Re: XFS reports lchmod failure, but changes file system contents

[Al Viro <viro at zeniv dot linux dot org dot uk>]

On Wed, Feb 12, 2020 at 03:19:51PM -0500, Rich Felker wrote:
> On Wed, Feb 12, 2020 at 09:17:41PM +0100, Andreas Schwab wrote:
> > On Feb 12 2020, Florian Weimer wrote:
> > 
> > > * Al Viro:
> > >
> > >> On Wed, Feb 12, 2020 at 08:15:08PM +0100, Florian Weimer wrote:
> > >>
> > >>> | Further, I've found some inconsistent behavior with ext4: chmod on the
> > >>> | magic symlink fails with EOPNOTSUPP as in Florian's test, but fchmod
> > >>> | on the O_PATH fd succeeds and changes the symlink mode. This is with
> > >>> | 5.4. Cany anyone else confirm this? Is it a problem?
> > >>> 
> > >>> It looks broken to me because fchmod (as an inode-changing operation)
> > >>> is not supposed to work on O_PATH descriptors.
> > >>
> > >> Why?  O_PATH does have an associated inode just fine; where does
> > >> that "not supposed to" come from?
> > >
> > > It fails on most file systems right now.  I thought that was expected.
> > > Other system calls (fsetxattr IIRC) do not work on O_PATH descriptors,
> > > either.  I assumed that an O_PATH descriptor was not intending to
> > > confer that capability.  Even openat fails.
> > 
> > According to open(2), this is expected:
> > 
> >        O_PATH (since Linux 2.6.39)
> >               Obtain a file descriptor that can be used for two  purposes:  to
> >               indicate a location in the filesystem tree and to perform opera-
> >               tions that act purely at the file descriptor  level.   The  file
> >               itself  is not opened, and other file operations (e.g., read(2),
> >               write(2), fchmod(2), fchown(2), fgetxattr(2), ioctl(2), mmap(2))
> >               fail with the error EBADF.
> 
> That text is outdated and should be corrected. Fixing fchmod fchown,
> fstat, etc. to operate on O_PATH file descriptors was a very
> intentional change in the kernel.

Wait.  First of all, in the testcase it's chmod(2) applied to /proc/*/fd/*; that's
no different for O_PATH descriptors.  Location in the tree *is* associated with
O_PATH fd; that's the only thing they exist for.

fchmod(2) will certainly fail for those, as it always had:
int ksys_fchmod(unsigned int fd, umode_t mode)
{
        struct fd f = fdget(fd);
        int err = -EBADF;

        if (f.file) {
                audit_file(f.file);
                err = chmod_common(&f.file->f_path, mode);
                fdput(f);  
        }
        return err;
}

SYSCALL_DEFINE2(fchmod, unsigned int, fd, umode_t, mode)
{
        return ksys_fchmod(fd, mode);
}

... and that fdget() will give you -EBADF.  If you've managed to get
fchmod(2) the syscall to give you anything other than that, I want
to see details.