Hey folks,
I think we're approaching the end of the review for this patchset and we're
getting somewhat close to being ready for it being queued up. At this point,
I've incorporated all of the extremely helpful and instructive suggestions
from the list.
For this v5, we now accept u64[2] as the key, so that alignment is taken
care of naturally. For other alignment issues, we have both the fast aligned
version and the unaligned version, depending on what's necessary. We've
worked out the issues for struct padding. The functions now take a void
pointer to avoid ugly casting, which also helps us shed the inline helper
functions which were not very pretty. The replacements of MD5 have been
benchmarked and show a big increase in speed. We've even come up with a
better naming scheme for dword/qword. All and all it's shaping up nicely.
So, if this series looks good to you, please send along your Reviewed-by,
so we can begin to get this completed. If there are still lingering issues,
let me know and I'll incorporated them into a v6 if necessary.
Thanks,
Jason
Jason A. Donenfeld (4):
siphash: add cryptographically secure PRF
siphash: add Nu{32,64} helpers
secure_seq: use SipHash in place of MD5
random: use SipHash in place of MD5
drivers/char/random.c | 32 +++----
include/linux/siphash.h | 65 ++++++++++++++
lib/Kconfig.debug | 6 +-
lib/Makefile | 5 +-
lib/siphash.c | 223 ++++++++++++++++++++++++++++++++++++++++++++++++
lib/test_siphash.c | 101 ++++++++++++++++++++++
net/core/secure_seq.c | 133 +++++++++++------------------
7 files changed, 460 insertions(+), 105 deletions(-)
create mode 100644 include/linux/siphash.h
create mode 100644 lib/siphash.c
create mode 100644 lib/test_siphash.c
--
2.11.0
SipHash is a 64-bit keyed hash function that is actually a
cryptographically secure PRF, like HMAC. Except SipHash is super fast,
and is meant to be used as a hashtable keyed lookup function, or as a
general PRF for short input use cases, such as sequence numbers or RNG
chaining.
For the first usage:
There are a variety of attacks known as "hashtable poisoning" in which an
attacker forms some data such that the hash of that data will be the
same, and then preceeds to fill up all entries of a hashbucket. This is
a realistic and well-known denial-of-service vector. Currently
hashtables use jhash, which is fast but not secure, and some kind of
rotating key scheme (or none at all, which isn't good). SipHash is meant
as a replacement for jhash in these cases.
There are a modicum of places in the kernel that are vulnerable to
hashtable poisoning attacks, either via userspace vectors or network
vectors, and there's not a reliable mechanism inside the kernel at the
moment to fix it. The first step toward fixing these issues is actually
getting a secure primitive into the kernel for developers to use. Then
we can, bit by bit, port things over to it as deemed appropriate.
While SipHash is extremely fast for a cryptographically secure function,
it is likely a tiny bit slower than the insecure jhash, and so replacements
will be evaluated on a case-by-case basis based on whether or not the
difference in speed is negligible and whether or not the current jhash usage
poses a real security risk.
For the second usage:
A few places in the kernel are using MD5 for creating secure sequence
numbers, port numbers, or fast random numbers. SipHash is a faster, more
fitting, and more secure replacement for MD5 in those situations.
Replacing MD5 with SipHash for these uses is obvious and straight-
forward, and so is submitted along with this patch series. There
shouldn't be much of a debate over its efficacy.
Dozens of languages are already using this internally for their hash
tables and PRFs. Some of the BSDs already use this in their kernels.
SipHash is a widely known high-speed solution to a widely known set of
problems, and it's time we catch-up.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Jean-Philippe Aumasson <[email protected]>
Cc: Daniel J. Bernstein <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Eric Biggers <[email protected]>
Cc: David Laight <[email protected]>
---
include/linux/siphash.h | 32 +++++++++++
lib/Kconfig.debug | 6 +--
lib/Makefile | 5 +-
lib/siphash.c | 138 ++++++++++++++++++++++++++++++++++++++++++++++++
lib/test_siphash.c | 83 +++++++++++++++++++++++++++++
5 files changed, 259 insertions(+), 5 deletions(-)
create mode 100644 include/linux/siphash.h
create mode 100644 lib/siphash.c
create mode 100644 lib/test_siphash.c
diff --git a/include/linux/siphash.h b/include/linux/siphash.h
new file mode 100644
index 000000000000..145cf5667078
--- /dev/null
+++ b/include/linux/siphash.h
@@ -0,0 +1,32 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#ifndef _LINUX_SIPHASH_H
+#define _LINUX_SIPHASH_H
+
+#include <linux/types.h>
+
+#define SIPHASH_ALIGNMENT 8
+
+typedef u64 siphash_key_t[2];
+
+u64 siphash(const void *data, size_t len, const siphash_key_t key);
+
+#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+static inline u64 siphash_unaligned(const void *data, size_t len,
+ const siphash_key_t key)
+{
+ return siphash(data, len, key);
+}
+#else
+u64 siphash_unaligned(const void *data, size_t len, const siphash_key_t key);
+#endif
+
+#endif /* _LINUX_SIPHASH_H */
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 7446097f72bd..86254ea99b45 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1843,9 +1843,9 @@ config TEST_HASH
tristate "Perform selftest on hash functions"
default n
help
- Enable this option to test the kernel's integer (<linux/hash,h>)
- and string (<linux/stringhash.h>) hash functions on boot
- (or module load).
+ Enable this option to test the kernel's integer (<linux/hash.h>),
+ string (<linux/stringhash.h>), and siphash (<linux/siphash.h>)
+ hash functions on boot (or module load).
This is intended to help people writing architecture-specific
optimized versions. If unsure, say N.
diff --git a/lib/Makefile b/lib/Makefile
index 50144a3aeebd..71d398b04a74 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -22,7 +22,8 @@ lib-y := ctype.o string.o vsprintf.o cmdline.o \
sha1.o chacha20.o md5.o irq_regs.o argv_split.o \
flex_proportions.o ratelimit.o show_mem.o \
is_single_threaded.o plist.o decompress.o kobject_uevent.o \
- earlycpio.o seq_buf.o nmi_backtrace.o nodemask.o win_minmax.o
+ earlycpio.o seq_buf.o siphash.o \
+ nmi_backtrace.o nodemask.o win_minmax.o
lib-$(CONFIG_MMU) += ioremap.o
lib-$(CONFIG_SMP) += cpumask.o
@@ -44,7 +45,7 @@ obj-$(CONFIG_TEST_HEXDUMP) += test_hexdump.o
obj-y += kstrtox.o
obj-$(CONFIG_TEST_BPF) += test_bpf.o
obj-$(CONFIG_TEST_FIRMWARE) += test_firmware.o
-obj-$(CONFIG_TEST_HASH) += test_hash.o
+obj-$(CONFIG_TEST_HASH) += test_hash.o test_siphash.o
obj-$(CONFIG_TEST_KASAN) += test_kasan.o
obj-$(CONFIG_TEST_KSTRTOX) += test-kstrtox.o
obj-$(CONFIG_TEST_LKM) += test_module.o
diff --git a/lib/siphash.c b/lib/siphash.c
new file mode 100644
index 000000000000..afc13cbb1b78
--- /dev/null
+++ b/lib/siphash.c
@@ -0,0 +1,138 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#include <linux/siphash.h>
+#include <linux/kernel.h>
+#include <asm/unaligned.h>
+
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+#include <linux/dcache.h>
+#include <asm/word-at-a-time.h>
+#endif
+
+#define SIPROUND \
+ do { \
+ v0 += v1; v1 = rol64(v1, 13); v1 ^= v0; v0 = rol64(v0, 32); \
+ v2 += v3; v3 = rol64(v3, 16); v3 ^= v2; \
+ v0 += v3; v3 = rol64(v3, 21); v3 ^= v0; \
+ v2 += v1; v1 = rol64(v1, 17); v1 ^= v2; v2 = rol64(v2, 32); \
+ } while(0)
+
+/**
+ * siphash - compute 64-bit siphash PRF value
+ * @data: buffer to hash, must be aligned to SIPHASH_ALIGNMENT
+ * @size: size of @data
+ * @key: the siphash key
+ */
+u64 siphash(const void *data, size_t len, const siphash_key_t key)
+{
+ u64 v0 = 0x736f6d6570736575ULL;
+ u64 v1 = 0x646f72616e646f6dULL;
+ u64 v2 = 0x6c7967656e657261ULL;
+ u64 v3 = 0x7465646279746573ULL;
+ u64 b = ((u64)len) << 56;
+ u64 m;
+ const u8 *end = data + len - (len % sizeof(u64));
+ const u8 left = len & (sizeof(u64) - 1);
+ v3 ^= key[1];
+ v2 ^= key[0];
+ v1 ^= key[1];
+ v0 ^= key[0];
+ for (; data != end; data += sizeof(u64)) {
+ m = le64_to_cpup(data);
+ v3 ^= m;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= m;
+ }
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+ if (left)
+ b |= le64_to_cpu((__force __le64)(load_unaligned_zeropad(data) &
+ bytemask_from_count(left)));
+#else
+ switch (left) {
+ case 7: b |= ((u64)end[6]) << 48;
+ case 6: b |= ((u64)end[5]) << 40;
+ case 5: b |= ((u64)end[4]) << 32;
+ case 4: b |= le32_to_cpup(data); break;
+ case 3: b |= ((u64)end[2]) << 16;
+ case 2: b |= le16_to_cpup(data); break;
+ case 1: b |= end[0];
+ }
+#endif
+ v3 ^= b;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= b;
+ v2 ^= 0xff;
+ SIPROUND;
+ SIPROUND;
+ SIPROUND;
+ SIPROUND;
+ return (v0 ^ v1) ^ (v2 ^ v3);
+}
+EXPORT_SYMBOL(siphash);
+
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+/**
+ * siphash - compute 64-bit siphash PRF value, without alignment requirements
+ * @data: buffer to hash
+ * @size: size of @data
+ * @key: the siphash key
+ */
+u64 siphash_unaligned(const void *data, size_t len, const siphash_key_t key)
+{
+ u64 v0 = 0x736f6d6570736575ULL;
+ u64 v1 = 0x646f72616e646f6dULL;
+ u64 v2 = 0x6c7967656e657261ULL;
+ u64 v3 = 0x7465646279746573ULL;
+ u64 b = ((u64)len) << 56;
+ u64 m;
+ const u8 *end = data + len - (len % sizeof(u64));
+ const u8 left = len & (sizeof(u64) - 1);
+ v3 ^= key[1];
+ v2 ^= key[0];
+ v1 ^= key[1];
+ v0 ^= key[0];
+ for (; data != end; data += sizeof(u64)) {
+ m = get_unaligned_le64(data);
+ v3 ^= m;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= m;
+ }
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+ if (left)
+ b |= le64_to_cpu((__force __le64)(load_unaligned_zeropad(data) &
+ bytemask_from_count(left)));
+#else
+ switch (left) {
+ case 7: b |= ((u64)end[6]) << 48;
+ case 6: b |= ((u64)end[5]) << 40;
+ case 5: b |= ((u64)end[4]) << 32;
+ case 4: b |= get_unaligned_le32(end); break;
+ case 3: b |= ((u64)end[2]) << 16;
+ case 2: b |= get_unaligned_le16(end); break;
+ case 1: b |= bytes[0];
+ }
+#endif
+ v3 ^= b;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= b;
+ v2 ^= 0xff;
+ SIPROUND;
+ SIPROUND;
+ SIPROUND;
+ SIPROUND;
+ return (v0 ^ v1) ^ (v2 ^ v3);
+}
+EXPORT_SYMBOL(siphash_unaligned);
+#endif
diff --git a/lib/test_siphash.c b/lib/test_siphash.c
new file mode 100644
index 000000000000..93549e4e22c5
--- /dev/null
+++ b/lib/test_siphash.c
@@ -0,0 +1,83 @@
+/* Test cases for siphash.c
+ *
+ * Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/siphash.h>
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/module.h>
+
+/* Test vectors taken from official reference source available at:
+ * https://131002.net/siphash/siphash24.c
+ */
+static const u64 test_vectors[64] = {
+ 0x726fdb47dd0e0e31ULL, 0x74f839c593dc67fdULL, 0x0d6c8009d9a94f5aULL,
+ 0x85676696d7fb7e2dULL, 0xcf2794e0277187b7ULL, 0x18765564cd99a68dULL,
+ 0xcbc9466e58fee3ceULL, 0xab0200f58b01d137ULL, 0x93f5f5799a932462ULL,
+ 0x9e0082df0ba9e4b0ULL, 0x7a5dbbc594ddb9f3ULL, 0xf4b32f46226bada7ULL,
+ 0x751e8fbc860ee5fbULL, 0x14ea5627c0843d90ULL, 0xf723ca908e7af2eeULL,
+ 0xa129ca6149be45e5ULL, 0x3f2acc7f57c29bdbULL, 0x699ae9f52cbe4794ULL,
+ 0x4bc1b3f0968dd39cULL, 0xbb6dc91da77961bdULL, 0xbed65cf21aa2ee98ULL,
+ 0xd0f2cbb02e3b67c7ULL, 0x93536795e3a33e88ULL, 0xa80c038ccd5ccec8ULL,
+ 0xb8ad50c6f649af94ULL, 0xbce192de8a85b8eaULL, 0x17d835b85bbb15f3ULL,
+ 0x2f2e6163076bcfadULL, 0xde4daaaca71dc9a5ULL, 0xa6a2506687956571ULL,
+ 0xad87a3535c49ef28ULL, 0x32d892fad841c342ULL, 0x7127512f72f27cceULL,
+ 0xa7f32346f95978e3ULL, 0x12e0b01abb051238ULL, 0x15e034d40fa197aeULL,
+ 0x314dffbe0815a3b4ULL, 0x027990f029623981ULL, 0xcadcd4e59ef40c4dULL,
+ 0x9abfd8766a33735cULL, 0x0e3ea96b5304a7d0ULL, 0xad0c42d6fc585992ULL,
+ 0x187306c89bc215a9ULL, 0xd4a60abcf3792b95ULL, 0xf935451de4f21df2ULL,
+ 0xa9538f0419755787ULL, 0xdb9acddff56ca510ULL, 0xd06c98cd5c0975ebULL,
+ 0xe612a3cb9ecba951ULL, 0xc766e62cfcadaf96ULL, 0xee64435a9752fe72ULL,
+ 0xa192d576b245165aULL, 0x0a8787bf8ecb74b2ULL, 0x81b3e73d20b49b6fULL,
+ 0x7fa8220ba3b2eceaULL, 0x245731c13ca42499ULL, 0xb78dbfaf3a8d83bdULL,
+ 0xea1ad565322a1a0bULL, 0x60e61c23a3795013ULL, 0x6606d7e446282b93ULL,
+ 0x6ca4ecb15c5f91e1ULL, 0x9f626da15c9625f3ULL, 0xe51b38608ef25f57ULL,
+ 0x958a324ceb064572ULL
+};
+static const siphash_key_t test_key =
+ { 0x0706050403020100ULL , 0x0f0e0d0c0b0a0908ULL };
+
+static int __init siphash_test_init(void)
+{
+ u8 in[64] __aligned(SIPHASH_ALIGNMENT);
+ u8 in_unaligned[65];
+ u8 i;
+ int ret = 0;
+
+ for (i = 0; i < 64; ++i) {
+ in[i] = i;
+ in_unaligned[i + 1] = i;
+ if (siphash(in, i, test_key) != test_vectors[i]) {
+ pr_info("self-test aligned %u: FAIL\n", i + 1);
+ ret = -EINVAL;
+ }
+ if (siphash_unaligned(in_unaligned + 1, i, test_key) != test_vectors[i]) {
+ pr_info("self-test unaligned %u: FAIL\n", i + 1);
+ ret = -EINVAL;
+ }
+ }
+ if (!ret)
+ pr_info("self-tests: pass\n");
+ return ret;
+}
+
+static void __exit siphash_test_exit(void)
+{
+}
+
+module_init(siphash_test_init);
+module_exit(siphash_test_exit);
+
+MODULE_AUTHOR("Jason A. Donenfeld <[email protected]>");
+MODULE_LICENSE("Dual BSD/GPL");
--
2.11.0
These restore parity with the jhash interface by providing high
performance helpers for common input sizes.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Tom Herbert <[email protected]>
---
include/linux/siphash.h | 33 ++++++++++
lib/siphash.c | 157 +++++++++++++++++++++++++++++++++++++-----------
lib/test_siphash.c | 18 ++++++
3 files changed, 172 insertions(+), 36 deletions(-)
diff --git a/include/linux/siphash.h b/include/linux/siphash.h
index 145cf5667078..6f5a08a0fc7e 100644
--- a/include/linux/siphash.h
+++ b/include/linux/siphash.h
@@ -29,4 +29,37 @@ static inline u64 siphash_unaligned(const void *data, size_t len,
u64 siphash_unaligned(const void *data, size_t len, const siphash_key_t key);
#endif
+u64 siphash_1u64(const u64 a, const siphash_key_t key);
+u64 siphash_2u64(const u64 a, const u64 b, const siphash_key_t key);
+u64 siphash_3u64(const u64 a, const u64 b, const u64 c,
+ const siphash_key_t key);
+u64 siphash_4u64(const u64 a, const u64 b, const u64 c, const u64 d,
+ const siphash_key_t key);
+
+static inline u64 siphash_2u32(const u32 a, const u32 b, const siphash_key_t key)
+{
+ return siphash_1u64((u64)b << 32 | a, key);
+}
+
+static inline u64 siphash_4u32(const u32 a, const u32 b, const u32 c, const u32 d,
+ const siphash_key_t key)
+{
+ return siphash_2u64((u64)b << 32 | a, (u64)d << 32 | c, key);
+}
+
+static inline u64 siphash_6u32(const u32 a, const u32 b, const u32 c, const u32 d,
+ const u32 e, const u32 f, const siphash_key_t key)
+{
+ return siphash_3u64((u64)b << 32 | a, (u64)d << 32 | c, (u64)f << 32 | e,
+ key);
+}
+
+static inline u64 siphash_8u32(const u32 a, const u32 b, const u32 c, const u32 d,
+ const u32 e, const u32 f, const u32 g, const u32 h,
+ const siphash_key_t key)
+{
+ return siphash_4u64((u64)b << 32 | a, (u64)d << 32 | c, (u64)f << 32 | e,
+ (u64)h << 32 | g, key);
+}
+
#endif /* _LINUX_SIPHASH_H */
diff --git a/lib/siphash.c b/lib/siphash.c
index afc13cbb1b78..970c083ab06a 100644
--- a/lib/siphash.c
+++ b/lib/siphash.c
@@ -25,6 +25,29 @@
v2 += v1; v1 = rol64(v1, 17); v1 ^= v2; v2 = rol64(v2, 32); \
} while(0)
+#define PREAMBLE(len) \
+ u64 v0 = 0x736f6d6570736575ULL; \
+ u64 v1 = 0x646f72616e646f6dULL; \
+ u64 v2 = 0x6c7967656e657261ULL; \
+ u64 v3 = 0x7465646279746573ULL; \
+ u64 b = ((u64)len) << 56; \
+ v3 ^= key[1]; \
+ v2 ^= key[0]; \
+ v1 ^= key[1]; \
+ v0 ^= key[0];
+
+#define POSTAMBLE \
+ v3 ^= b; \
+ SIPROUND; \
+ SIPROUND; \
+ v0 ^= b; \
+ v2 ^= 0xff; \
+ SIPROUND; \
+ SIPROUND; \
+ SIPROUND; \
+ SIPROUND; \
+ return (v0 ^ v1) ^ (v2 ^ v3);
+
/**
* siphash - compute 64-bit siphash PRF value
* @data: buffer to hash, must be aligned to SIPHASH_ALIGNMENT
@@ -33,18 +56,10 @@
*/
u64 siphash(const void *data, size_t len, const siphash_key_t key)
{
- u64 v0 = 0x736f6d6570736575ULL;
- u64 v1 = 0x646f72616e646f6dULL;
- u64 v2 = 0x6c7967656e657261ULL;
- u64 v3 = 0x7465646279746573ULL;
- u64 b = ((u64)len) << 56;
- u64 m;
const u8 *end = data + len - (len % sizeof(u64));
const u8 left = len & (sizeof(u64) - 1);
- v3 ^= key[1];
- v2 ^= key[0];
- v1 ^= key[1];
- v0 ^= key[0];
+ u64 m;
+ PREAMBLE(len)
for (; data != end; data += sizeof(u64)) {
m = le64_to_cpup(data);
v3 ^= m;
@@ -67,16 +82,7 @@ u64 siphash(const void *data, size_t len, const siphash_key_t key)
case 1: b |= end[0];
}
#endif
- v3 ^= b;
- SIPROUND;
- SIPROUND;
- v0 ^= b;
- v2 ^= 0xff;
- SIPROUND;
- SIPROUND;
- SIPROUND;
- SIPROUND;
- return (v0 ^ v1) ^ (v2 ^ v3);
+ POSTAMBLE
}
EXPORT_SYMBOL(siphash);
@@ -89,18 +95,10 @@ EXPORT_SYMBOL(siphash);
*/
u64 siphash_unaligned(const void *data, size_t len, const siphash_key_t key)
{
- u64 v0 = 0x736f6d6570736575ULL;
- u64 v1 = 0x646f72616e646f6dULL;
- u64 v2 = 0x6c7967656e657261ULL;
- u64 v3 = 0x7465646279746573ULL;
- u64 b = ((u64)len) << 56;
- u64 m;
const u8 *end = data + len - (len % sizeof(u64));
const u8 left = len & (sizeof(u64) - 1);
- v3 ^= key[1];
- v2 ^= key[0];
- v1 ^= key[1];
- v0 ^= key[0];
+ u64 m;
+ PREAMBLE(len)
for (; data != end; data += sizeof(u64)) {
m = get_unaligned_le64(data);
v3 ^= m;
@@ -123,16 +121,103 @@ u64 siphash_unaligned(const void *data, size_t len, const siphash_key_t key)
case 1: b |= bytes[0];
}
#endif
- v3 ^= b;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_unaligned);
+#endif
+
+/**
+ * siphash_1u64 - compute 64-bit siphash PRF value of a u64
+ * @first: first u64
+ * @key: the siphash key
+ */
+u64 siphash_1u64(const u64 first, const siphash_key_t key)
+{
+ PREAMBLE(8)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_1u64);
+
+/**
+ * siphash_2u64 - compute 64-bit siphash PRF value of 2 u64
+ * @first: first u64
+ * @second: second u64
+ * @key: the siphash key
+ */
+u64 siphash_2u64(const u64 first, const u64 second, const siphash_key_t key)
+{
+ PREAMBLE(16)
+ v3 ^= first;
SIPROUND;
SIPROUND;
- v0 ^= b;
- v2 ^= 0xff;
+ v0 ^= first;
+ v3 ^= second;
SIPROUND;
SIPROUND;
+ v0 ^= second;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_2u64);
+
+/**
+ * siphash_3u64 - compute 64-bit siphash PRF value of 3 u64
+ * @first: first u64
+ * @second: second u64
+ * @third: third u64
+ * @key: the siphash key
+ */
+u64 siphash_3u64(const u64 first, const u64 second, const u64 third,
+ const siphash_key_t key)
+{
+ PREAMBLE(24)
+ v3 ^= first;
SIPROUND;
SIPROUND;
- return (v0 ^ v1) ^ (v2 ^ v3);
+ v0 ^= first;
+ v3 ^= second;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= second;
+ v3 ^= third;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= third;
+ POSTAMBLE
}
-EXPORT_SYMBOL(siphash_unaligned);
-#endif
+EXPORT_SYMBOL(siphash_3u64);
+
+/**
+ * siphash_4u64 - compute 64-bit siphash PRF value of 4 u64
+ * @first: first u64
+ * @second: second u64
+ * @third: third u64
+ * @forth: forth u64
+ * @key: the siphash key
+ */
+u64 siphash_4u64(const u64 first, const u64 second, const u64 third,
+ const u64 forth, const siphash_key_t key)
+{
+ PREAMBLE(32)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= second;
+ v3 ^= third;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= third;
+ v3 ^= forth;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= forth;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_4u64);
diff --git a/lib/test_siphash.c b/lib/test_siphash.c
index 93549e4e22c5..1635189c171f 100644
--- a/lib/test_siphash.c
+++ b/lib/test_siphash.c
@@ -67,6 +67,24 @@ static int __init siphash_test_init(void)
ret = -EINVAL;
}
}
+ if (siphash_1u64(0x0706050403020100ULL, test_key) != test_vectors[8]) {
+ pr_info("self-test 1u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_2u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL, test_key) != test_vectors[16]) {
+ pr_info("self-test 2u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_3u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL,
+ 0x1716151413121110ULL, test_key) != test_vectors[24]) {
+ pr_info("self-test 3u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_4u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL,
+ 0x1716151413121110ULL, 0x1f1e1d1c1b1a1918ULL, test_key) != test_vectors[32]) {
+ pr_info("self-test 4u64: FAIL\n");
+ ret = -EINVAL;
+ }
if (!ret)
pr_info("self-tests: pass\n");
return ret;
--
2.11.0
This gives a clear speed and security improvement. Siphash is both
faster and is more solid crypto than the aging MD5.
Rather than manually filling MD5 buffers, for IPv6, we simply create
a layout by a simple anonymous struct, for which gcc generates
rather efficient code. For IPv4, we pass the values directly to the
short input convenience functions.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: David Miller <[email protected]>
Cc: David Laight <[email protected]>
Cc: Tom Herbert <[email protected]>
Cc: Hannes Frederic Sowa <[email protected]>
---
net/core/secure_seq.c | 133 ++++++++++++++++++++------------------------------
1 file changed, 52 insertions(+), 81 deletions(-)
diff --git a/net/core/secure_seq.c b/net/core/secure_seq.c
index 88a8e429fc3e..c80583bf3213 100644
--- a/net/core/secure_seq.c
+++ b/net/core/secure_seq.c
@@ -1,3 +1,5 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved. */
+
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/cryptohash.h>
@@ -8,14 +10,14 @@
#include <linux/ktime.h>
#include <linux/string.h>
#include <linux/net.h>
-
+#include <linux/siphash.h>
#include <net/secure_seq.h>
#if IS_ENABLED(CONFIG_IPV6) || IS_ENABLED(CONFIG_INET)
+#include <linux/in6.h>
#include <net/tcp.h>
-#define NET_SECRET_SIZE (MD5_MESSAGE_BYTES / 4)
-static u32 net_secret[NET_SECRET_SIZE] ____cacheline_aligned;
+static siphash_key_t net_secret;
static __always_inline void net_secret_init(void)
{
@@ -44,44 +46,42 @@ static u32 seq_scale(u32 seq)
u32 secure_tcpv6_sequence_number(const __be32 *saddr, const __be32 *daddr,
__be16 sport, __be16 dport, u32 *tsoff)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
- u32 i;
-
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 sport;
+ __be16 dport;
+ u32 padding;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .sport = sport,
+ .dport = dport
+ };
+ u64 hash;
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32)daddr[i];
- secret[4] = net_secret[4] +
- (((__force u16)sport << 16) + (__force u16)dport);
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- *tsoff = sysctl_tcp_timestamps == 1 ? hash[1] : 0;
- return seq_scale(hash[0]);
+ hash = siphash(&combined, sizeof(combined), net_secret);
+ *tsoff = sysctl_tcp_timestamps == 1 ? (hash >> 32) : 0;
+ return seq_scale(hash);
}
EXPORT_SYMBOL(secure_tcpv6_sequence_number);
u32 secure_ipv6_port_ephemeral(const __be32 *saddr, const __be32 *daddr,
__be16 dport)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
- u32 i;
-
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 dport;
+ u16 padding1;
+ u32 padding2;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .dport = dport
+ };
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32) daddr[i];
- secret[4] = net_secret[4] + (__force u32)dport;
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- return hash[0];
+ return siphash(&combined, sizeof(combined), net_secret);
}
EXPORT_SYMBOL(secure_ipv6_port_ephemeral);
#endif
@@ -91,33 +91,17 @@ EXPORT_SYMBOL(secure_ipv6_port_ephemeral);
u32 secure_tcp_sequence_number(__be32 saddr, __be32 daddr,
__be16 sport, __be16 dport, u32 *tsoff)
{
- u32 hash[MD5_DIGEST_WORDS];
-
+ u64 hash;
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = ((__force u16)sport << 16) + (__force u16)dport;
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- *tsoff = sysctl_tcp_timestamps == 1 ? hash[1] : 0;
- return seq_scale(hash[0]);
+ hash = siphash_4u32(saddr, daddr, sport, dport, net_secret);
+ *tsoff = sysctl_tcp_timestamps == 1 ? (hash >> 32) : 0;
+ return seq_scale(hash);
}
u32 secure_ipv4_port_ephemeral(__be32 saddr, __be32 daddr, __be16 dport)
{
- u32 hash[MD5_DIGEST_WORDS];
-
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = (__force u32)dport ^ net_secret[14];
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- return hash[0];
+ return siphash_4u32(saddr, daddr, dport, 0, net_secret);
}
EXPORT_SYMBOL_GPL(secure_ipv4_port_ephemeral);
#endif
@@ -126,21 +110,11 @@ EXPORT_SYMBOL_GPL(secure_ipv4_port_ephemeral);
u64 secure_dccp_sequence_number(__be32 saddr, __be32 daddr,
__be16 sport, __be16 dport)
{
- u32 hash[MD5_DIGEST_WORDS];
u64 seq;
-
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = ((__force u16)sport << 16) + (__force u16)dport;
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- seq = hash[0] | (((u64)hash[1]) << 32);
+ seq = siphash_4u32(saddr, daddr, sport, dport, net_secret);
seq += ktime_get_real_ns();
seq &= (1ull << 48) - 1;
-
return seq;
}
EXPORT_SYMBOL(secure_dccp_sequence_number);
@@ -149,26 +123,23 @@ EXPORT_SYMBOL(secure_dccp_sequence_number);
u64 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr,
__be16 sport, __be16 dport)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 sport;
+ __be16 dport;
+ u32 padding;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .sport = sport,
+ .dport = dport
+ };
u64 seq;
- u32 i;
-
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32)daddr[i];
- secret[4] = net_secret[4] +
- (((__force u16)sport << 16) + (__force u16)dport);
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- seq = hash[0] | (((u64)hash[1]) << 32);
+ seq = siphash(&combined, sizeof(combined), net_secret);
seq += ktime_get_real_ns();
seq &= (1ull << 48) - 1;
-
return seq;
}
EXPORT_SYMBOL(secure_dccpv6_sequence_number);
--
2.11.0
This duplicates the current algorithm for get_random_int/long, but uses
siphash instead. This comes with several benefits. It's certainly
faster and more cryptographically secure than MD5. This patch also
separates hashed fields into three values instead of one, in order to
increase diffusion.
The previous MD5 algorithm used a per-cpu MD5 state, which caused
successive calls to the function to chain upon each other. While it's
not entirely clear that this kind of chaining is absolutely necessary
when using a secure PRF like siphash, it can't hurt, and the timing of
the call chain does add a degree of natural entropy. So, in keeping with
this design, instead of the massive per-cpu 64-byte MD5 state, there is
instead a per-cpu previously returned value for chaining.
The speed benefits are substantial:
| siphash | md5 | speedup |
------------------------------
get_random_long | 137130 | 415983 | 3.03x |
get_random_int | 86384 | 343323 | 3.97x |
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Jean-Philippe Aumasson <[email protected]>
Cc: Ted Tso <[email protected]>
---
drivers/char/random.c | 32 +++++++++++++-------------------
1 file changed, 13 insertions(+), 19 deletions(-)
diff --git a/drivers/char/random.c b/drivers/char/random.c
index d6876d506220..a51f0ff43f00 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -262,6 +262,7 @@
#include <linux/syscalls.h>
#include <linux/completion.h>
#include <linux/uuid.h>
+#include <linux/siphash.h>
#include <crypto/chacha20.h>
#include <asm/processor.h>
@@ -2042,7 +2043,7 @@ struct ctl_table random_table[] = {
};
#endif /* CONFIG_SYSCTL */
-static u32 random_int_secret[MD5_MESSAGE_BYTES / 4] ____cacheline_aligned;
+static siphash_key_t random_int_secret;
int random_int_secret_init(void)
{
@@ -2050,8 +2051,7 @@ int random_int_secret_init(void)
return 0;
}
-static DEFINE_PER_CPU(__u32 [MD5_DIGEST_WORDS], get_random_int_hash)
- __aligned(sizeof(unsigned long));
+static DEFINE_PER_CPU(u64, get_random_int_chaining);
/*
* Get a random word for internal kernel use only. Similar to urandom but
@@ -2061,19 +2061,16 @@ static DEFINE_PER_CPU(__u32 [MD5_DIGEST_WORDS], get_random_int_hash)
*/
unsigned int get_random_int(void)
{
- __u32 *hash;
unsigned int ret;
+ u64 *chaining;
if (arch_get_random_int(&ret))
return ret;
- hash = get_cpu_var(get_random_int_hash);
-
- hash[0] += current->pid + jiffies + random_get_entropy();
- md5_transform(hash, random_int_secret);
- ret = hash[0];
- put_cpu_var(get_random_int_hash);
-
+ chaining = &get_cpu_var(get_random_int_chaining);
+ ret = *chaining = siphash_3u64(*chaining, jiffies, random_get_entropy() +
+ current->pid, random_int_secret);
+ put_cpu_var(get_random_int_chaining);
return ret;
}
EXPORT_SYMBOL(get_random_int);
@@ -2083,19 +2080,16 @@ EXPORT_SYMBOL(get_random_int);
*/
unsigned long get_random_long(void)
{
- __u32 *hash;
unsigned long ret;
+ u64 *chaining;
if (arch_get_random_long(&ret))
return ret;
- hash = get_cpu_var(get_random_int_hash);
-
- hash[0] += current->pid + jiffies + random_get_entropy();
- md5_transform(hash, random_int_secret);
- ret = *(unsigned long *)hash;
- put_cpu_var(get_random_int_hash);
-
+ chaining = &get_cpu_var(get_random_int_chaining);
+ ret = *chaining = siphash_3u64(*chaining, jiffies, random_get_entropy() +
+ current->pid, random_int_secret);
+ put_cpu_var(get_random_int_chaining);
return ret;
}
EXPORT_SYMBOL(get_random_long);
--
2.11.0
> While SipHash is extremely fast for a cryptographically secure function,
> it is likely a tiny bit slower than the insecure jhash, and so replacements
> will be evaluated on a case-by-case basis based on whether or not the
> difference in speed is negligible and whether or not the current jhash usage
> poses a real security risk.
To quantify that, jhash is 27 instructions per 12 bytes of input, with a
dependency path length of 13 instructions. (24/12 in __jash_mix, plus
3/1 for adding the input to the state.) The final add + __jhash_final
is 24 instructions with a path length of 15, which is close enough for
this handwaving. Call it 18n instructions and 8n cycles for 8n bytes.
SipHash (on a 64-bit machine) is 14 instructions with a dependency path
length of 4 *per round*. Two rounds per 8 bytes, plus plus two adds
and one cycle per input word, plus four rounds to finish makes 30n+46
instructions and 9n+16 cycles for 8n bytes.
So *if* you have a 64-bit 4-way superscalar machine, it's not that much
slower once it gets going, but the four-round finalization is quite
noticeable for short inputs.
For typical kernel input lengths "within a factor of 2" is
probably more accurate than "a tiny bit".
You lose a factor of 2 if you machine is 2-way or non-superscalar,
and a second factor of 2 if it's a 32-bit machine.
I mention this because there are a lot of home routers and other netwoek
appliances running Linux on 32-bit ARM and MIPS processors. For those,
it's a factor of *eight*, which is a lot more than "a tiny bit".
The real killer is if you don't have enough registers; SipHash performs
horribly on i386 because it uses more state than i386 has registers.
(If i386 performance is desired, you might ask Jean-Philippe for some
rotate constants for a 32-bit variant with 64 bits of key. Note that
SipHash's security proof requires that key length + input length is
strictly less than the state size, so for a 4x32-bit variant, while
you could stretch the key length a little, you'd have a hard limit at
95 bits.)
A second point, the final XOR in SipHash is either a (very minor) design
mistake, or an opportunity for optimization, depending on how you look
at it. Look at the end of the function:
>+ SIPROUND;
>+ SIPROUND;
>+ return (v0 ^ v1) ^ (v2 ^ v3);
Expanding that out, you get:
+ v0 += v1; v1 = rol64(v1, 13); v1 ^= v0; v0 = rol64(v0, 32);
+ v2 += v3; v3 = rol64(v3, 16); v3 ^= v2;
+ v0 += v3; v3 = rol64(v3, 21); v3 ^= v0;
+ v2 += v1; v1 = rol64(v1, 17); v1 ^= v2; v2 = rol64(v2, 32);
+ return v0 ^ v1 ^ v2 ^ v3;
Since the final XOR includes both v0 and v3, it's undoing the "v3 ^= v0"
two lines earlier, so the value of v0 doesn't matter after its XOR into
v1 on line one.
The final SIPROUND and return can then be optimized to
+ v0 += v1; v1 = rol64(v1, 13); v1 ^= v0;
+ v2 += v3; v3 = rol64(v3, 16); v3 ^= v2;
+ v3 = rol64(v3, 21);
+ v2 += v1; v1 = rol64(v1, 17); v1 ^= v2; v2 = rol64(v2, 32);
+ return v1 ^ v2 ^ v3;
A 32-bit implementation could further tweak the 4 instructions of
v1 ^= v2; v2 = rol64(v2, 32); v1 ^= v2;
gcc 6.2.1 -O3 compiles it to basically:
v1.low ^= v2.low;
v1.high ^= v2.high;
v1.low ^= v2.high;
v1.high ^= v2.low;
but it could be written as:
v2.low ^= v2.high;
v1.low ^= v2.low;
v1.high ^= v2.low;
Alternatively, if it's for private use only (key not shared with other
systems), a slightly stronger variant would "return v1 ^ v3;".
(The final swap of v2 is dead code, but a compiler can spot that easily.)
If a halved version of SipHash can bring significant performance boost
(with 32b words instead of 64b words) with an acceptable security level
(64-bit enough?) then we may design such a version.
Regarding output size, are 64 bits sufficient?
On Thu, 15 Dec 2016 at 23:42, George Spelvin <[email protected]>
wrote:
> > While SipHash is extremely fast for a cryptographically secure function,
> > it is likely a tiny bit slower than the insecure jhash, and so
> replacements
> > will be evaluated on a case-by-case basis based on whether or not the
> > difference in speed is negligible and whether or not the current jhash
> usage
> > poses a real security risk.
>
> To quantify that, jhash is 27 instructions per 12 bytes of input, with a
> dependency path length of 13 instructions. (24/12 in __jash_mix, plus
> 3/1 for adding the input to the state.) The final add + __jhash_final
> is 24 instructions with a path length of 15, which is close enough for
> this handwaving. Call it 18n instructions and 8n cycles for 8n bytes.
>
> SipHash (on a 64-bit machine) is 14 instructions with a dependency path
> length of 4 *per round*. Two rounds per 8 bytes, plus plus two adds
> and one cycle per input word, plus four rounds to finish makes 30n+46
> instructions and 9n+16 cycles for 8n bytes.
>
> So *if* you have a 64-bit 4-way superscalar machine, it's not that much
> slower once it gets going, but the four-round finalization is quite
> noticeable for short inputs.
>
> For typical kernel input lengths "within a factor of 2" is
> probably more accurate than "a tiny bit".
>
> You lose a factor of 2 if you machine is 2-way or non-superscalar,
> and a second factor of 2 if it's a 32-bit machine.
>
> I mention this because there are a lot of home routers and other netwoek
> appliances running Linux on 32-bit ARM and MIPS processors. For those,
> it's a factor of *eight*, which is a lot more than "a tiny bit".
>
> The real killer is if you don't have enough registers; SipHash performs
> horribly on i386 because it uses more state than i386 has registers.
>
> (If i386 performance is desired, you might ask Jean-Philippe for some
> rotate constants for a 32-bit variant with 64 bits of key. Note that
> SipHash's security proof requires that key length + input length is
> strictly less than the state size, so for a 4x32-bit variant, while
> you could stretch the key length a little, you'd have a hard limit at
> 95 bits.)
>
>
> A second point, the final XOR in SipHash is either a (very minor) design
> mistake, or an opportunity for optimization, depending on how you look
> at it. Look at the end of the function:
>
> >+ SIPROUND;
> >+ SIPROUND;
> >+ return (v0 ^ v1) ^ (v2 ^ v3);
>
> Expanding that out, you get:
> + v0 += v1; v1 = rol64(v1, 13); v1 ^= v0; v0 = rol64(v0, 32);
> + v2 += v3; v3 = rol64(v3, 16); v3 ^= v2;
> + v0 += v3; v3 = rol64(v3, 21); v3 ^= v0;
> + v2 += v1; v1 = rol64(v1, 17); v1 ^= v2; v2 = rol64(v2, 32);
> + return v0 ^ v1 ^ v2 ^ v3;
>
> Since the final XOR includes both v0 and v3, it's undoing the "v3 ^= v0"
> two lines earlier, so the value of v0 doesn't matter after its XOR into
> v1 on line one.
>
> The final SIPROUND and return can then be optimized to
>
> + v0 += v1; v1 = rol64(v1, 13); v1 ^= v0;
> + v2 += v3; v3 = rol64(v3, 16); v3 ^= v2;
> + v3 = rol64(v3, 21);
> + v2 += v1; v1 = rol64(v1, 17); v1 ^= v2; v2 = rol64(v2, 32);
> + return v1 ^ v2 ^ v3;
>
> A 32-bit implementation could further tweak the 4 instructions of
> v1 ^= v2; v2 = rol64(v2, 32); v1 ^= v2;
>
> gcc 6.2.1 -O3 compiles it to basically:
> v1.low ^= v2.low;
> v1.high ^= v2.high;
> v1.low ^= v2.high;
> v1.high ^= v2.low;
> but it could be written as:
> v2.low ^= v2.high;
> v1.low ^= v2.low;
> v1.high ^= v2.low;
>
> Alternatively, if it's for private use only (key not shared with other
> systems), a slightly stronger variant would "return v1 ^ v3;".
> (The final swap of v2 is dead code, but a compiler can spot that easily.)
>
> If a halved version of SipHash can bring significant performance boost
> (with 32b words instead of 64b words) with an acceptable security level
> (64-bit enough?) then we may design such a version.
I was thinking if the key could be pushed to 80 bits, that would be nice,
but honestly 64 bits is fine. This is DoS protection, and while it's
possible to brute-force a 64 bit secret, there are more effective (DDoS)
attacks possible for the same cost.
(I'd suggest a name of "HalfSipHash" to convey the reduced security
effectively.)
> Regarding output size, are 64 bits sufficient?
As a replacement for jhash, 32 bits are sufficient. It's for
indexing an in-memory hash table on a 32-bit machine.
(When you're done thinking about this, as a matter of personal interest
I'd love a hash expert's opinion on
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=2a18da7a9c7886f1c7307f8d3f23f24318583f03
and
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=8387ff2577eb9ed245df9a39947f66976c6bcd02
which is a non-cryptographic hash function of novel design that's
inspired by SipHash.)
Hi Jason,
[auto build test ERROR on linus/master]
[also build test ERROR on v4.9 next-20161215]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Jason-A-Donenfeld/siphash-add-cryptographically-secure-PRF/20161216-092837
config: ia64-allmodconfig (attached as .config)
compiler: ia64-linux-gcc (GCC) 6.2.0
reproduce:
wget https://git.kernel.org/cgit/linux/kernel/git/wfg/lkp-tests.git/plain/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
make.cross ARCH=ia64
All errors (new ones prefixed by >>):
lib/siphash.c: In function 'siphash_unaligned':
>> lib/siphash.c:123:15: error: 'bytes' undeclared (first use in this function)
case 1: b |= bytes[0];
^~~~~
lib/siphash.c:123:15: note: each undeclared identifier is reported only once for each function it appears in
vim +/bytes +123 lib/siphash.c
117 case 7: b |= ((u64)end[6]) << 48;
118 case 6: b |= ((u64)end[5]) << 40;
119 case 5: b |= ((u64)end[4]) << 32;
120 case 4: b |= get_unaligned_le32(end); break;
121 case 3: b |= ((u64)end[2]) << 16;
122 case 2: b |= get_unaligned_le16(end); break;
> 123 case 1: b |= bytes[0];
124 }
125 #endif
126 v3 ^= b;
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
Hey again,
This keeps getting more ambitious, which is good I suppose. If the frequency
of new versioned patchsets is too high for LKML and not customary, please let
me know. Otherwise, read on to see what's new this time...
With Hannes' suggestion, there is now only one siphash() function, which will
use the faster aligned version by compile-time constant folding. Additionally,
I now use constant folding to optionally switch to the helper siphash_Nu64
functions that are a bit faster for data of length 8, 16, 24, and 32. So, the
result is that you use siphash(data, len, key) if you have a buffer of sorts,
and then everything is taken care of for you. Or, if you have a series of
integers, you can opt to use siphash_Nu{32,64} functions instead. The basic
API is now complete.
After replacing MD5 in secure sequence number generation and the RNG, it
turned out that md5_transform wasn't used any place else in the tree, so
finally -- this is something to rejoice over -- lib/md5.c has been deleted and
now that function lives as a static function in crypto/md5.c where it belongs.
Meanwhile, it seems that sha_transform is used in places where SipHash would
be more fitting, so the IPv4 and IPv6 syncookies implementation now uses
SipHash, which should speed up TCP performance. Some BSDs already do this.
I'd like to replace sha_transform in addrconf, but that code is a bit gnarley,
so I don't want to be too meddlesome. I'm not entirely convinced either that
SipHash is a good choice for it. But I'm open to discussion here, so if you
have an opinion, please speak up.
If you've been following the evolution of this patchset, and think that
certain patches in it are fine, please do lend me your Reviewed-by to carry
into any subsequent versions, so that in case you disappear your useful
reviews will still keep the ball moving.
Thanks for all the great feedback thus far.
Jason
Jason A. Donenfeld (5):
siphash: add cryptographically secure PRF
secure_seq: use SipHash in place of MD5
random: use SipHash in place of MD5
md5: remove from lib and only live in crypto
syncookies: use SipHash in place of SHA1
MAINTAINERS | 7 ++
crypto/md5.c | 95 +++++++++++++++++++++-
drivers/char/random.c | 32 +++-----
include/linux/siphash.h | 86 ++++++++++++++++++++
lib/Kconfig.debug | 6 +-
lib/Makefile | 7 +-
lib/md5.c | 95 ----------------------
lib/siphash.c | 210 ++++++++++++++++++++++++++++++++++++++++++++++++
lib/test_siphash.c | 101 +++++++++++++++++++++++
net/core/secure_seq.c | 133 ++++++++++++------------------
net/ipv4/syncookies.c | 20 +----
net/ipv6/syncookies.c | 37 ++++-----
12 files changed, 590 insertions(+), 239 deletions(-)
create mode 100644 include/linux/siphash.h
delete mode 100644 lib/md5.c
create mode 100644 lib/siphash.c
create mode 100644 lib/test_siphash.c
--
2.11.0
This gives a clear speed and security improvement. Siphash is both
faster and is more solid crypto than the aging MD5.
Rather than manually filling MD5 buffers, for IPv6, we simply create
a layout by a simple anonymous struct, for which gcc generates
rather efficient code. For IPv4, we pass the values directly to the
short input convenience functions.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: David Miller <[email protected]>
Cc: David Laight <[email protected]>
Cc: Tom Herbert <[email protected]>
Cc: Hannes Frederic Sowa <[email protected]>
---
net/core/secure_seq.c | 133 ++++++++++++++++++++------------------------------
1 file changed, 52 insertions(+), 81 deletions(-)
diff --git a/net/core/secure_seq.c b/net/core/secure_seq.c
index 88a8e429fc3e..c80583bf3213 100644
--- a/net/core/secure_seq.c
+++ b/net/core/secure_seq.c
@@ -1,3 +1,5 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved. */
+
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/cryptohash.h>
@@ -8,14 +10,14 @@
#include <linux/ktime.h>
#include <linux/string.h>
#include <linux/net.h>
-
+#include <linux/siphash.h>
#include <net/secure_seq.h>
#if IS_ENABLED(CONFIG_IPV6) || IS_ENABLED(CONFIG_INET)
+#include <linux/in6.h>
#include <net/tcp.h>
-#define NET_SECRET_SIZE (MD5_MESSAGE_BYTES / 4)
-static u32 net_secret[NET_SECRET_SIZE] ____cacheline_aligned;
+static siphash_key_t net_secret;
static __always_inline void net_secret_init(void)
{
@@ -44,44 +46,42 @@ static u32 seq_scale(u32 seq)
u32 secure_tcpv6_sequence_number(const __be32 *saddr, const __be32 *daddr,
__be16 sport, __be16 dport, u32 *tsoff)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
- u32 i;
-
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 sport;
+ __be16 dport;
+ u32 padding;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .sport = sport,
+ .dport = dport
+ };
+ u64 hash;
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32)daddr[i];
- secret[4] = net_secret[4] +
- (((__force u16)sport << 16) + (__force u16)dport);
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- *tsoff = sysctl_tcp_timestamps == 1 ? hash[1] : 0;
- return seq_scale(hash[0]);
+ hash = siphash(&combined, sizeof(combined), net_secret);
+ *tsoff = sysctl_tcp_timestamps == 1 ? (hash >> 32) : 0;
+ return seq_scale(hash);
}
EXPORT_SYMBOL(secure_tcpv6_sequence_number);
u32 secure_ipv6_port_ephemeral(const __be32 *saddr, const __be32 *daddr,
__be16 dport)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
- u32 i;
-
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 dport;
+ u16 padding1;
+ u32 padding2;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .dport = dport
+ };
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32) daddr[i];
- secret[4] = net_secret[4] + (__force u32)dport;
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- return hash[0];
+ return siphash(&combined, sizeof(combined), net_secret);
}
EXPORT_SYMBOL(secure_ipv6_port_ephemeral);
#endif
@@ -91,33 +91,17 @@ EXPORT_SYMBOL(secure_ipv6_port_ephemeral);
u32 secure_tcp_sequence_number(__be32 saddr, __be32 daddr,
__be16 sport, __be16 dport, u32 *tsoff)
{
- u32 hash[MD5_DIGEST_WORDS];
-
+ u64 hash;
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = ((__force u16)sport << 16) + (__force u16)dport;
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- *tsoff = sysctl_tcp_timestamps == 1 ? hash[1] : 0;
- return seq_scale(hash[0]);
+ hash = siphash_4u32(saddr, daddr, sport, dport, net_secret);
+ *tsoff = sysctl_tcp_timestamps == 1 ? (hash >> 32) : 0;
+ return seq_scale(hash);
}
u32 secure_ipv4_port_ephemeral(__be32 saddr, __be32 daddr, __be16 dport)
{
- u32 hash[MD5_DIGEST_WORDS];
-
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = (__force u32)dport ^ net_secret[14];
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- return hash[0];
+ return siphash_4u32(saddr, daddr, dport, 0, net_secret);
}
EXPORT_SYMBOL_GPL(secure_ipv4_port_ephemeral);
#endif
@@ -126,21 +110,11 @@ EXPORT_SYMBOL_GPL(secure_ipv4_port_ephemeral);
u64 secure_dccp_sequence_number(__be32 saddr, __be32 daddr,
__be16 sport, __be16 dport)
{
- u32 hash[MD5_DIGEST_WORDS];
u64 seq;
-
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = ((__force u16)sport << 16) + (__force u16)dport;
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- seq = hash[0] | (((u64)hash[1]) << 32);
+ seq = siphash_4u32(saddr, daddr, sport, dport, net_secret);
seq += ktime_get_real_ns();
seq &= (1ull << 48) - 1;
-
return seq;
}
EXPORT_SYMBOL(secure_dccp_sequence_number);
@@ -149,26 +123,23 @@ EXPORT_SYMBOL(secure_dccp_sequence_number);
u64 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr,
__be16 sport, __be16 dport)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 sport;
+ __be16 dport;
+ u32 padding;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .sport = sport,
+ .dport = dport
+ };
u64 seq;
- u32 i;
-
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32)daddr[i];
- secret[4] = net_secret[4] +
- (((__force u16)sport << 16) + (__force u16)dport);
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- seq = hash[0] | (((u64)hash[1]) << 32);
+ seq = siphash(&combined, sizeof(combined), net_secret);
seq += ktime_get_real_ns();
seq &= (1ull << 48) - 1;
-
return seq;
}
EXPORT_SYMBOL(secure_dccpv6_sequence_number);
--
2.11.0
SipHash is a 64-bit keyed hash function that is actually a
cryptographically secure PRF, like HMAC. Except SipHash is super fast,
and is meant to be used as a hashtable keyed lookup function, or as a
general PRF for short input use cases, such as sequence numbers or RNG
chaining.
For the first usage:
There are a variety of attacks known as "hashtable poisoning" in which an
attacker forms some data such that the hash of that data will be the
same, and then preceeds to fill up all entries of a hashbucket. This is
a realistic and well-known denial-of-service vector. Currently
hashtables use jhash, which is fast but not secure, and some kind of
rotating key scheme (or none at all, which isn't good). SipHash is meant
as a replacement for jhash in these cases.
There are a modicum of places in the kernel that are vulnerable to
hashtable poisoning attacks, either via userspace vectors or network
vectors, and there's not a reliable mechanism inside the kernel at the
moment to fix it. The first step toward fixing these issues is actually
getting a secure primitive into the kernel for developers to use. Then
we can, bit by bit, port things over to it as deemed appropriate.
While SipHash is extremely fast for a cryptographically secure function,
it is likely a bit slower than the insecure jhash, and so replacements
will be evaluated on a case-by-case basis based on whether or not the
difference in speed is negligible and whether or not the current jhash usage
poses a real security risk.
For the second usage:
A few places in the kernel are using MD5 or SHA1 for creating secure
sequence numbers, syn cookies, port numbers, or fast random numbers.
SipHash is a faster and more fitting, and more secure replacement for MD5
in those situations. Replacing MD5 and SHA1 with SipHash for these uses is
obvious and straight-forward, and so is submitted along with this patch
series. There shouldn't be much of a debate over its efficacy.
Dozens of languages are already using this internally for their hash
tables and PRFs. Some of the BSDs already use this in their kernels.
SipHash is a widely known high-speed solution to a widely known set of
problems, and it's time we catch-up.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Jean-Philippe Aumasson <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Eric Biggers <[email protected]>
Cc: David Laight <[email protected]>
---
MAINTAINERS | 7 ++
include/linux/siphash.h | 86 ++++++++++++++++++++
lib/Kconfig.debug | 6 +-
lib/Makefile | 5 +-
lib/siphash.c | 210 ++++++++++++++++++++++++++++++++++++++++++++++++
lib/test_siphash.c | 101 +++++++++++++++++++++++
6 files changed, 410 insertions(+), 5 deletions(-)
create mode 100644 include/linux/siphash.h
create mode 100644 lib/siphash.c
create mode 100644 lib/test_siphash.c
diff --git a/MAINTAINERS b/MAINTAINERS
index 59c9895d73d5..5d87a8c1056a 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -11231,6 +11231,13 @@ F: arch/arm/mach-s3c24xx/mach-bast.c
F: arch/arm/mach-s3c24xx/bast-ide.c
F: arch/arm/mach-s3c24xx/bast-irq.c
+SIPHASH PRF ROUTINES
+M: Jason A. Donenfeld <[email protected]>
+S: Maintained
+F: lib/siphash.c
+F: lib/test_siphash.c
+F: include/linux/siphash.h
+
TI DAVINCI MACHINE SUPPORT
M: Sekhar Nori <[email protected]>
M: Kevin Hilman <[email protected]>
diff --git a/include/linux/siphash.h b/include/linux/siphash.h
new file mode 100644
index 000000000000..e82fce48a0f6
--- /dev/null
+++ b/include/linux/siphash.h
@@ -0,0 +1,86 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#ifndef _LINUX_SIPHASH_H
+#define _LINUX_SIPHASH_H
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+
+#define SIPHASH_ALIGNMENT 8
+typedef u64 siphash_key_t[2];
+
+u64 __siphash_aligned(const void *data, size_t len, const siphash_key_t key);
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+u64 __siphash_unaligned(const void *data, size_t len, const siphash_key_t key);
+#endif
+
+u64 siphash_1u64(const u64 a, const siphash_key_t key);
+u64 siphash_2u64(const u64 a, const u64 b, const siphash_key_t key);
+u64 siphash_3u64(const u64 a, const u64 b, const u64 c,
+ const siphash_key_t key);
+u64 siphash_4u64(const u64 a, const u64 b, const u64 c, const u64 d,
+ const siphash_key_t key);
+
+static inline u64 ___siphash_aligned(const u64 *data, size_t len, const siphash_key_t key)
+{
+ if (__builtin_constant_p(len) && len == 8)
+ return siphash_1u64(data[0], key);
+ if (__builtin_constant_p(len) && len == 16)
+ return siphash_2u64(data[0], data[1], key);
+ if (__builtin_constant_p(len) && len == 24)
+ return siphash_3u64(data[0], data[1], data[2], key);
+ if (__builtin_constant_p(len) && len == 32)
+ return siphash_4u64(data[0], data[1], data[2], data[3], key);
+ return __siphash_aligned(data, len, key);
+}
+
+/**
+ * siphash - compute 64-bit siphash PRF value
+ * @data: buffer to hash
+ * @size: size of @data
+ * @key: the siphash key
+ */
+static inline u64 siphash(const void *data, size_t len, const siphash_key_t key)
+{
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+ if (!IS_ALIGNED((unsigned long)data, SIPHASH_ALIGNMENT))
+ return __siphash_unaligned(data, len, key);
+#endif
+ return ___siphash_aligned(data, len, key);
+}
+
+static inline u64 siphash_2u32(const u32 a, const u32 b, const siphash_key_t key)
+{
+ return siphash_1u64((u64)b << 32 | a, key);
+}
+
+static inline u64 siphash_4u32(const u32 a, const u32 b, const u32 c, const u32 d,
+ const siphash_key_t key)
+{
+ return siphash_2u64((u64)b << 32 | a, (u64)d << 32 | c, key);
+}
+
+static inline u64 siphash_6u32(const u32 a, const u32 b, const u32 c, const u32 d,
+ const u32 e, const u32 f, const siphash_key_t key)
+{
+ return siphash_3u64((u64)b << 32 | a, (u64)d << 32 | c, (u64)f << 32 | e,
+ key);
+}
+
+static inline u64 siphash_8u32(const u32 a, const u32 b, const u32 c, const u32 d,
+ const u32 e, const u32 f, const u32 g, const u32 h,
+ const siphash_key_t key)
+{
+ return siphash_4u64((u64)b << 32 | a, (u64)d << 32 | c, (u64)f << 32 | e,
+ (u64)h << 32 | g, key);
+}
+
+#endif /* _LINUX_SIPHASH_H */
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 7446097f72bd..86254ea99b45 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1843,9 +1843,9 @@ config TEST_HASH
tristate "Perform selftest on hash functions"
default n
help
- Enable this option to test the kernel's integer (<linux/hash,h>)
- and string (<linux/stringhash.h>) hash functions on boot
- (or module load).
+ Enable this option to test the kernel's integer (<linux/hash.h>),
+ string (<linux/stringhash.h>), and siphash (<linux/siphash.h>)
+ hash functions on boot (or module load).
This is intended to help people writing architecture-specific
optimized versions. If unsure, say N.
diff --git a/lib/Makefile b/lib/Makefile
index 50144a3aeebd..71d398b04a74 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -22,7 +22,8 @@ lib-y := ctype.o string.o vsprintf.o cmdline.o \
sha1.o chacha20.o md5.o irq_regs.o argv_split.o \
flex_proportions.o ratelimit.o show_mem.o \
is_single_threaded.o plist.o decompress.o kobject_uevent.o \
- earlycpio.o seq_buf.o nmi_backtrace.o nodemask.o win_minmax.o
+ earlycpio.o seq_buf.o siphash.o \
+ nmi_backtrace.o nodemask.o win_minmax.o
lib-$(CONFIG_MMU) += ioremap.o
lib-$(CONFIG_SMP) += cpumask.o
@@ -44,7 +45,7 @@ obj-$(CONFIG_TEST_HEXDUMP) += test_hexdump.o
obj-y += kstrtox.o
obj-$(CONFIG_TEST_BPF) += test_bpf.o
obj-$(CONFIG_TEST_FIRMWARE) += test_firmware.o
-obj-$(CONFIG_TEST_HASH) += test_hash.o
+obj-$(CONFIG_TEST_HASH) += test_hash.o test_siphash.o
obj-$(CONFIG_TEST_KASAN) += test_kasan.o
obj-$(CONFIG_TEST_KSTRTOX) += test-kstrtox.o
obj-$(CONFIG_TEST_LKM) += test_module.o
diff --git a/lib/siphash.c b/lib/siphash.c
new file mode 100644
index 000000000000..7efc273de5d0
--- /dev/null
+++ b/lib/siphash.c
@@ -0,0 +1,210 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#include <linux/siphash.h>
+#include <asm/unaligned.h>
+
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+#include <linux/dcache.h>
+#include <asm/word-at-a-time.h>
+#endif
+
+#define SIPROUND \
+ do { \
+ v0 += v1; v1 = rol64(v1, 13); v1 ^= v0; v0 = rol64(v0, 32); \
+ v2 += v3; v3 = rol64(v3, 16); v3 ^= v2; \
+ v0 += v3; v3 = rol64(v3, 21); v3 ^= v0; \
+ v2 += v1; v1 = rol64(v1, 17); v1 ^= v2; v2 = rol64(v2, 32); \
+ } while(0)
+
+#define PREAMBLE(len) \
+ u64 v0 = 0x736f6d6570736575ULL; \
+ u64 v1 = 0x646f72616e646f6dULL; \
+ u64 v2 = 0x6c7967656e657261ULL; \
+ u64 v3 = 0x7465646279746573ULL; \
+ u64 b = ((u64)len) << 56; \
+ v3 ^= key[1]; \
+ v2 ^= key[0]; \
+ v1 ^= key[1]; \
+ v0 ^= key[0];
+
+#define POSTAMBLE \
+ v3 ^= b; \
+ SIPROUND; \
+ SIPROUND; \
+ v0 ^= b; \
+ v2 ^= 0xff; \
+ SIPROUND; \
+ SIPROUND; \
+ SIPROUND; \
+ SIPROUND; \
+ return (v0 ^ v1) ^ (v2 ^ v3);
+
+u64 __siphash_aligned(const void *data, size_t len, const siphash_key_t key)
+{
+ const u8 *end = data + len - (len % sizeof(u64));
+ const u8 left = len & (sizeof(u64) - 1);
+ u64 m;
+ PREAMBLE(len)
+ for (; data != end; data += sizeof(u64)) {
+ m = le64_to_cpup(data);
+ v3 ^= m;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= m;
+ }
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+ if (left)
+ b |= le64_to_cpu((__force __le64)(load_unaligned_zeropad(data) &
+ bytemask_from_count(left)));
+#else
+ switch (left) {
+ case 7: b |= ((u64)end[6]) << 48;
+ case 6: b |= ((u64)end[5]) << 40;
+ case 5: b |= ((u64)end[4]) << 32;
+ case 4: b |= le32_to_cpup(data); break;
+ case 3: b |= ((u64)end[2]) << 16;
+ case 2: b |= le16_to_cpup(data); break;
+ case 1: b |= end[0];
+ }
+#endif
+ POSTAMBLE
+}
+EXPORT_SYMBOL(__siphash_aligned);
+
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+u64 __siphash_unaligned(const void *data, size_t len, const siphash_key_t key)
+{
+ const u8 *end = data + len - (len % sizeof(u64));
+ const u8 left = len & (sizeof(u64) - 1);
+ u64 m;
+ PREAMBLE(len)
+ for (; data != end; data += sizeof(u64)) {
+ m = get_unaligned_le64(data);
+ v3 ^= m;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= m;
+ }
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+ if (left)
+ b |= le64_to_cpu((__force __le64)(load_unaligned_zeropad(data) &
+ bytemask_from_count(left)));
+#else
+ switch (left) {
+ case 7: b |= ((u64)end[6]) << 48;
+ case 6: b |= ((u64)end[5]) << 40;
+ case 5: b |= ((u64)end[4]) << 32;
+ case 4: b |= get_unaligned_le32(end); break;
+ case 3: b |= ((u64)end[2]) << 16;
+ case 2: b |= get_unaligned_le16(end); break;
+ case 1: b |= end[0];
+ }
+#endif
+ POSTAMBLE
+}
+EXPORT_SYMBOL(__siphash_unaligned);
+#endif
+
+/**
+ * siphash_1u64 - compute 64-bit siphash PRF value of a u64
+ * @first: first u64
+ * @key: the siphash key
+ */
+u64 siphash_1u64(const u64 first, const siphash_key_t key)
+{
+ PREAMBLE(8)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_1u64);
+
+/**
+ * siphash_2u64 - compute 64-bit siphash PRF value of 2 u64
+ * @first: first u64
+ * @second: second u64
+ * @key: the siphash key
+ */
+u64 siphash_2u64(const u64 first, const u64 second, const siphash_key_t key)
+{
+ PREAMBLE(16)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= second;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_2u64);
+
+/**
+ * siphash_3u64 - compute 64-bit siphash PRF value of 3 u64
+ * @first: first u64
+ * @second: second u64
+ * @third: third u64
+ * @key: the siphash key
+ */
+u64 siphash_3u64(const u64 first, const u64 second, const u64 third,
+ const siphash_key_t key)
+{
+ PREAMBLE(24)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= second;
+ v3 ^= third;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= third;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_3u64);
+
+/**
+ * siphash_4u64 - compute 64-bit siphash PRF value of 4 u64
+ * @first: first u64
+ * @second: second u64
+ * @third: third u64
+ * @forth: forth u64
+ * @key: the siphash key
+ */
+u64 siphash_4u64(const u64 first, const u64 second, const u64 third,
+ const u64 forth, const siphash_key_t key)
+{
+ PREAMBLE(32)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= second;
+ v3 ^= third;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= third;
+ v3 ^= forth;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= forth;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_4u64);
diff --git a/lib/test_siphash.c b/lib/test_siphash.c
new file mode 100644
index 000000000000..906e58a2c946
--- /dev/null
+++ b/lib/test_siphash.c
@@ -0,0 +1,101 @@
+/* Test cases for siphash.c
+ *
+ * Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/siphash.h>
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/module.h>
+
+/* Test vectors taken from official reference source available at:
+ * https://131002.net/siphash/siphash24.c
+ */
+static const u64 test_vectors[64] = {
+ 0x726fdb47dd0e0e31ULL, 0x74f839c593dc67fdULL, 0x0d6c8009d9a94f5aULL,
+ 0x85676696d7fb7e2dULL, 0xcf2794e0277187b7ULL, 0x18765564cd99a68dULL,
+ 0xcbc9466e58fee3ceULL, 0xab0200f58b01d137ULL, 0x93f5f5799a932462ULL,
+ 0x9e0082df0ba9e4b0ULL, 0x7a5dbbc594ddb9f3ULL, 0xf4b32f46226bada7ULL,
+ 0x751e8fbc860ee5fbULL, 0x14ea5627c0843d90ULL, 0xf723ca908e7af2eeULL,
+ 0xa129ca6149be45e5ULL, 0x3f2acc7f57c29bdbULL, 0x699ae9f52cbe4794ULL,
+ 0x4bc1b3f0968dd39cULL, 0xbb6dc91da77961bdULL, 0xbed65cf21aa2ee98ULL,
+ 0xd0f2cbb02e3b67c7ULL, 0x93536795e3a33e88ULL, 0xa80c038ccd5ccec8ULL,
+ 0xb8ad50c6f649af94ULL, 0xbce192de8a85b8eaULL, 0x17d835b85bbb15f3ULL,
+ 0x2f2e6163076bcfadULL, 0xde4daaaca71dc9a5ULL, 0xa6a2506687956571ULL,
+ 0xad87a3535c49ef28ULL, 0x32d892fad841c342ULL, 0x7127512f72f27cceULL,
+ 0xa7f32346f95978e3ULL, 0x12e0b01abb051238ULL, 0x15e034d40fa197aeULL,
+ 0x314dffbe0815a3b4ULL, 0x027990f029623981ULL, 0xcadcd4e59ef40c4dULL,
+ 0x9abfd8766a33735cULL, 0x0e3ea96b5304a7d0ULL, 0xad0c42d6fc585992ULL,
+ 0x187306c89bc215a9ULL, 0xd4a60abcf3792b95ULL, 0xf935451de4f21df2ULL,
+ 0xa9538f0419755787ULL, 0xdb9acddff56ca510ULL, 0xd06c98cd5c0975ebULL,
+ 0xe612a3cb9ecba951ULL, 0xc766e62cfcadaf96ULL, 0xee64435a9752fe72ULL,
+ 0xa192d576b245165aULL, 0x0a8787bf8ecb74b2ULL, 0x81b3e73d20b49b6fULL,
+ 0x7fa8220ba3b2eceaULL, 0x245731c13ca42499ULL, 0xb78dbfaf3a8d83bdULL,
+ 0xea1ad565322a1a0bULL, 0x60e61c23a3795013ULL, 0x6606d7e446282b93ULL,
+ 0x6ca4ecb15c5f91e1ULL, 0x9f626da15c9625f3ULL, 0xe51b38608ef25f57ULL,
+ 0x958a324ceb064572ULL
+};
+static const siphash_key_t test_key =
+ { 0x0706050403020100ULL , 0x0f0e0d0c0b0a0908ULL };
+
+static int __init siphash_test_init(void)
+{
+ u8 in[64] __aligned(SIPHASH_ALIGNMENT);
+ u8 in_unaligned[65];
+ u8 i;
+ int ret = 0;
+
+ for (i = 0; i < 64; ++i) {
+ in[i] = i;
+ in_unaligned[i + 1] = i;
+ if (siphash(in, i, test_key) != test_vectors[i]) {
+ pr_info("self-test aligned %u: FAIL\n", i + 1);
+ ret = -EINVAL;
+ }
+ if (siphash(in_unaligned + 1, i, test_key) != test_vectors[i]) {
+ pr_info("self-test unaligned %u: FAIL\n", i + 1);
+ ret = -EINVAL;
+ }
+ }
+ if (siphash_1u64(0x0706050403020100ULL, test_key) != test_vectors[8]) {
+ pr_info("self-test 1u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_2u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL, test_key) != test_vectors[16]) {
+ pr_info("self-test 2u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_3u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL,
+ 0x1716151413121110ULL, test_key) != test_vectors[24]) {
+ pr_info("self-test 3u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_4u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL,
+ 0x1716151413121110ULL, 0x1f1e1d1c1b1a1918ULL, test_key) != test_vectors[32]) {
+ pr_info("self-test 4u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (!ret)
+ pr_info("self-tests: pass\n");
+ return ret;
+}
+
+static void __exit siphash_test_exit(void)
+{
+}
+
+module_init(siphash_test_init);
+module_exit(siphash_test_exit);
+
+MODULE_AUTHOR("Jason A. Donenfeld <[email protected]>");
+MODULE_LICENSE("Dual BSD/GPL");
--
2.11.0
This duplicates the current algorithm for get_random_int/long, but uses
siphash instead. This comes with several benefits. It's certainly
faster and more cryptographically secure than MD5. This patch also
separates hashed fields into three values instead of one, in order to
increase diffusion.
The previous MD5 algorithm used a per-cpu MD5 state, which caused
successive calls to the function to chain upon each other. While it's
not entirely clear that this kind of chaining is absolutely necessary
when using a secure PRF like siphash, it can't hurt, and the timing of
the call chain does add a degree of natural entropy. So, in keeping with
this design, instead of the massive per-cpu 64-byte MD5 state, there is
instead a per-cpu previously returned value for chaining.
The speed benefits are substantial:
| siphash | md5 | speedup |
------------------------------
get_random_long | 137130 | 415983 | 3.03x |
get_random_int | 86384 | 343323 | 3.97x |
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Jean-Philippe Aumasson <[email protected]>
Cc: Ted Tso <[email protected]>
---
drivers/char/random.c | 32 +++++++++++++-------------------
1 file changed, 13 insertions(+), 19 deletions(-)
diff --git a/drivers/char/random.c b/drivers/char/random.c
index d6876d506220..a51f0ff43f00 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -262,6 +262,7 @@
#include <linux/syscalls.h>
#include <linux/completion.h>
#include <linux/uuid.h>
+#include <linux/siphash.h>
#include <crypto/chacha20.h>
#include <asm/processor.h>
@@ -2042,7 +2043,7 @@ struct ctl_table random_table[] = {
};
#endif /* CONFIG_SYSCTL */
-static u32 random_int_secret[MD5_MESSAGE_BYTES / 4] ____cacheline_aligned;
+static siphash_key_t random_int_secret;
int random_int_secret_init(void)
{
@@ -2050,8 +2051,7 @@ int random_int_secret_init(void)
return 0;
}
-static DEFINE_PER_CPU(__u32 [MD5_DIGEST_WORDS], get_random_int_hash)
- __aligned(sizeof(unsigned long));
+static DEFINE_PER_CPU(u64, get_random_int_chaining);
/*
* Get a random word for internal kernel use only. Similar to urandom but
@@ -2061,19 +2061,16 @@ static DEFINE_PER_CPU(__u32 [MD5_DIGEST_WORDS], get_random_int_hash)
*/
unsigned int get_random_int(void)
{
- __u32 *hash;
unsigned int ret;
+ u64 *chaining;
if (arch_get_random_int(&ret))
return ret;
- hash = get_cpu_var(get_random_int_hash);
-
- hash[0] += current->pid + jiffies + random_get_entropy();
- md5_transform(hash, random_int_secret);
- ret = hash[0];
- put_cpu_var(get_random_int_hash);
-
+ chaining = &get_cpu_var(get_random_int_chaining);
+ ret = *chaining = siphash_3u64(*chaining, jiffies, random_get_entropy() +
+ current->pid, random_int_secret);
+ put_cpu_var(get_random_int_chaining);
return ret;
}
EXPORT_SYMBOL(get_random_int);
@@ -2083,19 +2080,16 @@ EXPORT_SYMBOL(get_random_int);
*/
unsigned long get_random_long(void)
{
- __u32 *hash;
unsigned long ret;
+ u64 *chaining;
if (arch_get_random_long(&ret))
return ret;
- hash = get_cpu_var(get_random_int_hash);
-
- hash[0] += current->pid + jiffies + random_get_entropy();
- md5_transform(hash, random_int_secret);
- ret = *(unsigned long *)hash;
- put_cpu_var(get_random_int_hash);
-
+ chaining = &get_cpu_var(get_random_int_chaining);
+ ret = *chaining = siphash_3u64(*chaining, jiffies, random_get_entropy() +
+ current->pid, random_int_secret);
+ put_cpu_var(get_random_int_chaining);
return ret;
}
EXPORT_SYMBOL(get_random_long);
--
2.11.0
The md5_transform function is no longer used any where in the tree,
except for the crypto api's actual implementation of md5, so we can drop
the function from lib and put it as a static function of the crypto
file, where it belongs. There should be no new users of md5_transform,
anyway, since there are more modern ways of doing what it once achieved.
Signed-off-by: Jason A. Donenfeld <[email protected]>
---
crypto/md5.c | 95 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
lib/Makefile | 2 +-
lib/md5.c | 95 ------------------------------------------------------------
3 files changed, 95 insertions(+), 97 deletions(-)
delete mode 100644 lib/md5.c
diff --git a/crypto/md5.c b/crypto/md5.c
index 2355a7c25c45..f7ae1a48225b 100644
--- a/crypto/md5.c
+++ b/crypto/md5.c
@@ -21,9 +21,11 @@
#include <linux/module.h>
#include <linux/string.h>
#include <linux/types.h>
-#include <linux/cryptohash.h>
#include <asm/byteorder.h>
+#define MD5_DIGEST_WORDS 4
+#define MD5_MESSAGE_BYTES 64
+
const u8 md5_zero_message_hash[MD5_DIGEST_SIZE] = {
0xd4, 0x1d, 0x8c, 0xd9, 0x8f, 0x00, 0xb2, 0x04,
0xe9, 0x80, 0x09, 0x98, 0xec, 0xf8, 0x42, 0x7e,
@@ -47,6 +49,97 @@ static inline void cpu_to_le32_array(u32 *buf, unsigned int words)
}
}
+#define F1(x, y, z) (z ^ (x & (y ^ z)))
+#define F2(x, y, z) F1(z, x, y)
+#define F3(x, y, z) (x ^ y ^ z)
+#define F4(x, y, z) (y ^ (x | ~z))
+
+#define MD5STEP(f, w, x, y, z, in, s) \
+ (w += f(x, y, z) + in, w = (w<<s | w>>(32-s)) + x)
+
+static void md5_transform(__u32 *hash, __u32 const *in)
+{
+ u32 a, b, c, d;
+
+ a = hash[0];
+ b = hash[1];
+ c = hash[2];
+ d = hash[3];
+
+ MD5STEP(F1, a, b, c, d, in[0] + 0xd76aa478, 7);
+ MD5STEP(F1, d, a, b, c, in[1] + 0xe8c7b756, 12);
+ MD5STEP(F1, c, d, a, b, in[2] + 0x242070db, 17);
+ MD5STEP(F1, b, c, d, a, in[3] + 0xc1bdceee, 22);
+ MD5STEP(F1, a, b, c, d, in[4] + 0xf57c0faf, 7);
+ MD5STEP(F1, d, a, b, c, in[5] + 0x4787c62a, 12);
+ MD5STEP(F1, c, d, a, b, in[6] + 0xa8304613, 17);
+ MD5STEP(F1, b, c, d, a, in[7] + 0xfd469501, 22);
+ MD5STEP(F1, a, b, c, d, in[8] + 0x698098d8, 7);
+ MD5STEP(F1, d, a, b, c, in[9] + 0x8b44f7af, 12);
+ MD5STEP(F1, c, d, a, b, in[10] + 0xffff5bb1, 17);
+ MD5STEP(F1, b, c, d, a, in[11] + 0x895cd7be, 22);
+ MD5STEP(F1, a, b, c, d, in[12] + 0x6b901122, 7);
+ MD5STEP(F1, d, a, b, c, in[13] + 0xfd987193, 12);
+ MD5STEP(F1, c, d, a, b, in[14] + 0xa679438e, 17);
+ MD5STEP(F1, b, c, d, a, in[15] + 0x49b40821, 22);
+
+ MD5STEP(F2, a, b, c, d, in[1] + 0xf61e2562, 5);
+ MD5STEP(F2, d, a, b, c, in[6] + 0xc040b340, 9);
+ MD5STEP(F2, c, d, a, b, in[11] + 0x265e5a51, 14);
+ MD5STEP(F2, b, c, d, a, in[0] + 0xe9b6c7aa, 20);
+ MD5STEP(F2, a, b, c, d, in[5] + 0xd62f105d, 5);
+ MD5STEP(F2, d, a, b, c, in[10] + 0x02441453, 9);
+ MD5STEP(F2, c, d, a, b, in[15] + 0xd8a1e681, 14);
+ MD5STEP(F2, b, c, d, a, in[4] + 0xe7d3fbc8, 20);
+ MD5STEP(F2, a, b, c, d, in[9] + 0x21e1cde6, 5);
+ MD5STEP(F2, d, a, b, c, in[14] + 0xc33707d6, 9);
+ MD5STEP(F2, c, d, a, b, in[3] + 0xf4d50d87, 14);
+ MD5STEP(F2, b, c, d, a, in[8] + 0x455a14ed, 20);
+ MD5STEP(F2, a, b, c, d, in[13] + 0xa9e3e905, 5);
+ MD5STEP(F2, d, a, b, c, in[2] + 0xfcefa3f8, 9);
+ MD5STEP(F2, c, d, a, b, in[7] + 0x676f02d9, 14);
+ MD5STEP(F2, b, c, d, a, in[12] + 0x8d2a4c8a, 20);
+
+ MD5STEP(F3, a, b, c, d, in[5] + 0xfffa3942, 4);
+ MD5STEP(F3, d, a, b, c, in[8] + 0x8771f681, 11);
+ MD5STEP(F3, c, d, a, b, in[11] + 0x6d9d6122, 16);
+ MD5STEP(F3, b, c, d, a, in[14] + 0xfde5380c, 23);
+ MD5STEP(F3, a, b, c, d, in[1] + 0xa4beea44, 4);
+ MD5STEP(F3, d, a, b, c, in[4] + 0x4bdecfa9, 11);
+ MD5STEP(F3, c, d, a, b, in[7] + 0xf6bb4b60, 16);
+ MD5STEP(F3, b, c, d, a, in[10] + 0xbebfbc70, 23);
+ MD5STEP(F3, a, b, c, d, in[13] + 0x289b7ec6, 4);
+ MD5STEP(F3, d, a, b, c, in[0] + 0xeaa127fa, 11);
+ MD5STEP(F3, c, d, a, b, in[3] + 0xd4ef3085, 16);
+ MD5STEP(F3, b, c, d, a, in[6] + 0x04881d05, 23);
+ MD5STEP(F3, a, b, c, d, in[9] + 0xd9d4d039, 4);
+ MD5STEP(F3, d, a, b, c, in[12] + 0xe6db99e5, 11);
+ MD5STEP(F3, c, d, a, b, in[15] + 0x1fa27cf8, 16);
+ MD5STEP(F3, b, c, d, a, in[2] + 0xc4ac5665, 23);
+
+ MD5STEP(F4, a, b, c, d, in[0] + 0xf4292244, 6);
+ MD5STEP(F4, d, a, b, c, in[7] + 0x432aff97, 10);
+ MD5STEP(F4, c, d, a, b, in[14] + 0xab9423a7, 15);
+ MD5STEP(F4, b, c, d, a, in[5] + 0xfc93a039, 21);
+ MD5STEP(F4, a, b, c, d, in[12] + 0x655b59c3, 6);
+ MD5STEP(F4, d, a, b, c, in[3] + 0x8f0ccc92, 10);
+ MD5STEP(F4, c, d, a, b, in[10] + 0xffeff47d, 15);
+ MD5STEP(F4, b, c, d, a, in[1] + 0x85845dd1, 21);
+ MD5STEP(F4, a, b, c, d, in[8] + 0x6fa87e4f, 6);
+ MD5STEP(F4, d, a, b, c, in[15] + 0xfe2ce6e0, 10);
+ MD5STEP(F4, c, d, a, b, in[6] + 0xa3014314, 15);
+ MD5STEP(F4, b, c, d, a, in[13] + 0x4e0811a1, 21);
+ MD5STEP(F4, a, b, c, d, in[4] + 0xf7537e82, 6);
+ MD5STEP(F4, d, a, b, c, in[11] + 0xbd3af235, 10);
+ MD5STEP(F4, c, d, a, b, in[2] + 0x2ad7d2bb, 15);
+ MD5STEP(F4, b, c, d, a, in[9] + 0xeb86d391, 21);
+
+ hash[0] += a;
+ hash[1] += b;
+ hash[2] += c;
+ hash[3] += d;
+}
+
static inline void md5_transform_helper(struct md5_state *ctx)
{
le32_to_cpu_array(ctx->block, sizeof(ctx->block) / sizeof(u32));
diff --git a/lib/Makefile b/lib/Makefile
index 71d398b04a74..1079152607e0 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -19,7 +19,7 @@ KCOV_INSTRUMENT_dynamic_debug.o := n
lib-y := ctype.o string.o vsprintf.o cmdline.o \
rbtree.o radix-tree.o dump_stack.o timerqueue.o\
idr.o int_sqrt.o extable.o \
- sha1.o chacha20.o md5.o irq_regs.o argv_split.o \
+ sha1.o chacha20.o irq_regs.o argv_split.o \
flex_proportions.o ratelimit.o show_mem.o \
is_single_threaded.o plist.o decompress.o kobject_uevent.o \
earlycpio.o seq_buf.o siphash.o \
diff --git a/lib/md5.c b/lib/md5.c
deleted file mode 100644
index bb0cd01d356d..000000000000
--- a/lib/md5.c
+++ /dev/null
@@ -1,95 +0,0 @@
-#include <linux/compiler.h>
-#include <linux/export.h>
-#include <linux/cryptohash.h>
-
-#define F1(x, y, z) (z ^ (x & (y ^ z)))
-#define F2(x, y, z) F1(z, x, y)
-#define F3(x, y, z) (x ^ y ^ z)
-#define F4(x, y, z) (y ^ (x | ~z))
-
-#define MD5STEP(f, w, x, y, z, in, s) \
- (w += f(x, y, z) + in, w = (w<<s | w>>(32-s)) + x)
-
-void md5_transform(__u32 *hash, __u32 const *in)
-{
- u32 a, b, c, d;
-
- a = hash[0];
- b = hash[1];
- c = hash[2];
- d = hash[3];
-
- MD5STEP(F1, a, b, c, d, in[0] + 0xd76aa478, 7);
- MD5STEP(F1, d, a, b, c, in[1] + 0xe8c7b756, 12);
- MD5STEP(F1, c, d, a, b, in[2] + 0x242070db, 17);
- MD5STEP(F1, b, c, d, a, in[3] + 0xc1bdceee, 22);
- MD5STEP(F1, a, b, c, d, in[4] + 0xf57c0faf, 7);
- MD5STEP(F1, d, a, b, c, in[5] + 0x4787c62a, 12);
- MD5STEP(F1, c, d, a, b, in[6] + 0xa8304613, 17);
- MD5STEP(F1, b, c, d, a, in[7] + 0xfd469501, 22);
- MD5STEP(F1, a, b, c, d, in[8] + 0x698098d8, 7);
- MD5STEP(F1, d, a, b, c, in[9] + 0x8b44f7af, 12);
- MD5STEP(F1, c, d, a, b, in[10] + 0xffff5bb1, 17);
- MD5STEP(F1, b, c, d, a, in[11] + 0x895cd7be, 22);
- MD5STEP(F1, a, b, c, d, in[12] + 0x6b901122, 7);
- MD5STEP(F1, d, a, b, c, in[13] + 0xfd987193, 12);
- MD5STEP(F1, c, d, a, b, in[14] + 0xa679438e, 17);
- MD5STEP(F1, b, c, d, a, in[15] + 0x49b40821, 22);
-
- MD5STEP(F2, a, b, c, d, in[1] + 0xf61e2562, 5);
- MD5STEP(F2, d, a, b, c, in[6] + 0xc040b340, 9);
- MD5STEP(F2, c, d, a, b, in[11] + 0x265e5a51, 14);
- MD5STEP(F2, b, c, d, a, in[0] + 0xe9b6c7aa, 20);
- MD5STEP(F2, a, b, c, d, in[5] + 0xd62f105d, 5);
- MD5STEP(F2, d, a, b, c, in[10] + 0x02441453, 9);
- MD5STEP(F2, c, d, a, b, in[15] + 0xd8a1e681, 14);
- MD5STEP(F2, b, c, d, a, in[4] + 0xe7d3fbc8, 20);
- MD5STEP(F2, a, b, c, d, in[9] + 0x21e1cde6, 5);
- MD5STEP(F2, d, a, b, c, in[14] + 0xc33707d6, 9);
- MD5STEP(F2, c, d, a, b, in[3] + 0xf4d50d87, 14);
- MD5STEP(F2, b, c, d, a, in[8] + 0x455a14ed, 20);
- MD5STEP(F2, a, b, c, d, in[13] + 0xa9e3e905, 5);
- MD5STEP(F2, d, a, b, c, in[2] + 0xfcefa3f8, 9);
- MD5STEP(F2, c, d, a, b, in[7] + 0x676f02d9, 14);
- MD5STEP(F2, b, c, d, a, in[12] + 0x8d2a4c8a, 20);
-
- MD5STEP(F3, a, b, c, d, in[5] + 0xfffa3942, 4);
- MD5STEP(F3, d, a, b, c, in[8] + 0x8771f681, 11);
- MD5STEP(F3, c, d, a, b, in[11] + 0x6d9d6122, 16);
- MD5STEP(F3, b, c, d, a, in[14] + 0xfde5380c, 23);
- MD5STEP(F3, a, b, c, d, in[1] + 0xa4beea44, 4);
- MD5STEP(F3, d, a, b, c, in[4] + 0x4bdecfa9, 11);
- MD5STEP(F3, c, d, a, b, in[7] + 0xf6bb4b60, 16);
- MD5STEP(F3, b, c, d, a, in[10] + 0xbebfbc70, 23);
- MD5STEP(F3, a, b, c, d, in[13] + 0x289b7ec6, 4);
- MD5STEP(F3, d, a, b, c, in[0] + 0xeaa127fa, 11);
- MD5STEP(F3, c, d, a, b, in[3] + 0xd4ef3085, 16);
- MD5STEP(F3, b, c, d, a, in[6] + 0x04881d05, 23);
- MD5STEP(F3, a, b, c, d, in[9] + 0xd9d4d039, 4);
- MD5STEP(F3, d, a, b, c, in[12] + 0xe6db99e5, 11);
- MD5STEP(F3, c, d, a, b, in[15] + 0x1fa27cf8, 16);
- MD5STEP(F3, b, c, d, a, in[2] + 0xc4ac5665, 23);
-
- MD5STEP(F4, a, b, c, d, in[0] + 0xf4292244, 6);
- MD5STEP(F4, d, a, b, c, in[7] + 0x432aff97, 10);
- MD5STEP(F4, c, d, a, b, in[14] + 0xab9423a7, 15);
- MD5STEP(F4, b, c, d, a, in[5] + 0xfc93a039, 21);
- MD5STEP(F4, a, b, c, d, in[12] + 0x655b59c3, 6);
- MD5STEP(F4, d, a, b, c, in[3] + 0x8f0ccc92, 10);
- MD5STEP(F4, c, d, a, b, in[10] + 0xffeff47d, 15);
- MD5STEP(F4, b, c, d, a, in[1] + 0x85845dd1, 21);
- MD5STEP(F4, a, b, c, d, in[8] + 0x6fa87e4f, 6);
- MD5STEP(F4, d, a, b, c, in[15] + 0xfe2ce6e0, 10);
- MD5STEP(F4, c, d, a, b, in[6] + 0xa3014314, 15);
- MD5STEP(F4, b, c, d, a, in[13] + 0x4e0811a1, 21);
- MD5STEP(F4, a, b, c, d, in[4] + 0xf7537e82, 6);
- MD5STEP(F4, d, a, b, c, in[11] + 0xbd3af235, 10);
- MD5STEP(F4, c, d, a, b, in[2] + 0x2ad7d2bb, 15);
- MD5STEP(F4, b, c, d, a, in[9] + 0xeb86d391, 21);
-
- hash[0] += a;
- hash[1] += b;
- hash[2] += c;
- hash[3] += d;
-}
-EXPORT_SYMBOL(md5_transform);
--
2.11.0
SHA1 is slower and less secure than SipHash, and so replacing syncookie
generation with SipHash makes natural sense. Some BSDs have been doing
this for several years in fact.
Signed-off-by: Jason A. Donenfeld <[email protected]>
---
net/ipv4/syncookies.c | 20 ++++----------------
net/ipv6/syncookies.c | 37 ++++++++++++++++---------------------
2 files changed, 20 insertions(+), 37 deletions(-)
diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c
index 3e88467d70ee..03bb068f8888 100644
--- a/net/ipv4/syncookies.c
+++ b/net/ipv4/syncookies.c
@@ -13,13 +13,13 @@
#include <linux/tcp.h>
#include <linux/slab.h>
#include <linux/random.h>
-#include <linux/cryptohash.h>
+#include <linux/siphash.h>
#include <linux/kernel.h>
#include <linux/export.h>
#include <net/tcp.h>
#include <net/route.h>
-static u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS] __read_mostly;
+static siphash_key_t syncookie_secret[2] __read_mostly;
#define COOKIEBITS 24 /* Upper bits store count */
#define COOKIEMASK (((__u32)1 << COOKIEBITS) - 1)
@@ -48,24 +48,12 @@ static u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS] __read_mostly;
#define TSBITS 6
#define TSMASK (((__u32)1 << TSBITS) - 1)
-static DEFINE_PER_CPU(__u32 [16 + 5 + SHA_WORKSPACE_WORDS], ipv4_cookie_scratch);
-
static u32 cookie_hash(__be32 saddr, __be32 daddr, __be16 sport, __be16 dport,
u32 count, int c)
{
- __u32 *tmp;
-
net_get_random_once(syncookie_secret, sizeof(syncookie_secret));
-
- tmp = this_cpu_ptr(ipv4_cookie_scratch);
- memcpy(tmp + 4, syncookie_secret[c], sizeof(syncookie_secret[c]));
- tmp[0] = (__force u32)saddr;
- tmp[1] = (__force u32)daddr;
- tmp[2] = ((__force u32)sport << 16) + (__force u32)dport;
- tmp[3] = count;
- sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
-
- return tmp[17];
+ return siphash_4u32(saddr, daddr, (u32)sport << 16 | dport, count,
+ syncookie_secret[c]);
}
diff --git a/net/ipv6/syncookies.c b/net/ipv6/syncookies.c
index a4d49760bf43..04d19e89a3e0 100644
--- a/net/ipv6/syncookies.c
+++ b/net/ipv6/syncookies.c
@@ -16,7 +16,7 @@
#include <linux/tcp.h>
#include <linux/random.h>
-#include <linux/cryptohash.h>
+#include <linux/siphash.h>
#include <linux/kernel.h>
#include <net/ipv6.h>
#include <net/tcp.h>
@@ -24,7 +24,7 @@
#define COOKIEBITS 24 /* Upper bits store count */
#define COOKIEMASK (((__u32)1 << COOKIEBITS) - 1)
-static u32 syncookie6_secret[2][16-4+SHA_DIGEST_WORDS] __read_mostly;
+static siphash_key_t syncookie6_secret[2] __read_mostly;
/* RFC 2460, Section 8.3:
* [ipv6 tcp] MSS must be computed as the maximum packet size minus 60 [..]
@@ -41,30 +41,25 @@ static __u16 const msstab[] = {
9000 - 60,
};
-static DEFINE_PER_CPU(__u32 [16 + 5 + SHA_WORKSPACE_WORDS], ipv6_cookie_scratch);
-
static u32 cookie_hash(const struct in6_addr *saddr, const struct in6_addr *daddr,
__be16 sport, __be16 dport, u32 count, int c)
{
- __u32 *tmp;
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ u32 count;
+ u16 sport;
+ u16 dport;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *saddr,
+ .daddr = *daddr,
+ .count = count,
+ .sport = sport,
+ .dport = dport
+ };
net_get_random_once(syncookie6_secret, sizeof(syncookie6_secret));
-
- tmp = this_cpu_ptr(ipv6_cookie_scratch);
-
- /*
- * we have 320 bits of information to hash, copy in the remaining
- * 192 bits required for sha_transform, from the syncookie6_secret
- * and overwrite the digest with the secret
- */
- memcpy(tmp + 10, syncookie6_secret[c], 44);
- memcpy(tmp, saddr, 16);
- memcpy(tmp + 4, daddr, 16);
- tmp[8] = ((__force u32)sport << 16) + (__force u32)dport;
- tmp[9] = count;
- sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
-
- return tmp[17];
+ return siphash(&combined, sizeof(combined), syncookie6_secret[c]);
}
static __u32 secure_tcp_syn_cookie(const struct in6_addr *saddr,
--
2.11.0
Jean-Philippe Aumasson wrote:
> If a halved version of SipHash can bring significant performance boost
> (with 32b words instead of 64b words) with an acceptable security level
> (64-bit enough?) then we may design such a version.
It would be fairly significant, a 2x speed benefit on a lot of 32-bit
machines.
First is the fact that a 64-bit SipHash round on a generic 32-bit machine
requires not twice as many instructions, but more than three.
Consider the core SipHash quarter-round operation:
a += b;
b = rotate_left(b, k)
b ^= a
The add and xor are equivalent between 32- and 64-bit rounds; twice the
instructions do twice the work. (There's a dependency via the carry
bit between the two halves of the add, but it ends up not being on the
critical path even in a superscalar implementation.)
The problem is the rotates. Although some particularly nice code is
possible on 32-bit ARM due to its support for shift-and-xor operations,
on a generic 32-bit CPU the rotate grows to 6 instructions with a 2-cycle
dependency chain (more in practice because barrel shifters are large and
even quad-issue CPUs can't do 4 shifts per cycle):
temp_lo = b_lo >> (32-k)
temp_hi = b_hi >> (32-k)
b_lo <<= k
b_hi <<= k
b_lo ^= temp_hi
b_hi ^= temp_lo
The resultant instruction counts and (assuming wide issue)
latencies are:
64-bit SipHash "Half" SipHash
Inst. Latency Inst. Latency
10 3 3 2 Quarter round
40 6 12 4 Full round
80 12 24 8 Two rounds
82 13 26 9 Mix in one word
82 13 52 18 Mix in 64 bits
166 26 61 18 Four round finalization + final XOR
248 39 113 36 Hash 64 bits
330 52 165 54 Hash 128 bits
412 65 217 72 Hash 192 bits
While the ideal latencies are actually better for the 64-bit algorithm,
that requires an unrealistic 6+-wide superscalar implementation that's
more than twice as wide as the 64-bit code requires (which is already
optimized for quad-issue). For a 1- or 2-wide processor, the instruction
counts dominate, and not only does the 64-bit algorithm take 60% more
time to mix in the same number of bytes, but the finalization rounds
bring the ratio to 2:1 for small inputs.
(And I haven't included the possible savings if the input size is an odd
number of 32-bit words, such as networking applications which include
the source/dest port numbers.)
Notes on particular processors:
- x86 can do a 64-bit rotate in 3 instructions and 2 cycles using
the SHLD/SHRD instructions instead:
movl %b_hi, %temp
shldl $k, %b_lo, %b_hi
shldl $k, %temp, %b_lo
... but as I mentioned the problem is registers. SipHash needs 8 32-bit
words plus at least one temporary, and 32-bit x86 has only 7 available.
(And compilers can rarely manage to keep more than 6 of them busy.)
- 64-bit SipHash is particularly efficient on 32-bit ARM due to its
support for shift-and-op instructions. The 64-bit shift and following
xor can be done in 4 instructions. So the only benefit is from the
reduced finalization.
- Double-width adds cost a little more on CPUs like MIPS and RISC-V without
condition codes.
- Certain particularly crappy uClinux processors with slow shifts
(68000, anyone?) really suffer from extra shifts.
One *weakly* requested feature: It might simplify some programming
interfaces if we could use the same key for multiple hash tables with a
1-word "tweak" (e.g. pointer to the hash table, so it could be assumed
non-zero if that helped) to make distinct functions. That would let us
more safely use a global key for multiple small hash tables without the
need to add code to generate and store key material for each place that
an unkeyed hash is replaced.
Here's a tentative HalfSipHash:
https://github.com/veorq/SipHash/blob/halfsiphash/halfsiphash.c
Haven't computed the cycle count nor measured its speed.
On Fri, Dec 16, 2016 at 4:46 AM George Spelvin <[email protected]>
wrote:
> Jean-Philippe Aumasson wrote:
> > If a halved version of SipHash can bring significant performance boost
> > (with 32b words instead of 64b words) with an acceptable security level
> > (64-bit enough?) then we may design such a version.
>
> It would be fairly significant, a 2x speed benefit on a lot of 32-bit
> machines.
>
> First is the fact that a 64-bit SipHash round on a generic 32-bit machine
> requires not twice as many instructions, but more than three.
>
> Consider the core SipHash quarter-round operation:
> a += b;
> b = rotate_left(b, k)
> b ^= a
>
> The add and xor are equivalent between 32- and 64-bit rounds; twice the
> instructions do twice the work. (There's a dependency via the carry
> bit between the two halves of the add, but it ends up not being on the
> critical path even in a superscalar implementation.)
>
> The problem is the rotates. Although some particularly nice code is
> possible on 32-bit ARM due to its support for shift-and-xor operations,
> on a generic 32-bit CPU the rotate grows to 6 instructions with a 2-cycle
> dependency chain (more in practice because barrel shifters are large and
> even quad-issue CPUs can't do 4 shifts per cycle):
>
> temp_lo = b_lo >> (32-k)
> temp_hi = b_hi >> (32-k)
> b_lo <<= k
> b_hi <<= k
> b_lo ^= temp_hi
> b_hi ^= temp_lo
>
> The resultant instruction counts and (assuming wide issue)
> latencies are:
>
> 64-bit SipHash "Half" SipHash
> Inst. Latency Inst. Latency
> 10 3 3 2 Quarter round
> 40 6 12 4 Full round
> 80 12 24 8 Two rounds
> 82 13 26 9 Mix in one word
> 82 13 52 18 Mix in 64 bits
> 166 26 61 18 Four round finalization + final XOR
> 248 39 113 36 Hash 64 bits
> 330 52 165 54 Hash 128 bits
> 412 65 217 72 Hash 192 bits
>
> While the ideal latencies are actually better for the 64-bit algorithm,
> that requires an unrealistic 6+-wide superscalar implementation that's
> more than twice as wide as the 64-bit code requires (which is already
> optimized for quad-issue). For a 1- or 2-wide processor, the instruction
> counts dominate, and not only does the 64-bit algorithm take 60% more
> time to mix in the same number of bytes, but the finalization rounds
> bring the ratio to 2:1 for small inputs.
>
> (And I haven't included the possible savings if the input size is an odd
> number of 32-bit words, such as networking applications which include
> the source/dest port numbers.)
>
>
> Notes on particular processors:
> - x86 can do a 64-bit rotate in 3 instructions and 2 cycles using
> the SHLD/SHRD instructions instead:
> movl %b_hi, %temp
> shldl $k, %b_lo, %b_hi
> shldl $k, %temp, %b_lo
> ... but as I mentioned the problem is registers. SipHash needs 8 32-bit
> words plus at least one temporary, and 32-bit x86 has only 7 available.
> (And compilers can rarely manage to keep more than 6 of them busy.)
> - 64-bit SipHash is particularly efficient on 32-bit ARM due to its
> support for shift-and-op instructions. The 64-bit shift and following
> xor can be done in 4 instructions. So the only benefit is from the
> reduced finalization.
> - Double-width adds cost a little more on CPUs like MIPS and RISC-V without
> condition codes.
> - Certain particularly crappy uClinux processors with slow shifts
> (68000, anyone?) really suffer from extra shifts.
>
> One *weakly* requested feature: It might simplify some programming
> interfaces if we could use the same key for multiple hash tables with a
> 1-word "tweak" (e.g. pointer to the hash table, so it could be assumed
> non-zero if that helped) to make distinct functions. That would let us
> more safely use a global key for multiple small hash tables without the
> need to add code to generate and store key material for each place that
> an unkeyed hash is replaced.
>
From: Jason A. Donenfeld
> Sent: 15 December 2016 20:30
> This gives a clear speed and security improvement. Siphash is both
> faster and is more solid crypto than the aging MD5.
>
> Rather than manually filling MD5 buffers, for IPv6, we simply create
> a layout by a simple anonymous struct, for which gcc generates
> rather efficient code. For IPv4, we pass the values directly to the
> short input convenience functions.
...
> diff --git a/net/core/secure_seq.c b/net/core/secure_seq.c
> index 88a8e429fc3e..c80583bf3213 100644
...
> + const struct {
> + struct in6_addr saddr;
> + struct in6_addr daddr;
> + __be16 sport;
> + __be16 dport;
> + u32 padding;
> + } __aligned(SIPHASH_ALIGNMENT) combined = {
> + .saddr = *(struct in6_addr *)saddr,
> + .daddr = *(struct in6_addr *)daddr,
> + .sport = sport,
> + .dport = dport
> + };
I think you should explicitly initialise the 'padding'.
It can do no harm and makes it obvious that it is necessary.
You are still putting over-aligned data on stack.
You only need to align it to the alignment of u64 (not the size of u64).
If an on-stack item has a stronger alignment requirement than the stack
the gcc has to generate two stack frames for the function.
If you assign to each field (instead of using initialisers) then you
can get the alignment by making the first member an anonymous union
of in6_addr and u64.
Oh - and wait a bit longer between revisions.
David
From: Jason A. Donenfeld
> Sent: 15 December 2016 20:30
> These restore parity with the jhash interface by providing high
> performance helpers for common input sizes.
...
> +#define PREAMBLE(len) \
> + u64 v0 = 0x736f6d6570736575ULL; \
> + u64 v1 = 0x646f72616e646f6dULL; \
> + u64 v2 = 0x6c7967656e657261ULL; \
> + u64 v3 = 0x7465646279746573ULL; \
> + u64 b = ((u64)len) << 56; \
> + v3 ^= key[1]; \
> + v2 ^= key[0]; \
> + v1 ^= key[1]; \
> + v0 ^= key[0];
Isn't that equivalent to:
v0 = key[0];
v1 = key[1];
v2 = key[0] ^ (0x736f6d6570736575ULL ^ 0x646f72616e646f6dULL);
v3 = key[1] ^ (0x646f72616e646f6dULL ^ 0x7465646279746573ULL);
Those constants also look like ASCII strings.
What cryptographic analysis has been done on the values?
David
Hey JP,
On Fri, Dec 16, 2016 at 9:08 AM, Jean-Philippe Aumasson
<[email protected]> wrote:
> Here's a tentative HalfSipHash:
> https://github.com/veorq/SipHash/blob/halfsiphash/halfsiphash.c
>
> Haven't computed the cycle count nor measured its speed.
This is incredible. Really. Wow!
I'll integrate this into my patchset and will write up some
documentation about when one should be used over the other.
Thanks again. Quite exciting.
Jason
It needs some basic security review, which I'll try do next week (check for
security margin, optimality of rotation counts, etc.). But after a lot of
experience with this kind of construction (BLAKE, SipHash, NORX), I'm
confident it will be safe as it is.
On Fri, Dec 16, 2016 at 1:44 PM Jason A. Donenfeld <[email protected]> wrote:
> Hey JP,
>
> On Fri, Dec 16, 2016 at 9:08 AM, Jean-Philippe Aumasson
> <[email protected]> wrote:
> > Here's a tentative HalfSipHash:
> > https://github.com/veorq/SipHash/blob/halfsiphash/halfsiphash.c
> >
> > Haven't computed the cycle count nor measured its speed.
>
> This is incredible. Really. Wow!
>
> I'll integrate this into my patchset and will write up some
> documentation about when one should be used over the other.
>
> Thanks again. Quite exciting.
>
> Jason
>
Jason A. Donenfeld wrote:
> Isn't that equivalent to:
> v0 = key[0];
> v1 = key[1];
> v2 = key[0] ^ (0x736f6d6570736575ULL ^ 0x646f72616e646f6dULL);
> v3 = key[1] ^ (0x646f72616e646f6dULL ^ 0x7465646279746573ULL);
(Pre-XORing key[] with the first two constants which, if the constants
are random in the first place, can be a no-op.) Other than the typo
in the v2 line, yes. If they key is non-public, then you can xor an
arbitrary constant in to both halves to slightly speed up the startup.
(Nits: There's a typo in the v2 line, you don't need to parenthesize
associative operators like xor, and the "ull" suffix is redundant here.)
> Those constants also look like ASCII strings.
They are. The ASCII is "somepseudorandomlygeneratedbytes".
> What cryptographic analysis has been done on the values?
They're "nothing up my sleeve numbers".
They're arbitrary numbers, and almost any other values would do exactly
as well. The main properties are:
1) They're different (particulatly v0 != v2 and v1 != v3), and
2) Neither they, nor their xor, is rotationally symmetric like 0x55555555.
(Because SipHash is mostly rotationally symmetric, broken only by the
interruption of the carry chain at the msbit, it helps slightly
to break this up at the beginning.)
Those exact values only matter for portability. If you don't need anyone
else to be able to compute matching outputs, then you could use any other
convenient constants (like the MD5 round constants).
Hi JP & George,
My function names:
- SipHash -> siphash
- HalfSipHash -> hsiphash
It appears that hsiphash can produce either 32-bit output or 64-bit
output, with the output length parameter as part of the hash algorithm
in there. When I code this for my kernel patchset, I very likely will
only implement one output length size. Right now I'm leaning toward
32-bit. Questions:
- Is this a reasonable choice?
- When hsiphash is desired due to its faster speed, are there any
circumstances in which producing a 64-bit output would actually be
useful? Namely, are there any hashtables that could benefit from a
64-bit functions?
- Are there reasons why hsiphash with 64-bit output would be
reasonable? Or will we be fine sticking with 32-bit output only?
With both hsiphash and siphash, the division of usage will probably become:
- Use 64-bit output 128-bit key siphash for keyed RNG-like things,
such as syncookies and sequence numbers
- Use 64-bit output 128-bit key siphash for hashtables that must
absolutely be secure to an extremely high bandwidth attacker, such as
userspace directly DoSing a kernel hashtable
- Use 32-bit output 64-bit key hsiphash for quick hashtable functions
that still must be secure but do not require as large of a security
margin
Sound good?
Jason
Hi David,
On Fri, Dec 16, 2016 at 10:59 AM, David Laight <[email protected]> wrote:
> You are still putting over-aligned data on stack.
> You only need to align it to the alignment of u64 (not the size of u64).
> If an on-stack item has a stronger alignment requirement than the stack
> the gcc has to generate two stack frames for the function.
Yesterday, folks were saying that sometimes 32-bit platforms need
8-byte alignment for certain 64-bit operations, so I shouldn't fall
back to 4-byte alignment there. But actually, looking at this more
closely, I can just make SIPHASH_ALIGNMENT == __alignof__(u64), which
will take care of all possible concerns, since gcc knows best which
platforms need what alignment. Thanks for making this clear to me with
"the alignment of u64 (not the size of u64)".
> Oh - and wait a bit longer between revisions.
Okay. We can be turtles.
Jason
From: George Spelvin
> Sent: 15 December 2016 23:29
> > If a halved version of SipHash can bring significant performance boost
> > (with 32b words instead of 64b words) with an acceptable security level
> > (64-bit enough?) then we may design such a version.
>
> I was thinking if the key could be pushed to 80 bits, that would be nice,
> but honestly 64 bits is fine. This is DoS protection, and while it's
> possible to brute-force a 64 bit secret, there are more effective (DDoS)
> attacks possible for the same cost.
A 32bit hash would also remove all the issues about the alignment
of IP addresses (etc) on 64bit systems.
> (I'd suggest a name of "HalfSipHash" to convey the reduced security
> effectively.)
>
> > Regarding output size, are 64 bits sufficient?
>
> As a replacement for jhash, 32 bits are sufficient. It's for
> indexing an in-memory hash table on a 32-bit machine.
It is also worth remembering that if the intent is to generate
a hash table index (not a unique fingerprint) you will always
get collisions on the final value.
Randomness could still give overlong hash chains - which might
still need rehashing with a different key.
David
Hi David,
On Fri, Dec 16, 2016 at 6:06 PM, David Laight <[email protected]> wrote:
> A 32bit hash would also remove all the issues about the alignment
> of IP addresses (etc) on 64bit systems.
The current replacements of md5_transform with siphash in the v6 patch
series will continue to use the original siphash, since the 128-bit
key is rather important for these kinds of secrets. Additionally,
64-bit siphash is already faster than the md5_transform that it
replaces. So the alignment concerns (now, non-issues; problems have
been solved, I believe) still remain.
Jason
> It appears that hsiphash can produce either 32-bit output or 64-bit
> output, with the output length parameter as part of the hash algorithm
> in there. When I code this for my kernel patchset, I very likely will
> only implement one output length size. Right now I'm leaning toward
> 32-bit.
A 128-bit output option was added to SipHash after the initial publication;
this is just the equivalent in 32-bit.
> - Is this a reasonable choice?
Yes.
> - Are there reasons why hsiphash with 64-bit output would be
> reasonable? Or will we be fine sticking with 32-bit output only?
Personally, I'd put in a comment saying that "there's a 64-bit output
variant that's not implemented" and punt until someone find a need.
> With both hsiphash and siphash, the division of usage will probably become:
> - Use 64-bit output 128-bit key siphash for keyed RNG-like things,
> such as syncookies and sequence numbers
> - Use 64-bit output 128-bit key siphash for hashtables that must
> absolutely be secure to an extremely high bandwidth attacker, such as
> userspace directly DoSing a kernel hashtable
> - Use 32-bit output 64-bit key hsiphash for quick hashtable functions
> that still must be secure but do not require as large of a security
> margin.
On a 64-bit machine, 64-bit SipHash is *always* faster than 32-bit, and
should be used always. Don't even compile the 32-bit code, to prevent
anyone accidentally using it, and make hsiphash an alias for siphash.
On a 32-bit machine, it's a much trickier case. I'd be tempted to
use the 32-bit code always, but it needs examination.
Fortunately, the cost of brute-forcing hash functions can be fairly
exactly quantified, thanks to bitcoin miners. It currently takes 2^70
hashes to create one bitcoin block, worth 25 bitcoins ($19,500). Thus,
2^63 hashes cost $152.
Now, there are two factors that must be considered:
- That's a very very "wholesale" rate. That's assuming you're doing
large numbers of these and can put in the up-front effort designing
silicon ASICs to do the attack.
- That's for a more difficult hash (double sha-256) than SipHash.
That's a constant fator, but a pretty significant one. If the wholesale
assumption holds, that might bring the cost down another 6 or 7 bits,
to $1-2 per break.
If you're not the NSA and limited to general-purpose silicon, let's
assume a state of the art GPU (Radeon HD 7970; AMD GPUs seem do to better
than nVidia). The bitcoin mining rate for those is about 700M/second,
29.4 bits. So 63 bits is 152502 GPU-days, divided by some factor
to account for SipHash's high speed compared to two rounds of SHA-2.
Call it 1000 GPU-days.
It's very doable, but also very non-trivial. The question is, wouldn't
it be cheaper and easier just to do a brute-force flooding DDoS?
(This is why I wish the key size could be tweaked up to 80 bits.
That would take all these numbers out of the reasonable range.)
Let me consider your second example above, "secure against local users".
I should dig through your patchset and find the details, but what exactly
are the consequences of such an attack? Hasn't a local user already
got much better ways to DoS the system?
The thing to remember is that we're worried only about the combination
of a *new* Linux kernel (new build or under active maintenance) and a
32-bit host. You'd be hard-pressed to find a *single* machine fitting
that description which is hosting multiple users or VMs and is not 64-bit.
These days, 32-bit CPUs are for embedded applications: network appliances,
TVs, etc. That means basically single-user. Even phones are 64 bit.
Is this really a threat that needs to be defended against?
For your first case, network applications, the additional security
is definitely attractive. Syncookies are only a DoS, but sequence
numbers are a real security issue; they can let you inject data into a
TCP connection.
Hash tables are much harder to attack. The information you get back from
timing probes is statistical, and thus testing a key is more expensive.
With sequence numbers, large amounts (32 bits) the hash output is
directly observable.
I wish we could get away with 64-bit security, but given that the
modern internet involves attacks from NSA/Spetssvyaz/3PLA, I agree
it's just not enough.
Hi George,
On Fri, Dec 16, 2016 at 6:36 PM, George Spelvin
<[email protected]> wrote:
> A 128-bit output option was added to SipHash after the initial publication;
> this is just the equivalent in 32-bit.
> Personally, I'd put in a comment saying that "there's a 64-bit output
> variant that's not implemented" and punt until someone find a need.
That's a good way to think about it. Okay, I'll do precisely that.
> On a 64-bit machine, 64-bit SipHash is *always* faster than 32-bit, and
> should be used always. Don't even compile the 32-bit code, to prevent
> anyone accidentally using it, and make hsiphash an alias for siphash.
Fascinating! Okay. So I'll alias hsiphash to siphash on 64-bit then. I
like this arrangement.
> Fortunately, the cost of brute-forcing hash functions can be fairly
> exactly quantified, thanks to bitcoin miners. It currently takes 2^70
> hashes to create one bitcoin block, worth 25 bitcoins ($19,500). Thus,
> 2^63 hashes cost $152.
>
> Now, there are two factors that must be considered:
> - That's a very very "wholesale" rate. That's assuming you're doing
> large numbers of these and can put in the up-front effort designing
> silicon ASICs to do the attack.
> - That's for a more difficult hash (double sha-256) than SipHash.
> That's a constant fator, but a pretty significant one. If the wholesale
> assumption holds, that might bring the cost down another 6 or 7 bits,
> to $1-2 per break.
>
> If you're not the NSA and limited to general-purpose silicon, let's
> assume a state of the art GPU (Radeon HD 7970; AMD GPUs seem do to better
> than nVidia). The bitcoin mining rate for those is about 700M/second,
> 29.4 bits. So 63 bits is 152502 GPU-days, divided by some factor
> to account for SipHash's high speed compared to two rounds of SHA-2.
> Call it 1000 GPU-days.
>
> It's very doable, but also very non-trivial. The question is, wouldn't
> it be cheaper and easier just to do a brute-force flooding DDoS?
>
> (This is why I wish the key size could be tweaked up to 80 bits.
> That would take all these numbers out of the reasonable range.)
That's a nice analysis. Might one conclude from that that hsiphash is
not useful for our purposes? Or does it still remain useful for
network facing code?
> Let me consider your second example above, "secure against local users".
> I should dig through your patchset and find the details, but what exactly
> are the consequences of such an attack? Hasn't a local user already
> got much better ways to DoS the system?
For example, an unpriv'd user putting lots of entries in one hash
bucket for a shared resource that's used by root, like filesystems or
other lookup tables. If he can cause root to use more of root's cpu
schedule budget than otherwise in a directed way, then that's a bad
DoS.
> The thing to remember is that we're worried only about the combination
> of a *new* Linux kernel (new build or under active maintenance) and a
> 32-bit host. You'd be hard-pressed to find a *single* machine fitting
> that description which is hosting multiple users or VMs and is not 64-bit.
>
> These days, 32-bit CPUs are for embedded applications: network appliances,
> TVs, etc. That means basically single-user. Even phones are 64 bit.
> Is this really a threat that needs to be defended against?
I interpret this to indicate all the more reason to alias hsiphash to
siphash on 64-bit, and then the problem space collapses in a clear
way.
> For your first case, network applications, the additional security
> is definitely attractive. Syncookies are only a DoS, but sequence
> numbers are a real security issue; they can let you inject data into a
> TCP connection.
> With sequence numbers, large amounts (32 bits) the hash output is
> directly observable.
Right. Hence the need for always using full siphash and not hsiphash
for sequence numbers, per my earlier email to David.
>
> I wish we could get away with 64-bit security, but given that the
> modern internet involves attacks from NSA/Spetssvyaz/3PLA, I agree
> it's just not enough.
I take this comment to be relavent for the sequence number case.
For hashtables and hashtable flooding, is it still your opinion that
we will benefit from hsiphash? Or is this final conclusion a rejection
of hsiphash for that too? We're talking about two different use cases,
and your email kind of interleaved both into your analysis, so I'm not
certain so to precisely what your conclusion is for each use case. Can
you clear up the ambiguity?
Jason
On Fri, Dec 16, 2016 at 4:39 AM, Jason A. Donenfeld <[email protected]> wrote:
> Hey JP,
>
> On Fri, Dec 16, 2016 at 9:08 AM, Jean-Philippe Aumasson
> <[email protected]> wrote:
>> Here's a tentative HalfSipHash:
>> https://github.com/veorq/SipHash/blob/halfsiphash/halfsiphash.c
>>
>> Haven't computed the cycle count nor measured its speed.
>
Tested this. Distribution and avalanche effect are still good. Speed
wise I see about a 33% improvement over siphash (20 nsecs/op versus 32
nsecs). That's about 3x of jhash speed (7 nsecs). So that might closer
to a more palatable replacement for jhash. Do we lose any security
advantages with halfsiphash?
Tom
> This is incredible. Really. Wow!
>
> I'll integrate this into my patchset and will write up some
> documentation about when one should be used over the other.
>
> Thanks again. Quite exciting.
>
> Jason
>> On a 64-bit machine, 64-bit SipHash is *always* faster than 32-bit, and
>> should be used always. Don't even compile the 32-bit code, to prevent
>> anyone accidentally using it, and make hsiphash an alias for siphash.
> Fascinating! Okay. So I'll alias hsiphash to siphash on 64-bit then. I
> like this arrangement.
This is a basic assumption I make in the security analysis below:
on most machines, it's 128-bit-key SipHash everywhere and we can
consider security solved.
Our analysis *only* has to consider 32-bit machines. My big concern
is home routers, with IoT appliances coming second. The routers have
severe hardware cost constraints (so limited CPU power), but see a lot
of traffic and need to process (NAT) it.
> That's a nice analysis. Might one conclude from that that hsiphash is
> not useful for our purposes? Or does it still remain useful for
> network facing code?
I think for attacks where the threat is a DoS, it's usable. The point
is you only have to raise the cost to equal that of a packet flood.
(Just like in electronic warfare, the best you can possibly do is force
the enemy to use broadband jamming.)
Hash collision attacks just aren't that powerful. The original PoC
was against an application that implemented a hard limit on hash chain
length as a DoS defense, which the attack then exploited to turn it into
a hard DoS.
>> Let me consider your second example above, "secure against local users".
>> I should dig through your patchset and find the details, but what exactly
>> are the consequences of such an attack? Hasn't a local user already
>> got much better ways to DoS the system?
> For example, an unpriv'd user putting lots of entries in one hash
> bucket for a shared resource that's used by root, like filesystems or
> other lookup tables. If he can cause root to use more of root's cpu
> schedule budget than otherwise in a directed way, then that's a bad
> DoS.
This issue was recently discussed when we redesigned the dcache hash.
Even a successful attack doesn't slow things down all *that* much.
Before you overkill every hash table in the kernel, think about whether
it's a bigger problem than the dcache. (Hint: it's probably not.)
There's no point armor-plating the side door when the front door was
just upgraded from screen to wood.
>> These days, 32-bit CPUs are for embedded applications: network appliances,
>> TVs, etc. That means basically single-user. Even phones are 64 bit.
>> Is this really a threat that needs to be defended against?
> I interpret this to indicate all the more reason to alias hsiphash to
> siphash on 64-bit, and then the problem space collapses in a clear
> way.
Yes, exactly.
> Right. Hence the need for always using full siphash and not hsiphash
> for sequence numbers, per my earlier email to David.
>
>> I wish we could get away with 64-bit security, but given that the
>> modern internet involves attacks from NSA/Spetssvyaz/3PLA, I agree
>> it's just not enough.
>
> I take this comment to be relavent for the sequence number case.
Yes.
> For hashtables and hashtable flooding, is it still your opinion that
> we will benefit from hsiphash? Or is this final conclusion a rejection
> of hsiphash for that too? We're talking about two different use cases,
> and your email kind of interleaved both into your analysis, so I'm not
> certain so to precisely what your conclusion is for each use case. Can
> you clear up the ambiguity?
My (speaking enerally; I should walk through every hash table you've
converted) opinion is that:
- Hash tables, even network-facing ones, can all use hsiphash as long
as an attacker can only see collisions, i.e. ((H(x) ^ H(y)) & bits) ==
0, and the consequences of a successful attack is only more collisions
(timing). While the attack is only 2x the cost (two hashes rather than
one to test a key), the knowledge of the collision is statistical,
especially for network attackers, which raises the cost of guessing
beyond an even more brute-force attack.
- When the hash value directly visible (e.g. included in a network
packet), full SipHash should be the default.
- Syncookies *could* use hsiphash, especially as there are
two keys in there. Not sure if we need the performance.
- For TCP ISNs, I'd prefer to use full SipHash. I know this is
a very hot path, and if that's a performance bottleneck,
we can work harder on it.
In particular, TCP ISNs *used* to rotate the key periodically,
limiting the time available to an attacker to perform an
attack before the secret goes stale and is useless. commit
6e5714eaf77d79ae1c8b47e3e040ff5411b717ec upgraded to md5 and dropped
the key rotation.
If 2x hsiphash is faster than siphash, we could use a double-hashing
system like syncookies. One 32-bit hash with a permanent key, summed
with a k-bit counter and a (32-k)-bit hash, where the key is rotated
(and the counter incremented) periodically.
The requirement is that the increment rate of the counter hash doesn't
shorten the sequence number wraparound too much. The old code used an
8-bit counter and 24-bit hash, with the counter bumped every 5 minutes.
Current code uses a 64 ns tick for the ISN, so it counts 2^24 per second.
(32 bits wraps every 4.6 minutes.) A 4-bit counter and 28-bit hash
(or even 3+29) would work as long as the key is regenerated no more
than once per minute. (Just using the 4.6-minute ISN wrap time is the
obvious simple implementation.)
(Of course, I defer to DaveM's judgement on all network-related issues.)
Hi JP,
On Fri, Dec 16, 2016 at 2:22 PM, Jean-Philippe Aumasson
<[email protected]> wrote:
> It needs some basic security review, which I'll try do next week (check for
> security margin, optimality of rotation counts, etc.). But after a lot of
> experience with this kind of construction (BLAKE, SipHash, NORX), I'm
> confident it will be safe as it is.
I've implemented it in my siphash kernel branch:
https://git.zx2c4.com/linux-dev/log/?h=siphash
It's the commit that has "HalfSipHash" in the log message. As the
structure is nearly identical to SipHash, there wasn't a lot to
change, and so the same implementation strategy exists for each.
When you've finished your security review and feel good about it, some
test vectors using the same formula (key={0x03020100, 07060504},
input={0x0, 0x1, 0x2, 0x3...}, output=test_vectors) would be nice for
verification.
Jason
Tom Herbert wrote:
> Tested this. Distribution and avalanche effect are still good. Speed
> wise I see about a 33% improvement over siphash (20 nsecs/op versus 32
> nsecs). That's about 3x of jhash speed (7 nsecs). So that might closer
> to a more palatable replacement for jhash. Do we lose any security
> advantages with halfsiphash?
What are you testing on? And what input size? And does "33% improvement"
mean 4/3 the rate and 3/4 the time? Or 2/3 the time and 3/2 the rate?
These are very odd results. On a 64-bit machine, SipHash should be the
same speed per round, and faster because it hashes more data per round.
(Unless you're hitting some unexpected cache/decode effect due to REX
prefixes.)
On a 32-bit machine (other than ARM, where your results might make sense,
or maybe if you're hashing large amounts of data), the difference should
be larger.
And yes, there is a *significant* security loss. SipHash is 128 bits
("don't worry about it"). hsiphash is 64 bits, which is known breakable
("worry about it"), so we have to do a careful analysis of the cost of
a successful attack.
As mentioned in the e-mails that just flew by, hsiphash is intended
*only* for 32-bit machines which bog down on full SipHash. On all 64-bit
machines, it will be implemented as an alias for SipHash and the security
concerns will Just Go Away.
The place where hsiphash is expected to make a big difference is 32-bit
x86. If you only see 33% difference with "gcc -m32", I'm going to be
very confused.
On Fri, Dec 16, 2016 at 9:41 PM, George Spelvin
<[email protected]> wrote:
> What are you testing on? And what input size? And does "33% improvement"
> mean 4/3 the rate and 3/4 the time? Or 2/3 the time and 3/2 the rate?
How that I've published my hsiphash implementation to my tree, it
should be possible to conduct the tests back to back with nearly
identical implementation strategies, to remove a potential source of
error.
On Fri, Dec 16, 2016 at 03:17:39PM -0500, George Spelvin wrote:
> > That's a nice analysis. Might one conclude from that that hsiphash is
> > not useful for our purposes? Or does it still remain useful for
> > network facing code?
>
> I think for attacks where the threat is a DoS, it's usable. The point
> is you only have to raise the cost to equal that of a packet flood.
> (Just like in electronic warfare, the best you can possibly do is force
> the enemy to use broadband jamming.)
>
> Hash collision attacks just aren't that powerful. The original PoC
> was against an application that implemented a hard limit on hash chain
> length as a DoS defense, which the attack then exploited to turn it into
> a hard DoS.
What should we do with get_random_int() and get_random_long()? In
some cases it's being used in performance sensitive areas, and where
anti-DoS protection might be enough. In others, maybe not so much.
If we rekeyed the secret used by get_random_int() and
get_random_long() frequently (say, every minute or every 5 minutes),
would that be sufficient for current and future users of these
interfaces?
- Ted
P.S. I'll note that my performance figures when testing changes to
get_random_int() were done on a 32-bit x86; Jason, I'm guessing your
figures were using a 64-bit x86 system?. I haven't tried 32-bit ARM
or smaller CPU's (e.g., mips, et. al.) that might be more likely to be
used on IoT devices, but I'm worried about those too, of course.
On Fri, 2016-12-16 at 11:47 -0800, Tom Herbert wrote:
>
> That's about 3x of jhash speed (7 nsecs). So that might closer
> to a more palatable replacement for jhash. Do we lose any security
> advantages with halfsiphash?
Have you tested a lower round SipHash? Probably best to stick with the
usual construction for non-DoS mitigation, but why not try SipHash 1-3,
1-2, etc. for DoS mitigation?
Rust and Swift both went with SipHash 1-3 for hash tables.
On Fri, Dec 16, 2016 at 9:17 PM, George Spelvin
<[email protected]> wrote:
> My (speaking enerally; I should walk through every hash table you've
> converted) opinion is that:
>
> - Hash tables, even network-facing ones, can all use hsiphash as long
> as an attacker can only see collisions, i.e. ((H(x) ^ H(y)) & bits) ==
> 0, and the consequences of a successful attack is only more collisions
> (timing). While the attack is only 2x the cost (two hashes rather than
> one to test a key), the knowledge of the collision is statistical,
> especially for network attackers, which raises the cost of guessing
> beyond an even more brute-force attack.
> - When the hash value directly visible (e.g. included in a network
> packet), full SipHash should be the default.
> - Syncookies *could* use hsiphash, especially as there are
> two keys in there. Not sure if we need the performance.
> - For TCP ISNs, I'd prefer to use full SipHash. I know this is
> a very hot path, and if that's a performance bottleneck,
> we can work harder on it.
>
> In particular, TCP ISNs *used* to rotate the key periodically,
> limiting the time available to an attacker to perform an
> attack before the secret goes stale and is useless. commit
> 6e5714eaf77d79ae1c8b47e3e040ff5411b717ec upgraded to md5 and dropped
> the key rotation.
While I generally agree with this analysis for the most part, I do
think we should use SipHash and not HalfSipHash for syncookies.
Although the security risk is lower than with sequence numbers, it
previously used full MD5 for this, which means performance is not
generally a bottleneck and we'll get massive speedups no matter what,
whether using SipHash or HalfSipHash. In addition, using SipHash means
that the 128-bit key gives a larger margin and can be safe longterm.
So, I think we should err on the side of caution and stick with
SipHash in all cases in which we're upgrading from MD5.
In other words, only current jhash users should be potentially
eligible for hsiphash.
> Current code uses a 64 ns tick for the ISN, so it counts 2^24 per second.
> (32 bits wraps every 4.6 minutes.) A 4-bit counter and 28-bit hash
> (or even 3+29) would work as long as the key is regenerated no more
> than once per minute. (Just using the 4.6-minute ISN wrap time is the
> obvious simple implementation.)
>
> (Of course, I defer to DaveM's judgement on all network-related issues.)
I saw that jiffies addition in there and was wondering what it was all
about. It's currently added _after_ the siphash input, not before, to
keep with how the old algorithm worked. I'm not sure if this is
correct or if there's something wrong with that, as I haven't studied
how it works. If that jiffies should be part of the siphash input and
not added to the result, please tell me. Otherwise I'll keep things
how they are to avoid breaking something that seems to be working.
On Fri, Dec 16, 2016 at 12:41 PM, George Spelvin
<[email protected]> wrote:
> Tom Herbert wrote:
>> Tested this. Distribution and avalanche effect are still good. Speed
>> wise I see about a 33% improvement over siphash (20 nsecs/op versus 32
>> nsecs). That's about 3x of jhash speed (7 nsecs). So that might closer
>> to a more palatable replacement for jhash. Do we lose any security
>> advantages with halfsiphash?
>
> What are you testing on? And what input size? And does "33% improvement"
> mean 4/3 the rate and 3/4 the time? Or 2/3 the time and 3/2 the rate?
>
Sorry, that is over an IPv4 tuple. Intel(R) Xeon(R) CPU E5-2660 0 @
2.20GHz. Recoded the function I was using to look like more like 64
bit version and yes it is indeed slower.
> These are very odd results. On a 64-bit machine, SipHash should be the
> same speed per round, and faster because it hashes more data per round.
> (Unless you're hitting some unexpected cache/decode effect due to REX
> prefixes.)
>
> On a 32-bit machine (other than ARM, where your results might make sense,
> or maybe if you're hashing large amounts of data), the difference should
> be larger.
>
> And yes, there is a *significant* security loss. SipHash is 128 bits
> ("don't worry about it"). hsiphash is 64 bits, which is known breakable
> ("worry about it"), so we have to do a careful analysis of the cost of
> a successful attack.
>
> As mentioned in the e-mails that just flew by, hsiphash is intended
> *only* for 32-bit machines which bog down on full SipHash. On all 64-bit
> machines, it will be implemented as an alias for SipHash and the security
> concerns will Just Go Away.
>
> The place where hsiphash is expected to make a big difference is 32-bit
> x86. If you only see 33% difference with "gcc -m32", I'm going to be
> very confused.
Hi Ted,
On Fri, Dec 16, 2016 at 9:43 PM, Theodore Ts'o <[email protected]> wrote:
> What should we do with get_random_int() and get_random_long()? In
> some cases it's being used in performance sensitive areas, and where
> anti-DoS protection might be enough. In others, maybe not so much.
>
> If we rekeyed the secret used by get_random_int() and
> get_random_long() frequently (say, every minute or every 5 minutes),
> would that be sufficient for current and future users of these
> interfaces?
get_random_int() and get_random_long() should quite clearly use
SipHash with its secure 128-bit key and not HalfSipHash with its
64-bit key. HalfSipHash is absolutely insufficient for this use case.
Remember, we're already an order of magnitude or more faster than
md5...
With regard to periodic rekeying... since the secret is 128-bits, I
believe this is likely sufficient for _not_ rekeying. There's also the
chaining variable, to tie together invocations of the function. If
you'd prefer, instead of the chaining variable, we could use some
siphash output to mutate the original key, but I don't think this
approach is actually better and might introduce vulnerabilities. In my
opinion chaining+128bitkey is sufficient. On the other hand, rekeying
every X minutes is 3 or 4 lines of code. If you want (just say so),
I'll add this to my next revision.
You asked about the security requirements of these functions. The
comment says they're not cryptographically secure. And right now with
MD5 they're not. So the expectations are pretty low. Moving to siphash
adds some cryptographic security, certainly. Moving to siphash plus
rekeying adds a bit more. Of course, on recent x86, RDRAND is used
instead, so the cryptographic strength then depends on the thickness
of your tinfoil hat. So probably we shouldn't change what we advertise
these functions provide, even though we're certainly improving them
performance-wise and security-wise.
> P.S. I'll note that my performance figures when testing changes to
> get_random_int() were done on a 32-bit x86; Jason, I'm guessing your
> figures were using a 64-bit x86 system?. I haven't tried 32-bit ARM
> or smaller CPU's (e.g., mips, et. al.) that might be more likely to be
> used on IoT devices, but I'm worried about those too, of course.
Yes, on x86-64. But on i386 chacha20 incurs nearly the same kind of
slowdown as siphash, so I expect the comparison to be more or less
equal. There's another thing I really didn't like about your chacha20
approach which is that it uses the /dev/urandom pool, which means
various things need to kick in in the background to refill this.
Additionally, having to refill the buffered chacha output every 32 or
so longs isn't nice. These things together make for inconsistent and
hard to understand general operating system performance, because
get_random_long is called at every process startup for ASLR. So, in
the end, I believe there's another reason for going with the siphash
approach: deterministic performance.
Jason
Hi Daniel,
On Fri, Dec 16, 2016 at 9:44 PM, Daniel Micay <[email protected]> wrote:
> On Fri, 2016-12-16 at 11:47 -0800, Tom Herbert wrote:
>>
>> That's about 3x of jhash speed (7 nsecs). So that might closer
>> to a more palatable replacement for jhash. Do we lose any security
>> advantages with halfsiphash?
>
> Have you tested a lower round SipHash? Probably best to stick with the
> usual construction for non-DoS mitigation, but why not try SipHash 1-3,
> 1-2, etc. for DoS mitigation?
>
> Rust and Swift both went with SipHash 1-3 for hash tables.
Maybe not a bad idea.
SipHash2-4 for MD5 replacement, as we've done so far. This is when we
actually want things to be secure (and fast).
And then HalfSipHash1-3 for certain jhash replacements. This is for
when we're talking only about DoS or sort of just joking about
security, and want things to be very competitive with jhash. (Of
course for 64-bit we'd use SipHash1-3 instead of HalfSipHash for the
speedup.)
I need to think on this a bit more, but preliminarily, I guess this
would be maybe okay...
George, JP - what do you think?
Jason
On Fri, Dec 16, 2016, at 22:01, Jason A. Donenfeld wrote:
> Yes, on x86-64. But on i386 chacha20 incurs nearly the same kind of
> slowdown as siphash, so I expect the comparison to be more or less
> equal. There's another thing I really didn't like about your chacha20
> approach which is that it uses the /dev/urandom pool, which means
> various things need to kick in in the background to refill this.
> Additionally, having to refill the buffered chacha output every 32 or
> so longs isn't nice. These things together make for inconsistent and
> hard to understand general operating system performance, because
> get_random_long is called at every process startup for ASLR. So, in
> the end, I believe there's another reason for going with the siphash
> approach: deterministic performance.
*Hust*, so from where do you generate your key for siphash if called
early from ASLR?
Bye,
Hannes
Jason A. Donenfeld wrote:
> I saw that jiffies addition in there and was wondering what it was all
> about. It's currently added _after_ the siphash input, not before, to
> keep with how the old algorithm worked. I'm not sure if this is
> correct or if there's something wrong with that, as I haven't studied
> how it works. If that jiffies should be part of the siphash input and
> not added to the result, please tell me. Otherwise I'll keep things
> how they are to avoid breaking something that seems to be working.
Oh, geez, I didn't realize you didn't understand this code.
Full details at
https://en.wikipedia.org/wiki/TCP_sequence_prediction_attack
But yes, the sequence number is supposed to be (random base) + (timestamp).
In the old days before Canter & Siegel when the internet was a nice place,
people just used a counter that started at boot time.
But then someone observed that I can start a connection to host X,
see the sequence number it gives back to me, and thereby learn the
seauence number it's using on its connections to host Y.
And I can use that to inject forged data into an X-to-Y connection,
without ever seeing a single byte of the traffic! (If I *can* observe
the traffic, of course, none of this makes the slightest difference.)
So the random base was made a keyed hash of the endpoint identifiers.
(Practically only the hosts matter, but generally the ports are thrown
in for good measure.) That way, the ISN that host X sends to me
tells me nothing about the ISN it's using to talk to host Y. Now the
only way to inject forged data into the X-to-Y connection is to
send 2^32 bytes, which is a little less practical.
Hi George,
On Fri, Dec 16, 2016 at 10:25 PM, George Spelvin
<[email protected]> wrote:
> But yes, the sequence number is supposed to be (random base) + (timestamp).
> In the old days before Canter & Siegel when the internet was a nice place,
> people just used a counter that started at boot time.
>
> But then someone observed that I can start a connection to host X,
> see the sequence number it gives back to me, and thereby learn the
> seauence number it's using on its connections to host Y.
>
> And I can use that to inject forged data into an X-to-Y connection,
> without ever seeing a single byte of the traffic! (If I *can* observe
> the traffic, of course, none of this makes the slightest difference.)
>
> So the random base was made a keyed hash of the endpoint identifiers.
> (Practically only the hosts matter, but generally the ports are thrown
> in for good measure.) That way, the ISN that host X sends to me
> tells me nothing about the ISN it's using to talk to host Y. Now the
> only way to inject forged data into the X-to-Y connection is to
> send 2^32 bytes, which is a little less practical.
Oh, okay, that is exactly what I thought was going on. I just thought
you were implying that jiffies could be moved inside the hash, which
then confused my understanding of how things should be. In any case,
thanks for the explanation.
Jason
On Thu, Dec 15, 2016 at 7:03 PM, Jason A. Donenfeld <[email protected]> wrote:
> -static DEFINE_PER_CPU(__u32 [MD5_DIGEST_WORDS], get_random_int_hash)
> - __aligned(sizeof(unsigned long));
> +static DEFINE_PER_CPU(u64, get_random_int_chaining);
>
[...]
> unsigned long get_random_long(void)
> {
> - __u32 *hash;
> unsigned long ret;
> + u64 *chaining;
>
> if (arch_get_random_long(&ret))
> return ret;
>
> - hash = get_cpu_var(get_random_int_hash);
> -
> - hash[0] += current->pid + jiffies + random_get_entropy();
> - md5_transform(hash, random_int_secret);
> - ret = *(unsigned long *)hash;
> - put_cpu_var(get_random_int_hash);
> -
> + chaining = &get_cpu_var(get_random_int_chaining);
> + ret = *chaining = siphash_3u64(*chaining, jiffies, random_get_entropy() +
> + current->pid, random_int_secret);
> + put_cpu_var(get_random_int_chaining);
> return ret;
> }
I think it would be nice to try to strenghen the PRNG construction.
FWIW, I'm not an expert in PRNGs, and there's fairly extensive
literature, but I can at least try. Here are some properties I'd
like:
1. A one-time leak of memory contents doesn't ruin security until
reboot. This is especially value across suspend and/or hibernation.
2. An attack with a low work factor (2^64?) shouldn't break the scheme
until reboot.
This is effectively doing:
output = H(prev_output, weak "entropy", per-boot secret);
One unfortunately downside is that, if used in a context where an
attacker can see a single output, the attacker learns the chaining
value. If the attacker can guess the entropy, then, with 2^64 work,
they learn the secret, and they can predict future outputs.
I would advocate adding two types of improvements. First, re-seed it
every now and then (every 128 calls?) by just replacing both the
chaining value and the percpu secret with fresh CSPRNG output.
Second, change the mode so that an attacker doesn't learn so much
internal state. For example:
output = H(old_chain, entropy, secret);
new_chain = old_chain + entropy + output;
This increases the effort needed to brute-force the internal state
from 2^64 to 2^128 (barring any weaknesses in the scheme).
Also, can we not call this get_random_int()? get_random_int() sounds
too much like get_random_bytes(), and the latter is intended to be a
real CSPRNG. Can we call it get_weak_random_int() or similar?
--Andy
Hi Andy,
On Fri, Dec 16, 2016 at 10:31 PM, Andy Lutomirski <[email protected]> wrote:
> I think it would be nice to try to strenghen the PRNG construction.
> FWIW, I'm not an expert in PRNGs, and there's fairly extensive
> literature, but I can at least try.
In an effort to keep this patchset as initially as uncontroversial as
possible, I kept the same same construction as before and just swapped
out slow MD5 for fast Siphash. Additionally, the function
documentation says that it isn't cryptographically secure. But in the
end I certainly agree with you; we should most definitely improve
things, and seeing the eyeballs now on this series, I think we now
might have a mandate to do so.
> 1. A one-time leak of memory contents doesn't ruin security until
> reboot. This is especially value across suspend and/or hibernation.
Ted and I were discussing this in another thread, and it sounds like
he wants the same thing. I'll add re-generation of the secret every
once in a while. Perhaps time-based makes more sense than
counter-based for rekeying frequency?
> 2. An attack with a low work factor (2^64?) shouldn't break the scheme
> until reboot.
It won't. The key is 128-bits.
> This is effectively doing:
>
> output = H(prev_output, weak "entropy", per-boot secret);
>
> One unfortunately downside is that, if used in a context where an
> attacker can see a single output, the attacker learns the chaining
> value. If the attacker can guess the entropy, then, with 2^64 work,
> they learn the secret, and they can predict future outputs.
No, the secret is 128-bits, which isn't feasibly guessable. The secret
also isn't part of the hash, but rather is the generator of the hash
function. A keyed hash (a PRF) is a bit of a different construction
than just hashing a secret value into a hash function.
> Second, change the mode so that an attacker doesn't learn so much
> internal state. For example:
>
> output = H(old_chain, entropy, secret);
> new_chain = old_chain + entropy + output;
Happy to make this change, with making the chaining value additive
rather than a replacement.
However, I'm not sure adding entropy to the new_chain makes a
different. That entropy is basically just the cycle count plus the
jiffies count. If an attacker can already guess them, then adding them
again to the chaining value doesn't really add much.
Jason
Hi Andy, Ted,
I've made the requested changes. Keys now rotate and are per-CPU
based. The chaining value is now additive instead of replacing.
DavidL suggested I lower the velocity of `git-send-email` triggers, so
if you'd like to take a look at this before I post v7, you can follow
along at my git tree here:
https://git.zx2c4.com/linux-dev/log/?h=siphash
Choose the commit entitled "random: use SipHash in place of MD5"
Thanks,
Jason
> What should we do with get_random_int() and get_random_long()? In
> some cases it's being used in performance sensitive areas, and where
> anti-DoS protection might be enough. In others, maybe not so much.
This is tricky. The entire get_random_int() structure is an abuse of
the hash function and will need to be thoroughly rethought to convert
it to SipHash. Remember, SipHash's security goals are very different
from MD5, so there's no obvious way to do the conversion.
(It's *documented* as "not cryptographically secure", but we know
where that goes.)
> If we rekeyed the secret used by get_random_int() and
> get_random_long() frequently (say, every minute or every 5 minutes),
> would that be sufficient for current and future users of these
> interfaces?
Remembering that on "real" machines it's full SipHash, then I'd say that
64-bit security + rekeying seems reasonable.
The question is, the idea has recently been floated to make hsiphash =
SipHash-1-3 on 64-bit machines. Is *that* okay?
The annoying thing about the currently proposed patch is that the *only*
chaining is the returned value. What I'd *like* to do is the same
pattern as we do with md5, and remember v[0..3] between invocations.
But there's no partial SipHash primitive; we only get one word back.
Even
*chaining += ret = siphash_3u64(...)
would be an improvement.
Although we could do something like
c0 = chaining[0];
chaining[0] = c1 = chaining[1];
ret = hsiphash(c0, c1, ...)
chaining[1] = c0 + ret;
On Fri, Dec 16, 2016 at 1:45 PM, Jason A. Donenfeld <[email protected]> wrote:
> Hi Andy,
>
> On Fri, Dec 16, 2016 at 10:31 PM, Andy Lutomirski <[email protected]> wrote:
>> I think it would be nice to try to strenghen the PRNG construction.
>> FWIW, I'm not an expert in PRNGs, and there's fairly extensive
>> literature, but I can at least try.
>
> In an effort to keep this patchset as initially as uncontroversial as
> possible, I kept the same same construction as before and just swapped
> out slow MD5 for fast Siphash. Additionally, the function
> documentation says that it isn't cryptographically secure. But in the
> end I certainly agree with you; we should most definitely improve
> things, and seeing the eyeballs now on this series, I think we now
> might have a mandate to do so.
>
>> 1. A one-time leak of memory contents doesn't ruin security until
>> reboot. This is especially value across suspend and/or hibernation.
>
> Ted and I were discussing this in another thread, and it sounds like
> he wants the same thing. I'll add re-generation of the secret every
> once in a while. Perhaps time-based makes more sense than
> counter-based for rekeying frequency?
Counter may be faster -- you don't need to read a timer. Lots of CPUs
are surprisingly slow at timing. OTOH jiffies are fast.
>
>> 2. An attack with a low work factor (2^64?) shouldn't break the scheme
>> until reboot.
>
> It won't. The key is 128-bits.
Whoops, I thought the key was 64-bit...
>
>> This is effectively doing:
>>
>> output = H(prev_output, weak "entropy", per-boot secret);
>>
>> One unfortunately downside is that, if used in a context where an
>> attacker can see a single output, the attacker learns the chaining
>> value. If the attacker can guess the entropy, then, with 2^64 work,
>> they learn the secret, and they can predict future outputs.
>
> No, the secret is 128-bits, which isn't feasibly guessable. The secret
> also isn't part of the hash, but rather is the generator of the hash
> function. A keyed hash (a PRF) is a bit of a different construction
> than just hashing a secret value into a hash function.
I was thinking in the random oracle model, in which case the whole
function is just a PRF that takes a few input parameters, one of which
is a key.
>
>> Second, change the mode so that an attacker doesn't learn so much
>> internal state. For example:
>>
>> output = H(old_chain, entropy, secret);
>> new_chain = old_chain + entropy + output;
>
> Happy to make this change, with making the chaining value additive
> rather than a replacement.
>
> However, I'm not sure adding entropy to the new_chain makes a
> different. That entropy is basically just the cycle count plus the
> jiffies count. If an attacker can already guess them, then adding them
> again to the chaining value doesn't really add much.
Agreed. A simpler contruction would be:
chaining++;
output = H(chaining, secret);
And this looks a whole lot like Ted's ChaCha20 construction.
The benefit of my construction is that (in the random oracle model,
assuming my intuition is right), if an attacker misses ~128 samples
and entropy has at least one bit of independent min-entropy per
sample, then the attacker needs ~2^128 work to brute-force the
chaining value even fi the attacker knew both the original chaining
value and the secret.
--Andy
On Fri, Dec 16, 2016 at 2:13 PM, George Spelvin
<[email protected]> wrote:
>> What should we do with get_random_int() and get_random_long()? In
>> some cases it's being used in performance sensitive areas, and where
>> anti-DoS protection might be enough. In others, maybe not so much.
>
> This is tricky. The entire get_random_int() structure is an abuse of
> the hash function and will need to be thoroughly rethought to convert
> it to SipHash. Remember, SipHash's security goals are very different
> from MD5, so there's no obvious way to do the conversion.
>
> (It's *documented* as "not cryptographically secure", but we know
> where that goes.)
>
>> If we rekeyed the secret used by get_random_int() and
>> get_random_long() frequently (say, every minute or every 5 minutes),
>> would that be sufficient for current and future users of these
>> interfaces?
>
> Remembering that on "real" machines it's full SipHash, then I'd say that
> 64-bit security + rekeying seems reasonable.
>
> The question is, the idea has recently been floated to make hsiphash =
> SipHash-1-3 on 64-bit machines. Is *that* okay?
>
>
> The annoying thing about the currently proposed patch is that the *only*
> chaining is the returned value. What I'd *like* to do is the same
> pattern as we do with md5, and remember v[0..3] between invocations.
> But there's no partial SipHash primitive; we only get one word back.
>
> Even
> *chaining += ret = siphash_3u64(...)
>
> would be an improvement.
This is almost exactly what I suggested in my email on the other
thread from a few seconds ago :)
--Andy
On Fri, Dec 16, 2016 at 11:13 PM, George Spelvin
<[email protected]> wrote:
> Remembering that on "real" machines it's full SipHash, then I'd say that
> 64-bit security + rekeying seems reasonable.
64-bit security for an RNG is not reasonable even with rekeying. No no
no. Considering we already have a massive speed-up here with the
secure version, there's zero reason to start weakening the security
because we're trigger happy with our benchmarks. No no no.
Hi Andy,
> Agreed. A simpler contruction would be:
>
> chaining++;
> output = H(chaining, secret);
>
> And this looks a whole lot like Ted's ChaCha20 construction.
In that simpler construction with counter-based secret rekeying and in
Ted's ChaCha20 construction, the issue is that every X hits, there's a
call to get_random_bytes, which has variable performance and entropy
issues. Doing it my way with it being time based, in the event that
somebody runs ` :(){ :|:& };:`, system performance doesn't suffer
because ASLR is making repeated calls to get_random_bytes every 128 or
so process creations. In the time based way, the system performance
will not suffer.
Jason
An idea I had which mght be useful:
You could perhaps save two rounds in siphash_*u64.
The final word with the length (called "b" in your implementation)
only needs to be there if the input is variable-sized.
If every use of a given key is of a fixed-size input, you don't need
a length suffix. When the input is an even number of words, that can
save you two rounds.
This requires an audit of callers (e.g. you have to use different
keys for IPv4 and IPv6 ISNs), but can save time.
(This is crypto 101; search "MD-strengthening" or see the remark on
p. 101 on Damgaard's 1989 paper "A design principle for hash functions" at
http://saluc.engr.uconn.edu/refs/algorithms/hashalg/damgard89adesign.pdf
but I'm sure that Ted, Jean-Philippe, and/or DJB will confirm if you'd
like.)
Jason A. Donenfeld wrote:
> Oh, okay, that is exactly what I thought was going on. I just thought
> you were implying that jiffies could be moved inside the hash, which
> then confused my understanding of how things should be. In any case,
> thanks for the explanation.
No, the rekeying procedure is cleverer.
The thing is, all that matters is that the ISN increments fast enough,
but not wrap too soon.
It *is* permitted to change the random base, as long as it only
increases, and slower than the timestamp does.
So what you do is every few minutes, you increment the high 4 bits of the
random base and change the key used to generate the low 28 bits.
The base used for any particular host might change from 0x10000000
to 0x2fffffff, or from 0x1fffffff to 0x20000000, but either way, it's
increasing, and not too fast.
This has the downside that an attacker can see 4 bits of the base,
so only needs to send send 2^28 = 256 MB to flood the connection,
but the upside that the key used to generate the low bits changes
faster than it can be broken.
> 64-bit security for an RNG is not reasonable even with rekeying. No no
> no. Considering we already have a massive speed-up here with the
> secure version, there's zero reason to start weakening the security
> because we're trigger happy with our benchmarks. No no no.
Just to clarify, I was discussing the idea with Ted (who's in charge of
the whole thing, not me), not trying to make any sort of final decision
on the subject. I need to look at the various users (46 non-trivial ones
for get_random_int, 15 for get_random_long) and see what their security
requirements actually are.
I'm also trying to see if HalfSipHash can be used in a way that gives
slightly more than 64 bits of effective security.
The problem is that the old MD5-based transform had unclear, but
obviously ample, security. There were 64 bytes of global secret and
16 chaining bytes per CPU. Adapting SipHash (even the full version)
takes more thinking.
An actual HalfSipHash-based equivalent to the existing code would be:
#define RANDOM_INT_WORDS (64 / sizeof(long)) /* 16 or 8 */
static u32 random_int_secret[RANDOM_INT_WORDS]
____cacheline_aligned __read_mostly;
static DEFINE_PER_CPU(unsigned long[4], get_random_int_hash)
__aligned(sizeof(unsigned long));
unsigned long get_random_long(void)
{
unsigned long *hash = get_cpu_var(get_random_int_hash);
unsigned long v0 = hash[0], v1 = hash[1], v2 = hash[2], v3 = hash[3];
int i;
/* This could be improved, but it's equivalent */
v0 += current->pid + jiffies + random_get_entropy();
for (i = 0; i < RANDOM_INT_WORDS; i++) {
v3 ^= random_int_secret[i];
HSIPROUND;
HSIPROUND;
v0 ^= random_int_secret[i];
}
/* To be equivalent, we *don't* finalize the transform */
hash[0] = v0; hash[1] = v1; hash[2] = v2; hash[3] = v3;
put_cpu_var(get_random_int_hash);
return v0 ^ v1 ^ v2 ^ v3;
}
I don't think there's a 2^64 attack on that.
But 64 bytes of global secret is ridiculous if the hash function
doesn't require that minimum block size. It'll take some thinking.
Ths advice I'd give now is:
- Implement
unsigned long hsiphash(const void *data, size_t len, const unsigned long key[2])
.. as SipHash on 64-bit (maybe SipHash-1-3, still being discussed) and
HalfSipHash on 32-bit.
- Document when it may or may not be used carefully.
- #define get_random_int (unsigned)get_random_long
- Ted, Andy Lutorminski and I will try to figure out a construction of
get_random_long() that we all like.
('scuse me for a few hours, I have some unrelated things I really *should*
be working on...)
On Sat, Dec 17, 2016 at 12:44 AM, George Spelvin
<[email protected]> wrote:
> Ths advice I'd give now is:
> - Implement
> unsigned long hsiphash(const void *data, size_t len, const unsigned long key[2])
> .. as SipHash on 64-bit (maybe SipHash-1-3, still being discussed) and
> HalfSipHash on 32-bit.
I already did this. Check my branch.
> - Document when it may or may not be used carefully.
Good idea. I'll write up some extensive documentation about all of
this, detailing use cases and our various conclusions.
> - #define get_random_int (unsigned)get_random_long
That's a good idea, since ultimately the other just casts in the
return value. I wonder if this could also lead to a similar aliasing
with arch_get_random_int, since I'm pretty sure all rdrand-like
instructions return native word size anyway.
> - Ted, Andy Lutorminski and I will try to figure out a construction of
> get_random_long() that we all like.
And me, I hope... No need to make this exclusive.
Jason
> I already did this. Check my branch.
Do you think it should return "u32" (as you currently have it) or
"unsigned long"? I thought the latter, since it doesn't cost any
more and makes more
> I wonder if this could also lead to a similar aliasing
> with arch_get_random_int, since I'm pretty sure all rdrand-like
> instructions return native word size anyway.
Well, Intel's can return 16, 32 or 64 bits, and it makes a
small difference with reseed scheduling.
>> - Ted, Andy Lutorminski and I will try to figure out a construction of
>> get_random_long() that we all like.
> And me, I hope... No need to make this exclusive.
Gaah, engage brain before fingers. That was so obvious I didn't say
it, and the result came out sounding extremely rude.
A better (but longer) way to write it would be "I'm sorry that I, Ted,
and Andy are all arguing with you and each other about how to do this
and we can't finalize this part yet".
BTW, here's some SipHash code I wrote for Linux a while ago.
My target application was ext4 directory hashing, resulting in different
implementation choices, although I still think that a rolled-up
implementation like this is reasonable. Reducing I-cache impact speeds
up the calling code.
One thing I'd like to suggest you steal is the way it handles the
fetch of the final partial word. It's a lot smaller and faster than
an 8-way case statement.
#include <linux/bitops.h> /* For rol64 */
#include <linux/cryptohash.h>
#include <asm/byteorder.h>
#include <asm/unaligned.h>
/* The basic ARX mixing function, taken from Skein */
#define SIP_MIX(a, b, s) ((a) += (b), (b) = rol64(b, s), (b) ^= (a))
/*
* The complete SipRound. Note that, when unrolled twice like below,
* the 32-bit rotates drop out on 32-bit machines.
*/
#define SIP_ROUND(a, b, c, d) \
(SIP_MIX(a, b, 13), SIP_MIX(c, d, 16), (a) = rol64(a, 32), \
SIP_MIX(c, b, 17), SIP_MIX(a, d, 21), (c) = rol64(c, 32))
/*
* This is rolled up more than most implementations, resulting in about
* 55% the code size. Speed is a few precent slower. A crude benchmark
* (for (i=1; i <= max; i++) for (j = 0; j < 4096-i; j++) hash(buf+j, i);)
* produces the following timings (in usec):
*
* i386 i386 i386 x86_64 x86_64 x86_64 x86_64
* Length small unroll halfmd4 small unroll halfmd4 teahash
* 1..4 1069 1029 1608 195 160 399 690
* 1..8 2483 2381 3851 410 360 988 1659
* 1..12 4303 4152 6207 690 618 1642 2690
* 1..16 6122 5931 8668 968 876 2363 3786
* 1..20 8348 8137 11245 1323 1185 3162 5567
* 1..24 10580 10327 13935 1657 1504 4066 7635
* 1..28 13211 12956 16803 2069 1871 5028 9759
* 1..32 15843 15572 19725 2470 2260 6084 11932
* 1..36 18864 18609 24259 2934 2678 7566 14794
* 1..1024 5890194 6130242 10264816 881933 881244 3617392 7589036
*
* The performance penalty is quite minor, decreasing for long strings,
* and it's significantly faster than half_md4, so I'm going for the
* I-cache win.
*/
uint64_t
siphash24(char const *in, size_t len, uint32_t const seed[4])
{
uint64_t a = 0x736f6d6570736575; /* somepseu */
uint64_t b = 0x646f72616e646f6d; /* dorandom */
uint64_t c = 0x6c7967656e657261; /* lygenera */
uint64_t d = 0x7465646279746573; /* tedbytes */
uint64_t m = 0;
uint8_t padbyte = len;
/*
* Mix in the 128-bit hash seed. This is in a format convenient
* to the ext3/ext4 code. Please feel free to adapt the
* */
if (seed) {
m = seed[2] | (uint64_t)seed[3] << 32;
b ^= m;
d ^= m;
m = seed[0] | (uint64_t)seed[1] << 32;
/* a ^= m; is done in loop below */
c ^= m;
}
/*
* By using the same SipRound code for all iterations, we
* save space, at the expense of some branch prediction. But
* branch prediction is hard because of variable length anyway.
*/
len = len/8 + 3; /* Now number of rounds to perform */
do {
a ^= m;
switch (--len) {
unsigned bytes;
default: /* Full words */
d ^= m = get_unaligned_le64(in);
in += 8;
break;
case 2: /* Final partial word */
/*
* We'd like to do one 64-bit fetch rather than
* mess around with bytes, but reading past the end
* might hit a protection boundary. Fortunately,
* we know that protection boundaries are aligned,
* so we can consider only three cases:
* - The remainder occupies zero words
* - The remainder fits into one word
* - The remainder straddles two words
*/
bytes = padbyte & 7;
if (bytes == 0) {
m = 0;
} else {
unsigned offset = (unsigned)(uintptr_t)in & 7;
if (offset + bytes <= 8) {
m = le64_to_cpup((uint64_t const *)
(in - offset));
m >>= 8*offset;
} else {
m = get_unaligned_le64(in);
}
m &= ((uint64_t)1 << 8*bytes) - 1;
}
/* Could use | or +, but ^ allows associativity */
d ^= m ^= (uint64_t)padbyte << 56;
break;
case 1: /* Beginning of finalization */
m = 0;
c ^= 0xff;
/*FALLTHROUGH*/
case 0: /* Second half of finalization */
break;
}
SIP_ROUND(a, b, c, d);
SIP_ROUND(a, b, c, d);
} while (len);
return a ^ b ^ c ^ d;
}
#undef SIP_ROUND
#undef SIP_MIX
/*
* No objection to EXPORT_SYMBOL, but we should probably figure out
* how the seed[] array should work first. Homework for the first
* person to want to call it from a module!
*/
> diff --git a/lib/test_siphash.c b/lib/test_siphash.c
> new file mode 100644
> index 000000000000..93549e4e22c5
> --- /dev/null
> +++ b/lib/test_siphash.c
> @@ -0,0 +1,83 @@
> +/* Test cases for siphash.c
> + *
> + * Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
> + *
> + * This file is provided under a dual BSD/GPLv2 license.
> + *
> + * SipHash: a fast short-input PRF
> + * https://131002.net/siphash/
> + *
> + * This implementation is specifically for SipHash2-4.
> + */
> +
> +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
> +
> +#include <linux/siphash.h>
> +#include <linux/kernel.h>
> +#include <linux/string.h>
> +#include <linux/errno.h>
> +#include <linux/module.h>
> +
> +/* Test vectors taken from official reference source available at:
> + * https://131002.net/siphash/siphash24.c
> + */
> +static const u64 test_vectors[64] = {
> + 0x726fdb47dd0e0e31ULL, 0x74f839c593dc67fdULL, 0x0d6c8009d9a94f5aULL,
> + 0x85676696d7fb7e2dULL, 0xcf2794e0277187b7ULL, 0x18765564cd99a68dULL,
> + 0xcbc9466e58fee3ceULL, 0xab0200f58b01d137ULL, 0x93f5f5799a932462ULL,
> + 0x9e0082df0ba9e4b0ULL, 0x7a5dbbc594ddb9f3ULL, 0xf4b32f46226bada7ULL,
> + 0x751e8fbc860ee5fbULL, 0x14ea5627c0843d90ULL, 0xf723ca908e7af2eeULL,
> + 0xa129ca6149be45e5ULL, 0x3f2acc7f57c29bdbULL, 0x699ae9f52cbe4794ULL,
> + 0x4bc1b3f0968dd39cULL, 0xbb6dc91da77961bdULL, 0xbed65cf21aa2ee98ULL,
> + 0xd0f2cbb02e3b67c7ULL, 0x93536795e3a33e88ULL, 0xa80c038ccd5ccec8ULL,
> + 0xb8ad50c6f649af94ULL, 0xbce192de8a85b8eaULL, 0x17d835b85bbb15f3ULL,
> + 0x2f2e6163076bcfadULL, 0xde4daaaca71dc9a5ULL, 0xa6a2506687956571ULL,
> + 0xad87a3535c49ef28ULL, 0x32d892fad841c342ULL, 0x7127512f72f27cceULL,
> + 0xa7f32346f95978e3ULL, 0x12e0b01abb051238ULL, 0x15e034d40fa197aeULL,
> + 0x314dffbe0815a3b4ULL, 0x027990f029623981ULL, 0xcadcd4e59ef40c4dULL,
> + 0x9abfd8766a33735cULL, 0x0e3ea96b5304a7d0ULL, 0xad0c42d6fc585992ULL,
> + 0x187306c89bc215a9ULL, 0xd4a60abcf3792b95ULL, 0xf935451de4f21df2ULL,
> + 0xa9538f0419755787ULL, 0xdb9acddff56ca510ULL, 0xd06c98cd5c0975ebULL,
> + 0xe612a3cb9ecba951ULL, 0xc766e62cfcadaf96ULL, 0xee64435a9752fe72ULL,
> + 0xa192d576b245165aULL, 0x0a8787bf8ecb74b2ULL, 0x81b3e73d20b49b6fULL,
> + 0x7fa8220ba3b2eceaULL, 0x245731c13ca42499ULL, 0xb78dbfaf3a8d83bdULL,
> + 0xea1ad565322a1a0bULL, 0x60e61c23a3795013ULL, 0x6606d7e446282b93ULL,
> + 0x6ca4ecb15c5f91e1ULL, 0x9f626da15c9625f3ULL, 0xe51b38608ef25f57ULL,
> + 0x958a324ceb064572ULL
> +};
> +static const siphash_key_t test_key =
> + { 0x0706050403020100ULL , 0x0f0e0d0c0b0a0908ULL };
> +
> +static int __init siphash_test_init(void)
> +{
> + u8 in[64] __aligned(SIPHASH_ALIGNMENT);
> + u8 in_unaligned[65];
> + u8 i;
> + int ret = 0;
> +
> + for (i = 0; i < 64; ++i) {
> + in[i] = i;
> + in_unaligned[i + 1] = i;
> + if (siphash(in, i, test_key) != test_vectors[i]) {
> + pr_info("self-test aligned %u: FAIL\n", i + 1);
> + ret = -EINVAL;
> + }
> + if (siphash_unaligned(in_unaligned + 1, i, test_key) != test_vectors[i]) {
> + pr_info("self-test unaligned %u: FAIL\n", i + 1);
> + ret = -EINVAL;
> + }
> + }
> + if (!ret)
> + pr_info("self-tests: pass\n");
> + return ret;
> +}
> +
> +static void __exit siphash_test_exit(void)
> +{
> +}
> +
> +module_init(siphash_test_init);
> +module_exit(siphash_test_exit);
> +
> +MODULE_AUTHOR("Jason A. Donenfeld <[email protected]>");
> +MODULE_LICENSE("Dual BSD/GPL");
> --
> 2.11.0
>
I believe the output of SipHash depends upon endianness. Folks who
request a digest through the af_alg interface will likely expect a
byte array.
I think that means on little endian machines, values like element 0
must be reversed byte reversed:
0x726fdb47dd0e0e31ULL => 31,0e,0e,dd,47,db,6f,72
If I am not mistaken, that value (and other tv's) are returned here:
return (v0 ^ v1) ^ (v2 ^ v3);
It may be prudent to include the endian reversal in the test to ensure
big endian machines produce expected results. Some closely related
testing on an old Apple PowerMac G5 revealed that result needed to be
reversed before returning it to a caller.
Jeff
To follow up on my comments that your benchmark results were peculiar,
here's my benchmark code.
It just computes the hash of all n*(n+1)/2 possible non-empty substrings
of a buffer of n (called "max" below) bytes. "cpb" is "cycles per byte".
(The average length is (n+2)/3, c.f. https://oeis.org/A000292)
On x86-32, HSipHash is asymptotically twice the speed of SipHash,
rising to 2.5x for short strings:
SipHash/HSipHash benchmark, sizeof(long) = 4
SipHash: max= 4 cycles= 10495 cpb=524.7500 (sum=47a4f5554869fa97)
HSipHash: max= 4 cycles= 3400 cpb=170.0000 (sum=146a863e)
SipHash: max= 8 cycles= 24468 cpb=203.9000 (sum=21c41a86355affcc)
HSipHash: max= 8 cycles= 9237 cpb= 76.9750 (sum=d3b5e0cd)
SipHash: max= 16 cycles= 94622 cpb=115.9583 (sum=26d816b72721e48f)
HSipHash: max= 16 cycles= 34499 cpb= 42.2782 (sum=16bb7475)
SipHash: max= 32 cycles= 418767 cpb= 69.9811 (sum=dd5a97694b8a832d)
HSipHash: max= 32 cycles= 156695 cpb= 26.1857 (sum=eed00fcb)
SipHash: max= 64 cycles= 2119152 cpb= 46.3101 (sum=a2a725aecc09ed00)
HSipHash: max= 64 cycles= 1008678 cpb= 22.0428 (sum=99b9f4f)
SipHash: max= 128 cycles= 12728659 cpb= 35.5788 (sum=420878cd20272817)
HSipHash: max= 128 cycles= 5452931 cpb= 15.2419 (sum=f1f4ad18)
SipHash: max= 256 cycles= 38931946 cpb= 13.7615 (sum=e05dfb28b90dfd98)
HSipHash: max= 256 cycles= 13807312 cpb= 4.8805 (sum=ceeafcc1)
SipHash: max= 512 cycles= 205537380 cpb= 9.1346 (sum=7d129d4de145fbea)
HSipHash: max= 512 cycles= 103420960 cpb= 4.5963 (sum=7f15a313)
SipHash: max=1024 cycles=1540259472 cpb= 8.5817 (sum=cca7cbdc778ca8af)
HSipHash: max=1024 cycles= 796090824 cpb= 4.4355 (sum=d8f3374f)
On x86-64, SipHash is consistently faster, asymptotically approaching 2x
for long strings:
SipHash/HSipHash benchmark, sizeof(long) = 8
SipHash: max= 4 cycles= 2642 cpb=132.1000 (sum=47a4f5554869fa97)
HSipHash: max= 4 cycles= 2498 cpb=124.9000 (sum=146a863e)
SipHash: max= 8 cycles= 5270 cpb= 43.9167 (sum=21c41a86355affcc)
HSipHash: max= 8 cycles= 7140 cpb= 59.5000 (sum=d3b5e0cd)
SipHash: max= 16 cycles= 19950 cpb= 24.4485 (sum=26d816b72721e48f)
HSipHash: max= 16 cycles= 23546 cpb= 28.8554 (sum=16bb7475)
SipHash: max= 32 cycles= 80188 cpb= 13.4004 (sum=dd5a97694b8a832d)
HSipHash: max= 32 cycles= 101218 cpb= 16.9148 (sum=eed00fcb)
SipHash: max= 64 cycles= 373286 cpb= 8.1575 (sum=a2a725aecc09ed00)
HSipHash: max= 64 cycles= 535568 cpb= 11.7038 (sum=99b9f4f)
SipHash: max= 128 cycles= 2075224 cpb= 5.8006 (sum=420878cd20272817)
HSipHash: max= 128 cycles= 3336820 cpb= 9.3270 (sum=f1f4ad18)
SipHash: max= 256 cycles= 14276278 cpb= 5.0463 (sum=e05dfb28b90dfd98)
HSipHash: max= 256 cycles= 28847880 cpb= 10.1970 (sum=ceeafcc1)
SipHash: max= 512 cycles= 50135180 cpb= 2.2281 (sum=7d129d4de145fbea)
HSipHash: max= 512 cycles= 86145916 cpb= 3.8286 (sum=7f15a313)
SipHash: max=1024 cycles= 334111900 cpb= 1.8615 (sum=cca7cbdc778ca8af)
HSipHash: max=1024 cycles= 640432452 cpb= 3.5682 (sum=d8f3374f)
Here's the code; compile with -DSELFTEST. (The main purpose of
printing the sum is to prevent dead code elimination.)
#if SELFTEST
#include <stdint.h>
#include <stdlib.h>
static inline uint64_t rol64(uint64_t word, unsigned int shift)
{
return word << shift | word >> (64 - shift);
}
static inline uint32_t rol32(uint32_t word, unsigned int shift)
{
return word << shift | word >> (32 - shift);
}
static inline uint64_t get_unaligned_le64(void const *p)
{
return *(uint64_t const *)p;
}
static inline uint32_t get_unaligned_le32(void const *p)
{
return *(uint32_t const *)p;
}
static inline uint64_t le64_to_cpup(uint64_t const *p)
{
return *p;
}
static inline uint32_t le32_to_cpup(uint32_t const *p)
{
return *p;
}
#else
#include <linux/bitops.h> /* For rol64 */
#include <linux/cryptohash.h>
#include <asm/byteorder.h>
#include <asm/unaligned.h>
#endif
/* The basic ARX mixing function, taken from Skein */
#define SIP_MIX(a, b, s) ((a) += (b), (b) = rol64(b, s), (b) ^= (a))
/*
* The complete SipRound. Note that, when unrolled twice like below,
* the 32-bit rotates drop out on 32-bit machines.
*/
#define SIP_ROUND(a, b, c, d) \
(SIP_MIX(a, b, 13), SIP_MIX(c, d, 16), (a) = rol64(a, 32), \
SIP_MIX(c, b, 17), SIP_MIX(a, d, 21), (c) = rol64(c, 32))
/*
* This is rolled up more than most implementations, resulting in about
* 55% the code size. Speed is a few precent slower. A crude benchmark
* (for (i=1; i <= max; i++) for (j = 0; j < 4096-i; j++) hash(buf+j, i);)
* produces the following timings (in usec):
*
* i386 i386 i386 x86_64 x86_64 x86_64 x86_64
* Length small unroll halfmd4 small unroll halfmd4 teahash
* 1..4 1069 1029 1608 195 160 399 690
* 1..8 2483 2381 3851 410 360 988 1659
* 1..12 4303 4152 6207 690 618 1642 2690
* 1..16 6122 5931 8668 968 876 2363 3786
* 1..20 8348 8137 11245 1323 1185 3162 5567
* 1..24 10580 10327 13935 1657 1504 4066 7635
* 1..28 13211 12956 16803 2069 1871 5028 9759
* 1..32 15843 15572 19725 2470 2260 6084 11932
* 1..36 18864 18609 24259 2934 2678 7566 14794
* 1..1024 5890194 6130242 10264816 881933 881244 3617392 7589036
*
* The performance penalty is quite minor, decreasing for long strings,
* and it's significantly faster than half_md4, so I'm going for the
* I-cache win.
*/
uint64_t
siphash24(char const *in, size_t len, uint32_t const seed[4])
{
uint64_t a = 0x736f6d6570736575; /* somepseu */
uint64_t b = 0x646f72616e646f6d; /* dorandom */
uint64_t c = 0x6c7967656e657261; /* lygenera */
uint64_t d = 0x7465646279746573; /* tedbytes */
uint64_t m = 0;
uint8_t padbyte = len;
m = seed[2] | (uint64_t)seed[3] << 32;
b ^= m;
d ^= m;
m = seed[0] | (uint64_t)seed[1] << 32;
/* a ^= m; is done in loop below */
c ^= m;
/*
* By using the same SipRound code for all iterations, we
* save space, at the expense of some branch prediction. But
* branch prediction is hard because of variable length anyway.
*/
len = len/8 + 3; /* Now number of rounds to perform */
do {
a ^= m;
switch (--len) {
unsigned bytes;
default: /* Full words */
d ^= m = get_unaligned_le64(in);
in += 8;
break;
case 2: /* Final partial word */
/*
* We'd like to do one 64-bit fetch rather than
* mess around with bytes, but reading past the end
* might hit a protection boundary. Fortunately,
* we know that protection boundaries are aligned,
* so we can consider only three cases:
* - The remainder occupies zero words
* - The remainder fits into one word
* - The remainder straddles two words
*/
bytes = padbyte & 7;
if (bytes == 0) {
m = 0;
} else {
unsigned offset = (unsigned)(uintptr_t)in & 7;
if (offset + bytes <= 8) {
m = le64_to_cpup((uint64_t const *)
(in - offset));
m >>= 8*offset;
} else {
m = get_unaligned_le64(in);
}
m &= ((uint64_t)1 << 8*bytes) - 1;
}
/* Could use | or +, but ^ allows associativity */
d ^= m ^= (uint64_t)padbyte << 56;
break;
case 1: /* Beginning of finalization */
m = 0;
c ^= 0xff;
/*FALLTHROUGH*/
case 0: /* Second half of finalization */
break;
}
SIP_ROUND(a, b, c, d);
SIP_ROUND(a, b, c, d);
} while (len);
return a ^ b ^ c ^ d;
}
#undef SIP_ROUND
#undef SIP_MIX
#define HSIP_MIX(a, b, s) ((a) += (b), (b) = rol32(b, s), (b) ^= (a))
/*
* These are the PRELIMINARY rotate constants suggested by
* Jean-Philippe Aumasson. Update to final when available.
*/
#define HSIP_ROUND(a, b, c, d) \
(HSIP_MIX(a, b, 5), HSIP_MIX(c, d, 8), (a) = rol32(a, 16), \
HSIP_MIX(c, b, 7), HSIP_MIX(a, d, 13), (c) = rol32(c, 16))
uint32_t
hsiphash24(char const *in, size_t len, uint32_t const key[2])
{
uint32_t c = key[0];
uint32_t d = key[1];
uint32_t a = 0x6c796765 ^ 0x736f6d65;
uint32_t b = d ^ 0x74656462 ^ 0x646f7261;
uint32_t m = c;
uint8_t padbyte = len;
/*
* By using the same SipRound code for all iterations, we
* save space, at the expense of some branch prediction. But
* branch prediction is hard because of variable length anyway.
*/
len = len/sizeof(m) + 3; /* Now number of rounds to perform */
do {
a ^= m;
switch (--len) {
unsigned bytes;
default: /* Full words */
d ^= m = get_unaligned_le32(in);
in += sizeof(m);
break;
case 2: /* Final partial word */
/*
* We'd like to do one 32-bit fetch rather than
* mess around with bytes, but reading past the end
* might hit a protection boundary. Fortunately,
* we know that protection boundaries are aligned,
* so we can consider only three cases:
* - The remainder occupies zero words
* - The remainder fits into one word
* - The remainder straddles two words
*/
bytes = padbyte & 3;
if (bytes == 0) {
m = 0;
} else {
unsigned offset = (unsigned)(uintptr_t)in & 3;
if (offset + bytes <= 4) {
m = le32_to_cpup((uint32_t const *)
(in - offset));
m >>= 8*offset;
} else {
m = get_unaligned_le32(in);
}
m &= ((uint32_t)1 << 8*bytes) - 1;
}
/* Could use | or +, but ^ allows associativity */
d ^= m ^= (uint32_t)padbyte << 24;
break;
case 1: /* Beginning of finalization */
m = 0;
c ^= 0xff;
/*FALLTHROUGH*/
case 0: /* Second half of finalization */
break;
}
HSIP_ROUND(a, b, c, d);
HSIP_ROUND(a, b, c, d);
} while (len);
return a ^ b ^ c ^ d;
// return c + d;
}
#undef HSIP_ROUND
#undef HSIP_MIX
/*
* No objection to EXPORT_SYMBOL, but we should probably figure out
* how the seed[] array should work first. Homework for the first
* person to want to call it from a module!
*/
#if SELFTEST
#include <stdio.h>
static uint64_t rdtsc()
{
uint32_t eax, edx;
asm volatile ("rdtsc" : "=a" (eax), "=d" (edx));
return (uint64_t)edx << 32 | eax;
}
int
main(void)
{
static char const buf[1024] = { 0 };
unsigned max;
static const uint32_t key[4] = { 1, 2, 3, 4 };
printf("SipHash/HSipHash benchmark, sizeof(long) = %u\n",
(unsigned)sizeof(long));
for (unsigned max = 4; max <= 1024; max *= 2) {
uint64_t sum1 = 0;
uint32_t sum2 = 0;
uint64_t cycles;
uint32_t bytes = 0;
/* A less lazy person could figure out the closed form */
for (int i = 1; i <= max; i++)
bytes += i * (max + 1 - i);
cycles = rdtsc();
for (int i = 1; i <= max; i++)
for (int j = 0; j <= max-i; j++)
sum1 += siphash24(buf+j, i, key);
cycles = rdtsc() - cycles;
printf(" SipHash: max=%4u cycles=%10llu cpb=%8.4f (sum=%llx)\n",
max, cycles, (double)cycles/bytes, sum1);
cycles = rdtsc();
for (int i = 1; i <= max; i++)
for (int j = 0; j <= max-i; j++)
sum2 += hsiphash24(buf+j, i, key);
cycles = rdtsc() - cycles;
printf("HSipHash: max=%4u cycles=%10llu cpb=%8.4f (sum=%lx)\n",
max, cycles, (double)cycles/bytes, sum2);
}
return 0;
}
#endif
On Fri, Dec 16, 2016 at 09:15:03PM -0500, George Spelvin wrote:
> >> - Ted, Andy Lutorminski and I will try to figure out a construction of
> >> get_random_long() that we all like.
We don't have to find the most optimal solution right away; we can
approach this incrementally, after all.
So long as we replace get_random_{long,int}() with something which is
(a) strictly better in terms of security given today's use of MD5, and
(b) which is strictly *faster* than the current construction on 32-bit
and 64-bit systems, we can do that, and can try to make it be faster
while maintaining some minimum level of security which is sufficient
for all current users of get_random_{long,int}() and which can be
clearly artificulated for future users of get_random_{long,int}().
The main worry at this point I have is benchmarking siphash on a
32-bit system. It may be that simply batching the chacha20 output so
that we're using the urandom construction more efficiently is the
better way to go, since that *does* meet the criteron of strictly more
secure and strictly faster than the current MD5 solution. I'm open to
using siphash, but I want to see the the 32-bit numbers first.
As far as half-siphash is concerned, it occurs to me that the main
problem will be those users who need to guarantee that output can't be
guessed over a long period of time. For example, if you have a
long-running process, then the output needs to remain unguessable over
potentially months or years, or else you might be weakening the ASLR
protections. If on the other hand, the hash table or the process will
be going away in a matter of seconds or minutes, the requirements with
respect to cryptographic strength go down significantly.
Now, maybe this doesn't matter that much if we can guarantee (or make
assumptions) that the attacker doesn't have unlimited access the
output stream of get_random_{long,int}(), or if it's being used in an
anti-DOS use case where it ultimately only needs to be harder than
alternate ways of attacking the system.
Rekeying every five minutes doesn't necessarily help the with respect
to ASLR, but it might reduce the amount of the output stream that
would be available to the attacker in order to be able to attack the
get_random_{long,int}() generator, and it also reduces the value of
doing that attack to only compromising the ASLR for those processes
started within that five minute window.
Cheers,
- Ted
P.S. I'm using ASLR as an example use case, above; of course we will
need to make similar eximainations of the other uses of
get_random_{long,int}().
P.P.S. We might also want to think about potentially defining
get_random_{long,int}() to be unambiguously strong, and then creating
a get_weak_random_{long,int}() which on platforms where performance
might be a consideration, it uses a weaker algorithm perhaps with some
kind of rekeying interval.
> As far as half-siphash is concerned, it occurs to me that the main
> problem will be those users who need to guarantee that output can't be
> guessed over a long period of time. For example, if you have a
> long-running process, then the output needs to remain unguessable over
> potentially months or years, or else you might be weakening the ASLR
> protections. If on the other hand, the hash table or the process will
> be going away in a matter of seconds or minutes, the requirements with
> respect to cryptographic strength go down significantly.
Perhaps SipHash-4-8 should be used instead of SipHash-2-4. I believe
SipHash-4-8 is recommended for the security conscious who want to be
more conservative in their security estimates.
SipHash-4-8 does not add much more processing. If you are clocking
SipHash-2-4 at 2.0 or 2.5 cpb, then SipHash-4-8 will run at 3.0 to
4.0. Both are well below MD5 times. (At least with the data sets I've
tested).
> Now, maybe this doesn't matter that much if we can guarantee (or make
> assumptions) that the attacker doesn't have unlimited access the
> output stream of get_random_{long,int}(), or if it's being used in an
> anti-DOS use case where it ultimately only needs to be harder than
> alternate ways of attacking the system.
>
> Rekeying every five minutes doesn't necessarily help the with respect
> to ASLR, but it might reduce the amount of the output stream that
> would be available to the attacker in order to be able to attack the
> get_random_{long,int}() generator, and it also reduces the value of
> doing that attack to only compromising the ASLR for those processes
> started within that five minute window.
Forgive my ignorance... I did not find reading on using the primitive
in a PRNG. Does anyone know what Aumasson or Bernstein have to say?
Aumasson's site does not seem to discuss the use case:
https://www.google.com/search?q=siphash+rng+site%3A131002.net. (And
their paper only mentions random-number once in a different context).
Making the leap from internal hash tables and short-lived network
packets to the rng case may leave something to be desired, especially
if the bits get used in unanticipated ways, like creating long term
private keys.
Jeff
From: George Spelvin
> Sent: 17 December 2016 15:21
...
> uint32_t
> hsiphash24(char const *in, size_t len, uint32_t const key[2])
> {
> uint32_t c = key[0];
> uint32_t d = key[1];
> uint32_t a = 0x6c796765 ^ 0x736f6d65;
> uint32_t b = d ^ 0x74656462 ^ 0x646f7261;
I've not looked closely, but is that (in some sense) duplicating
the key length?
So you could set a = key[2] and b = key[3] and still have an
working hash - albeit not exactly the one specified.
I'll add another comment here...
Is it worth using the 32bit hash for IP addresses on 64bit systems that
can't do misaligned accessed?
David
On Sat, Dec 17, 2016 at 3:55 PM, Jeffrey Walton <[email protected]> wrote:
> It may be prudent to include the endian reversal in the test to ensure
> big endian machines produce expected results. Some closely related
> testing on an old Apple PowerMac G5 revealed that result needed to be
> reversed before returning it to a caller.
The function [1] returns a u64. Originally I had it returning a
__le64, but that was considered unnecessary by many prior reviewers on
the list. It returns an integer. If you want uniform bytes out of it,
then use the endian conversion function, the same as you would do with
any other type of integer.
Additionally, this function is *not* meant for af_alg or any of the
crypto/* code. It's very unlikely to find a use there.
> Forgive my ignorance... I did not find reading on using the primitive
> in a PRNG. Does anyone know what Aumasson or Bernstein have to say?
> Aumasson's site does not seem to discuss the use case:
He's on this thread so I suppose he can speak up for himself. But in
my conversations with him, the primary take-away was, "seems okay to
me!". But please -- JP - correct me if I've misinterpreted.
Yeah you can use the PRF properties to build a DRBG, but that may not be
optimal in terms of performance.
On Mon, 19 Dec 2016 at 18:08, Jason A. Donenfeld <[email protected]> wrote:
> On Sat, Dec 17, 2016 at 3:55 PM, Jeffrey Walton <[email protected]>
> wrote:
> > It may be prudent to include the endian reversal in the test to ensure
> > big endian machines produce expected results. Some closely related
> > testing on an old Apple PowerMac G5 revealed that result needed to be
> > reversed before returning it to a caller.
>
> The function [1] returns a u64. Originally I had it returning a
> __le64, but that was considered unnecessary by many prior reviewers on
> the list. It returns an integer. If you want uniform bytes out of it,
> then use the endian conversion function, the same as you would do with
> any other type of integer.
>
> Additionally, this function is *not* meant for af_alg or any of the
> crypto/* code. It's very unlikely to find a use there.
>
>
> > Forgive my ignorance... I did not find reading on using the primitive
> > in a PRNG. Does anyone know what Aumasson or Bernstein have to say?
> > Aumasson's site does not seem to discuss the use case:
>
> He's on this thread so I suppose he can speak up for himself. But in
> my conversations with him, the primary take-away was, "seems okay to
> me!". But please -- JP - correct me if I've misinterpreted.
>
Hi Ted,
On Sat, Dec 17, 2016 at 4:41 PM, Theodore Ts'o <[email protected]> wrote:
> On Fri, Dec 16, 2016 at 09:15:03PM -0500, George Spelvin wrote:
>> >> - Ted, Andy Lutorminski and I will try to figure out a construction of
>> >> get_random_long() that we all like.
>
> We don't have to find the most optimal solution right away; we can
> approach this incrementally, after all.
Thanks to your call for moderation. This is the impression I have too.
And with all the back and forth of these threads, I fear nothing will
get done. I'm going to collect the best ideas amongst all of these,
and try to get it merged. Then after that we can incrementally improve
on it.
David Miller -- would you merge something into your 4.11 tree? Seems
like you might be the guy for this, since the changes primarily affect
net/*.
Latest patches are here:
https://git.zx2c4.com/linux-dev/log/?h=siphash
> So long as we replace get_random_{long,int}() with something which is
> (a) strictly better in terms of security given today's use of MD5, and
> (b) which is strictly *faster* than the current construction on 32-bit
> and 64-bit systems, we can do that, and can try to make it be faster
> while maintaining some minimum level of security which is sufficient
> for all current users of get_random_{long,int}() and which can be
> clearly artificulated for future users of get_random_{long,int}().
>
> The main worry at this point I have is benchmarking siphash on a
> 32-bit system. It may be that simply batching the chacha20 output so
> that we're using the urandom construction more efficiently is the
> better way to go, since that *does* meet the criteron of strictly more
> secure and strictly faster than the current MD5 solution. I'm open to
> using siphash, but I want to see the the 32-bit numbers first.
Sure, I'll run some benchmarks and report back.
> As far as half-siphash is concerned, it occurs to me that the main
> problem will be those users who need to guarantee that output can't be
> guessed over a long period of time. For example, if you have a
> long-running process, then the output needs to remain unguessable over
> potentially months or years, or else you might be weakening the ASLR
> protections. If on the other hand, the hash table or the process will
> be going away in a matter of seconds or minutes, the requirements with
> respect to cryptographic strength go down significantly.
>
> Now, maybe this doesn't matter that much if we can guarantee (or make
> assumptions) that the attacker doesn't have unlimited access the
> output stream of get_random_{long,int}(), or if it's being used in an
> anti-DOS use case where it ultimately only needs to be harder than
> alternate ways of attacking the system.
The only acceptable usage of HalfSipHash is for hash table lookups
where top security isn't actually a goal. I'm still a bit queasy about
going with it, but George S has very aggressively been pursuing a
"save every last cycle" agenda, which makes sense given how
performance sensitive certain hash tables are, and so I suspect this
could be an okay compromise between performance and security. But,
only for hash tables. Certainly not for the RNG.
>
> Rekeying every five minutes doesn't necessarily help the with respect
> to ASLR, but it might reduce the amount of the output stream that
> would be available to the attacker in order to be able to attack the
> get_random_{long,int}() generator, and it also reduces the value of
> doing that attack to only compromising the ASLR for those processes
> started within that five minute window.
The current implemention of get_random_int/long in my branch uses
128-bit key siphash, so the chances of compromising the key are pretty
much nil. The periodic rekeying is to protect against direct-ram
attacks or other kernel exploits -- a concern brought up by Andy.
With siphash, which takes a 128-bit key, if you got an RNG output once
every picosecond, I believe it would take approximately 10^19 years...
Jason
David Laight wrote:
> From: George Spelvin
...
>> uint32_t
>> hsiphash24(char const *in, size_t len, uint32_t const key[2])
>> {
>> uint32_t c = key[0];
>> uint32_t d = key[1];
>> uint32_t a = 0x6c796765 ^ 0x736f6d65;
>> uint32_t b = d ^ 0x74656462 ^ 0x646f7261;
> I've not looked closely, but is that (in some sense) duplicating
> the key length?
> So you could set a = key[2] and b = key[3] and still have an
> working hash - albeit not exactly the one specified.
That's tempting, but not necessarily effective. (A similar unsuccesful
idea can be found in discussions of "DES with independent round keys".
Or see the design discussion of Salsa20 and the constants in its input.)
You can increase the key size, but that might not increase the *security*
any.
The big issue is that there are a *lot* of square root attack in
cryptanalysis. Because SipHash's state is twice the size of the key,
such an attack will have the same complexity as key exhaustion and need
not be considered. To make a stronger security claim, you need to start
working through them all and show that they don't apply.
For SipHash in particular, an important property is asymmetry of the
internal state. That's what duplicating the key with XORs guarantees.
If the two halves of the state end up identical, the mixing is much
weaker.
Now the probability of ending up in a "mirror state" is the square
root of the state size (1 in 2^64 for HalfSipHash's 128-bit state),
which is the same probability as guessing a key, so it's not a
problem that has to be considered when making a 64-bit security claim.
But if you want a higher security level, you have to think about
what can happen.
That said, I have been thinking very hard about
a = c ^ 0x48536970; /* 'HSip' */
d = key[2];
By guaranteeing that a and c are different, we get the desired
asymmetry, and the XOR of b and d is determined by the first word of
the message anyway, so this isn't weakening anything.
96 bits is far beyond the reach of any brute-force attack, and if a
more sophisticated 64-bit attack exists, it's at least out of the reach
of the script kiddies, and will almost certainly have a non-negligible
constant factor and more limits in when it can be applied.
> Is it worth using the 32bit hash for IP addresses on 64bit systems that
> can't do misaligned accessed?
Not a good idea. To hash 64 bits of input:
* Full SipHash has to do two loads, a shift, an or, and two rounds of mixing.
* HalfSipHash has to do a load, two rounds, another load, and two more rounds.
In other words, in addition to being less secure, it's half the speed.
Also, what systems are you thinking about? x86, ARMv8, PowerPC, and
S390 (and ia64, if anyone cares) all handle unaligned loads. MIPS has
efficient support. Alpha and HPPA are for retrocomputing fans, not
people who care about performance.
So you're down to SPARC. Which conveniently has the same maintainer as
the networking code, so I figure DaveM can take care of that himself. :-)
FTR, I agree that SipHash may benefit from security/performance
optimizations (there's always optimizations), but I'm note sure that it's
the right time/place to discuss it. Cryptanalysis is hard, and any major
change should be backed by a rigorous security analysis and/or security
proof. For example, the redundancy in the initial state prevents trivial
weaknesses that would occur if four key words were used (something I've
seen proposed in this thread).
In the paper at http://aumasson.jp/siphash/siphash.pdf we explain rationale
behind most of our design choices. The cryptanalysis paper at
https://eprint.iacr.org/2014/722.pdf analyzes differences propagations and
presents attacks on reduced versions of SipHash.
On Mon, Dec 19, 2016 at 7:10 PM George Spelvin <[email protected]>
wrote:
> David Laight wrote:
> > From: George Spelvin
> ...
> >> uint32_t
> >> hsiphash24(char const *in, size_t len, uint32_t const key[2])
> >> {
> >> uint32_t c = key[0];
> >> uint32_t d = key[1];
> >> uint32_t a = 0x6c796765 ^ 0x736f6d65;
> >> uint32_t b = d ^ 0x74656462 ^ 0x646f7261;
>
> > I've not looked closely, but is that (in some sense) duplicating
> > the key length?
> > So you could set a = key[2] and b = key[3] and still have an
> > working hash - albeit not exactly the one specified.
>
> That's tempting, but not necessarily effective. (A similar unsuccesful
> idea can be found in discussions of "DES with independent round keys".
> Or see the design discussion of Salsa20 and the constants in its input.)
>
> You can increase the key size, but that might not increase the *security*
> any.
>
> The big issue is that there are a *lot* of square root attack in
> cryptanalysis. Because SipHash's state is twice the size of the key,
> such an attack will have the same complexity as key exhaustion and need
> not be considered. To make a stronger security claim, you need to start
> working through them all and show that they don't apply.
>
> For SipHash in particular, an important property is asymmetry of the
> internal state. That's what duplicating the key with XORs guarantees.
> If the two halves of the state end up identical, the mixing is much
> weaker.
>
> Now the probability of ending up in a "mirror state" is the square
> root of the state size (1 in 2^64 for HalfSipHash's 128-bit state),
> which is the same probability as guessing a key, so it's not a
> problem that has to be considered when making a 64-bit security claim.
>
> But if you want a higher security level, you have to think about
> what can happen.
>
> That said, I have been thinking very hard about
>
> a = c ^ 0x48536970; /* 'HSip' */
> d = key[2];
>
> By guaranteeing that a and c are different, we get the desired
> asymmetry, and the XOR of b and d is determined by the first word of
> the message anyway, so this isn't weakening anything.
>
> 96 bits is far beyond the reach of any brute-force attack, and if a
> more sophisticated 64-bit attack exists, it's at least out of the reach
> of the script kiddies, and will almost certainly have a non-negligible
> constant factor and more limits in when it can be applied.
>
> > Is it worth using the 32bit hash for IP addresses on 64bit systems that
> > can't do misaligned accessed?
>
> Not a good idea. To hash 64 bits of input:
>
> * Full SipHash has to do two loads, a shift, an or, and two rounds of
> mixing.
> * HalfSipHash has to do a load, two rounds, another load, and two more
> rounds.
>
> In other words, in addition to being less secure, it's half the speed.
>
> Also, what systems are you thinking about? x86, ARMv8, PowerPC, and
> S390 (and ia64, if anyone cares) all handle unaligned loads. MIPS has
> efficient support. Alpha and HPPA are for retrocomputing fans, not
> people who care about performance.
>
> So you're down to SPARC. Which conveniently has the same maintainer as
> the networking code, so I figure DaveM can take care of that himself. :-)
>
SipHash is a 64-bit keyed hash function that is actually a
cryptographically secure PRF, like HMAC. Except SipHash is super fast,
and is meant to be used as a hashtable keyed lookup function, or as a
general PRF for short input use cases, such as sequence numbers or RNG
chaining.
For the first usage:
There are a variety of attacks known as "hashtable poisoning" in which an
attacker forms some data such that the hash of that data will be the
same, and then preceeds to fill up all entries of a hashbucket. This is
a realistic and well-known denial-of-service vector. Currently
hashtables use jhash, which is fast but not secure, and some kind of
rotating key scheme (or none at all, which isn't good). SipHash is meant
as a replacement for jhash in these cases.
There are a modicum of places in the kernel that are vulnerable to
hashtable poisoning attacks, either via userspace vectors or network
vectors, and there's not a reliable mechanism inside the kernel at the
moment to fix it. The first step toward fixing these issues is actually
getting a secure primitive into the kernel for developers to use. Then
we can, bit by bit, port things over to it as deemed appropriate.
While SipHash is extremely fast for a cryptographically secure function,
it is likely a bit slower than the insecure jhash, and so replacements
will be evaluated on a case-by-case basis based on whether or not the
difference in speed is negligible and whether or not the current jhash usage
poses a real security risk.
For the second usage:
A few places in the kernel are using MD5 or SHA1 for creating secure
sequence numbers, syn cookies, port numbers, or fast random numbers.
SipHash is a faster and more fitting, and more secure replacement for MD5
in those situations. Replacing MD5 and SHA1 with SipHash for these uses is
obvious and straight-forward, and so is submitted along with this patch
series. There shouldn't be much of a debate over its efficacy.
Dozens of languages are already using this internally for their hash
tables and PRFs. Some of the BSDs already use this in their kernels.
SipHash is a widely known high-speed solution to a widely known set of
problems, and it's time we catch-up.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Jean-Philippe Aumasson <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Eric Biggers <[email protected]>
Cc: David Laight <[email protected]>
Cc: Eric Dumazet <[email protected]>
---
Documentation/siphash.txt | 79 ++++++++++++++++
MAINTAINERS | 7 ++
include/linux/siphash.h | 79 ++++++++++++++++
lib/Kconfig.debug | 6 +-
lib/Makefile | 5 +-
lib/siphash.c | 232 ++++++++++++++++++++++++++++++++++++++++++++++
lib/test_siphash.c | 119 ++++++++++++++++++++++++
7 files changed, 522 insertions(+), 5 deletions(-)
create mode 100644 Documentation/siphash.txt
create mode 100644 include/linux/siphash.h
create mode 100644 lib/siphash.c
create mode 100644 lib/test_siphash.c
diff --git a/Documentation/siphash.txt b/Documentation/siphash.txt
new file mode 100644
index 000000000000..39ff7f0438e7
--- /dev/null
+++ b/Documentation/siphash.txt
@@ -0,0 +1,79 @@
+ SipHash - a short input PRF
+-----------------------------------------------
+Written by Jason A. Donenfeld <[email protected]>
+
+SipHash is a cryptographically secure PRF -- a keyed hash function -- that
+performs very well for short inputs, hence the name. It was designed by
+cryptographers Daniel J. Bernstein and Jean-Philippe Aumasson. It is intended
+as a replacement for some uses of: `jhash`, `md5_transform`, `sha_transform`,
+and so forth.
+
+SipHash takes a secret key filled with randomly generated numbers and either
+an input buffer or several input integers. It spits out an integer that is
+indistinguishable from random. You may then use that integer as part of secure
+sequence numbers, secure cookies, or mask it off for use in a hash table.
+
+1. Generating a key
+
+Keys should always be generated from a cryptographically secure source of
+random numbers, either using get_random_bytes or get_random_once:
+
+siphash_key_t key;
+get_random_bytes(key, sizeof(key));
+
+If you're not deriving your key from here, you're doing it wrong.
+
+2. Using the functions
+
+There are two variants of the function, one that takes a list of integers, and
+one that takes a buffer:
+
+u64 siphash(const void *data, size_t len, siphash_key_t key);
+
+And:
+
+u64 siphash_1u64(u64, siphash_key_t key);
+u64 siphash_2u64(u64, u64, siphash_key_t key);
+u64 siphash_3u64(u64, u64, u64, siphash_key_t key);
+u64 siphash_4u64(u64, u64, u64, u64, siphash_key_t key);
+u64 siphash_1u32(u32, siphash_key_t key);
+u64 siphash_2u32(u32, u32, siphash_key_t key);
+u64 siphash_3u32(u32, u32, u32, siphash_key_t key);
+u64 siphash_4u32(u32, u32, u32, u32, siphash_key_t key);
+
+If you pass the generic siphash function something of a constant length, it
+will constant fold at compile-time and automatically choose one of the
+optimized functions.
+
+3. Hashtable key function usage:
+
+struct some_hashtable {
+ DECLARE_HASHTABLE(hashtable, 8);
+ siphash_key_t key;
+};
+
+void init_hashtable(struct some_hashtable *table)
+{
+ get_random_bytes(table->key, sizeof(table->key));
+}
+
+static inline hlist_head *some_hashtable_bucket(struct some_hashtable *table, struct interesting_input *input)
+{
+ return &table->hashtable[siphash(input, sizeof(*input), table->key) & (HASH_SIZE(table->hashtable) - 1)];
+}
+
+You may then iterate like usual over the returned hash bucket.
+
+4. Security
+
+SipHash has a very high security margin, with its 128-bit key. So long as the
+key is kept secret, it is impossible for an attacker to guess the outputs of
+the function, even if being able to observe many outputs, since 2^128 outputs
+is significant.
+
+Linux implements the "2-4" variant of SipHash.
+
+5. Resources
+
+Read the SipHash paper if you're interested in learning more:
+https://131002.net/siphash/siphash.pdf
diff --git a/MAINTAINERS b/MAINTAINERS
index 59c9895d73d5..5d87a8c1056a 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -11231,6 +11231,13 @@ F: arch/arm/mach-s3c24xx/mach-bast.c
F: arch/arm/mach-s3c24xx/bast-ide.c
F: arch/arm/mach-s3c24xx/bast-irq.c
+SIPHASH PRF ROUTINES
+M: Jason A. Donenfeld <[email protected]>
+S: Maintained
+F: lib/siphash.c
+F: lib/test_siphash.c
+F: include/linux/siphash.h
+
TI DAVINCI MACHINE SUPPORT
M: Sekhar Nori <[email protected]>
M: Kevin Hilman <[email protected]>
diff --git a/include/linux/siphash.h b/include/linux/siphash.h
new file mode 100644
index 000000000000..7aa666eb00d9
--- /dev/null
+++ b/include/linux/siphash.h
@@ -0,0 +1,79 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#ifndef _LINUX_SIPHASH_H
+#define _LINUX_SIPHASH_H
+
+#include <linux/types.h>
+#include <linux/kernel.h>
+
+#define SIPHASH_ALIGNMENT __alignof__(u64)
+typedef u64 siphash_key_t[2];
+
+u64 __siphash_aligned(const void *data, size_t len, const siphash_key_t key);
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+u64 __siphash_unaligned(const void *data, size_t len, const siphash_key_t key);
+#endif
+
+u64 siphash_1u64(const u64 a, const siphash_key_t key);
+u64 siphash_2u64(const u64 a, const u64 b, const siphash_key_t key);
+u64 siphash_3u64(const u64 a, const u64 b, const u64 c,
+ const siphash_key_t key);
+u64 siphash_4u64(const u64 a, const u64 b, const u64 c, const u64 d,
+ const siphash_key_t key);
+u64 siphash_1u32(const u32 a, const siphash_key_t key);
+u64 siphash_3u32(const u32 a, const u32 b, const u32 c, const siphash_key_t key);
+
+static inline u64 siphash_2u32(const u32 a, const u32 b, const siphash_key_t key)
+{
+ return siphash_1u64((u64)b << 32 | a, key);
+}
+static inline u64 siphash_4u32(const u32 a, const u32 b, const u32 c, const u32 d,
+ const siphash_key_t key)
+{
+ return siphash_2u64((u64)b << 32 | a, (u64)d << 32 | c, key);
+}
+
+
+static inline u64 ___siphash_aligned(const __le64 *data, size_t len, const siphash_key_t key)
+{
+ if (__builtin_constant_p(len) && len == 4)
+ return siphash_1u32(le32_to_cpu(data[0]), key);
+ if (__builtin_constant_p(len) && len == 8)
+ return siphash_1u64(le64_to_cpu(data[0]), key);
+ if (__builtin_constant_p(len) && len == 16)
+ return siphash_2u64(le64_to_cpu(data[0]), le64_to_cpu(data[1]),
+ key);
+ if (__builtin_constant_p(len) && len == 24)
+ return siphash_3u64(le64_to_cpu(data[0]), le64_to_cpu(data[1]),
+ le64_to_cpu(data[2]), key);
+ if (__builtin_constant_p(len) && len == 32)
+ return siphash_4u64(le64_to_cpu(data[0]), le64_to_cpu(data[1]),
+ le64_to_cpu(data[2]), le64_to_cpu(data[3]),
+ key);
+ return __siphash_aligned(data, len, key);
+}
+
+/**
+ * siphash - compute 64-bit siphash PRF value
+ * @data: buffer to hash
+ * @size: size of @data
+ * @key: the siphash key
+ */
+static inline u64 siphash(const void *data, size_t len, const siphash_key_t key)
+{
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+ if (!IS_ALIGNED((unsigned long)data, SIPHASH_ALIGNMENT))
+ return __siphash_unaligned(data, len, key);
+#endif
+ return ___siphash_aligned(data, len, key);
+}
+
+#endif /* _LINUX_SIPHASH_H */
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 7446097f72bd..86254ea99b45 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1843,9 +1843,9 @@ config TEST_HASH
tristate "Perform selftest on hash functions"
default n
help
- Enable this option to test the kernel's integer (<linux/hash,h>)
- and string (<linux/stringhash.h>) hash functions on boot
- (or module load).
+ Enable this option to test the kernel's integer (<linux/hash.h>),
+ string (<linux/stringhash.h>), and siphash (<linux/siphash.h>)
+ hash functions on boot (or module load).
