net: add and use skb_unclone_keeptruesize() helper
authorEric Dumazet <edumazet@google.com>
Tue, 2 Nov 2021 00:45:55 +0000 (17:45 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 25 Nov 2021 08:49:08 +0000 (09:49 +0100)
commitfa5f8606353fd09af052d609e7988ee26fc1f2d5
treec3999a48f51724785ebc46e6e6b6646e12ba6235
parentf14c857331548f3b33000db8028c58d8c313f777
net: add and use skb_unclone_keeptruesize() helper

commit c4777efa751d293e369aec464ce6875e957be255 upstream.

While commit 097b9146c0e2 ("net: fix up truesize of cloned
skb in skb_prepare_for_shift()") fixed immediate issues found
when KFENCE was enabled/tested, there are still similar issues,
when tcp_trim_head() hits KFENCE while the master skb
is cloned.

This happens under heavy networking TX workloads,
when the TX completion might be delayed after incoming ACK.

This patch fixes the WARNING in sk_stream_kill_queues
when sk->sk_mem_queued/sk->sk_forward_alloc are not zero.

Fixes: d3fb45f370d9 ("mm, kfence: insert KFENCE hooks for SLAB")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Marco Elver <elver@google.com>
Link: https://lore.kernel.org/r/20211102004555.1359210-1-eric.dumazet@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
include/linux/skbuff.h
net/core/skbuff.c
net/ipv4/tcp_output.c