Skip to contentSkip to conversation

FAILED: patch "[PATCH] nfsd: Fix another OPEN stateid race" failed to apply to 4.4-stable tree

PATCH [email protected] 3 participants 4 messages Dec 4, 2017

Subsystems (2)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

The patch below does not apply to the 4.4-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

Quoted 1 line

From d8a1a000555ecd1b824ac1ed6df8fe364dfbbbb0 Mon Sep 17 00:00:00 2001

From: Trond Myklebust <[email protected]>
Date: Fri, 3 Nov 2017 08:00:11 -0400
Subject: [PATCH] nfsd: Fix another OPEN stateid race

If nfsd4_process_open2() is initialising a new stateid, and yet the call to nfs4_get_vfs_file() fails for some reason, then we must declare the stateid closed, and unhash it before dropping the mutex.

Right now, we unhash the stateid after dropping the mutex, and without changing the stateid type, meaning that another OPEN could theoretically look it up and attempt to use it.

fs/nfsd/nfs4state.c +13 -15
index ee8fde2dfa92..457f0e7ece74 100644
@@ -4502,6 +4502,7 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
 	struct nfs4_ol_stateid *stp = NULL;
 	struct nfs4_delegation *dp = NULL;
 	__be32 status;
+	bool new_stp = false;
 
 	/*
 	 * Lookup file; if found, lookup stateid and check open request,
@@ -4521,11 +4522,19 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
 			goto out;
 	}
 
+	if (!stp) {
+		stp = init_open_stateid(fp, open);
+		if (!open->op_stp)
+			new_stp = true;
+	}
+
 	/*
 	 * OPEN the file, or upgrade an existing OPEN.
 	 * If truncate fails, the OPEN fails.
+	 *
+	 * stp is already locked.
 	 */
-	if (stp) {
+	if (!new_stp) {
 		/* Stateid was found, this is an OPEN upgrade */
 		status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
 		if (status) {
@@ -4533,22 +4542,11 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
 			goto out;
 		}
 	} else {
-		/* stp is returned locked. */
-		stp = init_open_stateid(fp, open);
-		/* See if we lost the race to some other thread */
-		if (stp->st_access_bmap != 0) {
-			status = nfs4_upgrade_open(rqstp, fp, current_fh,
-						stp, open);
-			if (status) {
-				mutex_unlock(&stp->st_mutex);
-				goto out;
-			}
-			goto upgrade_out;
-		}
 		status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
 		if (status) {
-			mutex_unlock(&stp->st_mutex);
+			stp->st_stid.sc_type = NFS4_CLOSED_STID;
 			release_open_stateid(stp);
+			mutex_unlock(&stp->st_mutex);
 			goto out;
 		}
 
@@ -4557,7 +4555,7 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
 		if (stp->st_clnt_odstate == open->op_odstate)
 			open->op_odstate = NULL;
 	}
-upgrade_out:
+
 	nfs4_inc_and_copy_stateid(&open->op_stateid, &stp->st_stid);
 	mutex_unlock(&stp->st_mutex);
 
#1 lore raw
Quoted 8 lines · <[email protected]> writes:

The patch below does not apply to the 4.4-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

thanks,

greg k-h

Greg,

These backport cleanly if you cherry-pick 8c7245abda first:

commit 8c7245abda877d4689b3371db8ae2a4400d7d9ce
then
commit 15ca08d3299682dc49bad73251677b2c5017ef08
then
commit d8a1a000555ecd1b824ac1ed6df8fe364dfbbbb0

Thanks,

Andy

Quoted 89 lines

------------------ original commit in Linus's tree ------------------

Quoted 1 line
Quoted 1 line

From d8a1a000555ecd1b824ac1ed6df8fe364dfbbbb0 Mon Sep 17 00:00:00 2001

From: Trond Myklebust <[email protected]>
Date: Fri, 3 Nov 2017 08:00:11 -0400
Subject: [PATCH] nfsd: Fix another OPEN stateid race

If nfsd4_process_open2() is initialising a new stateid, and yet the call to nfs4_get_vfs_file() fails for some reason, then we must declare the stateid closed, and unhash it before dropping the mutex.

