Skip to contentSkip to conversation

Fix xoring of arch_get_random_long into crng->state array

PATCH In review Neil Horman 3 participants 8 messages Apr 2, 2019 – May 30, 2019 Random number driver

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix xoring of arch_get_random_long into crng->state array

When _crng_extract is called, any arch that has a registered arch_get_random_long method, attempts to mix an unsigned long value into the crng->state buffer, it only mixes in 32 of the 64 bits available, because the state buffer is an array of u32 values, even though 2 u32 are expected to be filled (owing to the fact that it expects indexes 14 and 15 to be filled).

Bring the expected behavior into alignment by casting index 14 to an unsignled long pointer, and xoring that in instead.

Tested successfully by myself

CC "Theodore Ts'o" <[email protected]>
CC Arnd Bergmann <[email protected]>
CC Greg Kroah-Hartman <[email protected]>
---
 drivers/char/random.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/char/random.c +5 -3
index 38c6d1af6d1c..8178618458ac 100644
@@ -975,14 +975,16 @@ static void _extract_crng(struct crng_state *crng,
 			  __u8 out[CHACHA_BLOCK_SIZE])
 {
 	unsigned long v, flags;
-
+	unsigned long *archrnd;
 	if (crng_ready() &&
 	    (time_after(crng_global_init_time, crng->init_time) ||
 	     time_after(jiffies, crng->init_time + CRNG_RESEED_INTERVAL)))
 		crng_reseed(crng, crng == &primary_crng ? &input_pool : NULL);
 	spin_lock_irqsave(&crng->lock, flags);
-	if (arch_get_random_long(&v))
-		crng->state[14] ^= v;
+	if (arch_get_random_long(&v)) {
+		archrnd = (unsigned long *)&crng->state[14];
+		*archrnd ^= v;
+	}
 	chacha20_block(&crng->state[0], out);
 	if (crng->state[12] == 0)
 		crng->state[13]++;
Signature · 1 line
2.20.1
#1 lore raw
Quoted 49 lines · Neil Horman

When _crng_extract is called, any arch that has a registered arch_get_random_long method, attempts to mix an unsigned long value into the crng->state buffer, it only mixes in 32 of the 64 bits available, because the state buffer is an array of u32 values, even though 2 u32 are expected to be filled (owing to the fact that it expects indexes 14 and 15 to be filled).

Bring the expected behavior into alignment by casting index 14 to an unsignled long pointer, and xoring that in instead.

Tested successfully by myself

CC "Theodore Ts'o" <[email protected]>
CC Arnd Bergmann <[email protected]>
CC Greg Kroah-Hartman <[email protected]>
---
 drivers/char/random.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
drivers/char/random.c +5 -3
index 38c6d1af6d1c..8178618458ac 100644
@@ -975,14 +975,16 @@ static void _extract_crng(struct crng_state *crng,
 			  __u8 out[CHACHA_BLOCK_SIZE])
 {
 	unsigned long v, flags;
-
+	unsigned long *archrnd;
 	if (crng_ready() &&
 	    (time_after(crng_global_init_time, crng->init_time) ||
 	     time_after(jiffies, crng->init_time + CRNG_RESEED_INTERVAL)))
 		crng_reseed(crng, crng == &primary_crng ? &input_pool : NULL);
 	spin_lock_irqsave(&crng->lock, flags);
-	if (arch_get_random_long(&v))
-		crng->state[14] ^= v;
+	if (arch_get_random_long(&v)) {
+		archrnd = (unsigned long *)&crng->state[14];
+		*archrnd ^= v;
+	}
 	chacha20_block(&crng->state[0], out);
 	if (crng->state[12] == 0)
 		crng->state[13]++;
Signature · 1 line
2.20.1

Ping, Arnd, Ted, Greg, any comment here?
Neil

#2 lore raw

From: Neil Horman

Quoted 11 lines

Sent: 29 May 2019 14:42

Quoted 9 lines · Neil Horman

When _crng_extract is called, any arch that has a registered arch_get_random_long method, attempts to mix an unsigned long value into the crng->state buffer, it only mixes in 32 of the 64 bits available, because the state buffer is an array of u32 values, even though 2 u32 are expected to be filled (owing to the fact that it expects indexes 14 and 15 to be filled).

