When the xe pm_notifier evicts for suspend / hibernate, there might be racing tasks trying to re-validate again. This can lead to suspend taking excessive time or get stuck in a live-lock. This behaviour becomes much worse with the fix that actually makes re-validation bring back bos to VRAM rather than letting them remain in TT.
Prevent that by having exec and the rebind worker waiting for a completion that is set to block by the pm_notifier before suspend and is signaled by the pm_notifier after resume / wakeup.
It's probably still possible to craft malicious applications that block suspending. More work is pending to fix that.
Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/4288 Fixes: c6a4d46ec1d7 ("drm/xe: evict user memory in PM notifier") Cc: Matthew Auld matthew.auld@intel.com Cc: Rodrigo Vivi rodrigo.vivi@intel.com Cc: stable@vger.kernel.org # v6.16+ Signed-off-by: Thomas Hellström thomas.hellstrom@linux.intel.com --- drivers/gpu/drm/xe/xe_device_types.h | 2 ++ drivers/gpu/drm/xe/xe_exec.c | 9 +++++++++ drivers/gpu/drm/xe/xe_pm.c | 4 ++++ drivers/gpu/drm/xe/xe_vm.c | 14 ++++++++++++++ 4 files changed, 29 insertions(+)
diff --git a/drivers/gpu/drm/xe/xe_device_types.h b/drivers/gpu/drm/xe/xe_device_types.h index 092004d14db2..6602bd678cbc 100644 --- a/drivers/gpu/drm/xe/xe_device_types.h +++ b/drivers/gpu/drm/xe/xe_device_types.h @@ -507,6 +507,8 @@ struct xe_device {
/** @pm_notifier: Our PM notifier to perform actions in response to various PM events. */ struct notifier_block pm_notifier; + /** @pm_block: Completion to block validating tasks on suspend / hibernate prepare */ + struct completion pm_block;
/** @pmt: Support the PMT driver callback interface */ struct { diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c index 44364c042ad7..374c831e691b 100644 --- a/drivers/gpu/drm/xe/xe_exec.c +++ b/drivers/gpu/drm/xe/xe_exec.c @@ -237,6 +237,15 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file) goto err_unlock_list; }
+ /* + * It's OK to block interruptible here with the vm lock held, since + * on task freezing during suspend / hibernate, the call will + * return -ERESTARTSYS and the IOCTL will be rerun. + */ + err = wait_for_completion_interruptible(&xe->pm_block); + if (err) + goto err_unlock_list; + vm_exec.vm = &vm->gpuvm; vm_exec.flags = DRM_EXEC_INTERRUPTIBLE_WAIT; if (xe_vm_in_lr_mode(vm)) { diff --git a/drivers/gpu/drm/xe/xe_pm.c b/drivers/gpu/drm/xe/xe_pm.c index bee9aacd82e7..f4c7a84270ea 100644 --- a/drivers/gpu/drm/xe/xe_pm.c +++ b/drivers/gpu/drm/xe/xe_pm.c @@ -306,6 +306,7 @@ static int xe_pm_notifier_callback(struct notifier_block *nb, switch (action) { case PM_HIBERNATION_PREPARE: case PM_SUSPEND_PREPARE: + reinit_completion(&xe->pm_block); xe_pm_runtime_get(xe); err = xe_bo_evict_all_user(xe); if (err) @@ -322,6 +323,7 @@ static int xe_pm_notifier_callback(struct notifier_block *nb, break; case PM_POST_HIBERNATION: case PM_POST_SUSPEND: + complete_all(&xe->pm_block); xe_bo_notifier_unprepare_all_pinned(xe); xe_pm_runtime_put(xe); break; @@ -348,6 +350,8 @@ int xe_pm_init(struct xe_device *xe) if (err) return err;
+ init_completion(&xe->pm_block); + complete_all(&xe->pm_block); /* For now suspend/resume is only allowed with GuC */ if (!xe_device_uc_enabled(xe)) return 0; diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index 3ff3c67aa79d..edcdb0528f2a 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -394,6 +394,9 @@ static int xe_gpuvm_validate(struct drm_gpuvm_bo *vm_bo, struct drm_exec *exec) list_move_tail(&gpuva_to_vma(gpuva)->combined_links.rebind, &vm->rebind_list);
+ if (!try_wait_for_completion(&vm->xe->pm_block)) + return -EAGAIN; + ret = xe_bo_validate(gem_to_xe_bo(vm_bo->obj), vm, false); if (ret) return ret; @@ -494,6 +497,12 @@ static void preempt_rebind_work_func(struct work_struct *w) xe_assert(vm->xe, xe_vm_in_preempt_fence_mode(vm)); trace_xe_vm_rebind_worker_enter(vm);
+ /* + * This blocks the wq during suspend / hibernate. + * Don't hold any locks. + */ +retry_pm: + wait_for_completion(&vm->xe->pm_block); down_write(&vm->lock);
if (xe_vm_is_closed_or_banned(vm)) { @@ -503,6 +512,11 @@ static void preempt_rebind_work_func(struct work_struct *w) }
retry: + if (!try_wait_for_completion(&vm->xe->pm_block)) { + up_write(&vm->lock); + goto retry_pm; + } + if (xe_vm_userptr_check_repin(vm)) { err = xe_vm_userptr_pin(vm); if (err)
linux-stable-mirror@lists.linaro.org