lkml.org 
[lkml]   [2014]   [Jul]   [22]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH v2] mm/highmem: make kmap cache coloring aware
From
Hi David,

On Tue, Jul 22, 2014 at 4:58 AM, David Rientjes <rientjes@google.com> wrote:
> On Thu, 17 Jul 2014, Max Filippov wrote:
>
>> From: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com>
>>
>> Provide hooks that allow architectures with aliasing cache to align
>> mapping address of high pages according to their color. Such architectures
>> may enforce similar coloring of low- and high-memory page mappings and
>> reuse existing cache management functions to support highmem.
>>
>
> Typically a change like this would be proposed along with a change to an
> architecture which would define this new ARCH_PKMAP_COLORING and have its
> own overriding definitions. Based on who you sent this patch to, it looks
> like that would be mips and xtensa. Now the only question is where are
> those patches to add the alternate definitions for those platforms?

I'm going to post xtensa series shortly, and I saw corresponding changes
for MIPS in the following patch:
http://permalink.gmane.org/gmane.linux.kernel.mm/107654

>> Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com>
>> [ Max: extract architecture-independent part of the original patch, clean
>> up checkpatch and build warnings. ]
>> Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
>> ---
>> Changes v1->v2:
>> - fix description
>>
>> mm/highmem.c | 19 ++++++++++++++++---
>> 1 file changed, 16 insertions(+), 3 deletions(-)
>>
>> diff --git a/mm/highmem.c b/mm/highmem.c
>> index b32b70c..6898a8b 100644
>> --- a/mm/highmem.c
>> +++ b/mm/highmem.c
>> @@ -44,6 +44,14 @@ DEFINE_PER_CPU(int, __kmap_atomic_idx);
>> */
>> #ifdef CONFIG_HIGHMEM
>>
>> +#ifndef ARCH_PKMAP_COLORING
>> +#define set_pkmap_color(pg, cl) /* */
>
> This is typically done with do {} while (0).

Will fix.

>> +#define get_last_pkmap_nr(p, cl) (p)
>> +#define get_next_pkmap_nr(p, cl) (((p) + 1) & LAST_PKMAP_MASK)
>> +#define is_no_more_pkmaps(p, cl) (!(p))
>
> That's not gramatically proper.

Just no_more_pkmaps maybe?

>> +#define get_next_pkmap_counter(c, cl) ((c) - 1)
>> +#endif
>> +
>> unsigned long totalhigh_pages __read_mostly;
>> EXPORT_SYMBOL(totalhigh_pages);
>>
>> @@ -161,19 +169,24 @@ static inline unsigned long map_new_virtual(struct page *page)
>> {
>> unsigned long vaddr;
>> int count;
>> + int color __maybe_unused;
>> +
>> + set_pkmap_color(page, color);
>> + last_pkmap_nr = get_last_pkmap_nr(last_pkmap_nr, color);
>>
>> start:
>> count = LAST_PKMAP;
>> /* Find an empty entry */
>> for (;;) {
>> - last_pkmap_nr = (last_pkmap_nr + 1) & LAST_PKMAP_MASK;
>> - if (!last_pkmap_nr) {
>> + last_pkmap_nr = get_next_pkmap_nr(last_pkmap_nr, color);
>> + if (is_no_more_pkmaps(last_pkmap_nr, color)) {
>> flush_all_zero_pkmaps();
>> count = LAST_PKMAP;
>> }
>> if (!pkmap_count[last_pkmap_nr])
>> break; /* Found a usable entry */
>> - if (--count)
>> + count = get_next_pkmap_counter(count, color);
>
> And that's not equivalent at all, --count decrements the auto variable and
> then tests it for being non-zero. Your get_next_pkmap_counter() never
> decrements count.

Yes, it's not literally equivalent, but it does the same thing: count gets
decreased by 1 and for loop iterates until count reaches zero.
get_next_pkmap_counter has no side effects, which is good. And count
is not supposed to go below zero anyway, but sure I can change the below
condition to 'if (count)'.

>> + if (count > 0)
>> continue;
>>
>> /*

--
Thanks.
-- Max


\
 
 \ /
  Last update: 2014-07-22 12:21    [W:0.102 / U:0.348 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site