em interrupt storm

John Baldwin jhb at freebsd.org
Mon Nov 28 21:37:24 GMT 2005


On Saturday 26 November 2005 12:03 am, Julian Elischer wrote:
> So I haven't been paying too much attention to this thread and now
> suddenly I find that
> I may have a variant of this problem.
>
> I have been tasked with adding a 4 port intel (em) based card into a
> Dell 2850 based system.
> It is based on the intel E7525 chipset and running 4.11.
>
> I have noticed that this system has irq's in use from 16 down (Maybe in
> -current it may use higher)
> but that some interrupt counts seem to move together, as if they are
> wired to gether.
>
> I also have the problem that the system seems to get completely wedged
> sometimes
> as it seems to get into the em driver interupt routine and never get out
> again,
> or if it does get out it gets called right back in again.
>
> Can someone who was following this tell me if this is the same problem?
>
> I can get into the kernel debugger and it usually shows the
> amr interrupt handler, having been interrupted by the em driver, (and
> that interrupted by
> the keyboard interrupt handler for the KDB entry.)
>
> It seems that no matter how hard it tries or how  often it is called the em
> driver is unable to clear this interrupt.
>
>
> Am I reading this correctly in that it seems to be the same problem as was
> being discussed here?

It could be.  4.x handles interrupt storms much more poorly (more prone to 
deadlock) than 5.x and later, and it still masks level-triggerred interrupts 
in the APIC while it waits for the handler to run.

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org


More information about the freebsd-current mailing list