On Sat, Jul 23, 2016 at 11:41 AM, Zbigniew Jędrzejewski-Szmek
wrote:
> On Fri, Jul 22, 2016 at 07:40:25PM -0600, Chris Murphy wrote:
>> qemu-system-x86 core dumped on me for some reason, and I see it in the
>> journal:
>>
>> Jul 22 13:24:30 f24m systemd-coredump[3914]: Process 3829
>> (qemu-system-x86) of user 107 dumped core.
>>
>> OK no problem.
>>
>> [chris@f24m ~]$ sudo coredumpctl gdb 3829
>> [...snip...]
>> BFD: Warning: /var/tmp/coredump-DtRBEA is truncated: expected core
>> file size >= 4115308544, found: 2147483648.
>>
>> OK let's extract the coredump to its own file:
>>
>> # sudo coredumpctl -o qemu-system-x86.coredump dump
>> /usr/bin/qemu-system-x86_64
>> # ls -l
>> -rw-r--r--. 1 chris chris 2147483648 Jul 22 14:08 qemu-system-x86.coredump
>>
>> Huh, 2147483648 is 0x8000 or exactly 2GiB. Suspicious? Is there a
>> misconfiguration of some kind causing this truncation? Or?
>>
>> This is Fedora 24, systemd-229-8.fc24.x86_64
>
> Sounds like a bug in the logic. 2GiB is the default value for ProcessSizeMax=
> and ExternalSizeMax=. coredump.conf(5) implies coredumps larger than that
> will not be stored. I don't think it's useful to have truncated coredumps,
> so I think we should:
> 1. make sure that coredumps which exceed the allowed size not saved
> 2. consider increasing the maximum coredump size. 2GiB feels a bit low
>in this day and age.
What's the significance of the path in the gdb complaint...
BFD: Warning: /var/tmp/coredump-S2qIGu is truncated: expected core
file size >= 4115308544, found: 2147483648.
...compared to the path coredumpctl info reports?
Coredump:
/var/lib/systemd/coredump/core.qemu-system-x86.107.b91161300395440f96b49cd0b879488d.3829.1469215461.lz4
Does 'coredumpctl gdb' grab the compressed coredump and decompresses
it into /var/tmp and then hands it off to gdb? I just want to make
sure there's no other source for truncation.
I'll file a bug.
--
Chris Murphy
___
systemd-devel mailing list
systemd-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/systemd-devel
