linux/crypto/vmac.c

698 lines
19 KiB
C
Raw Normal View History

/*
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
* VMAC: Message Authentication Code using Universal Hashing
*
* Reference: https://tools.ietf.org/html/draft-krovetz-vmac-01
*
* Copyright (c) 2009, Intel Corporation.
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
* Copyright (c) 2018, Google Inc.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms and conditions of the GNU General Public License,
* version 2, as published by the Free Software Foundation.
*
* This program is distributed in the hope it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License along with
* this program; if not, write to the Free Software Foundation, Inc., 59 Temple
* Place - Suite 330, Boston, MA 02111-1307 USA.
*/
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
/*
* Derived from:
* VMAC and VHASH Implementation by Ted Krovetz (tdk@acm.org) and Wei Dai.
* This implementation is herby placed in the public domain.
* The authors offers no warranty. Use at your own risk.
* Last modified: 17 APR 08, 1700 PDT
*/
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
#include <asm/unaligned.h>
#include <linux/init.h>
#include <linux/types.h>
#include <linux/crypto.h>
#include <linux/module.h>
#include <linux/scatterlist.h>
#include <asm/byteorder.h>
#include <crypto/scatterwalk.h>
#include <crypto/internal/cipher.h>
#include <crypto/internal/hash.h>
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
/*
* User definable settings.
*/
#define VMAC_TAG_LEN 64
#define VMAC_KEY_SIZE 128/* Must be 128, 192 or 256 */
#define VMAC_KEY_LEN (VMAC_KEY_SIZE/8)
#define VMAC_NHBYTES 128/* Must 2^i for any 3 < i < 13 Standard = 128*/
#define VMAC_NONCEBYTES 16
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
/* per-transform (per-key) context */
struct vmac_tfm_ctx {
struct crypto_cipher *cipher;
u64 nhkey[(VMAC_NHBYTES/8)+2*(VMAC_TAG_LEN/64-1)];
u64 polykey[2*VMAC_TAG_LEN/64];
u64 l3key[2*VMAC_TAG_LEN/64];
};
/* per-request context */
struct vmac_desc_ctx {
union {
u8 partial[VMAC_NHBYTES]; /* partial block */
__le64 partial_words[VMAC_NHBYTES / 8];
};
unsigned int partial_size; /* size of the partial block */
bool first_block_processed;
u64 polytmp[2*VMAC_TAG_LEN/64]; /* running total of L2-hash */
union {
u8 bytes[VMAC_NONCEBYTES];
__be64 pads[VMAC_NONCEBYTES / 8];
} nonce;
unsigned int nonce_size; /* nonce bytes filled so far */
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
};
/*
* Constants and masks
*/
#define UINT64_C(x) x##ULL
static const u64 p64 = UINT64_C(0xfffffffffffffeff); /* 2^64 - 257 prime */
static const u64 m62 = UINT64_C(0x3fffffffffffffff); /* 62-bit mask */
static const u64 m63 = UINT64_C(0x7fffffffffffffff); /* 63-bit mask */
static const u64 m64 = UINT64_C(0xffffffffffffffff); /* 64-bit mask */
static const u64 mpoly = UINT64_C(0x1fffffff1fffffff); /* Poly key mask */
#define pe64_to_cpup le64_to_cpup /* Prefer little endian */
#ifdef __LITTLE_ENDIAN
#define INDEX_HIGH 1
#define INDEX_LOW 0
#else
#define INDEX_HIGH 0
#define INDEX_LOW 1
#endif
/*
* The following routines are used in this implementation. They are
* written via macros to simulate zero-overhead call-by-reference.
*
* MUL64: 64x64->128-bit multiplication
* PMUL64: assumes top bits cleared on inputs
* ADD128: 128x128->128-bit addition
*/
#define ADD128(rh, rl, ih, il) \
do { \
u64 _il = (il); \
(rl) += (_il); \
if ((rl) < (_il)) \
(rh)++; \
(rh) += (ih); \
} while (0)
#define MUL32(i1, i2) ((u64)(u32)(i1)*(u32)(i2))
#define PMUL64(rh, rl, i1, i2) /* Assumes m doesn't overflow */ \
do { \
u64 _i1 = (i1), _i2 = (i2); \
u64 m = MUL32(_i1, _i2>>32) + MUL32(_i1>>32, _i2); \
rh = MUL32(_i1>>32, _i2>>32); \
rl = MUL32(_i1, _i2); \
ADD128(rh, rl, (m >> 32), (m << 32)); \
} while (0)
#define MUL64(rh, rl, i1, i2) \
do { \
u64 _i1 = (i1), _i2 = (i2); \
u64 m1 = MUL32(_i1, _i2>>32); \
u64 m2 = MUL32(_i1>>32, _i2); \
rh = MUL32(_i1>>32, _i2>>32); \
rl = MUL32(_i1, _i2); \
ADD128(rh, rl, (m1 >> 32), (m1 << 32)); \
ADD128(rh, rl, (m2 >> 32), (m2 << 32)); \
} while (0)
/*
* For highest performance the L1 NH and L2 polynomial hashes should be
* carefully implemented to take advantage of one's target architecture.
* Here these two hash functions are defined multiple time; once for
* 64-bit architectures, once for 32-bit SSE2 architectures, and once
* for the rest (32-bit) architectures.
* For each, nh_16 *must* be defined (works on multiples of 16 bytes).
* Optionally, nh_vmac_nhbytes can be defined (for multiples of
* VMAC_NHBYTES), and nh_16_2 and nh_vmac_nhbytes_2 (versions that do two
* NH computations at once).
*/
#ifdef CONFIG_64BIT
#define nh_16(mp, kp, nw, rh, rl) \
do { \
int i; u64 th, tl; \
rh = rl = 0; \
for (i = 0; i < nw; i += 2) { \
MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i], \
pe64_to_cpup((mp)+i+1)+(kp)[i+1]); \
ADD128(rh, rl, th, tl); \
} \
} while (0)
#define nh_16_2(mp, kp, nw, rh, rl, rh1, rl1) \
do { \
int i; u64 th, tl; \
rh1 = rl1 = rh = rl = 0; \
for (i = 0; i < nw; i += 2) { \
MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i], \
pe64_to_cpup((mp)+i+1)+(kp)[i+1]); \
ADD128(rh, rl, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2], \
pe64_to_cpup((mp)+i+1)+(kp)[i+3]); \
ADD128(rh1, rl1, th, tl); \
} \
} while (0)
#if (VMAC_NHBYTES >= 64) /* These versions do 64-bytes of message at a time */
#define nh_vmac_nhbytes(mp, kp, nw, rh, rl) \
do { \
int i; u64 th, tl; \
rh = rl = 0; \
for (i = 0; i < nw; i += 8) { \
MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i], \
pe64_to_cpup((mp)+i+1)+(kp)[i+1]); \
ADD128(rh, rl, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2], \
pe64_to_cpup((mp)+i+3)+(kp)[i+3]); \
ADD128(rh, rl, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4], \
pe64_to_cpup((mp)+i+5)+(kp)[i+5]); \
ADD128(rh, rl, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6], \
pe64_to_cpup((mp)+i+7)+(kp)[i+7]); \
ADD128(rh, rl, th, tl); \
} \
} while (0)
#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh1, rl1) \
do { \
int i; u64 th, tl; \
rh1 = rl1 = rh = rl = 0; \
for (i = 0; i < nw; i += 8) { \
MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i], \
pe64_to_cpup((mp)+i+1)+(kp)[i+1]); \
ADD128(rh, rl, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2], \
pe64_to_cpup((mp)+i+1)+(kp)[i+3]); \
ADD128(rh1, rl1, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2], \
pe64_to_cpup((mp)+i+3)+(kp)[i+3]); \
ADD128(rh, rl, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+4], \
pe64_to_cpup((mp)+i+3)+(kp)[i+5]); \
ADD128(rh1, rl1, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4], \
pe64_to_cpup((mp)+i+5)+(kp)[i+5]); \
ADD128(rh, rl, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+6], \
pe64_to_cpup((mp)+i+5)+(kp)[i+7]); \
ADD128(rh1, rl1, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6], \
pe64_to_cpup((mp)+i+7)+(kp)[i+7]); \
ADD128(rh, rl, th, tl); \
MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+8], \
pe64_to_cpup((mp)+i+7)+(kp)[i+9]); \
ADD128(rh1, rl1, th, tl); \
} \
} while (0)
#endif
#define poly_step(ah, al, kh, kl, mh, ml) \
do { \
u64 t1h, t1l, t2h, t2l, t3h, t3l, z = 0; \
/* compute ab*cd, put bd into result registers */ \
PMUL64(t3h, t3l, al, kh); \
PMUL64(t2h, t2l, ah, kl); \
PMUL64(t1h, t1l, ah, 2*kh); \
PMUL64(ah, al, al, kl); \
/* add 2 * ac to result */ \
ADD128(ah, al, t1h, t1l); \
/* add together ad + bc */ \
ADD128(t2h, t2l, t3h, t3l); \
/* now (ah,al), (t2l,2*t2h) need summing */ \
/* first add the high registers, carrying into t2h */ \
ADD128(t2h, ah, z, t2l); \
/* double t2h and add top bit of ah */ \
t2h = 2 * t2h + (ah >> 63); \
ah &= m63; \
/* now add the low registers */ \
ADD128(ah, al, mh, ml); \
ADD128(ah, al, z, t2h); \
} while (0)
#else /* ! CONFIG_64BIT */
#ifndef nh_16
#define nh_16(mp, kp, nw, rh, rl) \
do { \
u64 t1, t2, m1, m2, t; \
int i; \
rh = rl = t = 0; \
for (i = 0; i < nw; i += 2) { \
t1 = pe64_to_cpup(mp+i) + kp[i]; \
t2 = pe64_to_cpup(mp+i+1) + kp[i+1]; \
m2 = MUL32(t1 >> 32, t2); \
m1 = MUL32(t1, t2 >> 32); \
ADD128(rh, rl, MUL32(t1 >> 32, t2 >> 32), \
MUL32(t1, t2)); \
rh += (u64)(u32)(m1 >> 32) \
+ (u32)(m2 >> 32); \
t += (u64)(u32)m1 + (u32)m2; \
} \
ADD128(rh, rl, (t >> 32), (t << 32)); \
} while (0)
#endif
static void poly_step_func(u64 *ahi, u64 *alo,
const u64 *kh, const u64 *kl,
const u64 *mh, const u64 *ml)
{
#define a0 (*(((u32 *)alo)+INDEX_LOW))
#define a1 (*(((u32 *)alo)+INDEX_HIGH))
#define a2 (*(((u32 *)ahi)+INDEX_LOW))
#define a3 (*(((u32 *)ahi)+INDEX_HIGH))
#define k0 (*(((u32 *)kl)+INDEX_LOW))
#define k1 (*(((u32 *)kl)+INDEX_HIGH))
#define k2 (*(((u32 *)kh)+INDEX_LOW))
#define k3 (*(((u32 *)kh)+INDEX_HIGH))
u64 p, q, t;
u32 t2;
p = MUL32(a3, k3);
p += p;
p += *(u64 *)mh;
p += MUL32(a0, k2);
p += MUL32(a1, k1);
p += MUL32(a2, k0);
t = (u32)(p);
p >>= 32;
p += MUL32(a0, k3);
p += MUL32(a1, k2);
p += MUL32(a2, k1);
p += MUL32(a3, k0);
t |= ((u64)((u32)p & 0x7fffffff)) << 32;
p >>= 31;
p += (u64)(((u32 *)ml)[INDEX_LOW]);
p += MUL32(a0, k0);
q = MUL32(a1, k3);
q += MUL32(a2, k2);
q += MUL32(a3, k1);
q += q;
p += q;
t2 = (u32)(p);
p >>= 32;
p += (u64)(((u32 *)ml)[INDEX_HIGH]);
p += MUL32(a0, k1);
p += MUL32(a1, k0);
q = MUL32(a2, k3);
q += MUL32(a3, k2);
q += q;
p += q;
*(u64 *)(alo) = (p << 32) | t2;
p >>= 32;
*(u64 *)(ahi) = p + t;
#undef a0
#undef a1
#undef a2
#undef a3
#undef k0
#undef k1
#undef k2
#undef k3
}
#define poly_step(ah, al, kh, kl, mh, ml) \
poly_step_func(&(ah), &(al), &(kh), &(kl), &(mh), &(ml))
#endif /* end of specialized NH and poly definitions */
/* At least nh_16 is defined. Defined others as needed here */
#ifndef nh_16_2
#define nh_16_2(mp, kp, nw, rh, rl, rh2, rl2) \
do { \
nh_16(mp, kp, nw, rh, rl); \
nh_16(mp, ((kp)+2), nw, rh2, rl2); \
} while (0)
#endif
#ifndef nh_vmac_nhbytes
#define nh_vmac_nhbytes(mp, kp, nw, rh, rl) \
nh_16(mp, kp, nw, rh, rl)
#endif
#ifndef nh_vmac_nhbytes_2
#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh2, rl2) \
do { \
nh_vmac_nhbytes(mp, kp, nw, rh, rl); \
nh_vmac_nhbytes(mp, ((kp)+2), nw, rh2, rl2); \
} while (0)
#endif
static u64 l3hash(u64 p1, u64 p2, u64 k1, u64 k2, u64 len)
{
u64 rh, rl, t, z = 0;
/* fully reduce (p1,p2)+(len,0) mod p127 */
t = p1 >> 63;
p1 &= m63;
ADD128(p1, p2, len, t);
/* At this point, (p1,p2) is at most 2^127+(len<<64) */
t = (p1 > m63) + ((p1 == m63) && (p2 == m64));
ADD128(p1, p2, z, t);
p1 &= m63;
/* compute (p1,p2)/(2^64-2^32) and (p1,p2)%(2^64-2^32) */
t = p1 + (p2 >> 32);
t += (t >> 32);
t += (u32)t > 0xfffffffeu;
p1 += (t >> 32);
p2 += (p1 << 32);
/* compute (p1+k1)%p64 and (p2+k2)%p64 */
p1 += k1;
p1 += (0 - (p1 < k1)) & 257;
p2 += k2;
p2 += (0 - (p2 < k2)) & 257;
/* compute (p1+k1)*(p2+k2)%p64 */
MUL64(rh, rl, p1, p2);
t = rh >> 56;
ADD128(t, rl, z, rh);
rh <<= 8;
ADD128(t, rl, z, rh);
t += t << 8;
rl += t;
rl += (0 - (rl < t)) & 257;
rl += (0 - (rl > p64-1)) & 257;
return rl;
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
/* L1 and L2-hash one or more VMAC_NHBYTES-byte blocks */
static void vhash_blocks(const struct vmac_tfm_ctx *tctx,
struct vmac_desc_ctx *dctx,
const __le64 *mptr, unsigned int blocks)
{
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
const u64 *kptr = tctx->nhkey;
const u64 pkh = tctx->polykey[0];
const u64 pkl = tctx->polykey[1];
u64 ch = dctx->polytmp[0];
u64 cl = dctx->polytmp[1];
u64 rh, rl;
if (!dctx->first_block_processed) {
dctx->first_block_processed = true;
nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
rh &= m62;
ADD128(ch, cl, rh, rl);
mptr += (VMAC_NHBYTES/sizeof(u64));
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
blocks--;
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
while (blocks--) {
nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
rh &= m62;
poly_step(ch, cl, pkh, pkl, rh, rl);
mptr += (VMAC_NHBYTES/sizeof(u64));
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
dctx->polytmp[0] = ch;
dctx->polytmp[1] = cl;
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
static int vmac_setkey(struct crypto_shash *tfm,
const u8 *key, unsigned int keylen)
{
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
struct vmac_tfm_ctx *tctx = crypto_shash_ctx(tfm);
__be64 out[2];
u8 in[16] = { 0 };
unsigned int i;
int err;
crypto: remove CRYPTO_TFM_RES_BAD_KEY_LEN The CRYPTO_TFM_RES_BAD_KEY_LEN flag was apparently meant as a way to make the ->setkey() functions provide more information about errors. However, no one actually checks for this flag, which makes it pointless. Also, many algorithms fail to set this flag when given a bad length key. Reviewing just the generic implementations, this is the case for aes-fixed-time, cbcmac, echainiv, nhpoly1305, pcrypt, rfc3686, rfc4309, rfc7539, rfc7539esp, salsa20, seqiv, and xcbc. But there are probably many more in arch/*/crypto/ and drivers/crypto/. Some algorithms can even set this flag when the key is the correct length. For example, authenc and authencesn set it when the key payload is malformed in any way (not just a bad length), the atmel-sha and ccree drivers can set it if a memory allocation fails, and the chelsio driver sets it for bad auth tag lengths, not just bad key lengths. So even if someone actually wanted to start checking this flag (which seems unlikely, since it's been unused for a long time), there would be a lot of work needed to get it working correctly. But it would probably be much better to go back to the drawing board and just define different return values, like -EINVAL if the key is invalid for the algorithm vs. -EKEYREJECTED if the key was rejected by a policy like "no weak keys". That would be much simpler, less error-prone, and easier to test. So just remove this flag. Signed-off-by: Eric Biggers <ebiggers@google.com> Reviewed-by: Horia Geantă <horia.geanta@nxp.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2019-12-31 11:19:36 +08:00
if (keylen != VMAC_KEY_LEN)
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
return -EINVAL;
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
err = crypto_cipher_setkey(tctx->cipher, key, keylen);
if (err)
return err;
/* Fill nh key */
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
in[0] = 0x80;
for (i = 0; i < ARRAY_SIZE(tctx->nhkey); i += 2) {
crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
tctx->nhkey[i] = be64_to_cpu(out[0]);
tctx->nhkey[i+1] = be64_to_cpu(out[1]);
in[15]++;
}
/* Fill poly key */
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
in[0] = 0xC0;
in[15] = 0;
for (i = 0; i < ARRAY_SIZE(tctx->polykey); i += 2) {
crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
tctx->polykey[i] = be64_to_cpu(out[0]) & mpoly;
tctx->polykey[i+1] = be64_to_cpu(out[1]) & mpoly;
in[15]++;
}
/* Fill ip key */
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
in[0] = 0xE0;
in[15] = 0;
for (i = 0; i < ARRAY_SIZE(tctx->l3key); i += 2) {
do {
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
tctx->l3key[i] = be64_to_cpu(out[0]);
tctx->l3key[i+1] = be64_to_cpu(out[1]);
in[15]++;
} while (tctx->l3key[i] >= p64 || tctx->l3key[i+1] >= p64);
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
return 0;
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
static int vmac_init(struct shash_desc *desc)
{
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
dctx->partial_size = 0;
dctx->first_block_processed = false;
memcpy(dctx->polytmp, tctx->polykey, sizeof(dctx->polytmp));
dctx->nonce_size = 0;
return 0;
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
static int vmac_update(struct shash_desc *desc, const u8 *p, unsigned int len)
{
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
unsigned int n;
/* Nonce is passed as first VMAC_NONCEBYTES bytes of data */
if (dctx->nonce_size < VMAC_NONCEBYTES) {
n = min(len, VMAC_NONCEBYTES - dctx->nonce_size);
memcpy(&dctx->nonce.bytes[dctx->nonce_size], p, n);
dctx->nonce_size += n;
p += n;
len -= n;
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
if (dctx->partial_size) {
n = min(len, VMAC_NHBYTES - dctx->partial_size);
memcpy(&dctx->partial[dctx->partial_size], p, n);
dctx->partial_size += n;
p += n;
len -= n;
if (dctx->partial_size == VMAC_NHBYTES) {
vhash_blocks(tctx, dctx, dctx->partial_words, 1);
dctx->partial_size = 0;
}
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
if (len >= VMAC_NHBYTES) {
