All of lore.kernel.org
 help / color / mirror / Atom feed
* [RHEL7 drm 26/34] drm/amdkfd: unbind only existing processes
       [not found] <1465517793-14679-1-git-send-email-rclark@redhat.com>
@ 2016-06-10  0:16 ` Rob Clark
  2016-06-10  0:16 ` [RHEL7 drm 27/34] drm/amdkfd: destroy dbgmgr in notifier release Rob Clark
  1 sibling, 0 replies; 2+ messages in thread
From: Rob Clark @ 2016-06-10  0:16 UTC (permalink / raw)
  To: rhkernel-list; +Cc: Oded Gabbay, Stable, David Airlie, Adam Jackson, Rob Clark

From: Oded Gabbay <oded.gabbay@gmail.com>

Upstream: Posted
RHBZ: 1328491

When unbinding a process from a device (initiated by amd_iommu_v2), the
driver needs to make sure that process still exists in the process table.
There is a possibility that amdkfd's own notifier handler -
kfd_process_notifier_release() - was called before the unbind function
and it already removed the process from the process table.

v2:
Because there can be only one process with the specified pasid, and
because *p can't be NULL inside the hash_for_each_rcu macro, it is more
reasonable to just put the whole code inside the if statement that
compares the pasid value. That way, when we exit hash_for_each_rcu, we
simply exit the function as well.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
CC: Stable <stable@vger.kernel.org>
---
 drivers/gpu/drm/amd/amdkfd/kfd_process.c | 60 +++++++++++++++++++-------------
 1 file changed, 35 insertions(+), 25 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index ac00579..a64bc61 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -404,42 +404,52 @@ void kfd_unbind_process_from_device(struct kfd_dev *dev, unsigned int pasid)
 
 	idx = srcu_read_lock(&kfd_processes_srcu);
 
+	/*
+	 * Look for the process that matches the pasid. If there is no such
+	 * process, we either released it in amdkfd's own notifier, or there
+	 * is a bug. Unfortunately, there is no way to tell...
+	 */
 	hash_for_each_rcu(kfd_processes_table, i, p, kfd_processes)
-		if (p->pasid == pasid)
-			break;
+		if (p->pasid == pasid) {
 
-	srcu_read_unlock(&kfd_processes_srcu, idx);
+			srcu_read_unlock(&kfd_processes_srcu, idx);
 
-	BUG_ON(p->pasid != pasid);
+			pr_debug("Unbinding process %d from IOMMU\n", pasid);
 
-	mutex_lock(&p->mutex);
+			mutex_lock(&p->mutex);
 
-	if ((dev->dbgmgr) && (dev->dbgmgr->pasid == p->pasid))
-		kfd_dbgmgr_destroy(dev->dbgmgr);
+			if ((dev->dbgmgr) && (dev->dbgmgr->pasid == p->pasid))
+				kfd_dbgmgr_destroy(dev->dbgmgr);
 
-	pqm_uninit(&p->pqm);
+			pqm_uninit(&p->pqm);
 
-	pdd = kfd_get_process_device_data(dev, p);
+			pdd = kfd_get_process_device_data(dev, p);
 
-	if (!pdd) {
-		mutex_unlock(&p->mutex);
-		return;
-	}
+			if (!pdd) {
+				mutex_unlock(&p->mutex);
+				return;
+			}
 
-	if (pdd->reset_wavefronts) {
-		dbgdev_wave_reset_wavefronts(pdd->dev, p);
-		pdd->reset_wavefronts = false;
-	}
+			if (pdd->reset_wavefronts) {
+				dbgdev_wave_reset_wavefronts(pdd->dev, p);
+				pdd->reset_wavefronts = false;
+			}
 
-	/*
-	 * Just mark pdd as unbound, because we still need it to call
-	 * amd_iommu_unbind_pasid() in when the process exits.
-	 * We don't call amd_iommu_unbind_pasid() here
-	 * because the IOMMU called us.
-	 */
-	pdd->bound = false;
+			/*
+			 * Just mark pdd as unbound, because we still need it
+			 * to call amd_iommu_unbind_pasid() in when the
+			 * process exits.
+			 * We don't call amd_iommu_unbind_pasid() here
+			 * because the IOMMU called us.
+			 */
+			pdd->bound = false;
 
-	mutex_unlock(&p->mutex);
+			mutex_unlock(&p->mutex);
+
+			return;
+		}
+
+	srcu_read_unlock(&kfd_processes_srcu, idx);
 }
 
 struct kfd_process_device *kfd_get_first_process_device_data(struct kfd_process *p)
-- 
2.5.5


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* [RHEL7 drm 27/34] drm/amdkfd: destroy dbgmgr in notifier release
       [not found] <1465517793-14679-1-git-send-email-rclark@redhat.com>
  2016-06-10  0:16 ` [RHEL7 drm 26/34] drm/amdkfd: unbind only existing processes Rob Clark
@ 2016-06-10  0:16 ` Rob Clark
  1 sibling, 0 replies; 2+ messages in thread
From: Rob Clark @ 2016-06-10  0:16 UTC (permalink / raw)
  To: rhkernel-list; +Cc: Oded Gabbay, Stable, David Airlie, Adam Jackson, Rob Clark

From: Oded Gabbay <oded.gabbay@gmail.com>

Upstream: Posted
RHBZ: 1328491

amdkfd need to destroy the debug manager in case amdkfd's notifier
function is called before the unbind function, because in that case,
the unbind function will exit without destroying debug manager.

Signed-off-by: Oded Gabbay <oded.gabbay@gmail.com>
CC: Stable <stable@vger.kernel.org>
---
 drivers/gpu/drm/amd/amdkfd/kfd_process.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index a64bc61..7708d90 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -242,13 +242,19 @@ static void kfd_process_notifier_release(struct mmu_notifier *mn,
 	pqm_uninit(&p->pqm);
 
 	/* Iterate over all process device data structure and check
-	 * if we should reset all wavefronts */
-	list_for_each_entry(pdd, &p->per_device_data, per_device_list)
+	 * if we should delete debug managers and reset all wavefronts
+	 */
+	list_for_each_entry(pdd, &p->per_device_data, per_device_list) {
+		if ((pdd->dev->dbgmgr) &&
+				(pdd->dev->dbgmgr->pasid == p->pasid))
+			kfd_dbgmgr_destroy(pdd->dev->dbgmgr);
+
 		if (pdd->reset_wavefronts) {
 			pr_warn("amdkfd: Resetting all wave fronts\n");
 			dbgdev_wave_reset_wavefronts(pdd->dev, p);
 			pdd->reset_wavefronts = false;
 		}
+	}
 
 	mutex_unlock(&p->mutex);
 
-- 
2.5.5


^ permalink raw reply related	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2016-06-10  0:17 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <1465517793-14679-1-git-send-email-rclark@redhat.com>
2016-06-10  0:16 ` [RHEL7 drm 26/34] drm/amdkfd: unbind only existing processes Rob Clark
2016-06-10  0:16 ` [RHEL7 drm 27/34] drm/amdkfd: destroy dbgmgr in notifier release Rob Clark

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.