Re: Interrupt storm mitigation needed

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

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

> From what I see from the original message, there are two issues:
>
> (1) It's a shared interrupt.
> (2) uhci is generated interrupts when it should not.

Also ehci.  Here are the updated counters, with usage info:

interrupt                 total     rate    device      used?

cpu0 timer              6817487       99

ioapic0 pin 4               459        0    com0        yes (console)
ioapic0 pin 14                6        0    piixide0    no  (idle cd player)
ioapic0 pin 16           427117        6    uhci0       no
ioapic0 pin 18           706564       10    uhci2       yes (rs232 dongle)
                                            radeon0     no  (idle frame buf)
ioapic0 pin 23          1590273       23    ehci0       no
                                            cmdide0     no  (idle ide ctrl)
ioapic1 pin 14          2920953       42    amr0        yes (3 ld devices)
ioapic2 pin 0           6814339       99    wm0         yes (vlan trunk)

Total                  19277198      282

uhci0, uhci2, and ehci0 are all getting lots of erroneous interrupts,
while only uhci2 is actually used, and at a very, very much lower rate
than what is indicated here.  It is unknown to what extent amr0 and wm0
may be getting hit, but the data I have from captures of 'vmstat -i 10'
output when hangs occurred indicates that they aren't.

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