n = round_down(len, VMAC_NHBYTES);
/* TODO: 'p' may be misaligned here */
vhash_blocks(tctx, dctx, (const __le64 *)p, n / VMAC_NHBYTES);
p += n;
len -= n;
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
if (len) {
memcpy(dctx->partial, p, len);
dctx->partial_size = len;
}
return 0;
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
static u64 vhash_final(const struct vmac_tfm_ctx *tctx,
struct vmac_desc_ctx *dctx)
{
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
unsigned int partial = dctx->partial_size;
u64 ch = dctx->polytmp[0];
u64 cl = dctx->polytmp[1];
/* L1 and L2-hash the final block if needed */
if (partial) {
/* Zero-pad to next 128-bit boundary */
unsigned int n = round_up(partial, 16);
u64 rh, rl;
memset(&dctx->partial[partial], 0, n - partial);
nh_16(dctx->partial_words, tctx->nhkey, n / 8, rh, rl);
rh &= m62;
if (dctx->first_block_processed)
poly_step(ch, cl, tctx->polykey[0], tctx->polykey[1],
rh, rl);
else
ADD128(ch, cl, rh, rl);
}
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
/* L3-hash the 128-bit output of L2-hash */
return l3hash(ch, cl, tctx->l3key[0], tctx->l3key[1], partial * 8);
}
static int vmac_final(struct shash_desc *desc, u8 *out)
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
{
const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
int index;
u64 hash, pad;
if (dctx->nonce_size != VMAC_NONCEBYTES)
return -EINVAL;
/*
* The VMAC specification requires a nonce at least 1 bit shorter than
* the block cipher's block length, so we actually only accept a 127-bit
* nonce. We define the unused bit to be the first one and require that
* it be 0, so the needed prepending of a 0 bit is implicit.
*/
if (dctx->nonce.bytes[0] & 0x80)
return -EINVAL;
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
/* Finish calculating the VHASH of the message */
hash = vhash_final(tctx, dctx);
/* Generate pseudorandom pad by encrypting the nonce */
BUILD_BUG_ON(VMAC_NONCEBYTES != 2 * (VMAC_TAG_LEN / 8));
index = dctx->nonce.bytes[VMAC_NONCEBYTES - 1] & 1;
dctx->nonce.bytes[VMAC_NONCEBYTES - 1] &= ~1;
crypto_cipher_encrypt_one(tctx->cipher, dctx->nonce.bytes,
dctx->nonce.bytes);
pad = be64_to_cpu(dctx->nonce.pads[index]);
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
/* The VMAC is the sum of VHASH and the pseudorandom pad */
put_unaligned_be64(hash + pad, out);
return 0;
}
static int vmac_init_tfm(struct crypto_tfm *tfm)
{
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
struct crypto_instance *inst = crypto_tfm_alg_instance(tfm);
struct crypto_cipher_spawn *spawn = crypto_instance_ctx(inst);
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
struct vmac_tfm_ctx *tctx = crypto_tfm_ctx(tfm);
struct crypto_cipher *cipher;
cipher = crypto_spawn_cipher(spawn);
if (IS_ERR(cipher))
return PTR_ERR(cipher);
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
tctx->cipher = cipher;
return 0;
}
static void vmac_exit_tfm(struct crypto_tfm *tfm)
{
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
struct vmac_tfm_ctx *tctx = crypto_tfm_ctx(tfm);
crypto_free_cipher(tctx->cipher);
}
static int vmac_create(struct crypto_template *tmpl, struct rtattr **tb)
{
struct shash_instance *inst;
struct crypto_cipher_spawn *spawn;
struct crypto_alg *alg;
u32 mask;
int err;
err = crypto_check_attr_type(tb, CRYPTO_ALG_TYPE_SHASH, &mask);
if (err)
return err;
inst = kzalloc(sizeof(*inst) + sizeof(*spawn), GFP_KERNEL);
if (!inst)
return -ENOMEM;
spawn = shash_instance_ctx(inst);
err = crypto_grab_cipher(spawn, shash_crypto_instance(inst),
crypto_attr_alg_name(tb[1]), 0, mask);
if (err)
goto err_free_inst;
alg = crypto_spawn_cipher_alg(spawn);
err = -EINVAL;
if (alg->cra_blocksize != VMAC_NONCEBYTES)
goto err_free_inst;
err = crypto_inst_setname(shash_crypto_instance(inst), tmpl->name, alg);
if (err)
goto err_free_inst;
inst->alg.base.cra_priority = alg->cra_priority;
inst->alg.base.cra_blocksize = alg->cra_blocksize;
inst->alg.base.cra_alignmask = alg->cra_alignmask;
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
inst->alg.base.cra_ctxsize = sizeof(struct vmac_tfm_ctx);
inst->alg.base.cra_init = vmac_init_tfm;
inst->alg.base.cra_exit = vmac_exit_tfm;
crypto: vmac - separate tfm and request context syzbot reported a crash in vmac_final() when multiple threads concurrently use the same "vmac(aes)" transform through AF_ALG. The bug is pretty fundamental: the VMAC template doesn't separate per-request state from per-tfm (per-key) state like the other hash algorithms do, but rather stores it all in the tfm context. That's wrong. Also, vmac_final() incorrectly zeroes most of the state including the derived keys and cached pseudorandom pad. Therefore, only the first VMAC invocation with a given key calculates the correct digest. Fix these bugs by splitting the per-tfm state from the per-request state and using the proper init/update/final sequencing for requests. Reproducer for the crash: #include <linux/if_alg.h> #include <sys/socket.h> #include <unistd.h> int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "vmac(aes)", }; char buf[256] = { 0 }; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); setsockopt(fd, SOL_ALG, ALG_SET_KEY, buf, 16); fork(); fd = accept(fd, NULL, NULL); for (;;) write(fd, buf, 256); } The immediate cause of the crash is that vmac_ctx_t.partial_size exceeds VMAC_NHBYTES, causing vmac_final() to memset() a negative length. Reported-by: syzbot+264bca3a6e8d645550d3@syzkaller.appspotmail.com Fixes: f1939f7c5645 ("crypto: vmac - New hash algorithm for intel_txt support") Cc: <stable@vger.kernel.org> # v2.6.32+ Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2018-06-19 01:22:38 +08:00
inst->alg.descsize = sizeof(struct vmac_desc_ctx);
inst->alg.digestsize = VMAC_TAG_LEN / 8;
inst->alg.init = vmac_init;
inst->alg.update = vmac_update;
inst->alg.final = vmac_final;
inst->alg.setkey = vmac_setkey;
inst->free = shash_free_singlespawn_instance;
err = shash_register_instance(tmpl, inst);
if (err) {
err_free_inst:
shash_free_singlespawn_instance(inst);
}
return err;
}
static struct crypto_template vmac64_tmpl = {
.name = "vmac64",
.create = vmac_create,
.module = THIS_MODULE,
};
static int __init vmac_module_init(void)
{
return crypto_register_template(&vmac64_tmpl);
}
static void __exit vmac_module_exit(void)
{
crypto_unregister_template(&vmac64_tmpl);
}
subsys_initcall(vmac_module_init);
module_exit(vmac_module_exit);
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("VMAC hash algorithm");
MODULE_ALIAS_CRYPTO("vmac64");
MODULE_IMPORT_NS(CRYPTO_INTERNAL);