Created attachment 129878 [details] dmesg log with drm.debug=0xe The following tests are failing on BXT, attached dmesg log for two tests, all failures have similar output and dmesg, apparently running out of memory. igt@gem_shrink@execbuf1-oom igt@gem_shrink@execbufn igt@gem_shrink@execbufn-userptr igt@gem_shrink@get-pages-userptr igt@gem_shrink@mmap-cpu-oom igt@gem_shrink@mmap-gtt-oom igt@gem_shrink@pread igt@gem_shrink@pwrite-userptr Steps to reproduce ========================================= 1. Download and Install IGT suite. 2. go to intel-gpu-tools/tests/ folder 3. execute test ./gem_shrink --run-subtest execbuf1-oom 4. execute test ./gem_shrink --run-subtest execbufn 5. execute test ./gem_shrink --run-subtest execbufn-userptr 6. execute test ./gem_shrink --run-subtest get-pages-userptr 7. execute test ./gem_shrink --run-subtest mmap-cpu-oom 8. execute test ./gem_shrink --run-subtest mmap-gtt-oom 9. execute test ./gem_shrink --run-subtest pread 10. execute test ./gem_shrink --run-subtest pwrite-userptr Expected output ========================================= Tests should be marked as pass Actual output ========================================== Tests are marked as fail showing the following debug info ( more in dmesg) [ 995.656651] [ pid ] uid tgid total_vm rss nr_ptes nr_pmds swapents oom_score_adj name [ 995.656732] [ 1891] 0 1891 9922 84 23 3 0 1000 gem_shrink [ 995.656734] [ 1949] 0 1949 1844931 0 22 3 0 1000 gem_shrink [ 995.656736] [ 1950] 0 1950 1844931 0 22 3 0 1000 gem_shrink [ 995.656738] [ 1951] 0 1951 1844931 0 22 3 0 1000 gem_shrink [ 995.656740] [ 1952] 0 1952 1844931 0 22 3 0 1000 gem_shrink [ 995.656742] [ 1953] 0 1953 1844931 0 22 3 0 1000 gem_shrink [ 995.656744] [ 1954] 0 1954 1844931 84 22 3 0 1000 gem_shrink [ 995.656748] Out of memory: Kill process 1891 (gem_shrink) score 999 or sacrifice child [ 995.656761] Killed process 1954 (gem_shrink) total-vm:7379724kB, anon-rss:336kB, file-rss:0kB, shmem-rss:0kB [ 995.663642] oom_reaper: reaped process 1954 (gem_shrink), now anon-rss:0kB, file-rss:0kB, shmem-rss:0kB [ 995.736290] NetworkManager invoked oom-killer: gfp_mask=0x14200ca(GFP_HIGHUSER_MOVABLE), nodemask=0, order=0, oom_score_adj=0 [ 995.736292] NetworkManager cpuset=/ mems_allowed=0 [ 995.736298] CPU: 1 PID: 756 Comm: NetworkManager Not tainted 4.10.0-drm-tip-qa-ww8-commit-0a03ea9+ #1 ========== Kernel Information============ commit 0a03ea9496b49746b4964d05cc1f483385d1df8a Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Mon Feb 20 17:20:19 2017 +0000 drm-tip: 2017y-02m-20d-17h-19m-56s UTC integration manifest Kernel version : 4.10.0-0a03ea9 Architecture : source amd64 all ========== Software Information ========== Component : drm tag : libdrm-2.4.75-7-gf6499b1 commit : f6499b1 Component : cairo tag : 1.15.4-6-g5854dd9 commit : 5854dd9 Component : intel-gpu-tools tag : intel-gpu-tools-1.17-245-g2210c5b commit : 2210c5b ========================================= Ubuntu 16.10 yakkety
The tests are designed to fail, -oom definitely, the others as a deficiency of the kernel. The purpose is to ensure we don't crash under extreme mempressure. The new page allocation stalls are expected - since that is exactly the stress we are putting the system under. *** This bug has been marked as a duplicate of bug 95424 ***
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.