From: Eric Biggers <ebiggers@kernel.org>
To: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: linux-crypto@vger.kernel.org,
Herbert Xu <herbert@gondor.apana.org.au>,
David Miller <davem@davemloft.net>,
"Jason A . Donenfeld" <Jason@zx2c4.com>,
Samuel Neves <sneves@dei.uc.pt>, Arnd Bergmann <arnd@arndb.de>,
Andy Lutomirski <luto@kernel.org>,
Martin Willi <martin@strongswan.org>,
Rene van Dorst <opensource@vdorst.com>
Subject: Re: [PATCH v3 02/29] crypto: x86/chacha - depend on generic chacha library instead of crypto driver
Date: Thu, 10 Oct 2019 23:00:28 -0700 [thread overview]
Message-ID: <20191011060028.GA23882@sol.localdomain> (raw)
In-Reply-To: <20191007164610.6881-3-ard.biesheuvel@linaro.org>
On Mon, Oct 07, 2019 at 06:45:43PM +0200, Ard Biesheuvel wrote:
> In preparation of extending the x86 ChaCha driver to also expose the ChaCha
> library interface, drop the dependency on the chacha_generic crypto driver
> as a non-SIMD fallback, and depend on the generic ChaCha library directly.
> This way, we only pull in the code we actually need, without registering
> a set of ChaCha skciphers that we will never use.
>
> Since turning the FPU on and off is cheap these days, simplify the SIMD
> routine by dropping the per-page yield, which makes for a cleaner switch
> to the library API as well.
>
> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> ---
> arch/x86/crypto/chacha_glue.c | 77 ++++++++++----------
> crypto/Kconfig | 2 +-
> 2 files changed, 40 insertions(+), 39 deletions(-)
>
> diff --git a/arch/x86/crypto/chacha_glue.c b/arch/x86/crypto/chacha_glue.c
> index bc62daa8dafd..3a1a11a4326d 100644
> --- a/arch/x86/crypto/chacha_glue.c
> +++ b/arch/x86/crypto/chacha_glue.c
> @@ -127,32 +127,32 @@ static int chacha_simd_stream_xor(struct skcipher_walk *walk,
> const struct chacha_ctx *ctx, const u8 *iv)
> {
> u32 *state, state_buf[16 + 2] __aligned(8);
> - int next_yield = 4096; /* bytes until next FPU yield */
> + bool do_simd;
> int err = 0;
>
> BUILD_BUG_ON(CHACHA_STATE_ALIGN != 16);
> state = PTR_ALIGN(state_buf + 0, CHACHA_STATE_ALIGN);
>
> - crypto_chacha_init(state, ctx, iv);
> + chacha_init_generic(state, ctx->key, iv);
>
> + do_simd = (walk->total > CHACHA_BLOCK_SIZE) && crypto_simd_usable();
> while (walk->nbytes > 0) {
> unsigned int nbytes = walk->nbytes;
>
> - if (nbytes < walk->total) {
> + if (nbytes < walk->total)
> nbytes = round_down(nbytes, walk->stride);
> - next_yield -= nbytes;
> - }
> -
> - chacha_dosimd(state, walk->dst.virt.addr, walk->src.virt.addr,
> - nbytes, ctx->nrounds);
>
> - if (next_yield <= 0) {
> - /* temporarily allow preemption */
> - kernel_fpu_end();
> + if (!do_simd) {
> + chacha_crypt_generic(state, walk->dst.virt.addr,
> + walk->src.virt.addr, nbytes,
> + ctx->nrounds);
> + } else {
> kernel_fpu_begin();
> - next_yield = 4096;
> + chacha_dosimd(state, walk->dst.virt.addr,
> + walk->src.virt.addr, nbytes,
> + ctx->nrounds);
> + kernel_fpu_end();
Since the kernel_fpu_begin() and kernel_fpu_end() were moved here, it's now
possible to simplify the code by moving the call to skcipher_walk_virt() into
chacha_simd_stream_xor() rather than making the caller do it.
I.e., see what the code was like prior to the following commit:
commit f9c9bdb5131eee60dc3b92e5126d4c0e291703e2
Author: Eric Biggers <ebiggers@google.com>
Date: Sat Dec 15 12:40:17 2018 -0800
crypto: x86/chacha - avoid sleeping under kernel_fpu_begin()
> }
> -
> err = skcipher_walk_done(walk, walk->nbytes - nbytes);
> }
>
> @@ -164,19 +164,9 @@ static int chacha_simd(struct skcipher_request *req)
> struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
> struct chacha_ctx *ctx = crypto_skcipher_ctx(tfm);
> struct skcipher_walk walk;
> - int err;
>
> - if (req->cryptlen <= CHACHA_BLOCK_SIZE || !crypto_simd_usable())
> - return crypto_chacha_crypt(req);
> -
> - err = skcipher_walk_virt(&walk, req, true);
> - if (err)
> - return err;
> -
> - kernel_fpu_begin();
> - err = chacha_simd_stream_xor(&walk, ctx, req->iv);
> - kernel_fpu_end();
> - return err;
> + return skcipher_walk_virt(&walk, req, true) ?:
> + chacha_simd_stream_xor(&walk, ctx, req->iv);
> }
>
> static int xchacha_simd(struct skcipher_request *req)
> @@ -189,31 +179,42 @@ static int xchacha_simd(struct skcipher_request *req)
> u8 real_iv[16];
> int err;
>
> - if (req->cryptlen <= CHACHA_BLOCK_SIZE || !crypto_simd_usable())
> - return crypto_xchacha_crypt(req);
> -
> err = skcipher_walk_virt(&walk, req, true);
> if (err)
> return err;
>
> BUILD_BUG_ON(CHACHA_STATE_ALIGN != 16);
> state = PTR_ALIGN(state_buf + 0, CHACHA_STATE_ALIGN);
> - crypto_chacha_init(state, ctx, req->iv);
> -
> - kernel_fpu_begin();
> -
> - hchacha_block_ssse3(state, subctx.key, ctx->nrounds);
> + chacha_init_generic(state, ctx->key, req->iv);
> +
> + if (req->cryptlen > CHACHA_BLOCK_SIZE && crypto_simd_usable()) {
> + kernel_fpu_begin();
> + hchacha_block_ssse3(state, subctx.key, ctx->nrounds);
> + kernel_fpu_end();
> + } else {
> + hchacha_block_generic(state, subctx.key, ctx->nrounds);
> + }
> subctx.nrounds = ctx->nrounds;
>
> memcpy(&real_iv[0], req->iv + 24, 8);
> memcpy(&real_iv[8], req->iv + 16, 8);
> err = chacha_simd_stream_xor(&walk, &subctx, real_iv);
>
> - kernel_fpu_end();
> -
> return err;
Can use 'return chacha_simd_stream_xor(...') here.
- Eric
next prev parent reply other threads:[~2019-10-11 6:00 UTC|newest]
Thread overview: 51+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-07 16:45 [PATCH v3 00/29] crypto: crypto API library interfaces for WireGuard Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 01/29] crypto: chacha - move existing library code into lib/crypto Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 02/29] crypto: x86/chacha - depend on generic chacha library instead of crypto driver Ard Biesheuvel
2019-10-11 6:00 ` Eric Biggers [this message]
2019-10-15 10:00 ` Martin Willi
2019-10-15 10:12 ` Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 03/29] crypto: x86/chacha - expose SIMD ChaCha routine as library function Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 04/29] crypto: arm64/chacha - depend on generic chacha library instead of crypto driver Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 05/29] crypto: arm64/chacha - expose arm64 ChaCha routine as library function Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 06/29] crypto: arm/chacha - import Eric Biggers's scalar accelerated ChaCha code Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 07/29] crypto: arm/chacha - remove dependency on generic ChaCha driver Ard Biesheuvel
2019-10-11 6:12 ` Eric Biggers
2019-10-11 6:31 ` Eric Biggers
2019-10-07 16:45 ` [PATCH v3 08/29] crypto: arm/chacha - expose ARM ChaCha routine as library function Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 09/29] crypto: mips/chacha - import 32r2 ChaCha code from Zinc Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 10/29] crypto: mips/chacha - wire up accelerated 32r2 " Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 11/29] crypto: chacha - unexport chacha_generic routines Ard Biesheuvel
2019-10-11 6:04 ` Eric Biggers
2019-10-07 16:45 ` [PATCH v3 12/29] crypto: poly1305 - move core routines into a separate library Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 13/29] crypto: x86/poly1305 - unify Poly1305 state struct with generic code Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 14/29] crypto: poly1305 - expose init/update/final library interface Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 15/29] crypto: x86/poly1305 - depend on generic library not generic shash Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 16/29] crypto: x86/poly1305 - expose existing driver as poly1305 library Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 17/29] crypto: arm64/poly1305 - incorporate OpenSSL/CRYPTOGAMS NEON implementation Ard Biesheuvel
2019-10-07 16:45 ` [PATCH v3 18/29] crypto: arm/poly1305 " Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 19/29] crypto: mips/poly1305 - incorporate OpenSSL/CRYPTOGAMS optimized implementation Ard Biesheuvel
2019-10-07 21:02 ` René van Dorst
2019-10-08 5:55 ` Ard Biesheuvel
2019-10-08 11:38 ` Andy Polyakov
2019-10-08 17:46 ` René van Dorst
2019-10-11 14:14 ` Andy Polyakov
2019-10-11 17:21 ` René van Dorst
2019-10-11 18:49 ` Andy Polyakov
2019-10-11 21:38 ` Arnd Bergmann
2019-10-07 16:46 ` [PATCH v3 20/29] int128: move __uint128_t compiler test to Kconfig Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 21/29] crypto: BLAKE2s - generic C library implementation and selftest Ard Biesheuvel
2019-10-11 6:02 ` Eric Biggers
2019-10-11 16:45 ` Jason A. Donenfeld
2019-10-14 12:53 ` Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 22/29] crypto: BLAKE2s - x86_64 library implementation Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 23/29] crypto: Curve25519 - generic C library implementations and selftest Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 24/29] crypto: lib/curve25519 - work around Clang stack spilling issue Ard Biesheuvel
2019-10-14 14:13 ` Jason A. Donenfeld
2019-10-14 16:07 ` Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 25/29] crypto: Curve25519 - x86_64 library implementation Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 26/29] crypto: arm - import Bernstein and Schwabe's Curve25519 ARM implementation Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 27/29] crypto: arm/Curve25519 - wire up NEON implementation Ard Biesheuvel
2019-10-07 16:46 ` [PATCH v3 28/29] crypto: chacha20poly1305 - import construction and selftest from Zinc Ard Biesheuvel
2019-10-11 6:14 ` Eric Biggers
2019-10-07 16:46 ` [PATCH v3 29/29] crypto: lib/chacha20poly1305 - reimplement crypt_from_sg() routine Ard Biesheuvel
2019-10-14 14:33 ` [PATCH v3 00/29] crypto: crypto API library interfaces for WireGuard Jason A. Donenfeld
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20191011060028.GA23882@sol.localdomain \
--to=ebiggers@kernel.org \
--cc=Jason@zx2c4.com \
--cc=ard.biesheuvel@linaro.org \
--cc=arnd@arndb.de \
--cc=davem@davemloft.net \
--cc=herbert@gondor.apana.org.au \
--cc=linux-crypto@vger.kernel.org \
--cc=luto@kernel.org \
--cc=martin@strongswan.org \
--cc=opensource@vdorst.com \
--cc=sneves@dei.uc.pt \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).