On 7 December 2017 at 19:11, Michel Dänzer michel@daenzer.net wrote:
On 2017-12-07 10:35 AM, Xinliang Liu wrote:
Hi,
We are trying ATI HD 7470 card on arm64 platform D05 with kernel 4.1.40. But it failed with kernel log: radeon 0005:01:00: ring 0 stalled for more than 10000msec radeon 0005:01:00: GPU lockup (current fence id 0x0000000000000001 last fence id 0x0000000000000002 on ring0) radeon 0005:01:00: ring 0 stalled for more than 10500msec radeon 0005:01:00: GPU lockup (current fence id 0x0000000000000001 last fence id 0x0000000000000002 on ring0) radeon 0005:01:00: ring 0 stalled for more than 11000msec radeon 0005:01:00: GPU lockup (current fence id 0x0000000000000001 last fence id 0x0000000000000002 on ring0) radeon 0005:01:00: ring 0 stalled for more than 11500msec radeon 0005:01:00: GPU lockup (current fence id 0x0000000000000001 last fence id 0x0000000000000002 on ring0) ...
Any clue?
Any chance you can try a newer kernel? Somebody else (also @huawei.com incidentally) reported just this week on the mesa-dev mailing list that the amdgpu driver in the Ubuntu 4.13 kernel is working on an arm64 platform.
If a newer kernel still doesn't work on your platform, please provide the full dmesg output.
Hope we have change to try newer kernel. After tracking the mainline we had found that this commit[1] fixed this issue. The root cause is that on old kernel ttm can't succeed to set page writecombine property which make cpu/gpu in fatal status and stub.
Anyway, we should use newer kernel and keep on testing.
[1] f135b97 drm/ttm: recognize ARM64 arch in ioprot handler https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h...
BTW, in the other thread it was reported that Mesa has bugs which can be worked around by compiling it with -fsigned-char.
Thanks for letting us know this. Had looked into the mesa bug.
Thanks, Xinliang
-- Earthling Michel Dänzer | http://www.amd.com Libre software enthusiast | Mesa and X developer