linux-next.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Linus Torvalds <torvalds@linux-foundation.org>
To: Al Viro <viro@zeniv.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Stephen Rothwell <sfr@canb.auug.org.au>,
	linux-next <linux-next@vger.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Dave Chinner <dchinner@redhat.com>,
	Glauber Costa <glommer@openvz.org>
Subject: Re: linux-next: manual merge of the akpm tree with Linus' tree
Date: Fri, 13 Sep 2013 12:12:20 -0700	[thread overview]
Message-ID: <CA+55aFyaRsk-KkjppTnqDxy8+y5z9Yvi3TpMZwgGJ7ad7kwM4g@mail.gmail.com> (raw)
In-Reply-To: <CA+55aFxnq2hjx1Aw7+iFuGnZp4sRK3HYK0Kfipz1gEYubOdnYQ@mail.gmail.com>

[-- Attachment #1: Type: text/plain, Size: 1573 bytes --]

On Thu, Sep 12, 2013 at 6:12 PM, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> Damn, the code is too confused. I have to go to a highschool parent
> back-to-school meeting, so I won't get to this until maybe on a plane
> tomorrow. Al, can you please give this a look?

I'm on a plane. I have gogo. Here's my current TOTALLY UNTESTED patch.

It tries to consolidate the dentry LRU stuff into a few helper
functions that right now have anal checking of the flags. Maybe I
overdid it, but the code was really confusing, and I think we got the
free dentry counts wrong, and the bits wrong too, so I tried to be
extra careful.

There are several cases:
 - d_lru_add/del: fairly obvious
 - d_lru_isolate: this is when the LRU callbacks ask us to remove the
entry from the list. This is different from d_lru_del() only in that
it uses the raw list removal, not the lru list helper function. I'm
not sure that's right, but that's what the code used to do.
 - d_lru_shrink_move: move from the "global" lru list to a private shrinker list
 - d_shrink_add/del: fairly obvious.

And then "denty_lru_add/del" that actually take the current state into
account and do the right thing. Those we had before, I'm just
explaining the difference from the low-level operations that have
fixed "from this state to that" semantics

Hmm?

Does it work? Who knows.. But *if* it works, I think it has a higher
chance of getting all the rules for bits and free object counting
right.

Somebody not in a plane please double-check my low-oxygen-pressure thinking..

                  Linus

[-- Attachment #2: patch.diff --]
[-- Type: application/octet-stream, Size: 6399 bytes --]

 fs/dcache.c | 122 ++++++++++++++++++++++++++++++++++++++++++++++--------------
 1 file changed, 95 insertions(+), 27 deletions(-)

diff --git a/fs/dcache.c b/fs/dcache.c
index 1bd4614ce93b..6443b07f8a3c 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -357,15 +357,74 @@ static void dentry_unlink_inode(struct dentry * dentry)
 }
 
 /*
+ * The DCACHE_LRU_LIST bit is set whenever the 'd_lru' entry
+ * is in use - which includes both the "real" per-superblock
+ * LRU list _and_ the DCACHE_SHRINK_LIST use.
+ *
+ * The DCACHE_SHRINK_LIST bit is set whenever the dentry is
+ * on the shrink list (ie not on the superblock LRU list).
+ *
+ * The per-cpu "nd_dentry_unused" counters are updated with
+ * the DCACHE_LRU_LIST bit.
+ *
+ * These helper functions make sure we always follow the
+ * rules. d_lock must be held by the caller.
+ */
+#define D_FLAG_VERIFY(dentry,x) WARN_ON_ONCE(((dentry)->d_flags & (DCACHE_LRU_LIST | DCACHE_SHRINK_LIST)) != (x))
+static void d_lru_add(struct dentry *dentry)
+{
+	D_FLAG_VERIFY(dentry, 0);
+	dentry->d_flags |= DCACHE_LRU_LIST;
+	this_cpu_inc(nr_dentry_unused);
+	WARN_ON_ONCE(!list_lru_add(&dentry->d_sb->s_dentry_lru, &dentry->d_lru));
+}
+
+static void d_lru_del(struct dentry *dentry)
+{
+	D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
+	dentry->d_flags &= ~DCACHE_LRU_LIST;
+	this_cpu_dec(nr_dentry_unused);
+	WARN_ON_ONCE(!list_lru_del(&dentry->d_sb->s_dentry_lru, &dentry->d_lru));
+}
+
+static void d_lru_isolate(struct dentry *dentry)
+{
+	D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
+	dentry->d_flags &= ~DCACHE_LRU_LIST;
+	this_cpu_dec(nr_dentry_unused);
+	list_del_init(&dentry->d_lru);
+}
+
+static void d_shrink_del(struct dentry *dentry)
+{
+	D_FLAG_VERIFY(dentry, DCACHE_SHRINK_LIST | DCACHE_LRU_LIST);
+	list_del_init(&dentry->d_lru);
+	dentry->d_flags &= ~(DCACHE_SHRINK_LIST | DCACHE_LRU_LIST);
+	this_cpu_dec(nr_dentry_unused);
+}
+
+static void d_shrink_add(struct dentry *dentry, struct list_head *list)
+{
+	D_FLAG_VERIFY(dentry, 0);
+	list_add(&dentry->d_lru, list);
+	dentry->d_flags |= DCACHE_SHRINK_LIST | DCACHE_LRU_LIST;
+	this_cpu_inc(nr_dentry_unused);
+}
+
+static void d_lru_shrink_move(struct dentry *dentry, struct list_head *list)
+{
+	D_FLAG_VERIFY(dentry, DCACHE_LRU_LIST);
+	dentry->d_flags |= DCACHE_SHRINK_LIST;
+	list_move_tail(&dentry->d_lru, list);
+}
+
+/*
  * dentry_lru_(add|del)_list) must be called with d_lock held.
  */
 static void dentry_lru_add(struct dentry *dentry)
 {
-	if (unlikely(!(dentry->d_flags & DCACHE_LRU_LIST))) {
-		if (list_lru_add(&dentry->d_sb->s_dentry_lru, &dentry->d_lru))
-			this_cpu_inc(nr_dentry_unused);
-		dentry->d_flags |= DCACHE_LRU_LIST;
-	}
+	if (unlikely(!(dentry->d_flags & DCACHE_LRU_LIST)))
+		d_lru_add(dentry);
 }
 
 /*
@@ -377,15 +436,11 @@ static void dentry_lru_add(struct dentry *dentry)
  */
 static void dentry_lru_del(struct dentry *dentry)
 {
-	if (dentry->d_flags & DCACHE_SHRINK_LIST) {
-		list_del_init(&dentry->d_lru);
-		dentry->d_flags &= ~DCACHE_SHRINK_LIST;
-		return;
+	if (dentry->d_flags & DCACHE_LRU_LIST) {
+		if (dentry->d_flags & DCACHE_SHRINK_LIST)
+			return d_shrink_del(dentry);
+		d_lru_del(dentry);
 	}
-
-	if (list_lru_del(&dentry->d_sb->s_dentry_lru, &dentry->d_lru))
-		this_cpu_dec(nr_dentry_unused);
-	dentry->d_flags &= ~DCACHE_LRU_LIST;
 }
 
 /**
@@ -837,6 +892,13 @@ static void shrink_dentry_list(struct list_head *list)
 		dentry = list_entry_rcu(list->prev, struct dentry, d_lru);
 		if (&dentry->d_lru == list)
 			break; /* empty */
+
+
+		/*
+		 * Get the dentry lock, and re-verify that the dentry is
+		 * this on the shrinking list. If it is, we know that
+		 * DCACHE_SHRINK_LIST and DCACHE_LRU_LIST are set.
+		 */
 		spin_lock(&dentry->d_lock);
 		if (dentry != list_entry(list->prev, struct dentry, d_lru)) {
 			spin_unlock(&dentry->d_lock);
@@ -848,8 +910,7 @@ static void shrink_dentry_list(struct list_head *list)
 		 * to the LRU here, so we can simply remove it from the list
 		 * here regardless of whether it is referenced or not.
 		 */
-		list_del_init(&dentry->d_lru);
-		dentry->d_flags &= ~DCACHE_SHRINK_LIST;
+		d_shrink_del(dentry);
 
 		/*
 		 * We found an inuse dentry which was not removed from
@@ -861,12 +922,20 @@ static void shrink_dentry_list(struct list_head *list)
 		}
 		rcu_read_unlock();
 
+		/*
+		 * If 'try_to_prune()' returns a dentry, it will
+		 * be the same one we passed in, and d_lock will
+		 * have been held the whole time, so it will not
+		 * have been added to any other lists. We failed
+		 * to get the inode lock.
+		 *
+		 * We just add it back to the shrink list.
+		 */
 		dentry = try_prune_one_dentry(dentry);
 
 		rcu_read_lock();
 		if (dentry) {
-			dentry->d_flags |= DCACHE_SHRINK_LIST;
-			list_add(&dentry->d_lru, list);
+			d_shrink_add(dentry, list);
 			spin_unlock(&dentry->d_lock);
 		}
 	}
@@ -894,7 +963,7 @@ dentry_lru_isolate(struct list_head *item, spinlock_t *lru_lock, void *arg)
 	 * another pass through the LRU.
 	 */
 	if (dentry->d_lockref.count) {
-		list_del_init(&dentry->d_lru);
+		d_lru_isolate(dentry);
 		spin_unlock(&dentry->d_lock);
 		return LRU_REMOVED;
 	}
@@ -925,9 +994,7 @@ dentry_lru_isolate(struct list_head *item, spinlock_t *lru_lock, void *arg)
 		return LRU_ROTATE;
 	}
 
-	dentry->d_flags |= DCACHE_SHRINK_LIST;
-	list_move_tail(&dentry->d_lru, freeable);
-	this_cpu_dec(nr_dentry_unused);
+	d_lru_shrink_move(dentry, freeable);
 	spin_unlock(&dentry->d_lock);
 
 	return LRU_REMOVED;
@@ -972,9 +1039,7 @@ static enum lru_status dentry_lru_isolate_shrink(struct list_head *item,
 	if (!spin_trylock(&dentry->d_lock))
 		return LRU_SKIP;
 
-	dentry->d_flags |= DCACHE_SHRINK_LIST;
-	list_move_tail(&dentry->d_lru, freeable);
-	this_cpu_dec(nr_dentry_unused);
+	d_lru_shrink_move(dentry, freeable);
 	spin_unlock(&dentry->d_lock);
 
 	return LRU_REMOVED;
@@ -1362,9 +1427,12 @@ static enum d_walk_ret select_collect(void *_data, struct dentry *dentry)
 	if (dentry->d_lockref.count) {
 		dentry_lru_del(dentry);
 	} else if (!(dentry->d_flags & DCACHE_SHRINK_LIST)) {
-		dentry_lru_del(dentry);
-		list_add_tail(&dentry->d_lru, &data->dispose);
-		dentry->d_flags |= DCACHE_SHRINK_LIST;
+		/*
+		 * If we get here, lockref.count is zero, and
+		 * the dentry should be on the LRU list. So we
+		 * can just use the "move" helper.
+		 */
+		d_lru_shrink_move(dentry, &data->dispose);
 		data->found++;
 		ret = D_WALK_NORETRY;
 	}

  parent reply	other threads:[~2013-09-13 19:12 UTC|newest]

