Re: [PATCH RFC] random: introduce get_random_bytes_busy_wait_initialized

[Hannes Frederic Sowa <hannes@stressinduktion.org>]

On Wed, Sep 25, 2013 at 05:06:50AM -0700, Eric Dumazet wrote:
> On Wed, 2013-09-25 at 11:00 +0200, Hannes Frederic Sowa wrote:
> 
> >  /*
> > + * Busy loop until the nonblocking_pool is intialized and return
> > + * random data in buf of size nbytes.
> > + *
> > + * This is used by the network stack to defer the extraction of
> > + * entropy from the nonblocking_pool until the pool is initialized.
> > + *
> > + * We need to busy loop here, because we could be called from an
> > + * atomic section.
> > + */
> > +void get_random_bytes_busy_wait_initialized(void *buf, int nbytes)
> > +{
> > +	while (!nonblocking_pool.initialized)
> > +		cpu_relax();
> > +	get_random_bytes(buf, nbytes);
> > +}
> 
> No idea if this can work if called from IRQ context.
> 
> How is nonblocking_poll initialized if host has a single cpu ?

We increase the entropy_count and can initialize the random pool from
handle_irq_event.

We can document that it should not get called from irq context and
maybe add a WARN_ON(irqs_disabled())?