[PATCH v2 7/8] drm/amdgpu: Fix sdma code crash post device unplug

Christian König ckoenig.leichtzumerken at gmail.com
Tue Jun 23 07:14:50 UTC 2020


Am 23.06.20 um 07:11 schrieb Andrey Grodzovsky:
>
> On 6/22/20 3:40 PM, Christian König wrote:
>> Am 21.06.20 um 08:03 schrieb Andrey Grodzovsky:
>>> entity->rq becomes null aftre device unplugged so just return early
>>> in that case.
>>
>> Mhm, do you have a backtrace for this?
>>
>> This should only be called by an IOCTL and IOCTLs should already call 
>> drm_dev_enter()/exit() on their own...
>>
>> Christian.
>
>
> See bellow, it's not during IOCTL but during all GEM objects release 
> when releasing the device. entity->rq becomes null because all the gpu 
> schedulers are marked as not ready during the early pci remove stage 
> and so the next time sdma job tries to pick a scheduler to run nothing 
> is available and it's set to null.

I see. This should then probably go into amdgpu_gem_object_close() 
before we reserve the PD.

See drm_dev_enter()/exit() are kind of a read side lock and with this we 
create a nice lock inversion when we do it in the low level SDMA VM backend.

Christian.

>
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382648] BUG: kernel 
> NULL pointer dereference, address: 0000000000000038
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382651] #PF: 
> supervisor read access in kernel mode
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382652] #PF: 
> error_code(0x0000) - not-present page
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382653] PGD 0 P4D 0
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382656] Oops: 0000 
> [#1] SMP PTI
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382658] CPU: 6 PID: 
> 2598 Comm: llvmpipe-6 Tainted: G           OE     5.6.0-dev+ #51
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382659] Hardware name: 
> System manufacturer System Product Name/RAMPAGE IV FORMULA, BIOS 4804 
> 12/30/2013
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382700] RIP: 
> 0010:amdgpu_vm_sdma_commit+0x6c/0x270 [amdgpu]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382702] Code: 01 00 00 
> 48 89 ee 48 c7 c7 ef d4 85 c0 e8 fc 5f e8 ff 48 8b 75 10 48 c7 c7 fd 
> d4 85 c0 e8 ec 5f e8 ff 48 8b 45 10 41 8b 55 08 <48> 8b 40 38 85 d2 48 
> 8d b8 30 ff ff ff 0f 84 9b 01 00 00 48 8b 80
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382704] RSP: 
> 0018:ffffa88e40f57950 EFLAGS: 00010282
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382705] RAX: 
> 0000000000000000 RBX: ffffa88e40f579a8 RCX: 0000000000000001
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382707] RDX: 
> 0000000000000014 RSI: ffff94d4d62388e0 RDI: ffff94d4dbd98e30
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382708] RBP: 
> ffff94d4d2ad3288 R08: 0000000000000000 R09: 0000000000000001
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382709] R10: 
> 000000000000001f R11: 0000000000000000 R12: ffffa88e40f57a48
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382710] R13: 
> ffff94d4d627a5e8 R14: ffff94d4d424d978 R15: 0000000800100020
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382712] FS: 
> 00007f30ae694700(0000) GS:ffff94d4dbd80000(0000) knlGS:0000000000000000
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382713] CS:  0010 DS: 
> 0000 ES: 0000 CR0: 0000000080050033
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382714] CR2: 
> 0000000000000038 CR3: 0000000121810006 CR4: 00000000000606e0
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382716] Call Trace:
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382755] 
> amdgpu_vm_bo_update_mapping.constprop.30+0x16b/0x230 [amdgpu]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382795] 
> amdgpu_vm_clear_freed+0xd7/0x210 [amdgpu]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382833] 
> amdgpu_gem_object_close+0x200/0x2b0 [amdgpu]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382856]  ? 
> drm_gem_object_handle_put_unlocked+0x90/0x90 [drm]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382864]  ? 
