interrupt storm on current andy-tracking

Klaus Kurzmann mok at mnet-online.de
Thu Dec 11 19:07:38 CET 2008


* Andy Green <andy at openmoko.com> [081211 18:09]:
> Somebody in the thread at some point said:
> | Somebody in the thread at some point said:
> | | Hello,
> | |
> | | current andy-tracking (c59a8bdc7ca4b5470ebc43dfc31ed1d3d23a7c6f) seems
> | | to cause an interrupt storm for input2/3. Such that the system gets dead
> | | slow and the syslog filled with input event debugs.
> | |
> | | 9402b27b2a497f8eeb72f6fe1045f2c5be800fe2 works fine.
> | |
> | | Any idea?
> |
> | I guess it can be to do with the level interrupt processing change from
> | Balaji, or the GPIO fixes from Werner between those two?  Or maybe it is
> | config changes from me if this is that dumb input event debugging thing
> | come back again.
> |
> | Can you show what the system input event things look like?  I don't
> | think we need an "interrupt storm" to explain the system being slow if
> | in fact we're spewing 200 input event debugs a second to syslog.

> That'll be it I think...

> +CONFIG_INPUT_EVBUG=m

> It made trouble before that some rootfs will insert this module if it
> exists and then it spews for every service of 2 x 100Hz motion sensors.
> ~ I'll remove it from the configs, it got introduced again here
ahh, ok. will try with that disabled.

> http://git.openmoko.org/?p=kernel.git;a=commitdiff;h=201609904fb54abb87c9c30856a432d89567917c

> Thanks for the report, it's real nice getting the last working hash as
> well :-)
heh, and I just was about to start bisecting :-)

> -Andy

mrmoku




More information about the openmoko-kernel mailing list