From fc6ea4bee130710a77ec16a86d2013e964602503 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Tue, 25 Jan 2022 10:51:49 -0500 Subject: drm/amdgpu: Wipe all VRAM on free when RAS is enabled MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit On GPUs with RAS, poison can propagate between processes if VRAM is not cleared when it is freed or allocated. The reason is, that not all write accesses clear RAS poison. 32-byte writes by the SDMA engine do clear RAS poison. Clearing memory in the background when it is freed should avoid major performance impact. KFD has been doing this already for a long time. Signed-off-by: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_object.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 5661b82d84d4..ec29365d108d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -575,6 +575,9 @@ int amdgpu_bo_create(struct amdgpu_device *adev, if (!amdgpu_bo_support_uswc(bo->flags)) bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC; + if (adev->ras_enabled) + bo->flags |= AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE; + bo->tbo.bdev = &adev->mman.bdev; if (bp->domain & (AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA | AMDGPU_GEM_DOMAIN_GDS)) -- cgit From 447c7997b62a5115ba4da846dcdee4fc12298a6a Mon Sep 17 00:00:00 2001 From: Rajneesh Bhardwaj Date: Thu, 3 Feb 2022 21:18:21 -0500 Subject: drm/amdgpu: Fix recursive locking warning MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Noticed the below warning while running a pytorch workload on vega10 GPUs. Change to trylock to avoid conflicts with already held reservation locks. [ +0.000003] WARNING: possible recursive locking detected [ +0.000003] 5.13.0-kfd-rajneesh #1030 Not tainted [ +0.000004] -------------------------------------------- [ +0.000002] python/4822 is trying to acquire lock: [ +0.000004] ffff932cd9a259f8 (reservation_ww_class_mutex){+.+.}-{3:3}, at: amdgpu_bo_release_notify+0xc4/0x160 [amdgpu] [ +0.000203] but task is already holding lock: [ +0.000003] ffff932cbb7181f8 (reservation_ww_class_mutex){+.+.}-{3:3}, at: ttm_eu_reserve_buffers+0x270/0x470 [ttm] [ +0.000017] other info that might help us debug this: [ +0.000002] Possible unsafe locking scenario: [ +0.000003] CPU0 [ +0.000002] ---- [ +0.000002] lock(reservation_ww_class_mutex); [ +0.000004] lock(reservation_ww_class_mutex); [ +0.000003] *** DEADLOCK *** [ +0.000002] May be due to missing lock nesting notation [ +0.000003] 7 locks held by python/4822: [ +0.000003] #0: ffff932c4ac028d0 (&process->mutex){+.+.}-{3:3}, at: kfd_ioctl_map_memory_to_gpu+0x10b/0x320 [amdgpu] [ +0.000232] #1: ffff932c55e830a8 (&info->lock#2){+.+.}-{3:3}, at: amdgpu_amdkfd_gpuvm_map_memory_to_gpu+0x64/0xf60 [amdgpu] [ +0.000241] #2: ffff932cc45b5e68 (&(*mem)->lock){+.+.}-{3:3}, at: amdgpu_amdkfd_gpuvm_map_memory_to_gpu+0xdf/0xf60 [amdgpu] [ +0.000236] #3: ffffb2b35606fd28 (reservation_ww_class_acquire){+.+.}-{0:0}, at: amdgpu_amdkfd_gpuvm_map_memory_to_gpu+0x232/0xf60 [amdgpu] [ +0.000235] #4: ffff932cbb7181f8 (reservation_ww_class_mutex){+.+.}-{3:3}, at: ttm_eu_reserve_buffers+0x270/0x470 [ttm] [ +0.000015] #5: ffffffffc045f700 (*(sspp++)){....}-{0:0}, at: drm_dev_enter+0x5/0xa0 [drm] [ +0.000038] #6: ffff932c52da7078 (&vm->eviction_lock){+.