* [PATCH i-g-t] tests/amdgpu: move bad mem compute test to amd deadlock @ 2024-08-05 6:52 Jesse Zhang 2024-08-05 8:28 ` ✗ Fi.CI.BUILD: failure for " Patchwork 2024-08-05 10:08 ` [PATCH i-g-t] " Kamil Konieczny 0 siblings, 2 replies; 3+ messages in thread From: Jesse Zhang @ 2024-08-05 6:52 UTC (permalink / raw) To: igt-dev Cc: Vitaly Prosyak, Alex Deucher, Christian Koenig, Kamil Konieczny, Jesse Zhang 1. remove duplicate test case. 2. Since the bad memory test only has page faults and does not trigger job timeouts on mi300. So temporarily moved this test to amd deadlock. Signed-off-by: Jesse Zhang <jesse.zhang@amd.com> --- tests/amdgpu/amd_deadlock.c | 9 +++++++++ tests/amdgpu/amd_queue_reset.c | 11 +---------- 2 files changed, 10 insertions(+), 10 deletions(-) diff --git a/tests/amdgpu/amd_deadlock.c b/tests/amdgpu/amd_deadlock.c index 01185bfd3..6996855cb 100644 --- a/tests/amdgpu/amd_deadlock.c +++ b/tests/amdgpu/amd_deadlock.c @@ -94,6 +94,15 @@ igt_main } } + igt_describe("Test-GPU-reset-by-access-compute-illegal-mem-addr"); + igt_subtest_with_dynamic("amdgpu-compute-illegal-mem-access") { + if (arr_cap[AMD_IP_COMPUTE]) { + igt_dynamic_f("amdgpu-illegal-mem-access") + bad_access_helper(device, CMD_STREAM_TRANS_BAD_MEM_ADDRESS, + AMDGPU_HW_IP_COMPUTE); + } + } + igt_describe("Test-GPU-reset-by-flooding-compute-ring-with-jobs"); igt_subtest_with_dynamic("amdgpu-deadlock-compute") { if (arr_cap[AMD_IP_COMPUTE]) { diff --git a/tests/amdgpu/amd_queue_reset.c b/tests/amdgpu/amd_queue_reset.c index fb05aee35..df96c9f76 100644 --- a/tests/amdgpu/amd_queue_reset.c +++ b/tests/amdgpu/amd_queue_reset.c @@ -893,7 +893,6 @@ igt_main unsigned int arr_err[] = { CMD_STREAM_EXEC_INVALID_PACKET_LENGTH, CMD_STREAM_EXEC_INVALID_OPCODE, - CMD_STREAM_TRANS_BAD_MEM_ADDRESS, //CMD_STREAM_TRANS_BAD_MEM_ADDRESS_BY_SYNC,TODO not job timeout, debug why for n31 //CMD_STREAM_TRANS_BAD_REG_ADDRESS, TODO amdgpu: device lost from bus! for n31 BACKEND_SE_GC_SHADER_INVALID_PROGRAM_ADDR, @@ -975,14 +974,6 @@ igt_main } } - igt_describe("Stressful-and-multiple-cs-of-bad and good mem-operations-using-multiple-processes"); - igt_subtest_with_dynamic("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS") { - if (arr_cap[ip_test] && get_next_rings(ring_id_good, info.available_rings, &ring_id_good, &ring_id_bad)) { - igt_dynamic_f("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS") - set_next_test_to_run(sh_mem, CMD_STREAM_TRANS_BAD_MEM_ADDRESS, - ip_background, ip_test, ring_id_good, ring_id_bad); - } - } /* TODO not job timeout, debug why for nv32 *igt_describe("Stressful-and-multiple-cs-of-bad and good mem-sync-operations-using-multiple-processes"); *igt_subtest_with_dynamic("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS_BY_SYNC") { @@ -997,7 +988,7 @@ igt_main *igt_describe("Stressful-and-multiple-cs-of-bad and good reg-operations-using-multiple-processes"); *igt_subtest_with_dynamic("amdgpu-compute-CMD_STREAM_TRANS_BAD_REG_ADDRESS") { * if (arr_cap[ip_test] && get_next_rings(ring_id_good, info.available_rings, &ring_id_good, &ring_id_bad)) { - * igt_dynamic_f("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS_BY_SYNC") + * igt_dynamic_f("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS") * set_next_test_to_run(sh_mem, CMD_STREAM_TRANS_BAD_REG_ADDRESS, * ip_background, ip_test, ring_id_good, ring_id_bad); * } -- 2.25.1 ^ permalink raw reply related [flat|nested] 3+ messages in thread
