On Mon, Nov 9, 2020 at 8:37 PM Miklos Szeredi <miklos@xxxxxxxxxx> wrote:
>
> On Mon, Nov 9, 2020 at 11:04 AM Chirantan Ekbote <chirantan@xxxxxxxxxxxx> wrote:
> >
> > Implement support for O_TMPFILE by re-using the existing infrastructure
> > for mkdir, symlink, mknod, etc. The server should reply to the tmpfile
> > request by sending a fuse_entry_out describing the newly created
> > tmpfile.
> >
> > Signed-off-by: Chirantan Ekbote <chirantan@xxxxxxxxxxxx>
> > ---
> > fs/fuse/dir.c | 21 +++++++++++++++++++++
> > fs/fuse/file.c | 3 ++-
> > 2 files changed, 23 insertions(+), 1 deletion(-)
> >
> > diff --git a/fs/fuse/dir.c b/fs/fuse/dir.c
> > index ff7dbeb16f88d..1ab52e7ec1625 100644
> > --- a/fs/fuse/dir.c
> > +++ b/fs/fuse/dir.c
> > @@ -751,6 +751,26 @@ static int fuse_mkdir(struct inode *dir, struct dentry *entry, umode_t mode)
> > return create_new_entry(fm, &args, dir, entry, S_IFDIR);
> > }
> >
> > +static int fuse_tmpfile(struct inode *dir, struct dentry *entry, umode_t mode)
> > +{
> > + struct fuse_tmpfile_in inarg;
> > + struct fuse_mount *fm = get_fuse_mount(dir);
> > + FUSE_ARGS(args);
> > +
> > + if (!fm->fc->dont_mask)
> > + mode &= ~current_umask();
> > +
> > + memset(&inarg, 0, sizeof(inarg));
> > + inarg.mode = mode;
> > + inarg.umask = current_umask();
> > + args.opcode = FUSE_TMPFILE;
> > + args.in_numargs = 1;
> > + args.in_args[0].size = sizeof(inarg);
> > + args.in_args[0].value = &inarg;
> > +
> > + return create_new_entry(fm, &args, dir, entry, S_IFREG);
> > +}
> > +
> > static int fuse_symlink(struct inode *dir, struct dentry *entry,
> > const char *link)
> > {
> > @@ -1818,6 +1838,7 @@ static const struct inode_operations fuse_dir_inode_operations = {
> > .listxattr = fuse_listxattr,
> > .get_acl = fuse_get_acl,
> > .set_acl = fuse_set_acl,
> > + .tmpfile = fuse_tmpfile,
> > };
> >
> > static const struct file_operations fuse_dir_operations = {
> > diff --git a/fs/fuse/file.c b/fs/fuse/file.c
> > index c03034e8c1529..8ecf85699a014 100644
> > --- a/fs/fuse/file.c
> > +++ b/fs/fuse/file.c
> > @@ -39,7 +39,8 @@ static int fuse_send_open(struct fuse_mount *fm, u64 nodeid, struct file *file,
> > FUSE_ARGS(args);
> >
> > memset(&inarg, 0, sizeof(inarg));
> > - inarg.flags = file->f_flags & ~(O_CREAT | O_EXCL | O_NOCTTY);
> > + inarg.flags =
> > + file->f_flags & ~(O_CREAT | O_EXCL | O_NOCTTY | O_TMPFILE);
>
> Why did you add this? At this stage O_TMPFILE should not be in the
> flags, since that case was handled via the ->tmpfile() path.
>
That's not the behavior I observed. Without this, the O_TMPFILE flag
gets passed through to the server. The call stack is:
- do_filp_open
- path_openat
- do_tmpfile
- vfs_tmpfile
- dir->i_op->tmpfile
- finish_open
- do_dentry_open
- f->f_op->open
and I didn't see O_TMPFILE being removed anywhere in there.
Chirantan
