Re: [PATCH 03/23] thp: compile-time and sysfs knob for thp pagecache

[Ning Qu <quning@xxxxxxxxxx>]

Great!

Best wishes,

-- 
Ning Qu (曲宁) | Software Engineer | quning@xxxxxxxxxx | +1-408-418-6066

On Fri, Sep 6, 2013 at 4:33 AM, Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx> wrote:

Ning Qu wrote:
> One minor question inline.
>
> Best wishes,
> --
> Ning Qu (曲宁) | Software Engineer | quning@xxxxxxxxxx | +1-408-418-6066
>
>
> On Sat, Aug 3, 2013 at 7:17 PM, Kirill A. Shutemov <
> kirill.shutemov@xxxxxxxxxxxxxxx> wrote:
>
> > From: "Kirill A. Shutemov" <kirill.shutemov@xxxxxxxxxxxxxxx>
> >
> > For now, TRANSPARENT_HUGEPAGE_PAGECACHE is only implemented for x86_64.
> >
> > Radix tree perload overhead can be significant on BASE_SMALL systems, so
> > let's add dependency on !BASE_SMALL.
> >
> > /sys/kernel/mm/transparent_hugepage/page_cache is runtime knob for the
> > feature. It's enabled by default if TRANSPARENT_HUGEPAGE_PAGECACHE is
> > enabled.
> >
> > Signed-off-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
> > ---
> >  Documentation/vm/transhuge.txt |  9 +++++++++
> >  include/linux/huge_mm.h        |  9 +++++++++
> >  mm/Kconfig                     | 12 ++++++++++++
> >  mm/huge_memory.c               | 23 +++++++++++++++++++++++
> >  4 files changed, 53 insertions(+)
> >
> > diff --git a/Documentation/vm/transhuge.txt
> > b/Documentation/vm/transhuge.txt
> > index 4a63953..4cc15c4 100644
> > --- a/Documentation/vm/transhuge.txt
> > +++ b/Documentation/vm/transhuge.txt
> > @@ -103,6 +103,15 @@ echo always
> > >/sys/kernel/mm/transparent_hugepage/enabled
> >  echo madvise >/sys/kernel/mm/transparent_hugepage/enabled
> >  echo never >/sys/kernel/mm/transparent_hugepage/enabled
> >
> > +If TRANSPARENT_HUGEPAGE_PAGECACHE is enabled kernel will use huge pages in
> > +page cache if possible. It can be disable and re-enabled via sysfs:
> > +
> > +echo 0 >/sys/kernel/mm/transparent_hugepage/page_cache
> > +echo 1 >/sys/kernel/mm/transparent_hugepage/page_cache
> > +
> > +If it's disabled kernel will not add new huge pages to page cache and
> > +split them on mapping, but already mapped pages will stay intakt.
> > +
> >  It's also possible to limit defrag efforts in the VM to generate
> >  hugepages in case they're not immediately free to madvise regions or
> >  to never try to defrag memory and simply fallback to regular pages
> > diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> > index 3935428..1534e1e 100644
> > --- a/include/linux/huge_mm.h
> > +++ b/include/linux/huge_mm.h
> > @@ -40,6 +40,7 @@ enum transparent_hugepage_flag {
> >         TRANSPARENT_HUGEPAGE_DEFRAG_FLAG,
> >         TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG,
> >         TRANSPARENT_HUGEPAGE_DEFRAG_KHUGEPAGED_FLAG,
> > +       TRANSPARENT_HUGEPAGE_PAGECACHE,
> >         TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG,
> >  #ifdef CONFIG_DEBUG_VM
> >         TRANSPARENT_HUGEPAGE_DEBUG_COW_FLAG,
> > @@ -229,4 +230,12 @@ static inline int do_huge_pmd_numa_page(struct
> > mm_struct *mm, struct vm_area_str
> >
> >  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
> >
> > +static inline bool transparent_hugepage_pagecache(void)
> > +{
> > +       if (!IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE_PAGECACHE))
> > +               return false;
> > +       if (!(transparent_hugepage_flags & (1<<TRANSPARENT_HUGEPAGE_FLAG)))
> >
>
> Here, I suppose we should test the  TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG as
> well? E.g.
>         if (!(transparent_hugepage_flags &
>               ((1<<TRANSPARENT_HUGEPAGE_FLAG) |
>                (1<<TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG))))
>
> +               return false;

You're right. Fixed.

--
 Kirill A. Shutemov