[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-smp
Subject:    Semi-lockup with 2.1.40 Re: A semi-lockup with 2.1.39 and SMP
From:       Edward Welbon <welbon () bga ! com>
Date:       1997-05-23 15:39:56
[Download RAW message or body]

I have a similar problem.  Bonnie hangs in disk IO very reliably (though
unfortunately, not quickly). It seems that a disk interrupt has somehow
been lost or that a disk io request didn't actually start a physical IO.

On Wed, 21 May 1997, H.J. Lu wrote:

> When I was doing heavy disk I/O under SMP linux 2.1.39 on my dual PPro
> 150 with BT-958 SCSI, which is also my NFS server with the kernel-based
> NFS server, I got

I encountered the problem with 2.1.39 and 2.1.40 . I have a SMP P6DOF with
adaptec 2940 adapters. I had no NFS mounts at the time of any hang.  I
have the problem with and without the updated 2940 driver.

> May 21 10:50:16 ocean kernel: end_scsi_request: buffer-list destroyed
> May 21 10:50:16 ocean last message repeated 3 times

I have no log messages.

> and then all my processes, who were reading a particular file, stuck
> in kernel at 0xc011f550 and my System.map is
> 
> c011f4ac T __wait_on_page

  c011f550 <-- Lu's lockup 

> c011f5b0 T generic_file_read

c011e854 T __wait_on_page

c011e8f8 <-- address where bonnie stalled

c011e958 T generic_file_read
 	   
> At the same time, my NFS client machine was accessing the server.
> Although those processes were stuck, the machine was still running
> ok. But I couldn't kill those processes to do a cleanup shutdown.
> I had to press the reset button. When the machine came up again,
> I verified that the file those processes stuck with was ok. It
> seems it is a kernel bug.

Same, same, same. 

I am open to suggestions as to how to track this further.

Ed Welbon; welbon@bga.com;
<a href="http://www.bga.com/~welbon/arachnid.html">the arachnid mail list</a>

[prev in list] [next in list] [prev in thread] [next in thread]