Re: 2.0.pre31-2 + Werner patch of Aug 1: Bunch of Oopses.

Bill Hawes (whawes@star.net)
Mon, 04 Aug 1997 20:00:21 -0400


Philippe Strauss wrote:
> I've not yet been able to narrow down the exact patch combination
> that trigger the problem.
>
> Here's a messsage seen with the original faulty kernel, still while
> doing a tar xzvf. No oops. Seem harmless.
>
> Aug 5 00:49:04 lili kernel: Whoops: unlock_buffer: async io complete on unlocked page

Not entirely harmless -- this could be a prelude to the buffer-list
corruption. It means the page has been unlocked prematurely, so that
some of the page-scanning routines (e.g. shrink-mmap) might want to
harvest it. If I/O is really still pending, you've got problems.

It's a good clue though ...

Regards,
Bill