* [PATCH] staging: zcache: fix cleancache race condition with shrinker
@ 2012-08-29 21:58 Seth Jennings
2012-08-30 16:11 ` Konrad Rzeszutek Wilk
2012-08-30 17:55 ` Minchan Kim
0 siblings, 2 replies; 3+ messages in thread
From: Seth Jennings @ 2012-08-29 21:58 UTC (permalink / raw)
To: Greg Kroah-Hartman
Cc: Seth Jennings, Andrew Morton, Nitin Gupta, Minchan Kim,
Konrad Rzeszutek Wilk, Dan Magenheimer, Robert Jennings,
Xiao Guangrong, linux-mm, linux-kernel, devel
This patch fixes a race condition that results in memory
corruption when using cleancache.
The race exists between the zcache shrinker handler,
shrink_zcache_memory() and cleancache_get_page().
In most cases, the shrinker will both evict a zbpg
from its buddy list and flush it from tmem before a
cleancache_get_page() occurs on that page. A subsequent
cleancache_get_page() will fail in the tmem layer.
In the rare case that two occur together and the
cleancache_get_page() path gets through the tmem
layer before the shrinker path can flush tmem,
zbud_decompress() does a check to see if the zbpg is a
"zombie", i.e. not on a buddy list, which means the shrinker
is in the process of reclaiming it. If the zbpg is a zombie,
zbud_decompress() returns -EINVAL.
However, this return code is being ignored by the caller,
zcache_pampd_get_data_and_free(), which results in the
caller of cleancache_get_page() thinking that the page has
been properly retrieved when it has not.
This patch modifies zcache_pampd_get_data_and_free() to
convey the failure up the stack so that the caller of
cleancache_get_page() knows the page retrieval failed.
---
Based on v3.6-rc3.
This needs to be applied to stable trees as well.
zcache-main.c was named zcache.c before v3.1, so
I'm not sure how you want to handle trees earlier
than that.
Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
---
drivers/staging/zcache/zcache-main.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/drivers/staging/zcache/zcache-main.c b/drivers/staging/zcache/zcache-main.c
index c214977..52b43b7 100644
--- a/drivers/staging/zcache/zcache-main.c
+++ b/drivers/staging/zcache/zcache-main.c
@@ -1251,13 +1251,12 @@ static int zcache_pampd_get_data_and_free(char *data, size_t *bufsize, bool raw,
void *pampd, struct tmem_pool *pool,
struct tmem_oid *oid, uint32_t index)
{
- int ret = 0;
-
BUG_ON(!is_ephemeral(pool));
- zbud_decompress((struct page *)(data), pampd);
+ if (zbud_decompress((struct page *)(data), pampd) < 0)
+ return -EINVAL;
zbud_free_and_delist((struct zbud_hdr *)pampd);
atomic_dec(&zcache_curr_eph_pampd_count);
- return ret;
+ return 0;
}
/*
--
1.7.9.5
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] staging: zcache: fix cleancache race condition with shrinker
2012-08-29 21:58 [PATCH] staging: zcache: fix cleancache race condition with shrinker Seth Jennings
@ 2012-08-30 16:11 ` Konrad Rzeszutek Wilk
2012-08-30 17:55 ` Minchan Kim
1 sibling, 0 replies; 3+ messages in thread
From: Konrad Rzeszutek Wilk @ 2012-08-30 16:11 UTC (permalink / raw)
To: Seth Jennings
Cc: Greg Kroah-Hartman, Andrew Morton, Nitin Gupta, Minchan Kim,
Dan Magenheimer, Robert Jennings, Xiao Guangrong, linux-mm,
linux-kernel, devel
On Wed, Aug 29, 2012 at 04:58:45PM -0500, Seth Jennings wrote:
> This patch fixes a race condition that results in memory
> corruption when using cleancache.
>
> The race exists between the zcache shrinker handler,
> shrink_zcache_memory() and cleancache_get_page().
>
> In most cases, the shrinker will both evict a zbpg
> from its buddy list and flush it from tmem before a
> cleancache_get_page() occurs on that page. A subsequent
> cleancache_get_page() will fail in the tmem layer.
>
> In the rare case that two occur together and the
> cleancache_get_page() path gets through the tmem
> layer before the shrinker path can flush tmem,
> zbud_decompress() does a check to see if the zbpg is a
> "zombie", i.e. not on a buddy list, which means the shrinker
> is in the process of reclaiming it. If the zbpg is a zombie,
> zbud_decompress() returns -EINVAL.
