netdev - Re: [PATCH net-next 06/10] udp: update sk_rmem

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <willemdebruijn.kernel.896e45dd59fc@gmail.com>
Date: Wed, 17 Sep 2025 15:14:04 -0400
From: Willem de Bruijn <willemdebruijn.kernel@...il.com>
To: Eric Dumazet <edumazet@...gle.com>, 
 Willem de Bruijn <willemdebruijn.kernel@...il.com>
Cc: "David S . Miller" <davem@...emloft.net>, 
 Jakub Kicinski <kuba@...nel.org>, 
 Paolo Abeni <pabeni@...hat.com>, 
 Simon Horman <horms@...nel.org>, 
 Willem de Bruijn <willemb@...gle.com>, 
 Kuniyuki Iwashima <kuniyu@...gle.com>, 
 David Ahern <dsahern@...nel.org>, 
 netdev@...r.kernel.org, 
 eric.dumazet@...il.com
Subject: Re: [PATCH net-next 06/10] udp: update sk_rmem_alloc before busylock
 acquisition

Eric Dumazet wrote:
> On Wed, Sep 17, 2025 at 8:01 AM Willem de Bruijn
> <willemdebruijn.kernel@...il.com> wrote:
> >
> > Eric Dumazet wrote:
> > > Avoid piling too many producers on the busylock
> > > by updating sk_rmem_alloc before busylock acquisition.
> > >
> > > Signed-off-by: Eric Dumazet <edumazet@...gle.com>
> > > ---
> > >  net/ipv4/udp.c | 7 +++++--
> > >  1 file changed, 5 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
> > > index edd846fee90ff7850356a5cb3400ce96856e5429..658ae87827991a78c25c2172d52e772c94ea217f 100644
> > > --- a/net/ipv4/udp.c
> > > +++ b/net/ipv4/udp.c
> > > @@ -1753,13 +1753,16 @@ int __udp_enqueue_schedule_skb(struct sock *sk, struct sk_buff *skb)
> > >       if (rmem > (rcvbuf >> 1)) {
> > >               skb_condense(skb);
> > >               size = skb->truesize;
> > > +             rmem = atomic_add_return(size, &sk->sk_rmem_alloc);
> > > +             if (rmem > rcvbuf)
> > > +                     goto uncharge_drop;
> >
> > This does more than just reorganize code. Can you share some context
> > on the behavioral change?
> 
> Sure : If we update sk_rmem_alloc sooner, before waiting 50usec+ on the busylock
> other cpus trying to push packets might see sk_rmem_alloc being too
> big already and exit early,
> before even trying to acquire the spinlock.
> 
> Say you have many cpus coming there.
> 
> Before the patch :
> 
> They all spin on busylock, then update sk_rmem_alloc one at a time
> (while they hold busylock)
> 
> After :
> 
> They update sk_rmem_alloc :
> if too big, they immediately drop and return, no need to take any lock.
> 
> If not too big, then they acquire the busylock.

I see. So this likely also overshoots the rcvbuf less?

As they currently all check against the sk_rcvbuf limit before
waiting (50usec+) on the busylock, and only update rmem and enqueue
after acquiring the lock. With no drop path at all (aside from
forward alloc).

Reviewed-by: Willem de Bruijn <willemb@...gle.com>