[PATCH] zcache: fix deadlock condition

[PATCH] zcache: fix deadlock condition

[Dan Magenheimer]

I discovered this deadlock condition awhile ago working on RAMster
but it affects zcache as well.  The list spinlock must be
locked prior to the page spinlock and released after.  As
a result, the page copy must also be done while the locks are held.

Applies to 3.2.  Konrad, please push (via GregKH?)...
this is definitely a bug fix so need not be pushed during
a -rc0 window.

Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>

diff --git a/drivers/staging/zcache/zcache-main.c b/drivers/staging/zcache/zcache-main.c
index 56c1f9c..5b9f74e 100644
--- a/drivers/staging/zcache/zcache-main.c
+++ b/drivers/staging/zcache/zcache-main.c
@@ -358,8 +358,8 @@ static struct zbud_hdr *zbud_create(uint16_t client_id, uint16_t pool_id,
 	if (unlikely(zbpg == NULL))
 		goto out;
 	/* ok, have a page, now compress the data before taking locks */
-	spin_lock(&zbpg->lock);
 	spin_lock(&zbud_budlists_spinlock);
+	spin_lock(&zbpg->lock);
 	list_add_tail(&zbpg->bud_list, &zbud_unbuddied[nchunks].list);
 	zbud_unbuddied[nchunks].count++;
 	zh = &zbpg->buddy[0];
@@ -389,12 +389,11 @@ init_zh:
 	zh->oid = *oid;
 	zh->pool_id = pool_id;
 	zh->client_id = client_id;
-	/* can wait to copy the data until the list locks are dropped */
-	spin_unlock(&zbud_budlists_spinlock);
-
 	to = zbud_data(zh, size);
 	memcpy(to, cdata, size);
 	spin_unlock(&zbpg->lock);
+	spin_unlock(&zbud_budlists_spinlock);
+
 	zbud_cumul_chunk_counts[nchunks]++;
 	atomic_inc(&zcache_zbud_curr_zpages);
 	zcache_zbud_cumul_zpages++;

Re: [PATCH] zcache: fix deadlock condition

[Konrad Rzeszutek Wilk]

On Wed, Jan 25, 2012 at 02:32:51PM -0800, Dan Magenheimer wrote:
> I discovered this deadlock condition awhile ago working on RAMster
> but it affects zcache as well.  The list spinlock must be
> locked prior to the page spinlock and released after.  As
> a result, the page copy must also be done while the locks are held.
> 
> Applies to 3.2.  Konrad, please push (via GregKH?)...
> this is definitely a bug fix so need not be pushed during
> a -rc0 window.

Looks good to me. Greg, could you apply the patch to your tree
please? [Or would you rather I sent an GIT pull?]

Thanks!
> 
> Signed-off-by: Dan Magenheimer <dan.magenheimer@oracle.com>
> 
> diff --git a/drivers/staging/zcache/zcache-main.c b/drivers/staging/zcache/zcache-main.c
> index 56c1f9c..5b9f74e 100644
> --- a/drivers/staging/zcache/zcache-main.c
> +++ b/drivers/staging/zcache/zcache-main.c
> @@ -358,8 +358,8 @@ static struct zbud_hdr *zbud_create(uint16_t client_id, uint16_t pool_id,
>  	if (unlikely(zbpg == NULL))
>  		goto out;
>  	/* ok, have a page, now compress the data before taking locks */
> -	spin_lock(&zbpg->lock);
>  	spin_lock(&zbud_budlists_spinlock);
> +	spin_lock(&zbpg->lock);
>  	list_add_tail(&zbpg->bud_list, &zbud_unbuddied[nchunks].list);
>  	zbud_unbuddied[nchunks].count++;
>  	zh = &zbpg->buddy[0];
> @@ -389,12 +389,11 @@ init_zh:
>  	zh->oid = *oid;
>  	zh->pool_id = pool_id;
>  	zh->client_id = client_id;
> -	/* can wait to copy the data until the list locks are dropped */
> -	spin_unlock(&zbud_budlists_spinlock);
> -
>  	to = zbud_data(zh, size);
>  	memcpy(to, cdata, size);
>  	spin_unlock(&zbpg->lock);
> +	spin_unlock(&zbud_budlists_spinlock);
> +
>  	zbud_cumul_chunk_counts[nchunks]++;
>  	atomic_inc(&zcache_zbud_curr_zpages);
>  	zcache_zbud_cumul_zpages++;

Re: [PATCH] zcache: fix deadlock condition

[Greg KH]

On Fri, Jan 27, 2012 at 03:39:28PM -0500, Konrad Rzeszutek Wilk wrote:
> On Wed, Jan 25, 2012 at 02:32:51PM -0800, Dan Magenheimer wrote:
> > I discovered this deadlock condition awhile ago working on RAMster
> > but it affects zcache as well.  The list spinlock must be
> > locked prior to the page spinlock and released after.  As
> > a result, the page copy must also be done while the locks are held.
> > 
> > Applies to 3.2.  Konrad, please push (via GregKH?)...
> > this is definitely a bug fix so need not be pushed during
> > a -rc0 window.
> 
> Looks good to me. Greg, could you apply the patch to your tree
> please? [Or would you rather I sent an GIT pull?]

I will queue it up, thanks.

greg k-h