[v3,3/3] vfs: Use per-cpu list for superblock's inode list
diff mbox

Message ID 20160224082840.GB10096@quack.suse.cz
State New
Headers show

Commit Message

Jan Kara Feb. 24, 2016, 8:28 a.m. UTC
On Tue 23-02-16 14:04:32, Waiman Long wrote:
> When many threads are trying to add or delete inode to or from
> a superblock's s_inodes list, spinlock contention on the list can
> become a performance bottleneck.
> 
> This patch changes the s_inodes field to become a per-cpu list with
> per-cpu spinlocks. As a result, the following superblock inode list
> (sb->s_inodes) iteration functions in vfs are also being modified:
> 
>  1. iterate_bdevs()
>  2. drop_pagecache_sb()
>  3. wait_sb_inodes()
>  4. evict_inodes()
>  5. invalidate_inodes()
>  6. fsnotify_unmount_inodes()
>  7. add_dquot_ref()
>  8. remove_dquot_ref()
> 
> With an exit microbenchmark that creates a large number of threads,
> attachs many inodes to them and then exits. The runtimes of that
> microbenchmark with 1000 threads before and after the patch on a
> 4-socket Intel E7-4820 v3 system (40 cores, 80 threads) were as
> follows:
> 
>   Kernel            Elapsed Time    System Time
>   ------            ------------    -----------
>   Vanilla 4.5-rc4      65.29s         82m14s
>   Patched 4.5-rc4      22.81s         23m03s
> 
> Before the patch, spinlock contention at the inode_sb_list_add()
> function at the startup phase and the inode_sb_list_del() function at
> the exit phase were about 79% and 93% of total CPU time respectively
> (as measured by perf). After the patch, the percpu_list_add()
> function consumed only about 0.04% of CPU time at startup phase. The
> percpu_list_del() function consumed about 0.4% of CPU time at exit
> phase. There were still some spinlock contention, but they happened
> elsewhere.

While looking through this patch, I have noticed that the
list_for_each_entry_safe() iterations in evict_inodes() and
invalidate_inodes() are actually unnecessary. So if you first apply the
attached patch, you don't have to implement safe iteration variants at all.

As a second comment, I'd note that this patch grows struct inode by 1
pointer. It is probably acceptable for large machines given the speedup but
it should be noted in the changelog. Furthermore for UP or even small SMP
systems this is IMHO undesired bloat since the speedup won't be noticeable.

So for these small systems it would be good if per-cpu list magic would just
fall back to single linked list with a spinlock. Do you think that is
reasonably doable?

								Honza

> 
> Signed-off-by: Waiman Long <Waiman.Long@hpe.com>
> ---
>  fs/block_dev.c         |   13 +++++++------
>  fs/drop_caches.c       |   10 +++++-----
>  fs/fs-writeback.c      |   13 +++++++------
>  fs/inode.c             |   40 +++++++++++++++++-----------------------
>  fs/notify/inode_mark.c |   10 +++++-----
>  fs/quota/dquot.c       |   16 ++++++++--------
>  fs/super.c             |    7 ++++---
>  include/linux/fs.h     |    8 ++++----
>  8 files changed, 57 insertions(+), 60 deletions(-)
> 
> diff --git a/fs/block_dev.c b/fs/block_dev.c
> index 39b3a17..759d9b6 100644
> --- a/fs/block_dev.c
> +++ b/fs/block_dev.c
> @@ -1865,11 +1865,13 @@ EXPORT_SYMBOL(__invalidate_device);
>  void iterate_bdevs(void (*func)(struct block_device *, void *), void *arg)
>  {
>  	struct inode *inode, *old_inode = NULL;
