Re: start_kernel(): bug: interrupts were enabled early

[Matthew Wilcox <matthew@wil.cx>]

On Thu, Apr 01, 2010 at 10:41:11AM +0100, Jamie Lokier wrote:
> David Howells wrote:
> > Russell King <rmk@arm.linux.org.uk> wrote:
> > > We use the standard generic kernel implementation.  Is x86 different? ;)
> > 
> > The optimised fast paths used on x86 rwsems don't disable interrupts.
> 
> Any reason not to use the same technique for all the archs - plus the
> trick used in arch/armkernel/entry-armv.S:__kuser_cmpxchg for those
> archs which don't have atomic instructions or ll/sc?

Assuming you're talking about the __LINUX_ARM_ARCH__ < 6 + CONFIG_MMU
case, then this only works for uniprocessor machines.

> If the problem here is _only_ semaphores, and the above might make
> semaphores faster anyway, perhaps it's a solution.

You trade off a bit of overhead in the semaphore path for a bit of
overhead in the interrupt path.  We probably take more sempahores than
we do interrupts, so it's probably worthwhile.  Still, cmpxchg() needs
to be SMP-safe.

Realistically, this isn't something that can be done in generic code.
It has to be done in arch-specific code.

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."