Re: CPA patchset

[Ingo Molnar <mingo@elte.hu>]

* Dave Airlie <airlied@gmail.com> wrote:

> > - firstly, there's no rationale given. So we'll change ioremap()/etc.
> >   from doing a cflush-range instruction instead of a WBINVD. But why?
> >   WBINVD isnt particular fast (takes a few msecs), but why is that a
> >   problem? Drivers dont do high-frequency ioremap-ing. It's typically
> >   only done at driver/device startup and that's it. Whether module load
> >   time takes 1254 msecs instead of 1250 msecs is no big deal.
> 
> read graphics drivers, even though I think we may avoid the whole path 
> if we can and end up doing some of this in the drivers when they know 
> more about the situation so can avoid safeties..

by 'read graphics drivers' do you mean direct framebuffer access? In any 
case, a driver (or even userspace) can use cflush just fine, it's an 
unprivileged instruction. But this is about the ioremap() implementation 
using cflush instead of WBINVD, and that is a slowpath and is up to the 
kernel anyway - i'm not aware of many high-frequency ioremap() users, so 
robustness concerns control the policy here. So could you please explain 
your point in more detail?

	Ingo