[pulseaudio-discuss] alignment trap and pulseaudio being kiled

[Daniel Mack]

On Mon, Apr 4, 2011 at 8:45 PM, Baek Chang <baeksan at ccrma.stanford.edu> wrote:
> Hi,
> Trying to debug pulseaudio 0.9.22 and I am seeing some alignment trap
> warnings from kernel
> [ 1564.095562] Alignment trap: alsa-sink (5580) PC=0x2ab2ca3c
> Instr=0xe1ca00d0 Address=0x2af9302a FSR 0x011
> [ 1564.095597] Alignment trap: alsa-sink (5580) PC=0x2ab2ca68
> Instr=0xe0ca00f8 Address=0x2af9302a FSR 0x811
> [ 1564.113377] Alignment trap: alsa-sink (5580) PC=0x2ab2ca3c
> Instr=0xe1ca00d0 Address=0x2af93032 FSR 0x011
> [ 1564.122811] Alignment trap: alsa-sink (5580) PC=0x2ab2ca68
> Instr=0xe0ca00f8 Address=0x2af93032 FSR 0x811
> [ 1564.132240] Alignment trap: alsa-sink (5580) PC=0x2ab2ca3c
> Instr=0xe1ca00d0 Address=0x2af9303a FSR 0x011
> [ 1564.141703] Alignment trap: alsa-sink (5580) PC=0x2ab2ca68
> Instr=0xe0ca00f8 Address=0x2af9303a FSR 0x811
> [ 1564.151195] Alignment trap: alsa-sink (5580) PC=0x2ab2ca3c
> Instr=0xe1ca00d0 Address=0x2af93042 FSR 0x011
> [ 1564.160625] Alignment trap: alsa-sink (5580) PC=0x2ab2ca68
> Instr=0xe0ca00f8 Address=0x2af93042 FSR 0x811
> [ 1564.170065] Alignment trap: alsa-sink (5580) PC=0x2ab2ca3c
> Instr=0xe1ca00d0 Address=0x2af9304a FSR 0x011
> I tried connecting to gdb and reproducing the issue, the problem is that
> pulseaudio doesn't crash, but eventually terminates.
> Any ideas on how to debug this?

You can tell your kernel to terminate processes which cause an
alignment trap immediately: "echo 5 > /proc/cpu/alignment". That
should make gdb stop right at the instruction causing it. Also see
$kernelsrc/Documentation/arm/mem_alignment. But note that this setting
is for your whole system, and not done on a per-process level.

However, I dare to doubt that the alignment trap is your problem after
all. Such exceptions are normally just silently fixed in the
background, and the only effect you could possibly see is performance
drawbacks.  Anyway, it would be nice to fix them.

Daniel