[RFC PATCH 0/2] Introduce pte_isset for arm and arm64

[Steve Capper]

On 24 February 2014 17:36, Will Deacon <will.deacon at arm.com> wrote:
> Hi Steve,

Hey Will,

>
> On Mon, Feb 24, 2014 at 03:33:23PM +0000, Steve Capper wrote:
>> Long descriptors on ARM and all ptes on ARM64 are 64 bits. Some pte
>> attribute test functions such as pte_dirty return the bitwise-and of a
>> flag with the pte value. If the flag to be tested resides in the
>> upper 32 bits of the pte, then we run into the danger of the result
>> being dropped if downcast.
>>
>> For example:
>>       gather_stats(page, md, pte_dirty(*pte), 1);
>> where pte_dirty(*pte) is downcast to an int.
>>
>> Under ARM with LPAE, functions such as huge_pte_write also perform a
>> downcast to unsigned long (which is 32 bits).
>>
>> This series introduces a new macro pte_isset which performs the bitwise
>> and, then performs a double logical invert to ensure predictable
>> downcasting.
>>
>> The reasoning behind a new macro was that it creates an obvious
>> pattern which can be followed by future changes to the pte functions.
>> Rather than target specific pte functions, I opted to wrap all of them
>> with this macro for the sake of simplicity (also the flags may move in
>> future).
>>
>> I was toying with using (bool), but decided that the !! looked slightly
>> better. Under testing (with an Arndale and the Fast Model on 3.14-rc4)
>> both strategies worked as expected.
>
> Ok, but I still don't understand why this is needed over your previous
> series. Just inlining the '!!' into the existing accessors should be fine,
> no? Basically, I don't see what the new macro gains us.
>

I thought it would be slightly easier to read, but it probably is
somewhat superfluous. I will just expand out to !! for all the
accessors that don't have a logical inverse already.

Cheers,
-- 
Steve