Re: [PATCH] x86/paravirt: revert exports to restore old behaviour

[Takashi Iwai <tiwai@suse.de>]

At Mon, 19 Nov 2007 17:14:15 -0800,
Jeremy Fitzhardinge wrote:
> 
> Takashi Iwai wrote:
> > I took at this problem (as I have an nvidia card on one of my
> > workstations), and found out that the following suffer from
> > EXPORT_SYMBOL_GPL changes:
> >   
> 
> Which kernel version are you using?  This is different in .24-rc
> compared to .23.

24-rc2.  23 has no problem, as you know :)

> > * local_disable_irq(), local_irq_save*(), etc.
> >   
> 
> These should be OK either way.  pv_irq_ops is not _GPL.

Right.  I thought it somehow involved with other pv ops indirectly,
but it seems not.

> > * MSR-related macros like rdmsr(), wrmsr(), read_cr0(), etc.
> >   wbinvd(), too.
> >   
> 
> These could reasonably use the the native_* versions anyway, since the
> driver won't be being used in an environment where these won't work. 
> Perhaps they should be split out separate from the gdt/ldt operations,
> which they should have no business touching.

Yes, that's possible.

> > * pmd_val(), pgd_val(), etc are all involved with pv_mm_ops.
> >   pmd_large() and pmd_bad() is also indirectly involved.
> >   __flush_tlb() and friends suffer, too.
> >   
> 
> Yeah, I guess they can be expected to play with pagetables.
> 
> > The easiest workaround I found was to undefine CONFIG_PARAVIRT before
> > inclusion of linux kernel headers, but it is really ugly and hacky.
> >   
> 
> Yeah.  It will explode if you are running in a virtual environment which
> still gives the virtual machine graphics hardware access.

Yes.  More over, there is no guarantee that this will be built
properly in the future.  It's a kind of coincident that the driver is
built.  If any non-paravirt implementation accesses an exported symbol
instead of inlining, then this won't work, too.

> > Redefinig with raw_*() and native_*() is another way, but it takes
> > much more work than defining these primitive functions in assembly.
> >
> > So, in short, with EXPORT_SYMBOL_GPL change, it's pretty hard to write
> > a non-GPL driver in a same manner...
> >   
> 
> Yeah.  I think removing the difference between PARAVIRT and non-PARAVIRT
> is enough to justify the exports.  If we want to make the policy
> decision that modules can't use pagetable or msr operations at all, then
> that's a separate decision which can be applied uniformly to PARAVIRT
> and non-PARAVIRT.

Agreed.


thanks,

Takashi