[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20220726201600.1715505-1-dima@arista.com>
Date: Tue, 26 Jul 2022 21:15:54 +0100
From: Dmitry Safonov <dima@...sta.com>
To: linux-kernel@...r.kernel.org
Cc: Dmitry Safonov <0x7f454c46@...il.com>,
Dmitry Safonov <dima@...sta.com>,
Andy Lutomirski <luto@...capital.net>,
Ard Biesheuvel <ardb@...nel.org>,
David Ahern <dsahern@...nel.org>,
"David S. Miller" <davem@...emloft.net>,
Eric Biggers <ebiggers@...nel.org>,
Eric Dumazet <edumazet@...gle.com>,
Francesco Ruggeri <fruggeri@...sta.com>,
Herbert Xu <herbert@...dor.apana.org.au>,
Hideaki YOSHIFUJI <yoshfuji@...ux-ipv6.org>,
Jakub Kicinski <kuba@...nel.org>,
Leonard Crestez <cdleonard@...il.com>,
Paolo Abeni <pabeni@...hat.com>,
Salam Noureddine <noureddine@...sta.com>,
netdev@...r.kernel.org, linux-crypto@...r.kernel.org
Subject: [PATCH 0/6] net/crypto: Introduce crypto_pool
Add crypto_pool - an API for allocating per-CPU array of crypto requests
on slow-path (in sleep'able context) and to use them on a fast-path,
which is RX/TX for net/ users (or in any other bh-disabled users).
The design is based on the current implementations of md5sig_pool.
Previously, I've suggested to add such API on TCP-AO patch submission [1],
where Herbert kindly suggested to help with introducing new crypto API.
New API will allow:
- to reuse per-CPU ahash_request(s) for different users
- to allocate only one per-CPU scratch buffer rather than a new one for
each user
- to have a common API for net/ users that need ahash on RX/TX fast path
In this version I've wired up TCP-MD5 and IPv6-SR-HMAC as users.
Potentially, xfrm_ipcomp and xfrm_ah can be converted as well.
The initial reason for patches would be to have TCP-AO as a user, which
would let it share per-CPU crypto_request for any supported hashing
algorithm.
While at it, I've also made TCP-MD5 static key dynamically switchable.
This means that after TCP-MD5 was used and the last key got destroyed,
the static branch is disabled and any potential penalty for checking
tcp_md5sig_info is gone, and the system's tcp performance should be as
if it never had TCP-MD5 key defined.
[1]: http://lkml.kernel.org/r/20211106034334.GA18577@gondor.apana.org.au
Cc: Andy Lutomirski <luto@...capital.net>
Cc: Ard Biesheuvel <ardb@...nel.org>
Cc: David Ahern <dsahern@...nel.org>
Cc: "David S. Miller" <davem@...emloft.net>
Cc: Dmitry Safonov <0x7f454c46@...il.com>
Cc: Eric Biggers <ebiggers@...nel.org>
Cc: Eric Dumazet <edumazet@...gle.com>
Cc: Francesco Ruggeri <fruggeri@...sta.com>
Cc: Herbert Xu <herbert@...dor.apana.org.au>
Cc: Hideaki YOSHIFUJI <yoshfuji@...ux-ipv6.org>
Cc: Jakub Kicinski <kuba@...nel.org>
Cc: Leonard Crestez <cdleonard@...il.com>
Cc: Paolo Abeni <pabeni@...hat.com>
Cc: Salam Noureddine <noureddine@...sta.com>
Cc: netdev@...r.kernel.org
Cc: linux-crypto@...r.kernel.org
Cc: linux-kernel@...r.kernel.org
Dmitry Safonov (6):
crypto: Introduce crypto_pool
crypto_pool: Add crypto_pool_reserve_scratch()
net/tcp: Separate tcp_md5sig_info allocation into
tcp_md5sig_info_add()
net/tcp: Disable TCP-MD5 static key on tcp_md5sig_info destruction
net/tcp: Use crypto_pool for TCP-MD5
net/ipv6: sr: Switch to using crypto_pool
crypto/Kconfig | 12 ++
crypto/Makefile | 1 +
crypto/crypto_pool.c | 316 +++++++++++++++++++++++++++++++++++++++
include/crypto/pool.h | 33 ++++
include/net/seg6_hmac.h | 7 -
include/net/tcp.h | 32 ++--
net/ipv4/Kconfig | 2 +-
net/ipv4/tcp.c | 102 ++++---------
net/ipv4/tcp_ipv4.c | 153 ++++++++++++-------
net/ipv4/tcp_minisocks.c | 23 ++-
net/ipv4/tcp_output.c | 4 +-
net/ipv6/Kconfig | 2 +-
net/ipv6/seg6.c | 3 -
net/ipv6/seg6_hmac.c | 204 ++++++++++---------------
net/ipv6/tcp_ipv6.c | 63 ++++----
15 files changed, 634 insertions(+), 323 deletions(-)
create mode 100644 crypto/crypto_pool.c
create mode 100644 include/crypto/pool.h
base-commit: 058affafc65a74cf54499fb578b66ad0b18f939b
--
2.36.1
Powered by blists - more mailing lists