[Bug 110470] New: [drm/i915/icl]media driver see gpu hang issue caused by WaEnableStateCacheRedirectToCS context workaround configure for the L3 cache

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Fri Apr 19 08:33:52 UTC 2019


https://bugs.freedesktop.org/show_bug.cgi?id=110470

            Bug ID: 110470
           Summary: [drm/i915/icl]media driver see gpu hang issue caused
                    by WaEnableStateCacheRedirectToCS context workaround
                    configure for the L3 cache
           Product: DRI
           Version: XOrg git
          Hardware: x86 (IA32)
                OS: Linux (All)
            Status: NEW
          Severity: major
          Priority: medium
         Component: DRM/Intel
          Assignee: intel-gfx-bugs at lists.freedesktop.org
          Reporter: carry.lv at intel.com
        QA Contact: intel-gfx-bugs at lists.freedesktop.org
                CC: intel-gfx-bugs at lists.freedesktop.org

Created attachment 144044
  --> https://bugs.freedesktop.org/attachment.cgi?id=144044&action=edit
attached finding root cause

Media driver see gpu hang randomly for VME cases.

The issue is root caused. The reason is conflicting Settings of L3 were used in
UMD and KMD.
In picture 1 is the setting of KMD in function icl_ctx_workarounds_init, which
enabled  State Cache redirect to CS Command buffer section(bit11 of MMIO
0x731C). However,  the L3 configuration of UMD(in Picture2) didn't assign any
space for CS Command buffer.
I tried to remove the WaEnableStateCacheRedirectToCS in KMD, the issue is gone.



WaEnableStateCacheRedirectToCS context workaround configures the L3 cache to
benefit 3d workloads but media has different requirements.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the assignee for the bug.
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20190419/1f7c8f78/attachment.html>


More information about the intel-gfx-bugs mailing list