Re: [Patch bpf v2] skmsg: check sk_rcvbuf limit before queuing to ingress_skb

[John Fastabend <john.fastabend@gmail.com>]

Jakub Sitnicki wrote:
> On Thu, Jul 01, 2021 at 08:16 AM CEST, Cong Wang wrote:
> > From: Cong Wang <cong.wang@bytedance.com>
> >
> > Jiang observed OOM frequently when testing our AF_UNIX/UDP
> > proxy. This is due to the fact that we do not actually limit
> > the socket memory before queueing skb to ingress_skb. We
> > charge the skb memory later when handling the psock backlog,
> > but it is not limited either.
> >
> > This patch adds checks for sk->sk_rcvbuf right before queuing
> > to ingress_skb and drops packets if this limit exceeds. This
> > is very similar to UDP receive path. Ideally we should set the
> > skb owner before this check too, but it is hard to make TCP
> > happy about sk_forward_alloc.
> >
> > Reported-by: Jiang Wang <jiang.wang@bytedance.com>
> > Cc: Daniel Borkmann <daniel@iogearbox.net>
> > Cc: John Fastabend <john.fastabend@gmail.com>
> > Cc: Lorenz Bauer <lmb@cloudflare.com>
> > Cc: Jakub Sitnicki <jakub@cloudflare.com>
> > Signed-off-by: Cong Wang <cong.wang@bytedance.com>
> > ---
> 
> Acked-by: Jakub Sitnicki <jakub@cloudflare.com>
> 
> By saying that it is hard to make TCP happy about sk_forward_alloc, I'm
> guessing you're referring to problems described in 144748eb0c44 ("bpf,
> sockmap: Fix incorrect fwd_alloc accounting") [1]?

I have a couple fixes on my stack here I'm testing that clean up
the tear down logic. Once thats in place maybe its as simple
as adding the owner_r bits and calling the destructor to ensure
memory accounting happens earlier so these ingress_skb packets
are accounted for.

I'll flush those out today, maybe it will be clear then.

> 
> Thanks for the fix.
> 
> [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=144748eb0c445091466c9b741ebd0bfcc5914f3d
> 
> [...]