In order to immediately overwrite the old key on the stack, before
servicing a userspace request for bytes, we use the remaining 32 bytes
of block 0 as the key. This means moving indices 8,9,a,b,c,d,e,f ->
4,5,6,7,8,9,a,b. Since 4 < 8, for the kernel implementations of
memcpy(), this doesn't actually appear to be a problem in practice. But
relying on that characteristic seems a bit brittle. So let's change that
to a proper memmove(), which is the by-the-books way of handling
overlapping memory copies.

Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
---
 drivers/char/random.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/char/random.c b/drivers/char/random.c
index 6b01b2be9dd4..3a293f919af9 100644
--- a/drivers/char/random.c
+++ b/drivers/char/random.c
@@ -333,7 +333,7 @@ static void crng_fast_key_erasure(u8 key[CHACHA_KEY_SIZE],
 	chacha20_block(chacha_state, first_block);
 
 	memcpy(key, first_block, CHACHA_KEY_SIZE);
-	memcpy(random_data, first_block + CHACHA_KEY_SIZE, random_data_len);
+	memmove(random_data, first_block + CHACHA_KEY_SIZE, random_data_len);
 	memzero_explicit(first_block, sizeof(first_block));
 }
 
-- 
2.35.1

On Thu, Apr 14, 2022 at 01:56:49AM +0200, Jason A. Donenfeld wrote:
> In order to immediately overwrite the old key on the stack, before
> servicing a userspace request for bytes, we use the remaining 32 bytes
> of block 0 as the key. This means moving indices 8,9,a,b,c,d,e,f ->
> 4,5,6,7,8,9,a,b. Since 4 < 8, for the kernel implementations of
> memcpy(), this doesn't actually appear to be a problem in practice. But
> relying on that characteristic seems a bit brittle. So let's change that
> to a proper memmove(), which is the by-the-books way of handling
> overlapping memory copies.
> 
> Cc: Dominik Brodowski <linux@dominikbrodowski.net>
> Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
> ---
>  drivers/char/random.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/char/random.c b/drivers/char/random.c
> index 6b01b2be9dd4..3a293f919af9 100644
> --- a/drivers/char/random.c
> +++ b/drivers/char/random.c
> @@ -333,7 +333,7 @@ static void crng_fast_key_erasure(u8 key[CHACHA_KEY_SIZE],
>  	chacha20_block(chacha_state, first_block);
>  
>  	memcpy(key, first_block, CHACHA_KEY_SIZE);
> -	memcpy(random_data, first_block + CHACHA_KEY_SIZE, random_data_len);
> +	memmove(random_data, first_block + CHACHA_KEY_SIZE, random_data_len);
>  	memzero_explicit(first_block, sizeof(first_block));
>  }

first_block is on the stack, so this is never an overlapping copy.

It would be more important to document the fact that random_data can point into
chacha_state, as this is not obvious.

- Eric

Hey Eric,

On Mon, Apr 18, 2022 at 11:43:11AM -0700, Eric Biggers wrote:
> On Thu, Apr 14, 2022 at 01:56:49AM +0200, Jason A. Donenfeld wrote:
> > In order to immediately overwrite the old key on the stack, before
> > servicing a userspace request for bytes, we use the remaining 32 bytes
> > of block 0 as the key. This means moving indices 8,9,a,b,c,d,e,f ->
> > 4,5,6,7,8,9,a,b. Since 4 < 8, for the kernel implementations of
> > memcpy(), this doesn't actually appear to be a problem in practice. But
> > relying on that characteristic seems a bit brittle. So let's change that
> > to a proper memmove(), which is the by-the-books way of handling
> > overlapping memory copies.
> > 
> > Cc: Dominik Brodowski <linux@dominikbrodowski.net>
> > Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
> > ---
> >  drivers/char/random.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/char/random.c b/drivers/char/random.c
> > index 6b01b2be9dd4..3a293f919af9 100644
> > --- a/drivers/char/random.c
> > +++ b/drivers/char/random.c
> > @@ -333,7 +333,7 @@ static void crng_fast_key_erasure(u8 key[CHACHA_KEY_SIZE],
> >  	chacha20_block(chacha_state, first_block);
> >  
> >  	memcpy(key, first_block, CHACHA_KEY_SIZE);
> > -	memcpy(random_data, first_block + CHACHA_KEY_SIZE, random_data_len);
> > +	memmove(random_data, first_block + CHACHA_KEY_SIZE, random_data_len);
> >  	memzero_explicit(first_block, sizeof(first_block));
> >  }
> 
> first_block is on the stack, so this is never an overlapping copy.

Oh, grrr, yes you're right. I clearly over thought this into non-sense.
Will revert.

> 
> It would be more important to document the fact that random_data can point into
> chacha_state, as this is not obvious.

Good idea. I'll do that.

Jason

Am Thu, Apr 14, 2022 at 01:56:49AM +0200 schrieb Jason A. Donenfeld:
> In order to immediately overwrite the old key on the stack, before
> servicing a userspace request for bytes, we use the remaining 32 bytes
> of block 0 as the key. This means moving indices 8,9,a,b,c,d,e,f ->
> 4,5,6,7,8,9,a,b. Since 4 < 8, for the kernel implementations of
> memcpy(), this doesn't actually appear to be a problem in practice. But
> relying on that characteristic seems a bit brittle. So let's change that
> to a proper memmove(), which is the by-the-books way of handling
> overlapping memory copies.
> 
> Cc: Dominik Brodowski <linux@dominikbrodowski.net>
> Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>

	Reviewed-by: Dominik Brodowski <linux@dominikbrodowski.net>

Thanks,
	Dominik