Right now, we unhash the stateid after dropping the mutex, and without changing the stateid type, meaning that another OPEN could theoretically look it up and attempt to use it.

fs/nfsd/nfs4state.c +13 -15
index ee8fde2dfa92..457f0e7ece74 100644
@@ -4502,6 +4502,7 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
 	struct nfs4_ol_stateid *stp = NULL;
 	struct nfs4_delegation *dp = NULL;
 	__be32 status;
+	bool new_stp = false;
 
 	/*
 	 * Lookup file; if found, lookup stateid and check open request,
@@ -4521,11 +4522,19 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
 			goto out;
 	}
 
+	if (!stp) {
+		stp = init_open_stateid(fp, open);
+		if (!open->op_stp)
+			new_stp = true;
+	}
+
 	/*
 	 * OPEN the file, or upgrade an existing OPEN.
 	 * If truncate fails, the OPEN fails.
+	 *
+	 * stp is already locked.
 	 */
-	if (stp) {
+	if (!new_stp) {
 		/* Stateid was found, this is an OPEN upgrade */
 		status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
 		if (status) {
@@ -4533,22 +4542,11 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
 			goto out;
 		}
 	} else {
-		/* stp is returned locked. */
-		stp = init_open_stateid(fp, open);
-		/* See if we lost the race to some other thread */
-		if (stp->st_access_bmap != 0) {
-			status = nfs4_upgrade_open(rqstp, fp, current_fh,
-						stp, open);
-			if (status) {
-				mutex_unlock(&stp->st_mutex);
-				goto out;
-			}
-			goto upgrade_out;
-		}
 		status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
 		if (status) {
-			mutex_unlock(&stp->st_mutex);
+			stp->st_stid.sc_type = NFS4_CLOSED_STID;
 			release_open_stateid(stp);
+			mutex_unlock(&stp->st_mutex);
 			goto out;
 		}
 
@@ -4557,7 +4555,7 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
 		if (stp->st_clnt_odstate == open->op_odstate)
 			open->op_odstate = NULL;
 	}
-upgrade_out:
+
 	nfs4_inc_and_copy_stateid(&open->op_stateid, &stp->st_stid);
 	mutex_unlock(&stp->st_mutex);
 
Signature · 5 lines
Andrew W. Elble
[email protected]
Infrastructure Engineer, Communications Technical Lead
Rochester Institute of Technology
PGP: BFAD 8461 4CCF DC95 DA2C B0EB 965B 082E 863E C912
author maintainer
#2 lore raw
Quoted 13 lines · Andrew W Elble
Quoted 8 lines · <[email protected]> writes:

The patch below does not apply to the 4.4-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

thanks,

greg k-h

Greg,

Quoted 7 lines

These backport cleanly if you cherry-pick 8c7245abda first:

commit 8c7245abda877d4689b3371db8ae2a4400d7d9ce
then
commit 15ca08d3299682dc49bad73251677b2c5017ef08
then
commit d8a1a000555ecd1b824ac1ed6df8fe364dfbbbb0

Wonderful, that worked!

greg k-h

#3 lore raw
In reply to Andrew W Elble
Quoted 13 lines · Andrew W Elble
Quoted 8 lines · <[email protected]> writes:

The patch below does not apply to the 4.4-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

thanks,

greg k-h

Greg,

Quoted 7 lines

These backport cleanly if you cherry-pick 8c7245abda first:

commit 8c7245abda877d4689b3371db8ae2a4400d7d9ce
then
commit 15ca08d3299682dc49bad73251677b2c5017ef08
then
commit d8a1a000555ecd1b824ac1ed6df8fe364dfbbbb0

Looks right to me. Thanks, Andy!

--b.