Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 631002 - <dev-db/redis-{3.2.8-r5,4.0.2-r1}: privilege escalation via PID file manipulation
Summary: <dev-db/redis-{3.2.8-r5,4.0.2-r1}: privilege escalation via PID file manipula...
Status: RESOLVED FIXED
Alias: None
Product: Gentoo Security
Classification: Unclassified
Component: Vulnerabilities (show other bugs)
Hardware: All Linux
: Normal minor
Assignee: Gentoo Security
URL:
Whiteboard: B3 [noglsa]
Keywords:
Depends on:
Blocks:
 
Reported: 2017-09-14 19:18 UTC by Michael Orlitzky
Modified: 2018-11-23 23:30 UTC (History)
3 users (show)

See Also:
Package list:
dev-db/redis-3.2.8-r5 dev-db/redis-4.0.2-r1
Runtime testing required: ---
stable-bot: sanity-check+


Attachments
redis.initd-5 (redis,591 bytes, text/plain)
2017-09-14 19:18 UTC, Michael Orlitzky
no flags Details
redis.confd-r1 (redis,511 bytes, text/plain)
2017-09-14 19:19 UTC, Michael Orlitzky
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Michael Orlitzky gentoo-dev 2017-09-14 19:18:36 UTC
Created attachment 494550 [details]
redis.initd-5

The init script for redis gives ownership of its PID file directory to the daemon's runtime user:

  start_pre() {
      checkpath -d -m 0775 -o ${REDIS_USER}:${REDIS_GROUP} \
        $(dirname ${REDIS_PID})
  }

This can be exploited by $REDIS_USER to kill root processes, since when the service is stopped, root will send a SIGTERM to the contents of the PID file (which are under the control of $REDIS_USER).

I've rewritten the init script to work around this by having OpenRC manage the PID file itself, and store it directly in /run. I also utilized some other OpenRC variables to clean up the script and to eliminate everything except the depend() function.

Finally, I removed "use net" and added an rc_need line to the conf.d file. Since the actual interface we need is determined by redis.conf, the correct dependency needs to be set by the user. (The default 127.0.0.1 isn't even covered by "use net").
Comment 1 Michael Orlitzky gentoo-dev 2017-09-14 19:19:15 UTC
Created attachment 494552 [details]
redis.confd-r1
Comment 2 Tomáš Mózes 2017-10-06 20:42:49 UTC
https://github.com/gentoo/gentoo/pull/5875
Comment 3 Larry the Git Cow gentoo-dev 2017-10-08 23:24:34 UTC
The bug has been referenced in the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=f689ea635ecbb7290b1c914000bf83e9289cd5e9

commit f689ea635ecbb7290b1c914000bf83e9289cd5e9
Author:     Tomáš Mózes <hydrapolic@gmail.com>
AuthorDate: 2017-10-06 20:40:02 +0000
Commit:     Robin H. Johnson <robbat2@gentoo.org>
CommitDate: 2017-10-08 23:23:40 +0000

    dev-db/redis: bump to 4.0.2
    
    (cherry picked from commit 93eac00c9ae82a63fceacdcc5b7ded98e44a3a9c)
    Bug: https://bugs.gentoo.org/631002
    Closes: https://github.com/gentoo/gentoo/pull/5875#issuecomment-334886169
    Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

 dev-db/redis/Manifest                       |   1 +
 dev-db/redis/files/redis-3.2.3-config.patch |   2 +-
 dev-db/redis/files/redis.confd-r1           |  20 +++++
 dev-db/redis/files/redis.initd-5            |  23 ++++++
 dev-db/redis/redis-4.0.2.ebuild             | 122 ++++++++++++++++++++++++++++
 5 files changed, 167 insertions(+), 1 deletion(-)}
Comment 4 Larry the Git Cow gentoo-dev 2017-10-09 00:13:19 UTC
The bug has been referenced in the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=7733941faa2a1687dbc16d6cb88c214c0902a376

commit 7733941faa2a1687dbc16d6cb88c214c0902a376
Author:     Robin H. Johnson <robbat2@gentoo.org>
AuthorDate: 2017-10-09 00:03:11 +0000
Commit:     Robin H. Johnson <robbat2@gentoo.org>
CommitDate: 2017-10-09 00:13:07 +0000

    dev-db/redis: backport security & Lua-combo fixes from 4.x to 3.2.8 release.
    
    Bug: https://bugs.gentoo.org/show_bug.cgi?id=631002
    Package-Manager: Portage-2.3.8, Repoman-2.3.3

 dev-db/redis/redis-3.2.8-r4.ebuild | 131 +++++++++++++++++++++++++++++++++++++
 1 file changed, 131 insertions(+)}
Comment 5 Robin Johnson archtester Gentoo Infrastructure gentoo-dev Security 2017-10-09 00:18:44 UTC
arches, please test & stabilize redis-3.2.8-r4 AND redis-4.0.2.

target keywords: amd64, arm, hppa, ppc, ppc64, x86, arm64
test instructions: Package with FEATURES=test
Comment 6 Robin Johnson archtester Gentoo Infrastructure gentoo-dev Security 2017-10-09 00:19:29 UTC
arches, see comment #5 for stabilizing please.
Comment 7 jack_mort 2017-10-12 10:21:57 UTC
Hi,
Since the last update, I end with 2 pid files :

-rw-r--r--. 1 root  root  system_u:object_r:initrc_var_run_t  6 12 oct.  12:19 /run/redis.pid

/run/redis:
total 4
-rw-r--r--. 1 redis redis system_u:object_r:redis_var_run_t 6 12 oct.  12:19 redis.pid
srwxrwx---. 1 redis redis system_u:object_r:redis_var_run_t 0 12 oct.  12:19 redis.sock

The one monitored by OpenRC is /run/redis.pid, but if we look at the content of the file, the pid does not exist. The correct one is stored inside /run/redis/redis.pid.

Thus it leaves the redis service in a crashed state, and you have to manually kill the correct pid to actually stop redis.

Is this normal ?
Comment 8 Tomáš Mózes 2017-10-12 10:26:27 UTC
(In reply to jack_mort from comment #7)
> Hi,
> Since the last update, I end with 2 pid files :
> 
> -rw-r--r--. 1 root  root  system_u:object_r:initrc_var_run_t  6 12 oct. 
> 12:19 /run/redis.pid
> 
> /run/redis:
> total 4
> -rw-r--r--. 1 redis redis system_u:object_r:redis_var_run_t 6 12 oct.  12:19
> redis.pid
> srwxrwx---. 1 redis redis system_u:object_r:redis_var_run_t 0 12 oct.  12:19
> redis.sock
> 
> The one monitored by OpenRC is /run/redis.pid, but if we look at the content
> of the file, the pid does not exist. The correct one is stored inside
> /run/redis/redis.pid.
> 
> Thus it leaves the redis service in a crashed state, and you have to
> manually kill the correct pid to actually stop redis.
> 
> Is this normal ?

I forgot to change the pidfile in https://github.com/gentoo/gentoo/blob/master/dev-db/redis/files/redis-3.2.3-config.patch. This happens when you have supervised yes in your /etc/redis.conf, is that correct?
Comment 9 Tomáš Mózes 2017-10-12 11:10:31 UTC
On a second thought, this won't be fixed by adjusting the pidfile in /etc/redis.conf. Even if we set it to /run/redis.pid, redis will not be able to write there with user redis, that was the whole point of the bug.

So I'm afraid the setup with "daemonize yes" will not be supported. Sorry I wrongly asked about "supervised yes", it really is "daemonize yes".

Robbat2, how about to print a warning that redis setups with "daemonize yes" are not supported any more for security reasons?
Comment 10 Michael Orlitzky gentoo-dev 2017-10-12 16:57:14 UTC
"daemonize yes" never quite worked... if you try changing either $REDIS_PID (in the conf.d file) or "pidfile" (in redis.conf) in an older version of redis, then you'll still wind up with an extra junk PID file. By default, $REDIS_PID and "pidfile" agreed, so the junk file was fortuitously overwritten with the real one, and nobody noticed. But ultimately, OpenRC is trying to background the process itself (in order to run as a non-root user), so we've always wanted "daemonize no".

Fortunately, it looks like we can force "daemonize no" on the command-line, regardless of what the config file says. See "Passing arguments via the command line" in,

  https://redis.io/topics/config

I'm way too tired to test it right now, but it looks like "--daemonize no" in $command_args should do it.
Comment 11 Larry the Git Cow gentoo-dev 2017-10-12 20:37:08 UTC
The bug has been referenced in the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=b545fdf6233c35ef3173109dfcd3d0c2b05da205

commit b545fdf6233c35ef3173109dfcd3d0c2b05da205
Author:     Robin H. Johnson <robbat2@gentoo.org>
AuthorDate: 2017-10-12 20:30:01 +0000
Commit:     Robin H. Johnson <robbat2@gentoo.org>
CommitDate: 2017-10-12 20:36:58 +0000

    dev-db/redis: use s-s-d to background, disallow daemonize.
    
    Bug: https://bugs.gentoo.org/631002#c10
    Package-Manager: Portage-2.3.8, Repoman-2.3.3

 dev-db/redis/files/redis.initd-5                              | 4 +++-
 dev-db/redis/{redis-3.2.8-r4.ebuild => redis-3.2.8-r5.ebuild} | 0
 dev-db/redis/{redis-4.0.2.ebuild => redis-4.0.2-r1.ebuild}    | 0
 3 files changed, 3 insertions(+), 1 deletion(-)}
Comment 12 Robin Johnson archtester Gentoo Infrastructure gentoo-dev Security 2017-10-12 20:38:14 UTC
Implemented & tested.
Comment 13 Stabilization helper bot gentoo-dev 2017-10-12 21:01:20 UTC
An automated check of this bug failed - the following atoms are unknown:

dev-db/redis-3.2.8-r5
dev-db/redis-4.0.2-r1

Please verify the atom list.
Comment 14 Tomáš Mózes 2017-10-13 04:15:41 UTC
Thanks for pointing out the solution and fixing this.

However what to do with the /etc/redis.conf pidfile. I know it's not used, but to avoid confusion, we could possibly comment it. And also to print a warning that we enforce "daemonize no".
Comment 15 jack_mort 2017-10-13 10:10:20 UTC
Hi,
I was surprisingly using daemonize=yes, good catch.
Now it's all OK with -r1 : only one pidfile and rc-status shows a started process :-)
Thanks.
Comment 16 Thomas Deutschmann (RETIRED) gentoo-dev 2017-10-27 13:51:20 UTC
x86 stable
Comment 17 Sergei Trofimovich (RETIRED) gentoo-dev 2017-10-28 23:18:36 UTC
ppc64 stable
Comment 18 Manuel Rüger (RETIRED) gentoo-dev 2017-10-29 11:24:46 UTC
amd64 stable
Comment 19 Ultrabug gentoo-dev 2018-01-05 16:50:11 UTC
removed unstable ARM64 arch from list.
Comment 20 Markus Meier gentoo-dev 2018-02-05 21:20:01 UTC
arm stable
Comment 21 Mart Raudsepp gentoo-dev 2018-03-03 01:58:02 UTC
(In reply to Ultrabug from comment #19)
> removed unstable ARM64 arch from list.

Please do not do that when we have earlier versions stable.
Comment 22 Mart Raudsepp gentoo-dev 2018-03-04 11:53:24 UTC
arm64 stable despite test failure (bug 649556), as not a regression
Comment 23 Matt Turner gentoo-dev 2018-04-22 21:23:26 UTC
hppa stable
Comment 24 Mikle Kolyada (RETIRED) archtester Gentoo Infrastructure gentoo-dev Security 2018-05-27 11:24:44 UTC
ppc stable
Comment 25 Aaron Bauman (RETIRED) gentoo-dev 2018-06-11 15:55:53 UTC
@maintainer(s), please drop vulnerable.
Comment 26 Larry the Git Cow gentoo-dev 2018-08-15 03:09:51 UTC
The bug has been referenced in the following commit(s):

https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=248dbab6baed691d61011be1a4115b25e2f05e4f

commit 248dbab6baed691d61011be1a4115b25e2f05e4f
Author:     Robin H. Johnson <robbat2@gentoo.org>
AuthorDate: 2018-08-15 03:09:02 +0000
Commit:     Robin H. Johnson <robbat2@gentoo.org>
CommitDate: 2018-08-15 03:09:45 +0000

    dev-db/redis: cleanup old ebuilds
    
    Bug: https://bugs.gentoo.org/631002
    Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>
    Package-Manager: Portage-2.3.40, Repoman-2.3.9

 dev-db/redis/Manifest                          |   3 -
 dev-db/redis/files/configure.ac-2.2            |  58 -----------
 dev-db/redis/files/redis-2.8.17-config.patch   |  46 ---------
 dev-db/redis/files/redis-2.8.3-shared.patch    |  36 -------
 dev-db/redis/files/redis-3.0.0-sharedlua.patch |  44 ---------
 dev-db/redis/files/redis.confd                 |  20 ----
 dev-db/redis/files/redis.initd-4               |  31 ------
 dev-db/redis/redis-3.0.7-r1.ebuild             | 115 ----------------------
 dev-db/redis/redis-3.0.7.ebuild                | 115 ----------------------
 dev-db/redis/redis-3.2.5.ebuild                | 123 ------------------------
 dev-db/redis/redis-3.2.8-r2.ebuild             | 123 ------------------------
 dev-db/redis/redis-4.0.1-r1.ebuild             | 128 -------------------------
 dev-db/redis/redis-4.0.1.ebuild                | 122 -----------------------
 13 files changed, 964 deletions(-)