summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorTrond Myklebust <trond.myklebust@hammerspace.com>2025-04-28 01:21:06 +0300
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>2025-06-04 15:45:10 +0300
commitb6b3d8aeab9e6e8384c71908ff46f0f3c5b997d9 (patch)
tree48f308d2fd81a7ea0f1b7a262bc108d957cd3f2b
parente8fd089329f1e353d7b65894084596071c8a1263 (diff)
downloadlinux-b6b3d8aeab9e6e8384c71908ff46f0f3c5b997d9.tar.xz
NFS: Avoid flushing data while holding directory locks in nfs_rename()
[ Upstream commit dcd21b609d4abc7303f8683bce4f35d78d7d6830 ] The Linux client assumes that all filehandles are non-volatile for renames within the same directory (otherwise sillyrename cannot work). However, the existence of the Linux 'subtree_check' export option has meant that nfs_rename() has always assumed it needs to flush writes before attempting to rename. Since NFSv4 does allow the client to query whether or not the server exhibits this behaviour, and since knfsd does actually set the appropriate flag when 'subtree_check' is enabled on an export, it should be OK to optimise away the write flushing behaviour in the cases where it is clearly not needed. Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
-rw-r--r--fs/nfs/client.c2
-rw-r--r--fs/nfs/dir.c15
-rw-r--r--include/linux/nfs_fs_sb.h12
3 files changed, 25 insertions, 4 deletions
diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index 3b0918ade53c..a10d39150abc 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -1100,6 +1100,8 @@ struct nfs_server *nfs_create_server(struct fs_context *fc)
if (server->namelen == 0 || server->namelen > NFS2_MAXNAMLEN)
server->namelen = NFS2_MAXNAMLEN;
}
+ /* Linux 'subtree_check' borkenness mandates this setting */
+ server->fh_expire_type = NFS_FH_VOL_RENAME;
if (!(fattr->valid & NFS_ATTR_FATTR)) {
error = ctx->nfs_mod->rpc_ops->getattr(server, ctx->mntfh,
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index 2b04038b0e40..34f3471ce813 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -2678,6 +2678,18 @@ nfs_unblock_rename(struct rpc_task *task, struct nfs_renamedata *data)
unblock_revalidate(new_dentry);
}
+static bool nfs_rename_is_unsafe_cross_dir(struct dentry *old_dentry,
+ struct dentry *new_dentry)
+{
+ struct nfs_server *server = NFS_SB(old_dentry->d_sb);
+
+ if (old_dentry->d_parent != new_dentry->d_parent)
+ return false;
+ if (server->fh_expire_type & NFS_FH_RENAME_UNSAFE)
+ return !(server->fh_expire_type & NFS_FH_NOEXPIRE_WITH_OPEN);
+ return true;
+}
+
/*
* RENAME
* FIXME: Some nfsds, like the Linux user space nfsd, may generate a
@@ -2765,7 +2777,8 @@ int nfs_rename(struct mnt_idmap *idmap, struct inode *old_dir,
}
- if (S_ISREG(old_inode->i_mode))
+ if (S_ISREG(old_inode->i_mode) &&
+ nfs_rename_is_unsafe_cross_dir(old_dentry, new_dentry))
nfs_sync_inode(old_inode);
task = nfs_async_rename(old_dir, new_dir, old_dentry, new_dentry,
must_unblock ? nfs_unblock_rename : NULL);
diff --git a/include/linux/nfs_fs_sb.h b/include/linux/nfs_fs_sb.h
index 108862d81b57..8baaad2dfbe4 100644
--- a/include/linux/nfs_fs_sb.h
+++ b/include/linux/nfs_fs_sb.h
@@ -210,6 +210,15 @@ struct nfs_server {
char *fscache_uniq; /* Uniquifier (or NULL) */
#endif
+ /* The following #defines numerically match the NFSv4 equivalents */
+#define NFS_FH_NOEXPIRE_WITH_OPEN (0x1)
+#define NFS_FH_VOLATILE_ANY (0x2)
+#define NFS_FH_VOL_MIGRATION (0x4)
+#define NFS_FH_VOL_RENAME (0x8)
+#define NFS_FH_RENAME_UNSAFE (NFS_FH_VOLATILE_ANY | NFS_FH_VOL_RENAME)
+ u32 fh_expire_type; /* V4 bitmask representing file
+ handle volatility type for
+ this filesystem */
u32 pnfs_blksize; /* layout_blksize attr */
#if IS_ENABLED(CONFIG_NFS_V4)
u32 attr_bitmask[3];/* V4 bitmask representing the set
@@ -233,9 +242,6 @@ struct nfs_server {
u32 acl_bitmask; /* V4 bitmask representing the ACEs
that are supported on this
filesystem */
- u32 fh_expire_type; /* V4 bitmask representing file
- handle volatility type for
- this filesystem */
struct pnfs_layoutdriver_type *pnfs_curr_ld; /* Active layout driver */
struct rpc_wait_queue roc_rpcwaitq;
void *pnfs_ld_data; /* per mount point data */