Created attachment 114567 [details] dmesg ==System Environment== -------------------------- Regression: Yes it fails on 3.19.2, works well 3.18.0 Non-working platforms: HSW ==kernel== -------------------------- commit 660613d1a4e94144490850b6c3d350331860fac4 Author: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Date: Wed Mar 18 14:11:52 2015 +0100 Linux 3.19.2 ==Bug detailed description== ----------------------------- It fails on 3.19.2, works well 3.18.0 and the latest drm-intel-nightly kernel. igt@drm_read@short-buffer-nonblock also fails. output: IGT-Version: 1.10-ge7d26df (x86_64) (Linux: 3.19.2_v3.19.2+ x86_64) Test assertion failure function test_short_buffer, file drm_read.c:148: Failed assertion: read(fd, buffer, 4) == 0 Last errno: 11, Resource temporarily unavailable error: -1 != 0 Subtest short-buffer-block failed. **** DEBUG **** Test assertion failure function test_short_buffer, file drm_read.c:148: Failed assertion: read(fd, buffer, 4) == 0 Last errno: 11, Resource temporarily unavailable error: -1 != 0 **** END **** Subtest short-buffer-block: FAIL (0.000s) Reproduce steps: ------------------------- 1. ./drm_read --run-subtest short-buffer-block
The drm_read routine will return EAGAIN if the event list is empty, so this may be expected behavior if the event hasn't been queued yet. Could be a test bug or a race in the event delivery code in the kernel.
Hi Is this still happening on the 3.19 series? What about -nightly? Is this really HSW-specific? Thanks, Paulo
(In reply to Paulo Zanoni from comment #2) > Hi > > Is this still happening on the 3.19 series? What about -nightly? Is this > really HSW-specific? Hmm, I can reproduce this on Ubuntu's 3.19.0-23 Kernel, but not on -nightly.
I have no idea why they thought 3.18 worked, since the test was designed to expose the bugs in the read routine since its introduction.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.