Re: [PATCH RFC #2] hwrng: Add type categories

[Michael Buesch <mb@bu3sch.de>]

On Wednesday 27 June 2007 15:05:33 Michael Buesch wrote:
> On Wednesday 27 June 2007 04:48:56 Henrique de Moraes Holschuh wrote:
> > (a) is just broken, unless one is to take it as "never use it".  And I am
> > really not sure about (b).  It *is* better than just using whatever crap we
> > found first (or last), but it is the wrong solution for a problem that we
> > really should not have in the first place if someone had thought a bit
> > before adding a misc device for something that has no reason to be unique in
> > a system.
> 
> Well, we have that userspace ABI of one hwrng char device. I did not
> invent that. It's kind of broken, yes.
> And changing it in a compatible way is probably difficult.
> 
> > Instead of papering over the problem with borked solutions, maybe we should
> > just export ALL HRNGs to userspace.  While at it, please add whatever is
> 
> And then we would _still_ export some kind of hint for rngd that
> the CPU rng device should be preferred over the bcm43xx device.
> rngd needs some basic hint about the devices.
> How would you implement that? (We're back to my TYPE_XXX definitions ;) )
> 
> > needed so that userspace can talk to the kernel driver to get vital
> > information about the HRNG device the driver might have (the current
> > interface is a bad simplistic hack).
> 
> What is "vital information"? My TYPE_XXX categories? ;)
> 
> > Let userspace get the data from whichever HRNG it wants, process it in any
> 
> It _can_. We can switch the RNG in sysfs. So userspace _can_ get data
> from whichever HWRNG it wants.
> 
> > way it wants and pipe it back through /dev/random IOCTLs.  And let it do it
> > for as many HRNGs it wants at the same time.
> 
> That's an improvement, yes.
> 
> > And if you must have /dev/hw_random point somewhere, let udev scripts or
> > something else like that take care of it.
> 
> Could do that, yes.
> 
> 

Oh, and eh, if we _are_ really going to redesign the userspace interface
to show each RNG to userspace, we should merge my TYPE_XXX patch before
that nevertheless, as it fixes (works around) a real bug (IMO).

-- 
Greetings Michael.