drm/amdkfd: Fix some double free when destroy queue fails
Subsystems (3)
PATCH 1/2 drm/amdkfd: Fix some double free when destroy queue fails
Some resource are freed even destroy queue fails. That will cause double free when user-space issue another destroy_queue ioctl.
Paste some log below.
amdgpu: Can't create new usermode queue because -1 queues were already created
refcount_t: underflow; use-after-free.
Call Trace: kobject_put+0xe6/0x1b0 kfd_procfs_del_queue+0x37/0x50 [amdgpu] pqm_destroy_queue+0x17a/0x390 [amdgpu] kfd_ioctl_destroy_queue+0x57/0xc0 [amdgpu] kfd_ioctl+0x463/0x690 [amdgpu]
BUG kmalloc-32 (Tainted: G W ): Object already free
INFO: Allocated in allocate_sdma_mqd+0x30/0xb0 [amdgpu] age=4796 cpu=2 pid=2511
__slab_alloc+0x72/0x80 kmem_cache_alloc_trace+0x81f/0x8c0 allocate_sdma_mqd+0x30/0xb0 [amdgpu] create_queue_cpsch+0xbf/0x470 [amdgpu] pqm_create_queue+0x28d/0x6d0 [amdgpu] kfd_ioctl_create_queue+0x492/0xae0 [amdgpu] INFO: Freed in free_mqd_hiq_sdma+0x20/0x60 [amdgpu] age=2537 cpu=7 pid=2511 kfree+0x322/0x340 free_mqd_hiq_sdma+0x20/0x60 [amdgpu] destroy_queue_cpsch+0x20c/0x330 [amdgpu] pqm_destroy_queue+0x1a3/0x390 [amdgpu] kfd_ioctl_destroy_queue+0x57/0xc0 [amdgpu]
--- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 ++ drivers/gpu/drm/amd/amdkfd/kfd_process.c | 4 +++- drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 1 + 3 files changed, 6 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +2 -0
@@ -1529,6 +1529,8 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);if (retval == -ETIME ) qpd->reset_wavefronts = true; +else if (retval) + goto failed_try_destroy_debugged_queue;if (q->properties.is_gws) { dqm->gws_queue_count--; qpd->mapped_gws_queue = false;
drivers/gpu/drm/amd/amdkfd/kfd_process.c +3 -1
@@ -607,11 +607,13 @@ static int kfd_procfs_add_sysfs_files(struct kfd_process *p)void kfd_procfs_del_queue (struct queue *q) { -if (!q) +if (!q || !kobject_get_unless_zero (&q->kobj))return ;kobject_del (&q->kobj);kobject_put (&q->kobj); +/* paired with the get above */ +kobject_put (&q->kobj); }int kfd_process_create_wq (void )
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +1 -0
@@ -373,6 +373,7 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) dqm = pqn->kq->dev->dqm; dqm->ops.destroy_kernel_queue (dqm, pqn->kq, &pdd->qpd);kernel_queue_uninit (pqn->kq, false); + pqn->kq =NULL ; }if (pqn->q) {
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 2/2 drm/amdkfd: Walk thorugh list with dqm lock hold
To avoid any list corruption.
--- .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +8 -4
@@ -1704,7 +1704,7 @@ static int process_termination_cpsch(struct device_queue_manager *dqm, struct qcm_process_device *qpd) {int retval; -struct queue *q, *next; +struct queue *q;struct kernel_queue *kq, *kq_next;struct mqd_manager *mqd_mgr;struct device_process_node *cur, *next_dpn;
@@ -1739,8 +1739,6 @@ static int process_termination_cpsch(struct device_queue_manager *dqm, qpd->mapped_gws_queue = false; } } - - dqm->total_queue_count--; }/* Unregister process */
@@ -1772,13 +1770,19 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,/* Lastly, free mqd resources. * Do free_mqd() after dqm_unlock to avoid circular locking. */ -list_for_each_entry_safe (q, next, &qpd->queues_list, list) { +dqm_lock (dqm); +while (!list_empty (&qpd->queues_list)) { + q = list_first_entry(&qpd->queues_list, struct queue, list); mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type ( q->properties.type)];list_del (&q->list); qpd->queue_count--; + dqm->total_queue_count--; +dqm_unlock (dqm); mqd_mgr->free_mqd (mqd_mgr, q->mqd, q->mqd_mem_obj); +dqm_lock (dqm); } +dqm_unlock (dqm);return retval; }
Signature · 6 lines
2.25.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Felix Kuehling
maintainerRe: [PATCH 2/2] drm/amdkfd: Walk thorugh list with dqm lock hold
Quoted 21 lines · xinhui pan
To avoid any list corruption.