>
> However, this return code is being ignored by the caller,
> zcache_pampd_get_data_and_free(), which results in the
> caller of cleancache_get_page() thinking that the page has
> been properly retrieved when it has not.
>
> This patch modifies zcache_pampd_get_data_and_free() to
> convey the failure up the stack so that the caller of
> cleancache_get_page() knows the page retrieval failed.
>
> ---
> Based on v3.6-rc3.
>
> This needs to be applied to stable trees as well.
> zcache-main.c was named zcache.c before v3.1, so
> I'm not sure how you want to handle trees earlier
> than that.
>
> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Thanks for tracking this down!
> ---
> drivers/staging/zcache/zcache-main.c | 7 +++----
> 1 file changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/staging/zcache/zcache-main.c b/drivers/staging/zcache/zcache-main.c
> index c214977..52b43b7 100644
> --- a/drivers/staging/zcache/zcache-main.c
> +++ b/drivers/staging/zcache/zcache-main.c
> @@ -1251,13 +1251,12 @@ static int zcache_pampd_get_data_and_free(char *data, size_t *bufsize, bool raw,
> void *pampd, struct tmem_pool *pool,
> struct tmem_oid *oid, uint32_t index)
> {
> - int ret = 0;
> -
> BUG_ON(!is_ephemeral(pool));
> - zbud_decompress((struct page *)(data), pampd);
> + if (zbud_decompress((struct page *)(data), pampd) < 0)
> + return -EINVAL;
> zbud_free_and_delist((struct zbud_hdr *)pampd);
> atomic_dec(&zcache_curr_eph_pampd_count);
> - return ret;
> + return 0;
> }
>
> /*
> --
> 1.7.9.5
>
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] staging: zcache: fix cleancache race condition with shrinker
2012-08-29 21:58 [PATCH] staging: zcache: fix cleancache race condition with shrinker Seth Jennings
2012-08-30 16:11 ` Konrad Rzeszutek Wilk
@ 2012-08-30 17:55 ` Minchan Kim
1 sibling, 0 replies; 3+ messages in thread
From: Minchan Kim @ 2012-08-30 17:55 UTC (permalink / raw)
To: Seth Jennings
Cc: Greg Kroah-Hartman, Andrew Morton, Nitin Gupta,
Konrad Rzeszutek Wilk, Dan Magenheimer, Robert Jennings,
Xiao Guangrong, linux-mm, linux-kernel, devel
On Wed, Aug 29, 2012 at 04:58:45PM -0500, Seth Jennings wrote:
> This patch fixes a race condition that results in memory
> corruption when using cleancache.
>
> The race exists between the zcache shrinker handler,
> shrink_zcache_memory() and cleancache_get_page().
>
> In most cases, the shrinker will both evict a zbpg
> from its buddy list and flush it from tmem before a
> cleancache_get_page() occurs on that page. A subsequent
> cleancache_get_page() will fail in the tmem layer.
>
> In the rare case that two occur together and the
> cleancache_get_page() path gets through the tmem
> layer before the shrinker path can flush tmem,
> zbud_decompress() does a check to see if the zbpg is a
> "zombie", i.e. not on a buddy list, which means the shrinker
> is in the process of reclaiming it. If the zbpg is a zombie,
> zbud_decompress() returns -EINVAL.
>
> However, this return code is being ignored by the caller,
> zcache_pampd_get_data_and_free(), which results in the
> caller of cleancache_get_page() thinking that the page has
> been properly retrieved when it has not.
>
> This patch modifies zcache_pampd_get_data_and_free() to
> convey the failure up the stack so that the caller of
> cleancache_get_page() knows the page retrieval failed.
>
> ---
> Based on v3.6-rc3.
>
> This needs to be applied to stable trees as well.
> zcache-main.c was named zcache.c before v3.1, so
> I'm not sure how you want to handle trees earlier
> than that.
>
> Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Thanks!
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2012-08-30 17:55 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-08-29 21:58 [PATCH] staging: zcache: fix cleancache race condition with shrinker Seth Jennings
2012-08-30 16:11 ` Konrad Rzeszutek Wilk
2012-08-30 17:55 ` Minchan Kim
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).