On Wed, 21 Sept 2022 at 10:27, Christian Brauner <brauner@xxxxxxxxxx> wrote:
>
> On Tue, Sep 20, 2022 at 09:36:27PM +0200, Miklos Szeredi wrote:
> > Use the tmpfile_open() helper instead of doing tmpfile creation and opening
> > separately.
> >
> > Signed-off-by: Miklos Szeredi <mszeredi@xxxxxxxxxx>
> > ---
> > fs/cachefiles/namei.c | 26 ++++++++++----------------
> > 1 file changed, 10 insertions(+), 16 deletions(-)
> >
> > diff --git a/fs/cachefiles/namei.c b/fs/cachefiles/namei.c
> > index d3a5884fe5c9..44f575328af4 100644
> > --- a/fs/cachefiles/namei.c
> > +++ b/fs/cachefiles/namei.c
> > @@ -451,18 +451,19 @@ struct file *cachefiles_create_tmpfile(struct cachefiles_object *object)
> > const struct cred *saved_cred;
> > struct dentry *fan = volume->fanout[(u8)object->cookie->key_hash];
> > struct file *file;
> > - struct path path;
> > + struct path path = { .mnt = cache->mnt, .dentry = fan };
> > uint64_t ni_size;
> > long ret;
>
> Maybe we shouldn't use struct path to first refer to the parent path and
> then to the tmp path to avoid any potential confusion and instead rely
> on a compount initializer for the tmpfile_open() call (__not tested__)?:
>
> diff --git a/fs/cachefiles/namei.c b/fs/cachefiles/namei.c
> index 44f575328af4..979b2f173ac3 100644
> --- a/fs/cachefiles/namei.c
> +++ b/fs/cachefiles/namei.c
> @@ -451,7 +451,7 @@ struct file *cachefiles_create_tmpfile(struct cachefiles_object *object)
> const struct cred *saved_cred;
> struct dentry *fan = volume->fanout[(u8)object->cookie->key_hash];
> struct file *file;
> - struct path path = { .mnt = cache->mnt, .dentry = fan };
> + struct path path;
> uint64_t ni_size;
> long ret;
>
> @@ -460,8 +460,10 @@ struct file *cachefiles_create_tmpfile(struct cachefiles_object *object)
>
> ret = cachefiles_inject_write_error();
> if (ret == 0) {
> - file = tmpfile_open(&init_user_ns, &path, S_IFREG,
> - O_RDWR | O_LARGEFILE | O_DIRECT,
> + file = tmpfile_open(&init_user_ns,
> + &{const struct path} {.mnt = cache->mnt,
> + .dentry = fan},
This doesn't look nice. I fixed it with a separate "parentpath" variable.
Thanks,
Miklos
