Re: [PATCH] eni_vdpa: alibaba: select VIRTIO_PCI_LIB

From: Jason Wang
Date: Mon Dec 06 2021 - 03:41:22 EST


On Mon, Dec 6, 2021 at 4:14 PM Arnd Bergmann <arnd@xxxxxxxxxx> wrote:
>
> On Mon, Dec 6, 2021 at 4:12 AM Jason Wang <jasowang@xxxxxxxxxx> wrote:
> >
> > On Sat, Dec 4, 2021 at 2:55 AM Arnd Bergmann <arnd@xxxxxxxxxx> wrote:
> > >
> > > From: Arnd Bergmann <arnd@xxxxxxxx>
> > >
> > > When VIRTIO_PCI_LIB is not built-in but the alibaba driver is, the
> > > kernel runs into a link error:
> > >
> > > x86_64-linux-ld: drivers/vdpa/alibaba/eni_vdpa.o: in function `eni_vdpa_set_features':
> > > eni_vdpa.c:(.text+0x23f): undefined reference to `vp_legacy_set_features'
> > > x86_64-linux-ld: drivers/vdpa/alibaba/eni_vdpa.o: in function `eni_vdpa_set_vq_state':
> > > eni_vdpa.c:(.text+0x2fe): undefined reference to `vp_legacy_get_queue_enable'
> > > x86_64-linux-ld: drivers/vdpa/alibaba/eni_vdpa.o: in function `eni_vdpa_set_vq_address':
> > > eni_vdpa.c:(.text+0x376): undefined reference to `vp_legacy_set_queue_address'
> > > x86_64-linux-ld: drivers/vdpa/alibaba/eni_vdpa.o: in function `eni_vdpa_set_vq_ready':
> > > eni_vdpa.c:(.text+0x3b4): undefined reference to `vp_legacy_set_queue_address'
> > > x86_64-linux-ld: drivers/vdpa/alibaba/eni_vdpa.o: in function `eni_vdpa_free_irq':
> > > eni_vdpa.c:(.text+0x460): undefined reference to `vp_legacy_queue_vector'
> > > x86_64-linux-ld: eni_vdpa.c:(.text+0x4b7): undefined reference to `vp_legacy_config_vector'
> > > x86_64-linux-ld: drivers/vdpa/alibaba/eni_vdpa.o: in function `eni_vdpa_reset':
> >
> > Intersting, all those belongs to the legacy library.
> >
> > And I just have a try and I can complie alibaba eni without
> > VIRTIO_PCI_LIB is set.
>
> Ah, so the problem is in drivers/Makefile:
>
> obj-$(CONFIG_VIRTIO) += virtio/
> obj-$(CONFIG_VIRTIO_PCI_LIB) += virtio/
>
> We only enter this directory when one of these two symbols is set, but
> in my randconfig
> build, neither one is. I'll send a new patch.

Yes, we need to include when VIRTIO_PCI_LIB_LEGACY is set.

Thanks

>
> Arnd
>