Skip to contentSkip to conversation

:Fixed an issue with memory allocated using the GFP_KERNEL flag in spinlocks

PATCH In review xiaofeng.yan 3 participants 6 messages Nov 4, 2020

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 :Fixed an issue with memory allocated using the GFP_KERNEL flag in spinlocks

From: "xiaofeng.yan" <[email protected]>

function xa_store_irq() has a spinlock as follows:
xa_lock_irq()
-->spin_lock_irq(&(xa)->xa_lock)
GFP_KERNEL flag could cause sleep.
So change GFP_KERNEL to GFP_ATOMIC and Romve "gfp_t gfp" in function static inline void *xa_store_irq(struct xarray *xa, unsigned long index,
void *entry, gfp_t gfp)

---
 include/linux/xarray.h | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)
include/linux/xarray.h +2 -3
index 92c0160b3352..aeaf97d5642f 100644
@@ -595,7 +595,6 @@ static inline void *xa_store_bh(struct xarray *xa, unsigned long index,
  * @xa: XArray.
  * @index: Index into array.
  * @entry: New entry.
- * @gfp: Memory allocation flags.
  *
  * This function is like calling xa_store() except it disables interrupts
  * while holding the array lock.
@@ -605,12 +604,12 @@ static inline void *xa_store_bh(struct xarray *xa, unsigned long index,
  * Return: The old entry at this index or xa_err() if an error happened.
  */
 static inline void *xa_store_irq(struct xarray *xa, unsigned long index,
-		void *entry, gfp_t gfp)
+		void *entry)
 {
 	void *curr;
 
 	xa_lock_irq(xa);
-	curr = __xa_store(xa, index, entry, gfp);
+	curr = __xa_store(xa, index, entry, GFP_ATOMIC);
 	xa_unlock_irq(xa);
 
 	return curr;
Signature · 1 line
2.17.1
#1 lore raw

PATCH 2/2 infiniband: Modify the reference to xa_store_irq() because the parameter of this function has changed

From: "xiaofeng.yan" <[email protected]>

function xa_store_irq() has three parameters because of removing patameter "gfp_t gfp"

---
 drivers/infiniband/core/cm.c            | 2 +-
 drivers/infiniband/hw/hns/hns_roce_qp.c | 2 +-
 drivers/infiniband/hw/mlx5/srq_cmd.c    | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)
drivers/infiniband/core/cm.c +1 -1
index 5740d1ba3568..afcb5711270b 100644
@@ -879,7 +879,7 @@ static struct cm_id_private *cm_alloc_id_priv(struct ib_device *device,
 static void cm_finalize_id(struct cm_id_private *cm_id_priv)
 {
 	xa_store_irq(&cm.local_id_table, cm_local_id(cm_id_priv->id.local_id),
-		     cm_id_priv, GFP_KERNEL);
+		     cm_id_priv);
 }
 
 struct ib_cm_id *ib_create_cm_id(struct ib_device *device,
drivers/infiniband/hw/hns/hns_roce_qp.c +1 -1
index 6c081dd985fc..1876a51f9e08 100644
@@ -237,7 +237,7 @@ static int hns_roce_qp_store(struct hns_roce_dev *hr_dev,
 	if (!hr_qp->qpn)
 		return -EINVAL;
 
-	ret = xa_err(xa_store_irq(xa, hr_qp->qpn, hr_qp, GFP_KERNEL));
+	ret = xa_err(xa_store_irq(xa, hr_qp->qpn, hr_qp));
 	if (ret)
 		dev_err(hr_dev->dev, "Failed to xa store for QPC\n");
 	else
drivers/infiniband/hw/mlx5/srq_cmd.c +1 -1
index db889ec3fd48..f277e264ceab 100644
@@ -578,7 +578,7 @@ int mlx5_cmd_create_srq(struct mlx5_ib_dev *dev, struct mlx5_core_srq *srq,
 	refcount_set(&srq->common.refcount, 1);
 	init_completion(&srq->common.free);
 
-	err = xa_err(xa_store_irq(&table->array, srq->srqn, srq, GFP_KERNEL));
+	err = xa_err(xa_store_irq(&table->array, srq->srqn, srq));
 	if (err)
 		goto err_destroy_srq_split;
 
