Re: [PATCH] random: Initialize vsprintf's pointer hash once the random core is ready.

[Jason A. Donenfeld]

Hi Sebastian,

On Fri, Jul 29, 2022 at 10:52:58AM +0200, Sebastian Andrzej Siewior wrote:
> --- a/drivers/char/random.c
> +++ b/drivers/char/random.c
> @@ -202,6 +202,7 @@ static void extract_entropy(void *buf, size_t len);
>  /* This extracts a new crng key from the input pool. */
>  static void crng_reseed(void)
>  {
> +	bool init_hash_pointer = false;
>  	unsigned long flags;
>  	unsigned long next_gen;
>  	u8 key[CHACHA_KEY_SIZE];
> @@ -221,10 +222,15 @@ static void crng_reseed(void)
>  		++next_gen;
>  	WRITE_ONCE(base_crng.generation, next_gen);
>  	WRITE_ONCE(base_crng.birth, jiffies);
> -	if (!static_branch_likely(&crng_is_ready))
> +	if (!static_branch_likely(&crng_is_ready)) {
>  		crng_init = CRNG_READY;
> +		init_hash_pointer = true;
> +	}
>  	spin_unlock_irqrestore(&base_crng.lock, flags);
>  	memzero_explicit(key, sizeof(key));
> +
> +	if (init_hash_pointer)
> +		vsprintf_init_hash_pointer();
>  }

Gumming up random.c with these sorts of things isn't alright. vsprintf
isn't special in any regard here.

If you can't do this from ordinary context inside of vsprintf, just
launch a workqueue to do it. This is already needed for changing
vsprintf's static branch, so just move the get_random_bytes() call into
there on RT (leaving it alone on non-RT, I guess).

Jason