linux-kernel - Re: Is lockref_get_not_zero() really correct in dget

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <87ha1qwnky.fsf@x220.int.ebiederm.org>
Date:	Tue, 05 Aug 2014 15:58:37 -0700
From:	ebiederm@...ssion.com (Eric W. Biederman)
To:	Al Viro <viro@...IV.linux.org.uk>
Cc:	Linus Torvalds <torvalds@...ux-foundation.org>,
	NeilBrown <neilb@...e.de>,
	linux-fsdevel <linux-fsdevel@...r.kernel.org>,
	lkml <linux-kernel@...r.kernel.org>
Subject: Re: Is lockref_get_not_zero() really correct in dget_parent()

Al Viro <viro@...IV.linux.org.uk> writes:

> Speaking of bogosities, apparently nobody has noticed that automagical
> turning BSD process accounting off upon r/o remounts of the filesystem
> hosting the log got broken several years ago.  Suppose we find an opened
> log when remounting.  OK, it gets closed, which means filp_close(...).
> Which means that actual closing that sucker will happen just before we
> return to userland.  Return with -EBUSY, since the filesystem still has
> a file opened for write when we get around to checking that...
>
> I have kinda-sorta fixes for that (basically, restoring the situation we
> used to have before delayed fput() and being damn careful about avoiding
> deadlocks), but I would really love to just rip that idiocy out.  IOW,
> just let it fail with -EBUSY in such cases, same as for any other write-opened
> file on that fs.
>
> That, BTW, is the most painful part of the acct series, so being able to
> drop it would be very nice.  Comments?

Long ago in this discucssion someone mentioned that fedora and related
distro's don't have (or at least didn't have) code to close the
accounting files before unmount during shutdown.  I just checked the
debain acct package and it very clearly turns off accounting before
unmount (so debian should not care).

So the practical test is will returning -EBUSY cause fedora based
distro's it fail to unmount filesystems cleanly (looking at the code
and postualting bsd process accounting being left on) I think such
a change will cause such systems to fail to unmount their filesystems
on shutdown/reboot and cause problems because of that.

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/