lkml.org 
[lkml]   [2012]   [Jun]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH -V8 14/16] hugetlb/cgroup: add charge/uncharge calls for HugeTLB alloc/free
On Sat, Jun 09, 2012 at 06:39:06PM +0530, Aneesh Kumar K.V wrote:
> Johannes Weiner <hannes@cmpxchg.org> writes:
>
> > On Sat, Jun 09, 2012 at 02:29:59PM +0530, Aneesh Kumar K.V wrote:
> >> From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
> >>
> >> This adds necessary charge/uncharge calls in the HugeTLB code. We do
> >> hugetlb cgroup charge in page alloc and uncharge in compound page destructor.
> >>
> >> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
> >> ---
> >> mm/hugetlb.c | 16 +++++++++++++++-
> >> mm/hugetlb_cgroup.c | 7 +------
> >> 2 files changed, 16 insertions(+), 7 deletions(-)
> >>
> >> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> >> index bf79131..4ca92a9 100644
> >> --- a/mm/hugetlb.c
> >> +++ b/mm/hugetlb.c
> >> @@ -628,6 +628,8 @@ static void free_huge_page(struct page *page)
> >> BUG_ON(page_mapcount(page));
> >>
> >> spin_lock(&hugetlb_lock);
> >> + hugetlb_cgroup_uncharge_page(hstate_index(h),
> >> + pages_per_huge_page(h), page);
> >
> > hugetlb_cgroup_uncharge_page() takes the hugetlb_lock, no?
>
> Yes, But this patch also modifies it to not take the lock, because we
> hold spin_lock just below in the call site. I didn't want to drop the
> lock and take it again.

Sorry, I missed that.

> > It's quite hard to review code that is split up like this. Please
> > always keep the introduction of new functions in the same patch that
> > adds the callsite(s).
>
> One of the reason I split the charge/uncharge routines and the callers
> in separate patches is to make it easier for review. Irrespective of
> the call site charge/uncharge routines should be correct with respect
> to locking and other details. What I did in this patch is a small
> optimization of avoiding dropping and taking the lock again. May be the
> right approach would have been to name it __hugetlb_cgroup_uncharge_page
> and make sure the hugetlb_cgroup_uncharge_page still takes spin_lock.
> But then we don't have any callers for that.

I think this makes it needlessly complicated and there is no correct
or incorrect locking in (initially) dead code :-)

The callsites are just a few lines. It's harder to review if you
introduce an API and then change it again mid-patchset.

If there are no callers for a function that grabs the lock itself,
don't add it. Just add a note to the kerneldoc that explains the
requirement or put VM_BUG_ON(!spin_is_locked(&hugetlb_lock)); in
there or so.


\
 
 \ /
  Last update: 2012-06-09 16:41    [W:0.094 / U:0.808 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site