NFS: Add attribute update barriers to nfs_setattr_update_inode()

commit f044636d972246d451e06226cc1675d5da389762 upstream.

Ensure that other operations which raced with our setattr RPC call
cannot revert the file attribute changes that were made on the server.
To do so, we artificially bump the attribute generation counter on
the inode so that all calls to nfs_fattr_init() that precede ours
will be dropped.

The motivation for the patch came from Chuck Lever's reports of readaheads
racing with truncate operations and causing the file size to be reverted.

Reported-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Tested-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: NeilBrown <neilb@suse.de>
Signed-off-by: Jiri Slaby <jslaby@suse.cz>

4 files changed, 17 insertions, 10 deletions

diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index e5eb677ca9ce..127a6d9d81b7 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -545,6 +545,7 @@ EXPORT_SYMBOL_GPL(nfs_setattr);
  * This is a copy of the common vmtruncate, but with the locking
  * corrected to take into account the fact that NFS requires
  * inode->i_size to be updated under the inode->i_lock.
+ * Note: must be called with inode->i_lock held!
  */
 static int nfs_vmtruncate(struct inode * inode, loff_t offset)
 {
@@ -554,11 +555,11 @@ static int nfs_vmtruncate(struct inode * inode, loff_t offset)
 	if (err)
 		goto out;
 
-	spin_lock(&inode->i_lock);
 	i_size_write(inode, offset);
-	spin_unlock(&inode->i_lock);
 
+	spin_unlock(&inode->i_lock);
 	truncate_pagecache(inode, offset);
+	spin_lock(&inode->i_lock);
 out:
 	return err;
 }
@@ -571,10 +572,15 @@ out:
  * Note: we do this in the *proc.c in order to ensure that
  *       it works for things like exclusive creates too.
  */
-void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
+void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr,
+		struct nfs_fattr *fattr)
 {
+	/* Barrier: bump the attribute generation count. */
+	fattr->gencount = nfs_inc_attr_generation_counter();
+
+	spin_lock(&inode->i_lock);
+	NFS_I(inode)->attr_gencount = fattr->gencount;
 	if ((attr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0) {
-		spin_lock(&inode->i_lock);
 		if ((attr->ia_valid & ATTR_MODE) != 0) {
 			int mode = attr->ia_mode & S_IALLUGO;
 			mode |= inode->i_mode & ~S_IALLUGO;
@@ -585,12 +591,13 @@ void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr)
 		if ((attr->ia_valid & ATTR_GID) != 0)
 			inode->i_gid = attr->ia_gid;
 		NFS_I(inode)->cache_validity |= NFS_INO_INVALID_ACCESS|NFS_INO_INVALID_ACL;
-		spin_unlock(&inode->i_lock);
 	}
 	if ((attr->ia_valid & ATTR_SIZE) != 0) {
 		nfs_inc_stats(inode, NFSIOS_SETATTRTRUNC);
 		nfs_vmtruncate(inode, attr->ia_size);
 	}
+	nfs_update_inode(inode, fattr);
+	spin_unlock(&inode->i_lock);
 }
 EXPORT_SYMBOL_GPL(nfs_setattr_update_inode);
 
diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c
index 90cb10d7b693..8ef5276776f9 100644
--- a/fs/nfs/nfs3proc.c
+++ b/fs/nfs/nfs3proc.c
@@ -136,7 +136,7 @@ nfs3_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
 	nfs_fattr_init(fattr);
 	status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
 	if (status == 0)
-		nfs_setattr_update_inode(inode, sattr);
+		nfs_setattr_update_inode(inode, sattr, fattr);
 	dprintk("NFS reply setattr: %d\n", status);
 	return status;
 }
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 43c27110387a..36a72b59d7c8 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -2276,8 +2276,8 @@ static int _nfs4_do_open(struct inode *dir,
 				opendata->o_res.f_attr, sattr,
 				state, label, olabel);
 		if (status == 0) {
-			nfs_setattr_update_inode(state->inode, sattr);
-			nfs_post_op_update_inode(state->inode, opendata->o_res.f_attr);
+			nfs_setattr_update_inode(state->inode, sattr,
+					opendata->o_res.f_attr);
 			nfs_setsecurity(state->inode, opendata->o_res.f_attr, olabel);
 		}
 	}
@@ -3114,7 +3114,7 @@ nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
 
 	status = nfs4_do_setattr(inode, cred, fattr, sattr, state, NULL, label);
 	if (status == 0) {
-		nfs_setattr_update_inode(inode, sattr);
+		nfs_setattr_update_inode(inode, sattr, fattr);
 		nfs_setsecurity(inode, fattr, label);
 	}
 	nfs4_label_free(label);
diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c
index a8f57c728df5..7ffaeffc1330 100644
--- a/fs/nfs/proc.c
+++ b/fs/nfs/proc.c
@@ -139,7 +139,7 @@ nfs_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
 	nfs_fattr_init(fattr);
 	status = rpc_call_sync(NFS_CLIENT(inode), &msg, 0);
 	if (status == 0)
-		nfs_setattr_update_inode(inode, sattr);
+		nfs_setattr_update_inode(inode, sattr, fattr);
 	dprintk("NFS reply setattr: %d\n", status);
 	return status;
 }