iov_iter: optimize page_copy_sane()
authorEric Dumazet <edumazet@google.com>
Tue, 26 Feb 2019 18:42:39 +0000 (10:42 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 22 May 2019 05:37:45 +0000 (07:37 +0200)
commit627bb2d93b4d48f89e2087b3149c0a8444e5a1d4
tree9b6acc584f8b33f2915d7b9c61edf279985b1905
parent866f011181ffb9f4da5044dda316bbac26c78819
iov_iter: optimize page_copy_sane()

commit 6daef95b8c914866a46247232a048447fff97279 upstream.

Avoid cache line miss dereferencing struct page if we can.

page_copy_sane() mostly deals with order-0 pages.

Extra cache line miss is visible on TCP recvmsg() calls dealing
with GRO packets (typically 45 page frags are attached to one skb).

Bringing the 45 struct pages into cpu cache while copying the data
is not free, since the freeing of the skb (and associated
page frags put_page()) can happen after cache lines have been evicted.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
lib/iov_iter.c