Re: [Xen-devel] [PATCH 3/5] setup_irq/request_irq/free_irq fixups

[Espen Skoglund <espen.skoglund@xxxxxxxxxxxxx>]

[Keir Fraser]
> On 06/02/2009 11:44, "Espen Skoglund" <espen.skoglund@xxxxxxxxxxxxx> wrote:
>>> My reading of the patch was that you simply pushed irq_to_vector()
>>> out into every caller. I therefore just took the ->shutdown() fix
>>> for free_irq() and dropped the rest. I checked in the rest of your
>>> patches.
>> 
>> Hmmm... yeah.  The patch probably didn't make this very clear since
>> some of the callers just invoke the function with the vector as the
>> parameter (and hence these callers were not included in the patch).
>> 
>> Anyhow, if you drop the vector parameter the IOMMU related
>> interrupts will most certainly not work since we no longer update
>> the irq_to_vector() array with those entries.

> Oh yes, also your patch would certainly have broken IA64, since you
> did not update its irq.c, nor does IA64 even have an irq_to_vector()
> macro.

> I suggest you do something like:
>  * Rename the functions to {setup,request,free}_irq_vector().
>  * Make {setup,request,free}_irq() be simple wrappers around the above.
>  * Only use the new function names for the IOMMU callers for now.
>  * Think about how to not break IA64.

> And make a new patch, against c/s 19180 or later.

Whoops.  Completely forgot about ia64 (my source grep command by
default filters out non-x86 stuff).  Your suggestions make perfect
sense.  Will post a new patch once c/s 19180 has been propagated to
xen-unstable then.

        eSk

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel