Re: [PATCH] zsh/random module [UPDATED]

[Clinton Bunch <cdb_zsh@xxxxxxxxxxx>]

On 11/24/2022 4:07 AM, nimaje+zml@xxxxxxxxxxxxxx wrote:
> On 11/24/22 03:58, Clinton Bunch wrote:
> > [snip]
> > > > +/**/
> > > > +int
> > > > +setup_(UNUSED(Module m))
> > > > +{
> > > > +#ifdef USE_URANDOM
> > > > +    /* Check for the existence of /dev/urandom */
> > > > +
> > > > +    struct stat st;
> > > > +
> > > > +    if (lstat("/dev/urandom",&st) < 0) {
> > > Why not stat()?
> > Is it appropriate for /dev/urandom to be a symlink?
>
> Yes, for example that is the case on freebsd where urandom is just a 
> symlink to random for compatibility with programs that expect urandom 
> to be there (random blocks until the randomness pool is seeded and 
> then doesn't block anymore).
On FreeBSD it doesn't use urandom, so it didn't show up in my testing.  
It only uses urandom if arc4random() or getrandom() doesn't exist
> > > > +    zwarn("No kernel random pool found.");
> > > > +    return 1;
> > > > +    }
> > > > +
> > > > +    if (!(S_ISCHR(st.st_mode)) ) {
> > > > +    zwarn("No kernel random pool found.");
> > > > +    return 1;
> > > > +    }
>
> So this check would be wrong on freebsd (or use random directly and 
> give an option to use urandom on systems where random starts to block 
> for some reason after being seeded, but no idea if there is a good 
> reason for random being a symlink).
On most systems random is defined to block if the randomness pool falls 
below a certain threshold until more random data is available
> > [snip]