We use mmu_notifier_put to free the MMU notifier. That needs to be paired with mmu_notifier_get to work correctly. Othewrise the next patch would cause a kernel oops.

Signed-off-by: Felix Kuehling Felix.Kuehling@amd.com --- drivers/gpu/drm/amd/amdkfd/kfd_process.c | 33 +++++++++++++++++++----- 1 file changed, 27 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 2807e1c4d59b..145cd0a17d50 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -1011,6 +1011,16 @@ static void kfd_process_ref_release(struct kref *ref) queue_work(kfd_process_wq, &p->release_work); }

+static struct mmu_notifier *kfd_process_alloc_notifier(struct mm_struct *mm) +{ + int idx = srcu_read_lock(&kfd_processes_srcu); + struct kfd_process *p = find_process_by_mm(mm); + + srcu_read_unlock(&kfd_processes_srcu, idx); + + return p ? &p->mmu_notifier : ERR_PTR(-ESRCH); +} + static void kfd_process_free_notifier(struct mmu_notifier *mn) { kfd_unref_process(container_of(mn, struct kfd_process, mmu_notifier)); @@ -1075,6 +1085,7 @@ static void kfd_process_notifier_release(struct mmu_notifier *mn,

static const struct mmu_notifier_ops kfd_process_mmu_notifier_ops = { .release = kfd_process_notifier_release, + .alloc_notifier = kfd_process_alloc_notifier, .free_notifier = kfd_process_free_notifier, };

@@ -1152,6 +1163,7 @@ static int kfd_process_device_init_cwsr_dgpu(struct kfd_process_device *pdd) static struct kfd_process *create_process(const struct task_struct *thread) { struct kfd_process *process; + struct mmu_notifier *mn; int err = -ENOMEM;

process = kzalloc(sizeof(*process), GFP_KERNEL); @@ -1182,19 +1194,28 @@ static struct kfd_process *create_process(const struct task_struct *thread) if (err != 0) goto err_init_apertures;

- /* Must be last, have to use release destruction after this */ - process->mmu_notifier.ops = &kfd_process_mmu_notifier_ops; - err = mmu_notifier_register(&process->mmu_notifier, process->mm); - if (err) + /* alloc_notifier needs to find the process in the hash table */ + hash_add_rcu(kfd_processes_table, &process->kfd_processes, + (uintptr_t)process->mm); + + /* MMU notifier registration must be the last call that can fail + * because after this point we cannot unwind the process creation. + * After this point, mmu_notifier_put will trigger the cleanup by + * dropping the last process reference in the free_notifier. + */ + mn = mmu_notifier_get(&kfd_process_mmu_notifier_ops, process->mm); + if (IS_ERR(mn)) { + err = PTR_ERR(mn); goto err_register_notifier; + } + BUG_ON(mn != &process->mmu_notifier);

get_task_struct(process->lead_thread); - hash_add_rcu(kfd_processes_table, &process->kfd_processes, - (uintptr_t)process->mm);

return process;

err_register_notifier: + hash_del_rcu(&process->kfd_processes); kfd_process_free_outstanding_kfd_bos(process); kfd_process_destroy_pdds(process); err_init_apertures: