With two clients, each with NFSv3 mounts of the same directory, the sequence:
client1 client2
ls -l afile
echo hello there > afile
echo HELLO > afile
cat afile
will show
HELLO
there
because the O_TRUNC requested in the final 'echo' doesn't take effect.
This is because the "Negative dentry, just create a file" section in
lookup_open() assumes that the file *does* get created since the dentry
was negative, so it sets FMODE_CREATED, and this causes do_open() to
clear O_TRUNC and so the file doesn't get truncated.
Even mounting with -o lookupcache=none does not help as
nfs_neg_need_reval() always returns false if LOOKUP_CREATE is set.
This patch fixes the problem by providing an atomic_open inode operation
for NFSv3 (and v2). The code is largely the code from the branch in
lookup_open() when atomic_open is not provided. The significant change
is that the O_TRUNC flag is passed a new nfs_do_create() which add
'trunc' handling to nfs_create().
With this change we also optimise away an unnecessary LOOKUP before the
file is created.
Signed-off-by: NeilBrown <[email protected]>
---
fs/nfs/dir.c | 54 +++++++++++++++++++++++++++++++++++++++---
fs/nfs/nfs3proc.c | 1 +
fs/nfs/proc.c | 1 +
include/linux/nfs_fs.h | 3 +++
4 files changed, 56 insertions(+), 3 deletions(-)
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
index ac505671efbd..342930996226 100644
--- a/fs/nfs/dir.c
+++ b/fs/nfs/dir.c
@@ -56,6 +56,8 @@ static int nfs_readdir(struct file *, struct dir_context *);
static int nfs_fsync_dir(struct file *, loff_t, loff_t, int);
static loff_t nfs_llseek_dir(struct file *, loff_t, int);
static void nfs_readdir_clear_array(struct folio *);
+static int nfs_do_create(struct inode *dir, struct dentry *dentry,
+ umode_t mode, int open_flags);
const struct file_operations nfs_dir_operations = {
.llseek = nfs_llseek_dir,
@@ -2243,6 +2245,41 @@ static int nfs4_lookup_revalidate(struct dentry *dentry, unsigned int flags)
#endif /* CONFIG_NFSV4 */
+int nfs_atomic_open_v23(struct inode *dir, struct dentry *dentry,
+ struct file *file, unsigned int open_flags,
+ umode_t mode)
+{
+
+ /* Same as look+open from lookup_open(), but with different O_TRUNC
+ * handling.
+ */
+ int error = 0;
+
+ if (open_flags & O_CREAT) {
+ file->f_mode |= FMODE_CREATED;
+ error = nfs_do_create(dir, dentry, mode, open_flags);
+ if (error)
+ return error;
+ return finish_open(file, dentry, NULL);
+ } else if (d_in_lookup(dentry)) {
+ /* The only flags nfs_lookup considers are
+ * LOOKUP_EXCL and LOOKUP_RENAME_TARGET, and
+ * we want those to be zero so the lookup isn't skipped.
+ */
+ struct dentry *res = nfs_lookup(dir, dentry, 0);
+
+ d_lookup_done(dentry);
+ if (unlikely(res)) {
+ if (IS_ERR(res))
+ return PTR_ERR(res);
+ return finish_no_open(file, res);
+ }
+ }
+ return finish_no_open(file, NULL);
+
+}
+EXPORT_SYMBOL_GPL(nfs_atomic_open_v23);
+
struct dentry *
nfs_add_or_obtain(struct dentry *dentry, struct nfs_fh *fhandle,
struct nfs_fattr *fattr)
@@ -2303,18 +2340,23 @@ EXPORT_SYMBOL_GPL(nfs_instantiate);
* that the operation succeeded on the server, but an error in the
* reply path made it appear to have failed.
*/
-int nfs_create(struct mnt_idmap *idmap, struct inode *dir,
- struct dentry *dentry, umode_t mode, bool excl)
+static int nfs_do_create(struct inode *dir, struct dentry *dentry,
+ umode_t mode, int open_flags)
{
struct iattr attr;
- int open_flags = excl ? O_CREAT | O_EXCL : O_CREAT;
int error;
+ open_flags |= O_CREAT;
+
dfprintk(VFS, "NFS: create(%s/%lu), %pd\n",
dir->i_sb->s_id, dir->i_ino, dentry);
attr.ia_mode = mode;
attr.ia_valid = ATTR_MODE;
+ if (open_flags & O_TRUNC) {
+ attr.ia_size = 0;
+ attr.ia_valid |= ATTR_SIZE;
+ }
trace_nfs_create_enter(dir, dentry, open_flags);
error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags);
@@ -2326,6 +2368,12 @@ int nfs_create(struct mnt_idmap *idmap, struct inode *dir,
d_drop(dentry);
return error;
}
+
+int nfs_create(struct mnt_idmap *idmap, struct inode *dir,
+ struct dentry *dentry, umode_t mode, bool excl)
+{
+ return nfs_do_create(dir, dentry, mode, excl ? O_EXCL : 0);
+}
EXPORT_SYMBOL_GPL(nfs_create);
/*
diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c
index cbbe3f0193b8..74bda639a7cf 100644
--- a/fs/nfs/nfs3proc.c
+++ b/fs/nfs/nfs3proc.c
@@ -986,6 +986,7 @@ static int nfs3_have_delegation(struct inode *inode, fmode_t flags)
static const struct inode_operations nfs3_dir_inode_operations = {
.create = nfs_create,
+ .atomic_open = nfs_atomic_open_v23,
.lookup = nfs_lookup,
.link = nfs_link,
.unlink = nfs_unlink,
diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c
index ad3a321ae997..d105e5b2659d 100644
--- a/fs/nfs/proc.c
+++ b/fs/nfs/proc.c
@@ -695,6 +695,7 @@ static int nfs_have_delegation(struct inode *inode, fmode_t flags)
static const struct inode_operations nfs_dir_inode_operations = {
.create = nfs_create,
.lookup = nfs_lookup,
+ .atomic_open = nfs_atomic_open_v23,
.link = nfs_link,
.unlink = nfs_unlink,
.symlink = nfs_symlink,
diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
index d59116ac8209..039898d70954 100644
--- a/include/linux/nfs_fs.h
+++ b/include/linux/nfs_fs.h
@@ -561,6 +561,9 @@ extern int nfs_may_open(struct inode *inode, const struct cred *cred, int openfl
extern void nfs_access_zap_cache(struct inode *inode);
extern int nfs_access_get_cached(struct inode *inode, const struct cred *cred,
u32 *mask, bool may_block);
+extern int nfs_atomic_open_v23(struct inode *dir, struct dentry *dentry,
+ struct file *file, unsigned int open_flags,
+ umode_t mode);
/*
* linux/fs/nfs/symlink.c
--
2.44.0
Hi,
any thoughts about this patch and/or the problem it addresses?
Thanks,
NeilBrown
On Mon, 25 Mar 2024, NeilBrown wrote:
> With two clients, each with NFSv3 mounts of the same directory, the sequence:
>
> client1 client2
> ls -l afile
> echo hello there > afile
> echo HELLO > afile
> cat afile
>
> will show
> HELLO
> there
>
> because the O_TRUNC requested in the final 'echo' doesn't take effect.
> This is because the "Negative dentry, just create a file" section in
> lookup_open() assumes that the file *does* get created since the dentry
> was negative, so it sets FMODE_CREATED, and this causes do_open() to
> clear O_TRUNC and so the file doesn't get truncated.
>
> Even mounting with -o lookupcache=none does not help as
> nfs_neg_need_reval() always returns false if LOOKUP_CREATE is set.
>
> This patch fixes the problem by providing an atomic_open inode operation
> for NFSv3 (and v2). The code is largely the code from the branch in
> lookup_open() when atomic_open is not provided. The significant change
> is that the O_TRUNC flag is passed a new nfs_do_create() which add
> 'trunc' handling to nfs_create().
>
> With this change we also optimise away an unnecessary LOOKUP before the
> file is created.
>
> Signed-off-by: NeilBrown <[email protected]>
> ---
> fs/nfs/dir.c | 54 +++++++++++++++++++++++++++++++++++++++---
> fs/nfs/nfs3proc.c | 1 +
> fs/nfs/proc.c | 1 +
> include/linux/nfs_fs.h | 3 +++
> 4 files changed, 56 insertions(+), 3 deletions(-)
>
> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> index ac505671efbd..342930996226 100644
> --- a/fs/nfs/dir.c
> +++ b/fs/nfs/dir.c
> @@ -56,6 +56,8 @@ static int nfs_readdir(struct file *, struct dir_context *);
> static int nfs_fsync_dir(struct file *, loff_t, loff_t, int);
> static loff_t nfs_llseek_dir(struct file *, loff_t, int);
> static void nfs_readdir_clear_array(struct folio *);
> +static int nfs_do_create(struct inode *dir, struct dentry *dentry,
> + umode_t mode, int open_flags);
>
> const struct file_operations nfs_dir_operations = {
> .llseek = nfs_llseek_dir,
> @@ -2243,6 +2245,41 @@ static int nfs4_lookup_revalidate(struct dentry *dentry, unsigned int flags)
>
> #endif /* CONFIG_NFSV4 */
>
> +int nfs_atomic_open_v23(struct inode *dir, struct dentry *dentry,
> + struct file *file, unsigned int open_flags,
> + umode_t mode)
> +{
> +
> + /* Same as look+open from lookup_open(), but with different O_TRUNC
> + * handling.
