Date: Mon, 18 Mar 1996 22:32:16 +0200 From: Mark Murray <mark@grondar.za> To: Bruce Evans <bde@zeta.org.au> Cc: current@freebsd.org, mark@grondar.za Subject: Re: /dev/random - an idea/proposal Message-ID: <199603182032.WAA06095@grumble.grondar.za>
next in thread | raw e-mail | index | archive | help
Bruce Evans wrote: > >I am not completely happy with my current method of selecting the > >interrupts for /dev/random right now. (There is rndcontrol(8) with > >some command-line switches, and sysconfig has a line to edit). It > >is too easy to overlook, and I am very keen that /dev/random be seen > >to provide a decent source of entropy. > > Neither am I, for the opposite reasons :-). The default sysconfig > setting of rand_irqs is almost useless except to print an annoying > message to announce rndcontrol (add_keyboard_randomness() calls > add_timer_randomness() a nearly constant number of usec before > add_interrupt_randomness() for the keyboard interrupt calls > add_timer_randomness()). I had a hard time understanding this. It took a few re-reads and I'm still only half-way :-(. Could you please spell this out a bit better? What I think you are saying is invalid. (I suspect you do not like the fact that two calls to add_*_randomness happen on top of each other (a nearly constant time apart) but are ignoring the fact that individual keystrokes will occur at random times apart, which is the whole purpose of using this interrupt.) What is a better solution? > >I would like some way to be able to choose these at kernel build > >time - ie at the time the system manager configures her kernel, > >she nominates that a particular device's interrupts should be used. > > >This I would like to do by adding a keyword "random", syntactically > >similar to "conflicts" to the config(8) grammar. This will then set > > This would be far too much trouble. As well as hacking config, you > would have to: > > - fix the problems that stop "conflicts" from working for PCI and > EISA devices > > - update userconfig and dset to support the new flag > > A kernel option to set a bitmap of the interrupts used to generate > randomness would be easier to implement, but it would be easier to > overlook than rndcontrol(8). Thats not much use, then. I am looking for a solution that is elegant. That implies the positive ideals of useful, reasonably obvious, and clean, as well as being an improvement on what we have now. Suggestions? M -- Mark Murray 46 Harvey Rd, Claremont, Cape Town 7700, South Africa +27 21 61-3768 GMT+0200 Finger mark@grondar.za for PGP key
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199603182032.WAA06095>