl.h b/lib/xe/xe_ioctl.h
> > index b27c0053f0..6dfed1b330 100644
> > --- a/lib/xe/xe_ioctl.h
> > +++ b/lib/xe/xe_ioctl.h
> > @@ -35,6 +35,8 @@ void xe_vm_prefetch_async(int fd, uint32_t vm, uint32_t exec_queue,
> > void xe_vm_bind_async(int fd, uint32_t vm, uint32_t exec_queue, uint32_t bo,
> > uint64_t offset, uint64_t addr, uint64_t size,
> > struct drm_xe_sync *sync, uint32_t num_syncs);
> > +void xe_vm_bind_zero(int fd, uint32_t vm, uint32_t exec_queue,
> > + struct drm_xe_sync *sync, uint32_t num_syncs);
> > void xe_vm_bind_userptr_async(int fd, uint32_t vm, uint32_t exec_queue,
> > uint64_t userptr, uint64_t addr, uint64_t size,
> > struct drm_xe_sync *sync, uint32_t num_syncs);
> > diff --git a/tests/intel/xe_vm.c b/tests/intel/xe_vm.c
> > index a088bd5db8..ede75f72aa 100644
> > --- a/tests/intel/xe_vm.c
> > +++ b/tests/intel/xe_vm.c
> > @@ -868,6 +868,63 @@ static void xe_vm_bind_array_err(int fd, uint32_t vm, uint32_t exec_queue,
> > do_ioctl_err(fd, DRM_IOCTL_XE_VM_BIND, &bind, err);
> > }
> >
> > +/**
> > + * SUBTEST: zero-binds
> > + * Description: Test zero binds in IOCTL
> > + * Functionality: bind engines
> > + * Run type: BAT
> > + */
> > +
> > +static void test_zero_binds(int fd, struct drm_xe_engine_class_instance *eci)
> > +{
> > + uint32_t vm;
> > + uint64_t addr = 0x1a0000;
> > + struct drm_xe_sync sync[1] = {
> > + { .flags = DRM_XE_SYNC_TYPE_SYNCOBJ | DRM_XE_SYNC_FLAG_SIGNAL, },
> > + };
> > + size_t bo_size;
> > + uint32_t bo = 0;
> > + struct xe_cork cork;
> > +
> > + vm = xe_vm_create(fd, 0, 0);
> > + bo_size = ALIGN(xe_cs_prefetch_size(fd),
> > + xe_get_default_alignment(fd));
> > + bo = xe_bo_create(fd, vm, bo_size,
> > + vram_if_possible(fd, eci->gt_id),
> > + DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
> > +
> > + xe_cork_init(fd, eci, &cork);
> > + xe_cork_wait_started(&cork);
> > +
> > + /* Initial bind behind cork */
> > + sync[0].handle = cork.syncobj;
> > + sync[0].flags &= ~DRM_XE_SYNC_FLAG_SIGNAL;
> > + xe_vm_bind_async(fd, vm, 0, bo, 0, addr, bo_size, sync, 1);
> > +
> > + /* Bind /w num_binds == 0 */
> > + sync[0].handle = syncobj_create(fd, 0);
> > + sync[0].flags |= DRM_XE_SYNC_FLAG_SIGNAL;
> > + xe_vm_bind_zero(fd, vm, 0, sync, 1);
> > +
> > + /* Let jobs runs for a bit */
> > + usleep(100000);
> > +
> > + /* both binds are waiting */
> > + igt_assert(!syncobj_wait(fd, &sync[0].handle, 1, 0, 0, NULL));
> > +
> > + /* Release cork */
> > + xe_cork_end(&cork);
> > + xe_cork_wait_done(&cork);
> > + xe_cork_fini(&cork);
> > +
> > + /* both binds are done */
> > + igt_assert(syncobj_wait(fd, &sync[0].handle, 1, INT64_MAX, 0, NULL));
> > +
> > + syncobj_destroy(fd, sync[0].handle);
> > + gem_close(fd, bo);
> > + xe_vm_destroy(fd, vm);
> > +}
> > +
> > #define BIND_ARRAY_BIND_EXEC_QUEUE_FLAG (0x1 << 0)
> > #define BIND_ARRAY_ENOBUFS_FLAG (0x1 << 1)
> >
> > @@ -2416,6 +2473,12 @@ igt_main
> > xe_for_each_engine(fd, hwe)
> > test_bind_execqueues_independent(fd, hwe, CONFLICT);
> >
> > + igt_subtest("zero-binds")
> > + xe_for_each_engine(fd, hwe) {
> > + test_zero_binds(fd, hwe);
> > + break;
> > + }
> > +
> > igt_subtest("bind-array-twice")
> > xe_for_each_engine(fd, hwe)
> > test_bind_array(fd, hwe, 2, 0x1a0000, 0, 0);
> > --
> > 2.34.1
> >
prev parent reply other threads:[~2024-07-23 18:43 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-07-19 19:41 [PATCH v2 1/2] tests/xe/xe_vm: Add section to test zero number of VM binds Matthew Brost
2024-07-19 19:41 ` [PATCH v2 2/2] tests/xe/xe_exec_basic: Add section to test zero number of BB in exec IOCTL Matthew Brost
2024-07-23 12:27 ` Kamil Konieczny
2024-07-23 18:39 ` Matthew Brost
2024-07-19 20:33 ` ✗ Fi.CI.BUILD: failure for series starting with [v2,1/2] tests/xe/xe_vm: Add section to test zero number of VM binds Patchwork
2024-07-23 7:55 ` [PATCH v2 1/2] " Kamil Konieczny
2024-07-23 12:46 ` Kamil Konieczny
2024-07-23 18:41 ` Matthew Brost [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=Zp/5XP8eoNJ1CUyw@DUT025-TGLU.fm.intel.com \
--to=matthew.brost@intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=kamil.konieczny@linux.intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox