Re: [PATCH 3/4] Allow vif= to specify PCI address for each nic

[Don Slutz <don.slutz@gmail.com>]

On 06/16/15 06:32, Wei Liu wrote:
> On Mon, Jun 15, 2015 at 01:45:26PM -0400, Don Slutz wrote:
>> On 06/15/15 11:54, Wei Liu wrote:
>>> On Mon, Jun 15, 2015 at 10:15:51AM -0400, Don Slutz wrote:
>>>> This allows more then 32 nics.
>>>>
>>>
>>> How does this patch help? What prevents you from having more than 32
>>> nics?
>>>
>>
>> Without a way to put the emulated nics onto a PCI to PCI Bridge, you are
>> limited to 1 PCI bus (the host one).
>>
>> A PCI bus has a max of 32 PCI devices.
>>
>> So, since the host bus has some PCI devices that are not nics, the limit
>> is smaller then 32.
>>
> 
> Is there anything that you can't accomplish by using
> device_model_args_hvm?
> 

Yes, you cannot use xen-netfront/xen-netback (PV access).

>>
>>>> This can help with Windows finding nics at boot time.
>>>>
>>>> This allows changing config file:
>>>>
>>>>     builder = "hvm"
>>>>     device_model_args_hvm = [
>>>>      "-device",
>>>>      "pci-bridge,chassis_nr=2,msi=on,id=pciBridge5.0,multifunction=on,addr=0x15.0",
>>>>      "-device",
>>>>      "vmxnet3,id=nic3,netdev=net3,mac=00:0c:29:86:44:be,bus=pciBridge5.0,addr=0x4.0x0",
>>>>      "-netdev",
>>>>      "type=tap,id=net3,ifname=vif.3-emu,script=/etc/qemu-ifup,downscript=no",
>>>>     ]
>>>>     vif = [
>>>>     ]
>>>>
>>>> to:
>>>>
>>>>     builder = "hvm"
>>>>     device_model_args_hvm = [
>>>>      "-device",
>>>>      "pci-bridge,chassis_nr=2,msi=on,id=pciBridge5.0,multifunction=on,addr=0x15.0",
>>>>     ]
>>>>     vif = [
>>>>      "model=vmxnet3,bridge=xenbr0,mac=00:0c:29:86:44:a0,bus=pciBridge5.0,addr=0x4.0x0",
>>>>     ]
>>>>
> 
> The way that you construct PCI bridge via device_model_args_hvm while
> plumbing some other information via vif= is fragile and not general
> useful to others. I don't think I would accept a half-baked solution
> like this.
> 

Ok.  Will look to adding a pcibridge also.

>>>> which enables usage of xen-netback.
>>>>
>>>
>>> In any case, exposing HVM-only options to top-level vif configuration
>>> space doesn't look right.
>>
>>
>> There are already HVM-only options in vifs:
>>
>> ### type
>>
>> This keyword is valid for HVM guests only.
>> ...
>> ### model
>>
>> This keyword is valid for HVM guest devices with `type=ioemu` only.
>> ...
>>
>>> Why do you want to set bus and addr? The
>>> rationale should be stated in commit message.
>>
>>
>> That is why I said:
>>
>>>> This can help with Windows finding nics at boot time.
>>
>> Windows boot code is not as flexible as Linux.  Most versions of Windows
>> like to blue screen if the hardware changes enough.
>>
> 
> Looks like you're trying to migrate a guest from VMWare to Xen. If
> device_model_args_new is sufficient please just use that.
> 

It works, but slowly because you are prevented from using the faster PV
nics.

   -Don Slutz

> Wei.
>