Re: [PATCH net 2/5] hv_netvsc: Use vmbus_sendpacket_mpb_desc() to send VMBus messages

[Simon Horman <horms@kernel.org>]

On Wed, May 14, 2025 at 03:44:35PM +0000, Michael Kelley wrote:
> From: Simon Horman <horms@kernel.org> Sent: Wednesday, May 14, 2025 2:38 AM
> > 
> > On Mon, May 12, 2025 at 05:06:01PM -0700, mhkelley58@gmail.com wrote:
> > > From: Michael Kelley <mhklinux@outlook.com>
> > >
> > > netvsc currently uses vmbus_sendpacket_pagebuffer() to send VMBus
> > > messages. This function creates a series of GPA ranges, each of which
> > > contains a single PFN. However, if the rndis header in the VMBus
> > > message crosses a page boundary, the netvsc protocol with the host
> > > requires that both PFNs for the rndis header must be in a single "GPA
> > > range" data structure, which isn't possible with
> > > vmbus_sendpacket_pagebuffer(). As the first step in fixing this, add a
> > > new function netvsc_build_mpb_array() to build a VMBus message with
> > > multiple GPA ranges, each of which may contain multiple PFNs. Use
> > > vmbus_sendpacket_mpb_desc() to send this VMBus message to the host.
> > >
> > > There's no functional change since higher levels of netvsc don't
> > > maintain or propagate knowledge of contiguous PFNs. Based on its
> > > input, netvsc_build_mpb_array() still produces a separate GPA range
> > > for each PFN and the behavior is the same as with
> > > vmbus_sendpacket_pagebuffer(). But the groundwork is laid for a
> > > subsequent patch to provide the necessary grouping.
> > >
> > > Cc: <stable@vger.kernel.org> # 6.1.x
> > > Signed-off-by: Michael Kelley <mhklinux@outlook.com>
> > > ---
> > >  drivers/net/hyperv/netvsc.c | 50 +++++++++++++++++++++++++++++++++----
> > >  1 file changed, 45 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
> > > index d6f5b9ea3109..6d1705f87682 100644
> > > --- a/drivers/net/hyperv/netvsc.c
> > > +++ b/drivers/net/hyperv/netvsc.c
> > > @@ -1055,6 +1055,42 @@ static int netvsc_dma_map(struct hv_device *hv_dev,
> > >  	return 0;
> > >  }
> > >
> > > +/* Build an "array" of mpb entries describing the data to be transferred
> > > + * over VMBus. After the desc header fields, each "array" entry is variable
> > > + * size, and each entry starts after the end of the previous entry. The
> > > + * "offset" and "len" fields for each entry imply the size of the entry.
> > > + *
> > > + * The pfns are in HV_HYP_PAGE_SIZE, because all communication with Hyper-V
> > > + * uses that granularity, even if the system page size of the guest is larger.
> > > + * Each entry in the input "pb" array must describe a contiguous range of
> > > + * guest physical memory so that the pfns are sequential if the range crosses
> > > + * a page boundary. The offset field must be < HV_HYP_PAGE_SIZE.
> > 
> > Hi Michael,
> > 
> > Is there a guarantee that this constraint is met. And moreover, is there a
> > guarantee that all of the entries will fit in desc? I am slightly concerned
> > that there may be an overrun lurking here.
> > 
> 
> It is indeed up to the caller to ensure that the pb array is properly
> constructed. netvsc_build_mpb_array() doesn't do additional validation.
> There are only two sources of the pb array, both of which do the right
> thing, so additional validation seemed redundant.
> 
> An overrun is a concern, but again the callers do the right thing. As
> described in my response to Patch 3 of the series, netvsc_xmit()
> counts the number of pages ahead of time, and makes sure the count is
> within the limit of the amount space allocated in the "desc" argument
> to netvsc_build_mpb_array().

Thanks Michael,

I agree that is entirely reasonable for callers to be responsible
correctly constructing the pb array. And that it's not necessary
to add validation to netvsc_build_mpb_array().

Also, based on the above, I'm satisfied that the callers are correctly
constructing the pb array.

With the above clarified in my mind I'm now happy with this patch.

Reviewed-by: Simon Horman <horms@kernel.org>