Signature · 1 line
2.17.1
maintainer
#5 lore raw
In reply to xiaofeng.yan

include/linux/xarray.h static inline void *xa_store_bh(struct xarray *xa, unsigned long index, · line 613

Quoted 4 lines · xiaofeng.yan
include/linux/xarray.h +1 -1
 	xa_lock_irq(xa);
-	curr = __xa_store(xa, index, entry, gfp);
+	curr = __xa_store(xa, index, entry, GFP_ATOMIC);
 	xa_unlock_irq(xa);

You haven't actually seen a bug, have you? You just read the code wrongly.

void *__xa_store(struct xarray *xa, unsigned long index, void *entry, gfp_t gfp)
...

        } while (__xas_nomem(&xas, gfp));
...
        if (gfpflags_allow_blocking(gfp)) {
                xas_unlock_type(xas, lock_type);
                xas->xa_alloc = kmem_cache_alloc(radix_tree_node_cachep, gfp);
                xas_lock_type(xas, lock_type);
#2 lore raw

Re: [PATCH 2/2] infiniband: Modify the reference to xa_store_irq() because the parameter of this function has changed

In reply to xiaofeng.yan
Quoted 15 lines · xiaofeng.yan

From: "xiaofeng.yan" <[email protected]>

function xa_store_irq() has three parameters because of removing patameter "gfp_t gfp"

 drivers/infiniband/core/cm.c            | 2 +-
 drivers/infiniband/hw/hns/hns_roce_qp.c | 2 +-
 drivers/infiniband/hw/mlx5/srq_cmd.c    | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)
drivers/infiniband/core/cm.c +1 -0
index 5740d1ba3568..afcb5711270b 100644
+++ b/drivers/infiniband/core/cm.c

drivers/infiniband/core/cm.c static struct cm_id_private *cm_alloc_id_priv(struct ib_device *device, · line 883

Quoted 6 lines
drivers/infiniband/core/cm.c +1 -1
@@ -879,7 +879,7 @@ static struct cm_id_private *cm_alloc_id_priv(struct ib_device *device,
 static void cm_finalize_id(struct cm_id_private *cm_id_priv)
 {
 	xa_store_irq(&cm.local_id_table, cm_local_id(cm_id_priv->id.local_id),
-		     cm_id_priv, GFP_KERNEL);
+		     cm_id_priv);
 }

This one is almost a bug, the entry is preallocated with NULL though:

	ret = xa_alloc_cyclic_irq(&cm.local_id_table, &id, NULL, xa_limit_32b,
				  &cm.local_id_next, GFP_KERNEL);

so it should never allocate here:

static int cm_req_handler(struct cm_work *work)

{
	spin_lock_irq(&cm_id_priv->lock);
	cm_finalize_id(cm_id_priv);

Still, woops.

Matt, maybe a might_sleep is deserved in here someplace?

@@ -1534,6 +1534,8 @@ void *__xa_store(struct xarray *xa, unsigned long index, void *entry, gfp_t gfp)

        XA_STATE(xas, xa, index);
        void *curr;

+ might_sleep_if(gfpflags_allow_blocking(gfp));
+

        if (WARN_ON_ONCE(xa_is_advanced(entry)))
                return XA_ERROR(-EINVAL);
        if (xa_track_free(xa) && !entry)

And similar in the other places that conditionally call __xas_nomem() ?

I also still wish there was a proper 'xa store in already allocated but null' idiom - I remember you thought about using gfp flags == 0 at one point.

Jason

maintainer
#3 lore raw

Re: [PATCH 2/2] infiniband: Modify the reference to xa_store_irq() because the parameter of this function has changed

Quoted 12 lines · Jason Gunthorpe

drivers/infiniband/core/cm.c static struct cm_id_private *cm_alloc_id_priv(struct ib_device *device, · line 883

Quoted 6 lines

static void cm_finalize_id(struct cm_id_private *cm_id_priv) {

drivers/infiniband/core/cm.c +1 -1
 	xa_store_irq(&cm.local_id_table, cm_local_id(cm_id_priv->id.local_id),
-		     cm_id_priv, GFP_KERNEL);
+		     cm_id_priv);
 }

This one is almost a bug, the entry is preallocated with NULL though:

	ret = xa_alloc_cyclic_irq(&cm.local_id_table, &id, NULL, xa_limit_32b,
				  &cm.local_id_next, GFP_KERNEL);
Quoted 6 lines

so it should never allocate here:

static int cm_req_handler(struct cm_work *work)

{
	spin_lock_irq(&cm_id_priv->lock);
	cm_finalize_id(cm_id_priv);

Uhm. I think you want a different debugging check from this. The actual bug here is that you'll get back from calling cm_finalize_id() with interrupts enabled. Can you switch to xa_store(), or do we need an xa_store_irqsave()?

