On Tue, 2002-11-12 at 21:02, tom wang wrote:
>
> Just now, I tried that test on the cvs kernel again, it did not oops
> anymore,
> but it is locked after a while, I studied all the process trace of the
> system,
> it seems that only one process is locked, the trace is here.
> __down()
> __down_failed()
> [xfs].text.lock.xfs_log
> xlog_state_get_iclog_space()
> xlog_write()
> xfs_log_write()
> xfs_trans_commit()
> xfs_attr_rolltrans£¨£©
> xfs_attr_mode_addname()
> xfs_attr_set()
> linvfs_setxattr()
> setxattr()
> sys_lsetxattr()
> system_call()
> (It seems that some process was wirting the log to disk. but no other process
> was writing to disk at this time)
> I think it is easy reproduce in your box. (BTW if we compiled xfs with DEBUG
> option, the xfsidbg module can not been installed)
> thanks.
> tom
>
There are other issues with XFS in 2.4.20-rc1, which I am
working through right now. Changes in the way the elevator
works and the scsi code have caused some major issues with
XFS. One of the places things seem to at least pause is
here. Are you running scsi or ide here?
Steve
|