Signed-off-by xinhui pan <[email protected]>--- .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-)drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +1
-1@@ -1704,7 +1704,7 @@ static int process_termination_cpsch(struct device_queue_manager *dqm, struct qcm_process_device *qpd) {int retval; -struct queue *q, *next; +struct queue *q;struct kernel_queue *kq, *kq_next;struct mqd_manager *mqd_mgr;struct device_process_node *cur, *next_dpn;
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c static int process_termination_cpsch(struct device_queue_manager *dqm, · line 1743
Quoted 5 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +0
-2@@ -1739,8 +1739,6 @@ static int process_termination_cpsch(struct device_queue_manager *dqm, qpd->mapped_gws_queue = false; } } - - dqm->total_queue_count--;
I think this should stay here. This is only used to check the maximum user queue limit per-device, which is a HW limitation. As far as the HW is concerned, the queues are destroyed after the call to execute_queues_cpsch. So there is no need to delay this update.
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c static int process_termination_cpsch(struct device_queue_manager *dqm, · line 1773
Quoted 9 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +1
-1}/* Unregister process */ @@ -1772,13 +1770,19 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,/* Lastly, free mqd resources. * Do free_mqd() after dqm_unlock to avoid circular locking. */ -list_for_each_entry_safe (q, next, &qpd->queues_list, list) { +dqm_lock (dqm);
Instead of taking the dqm lock again, just move this up a couple of lines before the dqm_unlock call.
Regards,
Felix
Quoted 5 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +2
-0+while (!list_empty (&qpd->queues_list)) { + q = list_first_entry(&qpd->queues_list, struct queue, list); mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type ( q->properties.type)];list_del (&q->list);
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c static int process_termination_cpsch(struct device_queue_manager *dqm, · line 1788
Quoted 10 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +4
-0qpd->queue_count--; + dqm->total_queue_count--; +dqm_unlock (dqm); mqd_mgr->free_mqd (mqd_mgr, q->mqd, q->mqd_mem_obj); +dqm_lock (dqm); } +dqm_unlock (dqm);return retval; }
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
RE: [PATCH 2/2] drm/amdkfd: Walk thorugh list with dqm lock hold
[AMD Official Use Only]
BTW, there is an obvious typo in the subject, Walk thorugh => Walk through.
Regards,
Jiansong
-----Original Message-----
From: amd-gfx <[email protected]> On Behalf Of Felix Kuehling
Sent: Thursday, June 17, 2021 7:09 AM
To: Pan, Xinhui <[email protected]>; [email protected]
Cc: Deucher, Alexander <[email protected]>; Koenig, Christian <[email protected]>
Subject: Re: [PATCH 2/2] drm/amdkfd: Walk thorugh list with dqm lock hold
Quoted 25 lines · xinhui pan
To avoid any list corruption.
Signed-off-by xinhui pan <[email protected]>--- .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-)drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +1
-1@@ -1704,7 +1704,7 @@ static int process_termination_cpsch(struct device_queue_manager *dqm, struct qcm_process_device *qpd) {int retval; -struct queue *q, *next; +struct queue *q;struct kernel_queue *kq, *kq_next;struct mqd_manager *mqd_mgr; struct device_process_node *cur, *next_dpn; @@ -1739,8 +1739,6 @@static int process_termination_cpsch(struct
device_queue_manager*dqm,qpd->mapped_gws_queue = false; } }
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c static int process_termination_cpsch(struct device_queue_manager *dqm, · line 1743
Quoted 2 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +0
-2- - dqm->total_queue_count--;
I think this should stay here. This is only used to check the maximum user queue limit per-device, which is a HW limitation. As far as the HW is concerned, the queues are destroyed after the call to execute_queues_cpsch. So there is no need to delay this update.
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c static int process_termination_cpsch(struct device_queue_manager *dqm, · line 1773
Quoted 9 lines
} /* Unregister process */drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +1
-1@@ -1772,13 +1770,19 @@ static int process_termination_cpsch(struct device_queue_manager *dqm,/* Lastly, free mqd resources. * Do free_mqd() after dqm_unlock to avoid circular locking. */ -list_for_each_entry_safe (q, next, &qpd->queues_list, list) { +dqm_lock (dqm);
Instead of taking the dqm lock again, just move this up a couple of lines before the dqm_unlock call.