This is intended to help people writing architecture-specific
optimized versions. If unsure, say N.
diff --git a/lib/Makefile b/lib/Makefile
index 50144a3aeebd..71d398b04a74 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -22,7 +22,8 @@ lib-y := ctype.o string.o vsprintf.o cmdline.o \
sha1.o chacha20.o md5.o irq_regs.o argv_split.o \
flex_proportions.o ratelimit.o show_mem.o \
is_single_threaded.o plist.o decompress.o kobject_uevent.o \
- earlycpio.o seq_buf.o nmi_backtrace.o nodemask.o win_minmax.o
+ earlycpio.o seq_buf.o siphash.o \
+ nmi_backtrace.o nodemask.o win_minmax.o
lib-$(CONFIG_MMU) += ioremap.o
lib-$(CONFIG_SMP) += cpumask.o
@@ -44,7 +45,7 @@ obj-$(CONFIG_TEST_HEXDUMP) += test_hexdump.o
obj-y += kstrtox.o
obj-$(CONFIG_TEST_BPF) += test_bpf.o
obj-$(CONFIG_TEST_FIRMWARE) += test_firmware.o
-obj-$(CONFIG_TEST_HASH) += test_hash.o
+obj-$(CONFIG_TEST_HASH) += test_hash.o test_siphash.o
obj-$(CONFIG_TEST_KASAN) += test_kasan.o
obj-$(CONFIG_TEST_KSTRTOX) += test-kstrtox.o
obj-$(CONFIG_TEST_LKM) += test_module.o
diff --git a/lib/siphash.c b/lib/siphash.c
new file mode 100644
index 000000000000..ff2151313667
--- /dev/null
+++ b/lib/siphash.c
@@ -0,0 +1,232 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#include <linux/siphash.h>
+#include <asm/unaligned.h>
+
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+#include <linux/dcache.h>
+#include <asm/word-at-a-time.h>
+#endif
+
+#define SIPROUND \
+ do { \
+ v0 += v1; v1 = rol64(v1, 13); v1 ^= v0; v0 = rol64(v0, 32); \
+ v2 += v3; v3 = rol64(v3, 16); v3 ^= v2; \
+ v0 += v3; v3 = rol64(v3, 21); v3 ^= v0; \
+ v2 += v1; v1 = rol64(v1, 17); v1 ^= v2; v2 = rol64(v2, 32); \
+ } while(0)
+
+#define PREAMBLE(len) \
+ u64 v0 = 0x736f6d6570736575ULL; \
+ u64 v1 = 0x646f72616e646f6dULL; \
+ u64 v2 = 0x6c7967656e657261ULL; \
+ u64 v3 = 0x7465646279746573ULL; \
+ u64 b = ((u64)len) << 56; \
+ v3 ^= key[1]; \
+ v2 ^= key[0]; \
+ v1 ^= key[1]; \
+ v0 ^= key[0];
+
+#define POSTAMBLE \
+ v3 ^= b; \
+ SIPROUND; \
+ SIPROUND; \
+ v0 ^= b; \
+ v2 ^= 0xff; \
+ SIPROUND; \
+ SIPROUND; \
+ SIPROUND; \
+ SIPROUND; \
+ return (v0 ^ v1) ^ (v2 ^ v3);
+
+u64 __siphash_aligned(const void *data, size_t len, const siphash_key_t key)
+{
+ const u8 *end = data + len - (len % sizeof(u64));
+ const u8 left = len & (sizeof(u64) - 1);
+ u64 m;
+ PREAMBLE(len)
+ for (; data != end; data += sizeof(u64)) {
+ m = le64_to_cpup(data);
+ v3 ^= m;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= m;
+ }
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+ if (left)
+ b |= le64_to_cpu((__force __le64)(load_unaligned_zeropad(data) &
+ bytemask_from_count(left)));
+#else
+ switch (left) {
+ case 7: b |= ((u64)end[6]) << 48;
+ case 6: b |= ((u64)end[5]) << 40;
+ case 5: b |= ((u64)end[4]) << 32;
+ case 4: b |= le32_to_cpup(data); break;
+ case 3: b |= ((u64)end[2]) << 16;
+ case 2: b |= le16_to_cpup(data); break;
+ case 1: b |= end[0];
+ }
+#endif
+ POSTAMBLE
+}
+EXPORT_SYMBOL(__siphash_aligned);
+
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+u64 __siphash_unaligned(const void *data, size_t len, const siphash_key_t key)
+{
+ const u8 *end = data + len - (len % sizeof(u64));
+ const u8 left = len & (sizeof(u64) - 1);
+ u64 m;
+ PREAMBLE(len)
+ for (; data != end; data += sizeof(u64)) {
+ m = get_unaligned_le64(data);
+ v3 ^= m;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= m;
+ }
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+ if (left)
+ b |= le64_to_cpu((__force __le64)(load_unaligned_zeropad(data) &
+ bytemask_from_count(left)));
+#else
+ switch (left) {
+ case 7: b |= ((u64)end[6]) << 48;
+ case 6: b |= ((u64)end[5]) << 40;
+ case 5: b |= ((u64)end[4]) << 32;
+ case 4: b |= get_unaligned_le32(end); break;
+ case 3: b |= ((u64)end[2]) << 16;
+ case 2: b |= get_unaligned_le16(end); break;
+ case 1: b |= end[0];
+ }
+#endif
+ POSTAMBLE
+}
+EXPORT_SYMBOL(__siphash_unaligned);
+#endif
+
+/**
+ * siphash_1u64 - compute 64-bit siphash PRF value of a u64
+ * @first: first u64
+ * @key: the siphash key
+ */
+u64 siphash_1u64(const u64 first, const siphash_key_t key)
+{
+ PREAMBLE(8)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_1u64);
+
+/**
+ * siphash_2u64 - compute 64-bit siphash PRF value of 2 u64
+ * @first: first u64
+ * @second: second u64
+ * @key: the siphash key
+ */
+u64 siphash_2u64(const u64 first, const u64 second, const siphash_key_t key)
+{
+ PREAMBLE(16)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= second;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_2u64);
+
+/**
+ * siphash_3u64 - compute 64-bit siphash PRF value of 3 u64
+ * @first: first u64
+ * @second: second u64
+ * @third: third u64
+ * @key: the siphash key
+ */
+u64 siphash_3u64(const u64 first, const u64 second, const u64 third,
+ const siphash_key_t key)
+{
+ PREAMBLE(24)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= second;
+ v3 ^= third;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= third;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_3u64);
+
+/**
+ * siphash_4u64 - compute 64-bit siphash PRF value of 4 u64
+ * @first: first u64
+ * @second: second u64
+ * @third: third u64
+ * @forth: forth u64
+ * @key: the siphash key
+ */
+u64 siphash_4u64(const u64 first, const u64 second, const u64 third,
+ const u64 forth, const siphash_key_t key)
+{
+ PREAMBLE(32)
+ v3 ^= first;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= second;
+ v3 ^= third;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= third;
+ v3 ^= forth;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= forth;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_4u64);
+
+u64 siphash_1u32(const u32 first, const siphash_key_t key)
+{
+ PREAMBLE(4)
+ b |= first;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_1u32);
+
+u64 siphash_3u32(const u32 first, const u32 second, const u32 third,
+ const siphash_key_t key)
+{
+ u64 combined = (u64)second << 32 | first;
+ PREAMBLE(12)
+ v3 ^= combined;
+ SIPROUND;
+ SIPROUND;
+ v0 ^= combined;
+ b |= third;
+ POSTAMBLE
+}
+EXPORT_SYMBOL(siphash_3u32);
diff --git a/lib/test_siphash.c b/lib/test_siphash.c
new file mode 100644
index 000000000000..e0ba2cf8dc67
--- /dev/null
+++ b/lib/test_siphash.c
@@ -0,0 +1,119 @@
+/* Test cases for siphash.c
+ *
+ * Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.
+ *
+ * SipHash: a fast short-input PRF
+ * https://131002.net/siphash/
+ *
+ * This implementation is specifically for SipHash2-4.
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/siphash.h>
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/module.h>
+
+/* Test vectors taken from official reference source available at:
+ * https://131002.net/siphash/siphash24.c
+ */
+static const u64 test_vectors[64] = {
+ 0x726fdb47dd0e0e31ULL, 0x74f839c593dc67fdULL, 0x0d6c8009d9a94f5aULL,
+ 0x85676696d7fb7e2dULL, 0xcf2794e0277187b7ULL, 0x18765564cd99a68dULL,
+ 0xcbc9466e58fee3ceULL, 0xab0200f58b01d137ULL, 0x93f5f5799a932462ULL,
+ 0x9e0082df0ba9e4b0ULL, 0x7a5dbbc594ddb9f3ULL, 0xf4b32f46226bada7ULL,
+ 0x751e8fbc860ee5fbULL, 0x14ea5627c0843d90ULL, 0xf723ca908e7af2eeULL,
+ 0xa129ca6149be45e5ULL, 0x3f2acc7f57c29bdbULL, 0x699ae9f52cbe4794ULL,
+ 0x4bc1b3f0968dd39cULL, 0xbb6dc91da77961bdULL, 0xbed65cf21aa2ee98ULL,
+ 0xd0f2cbb02e3b67c7ULL, 0x93536795e3a33e88ULL, 0xa80c038ccd5ccec8ULL,
+ 0xb8ad50c6f649af94ULL, 0xbce192de8a85b8eaULL, 0x17d835b85bbb15f3ULL,
+ 0x2f2e6163076bcfadULL, 0xde4daaaca71dc9a5ULL, 0xa6a2506687956571ULL,
+ 0xad87a3535c49ef28ULL, 0x32d892fad841c342ULL, 0x7127512f72f27cceULL,
+ 0xa7f32346f95978e3ULL, 0x12e0b01abb051238ULL, 0x15e034d40fa197aeULL,
+ 0x314dffbe0815a3b4ULL, 0x027990f029623981ULL, 0xcadcd4e59ef40c4dULL,
+ 0x9abfd8766a33735cULL, 0x0e3ea96b5304a7d0ULL, 0xad0c42d6fc585992ULL,
+ 0x187306c89bc215a9ULL, 0xd4a60abcf3792b95ULL, 0xf935451de4f21df2ULL,
+ 0xa9538f0419755787ULL, 0xdb9acddff56ca510ULL, 0xd06c98cd5c0975ebULL,
+ 0xe612a3cb9ecba951ULL, 0xc766e62cfcadaf96ULL, 0xee64435a9752fe72ULL,
+ 0xa192d576b245165aULL, 0x0a8787bf8ecb74b2ULL, 0x81b3e73d20b49b6fULL,
+ 0x7fa8220ba3b2eceaULL, 0x245731c13ca42499ULL, 0xb78dbfaf3a8d83bdULL,
+ 0xea1ad565322a1a0bULL, 0x60e61c23a3795013ULL, 0x6606d7e446282b93ULL,
+ 0x6ca4ecb15c5f91e1ULL, 0x9f626da15c9625f3ULL, 0xe51b38608ef25f57ULL,
+ 0x958a324ceb064572ULL
+};
+static const siphash_key_t test_key =
+ { 0x0706050403020100ULL , 0x0f0e0d0c0b0a0908ULL };
+
+static int __init siphash_test_init(void)
+{
+ u8 in[64] __aligned(SIPHASH_ALIGNMENT);
+ u8 in_unaligned[65];
+ u8 i;
+ int ret = 0;
+
+ for (i = 0; i < 64; ++i) {
+ in[i] = i;
+ in_unaligned[i + 1] = i;
+ if (siphash(in, i, test_key) != test_vectors[i]) {
+ pr_info("self-test aligned %u: FAIL\n", i + 1);
+ ret = -EINVAL;
+ }
+ if (siphash(in_unaligned + 1, i, test_key) != test_vectors[i]) {
+ pr_info("self-test unaligned %u: FAIL\n", i + 1);
+ ret = -EINVAL;
+ }
+ }
+ if (siphash_1u64(0x0706050403020100ULL, test_key) != test_vectors[8]) {
+ pr_info("self-test 1u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_2u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL, test_key) != test_vectors[16]) {
+ pr_info("self-test 2u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_3u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL,
+ 0x1716151413121110ULL, test_key) != test_vectors[24]) {
+ pr_info("self-test 3u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_4u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL,
+ 0x1716151413121110ULL, 0x1f1e1d1c1b1a1918ULL, test_key) != test_vectors[32]) {
+ pr_info("self-test 4u64: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_1u32(0x03020100U, test_key) != test_vectors[4]) {
+ pr_info("self-test 1u32: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_2u32(0x03020100U, 0x07060504U, test_key) != test_vectors[8]) {
+ pr_info("self-test 2u32: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_3u32(0x03020100U, 0x07060504U,
+ 0x0b0a0908U, test_key) != test_vectors[12]) {
+ pr_info("self-test 3u32: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (siphash_4u32(0x03020100U, 0x07060504U,
+ 0x0b0a0908U, 0x0f0e0d0cU, test_key) != test_vectors[16]) {
+ pr_info("self-test 4u32: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (!ret)
+ pr_info("self-tests: pass\n");
+ return ret;
+}
+
+static void __exit siphash_test_exit(void)
+{
+}
+
+module_init(siphash_test_init);
+module_exit(siphash_test_exit);
+
+MODULE_AUTHOR("Jason A. Donenfeld <[email protected]>");
+MODULE_LICENSE("Dual BSD/GPL");
--
2.11.0
Hey folks,
Again we've made huge progress, with this latest version now shipping
Jean-Phillipe Aumasson's HalfSipHash, which should be much more competitive
with jhash (in addition to being more secure, of course).
There are dozens of little cleanups and improvements right and left throughout
this series, so I urge you to take a look at the whole thing. I've tried to
take into consideration lots of concerns and suggestions from many of you
over the last week.
There is also now documentation! And the test suite now has full coverage of
all functions.
Finally, there's been some significant benchmarking, and now a few commit
messages have some performance numbers.
Please send your Reviewed-by lines as you see fit.
Regards,
Jason
Jason A. Donenfeld (6):
siphash: add cryptographically secure PRF
secure_seq: use SipHash in place of MD5
random: use SipHash in place of MD5
md5: remove from lib and only live in crypto
syncookies: use SipHash in place of SHA1
siphash: implement HalfSipHash1-3 for hash tables
Documentation/siphash.txt | 154 +++++++++++++
MAINTAINERS | 7 +
crypto/md5.c | 95 +++++++-
drivers/char/random.c | 84 ++++---
include/linux/random.h | 1 -
include/linux/siphash.h | 133 +++++++++++
init/main.c | 1 -
lib/Kconfig.debug | 6 +-
lib/Makefile | 7 +-
lib/md5.c | 95 --------
lib/siphash.c | 548 ++++++++++++++++++++++++++++++++++++++++++++++
lib/test_siphash.c | 208 ++++++++++++++++++
net/core/secure_seq.c | 135 +++++-------
net/ipv4/syncookies.c | 20 +-
net/ipv6/syncookies.c | 37 ++--
15 files changed, 1274 insertions(+), 257 deletions(-)
create mode 100644 Documentation/siphash.txt
create mode 100644 include/linux/siphash.h
delete mode 100644 lib/md5.c
create mode 100644 lib/siphash.c
create mode 100644 lib/test_siphash.c
--
2.11.0
This duplicates the current algorithm for get_random_int/long, but uses
siphash instead. This comes with several benefits. It's certainly
faster and more cryptographically secure than MD5. This patch also
separates hashed fields into three values instead of one, in order to
increase diffusion.
The previous MD5 algorithm used a per-cpu MD5 state, which caused
successive calls to the function to chain upon each other. While it's
not entirely clear that this kind of chaining is absolutely necessary
when using a secure PRF like siphash, it can't hurt, and the timing of
the call chain does add a degree of natural entropy. So, in keeping with
this design, instead of the massive per-cpu 64-byte MD5 state, there is
instead a per-cpu previously returned value for chaining.
The speed benefits are substantial:
| siphash | md5 | speedup |
------------------------------
get_random_long | 137130 | 415983 | 3.03x |
get_random_int | 86384 | 343323 | 3.97x |
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Jean-Philippe Aumasson <[email protected]>
Cc: Ted Tso <[email protected]>
---
drivers/char/random.c | 84 +++++++++++++++++++++++++++++---------------------
include/linux/random.h | 1 -
init/main.c | 1 -
3 files changed, 49 insertions(+), 37 deletions(-)
diff --git a/drivers/char/random.c b/drivers/char/random.c
index d6876d506220..ea9858d9d8b4 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -262,6 +262,7 @@
#include <linux/syscalls.h>
#include <linux/completion.h>
#include <linux/uuid.h>
+#include <linux/siphash.h>
#include <crypto/chacha20.h>
#include <asm/processor.h>
@@ -2042,17 +2043,31 @@ struct ctl_table random_table[] = {
};
#endif /* CONFIG_SYSCTL */
-static u32 random_int_secret[MD5_MESSAGE_BYTES / 4] ____cacheline_aligned;
-int random_int_secret_init(void)
+struct random_int_secret {
+ siphash_key_t secret;
+ u64 chaining;
+ unsigned long birthdate;
+ bool initialized;
+};
+static DEFINE_PER_CPU(struct random_int_secret, random_int_secret);
+
+enum {
+ SECRET_ROTATION_TIME = HZ * 60 * 5
+};
+
+static struct random_int_secret *get_random_int_secret(void)
{
- get_random_bytes(random_int_secret, sizeof(random_int_secret));
- return 0;
+ struct random_int_secret *secret = &get_cpu_var(random_int_secret);
+ if (unlikely(!secret->initialized ||
+ !time_is_after_jiffies(secret->birthdate + SECRET_ROTATION_TIME))) {
+ secret->initialized = true;
+ secret->birthdate = jiffies;
+ get_random_bytes(secret->secret, sizeof(secret->secret));
+ }
+ return secret;
}
-static DEFINE_PER_CPU(__u32 [MD5_DIGEST_WORDS], get_random_int_hash)
- __aligned(sizeof(unsigned long));
-
/*
* Get a random word for internal kernel use only. Similar to urandom but
* with the goal of minimal entropy pool depletion. As a result, the random
@@ -2061,20 +2076,20 @@ static DEFINE_PER_CPU(__u32 [MD5_DIGEST_WORDS], get_random_int_hash)
*/
unsigned int get_random_int(void)
{
- __u32 *hash;
- unsigned int ret;
-
- if (arch_get_random_int(&ret))
- return ret;
-
- hash = get_cpu_var(get_random_int_hash);
-
- hash[0] += current->pid + jiffies + random_get_entropy();
- md5_transform(hash, random_int_secret);
- ret = hash[0];
- put_cpu_var(get_random_int_hash);
-
- return ret;
+ unsigned int arch_result;
+ u64 result;
+ struct random_int_secret *secret;
+
+ if (arch_get_random_int(&arch_result))
+ return arch_result;
+
+ secret = get_random_int_secret();
+ result = siphash_3u64(secret->chaining, jiffies,
+ (u64)random_get_entropy() + current->pid,
+ secret->secret);
+ secret->chaining += result;
+ put_cpu_var(secret);
+ return result;
}
EXPORT_SYMBOL(get_random_int);
@@ -2083,20 +2098,19 @@ EXPORT_SYMBOL(get_random_int);
*/
unsigned long get_random_long(void)
{
- __u32 *hash;
- unsigned long ret;
-
- if (arch_get_random_long(&ret))
- return ret;
-
- hash = get_cpu_var(get_random_int_hash);
-
- hash[0] += current->pid + jiffies + random_get_entropy();
- md5_transform(hash, random_int_secret);
- ret = *(unsigned long *)hash;
- put_cpu_var(get_random_int_hash);
-
- return ret;
+ unsigned long arch_result;
+ u64 result;
+ struct random_int_secret *secret;
+
+ if (arch_get_random_long(&arch_result))
+ return arch_result;
+
+ secret = get_random_int_secret();
+ result = siphash_3u64(secret->chaining, jiffies, random_get_entropy() +
+ current->pid, secret->secret);
+ secret->chaining += result;
+ put_cpu_var(secret);
+ return result;
}
EXPORT_SYMBOL(get_random_long);
diff --git a/include/linux/random.h b/include/linux/random.h
index 7bd2403e4fef..16ab429735a7 100644
--- a/include/linux/random.h
+++ b/include/linux/random.h
@@ -37,7 +37,6 @@ extern void get_random_bytes(void *buf, int nbytes);
extern int add_random_ready_callback(struct random_ready_callback *rdy);
extern void del_random_ready_callback(struct random_ready_callback *rdy);
extern void get_random_bytes_arch(void *buf, int nbytes);
-extern int random_int_secret_init(void);
#ifndef MODULE
extern const struct file_operations random_fops, urandom_fops;
diff --git a/init/main.c b/init/main.c
index 23c275cca73a..a3af9296cafd 100644
--- a/init/main.c
+++ b/init/main.c
@@ -879,7 +879,6 @@ static void __init do_basic_setup(void)
do_ctors();
usermodehelper_enable();
do_initcalls();
- random_int_secret_init();
}
static void __init do_pre_smp_initcalls(void)
--
2.11.0
This gives a clear speed and security improvement. Siphash is both
faster and is more solid crypto than the aging MD5.
Rather than manually filling MD5 buffers, for IPv6, we simply create
a layout by a simple anonymous struct, for which gcc generates
rather efficient code. For IPv4, we pass the values directly to the
short input convenience functions.
64-bit x86_64:
[ 1.683628] secure_tcpv6_sequence_number_md5# cycles: 99563527
[ 1.717350] secure_tcp_sequence_number_md5# cycles: 92890502
[ 1.741968] secure_tcpv6_sequence_number_siphash# cycles: 67825362
[ 1.762048] secure_tcp_sequence_number_siphash# cycles: 67485526
32-bit x86:
[ 1.600012] secure_tcpv6_sequence_number_md5# cycles: 103227892
[ 1.634219] secure_tcp_sequence_number_md5# cycles: 94732544
[ 1.669102] secure_tcpv6_sequence_number_siphash# cycles: 96299384
[ 1.700165] secure_tcp_sequence_number_siphash# cycles: 86015473
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: David Miller <[email protected]>
Cc: David Laight <[email protected]>
Cc: Tom Herbert <[email protected]>
Cc: Hannes Frederic Sowa <[email protected]>
Cc: Eric Dumazet <[email protected]>
---
net/core/secure_seq.c | 135 ++++++++++++++++++++------------------------------
1 file changed, 54 insertions(+), 81 deletions(-)
diff --git a/net/core/secure_seq.c b/net/core/secure_seq.c
index 88a8e429fc3e..3dc2689bcc64 100644
--- a/net/core/secure_seq.c
+++ b/net/core/secure_seq.c
@@ -1,3 +1,5 @@
+/* Copyright (C) 2016 Jason A. Donenfeld <[email protected]>. All Rights Reserved. */
+
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/cryptohash.h>
@@ -8,14 +10,14 @@
#include <linux/ktime.h>
#include <linux/string.h>
#include <linux/net.h>
-
+#include <linux/siphash.h>
#include <net/secure_seq.h>
#if IS_ENABLED(CONFIG_IPV6) || IS_ENABLED(CONFIG_INET)
+#include <linux/in6.h>
#include <net/tcp.h>
-#define NET_SECRET_SIZE (MD5_MESSAGE_BYTES / 4)
-static u32 net_secret[NET_SECRET_SIZE] ____cacheline_aligned;
+static siphash_key_t net_secret;
static __always_inline void net_secret_init(void)
{
@@ -44,80 +46,65 @@ static u32 seq_scale(u32 seq)
u32 secure_tcpv6_sequence_number(const __be32 *saddr, const __be32 *daddr,
__be16 sport, __be16 dport, u32 *tsoff)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
- u32 i;
-
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 sport;
+ __be16 dport;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .sport = sport,
+ .dport = dport
+ };
+ u64 hash;
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32)daddr[i];
- secret[4] = net_secret[4] +
- (((__force u16)sport << 16) + (__force u16)dport);
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- *tsoff = sysctl_tcp_timestamps == 1 ? hash[1] : 0;
- return seq_scale(hash[0]);
+ hash = siphash(&combined, offsetofend(typeof(combined), dport), net_secret);
+ *tsoff = sysctl_tcp_timestamps == 1 ? (hash >> 32) : 0;
+ return seq_scale(hash);
}
EXPORT_SYMBOL(secure_tcpv6_sequence_number);
u32 secure_ipv6_port_ephemeral(const __be32 *saddr, const __be32 *daddr,
__be16 dport)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
- u32 i;
-
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 dport;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .dport = dport
+ };
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32) daddr[i];
- secret[4] = net_secret[4] + (__force u32)dport;
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- return hash[0];
+ return siphash(&combined, offsetofend(typeof(combined), dport), net_secret);
}
EXPORT_SYMBOL(secure_ipv6_port_ephemeral);
#endif
#ifdef CONFIG_INET
+/* secure_tcp_sequence_number(a, b, 0, d) == secure_ipv4_port_ephemeral(a, b, d),
+ * but fortunately, `sport' cannot be 0 in any circumstances. If this changes,
+ * it would be easy enough to have the former function use siphash_4u32, passing
+ * the arguments as separate u32.
+ */
+
u32 secure_tcp_sequence_number(__be32 saddr, __be32 daddr,
__be16 sport, __be16 dport, u32 *tsoff)
{
- u32 hash[MD5_DIGEST_WORDS];
-
+ u64 hash;
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = ((__force u16)sport << 16) + (__force u16)dport;
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- *tsoff = sysctl_tcp_timestamps == 1 ? hash[1] : 0;
- return seq_scale(hash[0]);
+ hash = siphash_3u32(saddr, daddr, (u32)sport << 16 | dport, net_secret);
+ *tsoff = sysctl_tcp_timestamps == 1 ? (hash >> 32) : 0;
+ return seq_scale(hash);
}
u32 secure_ipv4_port_ephemeral(__be32 saddr, __be32 daddr, __be16 dport)
{
- u32 hash[MD5_DIGEST_WORDS];
-
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = (__force u32)dport ^ net_secret[14];
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- return hash[0];
+ return siphash_3u32(saddr, daddr, dport, net_secret);
}
EXPORT_SYMBOL_GPL(secure_ipv4_port_ephemeral);
#endif
@@ -126,21 +113,11 @@ EXPORT_SYMBOL_GPL(secure_ipv4_port_ephemeral);
u64 secure_dccp_sequence_number(__be32 saddr, __be32 daddr,
__be16 sport, __be16 dport)
{
- u32 hash[MD5_DIGEST_WORDS];
u64 seq;
-
net_secret_init();
- hash[0] = (__force u32)saddr;
- hash[1] = (__force u32)daddr;
- hash[2] = ((__force u16)sport << 16) + (__force u16)dport;
- hash[3] = net_secret[15];
-
- md5_transform(hash, net_secret);
-
- seq = hash[0] | (((u64)hash[1]) << 32);
+ seq = siphash_3u32(saddr, daddr, (u32)sport << 16 | dport, net_secret);
seq += ktime_get_real_ns();
seq &= (1ull << 48) - 1;
-
return seq;
}
EXPORT_SYMBOL(secure_dccp_sequence_number);
@@ -149,26 +126,22 @@ EXPORT_SYMBOL(secure_dccp_sequence_number);
u64 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr,
__be16 sport, __be16 dport)
{
- u32 secret[MD5_MESSAGE_BYTES / 4];
- u32 hash[MD5_DIGEST_WORDS];
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ __be16 sport;
+ __be16 dport;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *(struct in6_addr *)saddr,
+ .daddr = *(struct in6_addr *)daddr,
+ .sport = sport,
+ .dport = dport
+ };
u64 seq;
- u32 i;
-
net_secret_init();
- memcpy(hash, saddr, 16);
- for (i = 0; i < 4; i++)
- secret[i] = net_secret[i] + (__force u32)daddr[i];
- secret[4] = net_secret[4] +
- (((__force u16)sport << 16) + (__force u16)dport);
- for (i = 5; i < MD5_MESSAGE_BYTES / 4; i++)
- secret[i] = net_secret[i];
-
- md5_transform(hash, secret);
-
- seq = hash[0] | (((u64)hash[1]) << 32);
+ seq = siphash(&combined, offsetofend(typeof(combined), dport), net_secret);
seq += ktime_get_real_ns();
seq &= (1ull << 48) - 1;
-
return seq;
}
EXPORT_SYMBOL(secure_dccpv6_sequence_number);
--
2.11.0
The md5_transform function is no longer used any where in the tree,
except for the crypto api's actual implementation of md5, so we can drop
the function from lib and put it as a static function of the crypto
file, where it belongs. There should be no new users of md5_transform,
anyway, since there are more modern ways of doing what it once achieved.
Signed-off-by: Jason A. Donenfeld <[email protected]>
---
crypto/md5.c | 95 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
lib/Makefile | 2 +-
lib/md5.c | 95 ------------------------------------------------------------
3 files changed, 95 insertions(+), 97 deletions(-)
delete mode 100644 lib/md5.c
diff --git a/crypto/md5.c b/crypto/md5.c
index 2355a7c25c45..f7ae1a48225b 100644
--- a/crypto/md5.c
+++ b/crypto/md5.c
@@ -21,9 +21,11 @@
#include <linux/module.h>
#include <linux/string.h>
#include <linux/types.h>
-#include <linux/cryptohash.h>
#include <asm/byteorder.h>
+#define MD5_DIGEST_WORDS 4
+#define MD5_MESSAGE_BYTES 64
+
const u8 md5_zero_message_hash[MD5_DIGEST_SIZE] = {
0xd4, 0x1d, 0x8c, 0xd9, 0x8f, 0x00, 0xb2, 0x04,
0xe9, 0x80, 0x09, 0x98, 0xec, 0xf8, 0x42, 0x7e,
@@ -47,6 +49,97 @@ static inline void cpu_to_le32_array(u32 *buf, unsigned int words)
}
}
+#define F1(x, y, z) (z ^ (x & (y ^ z)))
+#define F2(x, y, z) F1(z, x, y)
+#define F3(x, y, z) (x ^ y ^ z)
+#define F4(x, y, z) (y ^ (x | ~z))
+
+#define MD5STEP(f, w, x, y, z, in, s) \
+ (w += f(x, y, z) + in, w = (w<<s | w>>(32-s)) + x)
+
+static void md5_transform(__u32 *hash, __u32 const *in)
+{
+ u32 a, b, c, d;
+
+ a = hash[0];
+ b = hash[1];
+ c = hash[2];
+ d = hash[3];
+
+ MD5STEP(F1, a, b, c, d, in[0] + 0xd76aa478, 7);
+ MD5STEP(F1, d, a, b, c, in[1] + 0xe8c7b756, 12);
+ MD5STEP(F1, c, d, a, b, in[2] + 0x242070db, 17);
+ MD5STEP(F1, b, c, d, a, in[3] + 0xc1bdceee, 22);
+ MD5STEP(F1, a, b, c, d, in[4] + 0xf57c0faf, 7);
+ MD5STEP(F1, d, a, b, c, in[5] + 0x4787c62a, 12);
+ MD5STEP(F1, c, d, a, b, in[6] + 0xa8304613, 17);
+ MD5STEP(F1, b, c, d, a, in[7] + 0xfd469501, 22);
+ MD5STEP(F1, a, b, c, d, in[8] + 0x698098d8, 7);
+ MD5STEP(F1, d, a, b, c, in[9] + 0x8b44f7af, 12);
+ MD5STEP(F1, c, d, a, b, in[10] + 0xffff5bb1, 17);
+ MD5STEP(F1, b, c, d, a, in[11] + 0x895cd7be, 22);
+ MD5STEP(F1, a, b, c, d, in[12] + 0x6b901122, 7);
+ MD5STEP(F1, d, a, b, c, in[13] + 0xfd987193, 12);
+ MD5STEP(F1, c, d, a, b, in[14] + 0xa679438e, 17);
+ MD5STEP(F1, b, c, d, a, in[15] + 0x49b40821, 22);
+
+ MD5STEP(F2, a, b, c, d, in[1] + 0xf61e2562, 5);
+ MD5STEP(F2, d, a, b, c, in[6] + 0xc040b340, 9);
+ MD5STEP(F2, c, d, a, b, in[11] + 0x265e5a51, 14);
+ MD5STEP(F2, b, c, d, a, in[0] + 0xe9b6c7aa, 20);
+ MD5STEP(F2, a, b, c, d, in[5] + 0xd62f105d, 5);
+ MD5STEP(F2, d, a, b, c, in[10] + 0x02441453, 9);
+ MD5STEP(F2, c, d, a, b, in[15] + 0xd8a1e681, 14);
+ MD5STEP(F2, b, c, d, a, in[4] + 0xe7d3fbc8, 20);
+ MD5STEP(F2, a, b, c, d, in[9] + 0x21e1cde6, 5);
+ MD5STEP(F2, d, a, b, c, in[14] + 0xc33707d6, 9);
+ MD5STEP(F2, c, d, a, b, in[3] + 0xf4d50d87, 14);
+ MD5STEP(F2, b, c, d, a, in[8] + 0x455a14ed, 20);
+ MD5STEP(F2, a, b, c, d, in[13] + 0xa9e3e905, 5);
+ MD5STEP(F2, d, a, b, c, in[2] + 0xfcefa3f8, 9);
+ MD5STEP(F2, c, d, a, b, in[7] + 0x676f02d9, 14);
+ MD5STEP(F2, b, c, d, a, in[12] + 0x8d2a4c8a, 20);
+
+ MD5STEP(F3, a, b, c, d, in[5] + 0xfffa3942, 4);
+ MD5STEP(F3, d, a, b, c, in[8] + 0x8771f681, 11);
+ MD5STEP(F3, c, d, a, b, in[11] + 0x6d9d6122, 16);
+ MD5STEP(F3, b, c, d, a, in[14] + 0xfde5380c, 23);
+ MD5STEP(F3, a, b, c, d, in[1] + 0xa4beea44, 4);
+ MD5STEP(F3, d, a, b, c, in[4] + 0x4bdecfa9, 11);
+ MD5STEP(F3, c, d, a, b, in[7] + 0xf6bb4b60, 16);
+ MD5STEP(F3, b, c, d, a, in[10] + 0xbebfbc70, 23);
+ MD5STEP(F3, a, b, c, d, in[13] + 0x289b7ec6, 4);
+ MD5STEP(F3, d, a, b, c, in[0] + 0xeaa127fa, 11);
+ MD5STEP(F3, c, d, a, b, in[3] + 0xd4ef3085, 16);
+ MD5STEP(F3, b, c, d, a, in[6] + 0x04881d05, 23);
+ MD5STEP(F3, a, b, c, d, in[9] + 0xd9d4d039, 4);
+ MD5STEP(F3, d, a, b, c, in[12] + 0xe6db99e5, 11);
+ MD5STEP(F3, c, d, a, b, in[15] + 0x1fa27cf8, 16);
+ MD5STEP(F3, b, c, d, a, in[2] + 0xc4ac5665, 23);
+
+ MD5STEP(F4, a, b, c, d, in[0] + 0xf4292244, 6);
+ MD5STEP(F4, d, a, b, c, in[7] + 0x432aff97, 10);
+ MD5STEP(F4, c, d, a, b, in[14] + 0xab9423a7, 15);
+ MD5STEP(F4, b, c, d, a, in[5] + 0xfc93a039, 21);
+ MD5STEP(F4, a, b, c, d, in[12] + 0x655b59c3, 6);
+ MD5STEP(F4, d, a, b, c, in[3] + 0x8f0ccc92, 10);
+ MD5STEP(F4, c, d, a, b, in[10] + 0xffeff47d, 15);
+ MD5STEP(F4, b, c, d, a, in[1] + 0x85845dd1, 21);
+ MD5STEP(F4, a, b, c, d, in[8] + 0x6fa87e4f, 6);
+ MD5STEP(F4, d, a, b, c, in[15] + 0xfe2ce6e0, 10);
+ MD5STEP(F4, c, d, a, b, in[6] + 0xa3014314, 15);
+ MD5STEP(F4, b, c, d, a, in[13] + 0x4e0811a1, 21);
+ MD5STEP(F4, a, b, c, d, in[4] + 0xf7537e82, 6);
+ MD5STEP(F4, d, a, b, c, in[11] + 0xbd3af235, 10);
+ MD5STEP(F4, c, d, a, b, in[2] + 0x2ad7d2bb, 15);
+ MD5STEP(F4, b, c, d, a, in[9] + 0xeb86d391, 21);
+
+ hash[0] += a;
+ hash[1] += b;
+ hash[2] += c;
+ hash[3] += d;
+}
+
static inline void md5_transform_helper(struct md5_state *ctx)
{
le32_to_cpu_array(ctx->block, sizeof(ctx->block) / sizeof(u32));
diff --git a/lib/Makefile b/lib/Makefile
index 71d398b04a74..1079152607e0 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -19,7 +19,7 @@ KCOV_INSTRUMENT_dynamic_debug.o := n
lib-y := ctype.o string.o vsprintf.o cmdline.o \
rbtree.o radix-tree.o dump_stack.o timerqueue.o\
idr.o int_sqrt.o extable.o \
- sha1.o chacha20.o md5.o irq_regs.o argv_split.o \
+ sha1.o chacha20.o irq_regs.o argv_split.o \
flex_proportions.o ratelimit.o show_mem.o \
is_single_threaded.o plist.o decompress.o kobject_uevent.o \
earlycpio.o seq_buf.o siphash.o \
diff --git a/lib/md5.c b/lib/md5.c
deleted file mode 100644
index bb0cd01d356d..000000000000
--- a/lib/md5.c
+++ /dev/null
@@ -1,95 +0,0 @@
-#include <linux/compiler.h>
-#include <linux/export.h>
-#include <linux/cryptohash.h>
-
-#define F1(x, y, z) (z ^ (x & (y ^ z)))
-#define F2(x, y, z) F1(z, x, y)
-#define F3(x, y, z) (x ^ y ^ z)
-#define F4(x, y, z) (y ^ (x | ~z))
-
-#define MD5STEP(f, w, x, y, z, in, s) \
- (w += f(x, y, z) + in, w = (w<<s | w>>(32-s)) + x)
-
-void md5_transform(__u32 *hash, __u32 const *in)
-{
- u32 a, b, c, d;
-
- a = hash[0];
- b = hash[1];
- c = hash[2];
- d = hash[3];
-
- MD5STEP(F1, a, b, c, d, in[0] + 0xd76aa478, 7);
- MD5STEP(F1, d, a, b, c, in[1] + 0xe8c7b756, 12);
- MD5STEP(F1, c, d, a, b, in[2] + 0x242070db, 17);
- MD5STEP(F1, b, c, d, a, in[3] + 0xc1bdceee, 22);
- MD5STEP(F1, a, b, c, d, in[4] + 0xf57c0faf, 7);
- MD5STEP(F1, d, a, b, c, in[5] + 0x4787c62a, 12);
- MD5STEP(F1, c, d, a, b, in[6] + 0xa8304613, 17);
- MD5STEP(F1, b, c, d, a, in[7] + 0xfd469501, 22);
- MD5STEP(F1, a, b, c, d, in[8] + 0x698098d8, 7);
- MD5STEP(F1, d, a, b, c, in[9] + 0x8b44f7af, 12);
- MD5STEP(F1, c, d, a, b, in[10] + 0xffff5bb1, 17);
- MD5STEP(F1, b, c, d, a, in[11] + 0x895cd7be, 22);
- MD5STEP(F1, a, b, c, d, in[12] + 0x6b901122, 7);
- MD5STEP(F1, d, a, b, c, in[13] + 0xfd987193, 12);
- MD5STEP(F1, c, d, a, b, in[14] + 0xa679438e, 17);
- MD5STEP(F1, b, c, d, a, in[15] + 0x49b40821, 22);
-
- MD5STEP(F2, a, b, c, d, in[1] + 0xf61e2562, 5);
- MD5STEP(F2, d, a, b, c, in[6] + 0xc040b340, 9);
- MD5STEP(F2, c, d, a, b, in[11] + 0x265e5a51, 14);
- MD5STEP(F2, b, c, d, a, in[0] + 0xe9b6c7aa, 20);
- MD5STEP(F2, a, b, c, d, in[5] + 0xd62f105d, 5);
- MD5STEP(F2, d, a, b, c, in[10] + 0x02441453, 9);
- MD5STEP(F2, c, d, a, b, in[15] + 0xd8a1e681, 14);
- MD5STEP(F2, b, c, d, a, in[4] + 0xe7d3fbc8, 20);
- MD5STEP(F2, a, b, c, d, in[9] + 0x21e1cde6, 5);
- MD5STEP(F2, d, a, b, c, in[14] + 0xc33707d6, 9);
- MD5STEP(F2, c, d, a, b, in[3] + 0xf4d50d87, 14);
- MD5STEP(F2, b, c, d, a, in[8] + 0x455a14ed, 20);
- MD5STEP(F2, a, b, c, d, in[13] + 0xa9e3e905, 5);
- MD5STEP(F2, d, a, b, c, in[2] + 0xfcefa3f8, 9);
- MD5STEP(F2, c, d, a, b, in[7] + 0x676f02d9, 14);
- MD5STEP(F2, b, c, d, a, in[12] + 0x8d2a4c8a, 20);
-
- MD5STEP(F3, a, b, c, d, in[5] + 0xfffa3942, 4);
- MD5STEP(F3, d, a, b, c, in[8] + 0x8771f681, 11);
- MD5STEP(F3, c, d, a, b, in[11] + 0x6d9d6122, 16);
- MD5STEP(F3, b, c, d, a, in[14] + 0xfde5380c, 23);
- MD5STEP(F3, a, b, c, d, in[1] + 0xa4beea44, 4);
- MD5STEP(F3, d, a, b, c, in[4] + 0x4bdecfa9, 11);
- MD5STEP(F3, c, d, a, b, in[7] + 0xf6bb4b60, 16);
- MD5STEP(F3, b, c, d, a, in[10] + 0xbebfbc70, 23);
- MD5STEP(F3, a, b, c, d, in[13] + 0x289b7ec6, 4);
- MD5STEP(F3, d, a, b, c, in[0] + 0xeaa127fa, 11);
- MD5STEP(F3, c, d, a, b, in[3] + 0xd4ef3085, 16);
- MD5STEP(F3, b, c, d, a, in[6] + 0x04881d05, 23);
- MD5STEP(F3, a, b, c, d, in[9] + 0xd9d4d039, 4);
- MD5STEP(F3, d, a, b, c, in[12] + 0xe6db99e5, 11);
- MD5STEP(F3, c, d, a, b, in[15] + 0x1fa27cf8, 16);
- MD5STEP(F3, b, c, d, a, in[2] + 0xc4ac5665, 23);
-
- MD5STEP(F4, a, b, c, d, in[0] + 0xf4292244, 6);
- MD5STEP(F4, d, a, b, c, in[7] + 0x432aff97, 10);
- MD5STEP(F4, c, d, a, b, in[14] + 0xab9423a7, 15);
- MD5STEP(F4, b, c, d, a, in[5] + 0xfc93a039, 21);
- MD5STEP(F4, a, b, c, d, in[12] + 0x655b59c3, 6);
- MD5STEP(F4, d, a, b, c, in[3] + 0x8f0ccc92, 10);
- MD5STEP(F4, c, d, a, b, in[10] + 0xffeff47d, 15);
- MD5STEP(F4, b, c, d, a, in[1] + 0x85845dd1, 21);
- MD5STEP(F4, a, b, c, d, in[8] + 0x6fa87e4f, 6);
- MD5STEP(F4, d, a, b, c, in[15] + 0xfe2ce6e0, 10);
- MD5STEP(F4, c, d, a, b, in[6] + 0xa3014314, 15);
- MD5STEP(F4, b, c, d, a, in[13] + 0x4e0811a1, 21);
- MD5STEP(F4, a, b, c, d, in[4] + 0xf7537e82, 6);
- MD5STEP(F4, d, a, b, c, in[11] + 0xbd3af235, 10);
- MD5STEP(F4, c, d, a, b, in[2] + 0x2ad7d2bb, 15);
- MD5STEP(F4, b, c, d, a, in[9] + 0xeb86d391, 21);
-
- hash[0] += a;
- hash[1] += b;
- hash[2] += c;
- hash[3] += d;
-}
-EXPORT_SYMBOL(md5_transform);
--
2.11.0
SHA1 is slower and less secure than SipHash, and so replacing syncookie
generation with SipHash makes natural sense. Some BSDs have been doing
this for several years in fact.
