Skip to contentSkip to conversation

vmd: Fix infinite loop executing irq's

PATCH In review Keith Busch 3 participants 4 messages Aug 8, 2016 – Aug 23, 2016

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 vmd: Fix infinite loop executing irq's

We can't initialize the list head on deletion as this causes the node to point to itself, looping infinitely if the vmd IRQ handler happened to be servicing that node.

The list initialization was trying fix a bug from multiple calls to disable the same IRQ. We can fix this instead by having the vmd driver track if the interrupt is enabled.

---
v1 -> v2:

Removed check for poisioned list pointer. This is abusing knowledge internal to the list api, so we need another way to know if the child irq is enabled. This patch uses a driver controlled boolean for this.

Added tags for the bug reporter and tester.

 arch/x86/pci/vmd.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/pci/vmd.c +8 -2
index e88b417..4b85837 100644
@@ -41,6 +41,7 @@ static DEFINE_RAW_SPINLOCK(list_lock);
  * @node:	list item for parent traversal.
  * @rcu:	RCU callback item for freeing.
  * @irq:	back pointer to parent.
+ * @enabled:	true if driver enabled irq
  * @virq:	the virtual IRQ value provided to the requesting driver.
  *
  * Every MSI/MSI-X IRQ requested for a device in a VMD domain will be mapped to
@@ -50,6 +51,7 @@ struct vmd_irq {
 	struct list_head	node;
 	struct rcu_head		rcu;
 	struct vmd_irq_list	*irq;
+	bool			enabled;
 	unsigned int		virq;
 };
 
@@ -122,7 +124,9 @@ static void vmd_irq_enable(struct irq_data *data)
 	unsigned long flags;
 
 	raw_spin_lock_irqsave(&list_lock, flags);
+	WARN_ON(vmdirq->enabled);
 	list_add_tail_rcu(&vmdirq->node, &vmdirq->irq->irq_list);
+	vmdirq->enabled = true;
 	raw_spin_unlock_irqrestore(&list_lock, flags);
 
 	data->chip->irq_unmask(data);
@@ -136,8 +140,10 @@ static void vmd_irq_disable(struct irq_data *data)
 	data->chip->irq_mask(data);
 
 	raw_spin_lock_irqsave(&list_lock, flags);
-	list_del_rcu(&vmdirq->node);
-	INIT_LIST_HEAD_RCU(&vmdirq->node);
+	if (vmdirq->enabled) {
+		list_del_rcu(&vmdirq->node);
+		vmdirq->enabled = false;
+	}
 	raw_spin_unlock_irqrestore(&list_lock, flags);
 }
 
Signature · 1 line
2.7.2
#1 lore raw

Seems fine and didn't grow the struct, which is always nice

Acked-by Jon Derrick: <[email protected]>

Quoted 74 lines · Keith Busch

We can't initialize the list head on deletion as this causes the node to point to itself, looping infinitely if the vmd IRQ handler happened to be servicing that node.

The list initialization was trying fix a bug from multiple calls to disable the same IRQ. We can fix this instead by having the vmd driver track if the interrupt is enabled.

---
v1 -> v2:

Removed check for poisioned list pointer. This is abusing knowledge internal to the list api, so we need another way to know if the child irq is enabled. This patch uses a driver controlled boolean for this.

Added tags for the bug reporter and tester.

 arch/x86/pci/vmd.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/pci/vmd.c +8 -2
index e88b417..4b85837 100644
@@ -41,6 +41,7 @@ static DEFINE_RAW_SPINLOCK(list_lock);
  * @node:	list item for parent traversal.
  * @rcu:	RCU callback item for freeing.
  * @irq:	back pointer to parent.
+ * @enabled:	true if driver enabled irq
  * @virq:	the virtual IRQ value provided to the requesting driver.
  *
  * Every MSI/MSI-X IRQ requested for a device in a VMD domain will be mapped to
@@ -50,6 +51,7 @@ struct vmd_irq {
 	struct list_head	node;
 	struct rcu_head		rcu;
 	struct vmd_irq_list	*irq;
+	bool			enabled;
 	unsigned int		virq;
 };
 
