EljakimHerrewijnen/Openwrt
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
a1735fe73c
Openwrt/target/linux/generic/backport-5.4/080-wireguard-0057-crypto-arch-lib-limit-simd-usage-to-4k-chunks.patch
Jason A. Donenfeld 3888fa7880 kernel: 5.4: import wireguard backport 
Rather than using the clunky, old, slower wireguard-linux-compat out of
tree module, this commit does a patch-by-patch backport of upstream's
wireguard to 5.4. This specific backport is in widespread use, being
part of SUSE's enterprise kernel, Oracle's enterprise kernel, Google's
Android kernel, Gentoo's distro kernel, and probably more I've forgotten
about. It's definately the "more proper" way of adding wireguard to a
kernel than the ugly compat.h hell of the wireguard-linux-compat repo.
And most importantly for OpenWRT, it allows using the same module
configuration code for 5.10 as for 5.4, with no need for bifurcation.

These patches are from the backport tree which is maintained in the
open here: https://git.zx2c4.com/wireguard-linux/log/?h=backport-5.4.y
I'll be sending PRs to update this as needed.

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
2021-02-26 20:41:01 +01:00

244 lines
7.7 KiB
Diff
Raw Blame History

From af386d2b1f9207290a12aa97ecec8b428f3bebb2 Mon Sep 17 00:00:00 2001
From: "Jason A. Donenfeld" <Jason@zx2c4.com>
Date: Thu, 23 Apr 2020 15:54:04 -0600
Subject: [PATCH 057/124] crypto: arch/lib - limit simd usage to 4k chunks
commit 706024a52c614b478b63f7728d202532ce6591a9 upstream.
The initial Zinc patchset, after some mailing list discussion, contained
code to ensure that kernel_fpu_enable would not be kept on for more than
a 4k chunk, since it disables preemption. The choice of 4k isn't totally
scientific, but it's not a bad guess either, and it's what's used in
both the x86 poly1305, blake2s, and nhpoly1305 code already (in the form
of PAGE_SIZE, which this commit corrects to be explicitly 4k for the
former two).
Ard did some back of the envelope calculations and found that
at 5 cycles/byte (overestimate) on a 1ghz processor (pretty slow), 4k
means we have a maximum preemption disabling of 20us, which Sebastian
confirmed was probably a good limit.
Unfortunately the chunking appears to have been left out of the final
patchset that added the glue code. So, this commit adds it back in.
Fixes: 84e03fa39fbe ("crypto: x86/chacha - expose SIMD ChaCha routine as library function")
Fixes: b3aad5bad26a ("crypto: arm64/chacha - expose arm64 ChaCha routine as library function")
Fixes: a44a3430d71b ("crypto: arm/chacha - expose ARM ChaCha routine as library function")
Fixes: d7d7b8535662 ("crypto: x86/poly1305 - wire up faster implementations for kernel")
Fixes: f569ca164751 ("crypto: arm64/poly1305 - incorporate OpenSSL/CRYPTOGAMS NEON implementation")
Fixes: a6b803b3ddc7 ("crypto: arm/poly1305 - incorporate OpenSSL/CRYPTOGAMS NEON implementation")
Fixes: ed0356eda153 ("crypto: blake2s - x86_64 SIMD implementation")
Cc: Eric Biggers <ebiggers@google.com>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: stable@vger.kernel.org
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
---
arch/arm/crypto/chacha-glue.c | 14 +++++++++++---
arch/arm/crypto/poly1305-glue.c | 15 +++++++++++----
arch/arm64/crypto/chacha-neon-glue.c | 14 +++++++++++---
arch/arm64/crypto/poly1305-glue.c | 15 +++++++++++----
arch/x86/crypto/blake2s-glue.c | 10 ++++------
arch/x86/crypto/chacha_glue.c | 14 +++++++++++---
arch/x86/crypto/poly1305_glue.c | 13 ++++++-------
7 files changed, 65 insertions(+), 30 deletions(-)
--- a/arch/arm/crypto/chacha-glue.c
+++ b/arch/arm/crypto/chacha-glue.c
@@ -91,9 +91,17 @@ void chacha_crypt_arch(u32 *state, u8 *d
return;
}
- kernel_neon_begin();
- chacha_doneon(state, dst, src, bytes, nrounds);
- kernel_neon_end();
+ do {
+ unsigned int todo = min_t(unsigned int, bytes, SZ_4K);
+
+ kernel_neon_begin();
+ chacha_doneon(state, dst, src, todo, nrounds);
+ kernel_neon_end();
+
+ bytes -= todo;
+ src += todo;
+ dst += todo;
+ } while (bytes);
}
EXPORT_SYMBOL(chacha_crypt_arch);
--- a/arch/arm/crypto/poly1305-glue.c
+++ b/arch/arm/crypto/poly1305-glue.c
@@ -160,13 +160,20 @@ void poly1305_update_arch(struct poly130
unsigned int len = round_down(nbytes, POLY1305_BLOCK_SIZE);
if (static_branch_likely(&have_neon) && do_neon) {
- kernel_neon_begin();
- poly1305_blocks_neon(&dctx->h, src, len, 1);
- kernel_neon_end();
+ do {
+ unsigned int todo = min_t(unsigned int, len, SZ_4K);
+
+ kernel_neon_begin();
+ poly1305_blocks_neon(&dctx->h, src, todo, 1);
+ kernel_neon_end();
+
+ len -= todo;
+ src += todo;
+ } while (len);
} else {
poly1305_blocks_arm(&dctx->h, src, len, 1);
+ src += len;
}
- src += len;
nbytes %= POLY1305_BLOCK_SIZE;
}
--- a/arch/arm64/crypto/chacha-neon-glue.c
+++ b/arch/arm64/crypto/chacha-neon-glue.c
@@ -87,9 +87,17 @@ void chacha_crypt_arch(u32 *state, u8 *d
!crypto_simd_usable())
return chacha_crypt_generic(state, dst, src, bytes, nrounds);
- kernel_neon_begin();
- chacha_doneon(state, dst, src, bytes, nrounds);
- kernel_neon_end();
+ do {
+ unsigned int todo = min_t(unsigned int, bytes, SZ_4K);
+
+ kernel_neon_begin();
+ chacha_doneon(state, dst, src, todo, nrounds);
+ kernel_neon_end();
+
+ bytes -= todo;
+ src += todo;
+ dst += todo;
+ } while (bytes);
}
EXPORT_SYMBOL(chacha_crypt_arch);
--- a/arch/arm64/crypto/poly1305-glue.c
+++ b/arch/arm64/crypto/poly1305-glue.c
@@ -143,13 +143,20 @@ void poly1305_update_arch(struct poly130
unsigned int len = round_down(nbytes, POLY1305_BLOCK_SIZE);
if (static_branch_likely(&have_neon) && crypto_simd_usable()) {
- kernel_neon_begin();
- poly1305_blocks_neon(&dctx->h, src, len, 1);
- kernel_neon_end();
+ do {
+ unsigned int todo = min_t(unsigned int, len, SZ_4K);
+
+ kernel_neon_begin();
+ poly1305_blocks_neon(&dctx->h, src, todo, 1);
+ kernel_neon_end();
+
+ len -= todo;
+ src += todo;
+ } while (len);
} else {
poly1305_blocks(&dctx->h, src, len, 1);
+ src += len;
}
- src += len;
nbytes %= POLY1305_BLOCK_SIZE;
}
--- a/arch/x86/crypto/blake2s-glue.c
+++ b/arch/x86/crypto/blake2s-glue.c
@@ -32,16 +32,16 @@ void blake2s_compress_arch(struct blake2
const u32 inc)
{
/* SIMD disables preemption, so relax after processing each page. */
- BUILD_BUG_ON(PAGE_SIZE / BLAKE2S_BLOCK_SIZE < 8);
+ BUILD_BUG_ON(SZ_4K / BLAKE2S_BLOCK_SIZE < 8);
if (!static_branch_likely(&blake2s_use_ssse3) || !crypto_simd_usable()) {
blake2s_compress_generic(state, block, nblocks, inc);
return;
}
- for (;;) {
+ do {
const size_t blocks = min_t(size_t, nblocks,
- PAGE_SIZE / BLAKE2S_BLOCK_SIZE);
+ SZ_4K / BLAKE2S_BLOCK_SIZE);
kernel_fpu_begin();
if (IS_ENABLED(CONFIG_AS_AVX512) &&
@@ -52,10 +52,8 @@ void blake2s_compress_arch(struct blake2
kernel_fpu_end();
nblocks -= blocks;
- if (!nblocks)
- break;
block += blocks * BLAKE2S_BLOCK_SIZE;
- }
+ } while (nblocks);
}
EXPORT_SYMBOL(blake2s_compress_arch);
--- a/arch/x86/crypto/chacha_glue.c
+++ b/arch/x86/crypto/chacha_glue.c
@@ -154,9 +154,17 @@ void chacha_crypt_arch(u32 *state, u8 *d
bytes <= CHACHA_BLOCK_SIZE)
return chacha_crypt_generic(state, dst, src, bytes, nrounds);
- kernel_fpu_begin();
- chacha_dosimd(state, dst, src, bytes, nrounds);
- kernel_fpu_end();
+ do {
+ unsigned int todo = min_t(unsigned int, bytes, SZ_4K);
+
+ kernel_fpu_begin();
+ chacha_dosimd(state, dst, src, todo, nrounds);
+ kernel_fpu_end();
+
+ bytes -= todo;
+ src += todo;
+ dst += todo;
+ } while (bytes);
}
EXPORT_SYMBOL(chacha_crypt_arch);
--- a/arch/x86/crypto/poly1305_glue.c
+++ b/arch/x86/crypto/poly1305_glue.c
@@ -91,8 +91,8 @@ static void poly1305_simd_blocks(void *c
struct poly1305_arch_internal *state = ctx;
/* SIMD disables preemption, so relax after processing each page. */
- BUILD_BUG_ON(PAGE_SIZE < POLY1305_BLOCK_SIZE ||
- PAGE_SIZE % POLY1305_BLOCK_SIZE);
+ BUILD_BUG_ON(SZ_4K < POLY1305_BLOCK_SIZE ||
+ SZ_4K % POLY1305_BLOCK_SIZE);
if (!IS_ENABLED(CONFIG_AS_AVX) || !static_branch_likely(&poly1305_use_avx) ||
(len < (POLY1305_BLOCK_SIZE * 18) && !state->is_base2_26) ||
@@ -102,8 +102,8 @@ static void poly1305_simd_blocks(void *c
return;
}
- for (;;) {
- const size_t bytes = min_t(size_t, len, PAGE_SIZE);
+ do {
+ const size_t bytes = min_t(size_t, len, SZ_4K);
kernel_fpu_begin();
if (IS_ENABLED(CONFIG_AS_AVX512) && static_branch_likely(&poly1305_use_avx512))
@@ -113,11 +113,10 @@ static void poly1305_simd_blocks(void *c
else
poly1305_blocks_avx(ctx, inp, bytes, padbit);
kernel_fpu_end();
+
len -= bytes;
- if (!len)
- break;
inp += bytes;
- }
+ } while (len);
}
static void poly1305_simd_emit(void *ctx, u8 mac[POLY1305_DIGEST_SIZE],
Reference in New Issue View Git Blame Copy Permalink