Re: [PATCH 2.6.13] lockless pagecache 2/7

[Nick Piggin <nickpiggin@yahoo.com.au>]

David S. Miller wrote:
> From: Nick Piggin <nickpiggin@yahoo.com.au>
> Date: Sat, 03 Sep 2005 07:22:18 +1000
> 
> 
>>This atomic_cmpxchg, unlike a "regular" cmpxchg, has the advantage
>>that the memory altered should always be going through the atomic_
>>accessors, and thus should be implementable with spinlocks.
>>
>>See for example, arch/sparc/lib/atomic32.c
>>
>>At least, that's what I'm hoping for.
> 
> 
> Ok, as long as the rule is that all accesses have to go
> through accessor macros, it would work.  This is not true
> for existing uses of cmpxchg() btw, userland accesses shared
> locks with the kernel would using any kind of accessors we
> can control.
> 
> This means that your atomic_cmpxchg() cannot be used for locking
> objects shared with userland, as DRM wants, since the hashed spinlock
> trick does not work in such a case.
> 

So neither could currently supported atomic_t ops be shared with
userland accesses?

Then I think it would not be breaking any interface rule to do an
atomic_t atomic_cmpxchg either. Definitely for my usage it will
not be shared with userland.

Thanks,
Nick

-- 
SUSE Labs, Novell Inc.

Send instant messages to your online friends http://au.messenger.yahoo.com