+.}-{3:3}, at: amdgpu_vm_bo_update_mapping+0xd5/0x4f0 [amdgpu] [ +0.000195] stack backtrace: [ +0.000003] CPU: 11 PID: 4822 Comm: python Not tainted 5.13.0-kfd-rajneesh #1030 [ +0.000005] Hardware name: GIGABYTE MZ01-CE0-00/MZ01-CE0-00, BIOS F02 08/29/2018 [ +0.000003] Call Trace: [ +0.000003] dump_stack+0x6d/0x89 [ +0.000010] __lock_acquire+0xb93/0x1a90 [ +0.000009] lock_acquire+0x25d/0x2d0 [ +0.000005] ? amdgpu_bo_release_notify+0xc4/0x160 [amdgpu] [ +0.000184] ? lock_is_held_type+0xa2/0x110 [ +0.000006] ? amdgpu_bo_release_notify+0xc4/0x160 [amdgpu] [ +0.000184] __ww_mutex_lock.constprop.17+0xca/0x1060 [ +0.000007] ? amdgpu_bo_release_notify+0xc4/0x160 [amdgpu] [ +0.000183] ? lock_release+0x13f/0x270 [ +0.000005] ? lock_is_held_type+0xa2/0x110 [ +0.000006] ? amdgpu_bo_release_notify+0xc4/0x160 [amdgpu] [ +0.000183] amdgpu_bo_release_notify+0xc4/0x160 [amdgpu] [ +0.000185] ttm_bo_release+0x4c6/0x580 [ttm] [ +0.000010] amdgpu_bo_unref+0x1a/0x30 [amdgpu] [ +0.000183] amdgpu_vm_free_table+0x76/0xa0 [amdgpu] [ +0.000189] amdgpu_vm_free_pts+0xb8/0xf0 [amdgpu] [ +0.000189] amdgpu_vm_update_ptes+0x411/0x770 [amdgpu] [ +0.000191] amdgpu_vm_bo_update_mapping+0x324/0x4f0 [amdgpu] [ +0.000191] amdgpu_vm_bo_update+0x251/0x610 [amdgpu] [ +0.000191] update_gpuvm_pte+0xcc/0x290 [amdgpu] [ +0.000229] ? amdgpu_vm_bo_map+0xd7/0x130 [amdgpu] [ +0.000190] amdgpu_amdkfd_gpuvm_map_memory_to_gpu+0x912/0xf60 [amdgpu] [ +0.000234] kfd_ioctl_map_memory_to_gpu+0x182/0x320 [amdgpu] [ +0.000218] kfd_ioctl+0x2b9/0x600 [amdgpu] [ +0.000216] ? kfd_ioctl_unmap_memory_from_gpu+0x270/0x270 [amdgpu] [ +0.000216] ? lock_release+0x13f/0x270 [ +0.000006] ? __fget_files+0x107/0x1e0 [ +0.000007] __x64_sys_ioctl+0x8b/0xd0 [ +0.000007] do_syscall_64+0x36/0x70 [ +0.000004] entry_SYSCALL_64_after_hwframe+0x44/0xae [ +0.000007] RIP: 0033:0x7fbff90a7317 [ +0.000004] Code: b3 66 90 48 8b 05 71 4b 2d 00 64 c7 00 26 00 00 00 48 c7 c0 ff ff ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 41 4b 2d 00 f7 d8 64 89 01 48 [ +0.000005] RSP: 002b:00007fbe301fe648 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 [ +0.000006] RAX: ffffffffffffffda RBX: 00007fbcc402d820 RCX: 00007fbff90a7317 [ +0.000003] RDX: 00007fbe301fe690 RSI: 00000000c0184b18 RDI: 0000000000000004 [ +0.000003] RBP: 00007fbe301fe690 R08: 0000000000000000 R09: 00007fbcc402d880 [ +0.000003] R10: 0000000002001000 R11: 0000000000000246 R12: 00000000c0184b18 [ +0.000003] R13: 0000000000000004 R14: 00007fbf689593a0 R15: 00007fbcc402d820 Cc: Christian König Cc: Felix Kuehling Cc: Alex Deucher Reviewed-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Rajneesh Bhardwaj Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_object.c') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index ec29365d108d..23c9a60693ee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -1306,7 +1306,8 @@ void amdgpu_bo_release_notify(struct ttm_buffer_object *bo) !(abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE)) return; - dma_resv_lock(bo->base.resv, NULL); + if (WARN_ON_ONCE(!dma_resv_trylock(bo->base.resv))) + return; r = amdgpu_fill_buffer(abo, AMDGPU_POISON, bo->base.resv, &fence); if (!WARN_ON(r)) { -- cgit