tcp: seq_file: Avoid skipping sk during tcp_seek_last_pos
authorMartin KaFai Lau <kafai@fb.com>
Thu, 1 Jul 2021 20:05:41 +0000 (13:05 -0700)
committerAndrii Nakryiko <andrii@kernel.org>
Fri, 23 Jul 2021 23:43:24 +0000 (16:43 -0700)
commit525e2f9fd0229eb10cb460a9e6d978257f24804e
treeba158dcde9d92518e42a730037b5f28eb8a6aa56
parentda97553ec6e1ba229f5b90c0e25799ea8afede51
tcp: seq_file: Avoid skipping sk during tcp_seek_last_pos

st->bucket stores the current bucket number.
st->offset stores the offset within this bucket that is the sk to be
seq_show().  Thus, st->offset only makes sense within the same
st->bucket.

These two variables are an optimization for the common no-lseek case.
When resuming the seq_file iteration (i.e. seq_start()),
tcp_seek_last_pos() tries to continue from the st->offset
at bucket st->bucket.

However, it is possible that the bucket pointed by st->bucket
has changed and st->offset may end up skipping the whole st->bucket
without finding a sk.  In this case, tcp_seek_last_pos() currently
continues to satisfy the offset condition in the next (and incorrect)
bucket.  Instead, regardless of the offset value, the first sk of the
next bucket should be returned.  Thus, "bucket == st->bucket" check is
added to tcp_seek_last_pos().

The chance of hitting this is small and the issue is a decade old,
so targeting for the next tree.

Fixes: a8b690f98baf ("tcp: Fix slowness in read /proc/net/tcp")
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Acked-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20210701200541.1033917-1-kafai@fb.com
net/ipv4/tcp_ipv4.c