The speedup should be similar -- and even more impressive -- to the
speedup from the sequence number fix in this series.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Eric Dumazet <[email protected]>
Cc: David Miller <[email protected]>
---
net/ipv4/syncookies.c | 20 ++++----------------
net/ipv6/syncookies.c | 37 ++++++++++++++++---------------------
2 files changed, 20 insertions(+), 37 deletions(-)
diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c
index 3e88467d70ee..03bb068f8888 100644
--- a/net/ipv4/syncookies.c
+++ b/net/ipv4/syncookies.c
@@ -13,13 +13,13 @@
#include <linux/tcp.h>
#include <linux/slab.h>
#include <linux/random.h>
-#include <linux/cryptohash.h>
+#include <linux/siphash.h>
#include <linux/kernel.h>
#include <linux/export.h>
#include <net/tcp.h>
#include <net/route.h>
-static u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS] __read_mostly;
+static siphash_key_t syncookie_secret[2] __read_mostly;
#define COOKIEBITS 24 /* Upper bits store count */
#define COOKIEMASK (((__u32)1 << COOKIEBITS) - 1)
@@ -48,24 +48,12 @@ static u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS] __read_mostly;
#define TSBITS 6
#define TSMASK (((__u32)1 << TSBITS) - 1)
-static DEFINE_PER_CPU(__u32 [16 + 5 + SHA_WORKSPACE_WORDS], ipv4_cookie_scratch);
-
static u32 cookie_hash(__be32 saddr, __be32 daddr, __be16 sport, __be16 dport,
u32 count, int c)
{
- __u32 *tmp;
-
net_get_random_once(syncookie_secret, sizeof(syncookie_secret));
-
- tmp = this_cpu_ptr(ipv4_cookie_scratch);
- memcpy(tmp + 4, syncookie_secret[c], sizeof(syncookie_secret[c]));
- tmp[0] = (__force u32)saddr;
- tmp[1] = (__force u32)daddr;
- tmp[2] = ((__force u32)sport << 16) + (__force u32)dport;
- tmp[3] = count;
- sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
-
- return tmp[17];
+ return siphash_4u32(saddr, daddr, (u32)sport << 16 | dport, count,
+ syncookie_secret[c]);
}
diff --git a/net/ipv6/syncookies.c b/net/ipv6/syncookies.c
index a4d49760bf43..be51fc0d99ad 100644
--- a/net/ipv6/syncookies.c
+++ b/net/ipv6/syncookies.c
@@ -16,7 +16,7 @@
#include <linux/tcp.h>
#include <linux/random.h>
-#include <linux/cryptohash.h>
+#include <linux/siphash.h>
#include <linux/kernel.h>
#include <net/ipv6.h>
#include <net/tcp.h>
@@ -24,7 +24,7 @@
#define COOKIEBITS 24 /* Upper bits store count */
#define COOKIEMASK (((__u32)1 << COOKIEBITS) - 1)
-static u32 syncookie6_secret[2][16-4+SHA_DIGEST_WORDS] __read_mostly;
+static siphash_key_t syncookie6_secret[2] __read_mostly;
/* RFC 2460, Section 8.3:
* [ipv6 tcp] MSS must be computed as the maximum packet size minus 60 [..]
@@ -41,30 +41,25 @@ static __u16 const msstab[] = {
9000 - 60,
};
-static DEFINE_PER_CPU(__u32 [16 + 5 + SHA_WORKSPACE_WORDS], ipv6_cookie_scratch);
-
static u32 cookie_hash(const struct in6_addr *saddr, const struct in6_addr *daddr,
__be16 sport, __be16 dport, u32 count, int c)
{
- __u32 *tmp;
+ const struct {
+ struct in6_addr saddr;
+ struct in6_addr daddr;
+ u32 count;
+ u16 sport;
+ u16 dport;
+ } __aligned(SIPHASH_ALIGNMENT) combined = {
+ .saddr = *saddr,
+ .daddr = *daddr,
+ .count = count,
+ .sport = sport,
+ .dport = dport
+ };
net_get_random_once(syncookie6_secret, sizeof(syncookie6_secret));
-
- tmp = this_cpu_ptr(ipv6_cookie_scratch);
-
- /*
- * we have 320 bits of information to hash, copy in the remaining
- * 192 bits required for sha_transform, from the syncookie6_secret
- * and overwrite the digest with the secret
- */
- memcpy(tmp + 10, syncookie6_secret[c], 44);
- memcpy(tmp, saddr, 16);
- memcpy(tmp + 4, daddr, 16);
- tmp[8] = ((__force u32)sport << 16) + (__force u32)dport;
- tmp[9] = count;
- sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
-
- return tmp[17];
+ return siphash(&combined, offsetofend(typeof(combined), dport), syncookie6_secret[c]);
}
static __u32 secure_tcp_syn_cookie(const struct in6_addr *saddr,
--
2.11.0
HalfSipHash, or hsiphash, is a shortened version of SipHash, which
generates 32-bit outputs using a weaker 64-bit key. It has *much* lower
security margins, and shouldn't be used for anything too sensitive, but
it could be used as a hashtable key function replacement, if the output
is never exposed, and if the security requirement is not too high.
The goal is to make this something that performance-critical jhash users
would be willing to use.
On 64-bit machines, HalfSipHash1-3 is slower than SipHash1-3, so we alias
SipHash1-3 to HalfSipHash1-3 on those systems.
64-bit x86_64:
[ 0.509409] test_siphash: SipHash2-4 cycles: 4049181
[ 0.510650] test_siphash: SipHash1-3 cycles: 2512884
[ 0.512205] test_siphash: HalfSipHash1-3 cycles: 3429920
[ 0.512904] test_siphash: JenkinsHash cycles: 978267
So, we map hsiphash() -> SipHash1-3
32-bit x86:
[ 0.509868] test_siphash: SipHash2-4 cycles: 14812892
[ 0.513601] test_siphash: SipHash1-3 cycles: 9510710
[ 0.515263] test_siphash: HalfSipHash1-3 cycles: 3856157
[ 0.515952] test_siphash: JenkinsHash cycles: 1148567
So, we map hsiphash() -> HalfSipHash1-3
hsiphash() is roughly 3 times slower than jhash(), but comes with a
considerable security improvement.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Cc: Jean-Philippe Aumasson <[email protected]>
---
Documentation/siphash.txt | 75 +++++++++++
include/linux/siphash.h | 56 +++++++-
lib/siphash.c | 318 +++++++++++++++++++++++++++++++++++++++++++++-
lib/test_siphash.c | 139 ++++++++++++++++----
4 files changed, 561 insertions(+), 27 deletions(-)
diff --git a/Documentation/siphash.txt b/Documentation/siphash.txt
index 39ff7f0438e7..f93c1d7104c4 100644
--- a/Documentation/siphash.txt
+++ b/Documentation/siphash.txt
@@ -77,3 +77,78 @@ Linux implements the "2-4" variant of SipHash.
Read the SipHash paper if you're interested in learning more:
https://131002.net/siphash/siphash.pdf
+
+
+~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~
+
+HalfSipHash - SipHash's insecure younger cousin
+-----------------------------------------------
+Written by Jason A. Donenfeld <[email protected]>
+
+On the off-chance that SipHash is not fast enough for your needs, you might be
+able to justify using HalfSipHash, a terrifying but potentially useful
+possibility. HalfSipHash cuts SipHash's rounds down from "2-4" to "1-3" and,
+even scarier, uses an easily brute-forcable 64-bit key (with a 32-bit output)
+instead of SipHash's 128-bit key. However, this may appeal to some
+high-performance `jhash` users.
+
+Danger!
+
+Do not ever use HalfSipHash except for as a hashtable key function, and only
+then when you can be absolutely certain that the outputs will never be
+transmitted out of the kernel. This is only remotely useful over `jhash` as a
+means of mitigating hashtable flooding denial of service attacks.
+
+1. Generating a key
+
+Keys should always be generated from a cryptographically secure source of
+random numbers, either using get_random_bytes or get_random_once:
+
+hsiphash_key_t key;
+get_random_bytes(key, sizeof(key));
+
+If you're not deriving your key from here, you're doing it wrong.
+
+2. Using the functions
+
+There are two variants of the function, one that takes a list of integers, and
+one that takes a buffer:
+
+u32 hsiphash(const void *data, size_t len, siphash_key_t key);
+
+And:
+
+u32 hsiphash_1u32(u32, hsiphash_key_t key);
+u32 hsiphash_2u32(u32, u32, hsiphash_key_t key);
+u32 hsiphash_3u32(u32, u32, u32, hsiphash_key_t key);
+u32 hsiphash_4u32(u32, u32, u32, u32, hsiphash_key_t key);
+
+If you pass the generic hsiphash function something of a constant length, it
+will constant fold at compile-time and automatically choose one of the
+optimized functions.
+
+3. Hashtable key function usage:
+
+struct some_hashtable {
+ DECLARE_HASHTABLE(hashtable, 8);
+ hsiphash_key_t key;
+};
+
+void init_hashtable(struct some_hashtable *table)
+{
+ get_random_bytes(table->key, sizeof(table->key));
+}
+
+static inline hlist_head *some_hashtable_bucket(struct some_hashtable *table, struct interesting_input *input)
+{
+ return &table->hashtable[hsiphash(input, sizeof(*input), table->key) & (HASH_SIZE(table->hashtable) - 1)];
+}
+
+You may then iterate like usual over the returned hash bucket.
+
+4. Performance
+
+HalfSipHash is roughly 3 times slower than JenkinsHash. For many replacements,
+this will not be a problem, as the hashtable lookup isn't the bottleneck. And
+in general, this is probably a good sacrifice to make for the security and DoS
+resistance of HalfSipHash.
diff --git a/include/linux/siphash.h b/include/linux/siphash.h
index 7aa666eb00d9..efab44c654f3 100644
--- a/include/linux/siphash.h
+++ b/include/linux/siphash.h
@@ -5,7 +5,9 @@
* SipHash: a fast short-input PRF
* https://131002.net/siphash/
*
- * This implementation is specifically for SipHash2-4.
+ * This implementation is specifically for SipHash2-4 for a secure PRF
+ * and HalfSipHash1-3/SipHash1-3 for an insecure PRF only suitable for
+ * hashtables.
*/
#ifndef _LINUX_SIPHASH_H
@@ -76,4 +78,56 @@ static inline u64 siphash(const void *data, size_t len, const siphash_key_t key)
return ___siphash_aligned(data, len, key);
}
+#if BITS_PER_LONG == 64
+typedef siphash_key_t hsiphash_key_t;
+#define HSIPHASH_ALIGNMENT SIPHASH_ALIGNMENT
+#else
+typedef u32 hsiphash_key_t[2];
+#define HSIPHASH_ALIGNMENT __alignof__(u32)
+#endif
+
+u32 __hsiphash_aligned(const void *data, size_t len, const hsiphash_key_t key);
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+u32 __hsiphash_unaligned(const void *data, size_t len, const hsiphash_key_t key);
+#endif
+
+u32 hsiphash_1u32(const u32 a, const hsiphash_key_t key);
+u32 hsiphash_2u32(const u32 a, const u32 b, const hsiphash_key_t key);
+u32 hsiphash_3u32(const u32 a, const u32 b, const u32 c,
+ const hsiphash_key_t key);
+u32 hsiphash_4u32(const u32 a, const u32 b, const u32 c, const u32 d,
+ const hsiphash_key_t key);
+
+static inline u32 ___hsiphash_aligned(const __le32 *data, size_t len, const hsiphash_key_t key)
+{
+ if (__builtin_constant_p(len) && len == 4)
+ return hsiphash_1u32(le32_to_cpu(data[0]), key);
+ if (__builtin_constant_p(len) && len == 8)
+ return hsiphash_2u32(le32_to_cpu(data[0]), le32_to_cpu(data[1]),
+ key);
+ if (__builtin_constant_p(len) && len == 12)
+ return hsiphash_3u32(le32_to_cpu(data[0]), le32_to_cpu(data[1]),
+ le32_to_cpu(data[2]), key);
+ if (__builtin_constant_p(len) && len == 16)
+ return hsiphash_4u32(le32_to_cpu(data[0]), le32_to_cpu(data[1]),
+ le32_to_cpu(data[2]), le32_to_cpu(data[3]),
+ key);
+ return __hsiphash_aligned(data, len, key);
+}
+
+/**
+ * hsiphash - compute 32-bit hsiphash PRF value
+ * @data: buffer to hash
+ * @size: size of @data
+ * @key: the hsiphash key
+ */
+static inline u32 hsiphash(const void *data, size_t len, const hsiphash_key_t key)
+{
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+ if (!IS_ALIGNED((unsigned long)data, HSIPHASH_ALIGNMENT))
+ return __hsiphash_unaligned(data, len, key);
+#endif
+ return ___hsiphash_aligned(data, len, key);
+}
+
#endif /* _LINUX_SIPHASH_H */
diff --git a/lib/siphash.c b/lib/siphash.c
index ff2151313667..e2481226d96c 100644
--- a/lib/siphash.c
+++ b/lib/siphash.c
@@ -5,7 +5,9 @@
* SipHash: a fast short-input PRF
* https://131002.net/siphash/
*
- * This implementation is specifically for SipHash2-4.
+ * This implementation is specifically for SipHash2-4 for a secure PRF
+ * and HalfSipHash1-3/SipHash1-3 for an insecure PRF only suitable for
+ * hashtables.
*/
#include <linux/siphash.h>
@@ -230,3 +232,317 @@ u64 siphash_3u32(const u32 first, const u32 second, const u32 third,
POSTAMBLE
}
EXPORT_SYMBOL(siphash_3u32);
+
+#if BITS_PER_LONG == 64
+/* Note that this HalfSipHash1-3 implementation on 64-bit
+ * isn't actually HalfSipHash1-3 but rather SipHash1-3. */
+
+#define HSIPROUND SIPROUND
+#define HPREAMBLE(len) PREAMBLE(len)
+#define HPOSTAMBLE \
+ v3 ^= b; \
+ HSIPROUND; \
+ v0 ^= b; \
+ v2 ^= 0xff; \
+ HSIPROUND; \
+ HSIPROUND; \
+ HSIPROUND; \
+ return (v0 ^ v1) ^ (v2 ^ v3);
+
+u32 __hsiphash_aligned(const void *data, size_t len, const hsiphash_key_t key)
+{
+ const u8 *end = data + len - (len % sizeof(u64));
+ const u8 left = len & (sizeof(u64) - 1);
+ u64 m;
+ HPREAMBLE(len)
+ for (; data != end; data += sizeof(u64)) {
+ m = le64_to_cpup(data);
+ v3 ^= m;
+ HSIPROUND;
+ v0 ^= m;
+ }
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+ if (left)
+ b |= le64_to_cpu((__force __le64)(load_unaligned_zeropad(data) &
+ bytemask_from_count(left)));
+#else
+ switch (left) {
+ case 7: b |= ((u64)end[6]) << 48;
+ case 6: b |= ((u64)end[5]) << 40;
+ case 5: b |= ((u64)end[4]) << 32;
+ case 4: b |= le32_to_cpup(data); break;
+ case 3: b |= ((u64)end[2]) << 16;
+ case 2: b |= le16_to_cpup(data); break;
+ case 1: b |= end[0];
+ }
+#endif
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(__hsiphash_aligned);
+
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+u32 __hsiphash_unaligned(const void *data, size_t len, const hsiphash_key_t key)
+{
+ const u8 *end = data + len - (len % sizeof(u64));
+ const u8 left = len & (sizeof(u64) - 1);
+ u64 m;
+ HPREAMBLE(len)
+ for (; data != end; data += sizeof(u64)) {
+ m = get_unaligned_le64(data);
+ v3 ^= m;
+ HSIPROUND;
+ v0 ^= m;
+ }
+#if defined(CONFIG_DCACHE_WORD_ACCESS) && BITS_PER_LONG == 64
+ if (left)
+ b |= le64_to_cpu((__force __le64)(load_unaligned_zeropad(data) &
+ bytemask_from_count(left)));
+#else
+ switch (left) {
+ case 7: b |= ((u64)end[6]) << 48;
+ case 6: b |= ((u64)end[5]) << 40;
+ case 5: b |= ((u64)end[4]) << 32;
+ case 4: b |= get_unaligned_le32(end); break;
+ case 3: b |= ((u64)end[2]) << 16;
+ case 2: b |= get_unaligned_le16(end); break;
+ case 1: b |= end[0];
+ }
+#endif
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(__hsiphash_unaligned);
+#endif
+
+/**
+ * hsiphash_1u32 - compute 64-bit hsiphash PRF value of a u32
+ * @first: first u32
+ * @key: the hsiphash key
+ */
+u32 hsiphash_1u32(const u32 first, const hsiphash_key_t key)
+{
+ HPREAMBLE(4)
+ b |= first;
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(hsiphash_1u32);
+
+/**
+ * hsiphash_2u32 - compute 32-bit hsiphash PRF value of 2 u32
+ * @first: first u32
+ * @second: second u32
+ * @key: the hsiphash key
+ */
+u32 hsiphash_2u32(const u32 first, const u32 second, const hsiphash_key_t key)
+{
+ u64 combined = (u64)second << 32 | first;
+ HPREAMBLE(8)
+ v3 ^= combined;
+ HSIPROUND;
+ v0 ^= combined;
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(hsiphash_2u32);
+
+/**
+ * hsiphash_3u32 - compute 32-bit hsiphash PRF value of 3 u32
+ * @first: first u32
+ * @second: second u32
+ * @third: third u32
+ * @key: the hsiphash key
+ */
+u32 hsiphash_3u32(const u32 first, const u32 second, const u32 third,
+ const hsiphash_key_t key)
+{
+ u64 combined = (u64)second << 32 | first;
+ HPREAMBLE(12)
+ v3 ^= combined;
+ HSIPROUND;
+ v0 ^= combined;
+ b |= third;
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(hsiphash_3u32);
+
+/**
+ * hsiphash_4u32 - compute 32-bit hsiphash PRF value of 4 u32
+ * @first: first u32
+ * @second: second u32
+ * @third: third u32
+ * @forth: forth u32
+ * @key: the hsiphash key
+ */
+u32 hsiphash_4u32(const u32 first, const u32 second, const u32 third,
+ const u32 forth, const hsiphash_key_t key)
+{
+ u64 combined = (u64)second << 32 | first;
+ HPREAMBLE(16)
+ v3 ^= combined;
+ HSIPROUND;
+ v0 ^= combined;
+ combined = (u64)forth << 32 | third;
+ v3 ^= combined;
+ HSIPROUND;
+ v0 ^= combined;
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(hsiphash_4u32);
+#else
+#define HSIPROUND \
+ do { \
+ v0 += v1; v1 = rol32(v1, 5); v1 ^= v0; v0 = rol32(v0, 16); \
+ v2 += v3; v3 = rol32(v3, 8); v3 ^= v2; \
+ v0 += v3; v3 = rol32(v3, 7); v3 ^= v0; \
+ v2 += v1; v1 = rol32(v1, 13); v1 ^= v2; v2 = rol32(v2, 16); \
+ } while(0)
+
+#define HPREAMBLE(len) \
+ u32 v0 = 0; \
+ u32 v1 = 0; \
+ u32 v2 = 0x6c796765U; \
+ u32 v3 = 0x74656462U; \
+ u32 b = ((u32)len) << 24; \
+ v3 ^= key[1]; \
+ v2 ^= key[0]; \
+ v1 ^= key[1]; \
+ v0 ^= key[0];
+
+#define HPOSTAMBLE \
+ v3 ^= b; \
+ HSIPROUND; \
+ v0 ^= b; \
+ v2 ^= 0xff; \
+ HSIPROUND; \
+ HSIPROUND; \
+ HSIPROUND; \
+ return v1 ^ v3;
+
+u32 __hsiphash_aligned(const void *data, size_t len, const hsiphash_key_t key)
+{
+ const u8 *end = data + len - (len % sizeof(u32));
+ const u8 left = len & (sizeof(u32) - 1);
+ u32 m;
+ HPREAMBLE(len)
+ for (; data != end; data += sizeof(u32)) {
+ m = le32_to_cpup(data);
+ v3 ^= m;
+ HSIPROUND;
+ v0 ^= m;
+ }
+ switch (left) {
+ case 3: b |= ((u32)end[2]) << 16;
+ case 2: b |= le16_to_cpup(data); break;
+ case 1: b |= end[0];
+ }
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(__hsiphash_aligned);
+
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
+u32 __hsiphash_unaligned(const void *data, size_t len, const hsiphash_key_t key)
+{
+ const u8 *end = data + len - (len % sizeof(u32));
+ const u8 left = len & (sizeof(u32) - 1);
+ u32 m;
+ HPREAMBLE(len)
+ for (; data != end; data += sizeof(u32)) {
+ m = get_unaligned_le32(data);
+ v3 ^= m;
+ HSIPROUND;
+ v0 ^= m;
+ }
+ switch (left) {
+ case 3: b |= ((u32)end[2]) << 16;
+ case 2: b |= get_unaligned_le16(end); break;
+ case 1: b |= end[0];
+ }
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(__hsiphash_unaligned);
+#endif
+
+/**
+ * hsiphash_1u32 - compute 32-bit hsiphash PRF value of a u32
+ * @first: first u32
+ * @key: the hsiphash key
+ */
+u32 hsiphash_1u32(const u32 first, const hsiphash_key_t key)
+{
+ HPREAMBLE(4)
+ v3 ^= first;
+ HSIPROUND;
+ v0 ^= first;
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(hsiphash_1u32);
+
+/**
+ * hsiphash_2u32 - compute 32-bit hsiphash PRF value of 2 u32
+ * @first: first u32
+ * @second: second u32
+ * @key: the hsiphash key
+ */
+u32 hsiphash_2u32(const u32 first, const u32 second, const hsiphash_key_t key)
+{
+ HPREAMBLE(8)
+ v3 ^= first;
+ HSIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ HSIPROUND;
+ v0 ^= second;
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(hsiphash_2u32);
+
+/**
+ * hsiphash_3u32 - compute 32-bit hsiphash PRF value of 3 u32
+ * @first: first u32
+ * @second: second u32
+ * @third: third u32
+ * @key: the hsiphash key
+ */
+u32 hsiphash_3u32(const u32 first, const u32 second, const u32 third,
+ const hsiphash_key_t key)
+{
+ HPREAMBLE(12)
+ v3 ^= first;
+ HSIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ HSIPROUND;
+ v0 ^= second;
+ v3 ^= third;
+ HSIPROUND;
+ v0 ^= third;
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(hsiphash_3u32);
+
+/**
+ * hsiphash_4u32 - compute 32-bit hsiphash PRF value of 4 u32
+ * @first: first u32
+ * @second: second u32
+ * @third: third u32
+ * @forth: forth u32
+ * @key: the hsiphash key
+ */
+u32 hsiphash_4u32(const u32 first, const u32 second, const u32 third,
+ const u32 forth, const hsiphash_key_t key)
+{
+ HPREAMBLE(16)
+ v3 ^= first;
+ HSIPROUND;
+ v0 ^= first;
+ v3 ^= second;
+ HSIPROUND;
+ v0 ^= second;
+ v3 ^= third;
+ HSIPROUND;
+ v0 ^= third;
+ v3 ^= forth;
+ HSIPROUND;
+ v0 ^= forth;
+ HPOSTAMBLE
+}
+EXPORT_SYMBOL(hsiphash_4u32);
+#endif
diff --git a/lib/test_siphash.c b/lib/test_siphash.c
index e0ba2cf8dc67..ac291ec27fb6 100644
--- a/lib/test_siphash.c
+++ b/lib/test_siphash.c
@@ -7,7 +7,9 @@
* SipHash: a fast short-input PRF
* https://131002.net/siphash/
*
- * This implementation is specifically for SipHash2-4.
+ * This implementation is specifically for SipHash2-4 for a secure PRF
+ * and HalfSipHash1-3/SipHash1-3 for an insecure PRF only suitable for
+ * hashtables.
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
@@ -18,10 +20,16 @@
#include <linux/errno.h>
#include <linux/module.h>
-/* Test vectors taken from official reference source available at:
- * https://131002.net/siphash/siphash24.c
+/* Test vectors taken from reference source available at:
+ * https://github.com/veorq/SipHash
*/
-static const u64 test_vectors[64] = {
+
+
+
+static const siphash_key_t test_key_siphash =
+ { 0x0706050403020100ULL , 0x0f0e0d0c0b0a0908ULL };
+
+static const u64 test_vectors_siphash[64] = {
0x726fdb47dd0e0e31ULL, 0x74f839c593dc67fdULL, 0x0d6c8009d9a94f5aULL,
0x85676696d7fb7e2dULL, 0xcf2794e0277187b7ULL, 0x18765564cd99a68dULL,
0xcbc9466e58fee3ceULL, 0xab0200f58b01d137ULL, 0x93f5f5799a932462ULL,
@@ -45,9 +53,64 @@ static const u64 test_vectors[64] = {
0x6ca4ecb15c5f91e1ULL, 0x9f626da15c9625f3ULL, 0xe51b38608ef25f57ULL,
0x958a324ceb064572ULL
};
-static const siphash_key_t test_key =
+#if BITS_PER_LONG == 64
+static const hsiphash_key_t test_key_hsiphash =
{ 0x0706050403020100ULL , 0x0f0e0d0c0b0a0908ULL };
+static const u32 test_vectors_hsiphash[64] = {
+ 0x050fc4dcU, 0x7d57ca93U, 0x4dc7d44dU,
+ 0xe7ddf7fbU, 0x88d38328U, 0x49533b67U,
+ 0xc59f22a7U, 0x9bb11140U, 0x8d299a8eU,
+ 0x6c063de4U, 0x92ff097fU, 0xf94dc352U,
+ 0x57b4d9a2U, 0x1229ffa7U, 0xc0f95d34U,
+ 0x2a519956U, 0x7d908b66U, 0x63dbd80cU,
+ 0xb473e63eU, 0x8d297d1cU, 0xa6cce040U,
+ 0x2b45f844U, 0xa320872eU, 0xdae6c123U,
+ 0x67349c8cU, 0x705b0979U, 0xca9913a5U,
+ 0x4ade3b35U, 0xef6cd00dU, 0x4ab1e1f4U,
+ 0x43c5e663U, 0x8c21d1bcU, 0x16a7b60dU,
+ 0x7a8ff9bfU, 0x1f2a753eU, 0xbf186b91U,
+ 0xada26206U, 0xa3c33057U, 0xae3a36a1U,
+ 0x7b108392U, 0x99e41531U, 0x3f1ad944U,
+ 0xc8138825U, 0xc28949a6U, 0xfaf8876bU,
+ 0x9f042196U, 0x68b1d623U, 0x8b5114fdU,
+ 0xdf074c46U, 0x12cc86b3U, 0x0a52098fU,
+ 0x9d292f9aU, 0xa2f41f12U, 0x43a71ed0U,
+ 0x73f0bce6U, 0x70a7e980U, 0x243c6d75U,
+ 0xfdb71513U, 0xa67d8a08U, 0xb7e8f148U,
+ 0xf7a644eeU, 0x0f1837f2U, 0x4b6694e0U,
+ 0xb7bbb3a8U
+};
+#else
+static const hsiphash_key_t test_key_hsiphash =
+ { 0x03020100U, 0x07060504U };
+
+static const u32 test_vectors_hsiphash[64] = {
+ 0x5814c896U, 0xe7e864caU, 0xbc4b0e30U,
+ 0x01539939U, 0x7e059ea6U, 0x88e3d89bU,
+ 0xa0080b65U, 0x9d38d9d6U, 0x577999b1U,
+ 0xc839caedU, 0xe4fa32cfU, 0x959246eeU,
+ 0x6b28096cU, 0x66dd9cd6U, 0x16658a7cU,
+ 0xd0257b04U, 0x8b31d501U, 0x2b1cd04bU,
+ 0x06712339U, 0x522aca67U, 0x911bb605U,
+ 0x90a65f0eU, 0xf826ef7bU, 0x62512debU,
+ 0x57150ad7U, 0x5d473507U, 0x1ec47442U,
+ 0xab64afd3U, 0x0a4100d0U, 0x6d2ce652U,
+ 0x2331b6a3U, 0x08d8791aU, 0xbc6dda8dU,
+ 0xe0f6c934U, 0xb0652033U, 0x9b9851ccU,
+ 0x7c46fb7fU, 0x732ba8cbU, 0xf142997aU,
+ 0xfcc9aa1bU, 0x05327eb2U, 0xe110131cU,
+ 0xf9e5e7c0U, 0xa7d708a6U, 0x11795ab1U,
+ 0x65671619U, 0x9f5fff91U, 0xd89c5267U,
+ 0x007783ebU, 0x95766243U, 0xab639262U,
+ 0x9c7e1390U, 0xc368dda6U, 0x38ddc455U,
+ 0xfa13d379U, 0x979ea4e8U, 0x53ecd77eU,
+ 0x2ee80657U, 0x33dbb66aU, 0xae3f0577U,
+ 0x88b4c4ccU, 0x3e7f480bU, 0x74c1ebf8U,
+ 0x87178304U
+};
+#endif
+
static int __init siphash_test_init(void)
{
u8 in[64] __aligned(SIPHASH_ALIGNMENT);
@@ -58,49 +121,75 @@ static int __init siphash_test_init(void)
for (i = 0; i < 64; ++i) {
in[i] = i;
in_unaligned[i + 1] = i;
- if (siphash(in, i, test_key) != test_vectors[i]) {
- pr_info("self-test aligned %u: FAIL\n", i + 1);
+ if (siphash(in, i, test_key_siphash) != test_vectors_siphash[i]) {
+ pr_info("siphash self-test aligned %u: FAIL\n", i + 1);
+ ret = -EINVAL;
+ }
+ if (siphash(in_unaligned + 1, i, test_key_siphash) != test_vectors_siphash[i]) {
+ pr_info("siphash self-test unaligned %u: FAIL\n", i + 1);
ret = -EINVAL;
}
- if (siphash(in_unaligned + 1, i, test_key) != test_vectors[i]) {
- pr_info("self-test unaligned %u: FAIL\n", i + 1);
+ if (hsiphash(in, i, test_key_hsiphash) != test_vectors_hsiphash[i]) {
+ pr_info("hsiphash self-test aligned %u: FAIL\n", i + 1);
+ ret = -EINVAL;
+ }
+ if (hsiphash(in_unaligned + 1, i, test_key_hsiphash) != test_vectors_hsiphash[i]) {
+ pr_info("hsiphash self-test unaligned %u: FAIL\n", i + 1);
ret = -EINVAL;
}
}
- if (siphash_1u64(0x0706050403020100ULL, test_key) != test_vectors[8]) {
- pr_info("self-test 1u64: FAIL\n");
+ if (siphash_1u64(0x0706050403020100ULL, test_key_siphash) != test_vectors_siphash[8]) {
+ pr_info("siphash self-test 1u64: FAIL\n");
ret = -EINVAL;
}
- if (siphash_2u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL, test_key) != test_vectors[16]) {
- pr_info("self-test 2u64: FAIL\n");
+ if (siphash_2u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL, test_key_siphash) != test_vectors_siphash[16]) {
+ pr_info("siphash self-test 2u64: FAIL\n");
ret = -EINVAL;
}
if (siphash_3u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL,
- 0x1716151413121110ULL, test_key) != test_vectors[24]) {
- pr_info("self-test 3u64: FAIL\n");
+ 0x1716151413121110ULL, test_key_siphash) != test_vectors_siphash[24]) {
+ pr_info("siphash self-test 3u64: FAIL\n");
ret = -EINVAL;
}
if (siphash_4u64(0x0706050403020100ULL, 0x0f0e0d0c0b0a0908ULL,
- 0x1716151413121110ULL, 0x1f1e1d1c1b1a1918ULL, test_key) != test_vectors[32]) {
- pr_info("self-test 4u64: FAIL\n");
+ 0x1716151413121110ULL, 0x1f1e1d1c1b1a1918ULL, test_key_siphash) != test_vectors_siphash[32]) {
+ pr_info("siphash self-test 4u64: FAIL\n");
ret = -EINVAL;
}
- if (siphash_1u32(0x03020100U, test_key) != test_vectors[4]) {
- pr_info("self-test 1u32: FAIL\n");
+ if (siphash_1u32(0x03020100U, test_key_siphash) != test_vectors_siphash[4]) {
+ pr_info("siphash self-test 1u32: FAIL\n");
ret = -EINVAL;
}
- if (siphash_2u32(0x03020100U, 0x07060504U, test_key) != test_vectors[8]) {
- pr_info("self-test 2u32: FAIL\n");
+ if (siphash_2u32(0x03020100U, 0x07060504U, test_key_siphash) != test_vectors_siphash[8]) {
+ pr_info("siphash self-test 2u32: FAIL\n");
ret = -EINVAL;
}
if (siphash_3u32(0x03020100U, 0x07060504U,
- 0x0b0a0908U, test_key) != test_vectors[12]) {
- pr_info("self-test 3u32: FAIL\n");
+ 0x0b0a0908U, test_key_siphash) != test_vectors_siphash[12]) {
+ pr_info("siphash self-test 3u32: FAIL\n");
ret = -EINVAL;
}
if (siphash_4u32(0x03020100U, 0x07060504U,
- 0x0b0a0908U, 0x0f0e0d0cU, test_key) != test_vectors[16]) {
- pr_info("self-test 4u32: FAIL\n");
+ 0x0b0a0908U, 0x0f0e0d0cU, test_key_siphash) != test_vectors_siphash[16]) {
+ pr_info("siphash self-test 4u32: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (hsiphash_1u32(0x03020100U, test_key_hsiphash) != test_vectors_hsiphash[4]) {
+ pr_info("hsiphash self-test 1u32: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (hsiphash_2u32(0x03020100U, 0x07060504U, test_key_hsiphash) != test_vectors_hsiphash[8]) {
+ pr_info("hsiphash self-test 2u32: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (hsiphash_3u32(0x03020100U, 0x07060504U,
+ 0x0b0a0908U, test_key_hsiphash) != test_vectors_hsiphash[12]) {
+ pr_info("hsiphash self-test 3u32: FAIL\n");
+ ret = -EINVAL;
+ }
+ if (hsiphash_4u32(0x03020100U, 0x07060504U,
+ 0x0b0a0908U, 0x0f0e0d0cU, test_key_hsiphash) != test_vectors_hsiphash[16]) {
+ pr_info("hsiphash self-test 4u32: FAIL\n");
ret = -EINVAL;
}
if (!ret)
--
2.11.0
Hi Ted,
On Thu, Dec 22, 2016 at 12:02 AM, Jason A. Donenfeld <[email protected]> wrote:
> This duplicates the current algorithm for get_random_int/long
I should have mentioned this directly in the commit message, which I
forgot to update: this v7 adds the time-based key rotation, which,
while not strictly necessary for ensuring the security of the RNG,
might help alleviate some concerns, as we talked about. Performance is
quite good on both 32-bit and 64-bit -- better than MD5 in both cases.
If you like this, terrific. If not, I'm happy to take this in whatever
direction you prefer, and implement whatever construction you think
best. There's been a lot of noise on this list about it; we can
continue to discuss more, or you can just tell me whatever you want to
do, and I'll implement it and that'll be the end of it. As you said,
we can always get something decent now and improve it later.
Alternatively, if you've decided in the end you prefer your batched
entropy approach using chacha, I'm happy to implement a polished
version of that here in this patch series (so that we can keep the `rm
lib/md5.c` commit.)
Just let me know how you'd like to proceed.
Thanks,
Jason
On Wed, Dec 21, 2016 at 3:02 PM, Jason A. Donenfeld <[email protected]> wrote:
> unsigned int get_random_int(void)
> {
> - __u32 *hash;
> - unsigned int ret;
> -
> - if (arch_get_random_int(&ret))
> - return ret;
> -
> - hash = get_cpu_var(get_random_int_hash);
> -
> - hash[0] += current->pid + jiffies + random_get_entropy();
> - md5_transform(hash, random_int_secret);
> - ret = hash[0];
> - put_cpu_var(get_random_int_hash);
> -
> - return ret;
> + unsigned int arch_result;
> + u64 result;
> + struct random_int_secret *secret;
> +
> + if (arch_get_random_int(&arch_result))
> + return arch_result;
> +
> + secret = get_random_int_secret();
> + result = siphash_3u64(secret->chaining, jiffies,
> + (u64)random_get_entropy() + current->pid,
> + secret->secret);
> + secret->chaining += result;
> + put_cpu_var(secret);
> + return result;
> }
> EXPORT_SYMBOL(get_random_int);
Hmm. I haven't tried to prove anything for real. But here goes (in
the random oracle model):
Suppose I'm an attacker and I don't know the secret or the chaining
value. Then, regardless of what the entropy is, I can't predict the
numbers.
Now suppose I do know the secret and the chaining value due to some
leak. If I want to deduce prior outputs, I think I'm stuck: I'd need
to find a value "result" such that prev_chaining + result = chaining
and result = H(prev_chaining, ..., secret);. I don't think this can
be done efficiently in the random oracle model regardless of what the
"..." is.
But, if I know the secret and chaining value, I can predict the next
output assuming I can guess the entropy. What's worse is that, even
if I can't guess the entropy, if I *observe* the next output then I
can calculate the next chaining value.
So this is probably good enough, and making it better is hard. Changing it to:
u64 entropy = (u64)random_get_entropy() + current->pid;
result = siphash(..., entropy, ...);
secret->chaining += result + entropy;
would reduce this problem by forcing an attacker to brute-force the
entropy on each iteration, which is probably an improvement.
To fully fix it, something like "catastrophic reseeding" would be
needed, but that's hard to get right.
(An aside: on x86 at least, using two percpu variables is faster
because directly percpu access is essentially free, whereas getting
the address of a percpu variable is not free.)
> 64-bit x86_64:
> [ 0.509409] test_siphash: SipHash2-4 cycles: 4049181
> [ 0.510650] test_siphash: SipHash1-3 cycles: 2512884
> [ 0.512205] test_siphash: HalfSipHash1-3 cycles: 3429920
> [ 0.512904] test_siphash: JenkinsHash cycles: 978267
I'm not sure what these numbers mean. Surely a single siphash2-4
does not take 4+ million cycles?
If you run them in a loop please divide by the iterations.
But generally running small code in a loop is often an unrealistic
benchmark strategy because it hides cache misses, primes
predictors, changes frequencies and changes memory costs,
but also can overload pipelines and oversubscribe
resources.
[see also page 46+ in http://halobates.de/applicative-mental-models.pdf]
So the numbers you get there are at least somewhat
dubious. It would be good to have at least some test which
is not just a tiny micro benchmark to compare before making
conclusions.
-Andi
On Thu, 22 Dec 2016 00:02:11 +0100
"Jason A. Donenfeld" <[email protected]> wrote:
> SipHash is a 64-bit keyed hash function that is actually a
> cryptographically secure PRF, like HMAC. Except SipHash is super fast,
> and is meant to be used as a hashtable keyed lookup function, or as a
> general PRF for short input use cases, such as sequence numbers or RNG
> chaining.
>
> For the first usage:
>
> There are a variety of attacks known as "hashtable poisoning" in which an
> attacker forms some data such that the hash of that data will be the
> same, and then preceeds to fill up all entries of a hashbucket. This is
> a realistic and well-known denial-of-service vector. Currently
> hashtables use jhash, which is fast but not secure, and some kind of
> rotating key scheme (or none at all, which isn't good). SipHash is meant
> as a replacement for jhash in these cases.
>
> There are a modicum of places in the kernel that are vulnerable to
> hashtable poisoning attacks, either via userspace vectors or network
> vectors, and there's not a reliable mechanism inside the kernel at the
> moment to fix it. The first step toward fixing these issues is actually
> getting a secure primitive into the kernel for developers to use. Then
> we can, bit by bit, port things over to it as deemed appropriate.
>
> While SipHash is extremely fast for a cryptographically secure function,
> it is likely a bit slower than the insecure jhash, and so replacements
> will be evaluated on a case-by-case basis based on whether or not the
> difference in speed is negligible and whether or not the current jhash usage
> poses a real security risk.
>
> For the second usage:
>
> A few places in the kernel are using MD5 or SHA1 for creating secure
> sequence numbers, syn cookies, port numbers, or fast random numbers.
> SipHash is a faster and more fitting, and more secure replacement for MD5
> in those situations. Replacing MD5 and SHA1 with SipHash for these uses is
> obvious and straight-forward, and so is submitted along with this patch
> series. There shouldn't be much of a debate over its efficacy.
>
> Dozens of languages are already using this internally for their hash
> tables and PRFs. Some of the BSDs already use this in their kernels.
> SipHash is a widely known high-speed solution to a widely known set of
> problems, and it's time we catch-up.
>
> Signed-off-by: Jason A. Donenfeld <[email protected]>
> Cc: Jean-Philippe Aumasson <[email protected]>
> Cc: Linus Torvalds <[email protected]>
> Cc: Eric Biggers <[email protected]>
> Cc: David Laight <[email protected]>
> Cc: Eric Dumazet <[email protected]>
The networking tree (net-next) which is where you are submitting to is technically
closed right now.
On Thu, Dec 22, 2016 at 2:40 AM, Stephen Hemminger
<[email protected]> wrote:
> The networking tree (net-next) which is where you are submitting to is technically
> closed right now.
That's okay. At some point in the future it will be open. By then v83
of this patch set will be shiny and done, just waiting for the merge
window to open. There's a lot to discuss with this, so getting the
feedback early is beneficial.
