Re: [PATCH net-next v8 4/4] netvsc: refactor notifier/event handling code to use the failover framework

["Samudrala, Sridhar" <sridhar.samudrala@intel.com>]

On 4/26/2018 5:09 PM, Michael S. Tsirkin wrote:
> On Thu, Apr 26, 2018 at 03:33:26PM -0700, Stephen Hemminger wrote:
> > On Thu, 26 Apr 2018 05:30:05 +0300
> > "Michael S. Tsirkin" <mst@redhat.com> wrote:
> >
> > > On Wed, Apr 25, 2018 at 05:08:37PM -0700, Stephen Hemminger wrote:
> > > > On Wed, 25 Apr 2018 16:59:28 -0700
> > > > Sridhar Samudrala <sridhar.samudrala@intel.com> wrote:
> > > >    
> > > > > Use the registration/notification framework supported by the generic
> > > > > failover infrastructure.
> > > > >
> > > > > Signed-off-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
> > > > NAK unless you prove this works on legacy distributions and with DPDK 18.05
> > > > without modification.
> > > It looks like it should work. What kind of proof are you looking for?
> >
> > I tried this with working Ubuntu 17 on WS2016.
> > It boots if the failover driver is configured in (as module).
> Oh so by "unless you prove" you meant "unless you test"?
>
> > But if the configuration has:
> >
> > $ grep FAILOVER .config
> > # CONFIG_NET_FAILOVER is not set
> > CONFIG_MAY_USE_NET_FAILOVER=y
> So the new option works, what's broken is when it's *not* selected.
> Looks like a bug.

Yes.  Looks like i need to make NET_FAILOVER  to be enabled automatically when
VIRTIO_NET or HYPERV_NET are selected.

Thanks Stephen for confirming that it works when NET_FAILOVER is enabled.



> > The netvsc driver fails on boot with:
> >
> > [    0.826447] hv_vmbus: registering driver hv_netvsc
> > [    0.829616] scsi 0:0:0:0: Direct-Access     Msft     Virtual Disk     1.0  PQ: 0 ANSI: 5
> > [    0.836291] input: Microsoft Vmbus HID-compliant Mouse as /devices/0006:045E:0621.0001/input/input1
> > [    0.839139] hid-generic 0006:045E:0621.0001: input: <UNKNOWN> HID v0.01 Mouse [Microsoft Vmbus HID-compliant Mouse] on
> > [    0.964897] hv_vmbus: probe failed for device 849a776e-8120-4e4a-9a36-7e3d95ac75b3 (-95)
> > [    0.968039] hv_netvsc: probe of 849a776e-8120-4e4a-9a36-7e3d95ac75b3 failed with error -95
> > [    1.112877] hv_vmbus: probe failed for device 53557f8e-057d-425b-9265-01c0fd7e273e (-95)
> > [    1.116064] hv_netvsc: probe of 53557f8e-057d-425b-9265-01c0fd7e273e failed with error -95
> >
> > The system has two virtual networks. eth0 is on vswitch for management.
> > eth1 is on vswitch with SR-IOV for performance tests.
> >
> > You probably need to just put the failover part in net/core and select it.
>  From all drivers. Yes. And it does not need to be visible in the menu
> imho.
>
>
> > It is trivial to get an evaluation version of Windows Server 2016 and setup a Linux VM.
> > Please try it.
> I presume some kind of test was done here. Sridhar, do you think you
> could include some notes about testing of this patch? Or in case you
> only build-tested this part, it's a good idea to notice this after ---
> in the patch, or in the cover letter.

I only did compile and build test of netvsc. I don't have a hyperv setup to test it out.
Would appreciate if Stephen or someone who has this setup will test it out when i
submit the next revision with the config fixes.

>