Summary: | [CI] igt@gem_ppgtt@blt-vs-render-ctxn - incomplete - gem_ppgtt: page allocation failure: order:0, mode:0x14204d2(GFP_HIGHUSER|__GFP_RETRY_MAYFAIL|__GFP_RECLAIMABLE), nodemask=(null) | ||
---|---|---|---|
Product: | DRI | Reporter: | Marta Löfstedt <marta.lofstedt> |
Component: | DRM/Intel | Assignee: | Intel GFX Bugs mailing list <intel-gfx-bugs> |
Status: | CLOSED WORKSFORME | QA Contact: | Intel GFX Bugs mailing list <intel-gfx-bugs> |
Severity: | normal | ||
Priority: | medium | CC: | intel-gfx-bugs |
Version: | DRI git | ||
Hardware: | Other | ||
OS: | All | ||
Whiteboard: | ReadyForDev | ||
i915 platform: | BXT | i915 features: | GEM/Other |
Description
Marta Löfstedt
2017-12-01 08:34:39 UTC
From dmesg, the problems start already on igt trying to init before gem_eio/throttle. multiring resets (dead gpu?) already before throttle starts and also then during throttle. (In reply to Mika Kuoppala from comment #1) > From dmesg, the problems start already on igt trying to init before > gem_eio/throttle. multiring resets (dead gpu?) already before throttle starts > and also then during throttle. That's gem_eio, it's purpose is to wedge the gpu and check abi for correct behaviour. Before gem_ppgtt started, it checked that the gpu was no longer wedged, so everything appears in order. What is not so happy is the <4>[ 193.388010] hpet1: lost 7161 rtc interrupts A dead timer. That would explain why we never were able to retire requests and so leaked everything, and everything remained marked as busy. (In reply to Chris Wilson from comment #2) > What is not so happy is the > > <4>[ 193.388010] hpet1: lost 7161 rtc interrupts > > A dead timer. That would explain why we never were able to retire requests > and so leaked everything, and everything remained marked as busy. No, it doesn't. If we thought everything was still active, we would not have filled swap (only inactive bo can be swapped). Which brings us back to a bo leak, but whose? Last seen: CI_DRM_3419: 2017-11-30 / 241 runs ago |
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.