Jason
On 22.12.2016 00:42, Andy Lutomirski wrote:
> On Wed, Dec 21, 2016 at 3:02 PM, Jason A. Donenfeld <[email protected]> wrote:
>> unsigned int get_random_int(void)
>> {
>> - __u32 *hash;
>> - unsigned int ret;
>> -
>> - if (arch_get_random_int(&ret))
>> - return ret;
>> -
>> - hash = get_cpu_var(get_random_int_hash);
>> -
>> - hash[0] += current->pid + jiffies + random_get_entropy();
>> - md5_transform(hash, random_int_secret);
>> - ret = hash[0];
>> - put_cpu_var(get_random_int_hash);
>> -
>> - return ret;
>> + unsigned int arch_result;
>> + u64 result;
>> + struct random_int_secret *secret;
>> +
>> + if (arch_get_random_int(&arch_result))
>> + return arch_result;
>> +
>> + secret = get_random_int_secret();
>> + result = siphash_3u64(secret->chaining, jiffies,
>> + (u64)random_get_entropy() + current->pid,
>> + secret->secret);
>> + secret->chaining += result;
>> + put_cpu_var(secret);
>> + return result;
>> }
>> EXPORT_SYMBOL(get_random_int);
>
> Hmm. I haven't tried to prove anything for real. But here goes (in
> the random oracle model):
>
> Suppose I'm an attacker and I don't know the secret or the chaining
> value. Then, regardless of what the entropy is, I can't predict the
> numbers.
>
> Now suppose I do know the secret and the chaining value due to some
> leak. If I want to deduce prior outputs, I think I'm stuck: I'd need
> to find a value "result" such that prev_chaining + result = chaining
> and result = H(prev_chaining, ..., secret);. I don't think this can
> be done efficiently in the random oracle model regardless of what the
> "..." is.
>
> But, if I know the secret and chaining value, I can predict the next
> output assuming I can guess the entropy. What's worse is that, even
> if I can't guess the entropy, if I *observe* the next output then I
> can calculate the next chaining value.
>
> So this is probably good enough, and making it better is hard. Changing it to:
>
> u64 entropy = (u64)random_get_entropy() + current->pid;
> result = siphash(..., entropy, ...);
> secret->chaining += result + entropy;
>
> would reduce this problem by forcing an attacker to brute-force the
> entropy on each iteration, which is probably an improvement.
>
> To fully fix it, something like "catastrophic reseeding" would be
> needed, but that's hard to get right.
I wonder if Ted's proposal was analyzed further in terms of performance
if get_random_int should provide cprng alike properties?
For reference: https://lkml.org/lkml/2016/12/14/351
The proposal made sense to me and would completely solve the above
mentioned problem on the cost of repeatedly reseeding from the crng.
Bye,
Hannes
On Wed, Dec 21, 2016 at 6:07 PM, Hannes Frederic Sowa
<[email protected]> wrote:
> On 22.12.2016 00:42, Andy Lutomirski wrote:
>> On Wed, Dec 21, 2016 at 3:02 PM, Jason A. Donenfeld <[email protected]> wrote:
>>> unsigned int get_random_int(void)
>>> {
>>> - __u32 *hash;
>>> - unsigned int ret;
>>> -
>>> - if (arch_get_random_int(&ret))
>>> - return ret;
>>> -
>>> - hash = get_cpu_var(get_random_int_hash);
>>> -
>>> - hash[0] += current->pid + jiffies + random_get_entropy();
>>> - md5_transform(hash, random_int_secret);
>>> - ret = hash[0];
>>> - put_cpu_var(get_random_int_hash);
>>> -
>>> - return ret;
>>> + unsigned int arch_result;
>>> + u64 result;
>>> + struct random_int_secret *secret;
>>> +
>>> + if (arch_get_random_int(&arch_result))
>>> + return arch_result;
>>> +
>>> + secret = get_random_int_secret();
>>> + result = siphash_3u64(secret->chaining, jiffies,
>>> + (u64)random_get_entropy() + current->pid,
>>> + secret->secret);
>>> + secret->chaining += result;
>>> + put_cpu_var(secret);
>>> + return result;
>>> }
>>> EXPORT_SYMBOL(get_random_int);
>>
>> Hmm. I haven't tried to prove anything for real. But here goes (in
>> the random oracle model):
>>
>> Suppose I'm an attacker and I don't know the secret or the chaining
>> value. Then, regardless of what the entropy is, I can't predict the
>> numbers.
>>
>> Now suppose I do know the secret and the chaining value due to some
>> leak. If I want to deduce prior outputs, I think I'm stuck: I'd need
>> to find a value "result" such that prev_chaining + result = chaining
>> and result = H(prev_chaining, ..., secret);. I don't think this can
>> be done efficiently in the random oracle model regardless of what the
>> "..." is.
>>
>> But, if I know the secret and chaining value, I can predict the next
>> output assuming I can guess the entropy. What's worse is that, even
>> if I can't guess the entropy, if I *observe* the next output then I
>> can calculate the next chaining value.
>>
>> So this is probably good enough, and making it better is hard. Changing it to:
>>
>> u64 entropy = (u64)random_get_entropy() + current->pid;
>> result = siphash(..., entropy, ...);
>> secret->chaining += result + entropy;
>>
>> would reduce this problem by forcing an attacker to brute-force the
>> entropy on each iteration, which is probably an improvement.
>>
>> To fully fix it, something like "catastrophic reseeding" would be
>> needed, but that's hard to get right.
>
> I wonder if Ted's proposal was analyzed further in terms of performance
> if get_random_int should provide cprng alike properties?
>
> For reference: https://lkml.org/lkml/2016/12/14/351
>
> The proposal made sense to me and would completely solve the above
> mentioned problem on the cost of repeatedly reseeding from the crng.
>
Unless I've misunderstood it, Ted's proposal causes get_random_int()
to return bytes straight from urandom (effectively), which should make
it very strong. And if urandom is competitively fast now, I don't see
the problem. ChaCha20 is designed for speed, after all.
Hi Andy,
On Thu, Dec 22, 2016 at 12:42 AM, Andy Lutomirski <[email protected]> wrote:
> So this is probably good enough, and making it better is hard. Changing it to:
>
> u64 entropy = (u64)random_get_entropy() + current->pid;
> result = siphash(..., entropy, ...);
> secret->chaining += result + entropy;
>
> would reduce this problem by forcing an attacker to brute-force the
> entropy on each iteration, which is probably an improvement.
Ahh, so that's the reasoning behind a similar suggestion of yours in a
previous email. Makes sense to me. I'll include this in the next merge
if we don't come up with a different idea before then. Your reasoning
seems good for it.
Part of what makes this process a bit goofy is that it's not all
together clear what the design goals are. Right now we're going for
"not worse than before", which we've nearly achieved. How good of an
RNG do we want? I'm willing to examine and analyze the security and
performance of all constructions we can come up with. One thing I
don't want to do, however, is start tweaking the primitives themselves
in ways not endorsed by the designers. So, I believe that precludes
things like carrying over SipHash's internal state (like what was done
with MD5), because there hasn't been a formal security analysis of
this like there has with other uses of SipHash. I also don't want to
change any internals of how SipHash actually works. I mention that
because of some of the suggestions on other threads, which make me
rather uneasy.
So with that said, while writing this reply to you, I was
simultaneously reading some other crypto code and was reminded that
there's a variant of SipHash which outputs an additional 64-bits; it's
part of the siphash reference code, which they call the "128-bit
mode". It has the benefit that we can return 64-bits to the caller and
save 64-bits for the chaining key. That way there's no correlation
between the returned secret and the chaining key, which I think would
completely alleviate all of your concerns, and simplify the analysis a
bit.
Here's what it looks like:
https://git.zx2c4.com/linux-dev/commit/?h=siphash&id=46fbe5b408e66b2d16b4447860f8083480e1c08d
The downside is that it takes 4 extra Sip rounds. This puts the
performance still better than MD5, though, and likely still better
than the other batched entropy solution. We could optimize this, I
suppose, by giving it only two parameters -- chaining,
jiffies+entropy+pid -- instead of the current three -- chaining,
jiffies, entropy+pid -- which would then shave off 2 Sip rounds. But I
liked the idea of having a bit more spread in the entropy input field.
Anyway, with this in mind, we now have three possibilities:
1. result = siphash(chaining, entropy, key); chaining += result + entropy
2. result = siphash_extra_output(chaining, entropy, key, &chaining);
3. Ted's batched entropy idea using chacha20
The more I think about this, the more I suspect that we should just
use chacha20. It will still be faster than MD5. I don't like the
non-determinism of it (some processes will start slower than others,
if the batched entropy has run out and ASLR demands more), but I guess
I can live with that. But, most importantly, it greatly simplifies
both the security analysis and what we can promise to callers about
the function. Right now in the comment documentation, we're coy with
callers about the security of the RNG. If we moved to a known
construction like chacha20/get_random_bytes_batched, then we could
just be straight up with a promise that the numbers it returns are
high quality.
Thoughts on 2 and 3, and on 1 vs 2 vs 3?
Jason
Hi Andy & Hannes,
On Thu, Dec 22, 2016 at 3:07 AM, Hannes Frederic Sowa
<[email protected]> wrote:
> I wonder if Ted's proposal was analyzed further in terms of performance
> if get_random_int should provide cprng alike properties?
>
> For reference: https://lkml.org/lkml/2016/12/14/351
>
> The proposal made sense to me and would completely solve the above
> mentioned problem on the cost of repeatedly reseeding from the crng.
On Thu, Dec 22, 2016 at 3:09 AM, Andy Lutomirski <[email protected]> wrote:
> Unless I've misunderstood it, Ted's proposal causes get_random_int()
> to return bytes straight from urandom (effectively), which should make
> it very strong. And if urandom is competitively fast now, I don't see
> the problem. ChaCha20 is designed for speed, after all.
Funny -- while you guys were sending this back & forth, I was writing
my reply to Andy which essentially arrives at the same conclusion.
Given that we're all arriving to the same thing, and that Ted shot in
this direction long before we all did, I'm leaning toward abandoning
SipHash for the de-MD5-ification of get_random_int/long, and working
on polishing Ted's idea into something shiny for this patchset.
I did have two objections to this. The first was that my SipHash
construction is faster. But in any case, they're both faster than the
current MD5, so it's just extra rice. The second, and the more
important one, was that batching entropy up like this means that 32
calls will be really fast, and then the 33rd will be slow, since it
has to do a whole ChaCha round, because get_random_bytes must be
called to refill the batch. Since get_random_long is called for every
process startup, I didn't really like there being inconsistent
performance on process startup. And I'm pretty sure that one ChaCha
whole block is slower than computing MD5, even though it lasts 32
times as long, though I need to measure this. But maybe that's dumb in
the end? Are these concerns that should point us toward the
determinism (and speed) of SipHash? Are these concerns that don't
matter and so we should roll with the simplicity of reusing ChaCha?
Jason
On Thu, Dec 22, 2016 at 3:49 AM, Jason A. Donenfeld <[email protected]> wrote:
> I did have two objections to this. The first was that my SipHash
> construction is faster. But in any case, they're both faster than the
> current MD5, so it's just extra rice. The second, and the more
> important one, was that batching entropy up like this means that 32
> calls will be really fast, and then the 33rd will be slow, since it
> has to do a whole ChaCha round, because get_random_bytes must be
> called to refill the batch. Since get_random_long is called for every
> process startup, I didn't really like there being inconsistent
> performance on process startup. And I'm pretty sure that one ChaCha
> whole block is slower than computing MD5, even though it lasts 32
> times as long, though I need to measure this. But maybe that's dumb in
> the end? Are these concerns that should point us toward the
> determinism (and speed) of SipHash? Are these concerns that don't
> matter and so we should roll with the simplicity of reusing ChaCha?
I ran some measurements in order to quantify what I'm talking about.
Repeatedly running md5_transform is about 2.3 times faster than
repeatedly running extract_crng. What does this mean?
One call to extract_crng gives us 32 times as many longs as one call
to md5_transform. This means that spread over 32 process creations,
chacha will be 13.9 times faster. However, every 32nd process will
take 2.3 times as long to generate its ASLR value as it would with the
old md5_transform code.
Personally, I don't think that 2.3 is a big deal. And I really like
how much this simplifies the analysis.
But if it's a big deal to you, then we can continue to discuss my
SipHash construction, which gives faster and more consistent
performance, at the cost of a more complicated and probably less
impressive security analysis.
Jason
On Thu, Dec 22, 2016 at 03:49:39AM +0100, Jason A. Donenfeld wrote:
>
> Funny -- while you guys were sending this back & forth, I was writing
> my reply to Andy which essentially arrives at the same conclusion.
> Given that we're all arriving to the same thing, and that Ted shot in
> this direction long before we all did, I'm leaning toward abandoning
> SipHash for the de-MD5-ification of get_random_int/long, and working
> on polishing Ted's idea into something shiny for this patchset.
here are my numbers comparing siphash (using the first three patches
of the v7 siphash patches) with my batched chacha20 implementation.
The results are taken by running get_random_* 10000 times, and then
dividing the numbers by 10000 to get the average number of cycles for
the call. I compiled 32-bit and 64-bit kernels, and ran the results
using kvm:
siphash batched chacha20
get_random_int get_random_long get_random_int get_random_long
32-bit 270 278 114 146
64-bit 75 75 106 186
> I did have two objections to this. The first was that my SipHash
> construction is faster.
Well, it's faster on everything except 32-bit x86. :-P
> The second, and the more
> important one, was that batching entropy up like this means that 32
> calls will be really fast, and then the 33rd will be slow, since it
> has to do a whole ChaCha round, because get_random_bytes must be
> called to refill the batch.
... and this will take 2121 cycles on 64-bit x86, and 2315 cycles on a
32-bit x86. Which on a 2.3 GHz processor, is just under a
microsecond. As far as being inconsistent on process startup, I very
much doubt a microsecond is really going to be visible to the user. :-)
The bottom line is that I think we're really "pixel peeping" at this
point --- which is what obsessed digital photographers will do when
debating the quality of a Canon vs Nikon DSLR by blowing up a photo by
a thousand times, and then trying to claim that this is visible to the
human eye. Or people who obsessing over the frequency response curves
of TH-X00 headphones with Mahogony vs Purpleheart wood, when it's
likely that in a blind head-to-head comparison, most people wouldn't
be able to tell the difference....
I think the main argument for using the batched getrandom approach is
that it, I would argue, simpler than introducing siphash into the
picture. On 64-bit platforms it is faster and more consistent, so
it's basically that versus complexity of having to adding siphash to
the things that people have to analyze when considering random number
security on Linux. But it's a close call either way, I think.
- Ted
P.S. My benchmarking code....
diff --git a/drivers/char/random.c b/drivers/char/random.c
index a51f0ff43f00..41860864b775 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -1682,6 +1682,55 @@ static int rand_initialize(void)
}
early_initcall(rand_initialize);
+static unsigned int get_random_int_new(void);
+static unsigned long get_random_long_new(void);
+
+#define NUM_CYCLES 10000
+#define AVG(finish, start) ((unsigned int)(finish - start + NUM_CYCLES/2) / NUM_CYCLES)
+
+static int rand_benchmark(void)
+{
+ cycles_t start,finish;
+ int i, out;
+
+ pr_crit("random benchmark!!\n");
+ start = get_cycles();
+ for (i = 0; i < NUM_CYCLES; i++) {
+ get_random_int();}
+ finish = get_cycles();
+ pr_err("get_random_int # cycles: %u\n", AVG(finish, start));
+
+ start = get_cycles();
+ for (i = 0; i < NUM_CYCLES; i++) {
+ get_random_int_new();
+ }
+ finish = get_cycles();
+ pr_err("get_random_int_new (batched chacha20) # cycles: %u\n", AVG(finish, start));
+
+ start = get_cycles();
+ for (i = 0; i < NUM_CYCLES; i++) {
+ get_random_long();
+ }
+ finish = get_cycles();
+ pr_err("get_random_long # cycles: %u\n", AVG(finish, start));
+
+ start = get_cycles();
+ for (i = 0; i < NUM_CYCLES; i++) {
+ get_random_long_new();
+ }
+ finish = get_cycles();
+ pr_err("get_random_long_new (batched chacha20) # cycles: %u\n", AVG(finish, start));
+
+ start = get_cycles();
+ for (i = 0; i < NUM_CYCLES; i++) {
+ get_random_bytes(&out, sizeof(out));
+ }
+ finish = get_cycles();
+ pr_err("get_random_bytes # cycles: %u\n", AVG(finish, start));
+ return 0;
+}
+device_initcall(rand_benchmark);
+
#ifdef CONFIG_BLOCK
void rand_initialize_disk(struct gendisk *disk)
{
@@ -2064,8 +2113,10 @@ unsigned int get_random_int(void)
unsigned int ret;
u64 *chaining;
+#if 0 // force slow path
if (arch_get_random_int(&ret))
return ret;
+#endif
chaining = &get_cpu_var(get_random_int_chaining);
ret = *chaining = siphash_3u64(*chaining, jiffies, random_get_entropy() +
@@ -2083,8 +2134,10 @@ unsigned long get_random_long(void)
unsigned long ret;
u64 *chaining;
+#if 0 // force slow path
if (arch_get_random_long(&ret))
return ret;
+#endif
chaining = &get_cpu_var(get_random_int_chaining);
ret = *chaining = siphash_3u64(*chaining, jiffies, random_get_entropy() +
@@ -2094,6 +2147,47 @@ unsigned long get_random_long(void)
}
EXPORT_SYMBOL(get_random_long);
+struct random_buf {
+ __u8 buf[CHACHA20_BLOCK_SIZE];
+ int ptr;
+};
+
+static DEFINE_PER_CPU(struct random_buf, batched_entropy);
+
+static void get_batched_entropy(void *buf, int n)
+{
+ struct random_buf *p;
+
+ p = &get_cpu_var(batched_entropy);
+
+ if ((p->ptr == 0) ||
+ (p->ptr + n >= CHACHA20_BLOCK_SIZE)) {
+ extract_crng(p->buf);
+ p->ptr = 0;
+ }
+ BUG_ON(n > CHACHA20_BLOCK_SIZE);
+ memcpy(buf, p->buf, n);
+ p->ptr += n;
+ put_cpu_var(batched_entropy);
+}
+
+static unsigned int get_random_int_new(void)
+{
+ unsigned int ret;
+
+ get_batched_entropy(&ret, sizeof(ret));
+ return ret;
+}
+
+static unsigned long get_random_long_new(void)
+{
+ unsigned long ret;
+
+ get_batched_entropy(&ret, sizeof(ret));
+ return ret;
+}
+
+
/**
* randomize_page - Generate a random, page aligned address
* @start: The smallest acceptable address the caller will take.
Hi Ted,
On Thu, Dec 22, 2016 at 6:41 AM, Theodore Ts'o <[email protected]> wrote:
> The bottom line is that I think we're really "pixel peeping" at this
> point --- which is what obsessed digital photographers will do when
> debating the quality of a Canon vs Nikon DSLR by blowing up a photo by
> a thousand times, and then trying to claim that this is visible to the
> human eye. Or people who obsessing over the frequency response curves
> of TH-X00 headphones with Mahogony vs Purpleheart wood, when it's
> likely that in a blind head-to-head comparison, most people wouldn't
> be able to tell the difference....
This is hilarious, thanks for the laugh. I believe you're right about this...
>
> I think the main argument for using the batched getrandom approach is
> that it, I would argue, simpler than introducing siphash into the
> picture. On 64-bit platforms it is faster and more consistent, so
> it's basically that versus complexity of having to adding siphash to
> the things that people have to analyze when considering random number
> security on Linux. But it's a close call either way, I think.
I find this compelling. We'll have one csprng for both
get_random_int/long and for /dev/urandom, and we'll be able to update
that silly warning on the comment of get_random_int/long to read
"gives output of either rdrand quality or of /dev/urandom quality",
which makes it more useful for other things. It introduces less error
prone code, and it lets the RNG analysis be spent on just one RNG, not
two.
So, with your blessing, I'm going to move ahead with implementing a
pretty version of this for v8.
Regards,
Jason
Hi Ted,
On Thu, 2016-12-22 at 00:41 -0500, Theodore Ts'o wrote:
> On Thu, Dec 22, 2016 at 03:49:39AM +0100, Jason A. Donenfeld wrote:
> >
> > Funny -- while you guys were sending this back & forth, I was writing
> > my reply to Andy which essentially arrives at the same conclusion.
> > Given that we're all arriving to the same thing, and that Ted shot in
> > this direction long before we all did, I'm leaning toward abandoning
> > SipHash for the de-MD5-ification of get_random_int/long, and working
> > on polishing Ted's idea into something shiny for this patchset.
>
> here are my numbers comparing siphash (using the first three patches
> of the v7 siphash patches) with my batched chacha20 implementation.
> The results are taken by running get_random_* 10000 times, and then
> dividing the numbers by 10000 to get the average number of cycles for
> the call. I compiled 32-bit and 64-bit kernels, and ran the results
> using kvm:
>
> siphash batched chacha20
> get_random_int get_random_long get_random_int get_random_long
>
> 32-bit 270 278 114 146
> 64-bit 75 75 106 186
>
> > I did have two objections to this. The first was that my SipHash
> > construction is faster.
>
> Well, it's faster on everything except 32-bit x86. :-P
>
> > The second, and the more
> > important one, was that batching entropy up like this means that 32
> > calls will be really fast, and then the 33rd will be slow, since it
> > has to do a whole ChaCha round, because get_random_bytes must be
> > called to refill the batch.
>
> ... and this will take 2121 cycles on 64-bit x86, and 2315 cycles on a
> 32-bit x86. Which on a 2.3 GHz processor, is just under a
> microsecond. As far as being inconsistent on process startup, I very
> much doubt a microsecond is really going to be visible to the user. :-)
>
> The bottom line is that I think we're really "pixel peeping" at this
> point --- which is what obsessed digital photographers will do when
> debating the quality of a Canon vs Nikon DSLR by blowing up a photo by
> a thousand times, and then trying to claim that this is visible to the
> human eye. Or people who obsessing over the frequency response curves
> of TH-X00 headphones with Mahogony vs Purpleheart wood, when it's
> likely that in a blind head-to-head comparison, most people wouldn't
> be able to tell the difference....
>
> I think the main argument for using the batched getrandom approach is
> that it, I would argue, simpler than introducing siphash into the
> picture. On 64-bit platforms it is faster and more consistent, so
> it's basically that versus complexity of having to adding siphash to
> the things that people have to analyze when considering random number
> security on Linux. But it's a close call either way, I think.
following up on what appears to be a random subject: ;)
IIRC, ext4 code by default still uses half_md4 for hashing of filenames
in the htree. siphash seems to fit this use case pretty good.
xfs could also need an update, as they don't seed the directory hash
tables at all (but not sure if they are vulnerable). I should improve
[1] a bit.
[1] http://oss.sgi.com/cgi-bin/gitweb.cgi?p=xfs/cmds/xfstests.git;a=blo
b;f=src/dirhash_collide.c;h=55cec872d5061ac2ca0f56d1f11e6bf349d5bb97;hb
=HEAD
Bye,
Hannes
On Thu, Dec 22, 2016 at 1:47 PM, Hannes Frederic Sowa
<[email protected]> wrote:
> following up on what appears to be a random subject: ;)
>
> IIRC, ext4 code by default still uses half_md4 for hashing of filenames
> in the htree. siphash seems to fit this use case pretty good.
I saw this too. I'll try to address it in v8 of this series.
On 22.12.2016 14:10, Jason A. Donenfeld wrote:
> On Thu, Dec 22, 2016 at 1:47 PM, Hannes Frederic Sowa
> <[email protected]> wrote:
>> following up on what appears to be a random subject: ;)
>>
>> IIRC, ext4 code by default still uses half_md4 for hashing of filenames
>> in the htree. siphash seems to fit this use case pretty good.
>
> I saw this too. I'll try to address it in v8 of this series.
This change would need a new version of the ext4 super block, because
you should not change existing file systems.
On Thu, Dec 22, 2016 at 4:05 PM, Hannes Frederic Sowa
<[email protected]> wrote:
> This change would need a new version of the ext4 super block, because
> you should not change existing file systems.
Right.
As a first step, I'm considering adding a patch to move halfmd4.c
inside the ext4 domain, or at the very least, simply remove it from
linux/cryptohash.h. That'll then leave the handful of bizarre sha1
usages to consider.
On Thu, Dec 22, 2016 at 4:12 PM, Jason A. Donenfeld <[email protected]> wrote:
> As a first step, I'm considering adding a patch to move halfmd4.c
> inside the ext4 domain, or at the very least, simply remove it from
> linux/cryptohash.h. That'll then leave the handful of bizarre sha1
> usages to consider.
Specifically something like this:
https://git.zx2c4.com/linux-dev/commit/?h=siphash&id=978213351f9633bd1e3d1fdc3f19d28e36eeac90
That only leaves two more uses of "cryptohash" to consider, but they
require a bit of help. First, sha_transform in net/ipv6/addrconf.c.
That might be a straight-forward conversion to SipHash, but perhaps
not; I need to look closely and think about it. The next is
sha_transform in kernel/bpf/core.c. I really have no idea what's going
on with the eBPF stuff, so that will take a bit longer to study. Maybe
sha1 is fine in the end there? I'm not sure yet.
On Thu, 2016-12-22 at 16:29 +0100, Jason A. Donenfeld wrote:
> On Thu, Dec 22, 2016 at 4:12 PM, Jason A. Donenfeld <[email protected]> wrote:
> > As a first step, I'm considering adding a patch to move halfmd4.c
> > inside the ext4 domain, or at the very least, simply remove it from
> > linux/cryptohash.h. That'll then leave the handful of bizarre sha1
> > usages to consider.
>
> Specifically something like this:
>
> https://git.zx2c4.com/linux-dev/commit/?h=siphash&id=978213351f9633bd1e3d1fdc3f19d28e36eeac90
>
> That only leaves two more uses of "cryptohash" to consider, but they
> require a bit of help. First, sha_transform in net/ipv6/addrconf.c.
> That might be a straight-forward conversion to SipHash, but perhaps
> not; I need to look closely and think about it. The next is
> sha_transform in kernel/bpf/core.c. I really have no idea what's going
> on with the eBPF stuff, so that will take a bit longer to study. Maybe
> sha1 is fine in the end there? I'm not sure yet.
IPv6 you cannot touch anymore. The hashing algorithm is part of uAPI.
You don't want to give people new IPv6 addresses with the same stable
secret (across reboots) after a kernel upgrade. Maybe they lose
connectivity then and it is extra work?
The bpf hash stuff can be changed during this merge window, as it is
not yet in a released kernel. Albeit I would probably have preferred
something like sha256 here, which can be easily replicated by user
space tools (minus the problem of patching out references to not
hashable data, which must be zeroed).
Bye,
Hannes
Hi Hannes,
On Thu, Dec 22, 2016 at 4:33 PM, Hannes Frederic Sowa
<[email protected]> wrote:
> IPv6 you cannot touch anymore. The hashing algorithm is part of uAPI.
> You don't want to give people new IPv6 addresses with the same stable
> secret (across reboots) after a kernel upgrade. Maybe they lose
> connectivity then and it is extra work?
Ahh, too bad. So it goes.
> The bpf hash stuff can be changed during this merge window, as it is
> not yet in a released kernel. Albeit I would probably have preferred
> something like sha256 here, which can be easily replicated by user
> space tools (minus the problem of patching out references to not
> hashable data, which must be zeroed).
Oh, interesting, so time is of the essence then. Do you want to handle
changing the new eBPF code to something not-SHA1 before it's too late,
as part of a new patchset that can fast track itself to David? And
then I can preserve my large series for the next merge window.
Jason
On Thu, 2016-12-22 at 16:41 +0100, Jason A. Donenfeld wrote:
> Hi Hannes,
>
> On Thu, Dec 22, 2016 at 4:33 PM, Hannes Frederic Sowa
> <[email protected]> wrote:
> > IPv6 you cannot touch anymore. The hashing algorithm is part of uAPI.
> > You don't want to give people new IPv6 addresses with the same stable
> > secret (across reboots) after a kernel upgrade. Maybe they lose
> > connectivity then and it is extra work?
>
> Ahh, too bad. So it goes.
If no other users survive we can put it into the ipv6 module.
> > The bpf hash stuff can be changed during this merge window, as it is
> > not yet in a released kernel. Albeit I would probably have preferred
> > something like sha256 here, which can be easily replicated by user
> > space tools (minus the problem of patching out references to not
> > hashable data, which must be zeroed).
>
> Oh, interesting, so time is of the essence then. Do you want to handle
> changing the new eBPF code to something not-SHA1 before it's too late,
> as part of a new patchset that can fast track itself to David? And
> then I can preserve my large series for the next merge window.
This algorithm should be a non-seeded algorithm, because the hashes
should be stable and verifiable by user space tooling. Thus this would
need a hashing algorithm that is hardened against pre-image
attacks/collision resistance, which siphash is not. I would prefer some
higher order SHA algorithm for that actually.
Bye,
Hannes
On Thu, Dec 22, 2016 at 4:51 PM, Hannes Frederic Sowa
<[email protected]> wrote:
> This algorithm should be a non-seeded algorithm, because the hashes
> should be stable and verifiable by user space tooling. Thus this would
> need a hashing algorithm that is hardened against pre-image
> attacks/collision resistance, which siphash is not. I would prefer some
> higher order SHA algorithm for that actually.
Right. SHA-256, SHA-512/256, Blake2s, or Blake2b would probably be
good candidates for this.
On Thu, Dec 22, 2016 at 02:10:33PM +0100, Jason A. Donenfeld wrote:
> On Thu, Dec 22, 2016 at 1:47 PM, Hannes Frederic Sowa
> <[email protected]> wrote:
> > following up on what appears to be a random subject: ;)
> >
> > IIRC, ext4 code by default still uses half_md4 for hashing of filenames
> > in the htree. siphash seems to fit this use case pretty good.
>
> I saw this too. I'll try to address it in v8 of this series.
This is a separate issue, and this series is getting a bit too
complex. So I'd suggest pushing this off to a separate change.
Changing the htree hash algorithm is an on-disk format change, and so
we couldn't roll it out until e2fsprogs gets updated and rolled out
pretty broadley. In fact George sent me patches to add siphash as a
hash algorithm for htree a while back (for both the kernel and
e2fsprogs), but I never got around to testing and applying them,
mainly because while it's technically faster, I had other higher
priority issues to work on --- and see previous comments regarding
pixel peeping. Improving the hash algorithm by tens or even hundreds
of nanoseconds isn't really going to matter since we only do a htree
lookup on a file creation or cold cache lookup, and the SSD or HDD I/O
times will dominate. And from the power perspective, saving
microwatts of CPU power isn't going to matter if you're going to be
spinning up the storage device....
- Ted
On Thu, Dec 22, 2016 at 07:03:29AM +0100, Jason A. Donenfeld wrote:
> I find this compelling. We'll have one csprng for both
> get_random_int/long and for /dev/urandom, and we'll be able to update
> that silly warning on the comment of get_random_int/long to read
> "gives output of either rdrand quality or of /dev/urandom quality",
> which makes it more useful for other things. It introduces less error
> prone code, and it lets the RNG analysis be spent on just one RNG, not
> two.
>
> So, with your blessing, I'm going to move ahead with implementing a
> pretty version of this for v8.
Can we do this as a separate series, please? At this point, it's a
completely separate change from a logical perspective, and we can take
in the change through the random.git tree.
Changes that touch files that are normally changed in several
different git trees leads to the potential for merge conflicts during
the linux-next integration and merge window processes. Which is why
it's generally best to try to isolate changes as much as possible.
Cheers,
- Ted
Hi Ted,
On Thu, Dec 22, 2016 at 4:58 PM, Theodore Ts'o <[email protected]> wrote:
> Can we do this as a separate series, please? At this point, it's a
> completely separate change from a logical perspective, and we can take
> in the change through the random.git tree.
>
> Changes that touch files that are normally changed in several
> different git trees leads to the potential for merge conflicts during
> the linux-next integration and merge window processes. Which is why
> it's generally best to try to isolate changes as much as possible.
Sure, I can separate things out.
Could you offer a bit of advice on how to manage dependencies between
patchsets during merge windows? I'm a bit new to the process.
Specifically, we how have 4 parts:
1. add siphash, and use it for some networking code. to: david miller's net-next
2. convert char/random to use siphash. to: ted ts'o's random-next
3. move lib/md5.c to static function in crypto/md5.c, remove entry
inside of linux/cryptohash.h. to: ??'s ??-next
4. move lib/halfmd4.c to static function in fs/ext/hash.c, remove
entry inside of linux/cryptohash.c. to: td ts'o's ext-next
Problem: 2 depends on 1, 3 depends on 1 & 2. But this can be
simplified into 3 parts:
1. add siphash, and use it for some networking code. to: david miller's net-next
2. convert char/random to use siphash, move lib/md5.c to static
function in crypto/md5.c, remove entry inside of linux/cryptohash.h.
to: ted ts'o's random-next
3. move lib/halfmd4.c to static function in fs/ext/hash.c, remove
entry inside of linux/cryptohash.c. to: td ts'o's ext-next
Problem: 2 depends on 1. Is that okay with you?
Also, would you like me to merge (3) and (2) of the second list into
one series for you?
Jason
On Thu, Dec 22, 2016 at 05:16:47PM +0100, Jason A. Donenfeld wrote:
> Could you offer a bit of advice on how to manage dependencies between
> patchsets during merge windows? I'm a bit new to the process.
>
> Specifically, we how have 4 parts:
> 1. add siphash, and use it for some networking code. to: david miller's net-next
I'd do this first, as one set. Adding a new file to crypto is
unlikely to cause merge conflicts.
> 2. convert char/random to use siphash. to: ted ts'o's random-next
I'm confused, I thought you had agreed to the batched chacha20
approach?
> 3. move lib/md5.c to static function in crypto/md5.c, remove entry
> inside of linux/cryptohash.h. to: ??'s ??-next
This is cleanup, so it doesn't matter that much when it happens. md5
changes to crypto is also unlikely to cause conflicts, so we could do
this at the same time as (2), if Herbert (the crypto maintainer) agrees.
> 4. move lib/halfmd4.c to static function in fs/ext/hash.c, remove
> entry inside of linux/cryptohash.c. to: td ts'o's ext-next
This is definitely separate.
One more thing. Can you add some test cases to lib/siphash.h?
Triggered off of a CONFIG_SIPHASH_REGRESSION_TEST config flag, with
some test inputs and known outputs? I'm going to need to add a
version of siphash to e2fsprogs, and I want to make sure the userspace
version is implementing the same algorithm as the kernel siphash.
- Ted
On Thu, Dec 22, 2016 at 5:30 PM, Theodore Ts'o <[email protected]> wrote:
> I'd do this first, as one set. Adding a new file to crypto is
> unlikely to cause merge conflicts.
Ack.
>
>> 2. convert char/random to use siphash. to: ted ts'o's random-next
>
> I'm confused, I thought you had agreed to the batched chacha20
> approach?
Sorry, I meant to write this. Long day, little sleep. Yes, of course.
Batched entropy.
>> 3. move lib/md5.c to static function in crypto/md5.c, remove entry
>> inside of linux/cryptohash.h. to: ??'s ??-next
>
> This is cleanup, so it doesn't matter that much when it happens. md5
> changes to crypto is also unlikely to cause conflicts, so we could do
> this at the same time as (2), if Herbert (the crypto maintainer) agrees.
Alright, sure.
>
>> 4. move lib/halfmd4.c to static function in fs/ext/hash.c, remove
>> entry inside of linux/cryptohash.c. to: td ts'o's ext-next
>
> This is definitely separate.
Okay, I'll submit it to you separately.
> One more thing. Can you add some test cases to lib/siphash.h?
> Triggered off of a CONFIG_SIPHASH_REGRESSION_TEST config flag, with
> some test inputs and known outputs? I'm going to need to add a
> version of siphash to e2fsprogs, and I want to make sure the userspace
> version is implementing the same algorithm as the kernel siphash.
I've already written these. They're behind TEST_HASH. They currently
test every single line of code of all implementations of siphash. I
spent a long time on these. The test vectors themselves were taken
from the SipHash creators' reference publication. Check out
lib/test_siphash.c in my tree.
Jason
On Thu, Dec 22, 2016 at 5:30 PM, Theodore Ts'o <[email protected]> wrote:
> On Thu, Dec 22, 2016 at 05:16:47PM +0100, Jason A. Donenfeld wrote:
>> Could you offer a bit of advice on how to manage dependencies between
>> patchsets during merge windows? I'm a bit new to the process.
>>
>> Specifically, we how have 4 parts:
>> 1. add siphash, and use it for some networking code. to: david miller's net-next
>
> I'd do this first, as one set. Adding a new file to crypto is
> unlikely to cause merge conflicts.
>
>> 2. convert char/random to use siphash. to: ted ts'o's random-next
>
> I'm confused, I thought you had agreed to the batched chacha20
> approach?
>
>> 3. move lib/md5.c to static function in crypto/md5.c, remove entry
>> inside of linux/cryptohash.h. to: ??'s ??-next
>
> This is cleanup, so it doesn't matter that much when it happens. md5
> changes to crypto is also unlikely to cause conflicts, so we could do
> this at the same time as (2), if Herbert (the crypto maintainer) agrees.
>
>> 4. move lib/halfmd4.c to static function in fs/ext/hash.c, remove
>> entry inside of linux/cryptohash.c. to: td ts'o's ext-next
>
> This is definitely separate.
>
> One more thing. Can you add some test cases to lib/siphash.h?
> Triggered off of a CONFIG_SIPHASH_REGRESSION_TEST config flag, with
> some test inputs and known outputs? I'm going to need to add a
> version of siphash to e2fsprogs, and I want to make sure the userspace
> version is implementing the same algorithm as the kernel siphash.
>
> - Ted
--
Jason A. Donenfeld
Deep Space Explorer
fr: +33 6 51 90 82 66
us: +1 513 476 1200
http://www.jasondonenfeld.com
http://www.zx2c4.com
zx2c4.com/keys/AB9942E6D4A4CFC3412620A749FC7012A5DE03AE.asc
On 22.12.2016 16:54, Theodore Ts'o wrote:
> On Thu, Dec 22, 2016 at 02:10:33PM +0100, Jason A. Donenfeld wrote:
>> On Thu, Dec 22, 2016 at 1:47 PM, Hannes Frederic Sowa
>> <[email protected]> wrote:
>>> following up on what appears to be a random subject: ;)
>>>
>>> IIRC, ext4 code by default still uses half_md4 for hashing of filenames
>>> in the htree. siphash seems to fit this use case pretty good.
>>
>> I saw this too. I'll try to address it in v8 of this series.
>
> This is a separate issue, and this series is getting a bit too
> complex. So I'd suggest pushing this off to a separate change.
>
> Changing the htree hash algorithm is an on-disk format change, and so
> we couldn't roll it out until e2fsprogs gets updated and rolled out
> pretty broadley. In fact George sent me patches to add siphash as a
> hash algorithm for htree a while back (for both the kernel and
> e2fsprogs), but I never got around to testing and applying them,
> mainly because while it's technically faster, I had other higher
> priority issues to work on --- and see previous comments regarding
> pixel peeping. Improving the hash algorithm by tens or even hundreds
> of nanoseconds isn't really going to matter since we only do a htree
> lookup on a file creation or cold cache lookup, and the SSD or HDD I/O
> times will dominate. And from the power perspective, saving
> microwatts of CPU power isn't going to matter if you're going to be
> spinning up the storage device....
I wasn't concerned about performance but more about DoS resilience. I
wonder how safe half md4 actually is in terms of allowing users to
generate long hash chains in the filesystem (in terms of length
extension attacks against half_md4).
In ext4, is it actually possible that a "disrupter" learns about the
hashing secret in the way how the inodes are returned during getdents?
Thanks,
Hannes
On Thu, Dec 22, 2016 at 7:08 PM, Hannes Frederic Sowa
<[email protected]> wrote:
> I wasn't concerned about performance but more about DoS resilience. I
> wonder how safe half md4 actually is in terms of allowing users to
> generate long hash chains in the filesystem (in terms of length
> extension attacks against half_md4).
AFAIK, this is a real vulnerability that needs to be addressed.
Judging by Ted's inquiry about my siphash testing suite, I assume he's
probably tinkering around with it as we speak. :)
Meanwhile I've separated things into several trees:
1. chacha20 rng, already submitted:
https://git.zx2c4.com/linux-dev/log/?h=random-next
2. md5 cleanup, not yet submitted:
https://git.zx2c4.com/linux-dev/log/?h=md5-cleanup
3. md4 cleanup, already submitted:
https://git.zx2c4.com/linux-dev/log/?h=ext4-next-md4-cleanup
4. siphash and networking, not yet submitted as a x/4 series:
https://git.zx2c4.com/linux-dev/log/?h=net-next-siphash
I'll submit (4) in a couple of days, waiting for any comments on the
existing patch-set.
Jason
On Thu, Dec 22, 2016 at 07:08:37PM +0100, Hannes Frederic Sowa wrote:
> I wasn't concerned about performance but more about DoS resilience. I
> wonder how safe half md4 actually is in terms of allowing users to
> generate long hash chains in the filesystem (in terms of length
> extension attacks against half_md4).
>
> In ext4, is it actually possible that a "disrupter" learns about the
> hashing secret in the way how the inodes are returned during getdents?
They'd have to be a local user, who can execute telldir(3) --- in
which case there are plenty of other denial of service attacks one
could carry out that would be far more devastating.
It might also be an issue if the file system is exposed via NFS, but
again, there are so many other ways an attacker could DoS a NFS server
that I don't think of it as a much of a concern.
Keep in mind that worst someone can do is cause directory inserts to
fail with an ENOSPC, and there are plenty of other ways of doing that
--- such as consuming all of the blocks and inodes in the file system,
for example.
So it's a threat, but not a high priority one as far as I'm concerned.
And if this was a problem in actual practice, users could switch to
the TEA based hash, which should be far harder to attack, and
available today.
- Ted