libmad causes mpd to seg fault on update

Status changed to 'Confirmed' because the bug affects multiple users.

Same behavior seen on a Pandaboard.

Output from gdb trying to play and mp3 file with herrie (http://herrie.info):

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0xad1fe3f0 (LWP 26380)]
0xb6d42760 in mad_layer_III () from /usr/lib/arm-linux-gnueabihf/libmad.so.0
(gdb) bt
#0 0xb6d42760 in mad_layer_III () from /usr/lib/arm-linux-gnueabihf/libmad.so.0
#1 0x00f4dd96 in ?? ()
#2 0x00f4dd96 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
(gdb) quit
A debugging session is active.

        Inferior 1 [process 26375] will be killed.

Pandaboard + mpd:

root@panda:~# gdb mpd 2>&1 | tee ~/gdb-mpd.txt
GNU gdb (Ubuntu/Linaro 7.4-2012.04-0ubuntu2) 7.4-2012.04
Copyright (C) 2012 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "arm-linux-gnueabihf".
For bug reporting instructions, please see:
<http://bugs.launchpad.net/gdb-linaro/>...
Reading symbols from /usr/bin/mpd...(no debugging symbols found)...done.
(gdb) handle SIG33 pass nostop noprint
Signal Stop Print Pass to program Description
SIG33 No No Yes Real-time event 33
(gdb) set pagination 0
(gdb) run --no-daemon /etc/mpd.conf
Starting program: /usr/bin/mpd --no-daemon /etc/mpd.conf
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/arm-linux-gnueabihf/libthread_db.so.1".
[New Thread 0xb4eb0440 (LWP 28658)]
[New Thread 0xb44ff440 (LWP 28659)]
[New Thread 0xb38ff440 (LWP 28660)]
[New Thread 0xb2eff440 (LWP 28661)]

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0xb2eff440 (LWP 28661)]
0xb6011760 in mad_layer_III () from /usr/lib/arm-linux-gnueabihf/libmad.so.0

I can confirm this bug in Kubuntu on pandaboard, too

Single-stepping in gdb revealed that things fell apart after calling III_imdct_l(). The ARM-specific implementation is in imdct_l_arm.S. Possibly a hardfp ABI incompatibility? Running configure with "--disable-aso" works around it.

I'm using a 12.04 Ubuntu armhf install on a Beaglebone and mpg321 is segfaulting.
Compiling libmad myself helped. Here are some notes: http://igw.tuwien.ac.at/ceat/node/2#comment-162

I can confirm the workaround from comment #7 - I build the package and added to --disable-aso to the debian/rules files and everything works file.

Indeed, it seems to be a problem in the architecture specific optimisation for ARM.

(gdb) s
_III_imdct_l () at imdct_l_arm.S:213
213 stmdb sp!, { r2, r4 - r11, lr } @ all callee saved regs, plus arg3
(gdb) s
Cannot access memory at address 0xfb30506a
(gdb) s

Program received signal SIGSEGV, Segmentation fault.
0x40107b12 in mad_layer_III (stream=<error reading variable: Cannot access memory at address 0xfb304f22>,
    frame=<error reading variable: Cannot access memory at address 0xfb304f1e>) at layer3.c:2560
2560 if (header->flags & MAD_FLAG_PROTECTION) {

I think for now I'll disable the ASO on arm. The assembler there is quite old, from 2001 according to copyright. If someone wants to update it they are more than welcome.

This bug causes rhythmbox to crash on ARM (e.g. Nexus 7) when gstreamer1.0-plugins-ugly (which links to libmad0) is installed.

This bug was fixed in the package libmad - 0.15.1b-7ubuntu2

---------------
libmad (0.15.1b-7ubuntu2) raring; urgency=low

  * Disable architecture specific optimisations on ARM, as there is a bug in
    this codepath which causes segfaults, and the assembler is very old
    (likely bitrotted). (LP: #989846)
 -- Iain Lane <email address hidden> Wed, 12 Dec 2012 12:10:33 +0000

What about putting the fix into the repos for 12.04 LTS?

Without this, most video players don't work on ARM machines. :(

What about putting the fix into the repos for 12.04 LTS?

Please, someone, nominate it for 12.04.

[Impact]

Without this, most video players don't work on ARM machines. :(

[Test Case]

From https://bugs.launchpad.net/ubuntu/+source/mpg321/+bug/1196736 :

mpg321 /usr/share/libubuntuoneui/1/javascript/empty.mp3

should not crash on ARM.

[Regression Potential]

The fix removes some optimizations. No regressions are expected.

follow https://wiki.ubuntu.com/StableReleaseUpdates to get a 12.04 update

Does --disable-aso (which fixes the problem) disable soemthing in the compiler tools, or this just disables some written beforehand assembler code which is supposed to work better?

If the bugs comes from the code generated by the compiler tools with soem optimizations on, shouldn't we report this to the compiler developers?

quantal has seen the end of its life and is no longer receiving any updates. Marking the quantal task for this ticket as "Won't Fix".

The Precise Pangolin has reached end of life, so this bug will not be fixed for that release