linux-kernel - Re: [PATCH 0/1] devpts: use dynamic

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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20170824022436.44adb497@mir>
Date:   Thu, 24 Aug 2017 02:24:36 +0200
From:   Stefan Lippers-Hollmann <s.l-h@....de>
To:     Linus Torvalds <torvalds@...ux-foundation.org>
Cc:     "Eric W. Biederman" <ebiederm@...ssion.com>,
        Christian Brauner <christian.brauner@...onical.com>,
        Christian Brauner <christian.brauner@...ntu.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        "Serge E. Hallyn" <serge@...lyn.com>,
        Al Viro <viro@...iv.linux.org.uk>,
        Thorsten Leemhuis <regressions@...mhuis.info>
Subject: Re: [PATCH 0/1] devpts: use dynamic_dname() to generate proc name

Hi

On 2017-08-16, Linus Torvalds wrote:
> On Wed, Aug 16, 2017 at 4:51 PM, Eric W. Biederman
> <ebiederm@...ssion.com> wrote:
[...]
> Maybe this attached patch is better anyway. It's smaller, because it
> keeps more closely to the old code, and just adds a mntput() in all
> the exit cases, and depends on the "path_get()" to have incremented
> the mnt refcount one extra time.
> 
> Can you find something in this one?
> 
> ENTIRELY UNTESTED!

This patch[1] as part of 4.13-rc6 (up to, at least, 
v4.13-rc6-45-g6470812e2226) introduces a regression for me when using
pbuilder 0.228.7[2] (a helper to build Debian packages in a chroot and
to create and update its chroots) when trying to umount /dev/ptmx (inside
the chroot) on Debian/ unstable (full log and pbuilder configuration 
file[3] attached).

[...]
Setting up build-essential (12.3) ...
Processing triggers for libc-bin (2.24-15) ...
I: unmounting dev/ptmx filesystem
W: Could not unmount dev/ptmx: umount: /var/cache/pbuilder/build/1340/dev/ptmx: target is busy
        (In some cases useful info about processes that
         use the device is found by lsof(8) or fuser(1).)
W: Retrying to unmount dev/ptmx in 5s
umount: /var/cache/pbuilder/build/1340/dev/ptmx: target is busy
        (In some cases useful info about processes that
         use the device is found by lsof(8) or fuser(1).)

  Could not unmount dev/ptmx, some programs might
  still be using files in /proc (klogd?).
  Please check and kill these processes manually
  so that I can unmount dev/ptmx.  Last umount error was:
umount: /var/cache/pbuilder/build/1340/dev/ptmx: target is busy
        (In some cases useful info about processes that
         use the device is found by lsof(8) or fuser(1).)
[...]

lsof isn't revealing (but this might point towards gvfs 1.30.4-1+b1 
involvement), fuser -k doesn't release the ressource.

Kernel v4.13-rc5 and before (at least 4.11.x and 4.12.x) are not
affected, but this problem is reliably reproducible on three different 
x86_64 systems running Debian/ unstable when the host is running a 
kernel >=4.6.13-rc6. Unfortunately I haven't really found an easier/
smaller way to reproduce this issue yet, but creating a new build
chroot[4] always triggers this problem, updating an existing build
chroot[5] (which also mounts and umounts /dev/ptmx) triggers most of 
the time, but not reliably - building a package (e.g. the kernel, also 
mounting and umounting /dev/ptmx) also triggered this issue at least 
once (but I didn't try this more often).

My git bisection log is this:

$ git bisect log
git bisect start
# good: [ef954844c7ace62f773f4f23e28d2d915adc419f] Linux 4.13-rc5
git bisect good ef954844c7ace62f773f4f23e28d2d915adc419f
# bad: [14ccee78fc82f5512908f4424f541549a5705b89] Linux 4.13-rc6
git bisect bad 14ccee78fc82f5512908f4424f541549a5705b89
# bad: [cb247857f3dae0bdb843362c35027a0066b963a4] Merge tag 'sound-4.13-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
git bisect bad cb247857f3dae0bdb843362c35027a0066b963a4
# good: [88a5c690b66110ad255380d8f629c629cf6ca559] bpf: fix bpf_trace_printk on 32 bit archs
git bisect good 88a5c690b66110ad255380d8f629c629cf6ca559
# bad: [3bc6c906eacec34f0d8dcfd3c7e4513edf152297] Merge branch 'parisc-4.13-5' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux
git bisect bad 3bc6c906eacec34f0d8dcfd3c7e4513edf152297
# good: [40c6d1b9e2fc4251ca19fa69398f6fa34e813e27] Merge tag 'linux-kselftest-4.13-rc6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest
git bisect good 40c6d1b9e2fc4251ca19fa69398f6fa34e813e27
# good: [ac9a40905a610fb02086a37b11ff4bf046825a88] Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
git bisect good ac9a40905a610fb02086a37b11ff4bf046825a88
# bad: [99f781b1bfc199ec8eb86d4e015920faf79d5d57] Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs
git bisect bad 99f781b1bfc199ec8eb86d4e015920faf79d5d57
# good: [41e327b586762833e48b3703d53312ac32f05f24] quota: correct space limit check
git bisect good 41e327b586762833e48b3703d53312ac32f05f24

Reverting just c8c03f1858331e85d397bacccd34ef409aae993c from
v4.13-rc6-65-g2acf097f16ab reliably fixes the problem for me.

Regards
	Stefan Lippers-Hollmann

[1]	commit c8c03f1858331e85d397bacccd34ef409aae993c (HEAD)
	Author: Linus Torvalds <torvalds@...ux-foundation.org>
	Date:   Wed Aug 16 17:08:07 2017 -0700
	Subject: pty: fix the cached path of the pty slave file 
	              descriptor in the master
[2]	https://packages.qa.debian.org/p/pbuilder.html
	https://pbuilder.alioth.debian.org/
	https://anonscm.debian.org/git/pbuilder/pbuilder.git
	mounting/ umounting /dev/ptmx happens in
	https://anonscm.debian.org/git/pbuilder/pbuilder.git/tree/pbuilder-modules
[3]	the configuration file as attached relies on BUILDUSERNAME="pbuilder"
	and BUILDUSERID="$(getent passwd $BUILDUSERNAME | cut -d\: -f3)",
	commenting those out should be possible, my full BUILDUSER* setup
	is:
	# addgroup --system pbuilder
	# adduser --system --disabled-password --ingroup pbuilder --gecos "pbuilder buildd user" --home /var/run/pbuilder pbuilder
	# chown pbuilder:pbuilder /var/cache/pbuilder/deps /var/cache/pbuilder/build-result
	# chmod 2775 /var/cache/pbuilder/deps /var/cache/pbuilder/build-result
[4]	sudo /usr/sbin/pbuilder create --configfile pbuilderrc.debian.sid.amd64
[5]	sudo /usr/sbin/pbuilder update --configfile pbuilderrc.debian.sid.amd64

View attachment "pbuilder.log" of type "text/x-log" (27194 bytes)

Download attachment "pbuilderrc.debian.sid.amd64" of type "application/octet-stream" (1703 bytes)

Content of type "application/pgp-signature" skipped