[libvirt] [PATCH 04/15] Code to return interface name or pci_addr of the VF in actualDevice

[Laine Stump]

On 08/15/2012 06:47 AM, Shradha Shah wrote:
> On 08/14/2012 06:36 AM, Laine Stump wrote:
>> On 08/10/2012 12:23 PM, Shradha Shah wrote:
>>> The network pool should be able to keep track of both, network device
>>> names nad PCI addresses, and return the appropriate one in the actualDevice
>>> when networkAllocateActualDevice is called.
>>>
>>> Signed-off-by: Shradha Shah <sshah at solarflare.com>
>>> ---
>>>  src/network/bridge_driver.c |   33 +++++++++++++++++++++++++++------
>>>  src/util/virnetdev.c        |   25 ++++++++++++-------------
>>>  src/util/virnetdev.h        |    4 +++-
>>>  3 files changed, 42 insertions(+), 20 deletions(-)
>>>
>>> diff --git a/src/network/bridge_driver.c b/src/network/bridge_driver.c
>>> index df3cc25..602e17d 100644
>>> --- a/src/network/bridge_driver.c
>>> +++ b/src/network/bridge_driver.c
>>> @@ -59,6 +59,7 @@
>>>  #include "dnsmasq.h"
>>>  #include "configmake.h"
>>>  #include "virnetdev.h"
>>> +#include "pci.h"
>>>  #include "virnetdevbridge.h"
>>>  #include "virnetdevtap.h"
>>>  
>>> @@ -2737,10 +2738,11 @@ static int
>>>  networkCreateInterfacePool(virNetworkDefPtr netdef) {
>>>      unsigned int num_virt_fns = 0;
>>>      char **vfname = NULL;
>>> +    struct pci_config_address **virt_fns;
>>>      int ret = -1, ii = 0;
>>>      
>>>      if ((virNetDevGetVirtualFunctions(netdef->forwardPfs->dev,
>>> -                                      &vfname, &num_virt_fns)) < 0) {
>>> +                                      &vfname, &virt_fns, &num_virt_fns)) < 0) { 
>>>          virReportError(VIR_ERR_INTERNAL_ERROR,
>>>                         _("Could not get Virtual functions on %s"),
>>>                         netdef->forwardPfs->dev);
>>> @@ -2762,19 +2764,38 @@ networkCreateInterfacePool(virNetworkDefPtr netdef) {
>>>      netdef->nForwardIfs = num_virt_fns;
>>>      
>>>      for (ii = 0; ii < netdef->nForwardIfs; ii++) {
>>> -        netdef->forwardIfs[ii].device.dev = strdup(vfname[ii]);
>>> -        if (!netdef->forwardIfs[ii].device.dev) {
>>> -            virReportOOMError();
>>> -            goto finish;
>> To be pure in the separation of patches, the following if else should be
>> removed from this patch, with just the contents of the "if" clause here.
>> Then the if else + body of the else should be added in the next patch.
>>
>> (And at any rate, the if() condition is incorrect here - really that
>> part should happen for all forwardTypes except HOSTDEV (BRIDGE, PRIVATE,
>> and VEPA also require netdev names.)
> I did not include the BRIDGE, PRIVATE and VEPA cases here because the networkCreateInterfacePool
> function is not called in those cases.
>
> Should I still include the conditions for BRIDGE, PRIVATE and VEPA?

Ah. It *should* be called for those cases as well. I hadn't noticed that
it wasn't. Those modes can also benefit from auto-creating the list of
devices.