tcp: call sk_wmem_schedule before sk_mem_charge in zerocopy path
authorTalal Ahmad <talalahmad@google.com>
Fri, 9 Jul 2021 15:43:06 +0000 (11:43 -0400)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 25 Jul 2021 12:36:21 +0000 (14:36 +0200)
commit638632997c3173e41a7e5fb22d802d9bc0522fbf
treeeab1173eeafaf9a8fd3e250d7455a43b4711c47f
parent2fee3cf4c97b8913c0c907dd7bb1988baa8faa01
tcp: call sk_wmem_schedule before sk_mem_charge in zerocopy path

commit 358ed624207012f03318235017ac6fb41f8af592 upstream.

sk_wmem_schedule makes sure that sk_forward_alloc has enough
bytes for charging that is going to be done by sk_mem_charge.

In the transmit zerocopy path, there is sk_mem_charge but there was
no call to sk_wmem_schedule. This change adds that call.

Without this call to sk_wmem_schedule, sk_forward_alloc can go
negetive which is a bug because sk_forward_alloc is a per-socket
space that has been forward charged so this can't be negative.

Fixes: f214f915e7db ("tcp: enable MSG_ZEROCOPY")
Signed-off-by: Talal Ahmad <talalahmad@google.com>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Reviewed-by: Wei Wang <weiwan@google.com>
Reviewed-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/ipv4/tcp.c