Re: wpi0 going offline frequently

["Steven M. Bellovin" <smb%cs.columbia.edu@localhost>]

On Tue, 19 Feb 2008 14:53:30 +0100
Joerg Sonnenberger <joerg%britannica.bec.de@localhost> wrote:

> On Mon, Feb 18, 2008 at 06:56:23PM +0000, Steven M. Bellovin wrote:
> > On Mon, 18 Feb 2008 19:15:37 +0100
> > Joerg Sonnenberger <joerg%britannica.bec.de@localhost> wrote:
> > 
> > > On Mon, Feb 18, 2008 at 06:36:56AM +0000, Steven M. Bellovin
> > > wrote:
> > > > Feb 18 01:16:41 yellowstone /netbsd: pckbport: command timeout
> > > > Feb 18 01:16:41 yellowstone /netbsd: wpi0: fatal firmware error
> > > > Feb 18 01:16:41 yellowstone /netbsd: pms_disable: command error
> > > 
> > > This is a somewhat known problem where some locks are hold way too
> > > long. Do you see this problems with a UP kernel as well?
> > > 
> > Which problem -- the command timeout or the wpi0 error?  The command
> > timeout is infrequent enough that I can live with the recovery; I'll
> > just have to remember not be too quick with the power button when
> > the machine appears to freeze...
> > 
> > The wpi0 problem is a major annoyance.
> 
> I think the fatal firmware error is just another victim of the same
> bug.
> 
Could be.  I stress, though, that the wpi0 bug *never* happens at home
unless there's something like starting X or the pckbport timeout above,
but very frequently happens -- several times an hour, at a minimum --
in my office.  In other words, while the underlying problem may be the
same, it's triggered by something environmental; my guess is traffic
levels.

When you ask about a UP kernel -- do I need to build one, or would
using cpuctl to turn off one CPU sufficient?

                --Steve Bellovin, http://www.cs.columbia.edu/~smb