diff mbox

mm: mempool: Fix a possible sleep-in-atomic-context bug in mempool_resize()

Message ID 20180621030714.10368-1-baijiaju1990@gmail.com (mailing list archive)
State New, archived
Headers show

Commit Message

Jia-Ju Bai June 21, 2018, 3:07 a.m. UTC
The kernel may sleep with holding a spinlock.
The function call path (from bottom to top) in Linux-4.16.7 is:

[FUNC] remove_element(GFP_KERNEL)
mm/mempool.c, 250: remove_element in mempool_resize
mm/mempool.c, 247: _raw_spin_lock_irqsave in mempool_resize

To fix this bug, GFP_KERNEL is replaced with GFP_ATOMIC.

This bug is found by my static analysis tool (DSAC-2) and checked by
my code review.

Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
---
 mm/mempool.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Matthew Wilcox (Oracle) June 21, 2018, 3:38 a.m. UTC | #1
On Thu, Jun 21, 2018 at 11:07:14AM +0800, Jia-Ju Bai wrote:
> The kernel may sleep with holding a spinlock.
> The function call path (from bottom to top) in Linux-4.16.7 is:
> 
> [FUNC] remove_element(GFP_KERNEL)
> mm/mempool.c, 250: remove_element in mempool_resize
> mm/mempool.c, 247: _raw_spin_lock_irqsave in mempool_resize
> 
> To fix this bug, GFP_KERNEL is replaced with GFP_ATOMIC.
> 
> This bug is found by my static analysis tool (DSAC-2) and checked by
> my code review.

But ... we don't use the flags argument.

static void *remove_element(mempool_t *pool, gfp_t flags)
{
        void *element = pool->elements[--pool->curr_nr];

        BUG_ON(pool->curr_nr < 0);
        kasan_unpoison_element(pool, element, flags);
        check_element(pool, element);
        return element;
}

...

static void kasan_unpoison_element(mempool_t *pool, void *element, gfp_t flags)
{
        if (pool->alloc == mempool_alloc_slab || pool->alloc == mempool_kmalloc)
                kasan_unpoison_slab(element);
        if (pool->alloc == mempool_alloc_pages)
                kasan_alloc_pages(element, (unsigned long)pool->pool_data);
}

So the correct patch would just remove this argument to remove_element() and
kasan_unpoison_element()?
Jia-Ju Bai June 21, 2018, 3:46 a.m. UTC | #2
On 2018/6/21 11:38, Matthew Wilcox wrote:
> On Thu, Jun 21, 2018 at 11:07:14AM +0800, Jia-Ju Bai wrote:
>> The kernel may sleep with holding a spinlock.
>> The function call path (from bottom to top) in Linux-4.16.7 is:
>>
>> [FUNC] remove_element(GFP_KERNEL)
>> mm/mempool.c, 250: remove_element in mempool_resize
>> mm/mempool.c, 247: _raw_spin_lock_irqsave in mempool_resize
>>
>> To fix this bug, GFP_KERNEL is replaced with GFP_ATOMIC.
>>
>> This bug is found by my static analysis tool (DSAC-2) and checked by
>> my code review.
> But ... we don't use the flags argument.
>
> static void *remove_element(mempool_t *pool, gfp_t flags)
> {
>          void *element = pool->elements[--pool->curr_nr];
>
>          BUG_ON(pool->curr_nr < 0);
>          kasan_unpoison_element(pool, element, flags);
>          check_element(pool, element);
>          return element;
> }
>
> ...
>
> static void kasan_unpoison_element(mempool_t *pool, void *element, gfp_t flags)
> {
>          if (pool->alloc == mempool_alloc_slab || pool->alloc == mempool_kmalloc)
>                  kasan_unpoison_slab(element);
>          if (pool->alloc == mempool_alloc_pages)
>                  kasan_alloc_pages(element, (unsigned long)pool->pool_data);
> }
>
> So the correct patch would just remove this argument to remove_element() and
> kasan_unpoison_element()?

Yes, I also find this.
I can submit a patch that removes the flag in:
Definitions of kasan_unpoison_element() and remove_element()
Three calls to remove_element() and one call to kasan_unpoison_element() 
in mempool.c.

Do you think it is okay?


Best wishes,
Jia-Ju Bai
Dmitry Vyukov June 21, 2018, 5:54 a.m. UTC | #3
On Thu, Jun 21, 2018 at 5:46 AM, Jia-Ju Bai <baijiaju1990@gmail.com> wrote:
> On 2018/6/21 11:38, Matthew Wilcox wrote:
>>
>> On Thu, Jun 21, 2018 at 11:07:14AM +0800, Jia-Ju Bai wrote:
>>>
>>> The kernel may sleep with holding a spinlock.
>>> The function call path (from bottom to top) in Linux-4.16.7 is:
>>>
>>> [FUNC] remove_element(GFP_KERNEL)
>>> mm/mempool.c, 250: remove_element in mempool_resize
>>> mm/mempool.c, 247: _raw_spin_lock_irqsave in mempool_resize
>>>
>>> To fix this bug, GFP_KERNEL is replaced with GFP_ATOMIC.
>>>
>>> This bug is found by my static analysis tool (DSAC-2) and checked by
>>> my code review.
>>
>> But ... we don't use the flags argument.
>>
>> static void *remove_element(mempool_t *pool, gfp_t flags)
>> {
>>          void *element = pool->elements[--pool->curr_nr];
>>
>>          BUG_ON(pool->curr_nr < 0);
>>          kasan_unpoison_element(pool, element, flags);
>>          check_element(pool, element);
>>          return element;
>> }
>>
>> ...
>>
>> static void kasan_unpoison_element(mempool_t *pool, void *element, gfp_t
>> flags)
>> {
>>          if (pool->alloc == mempool_alloc_slab || pool->alloc ==
>> mempool_kmalloc)
>>                  kasan_unpoison_slab(element);
>>          if (pool->alloc == mempool_alloc_pages)
>>                  kasan_alloc_pages(element, (unsigned
>> long)pool->pool_data);
>> }
>>
>> So the correct patch would just remove this argument to remove_element()
>> and
>> kasan_unpoison_element()?
>
>
> Yes, I also find this.
> I can submit a patch that removes the flag in:
> Definitions of kasan_unpoison_element() and remove_element()
> Three calls to remove_element() and one call to kasan_unpoison_element() in
> mempool.c.
>
> Do you think it is okay?

Hi Jia-Ju,

Removing an unused argument within a single file looks good to me.
diff mbox

Patch

diff --git a/mm/mempool.c b/mm/mempool.c
index 5c9dce34719b..d33bd5d622e7 100644
--- a/mm/mempool.c
+++ b/mm/mempool.c
@@ -247,7 +247,7 @@  int mempool_resize(mempool_t *pool, int new_min_nr)
 	spin_lock_irqsave(&pool->lock, flags);
 	if (new_min_nr <= pool->min_nr) {
 		while (new_min_nr < pool->curr_nr) {
-			element = remove_element(pool, GFP_KERNEL);
+			element = remove_element(pool, GFP_ATOMIC);
 			spin_unlock_irqrestore(&pool->lock, flags);
 			pool->free(element, pool->pool_data);
 			spin_lock_irqsave(&pool->lock, flags);