> drm_gem_object_release_handle+0x2c/0x90 [drm]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382872] 
> drm_gem_object_release_handle+0x2c/0x90 [drm]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382879]  ? 
> drm_gem_object_handle_put_unlocked+0x90/0x90 [drm]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382882] 
> idr_for_each+0x48/0xd0
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382885]  ? 
> _raw_spin_unlock_irqrestore+0x2d/0x50
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382893] 
> drm_gem_release+0x1c/0x30 [drm]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382901] 
> drm_file_free+0x21d/0x270 [drm]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382908] 
> drm_release+0x67/0xe0 [drm]
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382912] __fput+0xc6/0x260
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382916] 
> task_work_run+0x79/0xb0
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382919] 
> do_exit+0x3d0/0xc40
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382921]  ? 
> get_signal+0x13d/0xc30
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382924] 
> do_group_exit+0x47/0xb0
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382926] 
> get_signal+0x18b/0xc30
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382929] 
> do_signal+0x36/0x6a0
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382931]  ? 
> __set_task_comm+0x62/0x120
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382935]  ? 
> __x64_sys_futex+0x88/0x180
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382938] 
> exit_to_usermode_loop+0x6f/0xc0
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382941] 
> do_syscall_64+0x149/0x1c0
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382943] 
> entry_SYSCALL_64_after_hwframe+0x49/0xbe
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382944] RIP: 
> 0033:0x7f30f7f35360
> Jun  8 11:14:56 ubuntu-1604-test kernel: [   44.382947] Code: Bad RIP 
> value.
>
>
> Andrey
>
>
>>
>>>
>>> Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky at amd.com>
>>> ---
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c | 21 
>>> ++++++++++++++++-----
>>>   1 file changed, 16 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
>>> index 8d9c6fe..d252427 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
>>> @@ -24,6 +24,7 @@
>>>   #include "amdgpu_job.h"
>>>   #include "amdgpu_object.h"
>>>   #include "amdgpu_trace.h"
>>> +#include <drm/drm_drv.h>
>>>     #define AMDGPU_VM_SDMA_MIN_NUM_DW    256u
>>>   #define AMDGPU_VM_SDMA_MAX_NUM_DW    (16u * 1024u)
>>> @@ -94,7 +95,12 @@ static int amdgpu_vm_sdma_commit(struct 
>>> amdgpu_vm_update_params *p,
>>>       struct drm_sched_entity *entity;
>>>       struct amdgpu_ring *ring;
>>>       struct dma_fence *f;
>>> -    int r;
>>> +    int r, idx;
>>> +
>>> +    if (!drm_dev_enter(p->adev->ddev, &idx)) {
>>> +        r = -ENODEV;
>>> +        goto nodev;
>>> +    }
>>>         entity = p->immediate ? &p->vm->immediate : &p->vm->delayed;
>>>       ring = container_of(entity->rq->sched, struct amdgpu_ring, 
>>> sched);
>>> @@ -104,7 +110,7 @@ static int amdgpu_vm_sdma_commit(struct 
>>> amdgpu_vm_update_params *p,
>>>       WARN_ON(ib->length_dw > p->num_dw_left);
>>>       r = amdgpu_job_submit(p->job, entity, AMDGPU_FENCE_OWNER_VM, &f);
>>>       if (r)
>>> -        goto error;
>>> +        goto job_fail;
>>>         if (p->unlocked) {
>>>           struct dma_fence *tmp = dma_fence_get(f);
>>> @@ -118,10 +124,15 @@ static int amdgpu_vm_sdma_commit(struct 
>>> amdgpu_vm_update_params *p,
>>>       if (fence && !p->immediate)
>>>           swap(*fence, f);
>>>       dma_fence_put(f);
>>> -    return 0;
>>>   -error:
>>> -    amdgpu_job_free(p->job);
>>> +    r = 0;
>>> +
>>> +job_fail:
>>> +    drm_dev_exit(idx);
>>> +nodev:
>>> +    if (r)
>>> +        amdgpu_job_free(p->job);
>>> +
>>>       return r;
>>>   }
>>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx



More information about the amd-gfx mailing list