[PATCH] fix file counting

I have benchmarked this on an x86_64 NUMA system and see no significant
performance difference on kernbench.  Tested on both x86_64 and powerpc.

The way we do file struct accounting is not very suitable for batched
freeing.  For scalability reasons, file accounting was
constructor/destructor based.  This meant that nr_files was decremented
only when the object was removed from the slab cache.  This is susceptible
to slab fragmentation.  With RCU based file structure, consequent batched
freeing and a test program like Serge's, we just speed this up and end up
with a very fragmented slab -

llm22:~ # cat /proc/sys/fs/file-nr
587730  0       758844

At the same time, I see only a 2000+ objects in filp cache.  The following
patch I fixes this problem.

This patch changes the file counting by removing the filp_count_lock.
Instead we use a separate percpu counter, nr_files, for now and all
accesses to it are through get_nr_files() api.  In the sysctl handler for
nr_files, we populate files_stat.nr_files before returning to user.

Counting files as an when they are created and destroyed (as opposed to
inside slab) allows us to correctly count open files with RCU.

Signed-off-by: Dipankar Sarma <dipankar@in.ibm.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

6 files changed, 62 insertions, 37 deletions

diff --git a/fs/dcache.c b/fs/dcache.c
index a173bba..11dc830 100644
--- a/fs/dcache.c
+++ b/fs/dcache.c
@@ -1736,7 +1736,7 @@ void __init vfs_caches_init(unsigned long mempages)
 			SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
 
 	filp_cachep = kmem_cache_create("filp", sizeof(struct file), 0,
-			SLAB_HWCACHE_ALIGN|SLAB_PANIC, filp_ctor, filp_dtor);
+			SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
 
 	dcache_init(mempages);
 	inode_init(mempages);
diff --git a/fs/file_table.c b/fs/file_table.c
index 768b581..44fabea 100644
--- a/fs/file_table.c
+++ b/fs/file_table.c
@@ -5,6 +5,7 @@
  *  Copyright (C) 1997 David S. Miller (davem@caip.rutgers.edu)
  */
 
+#include <linux/config.h>
 #include <linux/string.h>
 #include <linux/slab.h>
 #include <linux/file.h>
@@ -19,52 +20,67 @@
 #include <linux/capability.h>
 #include <linux/cdev.h>
 #include <linux/fsnotify.h>
+#include <linux/sysctl.h>
+#include <linux/percpu_counter.h>
+
+#include <asm/atomic.h>
 
 /* sysctl tunables... */
 struct files_stat_struct files_stat = {
 	.max_files = NR_FILE
 };
 
-EXPORT_SYMBOL(files_stat); /* Needed by unix.o */
-
 /* public. Not pretty! */
- __cacheline_aligned_in_smp DEFINE_SPINLOCK(files_lock);
+__cacheline_aligned_in_smp DEFINE_SPINLOCK(files_lock);
 
-static DEFINE_SPINLOCK(filp_count_lock);
+static struct percpu_counter nr_files __cacheline_aligned_in_smp;
 
-/* slab constructors and destructors are called from arbitrary
- * context and must be fully threaded - use a local spinlock
- * to protect files_stat.nr_files
- */
-void filp_ctor(void *objp, struct kmem_cache *cachep, unsigned long cflags)
+static inline void file_free_rcu(struct rcu_head *head)
 {
-	if ((cflags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
-	    SLAB_CTOR_CONSTRUCTOR) {
-		unsigned long flags;
-		spin_lock_irqsave(&filp_count_lock, flags);
-		files_stat.nr_files++;
-		spin_unlock_irqrestore(&filp_count_lock, flags);
-	}
+	struct file *f =  container_of(head, struct file, f_u.fu_rcuhead);
+	kmem_cache_free(filp_cachep, f);
 }
 
-void filp_dtor(void *objp, struct kmem_cache *cachep, unsigned long dflags)
+static inline void file_free(struct file *f)
 {
-	unsigned long flags;
-	spin_lock_irqsave(&filp_count_lock, flags);
-	files_stat.nr_files--;
-	spin_unlock_irqrestore(&filp_count_lock, flags);
+	percpu_counter_dec(&nr_files);
+	call_rcu(&f->f_u.fu_rcuhead, file_free_rcu);
 }
 
-static inline void file_free_rcu(struct rcu_head *head)
+/*
+ * Return the total number of open files in the system
+ */
+static int get_nr_files(void)
 {
-	struct file *f =  container_of(head, struct file, f_u.fu_rcuhead);
-	kmem_cache_free(filp_cachep, f);
+	return percpu_counter_read_positive(&nr_files);
 }
 
-static inline void file_free(struct file *f)
+/*
+ * Return the maximum number of open files in the system
+ */
+int get_max_files(void)
 {
-	call_rcu(&f->f_u.fu_rcuhead, file_free_rcu);
+	return files_stat.max_files;
 }
+EXPORT_SYMBOL_GPL(get_max_files);
+
+/*
+ * Handle nr_files sysctl
+ */
+#if defined(CONFIG_SYSCTL) && defined(CONFIG_PROC_FS)
+int proc_nr_files(ctl_table *table, int write, struct file *filp,
+                     void __user *buffer, size_t *lenp, loff_t *ppos)
+{
+	files_stat.nr_files = get_nr_files();
+	return proc_dointvec(table, write, filp, buffer, lenp, ppos);
+}
+#else
+int proc_nr_files(ctl_table *table, int write, struct file *filp,
+                     void __user *buffer, size_t *lenp, loff_t *ppos)
+{
+	return -ENOSYS;
+}
+#endif
 
 /* Find an unused file structure and return a pointer to it.
  * Returns NULL, if there are no more free file structures or
@@ -78,14 +94,20 @@ struct file *get_empty_filp(void)
 	/*
 	 * Privileged users can go above max_files
 	 */
-	if (files_stat.nr_files >= files_stat.max_files &&
-				!capable(CAP_SYS_ADMIN))
-		goto over;
+	if (get_nr_files() >= files_stat.max_files && !capable(CAP_SYS_ADMIN)) {
+		/*
+		 * percpu_counters are inaccurate.  Do an expensive check before
+		 * we go and fail.
+		 */
+		if (percpu_counter_sum(&nr_files) >= files_stat.max_files)
+			goto over;
+	}
 
 	f = kmem_cache_alloc(filp_cachep, GFP_KERNEL);
 	if (f == NULL)
 		goto fail;
 
+	percpu_counter_inc(&nr_files);
 	memset(f, 0, sizeof(*f));
 	if (security_file_alloc(f))
 		goto fail_sec;
@@ -101,10 +123,10 @@ struct file *get_empty_filp(void)
 
 over:
 	/* Ran out of filps - report that */
-	if (files_stat.nr_files > old_max) {
+	if (get_nr_files() > old_max) {
 		printk(KERN_INFO "VFS: file-max limit %d reached\n",
-					files_stat.max_files);
-		old_max = files_stat.nr_files;
+					get_max_files());
+		old_max = get_nr_files();
 	}
 	goto fail;
 
@@ -276,4 +298,5 @@ void __init files_init(unsigned long mempages)
 	if (files_stat.max_files < NR_FILE)
 		files_stat.max_files = NR_FILE;
 	files_defer_init();
+	percpu_counter_init(&nr_files);
 } 
diff --git a/include/linux/file.h b/include/linux/file.h
index 418b610..9901b85 100644
--- a/include/linux/file.h
+++ b/include/linux/file.h
@@ -60,8 +60,6 @@ extern void put_filp(struct file *);
 extern int get_unused_fd(void);
 extern void FASTCALL(put_unused_fd(unsigned int fd));
 struct kmem_cache;
-extern void filp_ctor(void * objp, struct kmem_cache *cachep, unsigned long cflags);
-extern void filp_dtor(void * objp, struct kmem_cache *cachep, unsigned long dflags);
 
 extern struct file ** alloc_fd_array(int);
 extern void free_fd_array(struct file **, int);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 0cc34b1..51c0c93 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -35,6 +35,7 @@ struct files_stat_struct {
 	int max_files;		/* tunable */
 };
 extern struct files_stat_struct files_stat;
+extern int get_max_files(void);
 
 struct inodes_stat_t {
 	int nr_inodes;
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index de2d910..32b48e8 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -50,6 +50,9 @@
 #include <asm/uaccess.h>
 #include <asm/processor.h>
 
+extern int proc_nr_files(ctl_table *table, int write, struct file *filp,
+                     void __user *buffer, size_t *lenp, loff_t *ppos);
+
 #if defined(CONFIG_SYSCTL)
 
 /* External variables not in a header file. */
@@ -943,7 +946,7 @@ static ctl_table fs_table[] = {
 		.data		= &files_stat,
 		.maxlen		= 3*sizeof(int),
 		.mode		= 0444,
-		.proc_handler	= &proc_dointvec,
+		.proc_handler	= &proc_nr_files,
 	},
 	{
 		.ctl_name	= FS_MAXFILE,
diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index 1b5989b..c323cc6 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c
@@ -547,7 +547,7 @@ static struct sock * unix_create1(struct socket *sock)
 	struct sock *sk = NULL;
 	struct unix_sock *u;
 
-	if (atomic_read(&unix_nr_socks) >= 2*files_stat.max_files)
+	if (atomic_read(&unix_nr_socks) >= 2*get_max_files())
 		goto out;
 
 	sk = sk_alloc(PF_UNIX, GFP_KERNEL, &unix_proto, 1);