Regards,
Felix
Quoted 5 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +2
-0+while (!list_empty (&qpd->queues_list)) { + q = list_first_entry(&qpd->queues_list, struct queue, list); mqd_mgr = dqm->mqd_mgrs[get_mqd_type_from_queue_type ( q->properties.type)];list_del (&q->list);
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c static int process_termination_cpsch(struct device_queue_manager *dqm, · line 1788
Quoted 10 lines
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +4
-0qpd->queue_count--; + dqm->total_queue_count--; +dqm_unlock (dqm); mqd_mgr->free_mqd (mqd_mgr, q->mqd, q->mqd_mem_obj); +dqm_lock (dqm); } +dqm_unlock (dqm);return retval; }
_______________________________________________
amd-gfx mailing list
[email protected]
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Flists.freedesktop.org%2Fmailman%2Flistinfo%2Famd-gfx&data=04%7C01%7CJiansong.Chen%40amd.com%7C47251235ca70449c924608d9311bc4ce%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637594817623083340%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=h%2FzO6DHA%2F6%2Btw0iJp4aBFfw8KZPVgtmgkfj3VQho4pM%3D&reserved=0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Felix Kuehling
maintainerQuoted 1 line · xinhui pan
Some resource are freed even destroy queue fails.
Looks like you're keeping this behaviour for -ETIME. That is consistent with what pqn_destroy_queue does. What you're fixing here is the behaviour for non-timeout errors. Please make that clear in the patch description.
Out of curiosity, what kind of error were you getting? The only other ones that are not a fatal memory shortage, are some EINVAL cases in pm_unmap_queues_v*. But that would indicate some internal error, that a queue was created with an invalid type, or maybe the queue data structure was somehow corrupted.
Quoted 77 lines
That will cause double
free when user-space issue anotherdestroy_queueioctl.Paste some log below.
amdgpu: Can't create new usermode queue because -1 queues were already created
refcount_t: underflow; use-after-free.Call Trace: kobject_put+0xe6/0x1b0 kfd_procfs_del_queue+0x37/0x50 [amdgpu] pqm_destroy_queue+0x17a/0x390 [amdgpu] kfd_ioctl_destroy_queue+0x57/0xc0 [amdgpu] kfd_ioctl+0x463/0x690 [amdgpu]BUG kmalloc-32 (Tainted: G W ): Object already free
INFO: Allocated inallocate_sdma_mqd+0x30/0xb0[amdgpu] age=4796 cpu=2 pid=2511__slab_alloc+0x72/0x80 kmem_cache_alloc_trace+0x81f/0x8c0 allocate_sdma_mqd+0x30/0xb0 [amdgpu] create_queue_cpsch+0xbf/0x470 [amdgpu] pqm_create_queue+0x28d/0x6d0 [amdgpu] kfd_ioctl_create_queue+0x492/0xae0 [amdgpu] INFO: Freed in free_mqd_hiq_sdma+0x20/0x60 [amdgpu] age=2537 cpu=7 pid=2511 kfree+0x322/0x340 free_mqd_hiq_sdma+0x20/0x60 [amdgpu] destroy_queue_cpsch+0x20c/0x330 [amdgpu] pqm_destroy_queue+0x1a3/0x390 [amdgpu] kfd_ioctl_destroy_queue+0x57/0xc0 [amdgpu]Signed-off-by xinhui pan <[email protected]>--- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 ++ drivers/gpu/drm/amd/amdkfd/kfd_process.c | 4 +++- drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 1 + 3 files changed, 6 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +2
-0@@ -1529,6 +1529,8 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);if (retval == -ETIME ) qpd->reset_wavefronts = true; +else if (retval) + goto failed_try_destroy_debugged_queue;if (q->properties.is_gws) { dqm->gws_queue_count--; qpd->mapped_gws_queue = false;drivers/gpu/drm/amd/amdkfd/kfd_process.c +3
-1@@ -607,11 +607,13 @@ static int kfd_procfs_add_sysfs_files(struct kfd_process *p)void kfd_procfs_del_queue (struct queue *q) { -if (!q) +if (!q || !kobject_get_unless_zero (&q->kobj))return ;kobject_del (&q->kobj);kobject_put (&q->kobj); +/* paired with the get above */ +kobject_put (&q->kobj); }int kfd_process_create_wq (void )drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +0
-0
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) · line 376
Quoted 4 lines
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +1
-0@@ -373,6 +373,7 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) dqm = pqn->kq->dev->dqm; dqm->ops.destroy_kernel_queue (dqm, pqn->kq, &pdd->qpd);kernel_queue_uninit (pqn->kq, false); + pqn->kq =NULL ;
This seems unrelated to this patch. But if you're fixing this, I'd expect a similar fix after uninit_queue(pqn->q).