> +	DEFINE_PCPU_LIST_STATE(state);
>  
> -	spin_lock(&blockdev_superblock->s_inode_list_lock);
> -	list_for_each_entry(inode, &blockdev_superblock->s_inodes, i_sb_list) {
> -		struct address_space *mapping = inode->i_mapping;
> +	while (pcpu_list_iterate(blockdev_superblock->s_inodes, &state)) {
> +		struct address_space *mapping;
>  
> +		inode   = list_entry(state.curr, struct inode, i_sb_list);
> +		mapping = inode->i_mapping;
>  		spin_lock(&inode->i_lock);
>  		if (inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW) ||
>  		    mapping->nrpages == 0) {
> @@ -1878,7 +1880,7 @@ void iterate_bdevs(void (*func)(struct block_device *, void *), void *arg)
>  		}
>  		__iget(inode);
>  		spin_unlock(&inode->i_lock);
> -		spin_unlock(&blockdev_superblock->s_inode_list_lock);
> +		spin_unlock(state.lock);
>  		/*
>  		 * We hold a reference to 'inode' so it couldn't have been
>  		 * removed from s_inodes list while we dropped the
> @@ -1892,8 +1894,7 @@ void iterate_bdevs(void (*func)(struct block_device *, void *), void *arg)
>  
>  		func(I_BDEV(inode), arg);
>  
> -		spin_lock(&blockdev_superblock->s_inode_list_lock);
> +		spin_lock(state.lock);
>  	}
> -	spin_unlock(&blockdev_superblock->s_inode_list_lock);
>  	iput(old_inode);
>  }
> diff --git a/fs/drop_caches.c b/fs/drop_caches.c
> index d72d52b..ec272ed 100644
> --- a/fs/drop_caches.c
> +++ b/fs/drop_caches.c
> @@ -16,9 +16,10 @@ int sysctl_drop_caches;
>  static void drop_pagecache_sb(struct super_block *sb, void *unused)
>  {
>  	struct inode *inode, *toput_inode = NULL;
> +	DEFINE_PCPU_LIST_STATE(state);
>  
> -	spin_lock(&sb->s_inode_list_lock);
> -	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
> +	while (pcpu_list_iterate(sb->s_inodes, &state)) {
> +		inode = list_entry(state.curr, struct inode, i_sb_list);
>  		spin_lock(&inode->i_lock);
>  		if ((inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) ||
>  		    (inode->i_mapping->nrpages == 0)) {
> @@ -27,15 +28,14 @@ static void drop_pagecache_sb(struct super_block *sb, void *unused)
>  		}
>  		__iget(inode);
>  		spin_unlock(&inode->i_lock);
> -		spin_unlock(&sb->s_inode_list_lock);
> +		spin_unlock(state.lock);
>  
>  		invalidate_mapping_pages(inode->i_mapping, 0, -1);
>  		iput(toput_inode);
>  		toput_inode = inode;
>  
> -		spin_lock(&sb->s_inode_list_lock);
> +		spin_lock(state.lock);
>  	}
> -	spin_unlock(&sb->s_inode_list_lock);
>  	iput(toput_inode);
>  }
>  
> diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
> index 6915c95..05b3f85 100644
> --- a/fs/fs-writeback.c
> +++ b/fs/fs-writeback.c
> @@ -2107,6 +2107,7 @@ EXPORT_SYMBOL(__mark_inode_dirty);
>  static void wait_sb_inodes(struct super_block *sb)
>  {
>  	struct inode *inode, *old_inode = NULL;
> +	DEFINE_PCPU_LIST_STATE(state);
>  
>  	/*
>  	 * We need to be protected against the filesystem going from
> @@ -2115,7 +2116,6 @@ static void wait_sb_inodes(struct super_block *sb)
>  	WARN_ON(!rwsem_is_locked(&sb->s_umount));
>  
>  	mutex_lock(&sb->s_sync_lock);
> -	spin_lock(&sb->s_inode_list_lock);
>  
>  	/*
>  	 * Data integrity sync. Must wait for all pages under writeback,
> @@ -2124,9 +2124,11 @@ static void wait_sb_inodes(struct super_block *sb)
>  	 * In which case, the inode may not be on the dirty list, but
>  	 * we still have to wait for that writeout.
>  	 */
> -	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
> -		struct address_space *mapping = inode->i_mapping;
> +	while (pcpu_list_iterate(sb->s_inodes, &state)) {
> +		struct address_space *mapping;
>  
> +		inode   = list_entry(state.curr, struct inode, i_sb_list);
> +		mapping = inode->i_mapping;
>  		spin_lock(&inode->i_lock);
>  		if ((inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) ||
>  		    (mapping->nrpages == 0)) {
> @@ -2135,7 +2137,7 @@ static void wait_sb_inodes(struct super_block *sb)
>  		}
>  		__iget(inode);
>  		spin_unlock(&inode->i_lock);
> -		spin_unlock(&sb->s_inode_list_lock);
> +		spin_unlock(state.lock);
>  
>  		/*
>  		 * We hold a reference to 'inode' so it couldn't have been
> @@ -2157,9 +2159,8 @@ static void wait_sb_inodes(struct super_block *sb)
>  
>  		cond_resched();
>  
> -		spin_lock(&sb->s_inode_list_lock);
> +		spin_lock(state.lock);
>  	}
> -	spin_unlock(&sb->s_inode_list_lock);
>  	iput(old_inode);
>  	mutex_unlock(&sb->s_sync_lock);
>  }
> diff --git a/fs/inode.c b/fs/inode.c
> index 9f62db3..e6e41ef 100644
> --- a/fs/inode.c
> +++ b/fs/inode.c
> @@ -28,7 +28,7 @@
>   *   inode->i_state, inode->i_hash, __iget()
>   * Inode LRU list locks protect:
>   *   inode->i_sb->s_inode_lru, inode->i_lru
> - * inode->i_sb->s_inode_list_lock protects:
> + * inode->i_sb->s_inodes->lock protects:
>   *   inode->i_sb->s_inodes, inode->i_sb_list
>   * bdi->wb.list_lock protects:
>   *   bdi->wb.b_{dirty,io,more_io,dirty_time}, inode->i_io_list
> @@ -37,7 +37,7 @@
>   *
>   * Lock ordering:
>   *
> - * inode->i_sb->s_inode_list_lock
> + * inode->i_sb->s_inodes->lock
>   *   inode->i_lock
>   *     Inode LRU list locks
>   *
> @@ -45,7 +45,7 @@
>   *   inode->i_lock
>   *
>   * inode_hash_lock
> - *   inode->i_sb->s_inode_list_lock
> + *   inode->i_sb->s_inodes->lock
>   *   inode->i_lock
>   *
>   * iunique_lock
> @@ -424,19 +424,14 @@ static void inode_lru_list_del(struct inode *inode)
>   */
>  void inode_sb_list_add(struct inode *inode)
>  {
> -	spin_lock(&inode->i_sb->s_inode_list_lock);
> -	list_add(&inode->i_sb_list, &inode->i_sb->s_inodes);
> -	spin_unlock(&inode->i_sb->s_inode_list_lock);
> +	pcpu_list_add(&inode->i_sb_list, inode->i_sb->s_inodes);
>  }
>  EXPORT_SYMBOL_GPL(inode_sb_list_add);
>  
>  static inline void inode_sb_list_del(struct inode *inode)
>  {
> -	if (!list_empty(&inode->i_sb_list)) {
> -		spin_lock(&inode->i_sb->s_inode_list_lock);
> -		list_del_init(&inode->i_sb_list);
> -		spin_unlock(&inode->i_sb->s_inode_list_lock);
> -	}
> +	if (!list_empty(&inode->i_sb_list.list))
> +		pcpu_list_del(&inode->i_sb_list);
>  }
>  
>  static unsigned long hash(struct super_block *sb, unsigned long hashval)
> @@ -590,12 +585,14 @@ static void dispose_list(struct list_head *head)
>   */
>  void evict_inodes(struct super_block *sb)
>  {
> -	struct inode *inode, *next;
> +	struct inode *inode;
> +	struct pcpu_list_state state;
>  	LIST_HEAD(dispose);
>  
>  again:
> -	spin_lock(&sb->s_inode_list_lock);
> -	list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
> +	init_pcpu_list_state(&state);
> +	while (pcpu_list_iterate_safe(sb->s_inodes, &state)) {
> +		inode = list_entry(state.curr, struct inode, i_sb_list);
>  		if (atomic_read(&inode->i_count))
>  			continue;
>  
> @@ -616,13 +613,12 @@ again:
>  		 * bit so we don't livelock.
