kernel/linux.git/drivers/gpu/drm/amd/amdkfd, branch v6.6.16

drm/amdkfd: only flush mes process context if mes support is there

2024-02-05T20:14:34+00:00

[ Upstream commit 24149412dfc71f7f4a54868702e9145e396263d3 ] Fix up on mes process context flush to prevent non-mes devices from spamming error messages or running into undefined behaviour during process termination. Fixes: bd33bb1409b4 ("drm/amdkfd: fix mes set shader debugger process management") Signed-off-by: Jonathan Kim Reviewed-by: Eric Huang Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

drm/amdkfd: Fix 'node' NULL check in 'svm_range_get_range_boundaries()'

2024-02-05T20:14:33+00:00

[ Upstream commit d7a254fad873775ce6c32b77796c81e81e6b7f2e ] Range interval [start, last] is ordered by rb_tree, rb_prev, rb_next return value still needs NULL check, thus modified from "node" to "rb_node". Fixes the below: drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_svm.c:2691 svm_range_get_range_boundaries() warn: can 'node' even be NULL? Suggested-by: Philip Yang Cc: Felix Kuehling Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

drm/amdkfd: Fix lock dependency warning with srcu

2024-02-05T20:14:32+00:00

[ Upstream commit 2a9de42e8d3c82c6990d226198602be44f43f340 ] ====================================================== WARNING: possible circular locking dependency detected 6.5.0-kfd-yangp #2289 Not tainted ------------------------------------------------------ kworker/0:2/996 is trying to acquire lock: (srcu){.+.+}-{0:0}, at: __synchronize_srcu+0x5/0x1a0 but task is already holding lock: ((work_completion)(&svms->deferred_list_work)){+.+.}-{0:0}, at: process_one_work+0x211/0x560 which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #3 ((work_completion)(&svms->deferred_list_work)){+.+.}-{0:0}: __flush_work+0x88/0x4f0 svm_range_list_lock_and_flush_work+0x3d/0x110 [amdgpu] svm_range_set_attr+0xd6/0x14c0 [amdgpu] kfd_ioctl+0x1d1/0x630 [amdgpu] __x64_sys_ioctl+0x88/0xc0 -> #2 (&info->lock#2){+.+.}-{3:3}: __mutex_lock+0x99/0xc70 amdgpu_amdkfd_gpuvm_restore_process_bos+0x54/0x740 [amdgpu] restore_process_helper+0x22/0x80 [amdgpu] restore_process_worker+0x2d/0xa0 [amdgpu] process_one_work+0x29b/0x560 worker_thread+0x3d/0x3d0 -> #1 ((work_completion)(&(&process->restore_work)->work)){+.+.}-{0:0}: __flush_work+0x88/0x4f0 __cancel_work_timer+0x12c/0x1c0 kfd_process_notifier_release_internal+0x37/0x1f0 [amdgpu] __mmu_notifier_release+0xad/0x240 exit_mmap+0x6a/0x3a0 mmput+0x6a/0x120 do_exit+0x322/0xb90 do_group_exit+0x37/0xa0 __x64_sys_exit_group+0x18/0x20 do_syscall_64+0x38/0x80 -> #0 (srcu){.+.+}-{0:0}: __lock_acquire+0x1521/0x2510 lock_sync+0x5f/0x90 __synchronize_srcu+0x4f/0x1a0 __mmu_notifier_release+0x128/0x240 exit_mmap+0x6a/0x3a0 mmput+0x6a/0x120 svm_range_deferred_list_work+0x19f/0x350 [amdgpu] process_one_work+0x29b/0x560 worker_thread+0x3d/0x3d0 other info that might help us debug this: Chain exists of: srcu --> &info->lock#2 --> (work_completion)(&svms->deferred_list_work) Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock((work_completion)(&svms->deferred_list_work)); lock(&info->lock#2); lock((work_completion)(&svms->deferred_list_work)); sync(srcu); Signed-off-by: Philip Yang Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

drm/amdkfd: Fix lock dependency warning

2024-02-05T20:14:32+00:00

[ Upstream commit 47bf0f83fc86df1bf42b385a91aadb910137c5c9 ] ====================================================== WARNING: possible circular locking dependency detected 6.5.0-kfd-fkuehlin #276 Not tainted ------------------------------------------------------ kworker/8:2/2676 is trying to acquire lock: ffff9435aae95c88 ((work_completion)(&svm_bo->eviction_work)){+.+.}-{0:0}, at: __flush_work+0x52/0x550 but task is already holding lock: ffff9435cd8e1720 (&svms->lock){+.+.}-{3:3}, at: svm_range_deferred_list_work+0xe8/0x340 [amdgpu] which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #2 (&svms->lock){+.+.}-{3:3}: __mutex_lock+0x97/0xd30 kfd_ioctl_alloc_memory_of_gpu+0x6d/0x3c0 [amdgpu] kfd_ioctl+0x1b2/0x5d0 [amdgpu] __x64_sys_ioctl+0x86/0xc0 do_syscall_64+0x39/0x80 entry_SYSCALL_64_after_hwframe+0x63/0xcd -> #1 (&mm->mmap_lock){++++}-{3:3}: down_read+0x42/0x160 svm_range_evict_svm_bo_worker+0x8b/0x340 [amdgpu] process_one_work+0x27a/0x540 worker_thread+0x53/0x3e0 kthread+0xeb/0x120 ret_from_fork+0x31/0x50 ret_from_fork_asm+0x11/0x20 -> #0 ((work_completion)(&svm_bo->eviction_work)){+.+.}-{0:0}: __lock_acquire+0x1426/0x2200 lock_acquire+0xc1/0x2b0 __flush_work+0x80/0x550 __cancel_work_timer+0x109/0x190 svm_range_bo_release+0xdc/0x1c0 [amdgpu] svm_range_free+0x175/0x180 [amdgpu] svm_range_deferred_list_work+0x15d/0x340 [amdgpu] process_one_work+0x27a/0x540 worker_thread+0x53/0x3e0 kthread+0xeb/0x120 ret_from_fork+0x31/0x50 ret_from_fork_asm+0x11/0x20 other info that might help us debug this: Chain exists of: (work_completion)(&svm_bo->eviction_work) --> &mm->mmap_lock --> &svms->lock Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(&svms->lock); lock(&mm->mmap_lock); lock(&svms->lock); lock((work_completion)(&svm_bo->eviction_work)); I believe this cannot really lead to a deadlock in practice, because svm_range_evict_svm_bo_worker only takes the mmap_read_lock if the BO refcount is non-0. That means it's impossible that svm_range_bo_release is running concurrently. However, there is no good way to annotate this. To avoid the problem, take a BO reference in svm_range_schedule_evict_svm_bo instead of in the worker. That way it's impossible for a BO to get freed while eviction work is pending and the cancel_work_sync call in svm_range_bo_release can be eliminated. v2: Use svm_bo_ref_unless_zero and explained why that's safe. Also removed redundant checks that are already done in amdkfd_fence_enable_signaling. Signed-off-by: Felix Kuehling Reviewed-by: Philip Yang Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

