[PATCH] nfsd4: fix unlikely race in session replay case

[PATCH] nfsd4: fix unlikely race in session replay case

[J. Bruce Fields]

From: J. Bruce Fields <bfields@citi.umich.edu>

In the replay case, the

	renew_client(session->se_client);

happens after we've droppped the sessionid_lock, and without holding a
reference on the session; so there's nothing preventing the session
being freed before we get here.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
---
 fs/nfsd/nfs4proc.c  |    3 +++
 fs/nfsd/nfs4state.c |    7 +++----
 fs/nfsd/nfs4xdr.c   |    1 -
 3 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index e147dbc..e4a4a3e 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -1027,6 +1027,7 @@ nfsd4_proc_compound(struct svc_rqst *rqstp,
 	resp->rqstp = rqstp;
 	resp->cstate.minorversion = args->minorversion;
 	resp->cstate.replay_owner = NULL;
+	resp->cstate.session = NULL;
 	fh_init(&resp->cstate.current_fh, NFS4_FHSIZE);
 	fh_init(&resp->cstate.save_fh, NFS4_FHSIZE);
 	/* Use the deferral mechanism only for NFSv4.0 compounds */
@@ -1131,6 +1132,8 @@ encode_op:
 	fh_put(&resp->cstate.save_fh);
 	BUG_ON(resp->cstate.replay_owner);
 out:
+	if (resp->cstate.session)
+		nfsd4_put_session(resp->cstate.session);
 	nfsd4_release_compoundargs(args);
 	/* Reset deferral mechanism for RPC deferrals */
 	rqstp->rq_usedeferral = 1;
diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index bba9fff..737315c 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -1443,11 +1443,10 @@ nfsd4_sequence(struct svc_rqst *rqstp,
 	cstate->slot = slot;
 	cstate->session = session;
 
-	/* Hold a session reference until done processing the compound:
-	 * nfsd4_put_session called only if the cstate slot is set.
-	 */
-	nfsd4_get_session(session);
 out:
+	/* Hold a session reference until done processing the compound. */
+	if (cstate->session)
+		nfsd4_get_session(cstate->session);
 	spin_unlock(&sessionid_lock);
 	/* Renew the clientid on success and on replay */
 	if (cstate->session) {
diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
index fb27b1d..903c5a6 100644
--- a/fs/nfsd/nfs4xdr.c
+++ b/fs/nfsd/nfs4xdr.c
@@ -3310,7 +3310,6 @@ nfs4svc_encode_compoundres(struct svc_rqst *rqstp, __be32 *p, struct nfsd4_compo
 		nfsd4_store_cache_entry(resp);
 		dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
 		resp->cstate.slot->sl_inuse = false;
-		nfsd4_put_session(resp->cstate.session);
 	}
 	return 1;
 }
-- 
1.6.3.3

Re: [PATCH] nfsd4: fix unlikely race in session replay case

[Benny Halevy]

On Apr. 29, 2010, 0:31 +0300, "J. Bruce Fields" <bfields@fieldses.org> wrote:
> From: J. Bruce Fields <bfields@citi.umich.edu>
> 
> In the replay case, the
> 
> 	renew_client(session->se_client);
> 
> happens after we've droppped the sessionid_lock, and without holding a
> reference on the session; so there's nothing preventing the session
> being freed before we get here.
> 
> Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
> ---
>  fs/nfsd/nfs4proc.c  |    3 +++
>  fs/nfsd/nfs4state.c |    7 +++----
>  fs/nfsd/nfs4xdr.c   |    1 -
>  3 files changed, 6 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> index e147dbc..e4a4a3e 100644
> --- a/fs/nfsd/nfs4proc.c
> +++ b/fs/nfsd/nfs4proc.c
> @@ -1027,6 +1027,7 @@ nfsd4_proc_compound(struct svc_rqst *rqstp,
>  	resp->rqstp = rqstp;
>  	resp->cstate.minorversion = args->minorversion;
>  	resp->cstate.replay_owner = NULL;
> +	resp->cstate.session = NULL;
>  	fh_init(&resp->cstate.current_fh, NFS4_FHSIZE);
>  	fh_init(&resp->cstate.save_fh, NFS4_FHSIZE);
>  	/* Use the deferral mechanism only for NFSv4.0 compounds */
> @@ -1131,6 +1132,8 @@ encode_op:
>  	fh_put(&resp->cstate.save_fh);
>  	BUG_ON(resp->cstate.replay_owner);
>  out:
> +	if (resp->cstate.session)
> +		nfsd4_put_session(resp->cstate.session);

Bruce, I see what you wanted to do but unfortunately putting the session
here won't work since resp->cstate.slot is referred to later on in
nfs4svc_encode_compoundres and it points into the slot table
which is a member of the same session, therefore the current place
in which it gets dereferenced should be maintained.

