On Tue, Oct 27, 2015 at 01:51:20PM +0200, Michael S. Tsirkin wrote:
> On Tue, Oct 27, 2015 at 10:47:54AM +0200, Michael S. Tsirkin wrote:
> > TL;DR:
> > This fixes virtio in a way transparent to guest.
> > We should now be able to revert commits aa8580cd and df0acded19ec which
> > worked
> > around it in a way that's not transparent.
>
> I didn't check dataplane BTW. Igor? Stefan?
It doesn't handle it:
/* TODO handle non-contiguous memory across region boundaries */
iov->iov_base = vring_map(&mr, desc->addr, desc->len,
desc->flags & VRING_DESC_F_WRITE);
if (!iov->iov_base) {
error_report("Failed to map descriptor addr %#" PRIx64 " len %u",
(uint64_t)desc->addr, desc->len);
return -EFAULT;
}
The assumption is that vring_map() can map the full desc->len bytes.
The same logic that you're trying to add in this series could be added to
vring_pop() but it makes the number of sg list entries accepted by QEMU
variable and dependent on DIMMs :(. Previously guests could be sure that 1024
sgs work.
