This is a note to let you know that I've just added the patch titled
drm/amdkfd: Fixed kfd_process cleanup on module exit.
to the 6.2-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary
The filename of the patch is:
drm-amdkfd-fixed-kfd_process-cleanup-on-module-exit.patch
and it can be found in the queue-6.2 subdirectory.
If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@xxxxxxxxxxxxxxx> know about it.
commit 81e2ee644325118a3526cb75c61beb3c1ce48ff9
Author: David Belanger <david.belanger@xxxxxxx>
Date: Tue Feb 28 14:11:24 2023 -0500
drm/amdkfd: Fixed kfd_process cleanup on module exit.
[ Upstream commit 20bc9f76b6a2455c6b54b91ae7634f147f64987f ]
Handle case when module is unloaded (kfd_exit) before a process space
(mm_struct) is released.
v2: Fixed potential race conditions by removing all kfd_process from
the process table first, then working on releasing the resources.
v3: Fixed loop element access / synchronization. Fixed extra empty lines.
Signed-off-by: David Belanger <david.belanger@xxxxxxx>
Reviewed-by: Felix Kuehling <Felix.Kuehling@xxxxxxx>
Signed-off-by: Alex Deucher <alexander.deucher@xxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
index 09b966dc37681..aee2212e52f69 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c
@@ -77,6 +77,7 @@ static int kfd_init(void)
static void kfd_exit(void)
{
+ kfd_cleanup_processes();
kfd_debugfs_fini();
kfd_process_destroy_wq();
kfd_procfs_shutdown();
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
index 552c3ac85a132..7dc55919993c0 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h
@@ -926,6 +926,7 @@ bool kfd_dev_is_large_bar(struct kfd_dev *dev);
int kfd_process_create_wq(void);
void kfd_process_destroy_wq(void);
+void kfd_cleanup_processes(void);
struct kfd_process *kfd_create_process(struct file *filep);
struct kfd_process *kfd_get_process(const struct task_struct *task);
struct kfd_process *kfd_lookup_process_by_pasid(u32 pasid);
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 51b1683ac5c1e..4d9f2d1c49b1d 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1167,6 +1167,17 @@ static void kfd_process_free_notifier(struct mmu_notifier *mn)
kfd_unref_process(container_of(mn, struct kfd_process, mmu_notifier));
}
+static void kfd_process_notifier_release_internal(struct kfd_process *p)
+{
+ cancel_delayed_work_sync(&p->eviction_work);
+ cancel_delayed_work_sync(&p->restore_work);
+
+ /* Indicate to other users that MM is no longer valid */
+ p->mm = NULL;
+
+ mmu_notifier_put(&p->mmu_notifier);
+}
+
static void kfd_process_notifier_release(struct mmu_notifier *mn,
struct mm_struct *mm)
{
@@ -1181,17 +1192,22 @@ static void kfd_process_notifier_release(struct mmu_notifier *mn,
return;
mutex_lock(&kfd_processes_mutex);
+ /*
+ * Do early return if table is empty.
+ *
+ * This could potentially happen if this function is called concurrently
+ * by mmu_notifier and by kfd_cleanup_pocesses.
+ *
+ */
+ if (hash_empty(kfd_processes_table)) {
+ mutex_unlock(&kfd_processes_mutex);
+ return;
+ }
hash_del_rcu(&p->kfd_processes);
mutex_unlock(&kfd_processes_mutex);
synchronize_srcu(&kfd_processes_srcu);
- cancel_delayed_work_sync(&p->eviction_work);
- cancel_delayed_work_sync(&p->restore_work);
-
- /* Indicate to other users that MM is no longer valid */
- p->mm = NULL;
-
- mmu_notifier_put(&p->mmu_notifier);
+ kfd_process_notifier_release_internal(p);
}
static const struct mmu_notifier_ops kfd_process_mmu_notifier_ops = {
@@ -1200,6 +1216,43 @@ static const struct mmu_notifier_ops kfd_process_mmu_notifier_ops = {
.free_notifier = kfd_process_free_notifier,
};
+/*
+ * This code handles the case when driver is being unloaded before all
+ * mm_struct are released. We need to safely free the kfd_process and
+ * avoid race conditions with mmu_notifier that might try to free them.
+ *
+ */
+void kfd_cleanup_processes(void)
+{
+ struct kfd_process *p;
+ struct hlist_node *p_temp;
+ unsigned int temp;
+ HLIST_HEAD(cleanup_list);
+
+ /*
+ * Move all remaining kfd_process from the process table to a
+ * temp list for processing. Once done, callback from mmu_notifier
+ * release will not see the kfd_process in the table and do early return,
+ * avoiding double free issues.
+ */
+ mutex_lock(&kfd_processes_mutex);
+ hash_for_each_safe(kfd_processes_table, temp, p_temp, p, kfd_processes) {
+ hash_del_rcu(&p->kfd_processes);
+ synchronize_srcu(&kfd_processes_srcu);
+ hlist_add_head(&p->kfd_processes, &cleanup_list);
+ }
+ mutex_unlock(&kfd_processes_mutex);
+
+ hlist_for_each_entry_safe(p, p_temp, &cleanup_list, kfd_processes)
+ kfd_process_notifier_release_internal(p);
+
+ /*
+ * Ensures that all outstanding free_notifier get called, triggering
+ * the release of the kfd_process struct.
+ */
+ mmu_notifier_synchronize();
+}
+
static int kfd_process_init_cwsr_apu(struct kfd_process *p, struct file *filep)
{
unsigned long offset;
