Re: [PATCH 2/2 v2] lib/vsprintf: Initialize vsprintf's pointer hash once the random core is ready.

[Sebastian Andrzej Siewior <bigeasy@linutronix.de>]

On 2022-07-30 01:29:12 [+0200], Jason A. Donenfeld wrote:
> On Fri, Jul 29, 2022 at 05:47:16PM +0200, Sebastian Andrzej Siewior wrote:
> > +static void fill_ptr_key_workfn(struct work_struct *work)
> > +{
> > +	int ret;
> > +
> > +	ret = get_random_bytes_wait(&ptr_key, sizeof(ptr_key));
> 
> > +static int vsprintf_init_hashval(void)
> > +{
> > +	static DECLARE_WORK(fill_ptr_key_work, fill_ptr_key_workfn);
> > +
> > +	queue_work(system_unbound_wq, &fill_ptr_key_work);
> > +	return 0;
> > +}
> > +subsys_initcall(vsprintf_init_hashval)
> 
> I'm unsure how good of an idea this is; it'll wind up setting off the
> jitter entropy thing very early in init. It's probably a better idea to
> just schedule the worker the first time that the RNG is already
> initialized by some other means. Check `in_hardirq()` or something if
> you're worried about missing the first message.

I'm aware of in_hardirq() and this not the only I have to worry about.
The same is true for interrupts-off, preempt-off, BH-disabled, rcu-read
section.
If you don't want me to use get_random_bytes_wait(), would you prefer to
delay it to late_initcall() or would you rather prefer to schedule a worker
every other second until the RNG is ready?

> Jason

Sebastian