Re: [PATCH 1/2] af_packet: use vmalloc_to_page() instead for the addresss returned by vmalloc()

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mercredi 01 décembre 2010 à 14:38 +0100, Eric Dumazet a écrit :
> Le mercredi 01 décembre 2010 à 21:05 +0800, Changli Gao a écrit :
> > On Tue, Nov 30, 2010 at 10:37 PM, Neil Horman <nhorman@tuxdriver.com> wrote:
> > > Off the top of my head, I would think that pgv_to_page could be prototyped such
> > > that it could accept addr, offset and struct page ** arguments.  That way we can
> > > track the current page that we're mapped to, lowering the number of calls to
> > > vmalloc_to_page, and we can still use an increment like we do above (as long as
> > > its wrapped in a subsequent call to pgv_to_page)
> > 
> > I'll try to optimize pgv_to_page() after this patch series merged. I
> > am planning to call vmalloc_to_page() previously, and cache its result
> > in a per pgv array for future use. Thanks.
> > 
> > 
> 
> Hmm... fact is flush_dcache_page() is void on some arches.
> 
> Maybe the only thing to do is avoid pgv_to_page() calls if
> ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE is 0
> 
>     The ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE symbol was introduced to avoid
>     pointless empty cache-thrashing loops on architectures for which
>     flush_dcache_page() is a no-op.  Every architecture was provided with this
>     flush pages on architectires where ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE is
>     equal 1 or do nothing otherwise.
> 
> 

I guess using __pure attribute on pgv_to_page() should be enough ;)

static inline __pure struct page *pgv_to_page(void *addr)
{
       if (is_vmalloc_addr(addr))
               return vmalloc_to_page(addr);
       return virt_to_page(addr);
}

Compiler then should optimize away

flush_dcache_page(pgv_to_page(addr));


But the pgv_to_page() done in packet_sendmsg() will stay, of course.