* ✗ Fi.CI.BUILD: failure for tests/amdgpu: move bad mem compute test to amd deadlock 2024-08-05 6:52 [PATCH i-g-t] tests/amdgpu: move bad mem compute test to amd deadlock Jesse Zhang @ 2024-08-05 8:28 ` Patchwork 2024-08-05 10:08 ` [PATCH i-g-t] " Kamil Konieczny 1 sibling, 0 replies; 3+ messages in thread From: Patchwork @ 2024-08-05 8:28 UTC (permalink / raw) To: Jesse Zhang; +Cc: igt-dev == Series Details == Series: tests/amdgpu: move bad mem compute test to amd deadlock URL : https://patchwork.freedesktop.org/series/136877/ State : failure == Summary == Applying: tests/amdgpu: move bad mem compute test to amd deadlock Using index info to reconstruct a base tree... A tests/amdgpu/amd_queue_reset.c Falling back to patching base and 3-way merge... CONFLICT (modify/delete): tests/amdgpu/amd_queue_reset.c deleted in HEAD and modified in tests/amdgpu: move bad mem compute test to amd deadlock. Version tests/amdgpu: move bad mem compute test to amd deadlock of tests/amdgpu/amd_queue_reset.c left in tree. Patch failed at 0001 tests/amdgpu: move bad mem compute test to amd deadlock When you have resolved this problem, run "git am --continue". If you prefer to skip this patch, run "git am --skip" instead. To restore the original branch and stop patching, run "git am --abort". ^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH i-g-t] tests/amdgpu: move bad mem compute test to amd deadlock 2024-08-05 6:52 [PATCH i-g-t] tests/amdgpu: move bad mem compute test to amd deadlock Jesse Zhang 2024-08-05 8:28 ` ✗ Fi.CI.BUILD: failure for " Patchwork @ 2024-08-05 10:08 ` Kamil Konieczny 1 sibling, 0 replies; 3+ messages in thread From: Kamil Konieczny @ 2024-08-05 10:08 UTC (permalink / raw) To: igt-dev; +Cc: Jesse Zhang, Vitaly Prosyak, Alex Deucher, Christian Koenig Hi Jesse, On 2024-08-05 at 14:52:18 +0800, Jesse Zhang wrote: > 1. remove duplicate test case. > 2. Since the bad memory test only has page faults > and does not trigger job timeouts on mi300. > So temporarily moved this test to amd deadlock. > > Signed-off-by: Jesse Zhang <jesse.zhang@amd.com> > --- > tests/amdgpu/amd_deadlock.c | 9 +++++++++ > tests/amdgpu/amd_queue_reset.c | 11 +---------- amd_queue_reset was reverted, please prepare new version with your changes, also see note about subtests naming convetion below. > 2 files changed, 10 insertions(+), 10 deletions(-) > > diff --git a/tests/amdgpu/amd_deadlock.c b/tests/amdgpu/amd_deadlock.c > index 01185bfd3..6996855cb 100644 > --- a/tests/amdgpu/amd_deadlock.c > +++ b/tests/amdgpu/amd_deadlock.c > @@ -94,6 +94,15 @@ igt_main > } > } > > + igt_describe("Test-GPU-reset-by-access-compute-illegal-mem-addr"); > + igt_subtest_with_dynamic("amdgpu-compute-illegal-mem-access") { Why dynamic here? Dynamic tests are used for not-known number/names for its dynamic sub-subtests, not whe you do have static number of them. > + if (arr_cap[AMD_IP_COMPUTE]) { > + igt_dynamic_f("amdgpu-illegal-mem-access") Do not do such if(...) dynamic-name, imho