Jim Meyering (meyering) of Red Hat reports: glusterfs writes to predictably-named files in /tmp using the PID and other non random or easily guessed information to create file names. libglusterfs/src/statedump.c gf_proc_dump_open (char *dump_dir, char *brickname) { char path[PATH_MAX] = {0,}; int dump_fd = -1; snprintf (path, sizeof (path), "%s/%s.%d.dump", (dump_dir ? dump_dir : "/tmp"), brickname, getpid()); ... gf_proc_dump_options_init () { int ret = -1; FILE *fp = NULL; char buf[256]; char dumpbuf[GF_DUMP_MAX_BUF_LEN]; char *key = NULL, *value = NULL; char *saveptr = NULL; char dump_option_file[PATH_MAX]; snprintf (dump_option_file, sizeof (dump_option_file), "/tmp/glusterdump.%d.options", getpid ());
Created glusterfs tracking bugs for this issue Affects: fedora-all [bug 875888] Affects: epel-all [bug 875889]
This issue has been addressed in following products: Red Hat Storage 2.0 Native Client for RHEL 6 for Red Hat Storage Native Client for RHEL 5 for Red Hat Storage Via RHSA-2012:1456 https://rhn.redhat.com/errata/RHSA-2012-1456.html
Any word on an upstream fix?
commit 3d10587d9d6400c9141b1f278bb5e2027fa784b8 (http://review.gluster.org/4042) and commit 56d1f81949fde78615cd9fec048259d261f99c40 (http://review.gluster.org/4091) are done upstream to handle it.
(In reply to comment #17) > commit 3d10587d9d6400c9141b1f278bb5e2027fa784b8 > (http://review.gluster.org/4042) > and > commit 56d1f81949fde78615cd9fec048259d261f99c40 > (http://review.gluster.org/4091) > > are done upstream to handle it. Thanks, Amar.
This was partially addressed in http://rhn.redhat.com/errata/RHSA-2012-1456.html Please see bug #886364 for additional information
(In reply to comment #17) > commit 3d10587d9d6400c9141b1f278bb5e2027fa784b8 > (http://review.gluster.org/4042) > and > commit 56d1f81949fde78615cd9fec048259d261f99c40 > (http://review.gluster.org/4091) > > are done upstream to handle it. Amar - any chance of applying a patch to 3.2.x? This comes via this request: https://security-tracker.debian.org/tracker/CVE-2012-4417