Virtualization - Feb 2021 - [PATCH v7 bpf-next 6/6] xsk: build skb by page (aka generic zerocopy xmit)

Xuan Zhuo wrote:
> On Wed, 17 Feb 2021 16:46:04 -0800, John Fastabend <john.fastabend at
gmail.com> wrote:
> > Alexander Lobakin wrote:
> > > From: Xuan Zhuo <xuanzhuo at linux.alibaba.com>
> > >
> > > This patch is used to construct skb based on page to save memory
copy
> > > overhead.
> > >
> > > This function is implemented based on IFF_TX_SKB_NO_LINEAR. Only
the
> > > network card priv_flags supports IFF_TX_SKB_NO_LINEAR will use
page to
> > > directly construct skb. If this feature is not supported, it is
still
> > > necessary to copy data to construct skb.
> > >
> > > ---------------- Performance Testing ------------
> > >
> > > The test environment is Aliyun ECS server.
> > > Test cmd:
> > > ```
> > > xdpsock -i eth0 -t  -S -s <msg size>
> > > ```
> > >
> > > Test result data:
> > >
> > > size    64      512     1024    1500
> > > copy    1916747 1775988 1600203 1440054
> > > page    1974058 1953655 1945463 1904478
> > > percent 3.0%    10.0%   21.58%  32.3%
> > >
> > > Signed-off-by: Xuan Zhuo <xuanzhuo at linux.alibaba.com>
> > > Reviewed-by: Dust Li <dust.li at linux.alibaba.com>
> > > [ alobakin:
> > >  - expand subject to make it clearer;
> > >  - improve skb->truesize calculation;
> > >  - reserve some headroom in skb for drivers;
> > >  - tailroom is not needed as skb is non-linear ]
> > > Signed-off-by: Alexander Lobakin <alobakin at pm.me>
> > > Acked-by: Magnus Karlsson <magnus.karlsson at intel.com>
> > > ---
> >
> > [...]
> >
> > > +	buffer = xsk_buff_raw_get_data(pool, addr);
> > > +	offset = offset_in_page(buffer);
> > > +	addr = buffer - pool->addrs;
> > > +
> > > +	for (copied = 0, i = 0; copied < len; i++) {
> > > +		page = pool->umem->pgs[addr >> PAGE_SHIFT];
> >
> > Looks like we could walk off the end of pgs[] if len is larger than
> > the number of pgs? Do we need to guard against a misconfigured socket
> > causing a panic here? AFAIU len here is read from the user space
> > descriptor so is under user control. Or maybe I missed a check
somewhere.
> >
> > Thanks,
> > John
> >
> 
> Don't worry about this, the legality of desc has been checked.
> 
> xskq_cons_peek_desc -> xskq_cons_read_desc ->
>                    xskq_cons_is_valid_desc -> xp_validate_desc
Ah OK I didn't dig past the cons_read_desc(). In that case LGTM.

Acked-by: John Fastabend <john.fastabend at gmail.com>