Please see below for another suggestion

>  	nfsd4_release_compoundargs(args);
>  	/* Reset deferral mechanism for RPC deferrals */
>  	rqstp->rq_usedeferral = 1;
> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> index bba9fff..737315c 100644
> --- a/fs/nfsd/nfs4state.c
> +++ b/fs/nfsd/nfs4state.c
> @@ -1443,11 +1443,10 @@ nfsd4_sequence(struct svc_rqst *rqstp,
>  	cstate->slot = slot;
>  	cstate->session = session;
>  
> -	/* Hold a session reference until done processing the compound:
> -	 * nfsd4_put_session called only if the cstate slot is set.
> -	 */
> -	nfsd4_get_session(session);
>  out:
> +	/* Hold a session reference until done processing the compound. */
> +	if (cstate->session)
> +		nfsd4_get_session(cstate->session);
>  	spin_unlock(&sessionid_lock);
>  	/* Renew the clientid on success and on replay */
>  	if (cstate->session) {
> diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
> index fb27b1d..903c5a6 100644
> --- a/fs/nfsd/nfs4xdr.c
> +++ b/fs/nfsd/nfs4xdr.c
> @@ -3310,7 +3310,6 @@ nfs4svc_encode_compoundres(struct svc_rqst *rqstp, __be32 *p, struct nfsd4_compo
>  		nfsd4_store_cache_entry(resp);
>  		dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
>  		resp->cstate.slot->sl_inuse = false;
> -		nfsd4_put_session(resp->cstate.session);
>  	}
>  	return 1;
>  }


How about this instead:

diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
index 34ccf81..b04583c 100644
--- a/fs/nfsd/nfs4xdr.c
+++ b/fs/nfsd/nfs4xdr.c
@@ -3307,10 +3307,12 @@ nfs4svc_encode_compoundres(struct svc_rqst *rqstp, __be32 *p, struct nfsd4_compo
 		iov = &rqstp->rq_res.head[0];
 	iov->iov_len = ((char*)resp->p) - (char*)iov->iov_base;
 	BUG_ON(iov->iov_len > PAGE_SIZE);
-	if (nfsd4_has_session(cs) && cs->status != nfserr_replay_cache) {
-		nfsd4_store_cache_entry(resp);
-		dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
-		resp->cstate.slot->sl_inuse = false;
+	if (nfsd4_has_session(cs)) {
+		if (cs->status != nfserr_replay_cache) {
+			nfsd4_store_cache_entry(resp);
+			dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
+			resp->cstate.slot->sl_inuse = false;
+		}
 		nfsd4_put_session(resp->cstate.session);
 	}
 	return 1;

Re: [PATCH] nfsd4: fix unlikely race in session replay case

[J. Bruce Fields]

On Mon, May 03, 2010 at 02:20:12PM +0300, Benny Halevy wrote:
> On Apr. 29, 2010, 0:31 +0300, "J. Bruce Fields" <bfields@fieldses.org> wrote:
> > @@ -1131,6 +1132,8 @@ encode_op:
> >  	fh_put(&resp->cstate.save_fh);
> >  	BUG_ON(resp->cstate.replay_owner);
> >  out:
> > +	if (resp->cstate.session)
> > +		nfsd4_put_session(resp->cstate.session);
> 
> Bruce, I see what you wanted to do but unfortunately putting the session
> here won't work since resp->cstate.slot is referred to later on in
> nfs4svc_encode_compoundres and it points into the slot table
> which is a member of the same session, therefore the current place
> in which it gets dereferenced should be maintained.

Whoops, good catch!  (Since most of the xdr encoding is done before
nfsd4_proc_compound() exits, I think I must have forgotten that
nfs4svc_encode_compoundres() is an exception?)

