[3021] in linux-scsi channel archive
Re: AHA2940 kernel notices
daemon@ATHENA.MIT.EDU (Daniel M. Eischen)
Sun Jan 4 05:15:28 1998
Date: Sat, 3 Jan 1998 23:52:03 -0600 (CST)
From: "Daniel M. Eischen" <deischen@iworks.InterWorks.org>
To: jiivee@hut.fi, linux-scsi@vger.rutgers.edu
> Should I worry about kernel messages like those below?
>
> Jan 1 01:09:42 iguana kernel: (scsi0:-1:-1:-1) Command complete near Qfull count, qoutcnt = 16.
> Jan 1 01:14:31 iguana kernel: (scsi0:-1:-1:-1) Command complete near Qfull count, qoutcnt = 16.
> Jan 1 01:19:40 iguana kernel: (scsi0:-1:-1:-1) Command complete near Qfull count, qoutcnt = 15.
>
> This box is a pentium 75 MHz w/ 96 MB RAM, two SCSI disks on a
> AHA2940 and two IDE disks on separate buses. Kernel is
> Linux-2.0.33. Those messages happened during (supposedly) quite
> heavy network load (getting stuff in from one ethernet interface
> and saving it to NFS mounted disk through another ethernet
> interface. I suppose there were not very much local disk i/o at
> that time.
Those should be harmless messages. What they are telling you
is that there is 15/16 commands in the complete queue waiting
for processing. The queue is at it's limit and the sequencer
will have to wait for the kernel driver to start draining the
queue before it can perform any more work. You get this when
there is a high interrupt latency, which is because of the
high network load and/or IDE I/O processing.
I think earlier versions of the driver would have a problem
under this condition and lose commands because the sequencer
wouldn't wait for an opening in the complete queue.
Dan Eischen
deischen@iworks.InterWorks.org
