Re: [PATCH v2] xen: add support for MSI message groups

[David Vrabel <david.vrabel@citrix.com>]

On 28/02/14 18:36, Boris Ostrovsky wrote:
> On 02/28/2014 01:10 PM, Roger Pau Monné wrote:
>> On 28/02/14 19:00, Boris Ostrovsky wrote:
>>> On 02/28/2014 12:46 PM, Roger Pau Monné wrote:
>>>> On 28/02/14 18:20, Boris Ostrovsky wrote:
>>>>> On 02/27/2014 01:45 PM, Boris Ostrovsky wrote:
>>>>>> On 02/27/2014 01:15 PM, Roger Pau Monne wrote:
>>>>>>> Add support for MSI message groups for Xen Dom0 using the
>>>>>>> MAP_PIRQ_TYPE_MULTI_MSI pirq map type.
>>>>>>>
>>>>>>> In order to keep track of which pirq is the first one in the
>>>>>>> group all
>>>>>>> pirqs in the MSI group except for the first one have the newly
>>>>>>> introduced PIRQ_MSI_GROUP flag set. This prevents calling
>>>>>>> PHYSDEVOP_unmap_pirq on them, since the unmap must be done with the
>>>>>>> first pirq in the group.
>>>>>> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
>>>>>>
>>>>>>
>>>>> I was just looking at xen_setup_msi_irqs() (for a different reason)
>>>>> and
>>>>> I am no longer sure this patch does anything:
>>>>>
>>>>> static int xen_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
>>>>> {
>>>>>           int irq, ret, i;
>>>>>           struct msi_desc *msidesc;
>>>>>           int *v;
>>>>>
>>>>>           if (type == PCI_CAP_ID_MSI && nvec > 1)
>>>>>                   return 1;
>>>>> ....
>>>>>
>>>>> Same thing for xen_hvm_setup_msi_irqs().
>>>> As said in the commit message this is only for Dom0, so the function
>>>> modified is xen_initdom_setup_msi_irqs (were this check is removed).
>>> Then what is the reason for these changes:
>>>
>>> diff --git a/arch/x86/pci/xen.c b/arch/x86/pci/xen.c
>>> index 103e702..905956f 100644
>>> --- a/arch/x86/pci/xen.c
>>> +++ b/arch/x86/pci/xen.c
>>> @@ -178,6 +178,7 @@ static int xen_setup_msi_irqs(struct pci_dev *dev,
>>> int nvec, int type)
>>>       i = 0;
>>>       list_for_each_entry(msidesc, &dev->msi_list, list) {
>>>           irq = xen_bind_pirq_msi_to_irq(dev, msidesc, v[i],
>>> +                           (type == PCI_CAP_ID_MSI) ? nvec : 1,
>>>                              (type == PCI_CAP_ID_MSIX) ?
>>>                              "pcifront-msi-x" :
>>>                              "pcifront-msi",
>>> @@ -245,6 +246,7 @@ static int xen_hvm_setup_msi_irqs(struct pci_dev
>>> *dev, int nvec, int type)
>>>                   "xen: msi already bound to pirq=%d\n", pirq);
>>>           }
>>>           irq = xen_bind_pirq_msi_to_irq(dev, msidesc, pirq,
>>> +                           (type == PCI_CAP_ID_MSI) ? nvec : 1,
>>>                              (type == PCI_CAP_ID_MSIX) ?
>>>                              "msi-x" : "msi",
>>>                              DOMID_SELF);
>>>
>>> Should you simply pass 1?
>> Yes, but then if we implement MSI message groups for those cases we will
>> need to modify this line again, this way it's already correctly setup.
>> If you think it's best to hardcode it to 1, I can change it (I was also
>> in doubt about which way was better when modifying those lines).
> 
> 
> I think passing 1 explicitly this would be better. If we extend support
> for non-dom0 we would have to modify these routines anyway so making
> changes in both places simultaneously would make the commit more clear
> (IMO).

If we know now that this will need to be changed, it's better to do it
now than forget about it later.

Applied to devel/for-linus-3.15, thanks.

David