lkml.org 
[lkml]   [2013]   [Jul]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH] hugepage: allow parallelization of the hugepage fault path
From
Date
On Fri, 2013-07-19 at 17:14 +1000, David Gibson wrote:
> On Thu, Jul 18, 2013 at 05:42:35PM +0900, Joonsoo Kim wrote:
> > On Wed, Jul 17, 2013 at 12:50:25PM -0700, Davidlohr Bueso wrote:
> > > From: David Gibson <david@gibson.dropbear.id.au>
> > >
> > > At present, the page fault path for hugepages is serialized by a
> > > single mutex. This is used to avoid spurious out-of-memory conditions
> > > when the hugepage pool is fully utilized (two processes or threads can
> > > race to instantiate the same mapping with the last hugepage from the
> > > pool, the race loser returning VM_FAULT_OOM). This problem is
> > > specific to hugepages, because it is normal to want to use every
> > > single hugepage in the system - with normal pages we simply assume
> > > there will always be a few spare pages which can be used temporarily
> > > until the race is resolved.
> > >
> > > Unfortunately this serialization also means that clearing of hugepages
> > > cannot be parallelized across multiple CPUs, which can lead to very
> > > long process startup times when using large numbers of hugepages.
> > >
> > > This patch improves the situation by replacing the single mutex with a
> > > table of mutexes, selected based on a hash, which allows us to know
> > > which page in the file we're instantiating. For shared mappings, the
> > > hash key is selected based on the address space and file offset being faulted.
> > > Similarly, for private mappings, the mm and virtual address are used.
> > >
> >
> > Hello.
> >
> > With this table mutex, we cannot protect region tracking structure.
> > See below comment.
> >
> > /*
> > * Region tracking -- allows tracking of reservations and instantiated pages
> > * across the pages in a mapping.
> > *
> > * The region data structures are protected by a combination of the mmap_sem
> > * and the hugetlb_instantion_mutex. To access or modify a region the caller
> > * must either hold the mmap_sem for write, or the mmap_sem for read and
> > * the hugetlb_instantiation mutex:
> > *
> > * down_write(&mm->mmap_sem);
> > * or
> > * down_read(&mm->mmap_sem);
> > * mutex_lock(&hugetlb_instantiation_mutex);
> > */
>
> Ugh. Who the hell added that. I guess you'll need to split of
> another mutex for that purpose, afaict there should be no interaction
> with the actual, intended purpose of the instantiation mutex.

This was added in commit 84afd99b. One way to go would be to add a
spinlock to protect changes to the regions - however reading the
changelog, and based on David's previous explanation for the
instantiation mutex, I don't see why it was added. In fact several
places modify regions without holding the instantiation mutex, ie:
hugetlb_reserve_pages()

Am I missing something here?

Thanks,
Davidlohr






\
 
 \ /
  Last update: 2013-07-19 23:41    [W:0.057 / U:0.072 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site