Re: [Qemu-devel] [PATCH v4 2/9] util: Introduce vfio helpers

[Fam Zheng <famz@redhat.com>]

On Wed, 01/10 13:09, Alex Williamson wrote:
> On Wed, 10 Jan 2018 17:18:39 +0800
> Fam Zheng <famz@redhat.com> wrote:
> 
> > This is a library to manage the host vfio interface, which could be used
> > to implement userspace device driver code in QEMU such as NVMe or net
> > controllers.
> > 
> > Signed-off-by: Fam Zheng <famz@redhat.com>
> > ---
> >  include/qemu/vfio-helpers.h |  30 ++
> >  util/Makefile.objs          |   1 +
> >  util/trace-events           |  11 +
> >  util/vfio-helpers.c         | 723 ++++++++++++++++++++++++++++++++++++++++++++
> >  4 files changed, 765 insertions(+)
> >  create mode 100644 include/qemu/vfio-helpers.h
> >  create mode 100644 util/vfio-helpers.c
> > 
> > diff --git a/include/qemu/vfio-helpers.h b/include/qemu/vfio-helpers.h
> > new file mode 100644
> > index 0000000000..6bdba3b66e
> > --- /dev/null
> > +++ b/include/qemu/vfio-helpers.h
> ...
> > +/**
> > + * Map a PCI bar area.
> > + */
> > +void *qemu_vfio_pci_map_bar(QEMUVFIOState *s, int index, Error **errp)
> > +{
> > +    void *p;
> > +    assert_bar_index_valid(s, index);
> > +    p = mmap(NULL, MIN(8192, s->bar_region_info[index].size),
> > +             PROT_READ | PROT_WRITE, MAP_SHARED,
> > +             s->device, s->bar_region_info[index].offset);
> > +    if (p == MAP_FAILED) {
> > +        error_setg_errno(errp, errno, "Failed to map BAR region");
> > +        p = NULL;
> > +    }
> > +    return p;
> > +}
> > +
> > +/**
> > + * Unmap a PCI bar area.
> > + */
> > +void qemu_vfio_pci_unmap_bar(QEMUVFIOState *s, int index, void *bar)
> > +{
> > +    if (bar) {
> > +        munmap(bar, MIN(8192, s->bar_region_info[index].size));
> > +    }
> > +}
> 
> What's up with this 8KB thing?  Is it perhaps a hack to avoid
> un-mmap'able MSI-X sections of the BAR, which would make this general
> purpose library very specific to devices which only operate in the lower
> 8KB of their MMIO space.  Maybe the interface should have an offset and
> size so that the NVMe driver could implement that dependency.  We could
> also be testing if the region supports mmap, but I suppose trying and
> failing is just as good.  Thanks,

Sounds good, will add the offset and size parameters.

Fam