> + */
> + int error = 0;
> +
> + if (open_flags & O_CREAT) {
> + file->f_mode |= FMODE_CREATED;
> + error = nfs_do_create(dir, dentry, mode, open_flags);
> + if (error)
> + return error;
> + return finish_open(file, dentry, NULL);
> + } else if (d_in_lookup(dentry)) {
> + /* The only flags nfs_lookup considers are
> + * LOOKUP_EXCL and LOOKUP_RENAME_TARGET, and
> + * we want those to be zero so the lookup isn't skipped.
> + */
> + struct dentry *res = nfs_lookup(dir, dentry, 0);
> +
> + d_lookup_done(dentry);
> + if (unlikely(res)) {
> + if (IS_ERR(res))
> + return PTR_ERR(res);
> + return finish_no_open(file, res);
> + }
> + }
> + return finish_no_open(file, NULL);
> +
> +}
> +EXPORT_SYMBOL_GPL(nfs_atomic_open_v23);
> +
> struct dentry *
> nfs_add_or_obtain(struct dentry *dentry, struct nfs_fh *fhandle,
> struct nfs_fattr *fattr)
> @@ -2303,18 +2340,23 @@ EXPORT_SYMBOL_GPL(nfs_instantiate);
> * that the operation succeeded on the server, but an error in the
> * reply path made it appear to have failed.
> */
> -int nfs_create(struct mnt_idmap *idmap, struct inode *dir,
> - struct dentry *dentry, umode_t mode, bool excl)
> +static int nfs_do_create(struct inode *dir, struct dentry *dentry,
> + umode_t mode, int open_flags)
> {
> struct iattr attr;
> - int open_flags = excl ? O_CREAT | O_EXCL : O_CREAT;
> int error;
>
> + open_flags |= O_CREAT;
> +
> dfprintk(VFS, "NFS: create(%s/%lu), %pd\n",
> dir->i_sb->s_id, dir->i_ino, dentry);
>
> attr.ia_mode = mode;
> attr.ia_valid = ATTR_MODE;
> + if (open_flags & O_TRUNC) {
> + attr.ia_size = 0;
> + attr.ia_valid |= ATTR_SIZE;
> + }
>
> trace_nfs_create_enter(dir, dentry, open_flags);
> error = NFS_PROTO(dir)->create(dir, dentry, &attr, open_flags);
> @@ -2326,6 +2368,12 @@ int nfs_create(struct mnt_idmap *idmap, struct inode *dir,
> d_drop(dentry);
> return error;
> }
> +
> +int nfs_create(struct mnt_idmap *idmap, struct inode *dir,
> + struct dentry *dentry, umode_t mode, bool excl)
> +{
> + return nfs_do_create(dir, dentry, mode, excl ? O_EXCL : 0);
> +}
> EXPORT_SYMBOL_GPL(nfs_create);
>
> /*
> diff --git a/fs/nfs/nfs3proc.c b/fs/nfs/nfs3proc.c
> index cbbe3f0193b8..74bda639a7cf 100644
> --- a/fs/nfs/nfs3proc.c
> +++ b/fs/nfs/nfs3proc.c
> @@ -986,6 +986,7 @@ static int nfs3_have_delegation(struct inode *inode, fmode_t flags)
>
> static const struct inode_operations nfs3_dir_inode_operations = {
> .create = nfs_create,
> + .atomic_open = nfs_atomic_open_v23,
> .lookup = nfs_lookup,
> .link = nfs_link,
> .unlink = nfs_unlink,
> diff --git a/fs/nfs/proc.c b/fs/nfs/proc.c
> index ad3a321ae997..d105e5b2659d 100644
> --- a/fs/nfs/proc.c
> +++ b/fs/nfs/proc.c
> @@ -695,6 +695,7 @@ static int nfs_have_delegation(struct inode *inode, fmode_t flags)
> static const struct inode_operations nfs_dir_inode_operations = {
> .create = nfs_create,
> .lookup = nfs_lookup,
> + .atomic_open = nfs_atomic_open_v23,
> .link = nfs_link,
> .unlink = nfs_unlink,
> .symlink = nfs_symlink,
> diff --git a/include/linux/nfs_fs.h b/include/linux/nfs_fs.h
> index d59116ac8209..039898d70954 100644
> --- a/include/linux/nfs_fs.h
> +++ b/include/linux/nfs_fs.h
> @@ -561,6 +561,9 @@ extern int nfs_may_open(struct inode *inode, const struct cred *cred, int openfl
> extern void nfs_access_zap_cache(struct inode *inode);
> extern int nfs_access_get_cached(struct inode *inode, const struct cred *cred,
> u32 *mask, bool may_block);
> +extern int nfs_atomic_open_v23(struct inode *dir, struct dentry *dentry,
> + struct file *file, unsigned int open_flags,
> + umode_t mode);
>
> /*
> * linux/fs/nfs/symlink.c
> --
> 2.44.0
>
>
>