Bring the expected behavior into alignment by casting index 14 to an unsignled long pointer, and xoring that in instead.

...

Quoted 13 lines
Quoted 13 lines
drivers/char/random.c +1 -1
index 38c6d1af6d1c..8178618458ac 100644
@@ -975,14 +975,16 @@ static void _extract_crng(struct crng_state *crng,
 			  __u8 out[CHACHA_BLOCK_SIZE])
 {
 	unsigned long v, flags;
-
+	unsigned long *archrnd;
 	if (crng_ready() &&
 	    (time_after(crng_global_init_time, crng->init_time) ||
 	     time_after(jiffies, crng->init_time + CRNG_RESEED_INTERVAL)))
Quoted 8 lines

drivers/char/random.c static void _extract_crng(struct crng_state *crng, · line 987

Quoted 8 lines
drivers/char/random.c +4 -2
 		crng_reseed(crng, crng == &primary_crng ? &input_pool : NULL);
 	spin_lock_irqsave(&crng->lock, flags);
-	if (arch_get_random_long(&v))
-		crng->state[14] ^= v;
+	if (arch_get_random_long(&v)) {
+		archrnd = (unsigned long *)&crng->state[14];
+		*archrnd ^= v;
+	}

Isn't that likely to generate a misaligned memory access?

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

#3 lore raw
Quoted 28 lines · David Laight

From: Neil Horman

Quoted 11 lines

Sent: 29 May 2019 14:42

Quoted 9 lines · Neil Horman

When _crng_extract is called, any arch that has a registered arch_get_random_long method, attempts to mix an unsigned long value into the crng->state buffer, it only mixes in 32 of the 64 bits available, because the state buffer is an array of u32 values, even though 2 u32 are expected to be filled (owing to the fact that it expects indexes 14 and 15 to be filled).

Bring the expected behavior into alignment by casting index 14 to an unsignled long pointer, and xoring that in instead.

...

Quoted 15 lines
Quoted 15 lines
drivers/char/random.c +1 -1
index 38c6d1af6d1c..8178618458ac 100644
@@ -975,14 +975,16 @@ static void _extract_crng(struct crng_state *crng,
 			  __u8 out[CHACHA_BLOCK_SIZE])
 {
 	unsigned long v, flags;
-
+	unsigned long *archrnd;
 	if (crng_ready() &&
 	    (time_after(crng_global_init_time, crng->init_time) ||
 	     time_after(jiffies, crng->init_time + CRNG_RESEED_INTERVAL)))
 		crng_reseed(crng, crng == &primary_crng ? &input_pool : NULL);
 	spin_lock_irqsave(&crng->lock, flags);
Quoted 9 lines
Quoted 6 lines

drivers/char/random.c static void _extract_crng(struct crng_state *crng, · line 987

Quoted 6 lines
drivers/char/random.c +4 -2
-	if (arch_get_random_long(&v))
-		crng->state[14] ^= v;
+	if (arch_get_random_long(&v)) {
+		archrnd = (unsigned long *)&crng->state[14];
+		*archrnd ^= v;
+	}

Isn't that likely to generate a misaligned memory access?

I'm not quite sure how it would, crng->state is an array of _u32's, and so every even element should be on a 64 bit boundary.

Neil

Quoted 7 lines

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

#4 lore raw

From: Neil Horman [mailto:[email protected]]

Quoted 39 lines

Sent: 29 May 2019 16:52

Quoted 28 lines · David Laight

From: Neil Horman

Quoted 11 lines

Sent: 29 May 2019 14:42

Quoted 9 lines · Neil Horman

When _crng_extract is called, any arch that has a registered arch_get_random_long method, attempts to mix an unsigned long value into the crng->state buffer, it only mixes in 32 of the 64 bits available, because the state buffer is an array of u32 values, even though 2 u32 are expected to be filled (owing to the fact that it expects indexes 14 and 15 to be filled).

Bring the expected behavior into alignment by casting index 14 to an unsignled long pointer, and xoring that in instead.

...

Quoted 15 lines
Quoted 15 lines
drivers/char/random.c +1 -1
index 38c6d1af6d1c..8178618458ac 100644
@@ -975,14 +975,16 @@ static void _extract_crng(struct crng_state *crng,
 			  __u8 out[CHACHA_BLOCK_SIZE])
 {
 	unsigned long v, flags;
-
+	unsigned long *archrnd;
 	if (crng_ready() &&
 	    (time_after(crng_global_init_time, crng->init_time) ||
 	     time_after(jiffies, crng->init_time + CRNG_RESEED_INTERVAL)))
 		crng_reseed(crng, crng == &primary_crng ? &input_pool : NULL);
 	spin_lock_irqsave(&crng->lock, flags);
Quoted 9 lines
Quoted 6 lines

drivers/char/random.c static void _extract_crng(struct crng_state *crng, · line 987

Quoted 6 lines
drivers/char/random.c +4 -2
-	if (arch_get_random_long(&v))
-		crng->state[14] ^= v;
+	if (arch_get_random_long(&v)) {
+		archrnd = (unsigned long *)&crng->state[14];
+		*archrnd ^= v;
+	}

Isn't that likely to generate a misaligned memory access?

Quoted 2 lines

I'm not quite sure how it would, crng->state is an array of _u32's, and so every even element should be on a 64 bit boundary.

Only if the first item is aligned....
Add a u32 before it and you'll probably flip the alignment.

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

#5 lore raw
Quoted 40 lines · David Laight

From: Neil Horman [mailto:[email protected]]

Quoted 39 lines

Sent: 29 May 2019 16:52

Quoted 37 lines · David Laight

From: Neil Horman

Quoted 11 lines

Sent: 29 May 2019 14:42

Quoted 9 lines · Neil Horman

When _crng_extract is called, any arch that has a registered arch_get_random_long method, attempts to mix an unsigned long value into the crng->state buffer, it only mixes in 32 of the 64 bits available, because the state buffer is an array of u32 values, even though 2 u32 are expected to be filled (owing to the fact that it expects indexes 14 and 15 to be filled).

Bring the expected behavior into alignment by casting index 14 to an unsignled long pointer, and xoring that in instead.

...

Quoted 13 lines
Quoted 13 lines
drivers/char/random.c +1 -1
index 38c6d1af6d1c..8178618458ac 100644
@@ -975,14 +975,16 @@ static void _extract_crng(struct crng_state *crng,
 			  __u8 out[CHACHA_BLOCK_SIZE])
 {
 	unsigned long v, flags;
-
+	unsigned long *archrnd;
 	if (crng_ready() &&
 	    (time_after(crng_global_init_time, crng->init_time) ||
 	     time_after(jiffies, crng->init_time + CRNG_RESEED_INTERVAL)))
Quoted 8 lines

drivers/char/random.c static void _extract_crng(struct crng_state *crng, · line 987

Quoted 8 lines
drivers/char/random.c +4 -2
 		crng_reseed(crng, crng == &primary_crng ? &input_pool : NULL);
 	spin_lock_irqsave(&crng->lock, flags);
-	if (arch_get_random_long(&v))
-		crng->state[14] ^= v;
+	if (arch_get_random_long(&v)) {
+		archrnd = (unsigned long *)&crng->state[14];
+		*archrnd ^= v;
+	}

Isn't that likely to generate a misaligned memory access?

Quoted 6 lines
Quoted 2 lines

I'm not quite sure how it would, crng->state is an array of _u32's, and so every even element should be on a 64 bit boundary.

Only if the first item is aligned....
Add a u32 before it and you'll probably flip the alignment.

Sure (assuming no padding by the compiler of leading elements), but thats not the case here, state is the first element in the array. I suppose we could add an __attribute__((aligned,8)) to the element if you think it would help

Neil

Quoted 7 lines

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)

maintainer
#6 lore raw
In reply to Neil Horman
Quoted 6 lines · Neil Horman

When _crng_extract is called, any arch that has a registered arch_get_random_long method, attempts to mix an unsigned long value into the crng->state buffer, it only mixes in 32 of the 64 bits available, because the state buffer is an array of u32 values, even though 2 u32 are expected to be filled (owing to the fact that it expects indexes 14 and 15 to be filled).

Index 15 does get initialized; in fact, it's changed each time crng_reseed() is called.

The way things currently work is that we use state[12] and state[13] as 64-bit counter (it gets incremented each time we call _extract_crng), and state[14] and state[15] are nonce values. After crng->state has been in use for five minutes, we reseed the crng by grabbing randomness from the input pool, and using that to initialize state[4..15]. (State[0..3] are always set to the ChaCha20 constant of "expand 32-byte k".)

If the CPU provides and RDRAND-like instruction (which can be the case for x86, PPC, and S390), we xor it into state[14]. Whether we xor any extra entropy into state[15] to be honest, really doesn't matter much. I think I was trying to keep things simple, and it wasn't worth it to call RDRAND twice on a 32-bit x86. (And there isn't an arch_get_random_long_long. :-)

Why do we do this at all? Well, the goal was to feed in some contributing randomness from RDRAND when we turn the CRNG crank. (The reason why we don't just XOR in the RDRAND into the output ohf the CRNG is mainly to assuage critics that hypothetical RDRAND backdoor has access to the CPU registers. So we perturb the inputs to the CRNG, on the theory that if malicious firmware can reverse CHACHA20... we've got bigger problems. :-) We get up to 20 bytes out of a single turn of the CRNG crank, so whether we mix in 4 bytes or 8 bytes from RDRAND, we're never going to be depending on RDRAND completely in any case.

The bottom line is that I'm not at all convinced it worth the effort to mix in 8 bytes versus 4 bytes from RDRAND. This is really a CRNG, and the RDRAND inputs really don't change that.

- Ted

#7 lore raw
Quoted 37 lines · Theodore Ts'o
Quoted 6 lines · Neil Horman

When _crng_extract is called, any arch that has a registered arch_get_random_long method, attempts to mix an unsigned long value into the crng->state buffer, it only mixes in 32 of the 64 bits available, because the state buffer is an array of u32 values, even though 2 u32 are expected to be filled (owing to the fact that it expects indexes 14 and 15 to be filled).

Index 15 does get initialized; in fact, it's changed each time crng_reseed() is called.

The way things currently work is that we use state[12] and state[13] as 64-bit counter (it gets incremented each time we call _extract_crng), and state[14] and state[15] are nonce values. After crng->state has been in use for five minutes, we reseed the crng by grabbing randomness from the input pool, and using that to initialize state[4..15]. (State[0..3] are always set to the ChaCha20 constant of "expand 32-byte k".)

If the CPU provides and RDRAND-like instruction (which can be the case for x86, PPC, and S390), we xor it into state[14]. Whether we xor any extra entropy into state[15] to be honest, really doesn't matter much. I think I was trying to keep things simple, and it wasn't worth it to call RDRAND twice on a 32-bit x86. (And there isn't an arch_get_random_long_long. :-)

Why do we do this at all? Well, the goal was to feed in some contributing randomness from RDRAND when we turn the CRNG crank. (The reason why we don't just XOR in the RDRAND into the output ohf the CRNG is mainly to assuage critics that hypothetical RDRAND backdoor has access to the CPU registers. So we perturb the inputs to the CRNG, on the theory that if malicious firmware can reverse CHACHA20... we've got bigger problems. :-) We get up to 20 bytes out of a single turn of the CRNG crank, so whether we mix in 4 bytes or 8 bytes from RDRAND, we're never going to be depending on RDRAND completely in any case.

Quoted 4 lines

The bottom line is that I'm not at all convinced it worth the effort to mix in 8 bytes versus 4 bytes from RDRAND. This is really a CRNG, and the RDRAND inputs really don't change that.

Ok, so what I'm getting is that the exclusion of the second 32 bit word here from &crng->state[15], isn't an oversight, its just skipped because its not worth taking the time for the extra write there, and this is not a bug. I'm ok with that.

Thanks for the explination
Neil

Quoted 2 lines

- Ted