At last your patient try makes the problem solve
although it's from not your patch series.

Thanks for very patient try and testing until now, Kame. :)
I learned lot of things from this thread.

Thanks, all.

On Wed, Jan 6, 2010 at 4:06 PM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@jp.fujitsu.com> wrote:
> On Tue, 5 Jan 2010 20:20:56 -0800 (PST)
> Linus Torvalds <torvalds@linux-foundation.org> wrote:
>
>>
>>
>> On Wed, 6 Jan 2010, KAMEZAWA Hiroyuki wrote:
>> > >
>> > > Of course, your other load with MADV_DONTNEED seems to be horrible, and
>> > > has some nasty spinlock issues, but that looks like a separate deal (I
>> > > assume that load is just very hard on the pgtable lock).
>> >
>> > It's zone->lock, I guess. My test program avoids pgtable lock problem.
>>
>> Yeah, I should have looked more at your callchain. That's nasty. Much
>> worse than the per-mm lock. I thought the page buffering would avoid the
>> zone lock becoming a huge problem, but clearly not in this case.
>>
> For my mental peace, I rewrote test program as
>
> В while () {
> В  В  В  В touch memory
> В  В  В  В barrier
> В  В  В  В madvice DONTNEED all range by cpu 0
> В  В  В  В barrier
> В }
> And serialize madivce().
>
> Then, zone->lock disappears and I don't see big difference with XADD rwsem and
> my tricky patch. I think I got reasonable result and fixing rwsem is the sane way.
>
> next target will be clear_page()? hehe.
> What catches my eyes is cost of memcg... (>_<
>
> Thank you all,
> -Kame
> ==
> [XADD rwsem]
> [root@bluextal memory]# В /root/bin/perf stat -e page-faults,cache-misses --repeat 5 ./multi-fault-all 8
>
> В Performance counter stats for './multi-fault-all 8' (5 runs):
>
> В  В  В  33029186 В page-faults В  В  В  В  В  В  В  В ( +- В  0.146% )
> В  В  В 348698659 В cache-misses В  В  В  В  В  В  В  ( +- В  0.149% )
>
> В  60.002876268 В seconds time elapsed В  ( +- В  0.001% )
>
> # Samples: 815596419603
> #
> # Overhead В  В  В  В  В Command В  В  В  В  В  В  Shared Object В Symbol
> # ........ В ............... В ........................ В ......
> #
> В  В 41.51% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] clear_page_c
> В  В  9.08% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] down_read_trylock
> В  В  6.23% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] up_read
> В  В  6.17% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] __mem_cgroup_try_charg
> В  В  4.76% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] handle_mm_fault
> В  В  3.77% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] __mem_cgroup_commit_ch
> В  В  3.62% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] __rmqueue
> В  В  2.30% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] _raw_spin_lock
> В  В  2.30% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] page_fault
> В  В  2.12% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] mem_cgroup_charge_comm
> В  В  2.05% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] bad_range
> В  В  1.78% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] _raw_spin_lock_irq
> В  В  1.53% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] lookup_page_cgroup
> В  В  1.44% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] __mem_cgroup_uncharge_
> В  В  1.41% В multi-fault-all В ./multi-fault-all В  В  В  В  [.] worker
> В  В  1.30% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] get_page_from_freelist
> В  В  1.06% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] page_remove_rmap
>
>
>
> [async page fault]
> [root@bluextal memory]# В /root/bin/perf stat -e page-faults,cache-misses --repeat 5 ./multi-fault-all 8
>
> В Performance counter stats for './multi-fault-all 8' (5 runs):
>
> В  В  В  33345089 В page-faults В  В  В  В  В  В  В  В ( +- В  0.555% )
> В  В  В 357660074 В cache-misses В  В  В  В  В  В  В  ( +- В  1.438% )
>
> В  60.003711279 В seconds time elapsed В  ( +- В  0.002% )
>
>
> В  В 40.94% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] clear_page_c
> В  В  6.96% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] vma_put
> В  В  6.82% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] page_add_new_anon_rmap
> В  В  5.86% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] __mem_cgroup_try_charg
> В  В  4.40% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] __rmqueue
> В  В  4.14% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] find_vma_speculative
> В  В  3.97% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] handle_mm_fault
> В  В  3.52% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] _raw_spin_lock
> В  В  3.46% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] __mem_cgroup_commit_ch
> В  В  2.23% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] bad_range
> В  В  2.16% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] mem_cgroup_charge_comm
> В  В  1.96% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] _raw_spin_lock_irq
> В  В  1.75% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] mem_cgroup_add_lru_lis
> В  В  1.73% В multi-fault-all В [kernel] В  В  В  В  В  В  В  В  В [k] page_fault
>


-- 
Kind regards,
Minchan Kim
N‹§Іжмrё›zЗ§u©ћІЖ {­†йм№»®&Ю–)оЖiўћШ^n‡r¶‰љЋЉЭўj$Ѕ§$ўёў№Ё­и§~Љ'.)оДГ,yиm¶џяГ%Љ{±љj+ѓрићЧ¦j)Z†·џ