Skip to contentSkip to conversation

mm/slab.c: check pointer slabp before using it in alloc_slabmgmt()

PATCH In review ethan.zhao 4 participants 8 messages Dec 8, 2013 – Dec 14, 2013

Subsystems (2)
lore
Share & tools
lore

ethan.zhao

author
#0 lore raw

PATCH 1/1 mm/slab.c: check pointer slabp before using it in alloc_slabmgmt()

Move the NULL check of slabp to the right place before refer its memeber in function alloc_slabmgmt().

This bug may be introduced by rewriting of funcion kmemleak_scan_area(), the first parameter changed from slabp to &slabp->list.

---
 mm/slab.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
mm/slab.c +2 -2
index 2580db0..b6d27bc 100644
@@ -2612,6 +2612,8 @@ static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
 		/* Slab management obj is off-slab. */
 		slabp = kmem_cache_alloc_node(cachep->slabp_cache,
 					      local_flags, nodeid);
+		if (!slabp)
+			return NULL;
 		/*
 		 * If the first object in the slab is leaked (it's allocated
 		 * but no one has a reference to it), we want to make sure
@@ -2620,8 +2622,6 @@ static struct slab *alloc_slabmgmt(struct kmem_cache *cachep, void *objp,
 		 */
 		kmemleak_scan_area(&slabp->list, sizeof(struct list_head),
 				   local_flags);
-		if (!slabp)
-			return NULL;
 	} else {
 		slabp = objp + colour_off;
 		colour_off += cachep->slab_size;
Signature · 1 line
1.8.3.4 (Apple Git-47)
#1 lore raw
Quoted 2 lines · ethan.zhao

Move the NULL check of slabp to the right place before refer its memeber in function alloc_slabmgmt().

I am having trouble to find the code you are modifying. Which kernel release is this? The code you are referring to has been rewritten in the meantime or this is some other code basis.

author
#2 lore raw

Christoph,
Found in the latest stable release V3.12.3, yes, changed in
3.12.4. not needed for later release anymore.

Thanks,
Ethan

Quoted 9 lines · Christoph Lameter
Quoted 2 lines · ethan.zhao

Move the NULL check of slabp to the right place before refer its memeber in function alloc_slabmgmt().

I am having trouble to find the code you are modifying. Which kernel release is this? The code you are referring to has been rewritten in the meantime or this is some other code basis.

author
#3 lore raw

Christoph,

     No, stable 3.12.4 and 3.12.3 need this patch.  3.13RC doesn't

need it anymore.

Thanks,
Ethan

Quoted 18 lines · Ethan Zhao

Christoph,
Found in the latest stable release V3.12.3, yes, changed in
3.12.4. not needed for later release anymore.

Thanks,
Ethan

Quoted 9 lines · Christoph Lameter
Quoted 2 lines · ethan.zhao

Move the NULL check of slabp to the right place before refer its memeber in function alloc_slabmgmt().

I am having trouble to find the code you are modifying. Which kernel release is this? The code you are referring to has been rewritten in the meantime or this is some other code basis.

#4 lore raw
Quoted 2 lines · Ethan Zhao
     No, stable 3.12.4 and 3.12.3 need this patch.  3.13RC doesn't

need it anymore.

Hmmm.. Then this commit may have fixed it:

commit 0172f779e4314639a8ed440082cfe9e3450954e8

Author: Joonsoo Kim <[email protected]>
Date:   Wed Oct 30 19:04:00 2013 +0900

slab: fix to calm down kmemleak warning

After using struct page as slab management, we should not call kmemleak_scan_area(), since struct page isn't the tracking object of kmemleak. Without this patch and if CONFIG_DEBUG_KMEMLEAK is enabled, so many kmemleak warnings are printed.

mm/slab.c +0 -0
index af2db76..a8a9349 100644
@@ -2531,14 +2531,6 @@ static struct freelist *alloc_slabmgmt(struct

kmem_cache *cachep,

                /* Slab management obj is off-slab. */
                freelist = kmem_cache_alloc_node(cachep->freelist_cache,
                                              local_flags, nodeid);
-               /*
-                * If the first object in the slab is leaked (it's allocated
-                * but no one has a reference to it), we want to make sure
-                * kmemleak does not treat the ->s_mem pointer as a reference
-                * to the object. Otherwise we will not report the leak.
-                */
-               kmemleak_scan_area(&page->lru, sizeof(struct list_head),
-                                  local_flags);
                if (!freelist)
                        return NULL;
        } else {
#5 lore raw
Quoted 6 lines · Christoph Lameter
Quoted 2 lines · Ethan Zhao
      No, stable 3.12.4 and 3.12.3 need this patch.  3.13RC doesn't

need it anymore.

Hmmm.. Then this commit may have fixed it:

Cross-threaded from [Re: Slab BUG with DEBUG_* options ]

Quoted 4 lines · Meelis Roos

ldata alloc change + your second slab patch + slab debug: hang on boot after
console [tty0] enabled, bootconsole disabled
(after that, I should see all the dmesg again on serial but I do not).

Meelis,

Please grab this patch below and see if it fixes your 'hang on boot' with Christoph's other slab patch + slab debug.

Regards,
Peter Hurley

Quoted 35 lines

commit 0172f779e4314639a8ed440082cfe9e3450954e8

Author: Joonsoo Kim <[email protected]>
Date:   Wed Oct 30 19:04:00 2013 +0900

slab: fix to calm down kmemleak warning

After using struct page as slab management, we should not call kmemleak_scan_area(), since struct page isn't the tracking object of kmemleak. Without this patch and if CONFIG_DEBUG_KMEMLEAK is enabled, so many kmemleak warnings are printed.

mm/slab.c +0 -8
index af2db76..a8a9349 100644
@@ -2531,14 +2531,6 @@ static struct freelist *alloc_slabmgmt(struct
 kmem_cache *cachep,
                 /* Slab management obj is off-slab. */
                 freelist = kmem_cache_alloc_node(cachep->freelist_cache,
                                               local_flags, nodeid);
-               /*
-                * If the first object in the slab is leaked (it's allocated
-                * but no one has a reference to it), we want to make sure
-                * kmemleak does not treat the ->s_mem pointer as a reference
-                * to the object. Otherwise we will not report the leak.
-                */
-               kmemleak_scan_area(&page->lru, sizeof(struct list_head),
-                                  local_flags);
                 if (!freelist)
                         return NULL;
         } else {
#6 lore raw
Quoted 15 lines
Quoted 6 lines · Ethan Zhao
Quoted 3 lines
      No, stable 3.12.4 and 3.12.3 need this patch.  3.13RC doesn't

need it anymore.

Hmmm.. Then this commit may have fixed it:

Cross-threaded from [Re: Slab BUG with DEBUG_* options ]

Quoted 4 lines · Meelis Roos

ldata alloc change + your second slab patch + slab debug: hang on boot after
console [tty0] enabled, bootconsole disabled
(after that, I should see all the dmesg again on serial but I do not).

Quoted 4 lines

Meelis,

Please grab this patch below and see if it fixes your 'hang on boot' with Christoph's other slab patch + slab debug.

Tried it on top of 3.12 + second slab patch + ldata alloc change, and lsab debug conf. No change - still hangs on boot.

Signature · 1 line
Meelis Roos ([email protected])