[PATCH 1/1] doc: virtual function MTU settings has no meaning

[Slava Ovsiienko]

> -----Original Message-----
> From: Stephen Hemminger <stephen at networkplumber.org>
> Sent: Monday, October 28, 2024 5:52 PM
> To: Slava Ovsiienko <viacheslavo at nvidia.com>
> Cc: dev at dpdk.org; Raslan Darawsheh <rasland at nvidia.com>; Matan Azrad
> <matan at nvidia.com>; Suanming Mou <suanmingm at nvidia.com>
> Subject: Re: [PATCH 1/1] doc: virtual function MTU settings has no meaning
> 
> On Mon, 28 Oct 2024 16:45:09 +0200
> Viacheslav Ovsiienko <viacheslavo at nvidia.com> wrote:
> 
> > There is the mlx5 NIC limitations - configuring MTU for PCI Virtual
> > Function has no meaning. The actual maximal packet size in VF's
> > receiving is limited by MTU configured on the related PCI Physical
> > Function, the DPDK datapath running over VF should be prepared to
> > handle packets of this maximal size.
> >
> > Signed-off-by: Viacheslav Ovsiienko <viacheslavo at nvidia.com>
> > ---
> >  doc/guides/nics/mlx5.rst | 7 +++++++
> >  1 file changed, 7 insertions(+)
> >
> > diff --git a/doc/guides/nics/mlx5.rst b/doc/guides/nics/mlx5.rst index
> > 8d1a1311d4..c7dcb74da7 100644
> > --- a/doc/guides/nics/mlx5.rst
> > +++ b/doc/guides/nics/mlx5.rst
> > @@ -191,6 +191,13 @@ Limitations
> >      - IPv4/TCP with CVLAN filtering
> >      - L4 steering rules for port RSS of UDP, TCP and IP
> >
> > +- PCI Virtual Function MTU:
> > +
> > +  Configuring MTU for PCI Virtual Function has no meaning.
> > +  The actual maximal packet size in VF's receiving is limited by MTU
> > + configured  on the related PCI Physical Function, the DPDK datapath
> > + running over VF should be  prepared to handle packets of this maximal
> size.
> > +
> 
> This is true of many drivers not just MLX5.
> 
> And it is generally true that Maximum Receive Unit (MRU) can be larger than
> Maximum Transmit Unit (MTU).
> 
> I would rather see a more precise definition of MTU in DPDK show up in
> ethdev documentation than sprinkling bits in each vendor driver.

There is the specifics for mlx5 - it operates in "bifurcated mode",
mlx5 PMD inherits/configures the MTU system settings, not the hardware ones  directly.
So, MTU naming looks like to be correct.

And, from time to time, we see user risen questions about VF MTU,
especially in virtualized environment. MTU is configured (either with system utilities
or with DPDK) and has no effect, embarrassing the users.

I have no objections to update the generic rte_ethdev chapter,
but it would require attention from other vendors as well.

With best regards,
Slava