From: Daniel Lowengrub <lowdanie@gmail.com>
To: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org,
Nick Piggin <nickpiggin@yahoo.com.au>,
Ingo Molnar <mingo@elte.hu>,
KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Subject: Re: [PATCH 1/2] mm: use list.h for vma list
Date: Thu, 2 Apr 2009 12:56:22 +0300 [thread overview]
Message-ID: <8c5a844a0904020256r3e073d01o1991e5111127ce42@mail.gmail.com> (raw)
In-Reply-To: <1236851912.5090.93.camel@laptop>
I've been thinking for a while about the best way to implement Nick's
suggestions. Specifically, how to implement vma_next without all the
extra conditions. The main problem is that for vma_next to return a
vma and not a list_head it has to be know if vma->vm_list.next is
inside a vma or a mm so that it can call list_entry on it.
In the current code, something like next=vma->next runs the risk of
having next=NULL so the next pointer is usually used in a if statement
like:
vm_area_struct *next=vma->next;
if(next&&<check stuff with next>){<do stuff to next>}
What I want to do is to make vma_next return a list_head - without
calling list_entry, and a seperate function vma_entry return the
entry. This way, vma_entry would be a simple wrapper for
vma->vm_list.next. Then the above code snippet would read:
vm_area_struct *next;
list_head next_list =vma_next(vma);
if(in_list(mm, next_list) && next=vma_entry(next_list) && <check stuff
with next>)
{<do stuff to next>}
where in_list checks if next_list==mm->mm_vmas.
Or there could be a function called check_next which would do the
first two checks together so that we'd be able to write:
vm_area_struct *next;
if(check_next(vma, next) && <check stuff with next>){<do stuff to next>}
This does away with the redundant conditions that bothered Nick - the
check for the end of the list in vma_entry which returns NULL, and
then the check for next==NULL in the if statements.
This would lead to further optimizations based on the fact that we
could now pass around list_heads instead of vma's and only call
vma_entry after verifying that the list_head is really in the list and
not the mm_vmas list_head. This verification would be done in the
places that the current code checks for vma==NULL - as in the example
above.
What do you think?
Daniel
On Thu, Mar 12, 2009 at 12:58 PM, Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:
>
> On Wed, 2009-03-11 at 11:55 +0200, Daniel Lowengrub wrote:
> > Use the linked list defined list.h for the list of vmas that's stored
> > in the mm_struct structure. Wrapper functions "vma_next" and
> > "vma_prev" are also implemented. Functions that operate on more than
> > one vma are now given a list of vmas as input.
> >
> > Signed-off-by: Daniel Lowengrub
>
> While this is the approach I've taken for a patch I'm working on, a
> better solution has come up if you keep the RB tree (I don't).
>
> It is, however, even more invasive than the this one ;-)
>
> Wolfram has been working on implementing a threaded RB-tree. This means
> rb_prev() and rb_next() will be O(1) operations, so you could simply use
> those to iterate the vmas.
>
> The only draw-back is that each and every RB-tree user in the kernel
> needs to be adapted because its not quite possible to maintain the
> current API.
>
> I was planning to help Wolfram do that, but I'm utterly swamped atm. :-(
>
> What needs to be done is introduce rb_left(), rb_right() and rb_node()
> helpers that for now look like:
>
> static inline struct rb_node *rb_left(struct rb_node *n)
> {
> return n->rb_left;
> }
>
> static inline struct rb_node *rb_right(struct rb_node *n)
> {
> return n->rb_right;
> }
>
> static inline struct rb_node *rb_node(struct rb_node *n)
> {
> return n;
> }
>
> We need these because the left and right child pointers will be
> over-loaded with threading information.
>
> After that we can flip the implementation of the RB-tree.
>
>
>
>
prev parent reply other threads:[~2009-04-02 9:56 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-03-11 9:55 Daniel Lowengrub
2009-03-11 10:40 ` Alexey Dobriyan
2009-03-11 10:52 ` Ingo Molnar
2009-03-11 11:54 ` Nick Piggin
2009-03-11 13:25 ` Daniel Lowengrub
2009-03-12 3:00 ` Nick Piggin
2009-03-12 8:32 ` KOSAKI Motohiro
2009-03-12 9:58 ` Peter Zijlstra
2009-04-02 9:56 ` Daniel Lowengrub [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=8c5a844a0904020256r3e073d01o1991e5111127ce42@mail.gmail.com \
--to=lowdanie@gmail.com \
--cc=a.p.zijlstra@chello.nl \
--cc=kosaki.motohiro@jp.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mingo@elte.hu \
--cc=nickpiggin@yahoo.com.au \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®