Rod Evans wrote: > Darren J Moffat wrote: >> We have people starting to write drivers for to use the crypto and rng >> capability for the VIA CPUs with the Padlock functionality. Given that >> it doesn't require privilege I think it would be really nice if we >> could have $HWCAP tags for the functionality. >> >> The existence of the various features, RNG, RSA, SHA, AES can be >> tested for using the CPUID EDX flags. >> >> I believe this would require the following new HWCAP feature tags: >> >> RNG, ACE, ACE2, PHE, PMM >> >> These should probably be prefixed VIA since I believe they are unique >> to VIA at this time. >> >> See the developer guide for the VIA Padlock here: >> >> http://www.via.com.tw/en/initiatives/padlock/developercenter.jsp >> >> Specifically page 4 of the document "VIA C5J programming guide for >> Assembler" >> >> Thoughts ? > > Sounds fine, but it's not really a linking issue. > > A model for $HWCAP use has been developed for the link-editors, but > the individual capabilities are selected by those familiar with our > hardware and the needs of users (like yourself, or the compiler folks). > > The link-editors in effect, act as traffic cops in the whole scheme, > by matching hardware capabilities to dynamic object requirements. > The link-editors haven't a clue what each hardware capabilities flag > means :-) > > Additional flags typically arrive from the folks who deliver new > platforms. An ARC case seeds the delivery. Then you just need to update > the files in usr/src/common/elfcap, and the link-editors, together > with the various tools that display capabilities information, all get > magically updated.
Thanks for that. > I'm not being much help today am I? :-) Actually you are. 1) You do mostly userland linker stuff not krtld 2) That common/elfcap is what I need to update, I went looking at source after I posted my message. 3) What the involvement of the link-editors are with $HWCAP. Thanks Rod. -- Darren J Moffat