> Please see below for another suggestion

Done as you suggest.  I'd applied the patch locally but not pushed it
out yet, so I've updated it in place; the result is the following.  I'll
apply and push that to for-2.6.35 if you don't see any further probleme.

--b.

commit 44643622678d011642531915442d404ad14eb6c6
Author: J. Bruce Fields <bfields@citi.umich.edu>
Date:   Sat Apr 24 15:35:43 2010 -0400

    nfsd4: fix unlikely race in session replay case
    
    In the replay case, the
    
    	renew_client(session->se_client);
    
    happens after we've droppped the sessionid_lock, and without holding a
    reference on the session; so there's nothing preventing the session
    being freed before we get here.
    
    Thanks to Benny Halevy for catching a bug in an earlier version of this
    patch.
    
    Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
index e147dbc..61282f8 100644
--- a/fs/nfsd/nfs4proc.c
+++ b/fs/nfsd/nfs4proc.c
@@ -1027,6 +1027,7 @@ nfsd4_proc_compound(struct svc_rqst *rqstp,
 	resp->rqstp = rqstp;
 	resp->cstate.minorversion = args->minorversion;
 	resp->cstate.replay_owner = NULL;
+	resp->cstate.session = NULL;
 	fh_init(&resp->cstate.current_fh, NFS4_FHSIZE);
 	fh_init(&resp->cstate.save_fh, NFS4_FHSIZE);
 	/* Use the deferral mechanism only for NFSv4.0 compounds */
diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
index bba9fff..737315c 100644
--- a/fs/nfsd/nfs4state.c
+++ b/fs/nfsd/nfs4state.c
@@ -1443,11 +1443,10 @@ nfsd4_sequence(struct svc_rqst *rqstp,
 	cstate->slot = slot;
 	cstate->session = session;
 
-	/* Hold a session reference until done processing the compound:
-	 * nfsd4_put_session called only if the cstate slot is set.
-	 */
-	nfsd4_get_session(session);
 out:
+	/* Hold a session reference until done processing the compound. */
+	if (cstate->session)
+		nfsd4_get_session(cstate->session);
 	spin_unlock(&sessionid_lock);
 	/* Renew the clientid on success and on replay */
 	if (cstate->session) {
diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
index fb27b1d..05bc5bd 100644
--- a/fs/nfsd/nfs4xdr.c
+++ b/fs/nfsd/nfs4xdr.c
@@ -3306,10 +3306,12 @@ nfs4svc_encode_compoundres(struct svc_rqst *rqstp, __be32 *p, struct nfsd4_compo
 		iov = &rqstp->rq_res.head[0];
 	iov->iov_len = ((char*)resp->p) - (char*)iov->iov_base;
 	BUG_ON(iov->iov_len > PAGE_SIZE);
-	if (nfsd4_has_session(cs) && cs->status != nfserr_replay_cache) {
-		nfsd4_store_cache_entry(resp);
-		dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
-		resp->cstate.slot->sl_inuse = false;
+	if (nfsd4_has_session(cs)) {
+		if (cs->status != nfserr_replay_cache) {
+			nfsd4_store_cache_entry(resp);
+			dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
+			resp->cstate.slot->sl_inuse = false;
+		}
 		nfsd4_put_session(resp->cstate.session);
 	}
 	return 1;

Re: [PATCH] nfsd4: fix unlikely race in session replay case

[Benny Halevy]

