From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by gabe.freedesktop.org (Postfix) with ESMTPS id CD85E2BE93 for ; Tue, 26 Jul 2022 10:13:09 +0000 (UTC) From: Karolina Drobnik To: igt-dev@lists.freedesktop.org Date: Tue, 26 Jul 2022 12:13:10 +0200 Message-Id: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Subject: [igt-dev] [PATCH i-g-t v2 0/2] tests/gem_exec_fence: Fix test_fence_await for hanging workloads List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Tvrtko Ursulin , Chris Wilson Errors-To: igt-dev-bounces@lists.freedesktop.org Sender: "igt-dev" List-ID: This patch series fixes test_fence_await for failing gem_exec_fence subtests that inject GPU hangs (await-hang and nb-await-hang). The test assumed that the error notification happens after a hang is declared, which would be enough for the next set of fences to be submitted, making this test pass. But as we use the error-interrupt, we get immediate reset for the invalid command stream. This means that when the test checks for active fences, none can be found, leading to the test failure. To address this problem, the active fences check is only done for valid workloads. In addition to the fix, the series adds spinning before usleep() call to coordinate sleep with the start of the request. v2: No functional changes, fixed styling issues pointed out by Kamil: - Moved comment for fence_busy(spin->out_fence) assertion - Fixed formatting for flags assignment in spin definition Chris Wilson (2): tests/gem_exec_fence: Check stored values only for valid workloads tests/gem_exec_fence: Coordinate sleep with the start of the request tests/i915/gem_exec_fence.c | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-) -- 2.25.1