Re: ARMv5 and atomic operations

[Chris Tyler <chris@xxxxxxxxxxx>]

On Sun, 2012-04-22 at 18:16 -0700, Brendan Conoboy wrote:
> As we get closer to having 100% package coverage in F17-ARM we're 
> running into harder build failures due to the limitations of the chips 
> we're building for.  The problem I've noticed on many of the recent 
> failures is due to the lack of atomic operations (These didn't arrive 
> until ARMv6).  

What are some examples of these?

> How do we want to handle this?  I see a few options:
> 
> 1. Abandon armv5 and move to armv6 where some of the operations we need 
> are available.  This will still support the raspberry pi- what about 
> kirkwood *plugs?

That would kill the older plugs -- anything below a d2plug. 

However: do we care? Much? Going to v6 would let us optimize better for
the Raspi, which will have greater market penetration than the plugs
when existing orders are filled. Otoh, it's a whole 'nuther rebuild.

> 2. Excludearch armv5tel from the affected packages since they'll never 
> work atomically.
> 
> 3. Accept that these packages simply won't be available to 32 bit Fedora 
> systems (this is the result of inaction).
> 
> 4. Pretend operations are atomic when they are not.
> 
> 5. Create magic patch that implements atomic operations through software.

6. There is the kernel's "user space atomic helper" (kuser_cmpxchg64) at
0xffff0f60, see Documentation/arm/kernel_user_helpers.txt. The kernel
puts an instruction sequence here tuned for the current arch that can be
called by userland to provide an atomic compare/exchange -- if it can be
done natively, the instruction sequence does that, otherwise it does a
syscall (with IRQ protection etc). Would this solve the problems you've
identified?

-Chris

_______________________________________________
arm mailing list
arm@xxxxxxxxxxxxxxxxxxxxxxx
https://admin.fedoraproject.org/mailman/listinfo/arm