On May. 03, 2010, 15:22 +0300, "J. Bruce Fields" <bfields@fieldses.org> wrote:
> On Mon, May 03, 2010 at 02:20:12PM +0300, Benny Halevy wrote:
>> On Apr. 29, 2010, 0:31 +0300, "J. Bruce Fields" <bfields@fieldses.org> wrote:
>>> @@ -1131,6 +1132,8 @@ encode_op:
>>>  	fh_put(&resp->cstate.save_fh);
>>>  	BUG_ON(resp->cstate.replay_owner);
>>>  out:
>>> +	if (resp->cstate.session)
>>> +		nfsd4_put_session(resp->cstate.session);
>> Bruce, I see what you wanted to do but unfortunately putting the session
>> here won't work since resp->cstate.slot is referred to later on in
>> nfs4svc_encode_compoundres and it points into the slot table
>> which is a member of the same session, therefore the current place
>> in which it gets dereferenced should be maintained.
> 
> Whoops, good catch!  (Since most of the xdr encoding is done before
> nfsd4_proc_compound() exits, I think I must have forgotten that
> nfs4svc_encode_compoundres() is an exception?)
> 
>> Please see below for another suggestion
> 
> Done as you suggest.  I'd applied the patch locally but not pushed it
> out yet, so I've updated it in place; the result is the following.  I'll
> apply and push that to for-2.6.35 if you don't see any further probleme.
> 
> --b.
> 
> commit 44643622678d011642531915442d404ad14eb6c6
> Author: J. Bruce Fields <bfields@citi.umich.edu>
> Date:   Sat Apr 24 15:35:43 2010 -0400
> 
>     nfsd4: fix unlikely race in session replay case
>     
>     In the replay case, the
>     
>     	renew_client(session->se_client);
>     
>     happens after we've droppped the sessionid_lock, and without holding a
>     reference on the session; so there's nothing preventing the session
>     being freed before we get here.
>     
>     Thanks to Benny Halevy for catching a bug in an earlier version of this
>     patch.
>     
>     Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

Acked-by: Benny Halevy <bhalevy@panasas.com>

> 
> diff --git a/fs/nfsd/nfs4proc.c b/fs/nfsd/nfs4proc.c
> index e147dbc..61282f8 100644
> --- a/fs/nfsd/nfs4proc.c
> +++ b/fs/nfsd/nfs4proc.c
> @@ -1027,6 +1027,7 @@ nfsd4_proc_compound(struct svc_rqst *rqstp,
>  	resp->rqstp = rqstp;
>  	resp->cstate.minorversion = args->minorversion;
>  	resp->cstate.replay_owner = NULL;
> +	resp->cstate.session = NULL;
>  	fh_init(&resp->cstate.current_fh, NFS4_FHSIZE);
>  	fh_init(&resp->cstate.save_fh, NFS4_FHSIZE);
>  	/* Use the deferral mechanism only for NFSv4.0 compounds */
> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> index bba9fff..737315c 100644
> --- a/fs/nfsd/nfs4state.c
> +++ b/fs/nfsd/nfs4state.c
> @@ -1443,11 +1443,10 @@ nfsd4_sequence(struct svc_rqst *rqstp,
>  	cstate->slot = slot;
>  	cstate->session = session;
>  
> -	/* Hold a session reference until done processing the compound:
> -	 * nfsd4_put_session called only if the cstate slot is set.
> -	 */
> -	nfsd4_get_session(session);
>  out:
> +	/* Hold a session reference until done processing the compound. */
> +	if (cstate->session)
> +		nfsd4_get_session(cstate->session);
>  	spin_unlock(&sessionid_lock);
>  	/* Renew the clientid on success and on replay */
>  	if (cstate->session) {
> diff --git a/fs/nfsd/nfs4xdr.c b/fs/nfsd/nfs4xdr.c
> index fb27b1d..05bc5bd 100644
> --- a/fs/nfsd/nfs4xdr.c
> +++ b/fs/nfsd/nfs4xdr.c
> @@ -3306,10 +3306,12 @@ nfs4svc_encode_compoundres(struct svc_rqst *rqstp, __be32 *p, struct nfsd4_compo
>  		iov = &rqstp->rq_res.head[0];
>  	iov->iov_len = ((char*)resp->p) - (char*)iov->iov_base;
>  	BUG_ON(iov->iov_len > PAGE_SIZE);
> -	if (nfsd4_has_session(cs) && cs->status != nfserr_replay_cache) {
> -		nfsd4_store_cache_entry(resp);
> -		dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
> -		resp->cstate.slot->sl_inuse = false;
> +	if (nfsd4_has_session(cs)) {
> +		if (cs->status != nfserr_replay_cache) {
> +			nfsd4_store_cache_entry(resp);
> +			dprintk("%s: SET SLOT STATE TO AVAILABLE\n", __func__);
> +			resp->cstate.slot->sl_inuse = false;
> +		}
>  		nfsd4_put_session(resp->cstate.session);
>  	}
>  	return 1;
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html