svn commit: r361238 - head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs

[Kyle Evans]

On Tue, May 19, 2020 at 10:27 AM Rodney W. Grimes
<freebsd at gndrsh.dnsmgr.net> wrote:
>
> > On Tue, May 19, 2020 at 10:23 AM Rodney W. Grimes
> > <freebsd at gndrsh.dnsmgr.net> wrote:
> > >
> > > > Author: kevans
> > > > Date: Tue May 19 02:41:05 2020
> > > > New Revision: 361238
> > > > URL: https://svnweb.freebsd.org/changeset/base/361238
> > > >
> > > > Log:
> > > >   zfs: reject read(2) of a dirfd with EISDIR
> > > >
> > > >   This is independent of the recently-discussed global change, which is still
> > > >   in review/discussion stage.
> > > >
> > > >   This is effectively a measure for consistency in the ZFS world, where
> > > >   FreeBSD was the only platform (as far as I could find) that allowed this.
> > > >   What ZFS exposes is decidedly not useful for any real purposes, to
> > > >   paraphrase (hopefully faithfully) jhb's findings when exploring this:
> > > >
> > > >   The size of a directory in ZFS is the number of directory entries within.
> > > >   When reading a directory, you would instead get the leading part of its raw
> > > >   contents; the amount you get being dictated by the "size," i.e. number of
> > > >   directory entries. There's decidedly (luckily) no stack disclosure happening
> > > >   here, though the behavior is bizarre and almost certainly a historical
> > > >   accident.
> > > >
> > > >   This change has already been upstreamed to OpenZFS.
> > >
> > > Until the grep -d skip issue is addressed I object to this change as
> > > it is going to cause people who do grep with wildcards to see lots
> > > of errors that before where pretty much either silent (no match occured)
> > > or spit out a "binary file foo matches."
> > >
> >
> > That seems preferable to grepping random bytes that don't particularly
> > contain any strings? They'd never see "binary file foo matches" in
> > this case.
>
> The difference is you rarely get a hit, and now your gauranteed to
> get a hit on every single directory making grep * very noisy, where
> it was often silent or nearly silent before.
>

As you noted in the review for the larger change, -d skip is a good
option for the people that don't like this. It probably makes sense as
a default, but then we'd be diverging from the other popular grep that
defaults to -d read and spews out EISDIR more often than not.

> >
> > This isn't exactly divergent from the behavior they'd see with ZFS
> > anywhere else.
>
> It is extremly divergent from 42 years of behavior.
>

I don't think ZFS has been implemented on FreeBSD for 42 years, and I
don't find this grep argument compelling enough to restore peoples'
ability to read the raw znode of a directory.