[Avila] Numerous Interrupts on IRQ18

Burns, Darrell dburns1 at ibahn.com
Fri Nov 3 12:30:13 EST 2006


I'm with David on the IRQ18 question.  Why are there so many interrupts
per second on IRQ18?=20

Thanks,

Darrell Burns=20
Senior Firmware Engineer=20
10757 S River Front Parkway, Suite 300
South Jordan, UT 84095

Direct: (801) 563-2023
Email:dburns1 at ibahn.com

-----Original Message-----
From: David Acker [mailto:dacker at roinet.com]=20
Sent: Friday, October 20, 2006 11:57 AM
To: Avila
Subject: Re: [Avila] irq27: nobody cared
Subject: [Avila] Tim Harvey wrote:
> What are the details of the kernel your using?  Exactly how often are
you seeing this error?  I haven't noticed this on my 2.6.15 kernel
patched (the irq27 that is, I've had no dealings with USB mass storage)
with IAL 1.4 other than occasionally when the driver unloads.
>=20
> Tim
>=20

We are using the kernel provided by gateworks but with newer versions of

madwifi and hostap.  I can get the "nobody cared" spurious interrupt=20
error on irq27 with enough radio traffic on minipci J1 and on irq28=20
enough usb storage.  Also, I can put an intel e100 minipci adapter in=20
minipci J1 instead of an atheros radio and get the same message for irq=20
27.  The rate of occurrence depends on the traffic pattern.  irq27 will=20
happen several times in an overnight test with data one wireless client=20
sending TCP to the avila board which is then forwarding to an ethernet=20
port.  Rates are around 20 Mbps.  With USB, irq28 will happens around=20
every 10-15th time we copy a 20 MB file from USB to a tmpfs filesystem=20
in a loop.

The "Badness in dma_free_coherent" appears to be a known bug that=20
started in 2.6.15.  This occurs more often, like every 3rd time I copy a

file, although sometimes I get multiple messages for one copy attempt.=20
There have been several changes in the arm dma and mm areas since=20
2.6.15.  This is why I would like to know which 2.6.15 kernel was used=20
as a base for the gateworks CD.

Oh, and while we are on the topic of interrupts...why is the intel=20
ethernet port hooked to irq18 which is supposed to be the performance=20
monitor according to the gateworks docs?  It is generating LOTS of=20
interrupts, like thousands per second..even if no ethernet cable is=20
plugged in.

bash-3.1# cat /proc/interrupts
            CPU0
   0:          0   osal
   1:        192   osal
   2:        191   osal
   3:          0   ixp400_eth
   5:      26053   IXP4xx Timer Tick
  15:        263   serial
  18:     969402   ixp400_eth
  27:      40444   wifi0
  28:          0   ehci_hcd:usb1, ohci_hcd:usb2, ohci_hcd:usb3
Err:          0
bash-3.1# cat /proc/uptime
264.34 186.88
bash-3.1

That's around 3,672 interrupts a second without having ethernet plugged
in!
-Ack

---------------------------------------------------------------------
To unsubscribe, e-mail: avila-unsubscribe at lists.unixstudios.net
For additional commands, e-mail: avila-help at lists.unixstudios.net





More information about the Avila mailing list