drm/amdkfd: Fix iterator used outside loop in 'kfd_add_peer_prop()'

2024-02-05T20:14:30+00:00

[ Upstream commit b1a428b45dc7e47c7acc2ad0d08d8a6dda910c4c ] Fix the following about iterator use: drivers/gpu/drm/amd/amdgpu/../amdkfd/kfd_topology.c:1456 kfd_add_peer_prop() warn: iterator used outside loop: 'iolink3' Cc: Felix Kuehling Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

drm/amdkfd: fix mes set shader debugger process management

2024-02-05T20:14:28+00:00

[ Upstream commit bd33bb1409b494558a2935f7bbc7842def957fcd ] MES provides the driver a call to explicitly flush stale process memory within the MES to avoid a race condition that results in a fatal memory violation. When SET_SHADER_DEBUGGER is called, the driver passes a memory address that represents a process context address MES uses to keep track of future per-process calls. Normally, MES will purge its process context list when the last queue has been removed. The driver, however, can call SET_SHADER_DEBUGGER regardless of whether a queue has been added or not. If SET_SHADER_DEBUGGER has been called with no queues as the last call prior to process termination, the passed process context address will still reside within MES. On a new process call to SET_SHADER_DEBUGGER, the driver may end up passing an identical process context address value (based on per-process gpu memory address) to MES but is now pointing to a new allocated buffer object during KFD process creation. Since the MES is unaware of this, access of the passed address points to the stale object within MES and triggers a fatal memory violation. The solution is for KFD to explicitly flush the process context address from MES on process termination. Note that the flush call and the MES debugger calls use the same MES interface but are separated as KFD calls to avoid conflicting with each other. Signed-off-by: Jonathan Kim Tested-by: Alice Wong Reviewed-by: Eric Huang Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

drm/amdgpu/gfx11: set UNORD_DISPATCH in compute MQDs

2024-02-01T00:19:11+00:00

[ Upstream commit 3380fcad2c906872110d31ddf7aa1fdea57f9df6 ] This needs to be set to 1 to avoid a potential deadlock in the GC 10.x and newer. On GC 9.x and older, this needs to be set to 0. This can lead to hangs in some mixed graphics and compute workloads. Updated firmware is also required for AQL. Reviewed-by: Feifei Xu Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org Signed-off-by: Sasha Levin

drm/amdgpu/gfx10: set UNORD_DISPATCH in compute MQDs

2024-02-01T00:19:11+00:00

[ Upstream commit 03ff6d7238b77e5fb2b85dc5fe01d2db9eb893bd ] This needs to be set to 1 to avoid a potential deadlock in the GC 10.x and newer. On GC 9.x and older, this needs to be set to 0. This can lead to hangs in some mixed graphics and compute workloads. Updated firmware is also required for AQL. Reviewed-by: Feifei Xu Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org Signed-off-by: Sasha Levin

drm/amdkfd: fixes for HMM mem allocation

2024-01-25T23:35:58+00:00

[ Upstream commit 02eed83abc1395a1207591aafad9bcfc5cb1abcb ] Fix err return value and reset pgmap->type after checking it. Fixes: c83dee9b6394 ("drm/amdkfd: add SPM support for SVM") Reviewed-by: Felix Kuehling Signed-off-by: Dafna Hirschfeld Signed-off-by: Felix Kuehling Signed-off-by: Alex Deucher Signed-off-by: Sasha Levin

Revert "drm/amdkfd: Relocate TBA/TMA to opposite side of VM hole"

2024-01-25T23:35:41+00:00

commit 0f35b0a7b8fa402adbffa2565047cdcc4c480153 upstream. That commit causes NULL pointer dereferences in dmesgs when running applications using ROCm, including clinfo, blender, and PyTorch, since v6.6.1. Revert it to fix blender again. This reverts commit 96c211f1f9ef82183493f4ceed4e347b52849149. Closes: https://github.com/ROCm/ROCm/issues/2596 Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/2991 Reviewed-by: Jay Cornwall Signed-off-by: Kaibo Ma Signed-off-by: Alex Deucher Signed-off-by: Greg Kroah-Hartman