On Tue, Mar 20, 2018 at 02:18:33PM +0000, Peter Maydell wrote: > On 20 March 2018 at 03:16, Michael S. Tsirkin <m...@redhat.com> wrote: > > Changes from v1: > > - dropped include change for one generated file - proposed a tree-wide > > refactoring > > - dropped vhost used slot refactoring due to alignment issues found by clang > > - added vhost-user post-copy support > > > > The following changes since commit 026aaf47c02b79036feb830206cfebb2a726510d: > > > > Merge remote-tracking branch > > 'remotes/ehabkost/tags/python-next-pull-request' into staging (2018-03-13 > > 16:26:44 +0000) > > > > are available in the git repository at: > > > > git://git.kernel.org/pub/scm/virt/kvm/mst/qemu.git tags/for_upstream > > > > for you to fetch changes up to a466e2cdb09d7b1262e24bae8cc47a51550d3af3: > > > > postcopy shared docs (2018-03-20 05:03:30 +0200) > > > > ---------------------------------------------------------------- > > virtio,vhost,pci,pc: features, cleanups > > > > SRAT tables for DIMM devices > > new virtio net flags for speed/duplex > > post-copy migration support in vhost > > cleanups in pci > > > > Signed-off-by: Michael S. Tsirkin <m...@redhat.com> > > > > Link failure on aarch64 host: > LINK arm-softmmu/qemu-system-arm > hw/virtio/vhost-user.o: In function `vhost_user_postcopy_fault_handler': > /home/pm215/qemu/hw/virtio/vhost-user.c:1000: undefined reference to > `postcopy_request_shared_page' > > I think this host is taking the "don't have userfaultfd" #ifdef > path in postcopy-ram.c, which doesn't define a stub > for the postcopy_request_shared_page function, but there's > no equivalent guard in vhost-user.c > > thanks > -- PMM
I'm curious why does it take that path though. Here's the rule #if defined(__linux__) && defined(__NR_userfaultfd) && defined(CONFIG_EVENTFD) If it's Linux we'll pick our own copy of the header and so __NR_userfaultfd is defined. CONFIG_EVENTFD is not set? Actually vhost-user does not work without CONFIG_EVENTFD either. Maybe we should skip building it in that configuration. -- MST
