Re: Interrupt storm mitigation needed

[Tom Ivar Helbekkmo <tih%hamartun.priv.no@localhost>]

Joerg Sonnenberger <joerg%britannica.bec.de@localhost> writes:

> Well, let's try the first hack of keep uhci interrupts off and just poll
> via timeout.

Will try this in a few hours, when I'm back with the machine.  It'll be
interesting to see what happens, not least because there's another
symptom that I realize I haven't mentioned:

When I disable USB, either by turning it off in the BIOS configuration
of the machine, or by adding "no uhci* at pci?" to my config file (I
include GENERIC and adjust as needed), the machine will not boot.  It
will get through the kernel initialization, not mentioning uhci at all,
and attaching both amr0 and its ld? devices, but then, around when it
should be exec'ing init, the kernel crashes with a segmentation
violation, sometimes in amr_intr(), sometimes in amr_quartz_get_work()
(which is called from amr_intr()).  Whether this is the amr driver being
hit by an interrupt storm, or what, I don't know.

-tih
-- 
Popularity is the hallmark of mediocrity.  --Niles Crane, "Frasier"