Regards,
Felix
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) · line 379
Quoted 3 lines
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +0
-0}if (pqn->q) {
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Quoted 8 lines
2021年6月17日 06:55,Kuehling, Felix <[email protected]> 写道:
Quoted 1 line · xinhui pan
Some resource are freed even destroy queue fails.
Looks like you're keeping this behaviour for
-ETIME. That is consistent with whatpqn_destroy_queuedoes. What you're fixing here is the behaviour for non-timeout errors. Please make that clear in the patch description.
will do that in v2.
Quoted 3 lines
Out of curiosity, what kind of error were you getting? The only other ones that are not a fatal memory shortage, are some EINVAL cases in pm_unmap_queues_v*. But that would indicate some internal error, that a queue was created with an invalid type, or maybe the queue data structure was somehow corrupted.
This is just because amdkfd_fence_wait_timeout got timeout. execute_queues_cpsch return EIO as dqm->is_hws_hang is true. hit this issue with kfdtest --gtest_filter=*QM*
Quoted 82 lines
Quoted 75 lines
That will cause double
free when user-space issue anotherdestroy_queueioctl.Paste some log below.
amdgpu: Can't create new usermode queue because -1 queues were already created
refcount_t: underflow; use-after-free.Call Trace: kobject_put+0xe6/0x1b0 kfd_procfs_del_queue+0x37/0x50 [amdgpu] pqm_destroy_queue+0x17a/0x390 [amdgpu] kfd_ioctl_destroy_queue+0x57/0xc0 [amdgpu] kfd_ioctl+0x463/0x690 [amdgpu]BUG kmalloc-32 (Tainted: G W ): Object already free
INFO: Allocated inallocate_sdma_mqd+0x30/0xb0[amdgpu] age=4796 cpu=2 pid=2511__slab_alloc+0x72/0x80 kmem_cache_alloc_trace+0x81f/0x8c0 allocate_sdma_mqd+0x30/0xb0 [amdgpu] create_queue_cpsch+0xbf/0x470 [amdgpu] pqm_create_queue+0x28d/0x6d0 [amdgpu] kfd_ioctl_create_queue+0x492/0xae0 [amdgpu] INFO: Freed in free_mqd_hiq_sdma+0x20/0x60 [amdgpu] age=2537 cpu=7 pid=2511 kfree+0x322/0x340 free_mqd_hiq_sdma+0x20/0x60 [amdgpu] destroy_queue_cpsch+0x20c/0x330 [amdgpu] pqm_destroy_queue+0x1a3/0x390 [amdgpu] kfd_ioctl_destroy_queue+0x57/0xc0 [amdgpu]Signed-off-by xinhui pan <[email protected]>--- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 ++ drivers/gpu/drm/amd/amdkfd/kfd_process.c | 4 +++- drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 1 + 3 files changed, 6 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +2
-0@@ -1529,6 +1529,8 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm, KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES, 0);if (retval == -ETIME ) qpd->reset_wavefronts = true; +else if (retval) + goto failed_try_destroy_debugged_queue;if (q->properties.is_gws) { dqm->gws_queue_count--; qpd->mapped_gws_queue = false;drivers/gpu/drm/amd/amdkfd/kfd_process.c +3
-1@@ -607,11 +607,13 @@ static int kfd_procfs_add_sysfs_files(struct kfd_process *p)void kfd_procfs_del_queue (struct queue *q) { -if (!q) +if (!q || !kobject_get_unless_zero (&q->kobj))return ;kobject_del (&q->kobj);kobject_put (&q->kobj); +/* paired with the get above */ +kobject_put (&q->kobj); }int kfd_process_create_wq (void )drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +0
-0drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) · line 376
Quoted 4 lines
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +1
-0@@ -373,6 +373,7 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) dqm = pqn->kq->dev->dqm; dqm->ops.destroy_kernel_queue (dqm, pqn->kq, &pdd->qpd);kernel_queue_uninit (pqn->kq, false); + pqn->kq =NULL ;This seems unrelated to this patch. But if you're fixing this, I'd expect a similar fix after
uninit_queue(pqn->q).
Quoted 6 lines
Regards,
FelixQuoted 2 lines
drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +0
-0}if (pqn->q) {
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx