interrupt storm detected
Balanga Bar
balanga.bar at gmail.com
Thu Apr 18 09:37:49 UTC 2019
Wasn't sure how to check whether the correct PHY indexes were being used
but did notice differences in the boot logs between 11.2 and 12.0
Here's what I got under 11.2:-
Code:
FreeBSD clang version 6.0.0 (tags/RELEASE_600/final 326565) (based on
LLVM 6.0.0)
CPU: Feroceon 88FR131 rev 1 (**unknown 4** core)
Little-endian DC enabled IC disabled WA disabled DC streaming enabled
BTB disabled L2 enabled L2 prefetch enabled
WB enabled LABT branch prediction disabled
16KB/32B 4-way instruction cache
16KB/32B 4-way write-back-locking-C data cache
real memory = 134213632 (127 MB)
avail memory = 121294848 (115 MB)
and with 12.0:-
Code:
FreeBSD clang version 6.0.1 (tags/RELEASE_601/final 335540) (based on
LLVM 6.0.1)
CPU: Feroceon 88FR131 rev 1 (**unknown 4** core)
Little-endian DC enabled IC enabled WA disabled DC streaming enabled
BTB disabled L2 enabled L2 prefetch enabled
WB enabled LABT branch prediction enabled
16KB/32B 4-way instruction cache
16KB/32B 4-way write-back-locking-C data cache
real memory = 0 (0 MB)
avail memory = 120799232 (115 MB)
Could "IC (en/dis)abled" or "LABT branch prediction (en/dis)abled" be
responsible for the interrupt storm?
I've no idea what these terms refer to...
On Wed, Apr 17, 2019 at 6:33 PM Ian Lepore <ian at freebsd.org> wrote:
> On Wed, 2019-04-17 at 19:23 +0100, Balanga Bar wrote:
> > After building FreeBSD 12.0-RELEASE for ARM I managed to boot up my
> > GoFlex
> > Home booting up without getting kernel panics which I did with 11.2,
> > but
> > instead I get a msg every second or so saying:-
> >
> > interrupt storm detected on "intr13:" throttling interrupt source
> >
> > After some experimenting I managed to run vmstat -i which showed:-
> >
> > interrupt total rate
> > irq1: timer0 27224 201
> > irq12: mge0 6037 45
> > irq13: mge0 11003 81
> > irq19: ehci0 5665 42
> > irq33: uart0 247 2
> > Total 50176 371
> >
> >
> > Is there anyting I can do to stop this storm?
> >
> > I tried adding hw.intr_storm_threshold=4000 to /etc/sysctl.conf, but
> > that
> > made no difference.
> >
>
> Check whether the right PHY indexes are being used. I've seen kirkwood
> systems put the PHYs on 0 & 1, 0 & 4, and 0 & 8. I've seen interrupt
> storms on mge before. I'm not positive the PHY numbers and the irq
> storms are related, that's just two random things I remember from the
> past.
>
> -- Ian
>
>
More information about the freebsd-arm
mailing list