[haiku-development] Re: Tracker Crash post-initialization (was: Filesystem corruption and Alpha-1)
- From: "Axel Dörfler" <axeld@xxxxxxxxxxxxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Fri, 26 Jun 2009 20:50:43 +0200 CEST
David Himelright <david.himelright@xxxxxxxxx> wrote:
> One question however: is there anything that I can do to provide
> canned
> output for debugging of possible issues with the files themselves? I
> wasn't
> able to reproduce with two disk images under virtualization, however
> the
> copy from a VirtualBox vdi to a vmdk reference to a formatted
> partition on
> disk produces the error just after initialization sometimes once,
> sometimes
> a handful of times. This could be any number of things, it might be a
> problem with the format of the files on disk, missing metadata, or
> simply
> Tracker crashing due to unexpected latency due to the disk being
> marked busy
> by another active partition.
>
> Can anyone recommend a tool for investigating the possible filesystem
> issue
> if the crash points in that direction? This smells like it might be a
> pain
> to reproduce on another system.
I'm not sure what exactly the issue is anymore, but there are several
things you can do to have more information in a kernel debugger
session. You should enable tracing (see build/config_headers/
tracing_config.h) with a rather large buffer (like 100MB or more,
depending on how much you can spare), plus setting BFS_TRACING,
BLOCK_CACHE_BLOCK_TRACING, and BLOCK_CACHE_TRANSACTION_TRACING to 1.
This allows you to see the file system actions that happened.
Sometimes, it can be helpful to enable SYSCALL_TRACING as well, to see
what triggered a certain action.
HTH,
Bye,
Axel.
Other related posts: