[tulip] problem with tulip card ceasing to function - requires
ifup/ifdown to fix
Mon Nov 26 16:18:00 2001
On Mon, 26 Nov 2001 firstname.lastname@example.org wrote:
> On Mon, Nov 26, 2001 at 01:59:13PM -0500, Donald Becker wrote:
> > On Mon, 26 Nov 2001 email@example.com wrote:
> > Was this when the chip stopped? Or after you restarted?
> after the restart. sorry I didn't grab stats during the problem, the
> heat was on to get the firewall up again.
The diagnostic information reports the state of the chip. The driver
completely reinitializes the chip when the interface is restarted.
> > Hmmm, you either have an extreme burst of small packets, or some other
> > driver is hogging the machine. The status is reporting a normal receive
> > done interrupt.
> Is it expected behavior that a burst of small packets cause the card
> to cease passing traffic?
No. And your error log indicated that the driver did the Right Thing
when confronted with too much work: it repeatedly yeilded the machine to
other interrupt handlers. With a 21143-TD chip, it also enabled the
hardware support for interrupt mitigation.
> > Try changing the max_interrupt_work paramter from the default of 25 to 100
> > in /etc/modules.conf
> > options tulip max_interrupt_work=100
> Can you help me understand exactly what that number means?
It's the limit for actions taken during a single call of the interrupt
handler. Device driver should have a work limit on every loop,
especially interrupt handlers. But few non-network drivers are written
with checks. That means the network device driver is often the only
thing that complains when something unusual is happening on the system.
> Also, I have the driver compiled in, not loaded as a module. How
> could I change in in my situation?
Either a kernel debugger, or changing the source code and recompiling
Donald Becker firstname.lastname@example.org
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993