better skip: igt_subtest("amdgpu-compute-illegal-mem-access") { igt_skip_on_f(!arr_cap[AMD_IP_COMPUTE], "message why skipping\n"); bad_access_helper(device, CMD_STREAM_TRANS_BAD_MEM_ADDRESS, AMDGPU_HW_IP_COMPUTE); } > + bad_access_helper(device, CMD_STREAM_TRANS_BAD_MEM_ADDRESS, > + AMDGPU_HW_IP_COMPUTE); > + } > + } > + > igt_describe("Test-GPU-reset-by-flooding-compute-ring-with-jobs"); > igt_subtest_with_dynamic("amdgpu-deadlock-compute") { > if (arr_cap[AMD_IP_COMPUTE]) { > diff --git a/tests/amdgpu/amd_queue_reset.c b/tests/amdgpu/amd_queue_reset.c > index fb05aee35..df96c9f76 100644 > --- a/tests/amdgpu/amd_queue_reset.c > +++ b/tests/amdgpu/amd_queue_reset.c > @@ -893,7 +893,6 @@ igt_main > unsigned int arr_err[] = { > CMD_STREAM_EXEC_INVALID_PACKET_LENGTH, > CMD_STREAM_EXEC_INVALID_OPCODE, > - CMD_STREAM_TRANS_BAD_MEM_ADDRESS, > //CMD_STREAM_TRANS_BAD_MEM_ADDRESS_BY_SYNC,TODO not job timeout, debug why for n31 > //CMD_STREAM_TRANS_BAD_REG_ADDRESS, TODO amdgpu: device lost from bus! for n31 > BACKEND_SE_GC_SHADER_INVALID_PROGRAM_ADDR, > @@ -975,14 +974,6 @@ igt_main > } > } > > - igt_describe("Stressful-and-multiple-cs-of-bad and good mem-operations-using-multiple-processes"); > - igt_subtest_with_dynamic("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS") { > - if (arr_cap[ip_test] && get_next_rings(ring_id_good, info.available_rings, &ring_id_good, &ring_id_bad)) { > - igt_dynamic_f("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS") You cannot just repeat a name in a dynamic subtest, each name should be uniqe, I would suggest to rename it to something like: char name[32]; fill_in_name(name, sizeof(name) - 1, ring_id_good, ring_id_bad); igt_dynamic_f(name); So dynamic subtests should have dynamic names, not a static ones. Regards, Kamil > - set_next_test_to_run(sh_mem, CMD_STREAM_TRANS_BAD_MEM_ADDRESS, > - ip_background, ip_test, ring_id_good, ring_id_bad); > - } > - } > /* TODO not job timeout, debug why for nv32 > *igt_describe("Stressful-and-multiple-cs-of-bad and good mem-sync-operations-using-multiple-processes"); > *igt_subtest_with_dynamic("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS_BY_SYNC") { > @@ -997,7 +988,7 @@ igt_main > *igt_describe("Stressful-and-multiple-cs-of-bad and good reg-operations-using-multiple-processes"); > *igt_subtest_with_dynamic("amdgpu-compute-CMD_STREAM_TRANS_BAD_REG_ADDRESS") { > * if (arr_cap[ip_test] && get_next_rings(ring_id_good, info.available_rings, &ring_id_good, &ring_id_bad)) { > - * igt_dynamic_f("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS_BY_SYNC") > + * igt_dynamic_f("amdgpu-compute-CMD_STREAM_TRANS_BAD_MEM_ADDRESS") > * set_next_test_to_run(sh_mem, CMD_STREAM_TRANS_BAD_REG_ADDRESS, > * ip_background, ip_test, ring_id_good, ring_id_bad); > * } > -- > 2.25.1 > ^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2024-08-05 10:08 UTC | newest] Thread overview: 3+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2024-08-05 6:52 [PATCH i-g-t] tests/amdgpu: move bad mem compute test to amd deadlock Jesse Zhang 2024-08-05 8:28 ` ✗ Fi.CI.BUILD: failure for " Patchwork 2024-08-05 10:08 ` [PATCH i-g-t] " Kamil Konieczny
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox