On Fri, 13 Nov 2020 11:38:13 +0100 Paolo Abeni wrote:
> > Is there a chance someone could look into adding annotations to socket
> > locking?
>
> Annotating lock_sock_fast()/unlock_sock_fast() as they would
> unconditionally acquire/release the socket spinlock removes the warning
> related to fa
On Thu, 2020-11-12 at 15:12 -0800, Jakub Kicinski wrote:
> On Thu, 12 Nov 2020 15:08:31 -0800 Jakub Kicinski wrote:
> > On Thu, 12 Nov 2020 18:45:21 +0100 Paolo Abeni wrote:
> > > + skb = sk_stream_alloc_skb(sk, 0, sk->sk_allocation,
> > > + tcp_rtx_and_write_queues_
On Thu, 12 Nov 2020 15:08:31 -0800 Jakub Kicinski wrote:
> On Thu, 12 Nov 2020 18:45:21 +0100 Paolo Abeni wrote:
> > + skb = sk_stream_alloc_skb(sk, 0, sk->sk_allocation,
> > + tcp_rtx_and_write_queues_empty(sk));
>
> no good reason to misalign this AFAICT
Ma
On Thu, 12 Nov 2020 18:45:21 +0100 Paolo Abeni wrote:
> + skb = sk_stream_alloc_skb(sk, 0, sk->sk_allocation,
> + tcp_rtx_and_write_queues_empty(sk));
no good reason to misalign this AFAICT
Will be needed by the next patch, as MPTCP needs to handle
directly the error/memory-allocation-needed path.
No functional changes intended.
Additionally let MPTCP code access the tcp_remove_empty_skb()
helper.
Signed-off-by: Paolo Abeni
---
include/net/tcp.h | 3 ++
net/ipv4/tcp.c| 119