@@ -122,7 +124,9 @@ static void vmd_irq_enable(struct irq_data *data)
 	unsigned long flags;
 
 	raw_spin_lock_irqsave(&list_lock, flags);
+	WARN_ON(vmdirq->enabled);
 	list_add_tail_rcu(&vmdirq->node, &vmdirq->irq->irq_list);
+	vmdirq->enabled = true;
 	raw_spin_unlock_irqrestore(&list_lock, flags);
 
 	data->chip->irq_unmask(data);
@@ -136,8 +140,10 @@ static void vmd_irq_disable(struct irq_data *data)
 	data->chip->irq_mask(data);
 
 	raw_spin_lock_irqsave(&list_lock, flags);
-	list_del_rcu(&vmdirq->node);
-	INIT_LIST_HEAD_RCU(&vmdirq->node);
+	if (vmdirq->enabled) {
+		list_del_rcu(&vmdirq->node);
+		vmdirq->enabled = false;
+	}
 	raw_spin_unlock_irqrestore(&list_lock, flags);
 }
 
Signature · 6 lines
2.7.2

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#2 lore raw
In reply to Keith Busch

Hi Keith,

Quoted 4 lines · Keith Busch

We can't initialize the list head on deletion as this causes the node to point to itself, looping infinitely if the vmd IRQ handler happened to be servicing that node.

Quoted 8 lines

The list initialization was trying fix a bug from multiple calls to disable the same IRQ. We can fix this instead by having the vmd driver track if the interrupt is enabled.

I assume you want

Fixes 97e923063575 ("x86/PCI: VMD: Initialize list item in IRQ disable")

and you want this in for-linus so it goes in v4.8 instead of v4.9.

I did that for you, so let me know if it's not right.

Quoted 62 lines

---
v1 -> v2:

Removed check for poisioned list pointer. This is abusing knowledge internal to the list api, so we need another way to know if the child irq is enabled. This patch uses a driver controlled boolean for this.

Added tags for the bug reporter and tester.

 arch/x86/pci/vmd.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
arch/x86/pci/vmd.c +8 -2
index e88b417..4b85837 100644
@@ -41,6 +41,7 @@ static DEFINE_RAW_SPINLOCK(list_lock);
  * @node:	list item for parent traversal.
  * @rcu:	RCU callback item for freeing.
  * @irq:	back pointer to parent.
+ * @enabled:	true if driver enabled irq
  * @virq:	the virtual IRQ value provided to the requesting driver.
  *
  * Every MSI/MSI-X IRQ requested for a device in a VMD domain will be mapped to
@@ -50,6 +51,7 @@ struct vmd_irq {
 	struct list_head	node;
 	struct rcu_head		rcu;
 	struct vmd_irq_list	*irq;
+	bool			enabled;
 	unsigned int		virq;
 };
 
@@ -122,7 +124,9 @@ static void vmd_irq_enable(struct irq_data *data)
 	unsigned long flags;
 
 	raw_spin_lock_irqsave(&list_lock, flags);
+	WARN_ON(vmdirq->enabled);
 	list_add_tail_rcu(&vmdirq->node, &vmdirq->irq->irq_list);
+	vmdirq->enabled = true;
 	raw_spin_unlock_irqrestore(&list_lock, flags);
 
 	data->chip->irq_unmask(data);
@@ -136,8 +140,10 @@ static void vmd_irq_disable(struct irq_data *data)
 	data->chip->irq_mask(data);
 
 	raw_spin_lock_irqsave(&list_lock, flags);
-	list_del_rcu(&vmdirq->node);
-	INIT_LIST_HEAD_RCU(&vmdirq->node);
+	if (vmdirq->enabled) {
+		list_del_rcu(&vmdirq->node);
+		vmdirq->enabled = false;
+	}
 	raw_spin_unlock_irqrestore(&list_lock, flags);
 }
 
Signature · 6 lines
2.7.2

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#3 lore raw
Quoted 14 lines · Bjorn Helgaas
Quoted 4 lines · Keith Busch

We can't initialize the list head on deletion as this causes the node to point to itself, looping infinitely if the vmd IRQ handler happened to be servicing that node.

Quoted 8 lines

The list initialization was trying fix a bug from multiple calls to disable the same IRQ. We can fix this instead by having the vmd driver track if the interrupt is enabled.

Quoted 7 lines

I assume you want

Fixes 97e923063575 ("x86/PCI: VMD: Initialize list item in IRQ disable")

and you want this in for-linus so it goes in v4.8 instead of v4.9.

I did that for you, so let me know if it's not right.

Yes, that was our intention. Thank you for setting the "Fixes" and staging for 4.8.