Re: irq 21: nobody cared 2.6.24-rc1

[Mark Lord <lkml@rtr.ca>]

Bongani Hlope wrote:
> On Thursday 01 November 2007 22:32:38 Andrew Morton wrote:
>> On Thu, 25 Oct 2007 10:45:36 +0200
>>
>> Bongani Hlope <bonganilinux@mweb.co.za> wrote:
>>> Booting with irqpoll works
>>>
>>> ls /proc/irq/21/ (with irqpoll)
>>> ehci_hcd:usb1/  smp_affinity  uhci_hcd:usb2/  uhci_hcd:usb3/ 
>>> uhci_hcd:usb4/
>>>
>>>  Disabling IRQ #21
>> Was any earlier kernel version OK?  2.6.23?
> 
> Yes the 2.6.23 kernel works fine. This seems to happen when I boot the 
> 2.6.24-rc1 kernel with an iPod attached (hope this helps), but the 2.6.23 
> kernel doesn't experience this problem.
..

Something in 2.6.24-rc* seems to be a lot more fussy about this kind
of situation than in 2.6.23.   I have a driver (sata_qstor) that manages
a piece of hardware that unfortunately generates spurious interrupts
regularly.

With 2.6.23, it was only rarely bad enough to cause the "Disabling IRQ #nn"
event to occur.  But in 2.6.24-rc*, it happens almost immediately now
(so, yes, I've gone and put a workaround into the driver for it).

Something's different.