Bug#382214: multipath-tools: Multiple serious bugs in 0.4.7

[John Goerzen]

Package: multipath-tools
Version: 0.4.7-1
Severity: grave
Justification: renders package unusable

There are several problems in the current version that are fixed in the
git tree.  I have personally encountered some, but not all, of these.

Here are snippets from the git logs that document these:

commit 40b575955cc189aa993e6a030b66b5fef6bcf288
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Wed Jul 19 14:49:34 2006 +0200

    [libmultipath] make sure to terminate strftime buff properly

    Bernhard Fischer
commit 52d868ccb57be7d5a2dae9535c9e39caf2e6077e
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Fri Jun 16 12:36:55 2006 +0200

    Finish up the work on SG_IO timeout

    all over the place.

commit 08fc58d0ae7942a8a2dac5ddf144aef2eb79ee77
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Fri Jun 16 12:23:07 2006 +0200

    [libcheckers] workaround for rather profound path checking issue

    Overloaded storage response time can be very long.
    SG_IO timouts after DEF_TIMEOUT milliseconds, and checkers
interprets this
    as a path failure. multipathd then proactively evicts the path from
the DM
    multipath table in this case.

    This generaly snow balls and ends up in full eviction and IO errors
for end
    users. Bad. This may also cause SCSI bus resets, causing disruption
for all
    local and external storage hardware users.

    Provision a long timeout. Longer than any real-world application would cope
    with.

    If someone can suggest a real fix, like an evidence the ioctl timed out,
    voice up. The ioctl being blocking, I'd rather lower the timeout and retry
    than apply this workaround.

commit b1ff2a23937ec2a792fcf4a746d6d563cfd1eb4e
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Fri Jun 2 22:02:17 2006 +0200

    [libmultipath] fix hwtable brainos

    Matthias Rudolph remembered me the most simple regex rules. Shame.

commit c30f2a5bb7919d1482ff5d1de5e34b34786e3eec
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Fri Jun 2 21:56:16 2006 +0200

    [libmultipath] fix segv in load_config()

    segv when a vendor or product string was not set in the
    configuration file.

commit 4df495ad01614b3922b7474238b07d7d3e1316d7
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Thu May 4 23:34:40 2006 +0200

    [priority] pp_alua correctness fix

    Tom Kalus noted path_priority/pp_aula/spc.h misrepresented to spec
    regarding the product string lentgh. Credits for the fix too.

    Fix acked by Stefan Bader.

commit 3c73a28f02d8621e20819f37224834908b7baf0b
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Thu Apr 20 10:26:18 2006 +0200

    [libmultipath] push WWID_SIZE to 128

    64 was too small for "scsi_id -p 0x80" in most cases.

    As a consequence, the tools refuse to get the uid at all and the path was
    discarded ... which is still the good behaviour.

    The problem tracking was done by Tarik Sedlarevic, Visionbytes.

commit e31fb98d9d6c66ec88b334f363a05b247f87495f
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Thu Apr 13 23:03:08 2006 +0200

    [libmultipath] fix segfault when no config file

    Brian Wong noticed the segfault and tracked it down to free_keywords()
    trying to walk a NULL keywords vector.  Credits.

commit cea2dae4c8cb2584fb0de7cd6066862671b66d4b
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Wed Apr 12 21:49:39 2006 +0200

    [libmultipath] get_serial overflow fix

    Arsene Gschwind tracked down this bug. Credits.

    The patch also normalize the return codes, even if they are not
    used yet.

commit d98856cbb7d4bbafe10cfc402cba98409002ba86
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Fri Mar 31 23:51:43 2006 +0200

    [multipath] cosmetical fix

    I'd like to ask some cosmetical fix in libmultipath.
    In Redhat /etc/rc.sysinit calls /sbin/multipath -v 0 after modprobe
    dm-multipath to catch the disks already attached.

    Now it prints a small garbage on screen for every map, I think.
    Actually every multipath -F; multipath/multipath -v 0 sequence does
    that.

    The reason is neither print_multipath_topology nor
    snprint_multipath_topology initialize buff if verbosity is 0.

    Maxim Kozover

commit 1a67e8ba8e6a1c68662aaf5fa5d4f7ae898eccfe
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Fri Mar 31 23:08:08 2006 +0200

    [libmultipath] plug the keywords vector leak upon "reconfigure"

commit 31c986c9eb24a841a633bde624240fa24d7b6686
Author: Christophe Varoqui <root at xa-s05.(none)>
Date:   Fri Mar 31 23:00:56 2006 +0200

    [multipathd] plug a small leak on "reconfigure" CLI command

-- System Information:
Debian Release: testing/unstable
  APT prefers unstable
  APT policy: (500, 'unstable')
Architecture: i386 (i686)
Shell:  /bin/sh linked to /bin/bash
Kernel: Linux 2.6.17.2
Locale: LANG=C, LC_CTYPE=en_US (charmap=ISO-8859-1)