Re: [PATCH 1/5] drm/i915: audit bo->resource usage v3

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, 25 Jan 2023 at 14:20, Christian König
<ckoenig.leichtzumerken@xxxxxxxxx> wrote:
>
> Am 25.01.23 um 13:53 schrieb Matthew Auld:
> > On Wed, 25 Jan 2023 at 11:35, Christian König
> > <ckoenig.leichtzumerken@xxxxxxxxx> wrote:
> >> Am 25.01.23 um 11:21 schrieb Matthew Auld:
> >>> On Wed, 25 Jan 2023 at 10:07, Christian König
> >>> <ckoenig.leichtzumerken@xxxxxxxxx> wrote:
> >>>> Am 25.01.23 um 10:56 schrieb Matthew Auld:
> >>>>> On Tue, 24 Jan 2023 at 17:15, Matthew Auld
> >>>>> <matthew.william.auld@xxxxxxxxx> wrote:
> >>>>>> On Tue, 24 Jan 2023 at 13:48, Matthew Auld
> >>>>>> <matthew.william.auld@xxxxxxxxx> wrote:
> >>>>>>> On Tue, 24 Jan 2023 at 12:57, Christian König
> >>>>>>> <ckoenig.leichtzumerken@xxxxxxxxx> wrote:
> >>>>>>>> From: Christian König <ckoenig.leichtzumerken@xxxxxxxxx>
> >>>>>>>>
> >>>>>>>> Make sure we can at least move and alloc TT objects without backing store.
> >>>>>>>>
> >>>>>>>> v2: clear the tt object even when no resource is allocated.
> >>>>>>>> v3: add Matthews changes for i915 as well.
> >>>>>>>>
> >>>>>>>> Signed-off-by: Christian König <christian.koenig@xxxxxxx>
> >>>>>>> Reviewed-by: Matthew Auld <matthew.auld@xxxxxxxxx>
> >>>>>> Ofc that assumes intel-gfx CI is now happy with the series.
> >>>>> There are still some nasty failures it seems (in the extended test
> >>>>> list). But it looks like the series is already merged. Can we quickly
> >>>>> revert and try again?
> >>>> Ah, crap. I thought everything would be fine after the CI gave it's go.
> >>>>
> >>>> Which patch is causing the fallout?
> >>> I'm not sure. I think all of the patches kind of interact with each
> >>> other, but for sure there is an issue with the first patch. There is
> >>> one splat like:
> >> Well I would rather like to revert as less as possible.
> >>
> >> Are you sure that this isn't only on some i915 specific branch with not
> >> yet upstream changes?
> > Yeah, that splat is taken directly from the CI results reported with
> > this series. So it's just your series applied on top of drm-tip.
> >
> > Can you take a look at the first patch here:
> > https://patchwork.freedesktop.org/series/113332/
> >
> > Maybe you have a better idea? For reference the IGTs that we have for
> > verifying userspace object clearing are now failing, so hoping that
> > fixes it. The other two patches I'm hoping will fix the splat.
>
> The TTM change looks like a good idea to me. Feel free to add my rb to
> this one.
>
> I can't say much about the i915 changes.
>
> Maybe we should revert the two TTM patches to not allocate resources for
> now and fix i915 first?

>From what I can see, we would need to revert all three TTM patches,
keeping just the i915 one. Reverting for now I think makes sense.

>
> Christian.
>
> >
> >> I can't even find the i915_gem_obj_copy_ttm function in drm-misc-next
> >> nor drm-next.
> >>
> >> Regards,
> >> Christian.
> >>
> >>> <1>[  109.735148] BUG: kernel NULL pointer dereference, address:
> >>> 0000000000000010
> >>> <1>[  109.735151] #PF: supervisor read access in kernel mode
> >>> <1>[  109.735152] #PF: error_code(0x0000) - not-present page
> >>> <6>[  109.735153] PGD 0 P4D 0
> >>> <4>[  109.735155] Oops: 0000 [#1] PREEMPT SMP NOPTI
> >>> <4>[  109.735157] CPU: 1 PID: 92 Comm: kworker/u12:6 Not tainted
> >>> 6.2.0-rc5-Patchwork_113269v1-gc4d436608c4e+ #1
> >>> <4>[  109.735159] Hardware name: Gigabyte Technology Co., Ltd. GB-Z390
> >>> Garuda/GB-Z390 Garuda-CF, BIOS IG1c 11/19/2019
> >>> <4>[  109.735160] Workqueue: events_unbound async_run_entry_fn
> >>> <4>[  109.735163] RIP: 0010:i915_ttm_resource_mappable+0x4/0x30 [i915]
> >>> <4>[  109.735286] Code: b8 f9 ff ff ff eb c2 e8 aa 5e 52 e1 e9 4f 0f
> >>> 18 00 0f 1f 44 00 00 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90 90
> >>> 66 0f 1f 00 <8b> 57 10 b8 01 00 00 00 85 d2 74 15 48 8b 47 08 48 05 ff
> >>> 0f 00 00
> >>> <4>[  109.735288] RSP: 0018:ffffc90000f339a8 EFLAGS: 00010246
> >>> <4>[  109.735289] RAX: 0000000000000000 RBX: 0000000000000000 RCX:
> >>> ffff88810cea3a00
> >>> <4>[  109.735290] RDX: 0000000000000000 RSI: ffffc90000f33af0 RDI:
> >>> 0000000000000000
> >>> <4>[  109.735292] RBP: ffff88811645d7c0 R08: 0000000000000000 R09:
> >>> ffff888123afa940
> >>> <4>[  109.735292] R10: 0000000000000001 R11: ffff888104b70040 R12:
> >>> 0000000000000000
> >>> <4>[  109.735293] R13: 0000000000000000 R14: ffffc90000f33b08 R15:
> >>> ffffc90000f33af0
> >>> <4>[  109.735294] FS:  0000000000000000(0000)
> >>> GS:ffff8884ad680000(0000) knlGS:0000000000000000
> >>> <4>[  109.735295] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> >>> <4>[  109.735296] CR2: 0000000000000010 CR3: 000000011f9c6003 CR4:
> >>> 00000000003706e0
> >>> <4>[  109.735297] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
> >>> 0000000000000000
> >>> <4>[  109.735298] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
> >>> 0000000000000400
> >>> <4>[  109.735299] Call Trace:
> >>> <4>[  109.735300]  <TASK>
> >>> <4>[  109.735301]  __i915_ttm_move+0x128/0x940 [i915]
> >>> <4>[  109.735408]  ? dma_resv_iter_next+0x91/0xb0
> >>> <4>[  109.735412]  ? dma_resv_iter_first+0x42/0xb0
> >>> <4>[  109.735414]  ? i915_deps_add_resv+0x4c/0xc0 [i915]
> >>> <4>[  109.735520]  i915_gem_obj_copy_ttm+0x12f/0x250 [i915]
> >>> <4>[  109.735625]  i915_ttm_restore+0x167/0x250 [i915]
> >>> <4>[  109.735759]  i915_gem_process_region+0x27a/0x3b0 [i915]
> >>> <4>[  109.735881]  i915_ttm_restore_region+0x4b/0x70 [i915]
> >>> <4>[  109.735999]  lmem_restore+0x3a/0x60 [i915]
> >>> <4>[  109.736101]  i915_gem_resume+0x4c/0x100 [i915]
> >>> <4>[  109.736202]  i915_drm_resume+0xc2/0x170 [i915]
> >>>
> >>> Plus some other less obvious issue(s) with some tests failing.
> >>>
> >>>> Christian.
>




[Index of Archives]     [AMD Graphics]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux