Skip to contentSkip to conversation

split setattr operations take 2

PATCH In review Christoph Hellwig 4 participants 5 messages Feb 20, 2017 – Feb 21, 2017

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

This just splits the setattr operations and doesn't try to use more of the VFS infrastructure.

#1 lore raw

PATCH 1/1 nfsd: special case truncates some more

Both the NFS protocols and the Linux VFS use a setattr operation with a bitmap of attributs to set to set various file attributes including the file size and the uid/gid.

The Linux syscalls never mixes size updates with unrelated updates like the uid/gid, and some file systems like XFS and GFS2 rely on the fact that truncates might not update random other attributes, and many other file systems handle the case but do not update the different attributes in the same transaction. NFSD on the other hand passes the attributes it gets on the wire more or less directly through to the VFS, leading to updates the file systems don't expect. XFS at least has an assert on the allowed attributes, which caught an unusual NFS client setting the size and group at the same time.

To handle this issue properly this splits the notify_change call in nfsd_setattr into two separate ones.

---
 fs/nfsd/vfs.c | 59 +++++++++++++++++++++++++++++++++++++----------------------
 1 file changed, 37 insertions(+), 22 deletions(-)
fs/nfsd/vfs.c +37 -22
index 26c6fdb4bf67..3c36ed5a1f07 100644
@@ -377,7 +377,7 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 	__be32		err;
 	int		host_err;
 	bool		get_write_count;
-	int		size_change = 0;
+	bool		size_change = (iap->ia_valid & ATTR_SIZE);
 
 	if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
 		accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
@@ -390,11 +390,11 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 	/* Get inode */
 	err = fh_verify(rqstp, fhp, ftype, accmode);
 	if (err)
-		goto out;
+		return err;
 	if (get_write_count) {
 		host_err = fh_want_write(fhp);
 		if (host_err)
-			return nfserrno(host_err);
+			goto out;
 	}
 
 	dentry = fhp->fh_dentry;
@@ -405,20 +405,28 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 		iap->ia_valid &= ~ATTR_MODE;
 
 	if (!iap->ia_valid)
-		goto out;
+		return 0;
 
 	nfsd_sanitize_attrs(inode, iap);
 
+	if (check_guard && guardtime != inode->i_ctime.tv_sec)
+		return nfserr_notsync;
+
 	/*
 	 * The size case is special, it changes the file in addition to the
-	 * attributes.
+	 * attributes, and file systems don't expect it to be mixed with
+	 * "random" attribute changes.  We thus split out the size change
+	 * into a separate call to ->setattr, and do the rest as a separate
+	 * setattr call.
 	 */
-	if (iap->ia_valid & ATTR_SIZE) {
+	if (size_change) {
 		err = nfsd_get_write_access(rqstp, fhp, iap);
 		if (err)
-			goto out;
-		size_change = 1;
+			return err;
+	}
 
+	fh_lock(fhp);
+	if (size_change) {
 		/*
 		 * RFC5661, Section 18.30.4:
 		 *   Changing the size of a file with SETATTR indirectly
@@ -426,29 +434,36 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 		 *
 		 * (and similar for the older RFCs)
 		 */
-		if (iap->ia_size != i_size_read(inode))
-			iap->ia_valid |= ATTR_MTIME;
-	}
+		struct iattr size_attr = {
+			.ia_valid	= ATTR_SIZE | ATTR_CTIME | ATTR_MTIME,
+			.ia_size	= iap->ia_size,
+		};
 
-	iap->ia_valid |= ATTR_CTIME;
+		host_err = notify_change(dentry, &size_attr, NULL);
+		if (host_err)
+			goto out_unlock;
+		iap->ia_valid &= ~ATTR_SIZE;
 
