Re: [PATCH net-next v2 1/3] xsk: Remove non-zero 'dma_page' check in xp_assign_dev

[Magnus Karlsson]

On Thu, 29 Feb 2024 at 13:52, wangyunjian <wangyunjian@xxxxxxxxxx> wrote:
>
> > -----Original Message-----
> > From: Paolo Abeni [mailto:pabeni@xxxxxxxxxx]
> > Sent: Thursday, February 29, 2024 6:43 PM
> > To: wangyunjian <wangyunjian@xxxxxxxxxx>; mst@xxxxxxxxxx;
> > willemdebruijn.kernel@xxxxxxxxx; jasowang@xxxxxxxxxx; kuba@xxxxxxxxxx;
> > bjorn@xxxxxxxxxx; magnus.karlsson@xxxxxxxxx; maciej.fijalkowski@xxxxxxxxx;
> > jonathan.lemon@xxxxxxxxx; davem@xxxxxxxxxxxxx
> > Cc: bpf@xxxxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx;
> > linux-kernel@xxxxxxxxxxxxxxx; kvm@xxxxxxxxxxxxxxx;
> > virtualization@xxxxxxxxxxxxxxx; xudingke <xudingke@xxxxxxxxxx>; liwei (DT)
> > <liwei395@xxxxxxxxxx>
> > Subject: Re: [PATCH net-next v2 1/3] xsk: Remove non-zero 'dma_page' check in
> > xp_assign_dev
> >
> > On Wed, 2024-02-28 at 19:05 +0800, Yunjian Wang wrote:
> > > Now dma mappings are used by the physical NICs. However the vNIC maybe
> > > do not need them. So remove non-zero 'dma_page' check in
> > > xp_assign_dev.
> > >
> > > Signed-off-by: Yunjian Wang <wangyunjian@xxxxxxxxxx>
> > > ---
> > >  net/xdp/xsk_buff_pool.c | 7 -------
> > >  1 file changed, 7 deletions(-)
> > >
> > > diff --git a/net/xdp/xsk_buff_pool.c b/net/xdp/xsk_buff_pool.c index
> > > ce60ecd48a4d..a5af75b1f43c 100644
> > > --- a/net/xdp/xsk_buff_pool.c
> > > +++ b/net/xdp/xsk_buff_pool.c
> > > @@ -219,16 +219,9 @@ int xp_assign_dev(struct xsk_buff_pool *pool,
> > >     if (err)
> > >             goto err_unreg_pool;
> > >
> > > -   if (!pool->dma_pages) {
> > > -           WARN(1, "Driver did not DMA map zero-copy buffers");
> > > -           err = -EINVAL;
> > > -           goto err_unreg_xsk;
> > > -   }
> >
> > This would unconditionally remove an otherwise valid check for most NIC. What
> > about let the driver declare it wont need DMA map with a
> > (pool?) flag.
>
> This check is redundant. The NIC's driver determines whether a DMA map is required.
> If the NIC'driver requires the DMA map, it uses the xsk_pool_dma_map function, which
> initializes the DMA map and performs a check.

Just to provide some context: I put this check there many years ago to
guard against a zero-copy driver writer forgetting to call
xsk_pool_dma_map() during the implementation phase. A working driver
will always have pool->dma_pages != NULL. If you both think that this
check is too much of a precaution, then I have no problem getting rid
of it. Just thought that a text warning would be nicer than a crash
later.

Thanks: Magnus

> Thanks
>
> >
> > Cheers,
> >
> > Paolo
>