On Mon, Jul 19, 2021 at 10:36:34PM +0200, Pankaj Gupta wrote:
> > Update virtio-pmem to allow for the pmem region to be specified in either
> > guest absolute terms or as a PCI BAR-relative address. This is required
> > to support virtio-pmem in Hyper-V, since Hyper-V only allows PCI devices
> > to operate on PCI memory ranges defined via BARs.
> >
> > Virtio-pmem will check for a shared memory window and use that if found,
> > else it will fallback to using the guest absolute addresses in
> > virtio_pmem_config. This was chosen over defining a new feature bit,
> > since it's similar to how virtio-fs is configured.
> >
> > Signed-off-by: Taylor Stark <[email protected]>
> > ---
> > drivers/nvdimm/virtio_pmem.c | 21 +++++++++++++++++----
> > drivers/nvdimm/virtio_pmem.h | 3 +++
> > 2 files changed, 20 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/nvdimm/virtio_pmem.c b/drivers/nvdimm/virtio_pmem.c
> > index 726c7354d465..43c1d835a449 100644
> > --- a/drivers/nvdimm/virtio_pmem.c
> > +++ b/drivers/nvdimm/virtio_pmem.c
> > @@ -37,6 +37,8 @@ static int virtio_pmem_probe(struct virtio_device *vdev)
> > struct virtio_pmem *vpmem;
> > struct resource res;
> > int err = 0;
> > + bool have_shm_region;
> > + struct virtio_shm_region pmem_region;
> >
> > if (!vdev->config->get) {
> > dev_err(&vdev->dev, "%s failure: config access disabled\n",
> > @@ -58,10 +60,21 @@ static int virtio_pmem_probe(struct virtio_device *vdev)
> > goto out_err;
> > }
> >
> > - virtio_cread_le(vpmem->vdev, struct virtio_pmem_config,
> > - start, &vpmem->start);
> > - virtio_cread_le(vpmem->vdev, struct virtio_pmem_config,
> > - size, &vpmem->size);
> > + /* Retrieve the pmem device's address and size. It may have been
> > supplied
> > + * as a PCI BAR-relative shared memory region, or as a guest
> > absolute address.
> > + */
> > + have_shm_region = virtio_get_shm_region(vpmem->vdev, &pmem_region,
> > +
> > VIRTIO_PMEM_SHMCAP_ID_PMEM_REGION);
>
> Current implementation of Virtio pmem device in Qemu does not expose
> it as PCI BAR.
> So, can't test it. Just curious if device side implementation is also
> tested for asynchronous
> flush case?
>
> Thanks,
> Pankaj
Yes, I tested the async flush case as well. We basically call
FlushFileBuffers on the backing file, which is Windows' equivalent of
fsync. I also briefly tested with qemu to ensure that still works with
the patch.
Thanks,
Taylor
> > +
> > + if (have_shm_region) {
> > + vpmem->start = pmem_region.addr;
> > + vpmem->size = pmem_region.len;
> > + } else {
> > + virtio_cread_le(vpmem->vdev, struct virtio_pmem_config,
> > + start, &vpmem->start);
> > + virtio_cread_le(vpmem->vdev, struct virtio_pmem_config,
> > + size, &vpmem->size);
> > + }
> >
> > res.start = vpmem->start;
> > res.end = vpmem->start + vpmem->size - 1;
> > diff --git a/drivers/nvdimm/virtio_pmem.h b/drivers/nvdimm/virtio_pmem.h
> > index 0dddefe594c4..62bb564e81cb 100644
> > --- a/drivers/nvdimm/virtio_pmem.h
> > +++ b/drivers/nvdimm/virtio_pmem.h
> > @@ -50,6 +50,9 @@ struct virtio_pmem {
> > __u64 size;
> > };
> >
> > +/* For the id field in virtio_pci_shm_cap */
> > +#define VIRTIO_PMEM_SHMCAP_ID_PMEM_REGION 0
> > +
> > void virtio_pmem_host_ack(struct virtqueue *vq);
> > int async_pmem_flush(struct nd_region *nd_region, struct bio *bio);
> > #endif
> > --
> > 2.32.0
> >
> >