Quoted 14 lines

Still, woops.

Matt, maybe a might_sleep is deserved in here someplace?

(quoted diff) +2 -0
@@ -1534,6 +1534,8 @@ void *__xa_store(struct xarray *xa, unsigned long index, void *entry, gfp_t gfp)
        XA_STATE(xas, xa, index);
        void *curr;
 
+       might_sleep_if(gfpflags_allow_blocking(gfp));
+
        if (WARN_ON_ONCE(xa_is_advanced(entry)))
                return XA_ERROR(-EINVAL);
        if (xa_track_free(xa) && !entry)
Quoted 6 lines

And similar in the other places that conditionally call __xas_nomem() ?

I also still wish there was a proper 'xa store in already allocated but null' idiom - I remember you thought about using gfp flags == 0 at one point.

An xa_replace(), perhaps?

#4 lore raw

Re: [PATCH 2/2] infiniband: Modify the reference to xa_store_irq() because the parameter of this function has changed

Quoted 15 lines · Matthew Wilcox
Quoted 14 lines · Jason Gunthorpe

drivers/infiniband/core/cm.c static struct cm_id_private *cm_alloc_id_priv(struct ib_device *device, · line 883

Quoted 6 lines

static void cm_finalize_id(struct cm_id_private *cm_id_priv) {

drivers/infiniband/core/cm.c +1 -1
 	xa_store_irq(&cm.local_id_table, cm_local_id(cm_id_priv->id.local_id),
-		     cm_id_priv, GFP_KERNEL);
+		     cm_id_priv);
 }

This one is almost a bug, the entry is preallocated with NULL though:

	ret = xa_alloc_cyclic_irq(&cm.local_id_table, &id, NULL, xa_limit_32b,
				  &cm.local_id_next, GFP_KERNEL);

so it should never allocate here:

Quoted 8 lines
Quoted 4 lines

static int cm_req_handler(struct cm_work *work)

{
	spin_lock_irq(&cm_id_priv->lock);
	cm_finalize_id(cm_id_priv);

Uhm. I think you want a different debugging check from this. The actual bug here is that you'll get back from calling cm_finalize_id() with interrupts enabled.

Ooh, that is just no fun too :\

Again surprised some lockdep didn't catch wrongly nesting irq locks

Quoted 2 lines

Can you switch to xa_store(), or do we need an
xa_store_irqsave()?

Yes, it looks like there is no reason for this, all users of the xarray are from sleeping contexts, so it shouldn't need the IRQ version.. I made a patch for this thanks

The cm_id_priv->lock is probably also not needing to be irq either, but that is much harder to tell for sure

Quoted 8 lines
Quoted 8 lines

Still, woops.

Matt, maybe a might_sleep is deserved in here someplace?

(quoted diff) +0 -0
@@ -1534,6 +1534,8 @@ void *__xa_store(struct xarray *xa, unsigned long index, void *entry, gfp_t gfp)
        XA_STATE(xas, xa, index);
        void *curr;
 
Quoted 8 lines
Quoted 8 lines
(quoted diff) +2 -0
+       might_sleep_if(gfpflags_allow_blocking(gfp));
+
        if (WARN_ON_ONCE(xa_is_advanced(entry)))
                return XA_ERROR(-EINVAL);
        if (xa_track_free(xa) && !entry)

And similar in the other places that conditionally call __xas_nomem() ?

But this debugging would still catch the wrong nesting of a GFP_KERNEL inside a spinlock, you don't like it?

Quoted 5 lines
Quoted 3 lines

I also still wish there was a proper 'xa store in already allocated but null' idiom - I remember you thought about using gfp flags == 0 at one point.

An xa_replace(), perhaps?

Make sense.. But I've also done this with cmpxchg. A magic GFP flag, as you tried to do with 0, is appealing in many ways

Jason