raid5: delete another BUG_ON

PATCH In review Shaohua Li 2 participants 3 messages Apr 14, 2014 – Apr 15, 2014 MD RAID

Share & tools

Shaohua Li

author

2014-04-14 06:14 UTC #0 lore raw

PATCH 1/1 raid5: delete another BUG_ON

I hit another BUG_ON with e240c1839d11152b0355442. In __get_priority_stripe(), stripe count equals to 0 initially. Between atomic_inc and BUG_ON, get_active_stripe() finds the stripe. So the stripe count isn't 1 any more.

Signed-off-by Shaohua Li <[email protected]>

---
 drivers/md/raid5.c |    1 -
 1 file changed, 1 deletion(-)

drivers/md/raid5.c +0 -1

Index: linux/drivers/md/raid5.c
===================================================================
@@ -4371,7 +4371,6 @@ static struct stripe_head *__get_priorit
 	}
 	list_del_init(&sh->lru);
 	atomic_inc(&sh->count);
-	BUG_ON(atomic_read(&sh->count) != 1);
 	return sh;
 }

NeilBrown

2014-04-14 22:47 UTC #1 lore raw

Quoted 15 lines · On Mon, 14 Apr 2014 14:14:17 +0800 Shaohua Li <[email protected]> wrote:

I hit another BUG_ON with e240c1839d11152b0355442. In __get_priority_stripe(), stripe count equals to 0 initially. Between atomic_inc and BUG_ON, get_active_stripe() finds the stripe. So the stripe count isn't 1 any more.
Signed-off-by Shaohua Li <[email protected]>
---
 drivers/md/raid5.c |    1 -
 1 file changed, 1 deletion(-)
Index: linux/drivers/md/raid5.c
===================================================================

drivers/md/raid5.c static struct stripe_head *__get_priorit · line 4376

Quoted 7 lines

drivers/md/raid5.c +0 -1

@@ -4371,7 +4371,6 @@ static struct stripe_head *__get_priorit
 	}
 	list_del_init(&sh->lru);
 	atomic_inc(&sh->count);
-	BUG_ON(atomic_read(&sh->count) != 1);
 	return sh;
 }

What if we made those two lines:

  BUG_ON(atomic_inc_return(&sh->count) != 1);

NeilBrown

Shaohua Li

author

2014-04-15 01:12 UTC #2 lore raw

Quoted 27 lines · NeilBrown

Quoted 15 lines · On Mon, 14 Apr 2014 14:14:17 +0800 Shaohua Li <[email protected]> wrote:
I hit another BUG_ON with e240c1839d11152b0355442. In __get_priority_stripe(), stripe count equals to 0 initially. Between atomic_inc and BUG_ON, get_active_stripe() finds the stripe. So the stripe count isn't 1 any more.
Signed-off-by Shaohua Li <[email protected]>
---
 drivers/md/raid5.c |    1 -
 1 file changed, 1 deletion(-)
Index: linux/drivers/md/raid5.c
===================================================================
drivers/md/raid5.c static struct stripe_head *__get_priorit · line 4376
Quoted 7 lines
drivers/md/raid5.c +0 -1
@@ -4371,7 +4371,6 @@ static struct stripe_head *__get_priorit
 	}
 	list_del_init(&sh->lru);
 	atomic_inc(&sh->count);
-	BUG_ON(atomic_read(&sh->count) != 1);
 	return sh;
 }
 
What if we made those two lines:

Quoted 1 line

  BUG_ON(atomic_inc_return(&sh->count) != 1);

Ok if we want to keep the BUG_ON.

Subject: raid5: fix a race of stripe count check

V2: keeps the BUG_ON suggested by Neil.

Signed-off-by Shaohua Li <[email protected]>

---
 drivers/md/raid5.c |    3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

drivers/md/raid5.c +1 -2

Index: linux/drivers/md/raid5.c
===================================================================
@@ -4370,8 +4370,7 @@ static struct stripe_head *__get_priorit
 		sh->group = NULL;
 	}
 	list_del_init(&sh->lru);
-	atomic_inc(&sh->count);
-	BUG_ON(atomic_read(&sh->count) != 1);
+	BUG_ON(atomic_inc_return(&sh->count) != 1);
 	return sh;
 }