On Sun, May 24, 2020 at 06:56:54AM -0400, Greg Troxel wrote:
> Yorick Hardy <yorickha...@gmail.com> writes:
>
> (I realize you later say this isn't it.)
>
> >> @@ -4141,10 +4140,6 @@ sys_fdatasync(struct lwp *l, const struct
> >> sys_fdatasync_args *uap, register_t *r
> >> /* fd_getvnode() will use the descriptor for us */
> >> if ((error = fd_getvnode(SCARG(uap, fd), &fp)) != 0)
> >> return (error);
> >> - if ((fp->f_flag & FWRITE) == 0) {
> >> - fd_putfile(SCARG(uap, fd));
> >> - return (EBADF);
> >> - }
> >> vp = fp->f_vnode;
> >> vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
> >> error = VOP_FSYNC(vp, fp->f_cred, FSYNC_WAIT|FSYNC_DATAONLY, 0, 0);
>
> If you look at the function beyond what's in the diff, you will see (I
> think, but I really mean I see) that there is always a single
> fd_putfile. This was just doing the put before returning, rather than
> setting error and the usaul "goto out" where the end-of-routine cleanup
> happens. See also sys_fsync_range() in the same file.
>
> I could be reading this wrong.
I concur - this was just the fd_putfile to match the fd_getfile in
the early error path for read-only files. The code now falls through
and calls fd_putfile regardless, to remove the fd reference.
There should be no real behaviour change here apart from a relaxing
of the success conditions.
--
Paul Ripke
"Great minds discuss ideas, average minds discuss events, small minds
discuss people."
-- Disputed: Often attributed to Eleanor Roosevelt. 1948.
