net: Compute protocol sequence numbers and fragment IDs using MD5.

Computers have become a lot faster since we compromised on the
partial MD4 hash which we use currently for performance reasons.

MD5 is a much safer choice, and is inline with both RFC1948 and
other ISS generators (OpenBSD, Solaris, etc.)

Furthermore, only having 24-bits of the sequence number be truly
unpredictable is a very serious limitation.  So the periodic
regeneration and 8-bit counter have been removed.  We compute and
use a full 32-bit sequence number.

For ipv6, DCCP was found to use a 32-bit truncated initial sequence
number (it needs 43-bits) and that is fixed here as well.

Reported-by: Dan Kaminsky <dan@doxpara.com>
Tested-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>

14 files changed, 223 insertions, 361 deletions

diff --git a/drivers/char/random.c b/drivers/char/random.c
index 729281961f22..c35a785005b0 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -1300,345 +1300,14 @@ ctl_table random_table[] = {
 };
 #endif  /* CONFIG_SYSCTL */
 
-/********************************************************************
- *
- * Random functions for networking
- *
- ********************************************************************/
-
-/*
- * TCP initial sequence number picking.  This uses the random number
- * generator to pick an initial secret value.  This value is hashed
- * along with the TCP endpoint information to provide a unique
- * starting point for each pair of TCP endpoints.  This defeats
- * attacks which rely on guessing the initial TCP sequence number.
- * This algorithm was suggested by Steve Bellovin.
- *
- * Using a very strong hash was taking an appreciable amount of the total
- * TCP connection establishment time, so this is a weaker hash,
- * compensated for by changing the secret periodically.
- */
-
-/* F, G and H are basic MD4 functions: selection, majority, parity */
-#define F(x, y, z) ((z) ^ ((x) & ((y) ^ (z))))
-#define G(x, y, z) (((x) & (y)) + (((x) ^ (y)) & (z)))
-#define H(x, y, z) ((x) ^ (y) ^ (z))
-
-/*
- * The generic round function.  The application is so specific that
- * we don't bother protecting all the arguments with parens, as is generally
- * good macro practice, in favor of extra legibility.
- * Rotation is separate from addition to prevent recomputation
- */
-#define ROUND(f, a, b, c, d, x, s)      \
-        (a += f(b, c, d) + x, a = (a << s) | (a >> (32 - s)))
-#define K1 0
-#define K2 013240474631UL
-#define K3 015666365641UL
-
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
-
-static __u32 twothirdsMD4Transform(__u32 const buf[4], __u32 const in[12])
-{
-        __u32 a = buf[0], b = buf[1], c = buf[2], d = buf[3];
-
-        /* Round 1 */
-        ROUND(F, a, b, c, d, in[ 0] + K1,  3);
-        ROUND(F, d, a, b, c, in[ 1] + K1,  7);
-        ROUND(F, c, d, a, b, in[ 2] + K1, 11);
-        ROUND(F, b, c, d, a, in[ 3] + K1, 19);
-        ROUND(F, a, b, c, d, in[ 4] + K1,  3);
-        ROUND(F, d, a, b, c, in[ 5] + K1,  7);
-        ROUND(F, c, d, a, b, in[ 6] + K1, 11);
-        ROUND(F, b, c, d, a, in[ 7] + K1, 19);
-        ROUND(F, a, b, c, d, in[ 8] + K1,  3);
-        ROUND(F, d, a, b, c, in[ 9] + K1,  7);
-        ROUND(F, c, d, a, b, in[10] + K1, 11);
-        ROUND(F, b, c, d, a, in[11] + K1, 19);
-
-        /* Round 2 */
-        ROUND(G, a, b, c, d, in[ 1] + K2,  3);
-        ROUND(G, d, a, b, c, in[ 3] + K2,  5);
-        ROUND(G, c, d, a, b, in[ 5] + K2,  9);
-        ROUND(G, b, c, d, a, in[ 7] + K2, 13);
-        ROUND(G, a, b, c, d, in[ 9] + K2,  3);
-        ROUND(G, d, a, b, c, in[11] + K2,  5);
-        ROUND(G, c, d, a, b, in[ 0] + K2,  9);
-        ROUND(G, b, c, d, a, in[ 2] + K2, 13);
-        ROUND(G, a, b, c, d, in[ 4] + K2,  3);
-        ROUND(G, d, a, b, c, in[ 6] + K2,  5);
-        ROUND(G, c, d, a, b, in[ 8] + K2,  9);
-        ROUND(G, b, c, d, a, in[10] + K2, 13);
-
-        /* Round 3 */
-        ROUND(H, a, b, c, d, in[ 3] + K3,  3);
-        ROUND(H, d, a, b, c, in[ 7] + K3,  9);
-        ROUND(H, c, d, a, b, in[11] + K3, 11);
-        ROUND(H, b, c, d, a, in[ 2] + K3, 15);
-        ROUND(H, a, b, c, d, in[ 6] + K3,  3);
-        ROUND(H, d, a, b, c, in[10] + K3,  9);
-        ROUND(H, c, d, a, b, in[ 1] + K3, 11);
-        ROUND(H, b, c, d, a, in[ 5] + K3, 15);
-        ROUND(H, a, b, c, d, in[ 9] + K3,  3);
-        ROUND(H, d, a, b, c, in[ 0] + K3,  9);
-        ROUND(H, c, d, a, b, in[ 4] + K3, 11);
-        ROUND(H, b, c, d, a, in[ 8] + K3, 15);
-
-        return buf[1] + b; /* "most hashed" word */
-        /* Alternative: return sum of all words? */
-}
-#endif
-
-#undef ROUND
-#undef F
-#undef G
-#undef H
-#undef K1
-#undef K2
-#undef K3
-
-/* This should not be decreased so low that ISNs wrap too fast. */
-#define REKEY_INTERVAL (300 * HZ)
-/*
- * Bit layout of the tcp sequence numbers (before adding current time):
- * bit 24-31: increased after every key exchange
- * bit 0-23: hash(source,dest)
- *
- * The implementation is similar to the algorithm described
- * in the Appendix of RFC 1185, except that
- * - it uses a 1 MHz clock instead of a 250 kHz clock
- * - it performs a rekey every 5 minutes, which is equivalent
- *      to a (source,dest) tulple dependent forward jump of the
- *      clock by 0..2^(HASH_BITS+1)
- *
- * Thus the average ISN wraparound time is 68 minutes instead of
- * 4.55 hours.
- *
- * SMP cleanup and lock avoidance with poor man's RCU.
- *                      Manfred Spraul <manfred@colorfullife.com>
- *
- */
-#define COUNT_BITS 8
-#define COUNT_MASK ((1 << COUNT_BITS) - 1)
-#define HASH_BITS 24
-#define HASH_MASK ((1 << HASH_BITS) - 1)
+static u32 random_int_secret[MD5_MESSAGE_BYTES / 4] ____cacheline_aligned;
 
-static struct keydata {
-        __u32 count; /* already shifted to the final position */
-        __u32 secret[12];
-} ____cacheline_aligned ip_keydata[2];
-
-static unsigned int ip_cnt;
-
-static void rekey_seq_generator(struct work_struct *work);
-
-static DECLARE_DELAYED_WORK(rekey_work, rekey_seq_generator);
-
-/*
- * Lock avoidance:
- * The ISN generation runs lockless - it's just a hash over random data.
- * State changes happen every 5 minutes when the random key is replaced.
- * Synchronization is performed by having two copies of the hash function
- * state and rekey_seq_generator always updates the inactive copy.
- * The copy is then activated by updating ip_cnt.
- * The implementation breaks down if someone blocks the thread
- * that processes SYN requests for more than 5 minutes. Should never
- * happen, and even if that happens only a not perfectly compliant
- * ISN is generated, nothing fatal.
- */
-static void rekey_seq_generator(struct work_struct *work)
+static int __init random_int_secret_init(void)
 {
-        struct keydata *keyptr = &ip_keydata[1 ^ (ip_cnt & 1)];
-
-        get_random_bytes(keyptr->secret, sizeof(keyptr->secret));
-        keyptr->count = (ip_cnt & COUNT_MASK) << HASH_BITS;
-        smp_wmb();
-        ip_cnt++;
-        schedule_delayed_work(&rekey_work,
-                              round_jiffies_relative(REKEY_INTERVAL));
-}
-
-static inline struct keydata *get_keyptr(void)
-{
-        struct keydata *keyptr = &ip_keydata[ip_cnt & 1];
-
+        get_random_bytes(random_int_secret, sizeof(random_int_secret));