[haiku-development] Re: [BUG] File cache issue on hrev46464

  • From: Ingo Weinhold <ingo_weinhold@xxxxxx>
  • To: haiku-development@xxxxxxxxxxxxx
  • Date: Sat, 07 Dec 2013 03:38:41 +0100

On 12/07/2013 03:04 AM, François Revol wrote:
I finally updated my development VM to post-PM stuff, however it seems I
have a serious fs issue.

The netsurf-specific .bash_history now fails to load (with dev-perso).
Actually, trying to read the file gives an error (Value too large for
defined type), and syslog has:

KERN: file_cache: read pages failed: Value too large for defined type

Oddly, it works fine on my old gcc4 VM, so the file system itself is
fine (I checked it both in the old gcc4 VM and my antique Zeta VM).

I'll have to git bisect this I suppose...

The error code would be B_BUFFER_OVERFLOW and I guess there are less than 20 probable occurrence in the kernel (plus a few more in BFS) that are candidates for your issue. You could mark them with ktrace_printf()s and enable kernel tracing with ktrace_printf() stack traces. With sed it should be only a matter of minutes to replace the "return B_BUFFER_OVERFLOW" and "status = B_BUFFER_OVERFLOW" lines.

If you also enable ktracing for syscalls, you'll be able to narrow it down easily to a single tracing item, so that you can get a stack trace. From there it should be straightforward to analyze the problem with more debug output.

CU, Ingo


Other related posts: