On Wed, Sep 16, 2020 at 6:19 AM Eric Biggers <ebiggers@xxxxxxxxxx> wrote:
> The RNDRESEEDCRNG ioctl reseeds the primary_crng from itself, which
> doesn't make sense. Reseed it from the input_pool instead.
Good catch. (And its purpose is to ensure that entropy from
random_write() is plumbed all the way through such that getrandom()
and friends are guaranteed to actually use that entropy on the next
invocation; and random_write() just puts data into the input pool.)
But actually, looking at the surrounding code, I think there's another
small problem?
> Fixes: d848e5f8e1eb ("random: add new ioctl RNDRESEEDCRNG")
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Eric Biggers <ebiggers@xxxxxxxxxx>
Reviewed-by: Jann Horn <jannh@xxxxxxxxxx>
> ---
> 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 d20ba1b104ca3..a8b9e66f41435 100644
> --- a/drivers/char/random.c
> +++ b/drivers/char/random.c
> @@ -1973,7 +1973,7 @@ static long random_ioctl(struct file *f, unsigned int cmd, unsigned long arg)
> return -EPERM;
> if (crng_init < 2)
> return -ENODATA;
> - crng_reseed(&primary_crng, NULL);
> + crng_reseed(&primary_crng, &input_pool);
So now this will pull the data from the input_pool into the
primary_crng, so far so good...
> crng_global_init_time = jiffies - 1;
... and this will hopefully cause _extract_crng() to pull from the
primary_crng into crng_node_pool[numa_node_id()] afterwards. Unless
things are going too fast and therefore the jiffies haven't changed
since the last crng_reseed() on the crng_node_pool[numa_node_id()]...
a sequence number would probably be more robust than a timestamp.
And a plain write like this without holding any locks is also questionable.
The easiest way would probably be to make it an atomic_long_t, do
atomic_long_inc() instead of setting crng_global_init_time here, and
check atomic_long_read(...) against a copy stored in the crng_state on
_extract_crng()? And in crng_reseed(), grab the global sequence number
at the start, then do smp_rmb(), and then under the lock do the actual
reseeding and bump ->init_time? Or something like that?
> return 0;
> default:
