All of lore.kernel.org
 help / color / mirror / Atom feed
From: Andi Kleen <andi@firstfloor.org>
To: linux-mm@kvack.org
Cc: Andrea Arcangeli <aarcange@redhat.com>,
	Rik van Riel <riel@redhat.com>,
	tim.c.chen@linux.intel.com, torvalds@linux-foundation.org,
	lwoodman@redhat.com, mel@csn.ul.ie,
	Andi Kleen <ak@linux.intel.com>
Subject: [PATCH 1/4] VM/RMAP: Add infrastructure for batching the rmap chain locking
Date: Thu,  5 May 2011 12:32:49 -0700	[thread overview]
Message-ID: <1304623972-9159-2-git-send-email-andi@firstfloor.org> (raw)
In-Reply-To: <1304623972-9159-1-git-send-email-andi@firstfloor.org>

From: Andi Kleen <ak@linux.intel.com>

In fork and exit it's quite common to take same rmap chain locks
again and again when the whole address space is processed  for a
address space that has a lot of sharing. Also since the locking
has changed to always lock the root anon_vma this can be very
contended.

This patch adds a simple wrapper to batch these lock acquisitions
and only reaquire the lock when another is needed. The main
advantage is that when multiple processes are doing this in
parallel they will avoid a lot of communication overhead
on the lock cache line.

I added a simple lock break (100 locks) for paranoia reason,
but it's unclear if that's needed or not.

Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
 include/linux/rmap.h |   38 ++++++++++++++++++++++++++++++++++++++
 1 files changed, 38 insertions(+), 0 deletions(-)

diff --git a/include/linux/rmap.h b/include/linux/rmap.h
index 830e65d..d5bb9f8 100644
--- a/include/linux/rmap.h
+++ b/include/linux/rmap.h
@@ -113,6 +113,44 @@ static inline void anon_vma_unlock(struct anon_vma *anon_vma)
 	spin_unlock(&anon_vma->root->lock);
 }
 
+/* 
+ * Batched locking for anon VMA chains to avoid too much cache line 
+ * bouncing.
+ */
+
+#define AVL_LOCKBREAK 500
+
+struct anon_vma_lock_state {
+	struct anon_vma *root_anon_vma;
+	int counter;
+};
+
+static inline void init_anon_vma_lock_batch(struct anon_vma_lock_state *avs)
+{
+	avs->root_anon_vma = NULL;
+	avs->counter = 0;
+}
+
+static inline void anon_vma_lock_batch(struct anon_vma *anon_vma,
+				       struct anon_vma_lock_state *state)
+{
+	if (state->root_anon_vma == anon_vma->root &&
+	    state->counter++ < AVL_LOCKBREAK)
+		return;
+	if (state->root_anon_vma) {
+		state->counter = 0;
+		spin_unlock(&state->root_anon_vma->lock);
+	}
+	state->root_anon_vma = anon_vma->root;
+	spin_lock(&state->root_anon_vma->lock);
+}
+
+static inline void anon_vma_unlock_batch(struct anon_vma_lock_state *avs)
+{
+	if (avs->root_anon_vma)
+		spin_unlock(&avs->root_anon_vma->lock);
+}
+
 /*
  * anon_vma helper functions.
  */
-- 
1.7.4.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2011-05-05 19:33 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-05-05 19:32 Batch locking for rmap fork/exit processing Andi Kleen
2011-05-05 19:32 ` Andi Kleen [this message]
2011-05-06 15:59   ` [PATCH 1/4] VM/RMAP: Add infrastructure for batching the rmap chain locking Rik van Riel
2011-05-09 21:43   ` Andrew Morton
2011-05-09 22:23     ` Andi Kleen
2011-05-09 22:28       ` Andrew Morton
2011-05-09 23:02         ` Andi Kleen
2011-05-09 23:16           ` Andrew Morton
2011-05-10  0:05             ` Andi Kleen
2011-05-10  1:26               ` Andrew Morton
2011-05-10  1:27                 ` Rik van Riel
2011-05-10  1:48                   ` Andrew Morton
2011-05-05 19:32 ` [PATCH 2/4] VM/RMAP: Batch anon vma chain root locking in fork Andi Kleen
2011-05-06 15:59   ` Rik van Riel
2011-05-05 19:32 ` [PATCH 3/4] VM/RMAP: Batch anon_vma_unlink in exit Andi Kleen
2011-05-06 16:48   ` Rik van Riel
2011-05-05 19:32 ` [PATCH 4/4] VM/RMAP: Move avc freeing outside the lock Andi Kleen
2011-05-06 16:52   ` Rik van Riel
2011-05-06 16:44 ` Batch locking for rmap fork/exit processing Linus Torvalds
2011-05-06 17:05   ` Andi Kleen

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=1304623972-9159-2-git-send-email-andi@firstfloor.org \
    --to=andi@firstfloor.org \
    --cc=aarcange@redhat.com \
    --cc=ak@linux.intel.com \
    --cc=linux-mm@kvack.org \
    --cc=lwoodman@redhat.com \
    --cc=mel@csn.ul.ie \
    --cc=riel@redhat.com \
    --cc=tim.c.chen@linux.intel.com \
    --cc=torvalds@linux-foundation.org \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.