>  		 */
>  		if (need_resched()) {
> -			spin_unlock(&sb->s_inode_list_lock);
> +			spin_unlock(state.lock);
>  			cond_resched();
>  			dispose_list(&dispose);
>  			goto again;
>  		}
>  	}
> -	spin_unlock(&sb->s_inode_list_lock);
>  
>  	dispose_list(&dispose);
>  }
> @@ -640,11 +636,12 @@ again:
>  int invalidate_inodes(struct super_block *sb, bool kill_dirty)
>  {
>  	int busy = 0;
> -	struct inode *inode, *next;
> +	struct inode *inode;
>  	LIST_HEAD(dispose);
> +	DEFINE_PCPU_LIST_STATE(state);
>  
> -	spin_lock(&sb->s_inode_list_lock);
> -	list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
> +	while (pcpu_list_iterate_safe(sb->s_inodes, &state)) {
> +		inode = list_entry(state.curr, struct inode, i_sb_list);
>  		spin_lock(&inode->i_lock);
>  		if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
>  			spin_unlock(&inode->i_lock);
> @@ -666,7 +663,6 @@ int invalidate_inodes(struct super_block *sb, bool kill_dirty)
>  		spin_unlock(&inode->i_lock);
>  		list_add(&inode->i_lru, &dispose);
>  	}
> -	spin_unlock(&sb->s_inode_list_lock);
>  
>  	dispose_list(&dispose);
>  
> @@ -881,7 +877,7 @@ struct inode *new_inode_pseudo(struct super_block *sb)
>  		spin_lock(&inode->i_lock);
>  		inode->i_state = 0;
>  		spin_unlock(&inode->i_lock);
> -		INIT_LIST_HEAD(&inode->i_sb_list);
> +		init_pcpu_list_node(&inode->i_sb_list);
>  	}
>  	return inode;
>  }
> @@ -902,8 +898,6 @@ struct inode *new_inode(struct super_block *sb)
>  {
>  	struct inode *inode;
>  
> -	spin_lock_prefetch(&sb->s_inode_list_lock);
> -
>  	inode = new_inode_pseudo(sb);
>  	if (inode)
>  		inode_sb_list_add(inode);
> diff --git a/fs/notify/inode_mark.c b/fs/notify/inode_mark.c
> index a364524..12515a4 100644
> --- a/fs/notify/inode_mark.c
> +++ b/fs/notify/inode_mark.c
> @@ -151,14 +151,15 @@ int fsnotify_add_inode_mark(struct fsnotify_mark *mark,
>  void fsnotify_unmount_inodes(struct super_block *sb)
>  {
>  	struct inode *inode, *iput_inode = NULL;
> +	DEFINE_PCPU_LIST_STATE(state);
>  
> -	spin_lock(&sb->s_inode_list_lock);
> -	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
> +	while (pcpu_list_iterate(sb->s_inodes, &state)) {
>  		/*
>  		 * We cannot __iget() an inode in state I_FREEING,
>  		 * I_WILL_FREE, or I_NEW which is fine because by that point
>  		 * the inode cannot have any associated watches.
>  		 */
> +		inode = list_entry(state.curr, struct inode, i_sb_list);
>  		spin_lock(&inode->i_lock);
>  		if (inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) {
>  			spin_unlock(&inode->i_lock);
> @@ -178,7 +179,7 @@ void fsnotify_unmount_inodes(struct super_block *sb)
>  
>  		__iget(inode);
>  		spin_unlock(&inode->i_lock);
> -		spin_unlock(&sb->s_inode_list_lock);
> +		spin_unlock(state.lock);
>  
>  		if (iput_inode)
>  			iput(iput_inode);
> @@ -190,9 +191,8 @@ void fsnotify_unmount_inodes(struct super_block *sb)
>  
>  		iput_inode = inode;
>  
> -		spin_lock(&sb->s_inode_list_lock);
> +		spin_lock(state.lock);
>  	}
> -	spin_unlock(&sb->s_inode_list_lock);
>  
>  	if (iput_inode)
>  		iput(iput_inode);
> diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c
> index 3c3b81b..2564271 100644
> --- a/fs/quota/dquot.c
> +++ b/fs/quota/dquot.c
> @@ -924,12 +924,13 @@ static int dqinit_needed(struct inode *inode, int type)
>  static void add_dquot_ref(struct super_block *sb, int type)
>  {
>  	struct inode *inode, *old_inode = NULL;
> +	DEFINE_PCPU_LIST_STATE(state);
>  #ifdef CONFIG_QUOTA_DEBUG
>  	int reserved = 0;
>  #endif
>  
> -	spin_lock(&sb->s_inode_list_lock);
> -	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
> +	while (pcpu_list_iterate(sb->s_inodes, &state)) {
> +		inode = list_entry(state.curr, struct inode, i_sb_list);
>  		spin_lock(&inode->i_lock);
>  		if ((inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) ||
>  		    !atomic_read(&inode->i_writecount) ||
> @@ -939,7 +940,7 @@ static void add_dquot_ref(struct super_block *sb, int type)
>  		}
>  		__iget(inode);
>  		spin_unlock(&inode->i_lock);
> -		spin_unlock(&sb->s_inode_list_lock);
> +		spin_unlock(state.lock);
>  
>  #ifdef CONFIG_QUOTA_DEBUG
>  		if (unlikely(inode_get_rsv_space(inode) > 0))
> @@ -957,9 +958,8 @@ static void add_dquot_ref(struct super_block *sb, int type)
>  		 * later.
