Re: [Qemu-devel] [PULL 00/26] virtio, vhost, pci, pc: features, fixes and cleanups

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Feb 13, 2018 at 04:33:14PM +0000, Peter Maydell wrote:
> On 12 February 2018 at 09:35, Peter Maydell <peter.maydell@linaro.org> wrote:
> > This asserts in 'make check' for NetBSD, FreeBSD, OpenBSD, OSX:
> >
> > TEST: tests/device-introspect-test... (pid=19530)
> >   /aarch64/device/introspect/list:                                     OK
> >   /aarch64/device/introspect/list-fields:                              OK
> >   /aarch64/device/introspect/none:                                     OK
> >   /aarch64/device/introspect/abstract:                                 OK
> >   /aarch64/device/introspect/concrete:                                 **
> > ERROR:/root/qemu/qom/object.c:372:object_initialize_with_type:
> > assertion failed: (type != NULL)
> > Broken pipe
> > FAIL
> > GTester: last random seed: R02Sd4e2c04f6ac00d843a31ccac4de0d914
> > (pid=12686)
> >   /aarch64/device/introspect/abstract-interfaces:                      OK
> > FAIL: tests/device-introspect-test
> >
> > (other archs fail too, aarch64 is just the first one we hit). This
> > error is often "device A instantiates device B, but device B is
> > conditionally compiled in and A is always compiled; so test fails
> > trying to create device A on hosts where device B isn't built" I think.
> 
> This part is indeed that problem -- the 'virtio-crypto-device' object
> is built only if CONFIG_LINUX, but 'virtio-crypto-pci' is built if
> CONFIG_VIRTIO_PCI, so on systems where the latter is true but not the
> former you get a virtio-crypto-pci device that crashes when instantiated.
> 
> The fix should be
> --- a/hw/virtio/Makefile.objs
> +++ b/hw/virtio/Makefile.objs
> @@ -8,7 +8,7 @@ obj-y += virtio.o virtio-balloon.o
>  obj-$(CONFIG_LINUX) += vhost.o vhost-backend.o vhost-user.o
>  obj-$(CONFIG_VHOST_VSOCK) += vhost-vsock.o
>  obj-$(CONFIG_LINUX) += virtio-crypto.o
> -obj-$(CONFIG_VIRTIO_PCI) += virtio-crypto-pci.o
> +obj-$(call land,$(CONFIG_LINUX),$(CONFIG_VIRTIO_PCI)) += virtio-crypto-pci.o
>  endif
> 
>  common-obj-$(call lnot,$(CONFIG_LINUX)) += vhost-stub.o
> 
> 
> thanks
> -- PMM

Thanks for your help with this!

I think the root cause is that one of the patches disables
virtio-crypto build on non linux which used to be enabled.

This has been silently done by a patch which was supposed to
merely add a vhost crypto device, I'm sorry I missed that
in the review.

I've dropped the crypto vhost patches from the pull request for now.

Pushed with the same name - should be fine now.

-- 
MST