Re: [Xen-devel] [PATCH] xen: add support for MSI message groups

[Boris Ostrovsky <boris.ostrovsky@oracle.com>]

On 02/27/2014 10:45 AM, Roger Pau Monné wrote:
>>
>> @@ -291,7 +290,10 @@ static int xen_initdom_setup_msi_irqs(struct
>> pci_dev *dev, int nvec, int type)
>>                      (pci_domain_nr(dev->bus) << 16);
>>            map_irq.devfn = dev->devfn;
>>    -        if (type == PCI_CAP_ID_MSIX) {
>> +        if (type == PCI_CAP_ID_MSI && nvec > 1) {
>> +            map_irq.type = MAP_PIRQ_TYPE_MULTI_MSI;
>> +            map_irq.entry_nr = nvec;
>>
>> Are we overloading entry_nr here with a different meaning? I thought it
>> was meant to be entry number (in MSI-X table for example), not number of
>> entries.
> In the case of MSI message groups (MAP_PIRQ_TYPE_MULTI_MSI) entry_nr is
> the number of vectors to setup, so yes, it's an overloading of entry_nr.

Then I think we should at least make a note of this in physdev.h. (Or 
maybe even make entry_nr a union with nvec or some such, although that 
would look rather hacky).

> ....
>
>>
>> index 42721d1..eb13326d 100644
>> --- a/include/xen/interface/physdev.h
>> +++ b/include/xen/interface/physdev.h
>> @@ -131,6 +131,7 @@ struct physdev_irq {
>>    #define MAP_PIRQ_TYPE_GSI        0x1
>>    #define MAP_PIRQ_TYPE_UNKNOWN        0x2
>>    #define MAP_PIRQ_TYPE_MSI_SEG        0x3
>> +#define MAP_PIRQ_TYPE_MULTI_MSI        0x4
>> Formatting.
> I don't get the formatting problem, it's the same formatting that the
> other MAP_PIRQ_TYPE_* use, and if the patch is applied formatting is OK.
>

That's because my client messed up whitespaces. You can look for example 
at https://lkml.org/lkml/2014/2/26/352 to see the extra tab.

-boris