-	if (check_guard && guardtime != inode->i_ctime.tv_sec) {
-		err = nfserr_notsync;
-		goto out_put_write_access;
+		/*
+		 * Avoid the additional setattr call below if the only other
+		 * attribute that the client sends is the mtime, as we update
+		 * it as part of the size change above.
+		 */
+		if ((iap->ia_valid & ~ATTR_MTIME) == 0)
+			goto out_unlock;
 	}
 
-	fh_lock(fhp);
+	iap->ia_valid |= ATTR_CTIME;
 	host_err = notify_change(dentry, iap, NULL);
-	fh_unlock(fhp);
-	err = nfserrno(host_err);
 
-out_put_write_access:
+out_unlock:
+	fh_unlock(fhp);
 	if (size_change)
 		put_write_access(inode);
-	if (!err)
-		err = nfserrno(commit_metadata(fhp));
 out:
-	return err;
+	if (!host_err)
+		host_err = commit_metadata(fhp);
+	return nfserrno(host_err);
 }
 
 #if defined(CONFIG_NFSD_V4)
Signature · 1 line
2.11.0
#2 lore raw

Re: [PATCH] nfsd: special case truncates some more

Thanks! I split out the cleanup into a separate patch just to make sure I understood what the important change was.... Looks good to me.

--b.

Quoted 137 lines · Christoph Hellwig

Both the NFS protocols and the Linux VFS use a setattr operation with a bitmap of attributs to set to set various file attributes including the file size and the uid/gid.

The Linux syscalls never mixes size updates with unrelated updates like the uid/gid, and some file systems like XFS and GFS2 rely on the fact that truncates might not update random other attributes, and many other file systems handle the case but do not update the different attributes in the same transaction. NFSD on the other hand passes the attributes it gets on the wire more or less directly through to the VFS, leading to updates the file systems don't expect. XFS at least has an assert on the allowed attributes, which caught an unusual NFS client setting the size and group at the same time.

To handle this issue properly this splits the notify_change call in nfsd_setattr into two separate ones.

---
 fs/nfsd/vfs.c | 59 +++++++++++++++++++++++++++++++++++++----------------------
 1 file changed, 37 insertions(+), 22 deletions(-)
fs/nfsd/vfs.c +37 -22
index 26c6fdb4bf67..3c36ed5a1f07 100644
@@ -377,7 +377,7 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 	__be32		err;
 	int		host_err;
 	bool		get_write_count;
-	int		size_change = 0;
+	bool		size_change = (iap->ia_valid & ATTR_SIZE);
 
 	if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
 		accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
@@ -390,11 +390,11 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 	/* Get inode */
 	err = fh_verify(rqstp, fhp, ftype, accmode);
 	if (err)
-		goto out;
+		return err;
 	if (get_write_count) {
 		host_err = fh_want_write(fhp);
 		if (host_err)
-			return nfserrno(host_err);
+			goto out;
 	}
 
 	dentry = fhp->fh_dentry;
@@ -405,20 +405,28 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 		iap->ia_valid &= ~ATTR_MODE;
 
 	if (!iap->ia_valid)
-		goto out;
+		return 0;
 
 	nfsd_sanitize_attrs(inode, iap);
 
+	if (check_guard && guardtime != inode->i_ctime.tv_sec)
+		return nfserr_notsync;
+
 	/*
 	 * The size case is special, it changes the file in addition to the
-	 * attributes.
+	 * attributes, and file systems don't expect it to be mixed with
+	 * "random" attribute changes.  We thus split out the size change
+	 * into a separate call to ->setattr, and do the rest as a separate
+	 * setattr call.
 	 */
-	if (iap->ia_valid & ATTR_SIZE) {
+	if (size_change) {
 		err = nfsd_get_write_access(rqstp, fhp, iap);
 		if (err)
-			goto out;
-		size_change = 1;
+			return err;
+	}
 
+	fh_lock(fhp);
+	if (size_change) {
 		/*
 		 * RFC5661, Section 18.30.4:
 		 *   Changing the size of a file with SETATTR indirectly
@@ -426,29 +434,36 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 		 *
 		 * (and similar for the older RFCs)
 		 */
-		if (iap->ia_size != i_size_read(inode))
-			iap->ia_valid |= ATTR_MTIME;
-	}
+		struct iattr size_attr = {
+			.ia_valid	= ATTR_SIZE | ATTR_CTIME | ATTR_MTIME,
+			.ia_size	= iap->ia_size,
+		};
 
-	iap->ia_valid |= ATTR_CTIME;
+		host_err = notify_change(dentry, &size_attr, NULL);
+		if (host_err)
+			goto out_unlock;
+		iap->ia_valid &= ~ATTR_SIZE;
 
-	if (check_guard && guardtime != inode->i_ctime.tv_sec) {
-		err = nfserr_notsync;
-		goto out_put_write_access;
+		/*
+		 * Avoid the additional setattr call below if the only other
+		 * attribute that the client sends is the mtime, as we update
+		 * it as part of the size change above.
+		 */
+		if ((iap->ia_valid & ~ATTR_MTIME) == 0)
+			goto out_unlock;
 	}
 
-	fh_lock(fhp);
+	iap->ia_valid |= ATTR_CTIME;
 	host_err = notify_change(dentry, iap, NULL);
-	fh_unlock(fhp);
-	err = nfserrno(host_err);
 
-out_put_write_access:
+out_unlock:
+	fh_unlock(fhp);
 	if (size_change)
 		put_write_access(inode);
-	if (!err)
-		err = nfserrno(commit_metadata(fhp));
 out:
-	return err;
+	if (!host_err)
+		host_err = commit_metadata(fhp);
+	return nfserrno(host_err);
 }
 
 #if defined(CONFIG_NFSD_V4)
Signature · 1 line
2.11.0
Tested-by on v1 1/1
#3 lore raw

Re: [PATCH] nfsd: special case truncates some more

In reply to Christoph Hellwig
Quoted 17 lines · Christoph Hellwig

Both the NFS protocols and the Linux VFS use a setattr operation with a bitmap of attributs to set to set various file attributes including the file size and the uid/gid.

The Linux syscalls never mixes size updates with unrelated updates like the uid/gid, and some file systems like XFS and GFS2 rely on the fact that truncates might not update random other attributes, and many other file systems handle the case but do not update the different attributes in the same transaction. NFSD on the other hand passes the attributes it gets on the wire more or less directly through to the VFS, leading to updates the file systems don't expect. XFS at least has an assert on the allowed attributes, which caught an unusual NFS client setting the size and group at the same time.

Quoted 5 lines

To handle this issue properly this splits the notify_change call in nfsd_setattr into two separate ones.

Quoted 115 lines

---
fs/nfsd/vfs.c | 59 +++++++++++++++++++++++++++++++++++++----------------------

1 file changed, 37 insertions(+), 22 deletions(-)
fs/nfsd/vfs.c +34 -21
index 26c6fdb4bf67..3c36ed5a1f07 100644
@@ -377,7 +377,7 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 	__be32		err;
 	int		host_err;
 	bool		get_write_count;
-	int		size_change = 0;
+	bool		size_change = (iap->ia_valid & ATTR_SIZE);
 
 	if (iap->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_SIZE))
 		accmode |= NFSD_MAY_WRITE|NFSD_MAY_OWNER_OVERRIDE;
@@ -390,11 +390,11 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 	/* Get inode */
 	err = fh_verify(rqstp, fhp, ftype, accmode);
 	if (err)
-		goto out;
+		return err;
 	if (get_write_count) {
 		host_err = fh_want_write(fhp);
 		if (host_err)
-			return nfserrno(host_err);
+			goto out;
 	}
 
 	dentry = fhp->fh_dentry;
@@ -405,20 +405,28 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 		iap->ia_valid &= ~ATTR_MODE;
 
 	if (!iap->ia_valid)
-		goto out;
+		return 0;
 
 	nfsd_sanitize_attrs(inode, iap);
 
