| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Tue, 11 Jun 2013 09:57:42 -0500
From: Alex Thorlton <athorlton@....com>
To: "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>
Cc: linux-kernel@...r.kernel.org, Li Zefan <lizefan@...wei.com>,
Rob Landley <rob@...dley.net>,
Andrew Morton <akpm@...ux-foundation.org>,
Mel Gorman <mgorman@...e.de>, Rik van Riel <riel@...hat.com>,
Johannes Weiner <hannes@...xchg.org>,
Xiao Guangrong <xiaoguangrong@...ux.vnet.ibm.com>,
David Rientjes <rientjes@...gle.com>,
linux-doc@...r.kernel.org, linux-mm@...ck.org
Subject: Re: [PATCH] Make transparent hugepages cpuset aware
On Tue, Jun 11, 2013 at 09:55:18AM +0300, Kirill A. Shutemov wrote:
> Alex Thorlton wrote:
> > This patch adds the ability to control THPs on a per cpuset basis. Please see
> > the additions to Documentation/cgroups/cpusets.txt for more information.
> >
> > Signed-off-by: Alex Thorlton <athorlton@....com>
> > Reviewed-by: Robin Holt <holt@....com>
> > Cc: Li Zefan <lizefan@...wei.com>
> > Cc: Rob Landley <rob@...dley.net>
> > Cc: Andrew Morton <akpm@...ux-foundation.org>
> > Cc: Mel Gorman <mgorman@...e.de>
> > Cc: Rik van Riel <riel@...hat.com>
> > Cc: Kirill A. Shutemov <kirill.shutemov@...ux.intel.com>
> > Cc: Johannes Weiner <hannes@...xchg.org>
> > Cc: Xiao Guangrong <xiaoguangrong@...ux.vnet.ibm.com>
> > Cc: David Rientjes <rientjes@...gle.com>
> > Cc: linux-doc@...r.kernel.org
> > Cc: linux-mm@...ck.org
> > ---
> > Documentation/cgroups/cpusets.txt | 50 ++++++++++-
> > include/linux/cpuset.h | 5 ++
> > include/linux/huge_mm.h | 25 +++++-
> > kernel/cpuset.c | 181 ++++++++++++++++++++++++++++++++++++++
> > mm/huge_memory.c | 3 +
> > 5 files changed, 261 insertions(+), 3 deletions(-)
> >
> > diff --git a/Documentation/cgroups/cpusets.txt b/Documentation/cgroups/cpusets.txt
> > index 12e01d4..b7b2c83 100644
> > --- a/Documentation/cgroups/cpusets.txt
> > +++ b/Documentation/cgroups/cpusets.txt
> > @@ -22,12 +22,14 @@ CONTENTS:
> > 1.6 What is memory spread ?
> > 1.7 What is sched_load_balance ?
> > 1.8 What is sched_relax_domain_level ?
> > - 1.9 How do I use cpusets ?
> > + 1.9 What is thp_enabled ?
> > + 1.10 How do I use cpusets ?
> > 2. Usage Examples and Syntax
> > 2.1 Basic Usage
> > 2.2 Adding/removing cpus
> > 2.3 Setting flags
> > 2.4 Attaching processes
> > + 2.5 Setting thp_enabled flags
> > 3. Questions
> > 4. Contact
> >
> > @@ -581,7 +583,34 @@ If your situation is:
> > then increasing 'sched_relax_domain_level' would benefit you.
> >
> >
> > -1.9 How do I use cpusets ?
> > +1.9 What is thp_enabled ?
> > +-----------------------
> > +
> > +The thp_enabled file contained within each cpuset controls how transparent
> > +hugepages are handled within that cpuset.
> > +
> > +The root cpuset's thp_enabled flags mirror the flags set in
> > +/sys/kernel/mm/transparent_hugepage/enabled. The flags in the root cpuset can
> > +only be modified by changing /sys/kernel/mm/transparent_hugepage/enabled. The
> > +thp_enabled file for the root cpuset is read only. These flags cause the
> > +root cpuset to behave as one might expect:
> > +
> > +- When set to always, THPs are used whenever practical
> > +- When set to madvise, THPs are used only on chunks of memory that have the
> > + MADV_HUGEPAGE flag set
> > +- When set to never, THPs are never allowed for tasks in this cpuset
> > +
> > +The behavior of thp_enabled for children of the root cpuset is where things
> > +become a bit more interesting. The child cpusets accept the same flags as the
> > +root, but also have a default flag, which, when set, causes a cpuset to use the
> > +behavior of its parent. When a child cpuset is created, its default flag is
> > +always initially set.
> > +
> > +Since the flags on child cpusets are allowed to differ from the flags on their
> > +parents, we are able to enable THPs for tasks in specific cpusets, and disable
> > +them in others.
>
> Should we have a way for parent cgroup can enforce child behaviour?
> Like a mask of allowed thp_enabled values children can choose.
>
We don't have a use case for that particular scenario, so we didn't
include any such functionality. Our main goal here was to allow
cpusets to override the /sys/kernel/mm/transparent_hugepage/enabled
setting. If you have a use case for that scenario, then I think it
would be more suitable to add that functionality in a separate patch.
> > @@ -177,6 +177,29 @@ static inline struct page *compound_trans_head(struct page *page)
> > return page;
> > }
> >
> > +#ifdef CONFIG_CPUSETS
> > +extern int cpuset_thp_always(struct task_struct *p);
> > +extern int cpuset_thp_madvise(struct task_struct *p);
> > +
> > +static inline int transparent_hugepage_enabled(struct vm_area_struct *vma)
> > +{
> > + if (cpuset_thp_always(current))
> > + return 1;
>
> Why do you ignore VM_NOHUGEPAGE?
> And !is_vma_temporary_stack(__vma) is still relevant.
>
That was an oversight, on my part. I've fixed it and will submit the
corrected patch shortly. Thanks for pointing that out.
> > + else if (cpuset_thp_madvise(current) &&
> > + ((vma)->vm_flags & VM_HUGEPAGE) &&
> > + !((vma)->vm_flags & VM_NOHUGEPAGE) &&
> > + !is_vma_temporary_stack(vma))
> > + return 1;
> > + else
> > + return 0;
> > +}
> > +#else
> > +static inline int transparent_hugepage_enabled(struct vm_area_struct *vma)
> > +{
> > + return _transparent_hugepage_enabled(vma);
> > +}
> > +#endif
> > +
> > extern int do_huge_pmd_numa_page(struct mm_struct *mm, struct vm_area_struct *vma,
> > unsigned long addr, pmd_t pmd, pmd_t *pmdp);
> >
>
> --
> Kirill A. Shutemov
- Alex Thorlton
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists