[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] x86/IO-APIC: refine EOI-ing of migrating level interrupts



>>> On 15.11.11 at 14:35, Andrew Cooper <andrew.cooper3@xxxxxxxxxx> wrote:
> On 15/11/11 13:27, Jan Beulich wrote:
>>>>> On 15.11.11 at 14:19, Andrew Cooper <andrew.cooper3@xxxxxxxxxx> wrote:
>>> On 15/11/11 13:14, Jan Beulich wrote:
>>>>      if ( ioapic_has_eoi_reg(apic) )
>>>>      {
>>>>          /* If vector is unknown, read it from the IO-APIC */
>>>> -        if ( vector == -1 )
>>>> +        if ( vector == IRQ_VECTOR_UNASSIGNED )
>>> Quick style query:  I consider IRQ_VECTOR_UNASSIGNED logically different
>>> from passing -1 in as a value for vector, even though they are the are
>>> the same value.  Is it sensible to mix them?
>> I view it quite the other way around: One should explicitly pass
>> IRQ_VECTOR_UNASSIGNED when passing a literal value (which
>> currently doesn't happen anyway. Primarily because passing
>> desc->arch.vector or desc->arch.old_vector could happen to also
>> hold this very value.
>>
>> Jan
> 
> Ok.
> 
> Do you want any other patches to be tested on the problem server?

While not directly related, throwing in "x86/IRQ: prevent vector
sharing within IO-APICs" (which I want to apply only after this one)
would certainly be good.

Plus of course the experimental, yet to be written, use of
desc->arch.old_vector in end_level_io_apic_irq() (which anyway
should be done only after a good run with the above included).

Jan


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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.