Thread overview: 106+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-09-10  4:38 linux-next: manual merge of the akpm tree with Linus' tree Stephen Rothwell
2013-09-10 22:27 ` Andrew Morton
2013-09-10 22:29   ` Al Viro
2013-09-10 22:35     ` Andrew Morton
2013-09-10 22:36       ` Al Viro
2013-09-10 22:39         ` Al Viro
2013-09-10 22:41         ` Andrew Morton
2013-09-10 22:48           ` Al Viro
2013-09-10 22:59             ` Al Viro
2013-09-10 23:13               ` Andrew Morton
2013-09-10 23:55                 ` Al Viro
2013-09-11  4:30                 ` Stephen Rothwell
2013-09-10 23:37               ` Linus Torvalds
2013-09-10 23:53                 ` Al Viro
2013-09-11  0:01                   ` Linus Torvalds
2013-09-11  0:39                     ` Dave Chinner
2013-09-13  0:56                 ` Linus Torvalds
2013-09-13  1:12                   ` Linus Torvalds
2013-09-13  1:35                     ` Al Viro
2013-09-13 19:12                     ` Linus Torvalds [this message]
2013-09-13 19:28                       ` Linus Torvalds
2013-09-13 19:54                       ` Linus Torvalds
2013-09-13 20:00                       ` Al Viro
2013-09-13 20:18                         ` Al Viro
2013-09-13 20:23                           ` Al Viro
2013-09-13 20:25                         ` Linus Torvalds
2013-09-13 20:31                           ` Linus Torvalds
2013-09-13 20:31                           ` Al Viro
2013-09-13 20:34                             ` Linus Torvalds
2013-09-10 22:35   ` Linus Torvalds
2013-09-10 22:44     ` Andrew Morton
2013-09-11  0:30       ` Stephen Rothwell
2013-09-11  0:41         ` Linus Torvalds
  -- strict thread matches above, loose matches on Subject: below --
2019-12-02  2:17 Stephen Rothwell
2019-12-02  2:08 Stephen Rothwell
2018-10-15  7:22 Stephen Rothwell
2018-10-15  7:04 Stephen Rothwell
2018-03-26  8:56 Stephen Rothwell
2017-01-09  2:51 Stephen Rothwell
2016-12-12  5:49 Stephen Rothwell
2016-01-19  2:42 Stephen Rothwell
2016-01-19  2:36 Stephen Rothwell
2015-07-27  5:26 Stephen Rothwell
2013-09-10  4:21 Stephen Rothwell
2013-09-10  4:09 Stephen Rothwell
2013-09-10  4:12 ` Stephen Rothwell
2013-09-09  5:38 Stephen Rothwell
2013-09-09  5:22 Stephen Rothwell
2013-09-09  5:16 Stephen Rothwell
2013-08-30  8:44 Stephen Rothwell
2013-05-27  6:20 Stephen Rothwell
2013-05-20  4:04 Stephen Rothwell
2013-05-20 12:19 ` Chris Mason
2013-05-13  5:20 Eric Paris
2013-05-13  2:07 Stephen Rothwell
2013-05-13  2:11 ` Eric Paris
2013-05-13  4:16   ` Stephen Rothwell
2013-05-13  4:49   ` Kees Cook
2013-05-13  5:14     ` Eric Paris
2013-05-02  6:01 Stephen Rothwell
2013-04-29  8:38 Stephen Rothwell
2013-04-19  7:40 Stephen Rothwell
2013-04-16  7:25 Stephen Rothwell
2013-04-03  6:10 Stephen Rothwell
2013-03-25  4:22 Stephen Rothwell
2013-03-04  2:21 Stephen Rothwell
2013-01-04  3:27 Stephen Rothwell
2012-12-11  5:25 Stephen Rothwell
2012-12-11  5:22 Stephen Rothwell
2012-12-11  7:58 ` Glauber Costa
2012-12-07  6:39 Stephen Rothwell
2012-11-30  6:24 Stephen Rothwell
2012-11-26 12:52 Stephen Rothwell
2012-11-26 12:48 Stephen Rothwell
2012-11-26 13:25 ` Xiaotian Feng
2012-11-26 12:34 Stephen Rothwell
2012-10-15  2:07 Stephen Rothwell
2012-10-15 22:14 ` Catalin Marinas
2012-10-01 14:15 Stephen Rothwell
2012-08-22  5:59 Stephen Rothwell
2012-08-22  8:58 ` Mel Gorman
2012-07-27  3:57 Stephen Rothwell
2012-07-02  6:39 Stephen Rothwell
2012-06-04  4:58 Stephen Rothwell
2012-05-31  4:24 Stephen Rothwell
2012-05-31  4:13 Stephen Rothwell
2012-05-31  7:25 ` Johannes Weiner
2012-05-31  8:24   ` Stephen Rothwell
2012-05-31  8:27     ` Stephen Rothwell
2012-05-21  8:23 Stephen Rothwell
2012-05-21  8:13 Stephen Rothwell
2012-05-21  8:16 ` Cyrill Gorcunov
2012-03-08  6:53 Stephen Rothwell
2012-03-08  7:32 ` Andrew Morton
2012-03-08  7:41   ` Stephen Rothwell
2012-03-08  7:50     ` Andrew Morton
2012-03-08  7:50       ` Xiao Guangrong
2012-03-08  9:59       ` Xiao Guangrong
2012-03-08 21:24         ` Andrew Morton
2012-03-08 23:42           ` Linus Torvalds
2011-12-28  7:54 Stephen Rothwell
2011-11-08  3:24 Stephen Rothwell
2011-11-01  8:16 Stephen Rothwell
2011-11-01 10:47 ` Tao Ma
2011-09-16  6:09 Stephen Rothwell
2011-08-15  4:52 Stephen Rothwell

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=CA+55aFyaRsk-KkjppTnqDxy8+y5z9Yvi3TpMZwgGJ7ad7kwM4g@mail.gmail.com \
    --to=torvalds@linux-foundation.org \
    --cc=akpm@linux-foundation.org \
    --cc=dchinner@redhat.com \
    --cc=glommer@openvz.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-next@vger.kernel.org \
    --cc=sfr@canb.auug.org.au \
    --cc=viro@zeniv.linux.org.uk \
    /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;
as well as URLs for NNTP newsgroup(s).