From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga03.intel.com (mga03.intel.com [134.134.136.65]) by gabe.freedesktop.org (Postfix) with ESMTPS id 2949910E9FD for ; Fri, 13 Jan 2023 14:27:45 +0000 (UTC) Message-ID: Date: Fri, 13 Jan 2023 14:27:39 +0000 MIME-Version: 1.0 Content-Language: en-GB To: Niranjana Vishwanathapura , igt-dev@lists.freedesktop.org References: <20221212231254.2303-1-niranjana.vishwanathapura@intel.com> <20221212231254.2303-20-niranjana.vishwanathapura@intel.com> From: Matthew Auld In-Reply-To: <20221212231254.2303-20-niranjana.vishwanathapura@intel.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [igt-dev] [PATCH i-g-t v9 19/19] tests/i915/vm_bind: Add userptr invalidation test List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: tvrtko.ursulin@intel.com, thomas.hellstrom@intel.com, daniel.vetter@intel.com, petri.latvala@intel.com Errors-To: igt-dev-bounces@lists.freedesktop.org Sender: "igt-dev" List-ID: On 12/12/2022 23:12, Niranjana Vishwanathapura wrote: > Add vma-merge-execbuf3 subtest to gem_userptr_blits to > validate userptr invalidations with VM_BIND+execbuf3. > > NOTE: > The test is derived from vma-merge subtest. There is a known > fence timeout issue with vma-merge subtest which the > vma-merge-execbuf3 inherits, but the test still passes none > the less validates userptr invalidation of persistent mappings > and rebinding the same in the subsequent submission. > > Signed-off-by: Niranjana Vishwanathapura > --- > tests/i915/gem_userptr_blits.c | 71 ++++++++++++++++++++++++++++++++++ > 1 file changed, 71 insertions(+) > > diff --git a/tests/i915/gem_userptr_blits.c b/tests/i915/gem_userptr_blits.c > index 698508669..0ec54bd21 100644 > --- a/tests/i915/gem_userptr_blits.c > +++ b/tests/i915/gem_userptr_blits.c > @@ -60,7 +60,10 @@ > > #include "i915/gem.h" > #include "i915/gem_create.h" > +#include "i915/gem_vm.h" > +#include "i915/i915_vm_bind.h" > #include "igt.h" > +#include "igt_syncobj.h" > #include "igt_sysfs.h" > #include "igt_types.h" > #include "sw_sync.h" > @@ -740,6 +743,69 @@ static void test_vma_merge(int i915) > put_ahnd(ahnd); > } > > +static void test_vma_merge_execbuf3(int i915) > +{ > + struct drm_i915_gem_context_param param = { > + .param = I915_CONTEXT_PARAM_RECOVERABLE, > + .value = 0, > + }; > + const size_t sz = 2 * hugepagesize(); > + uint32_t handle, vm_id; > + const intel_ctx_t *ctx; > + igt_spin_t *spin; > + uint64_t ahnd; > + void *addr; > + > + vm_id = gem_vm_create_in_vm_bind_mode(i915); > + ctx = intel_ctx_create_all_physical(i915); > + param.ctx_id = ctx->id; > + gem_context_set_param(i915, ¶m); > + gem_context_set_vm(i915, ctx->id, vm_id); > + (void)gem_context_get_vm(i915, ctx->id); > + > + ahnd = get_simple_ahnd(i915, ctx->id); > + igt_require(ahnd); > + addr = mmap(NULL, sz, PROT_READ | PROT_WRITE, > + MAP_SHARED | MAP_ANONYMOUS, -1, 0); > + > + gem_userptr(i915, addr + sz / 2, 4096, 0, userptr_flags, &handle); > + > + spin = igt_spin_new(i915, .ctx = ctx, .vm_id = vm_id, > + .ahnd = ahnd, .dependency = handle, > + .flags = IGT_SPIN_EXECBUF3); > + > + for (size_t x = 0; x < sz; x += 4096) { > + if (x == sz / 2) > + continue; > + > + igt_assert(mmap(addr + x, 4096, PROT_READ | PROT_WRITE, > + MAP_FIXED | MAP_SHARED | MAP_ANON, -1, 0) != > + MAP_FAILED); > + } > + > + igt_spin_end(spin); > + igt_spin_reset(spin); > + igt_assert(syncobj_timeline_wait(i915, spin->out_syncobj, NULL, spin->nengines, > + gettime_ns() + (2 * NSEC_PER_SEC), > + DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT, NULL)); > + > + /* Another submission will cause rebind of invalidated mappings */ > + gem_execbuf3(i915, &spin->execbuf3); > + igt_spin_end(spin); > + igt_assert(syncobj_timeline_wait(i915, &spin->out_syncobj[spin->nengines - 1], > + NULL, 1, gettime_ns() + (2 * NSEC_PER_SEC), > + DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT, NULL)); > + gem_close(i915, handle); > + > + munmap(addr, sz); > + > + igt_spin_free(i915, spin); > + put_ahnd(ahnd); > + > + intel_ctx_destroy(i915, ctx); > + gem_vm_destroy(i915, vm_id); > +} > + > static void test_huge_split(int i915) > { > const size_t sz = 2 * hugepagesize(); > @@ -2628,6 +2694,11 @@ igt_main_args("c:", NULL, help_str, opt_handler, NULL) > igt_subtest("vma-merge") > test_vma_merge(fd); > > + igt_subtest("vma-merge-execbuf3") { I guess needs test description somewhere, since this is new subtest. Reviewed-by: Matthew Auld > + igt_require(i915_vm_bind_version(fd) == 1); > + test_vma_merge_execbuf3(fd); > + } > + > igt_subtest("huge-split") > test_huge_split(fd); > }