+	if (check_guard && guardtime != inode->i_ctime.tv_sec)
+		return nfserr_notsync;
+
 	/*
 	 * The size case is special, it changes the file in addition to the
-	 * attributes.
+	 * attributes, and file systems don't expect it to be mixed with
+	 * "random" attribute changes.  We thus split out the size change
+	 * into a separate call to ->setattr, and do the rest as a separate
+	 * setattr call.
 	 */
-	if (iap->ia_valid & ATTR_SIZE) {
+	if (size_change) {
 		err = nfsd_get_write_access(rqstp, fhp, iap);
 		if (err)
-			goto out;
-		size_change = 1;
+			return err;
+	}
 
+	fh_lock(fhp);
+	if (size_change) {
 		/*
 		 * RFC5661, Section 18.30.4:
 		 *   Changing the size of a file with SETATTR indirectly
@@ -426,29 +434,36 @@ nfsd_setattr(struct svc_rqst *rqstp, struct svc_fh *fhp, struct iattr *iap,
 		 *
 		 * (and similar for the older RFCs)
 		 */
-		if (iap->ia_size != i_size_read(inode))
-			iap->ia_valid |= ATTR_MTIME;
-	}
+		struct iattr size_attr = {
+			.ia_valid	= ATTR_SIZE | ATTR_CTIME | ATTR_MTIME,
+			.ia_size	= iap->ia_size,
+		};
 
-	iap->ia_valid |= ATTR_CTIME;
+		host_err = notify_change(dentry, &size_attr, NULL);
+		if (host_err)
+			goto out_unlock;
+		iap->ia_valid &= ~ATTR_SIZE;
 
-	if (check_guard && guardtime != inode->i_ctime.tv_sec) {
-		err = nfserr_notsync;
-		goto out_put_write_access;
+		/*
+		 * Avoid the additional setattr call below if the only other
+		 * attribute that the client sends is the mtime, as we update
+		 * it as part of the size change above.
+		 */
+		if ((iap->ia_valid & ~ATTR_MTIME) == 0)
+			goto out_unlock;
 	}
 
-	fh_lock(fhp);
+	iap->ia_valid |= ATTR_CTIME;
 	host_err = notify_change(dentry, iap, NULL);
-	fh_unlock(fhp);
-	err = nfserrno(host_err);
 
-out_put_write_access:
+out_unlock:
+	fh_unlock(fhp);
 	if (size_change)
 		put_write_access(inode);
-	if (!err)
-		err = nfserrno(commit_metadata(fhp));

out:

fs/nfsd/vfs.c +3 -1
-	return err;
+	if (!host_err)
+		host_err = commit_metadata(fhp);
+	return nfserrno(host_err);

}

Quoted 8 lines

#if defined(CONFIG_NFSD_V4)

Signature · 6 lines
2.11.0

--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
Chuck Lever

#4 lore raw

Re: [PATCH] nfsd: special case truncates some more

Quoted 17 lines · Chuck Lever
Quoted 17 lines · Christoph Hellwig

Both the NFS protocols and the Linux VFS use a setattr operation with a bitmap of attributs to set to set various file attributes including the file size and the uid/gid.

The Linux syscalls never mixes size updates with unrelated updates like the uid/gid, and some file systems like XFS and GFS2 rely on the fact that truncates might not update random other attributes, and many other file systems handle the case but do not update the different attributes in the same transaction. NFSD on the other hand passes the attributes it gets on the wire more or less directly through to the VFS, leading to updates the file systems don't expect. XFS at least has an assert on the allowed attributes, which caught an unusual NFS client setting the size and group at the same time.

Quoted 7 lines
Quoted 5 lines

To handle this issue properly this splits the notify_change call in nfsd_setattr into two separate ones.

Thanks.--b.