Re: [PATCH] S390: Replace calls to __get_free_pages() with __get_dma_pages().

["Satyam Sharma" <satyam.sharma@gmail.com>]

Hi,

On 6/6/07, Robert P. J. Day <rpjday@mindspring.com> wrote:
>
> Replace a number of calls to __get_free_pages() with the corresponding
> calls to __get_dma_pages().
> [...]
>   once the __GFP_DMA argument is removed, it does look weird to see
> the first argument of just 0.  should that be filled in with
> GFP_ATOMIC as christopher lameter suggested?

Yes, I suppose so ... GFP_ATOMIC can dip into the emergency
pools so would also make this code a bit more "robust" than using
"0" (== GFP_NOWAIT) and it's not that GFP_ATOMIC "waits" on
anything either ...

> -                       (void *)__get_free_pages(__GFP_DMA,
> +                       (void *)__get_dma_pages(0,

GFP_NOWAIT == 0, so the macro GFP_NOWAIT is the one to
use if you really don't want any change in behaviour (and as the
comment above GFP_NOWAIT says, it's much better to use that
name than simply specify "0").

Off-topic, but I wonder what are the valid usage cases / scenarios
for GFP_NOWAIT? The obvious answer is somebody might want to
be a way-too-polite citizen and stay off the emergency pools even
from atomic context, but why would anybody want to do /that/ ...
[ BTW there are 3 users of GFP_NOWAIT in kernel code, but there
could be more that simply specify "0" to get same behaviour. ]

Satyam

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>