>  		 */
>  		old_inode = inode;
> -		spin_lock(&sb->s_inode_list_lock);
> +		spin_lock(state.lock);
>  	}
> -	spin_unlock(&sb->s_inode_list_lock);
>  	iput(old_inode);
>  
>  #ifdef CONFIG_QUOTA_DEBUG
> @@ -1027,15 +1027,16 @@ static void remove_dquot_ref(struct super_block *sb, int type,
>  {
>  	struct inode *inode;
>  	int reserved = 0;
> +	DEFINE_PCPU_LIST_STATE(state);
>  
> -	spin_lock(&sb->s_inode_list_lock);
> -	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
> +	while (pcpu_list_iterate(sb->s_inodes, &state)) {
>  		/*
>  		 *  We have to scan also I_NEW inodes because they can already
>  		 *  have quota pointer initialized. Luckily, we need to touch
>  		 *  only quota pointers and these have separate locking
>  		 *  (dq_data_lock).
>  		 */
> +		inode = list_entry(state.curr, struct inode, i_sb_list);
>  		spin_lock(&dq_data_lock);
>  		if (!IS_NOQUOTA(inode)) {
>  			if (unlikely(inode_get_rsv_space(inode) > 0))
> @@ -1044,7 +1045,6 @@ static void remove_dquot_ref(struct super_block *sb, int type,
>  		}
>  		spin_unlock(&dq_data_lock);
>  	}
> -	spin_unlock(&sb->s_inode_list_lock);
>  #ifdef CONFIG_QUOTA_DEBUG
>  	if (reserved) {
>  		printk(KERN_WARNING "VFS (%s): Writes happened after quota"
> diff --git a/fs/super.c b/fs/super.c
> index 1182af8..7d44fad 100644
> --- a/fs/super.c
> +++ b/fs/super.c
> @@ -163,6 +163,7 @@ static void destroy_super(struct super_block *s)
>  {
>  	list_lru_destroy(&s->s_dentry_lru);
>  	list_lru_destroy(&s->s_inode_lru);
> +	free_pcpu_list_head(&s->s_inodes);
>  	security_sb_free(s);
>  	WARN_ON(!list_empty(&s->s_mounts));
>  	kfree(s->s_subtype);
> @@ -204,9 +205,9 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags)
>  	INIT_HLIST_NODE(&s->s_instances);
>  	INIT_HLIST_BL_HEAD(&s->s_anon);
>  	mutex_init(&s->s_sync_lock);
> -	INIT_LIST_HEAD(&s->s_inodes);
> -	spin_lock_init(&s->s_inode_list_lock);
>  
> +	if (init_pcpu_list_head(&s->s_inodes))
> +		goto fail;
>  	if (list_lru_init_memcg(&s->s_dentry_lru))
>  		goto fail;
>  	if (list_lru_init_memcg(&s->s_inode_lru))
> @@ -426,7 +427,7 @@ void generic_shutdown_super(struct super_block *sb)
>  		if (sop->put_super)
>  			sop->put_super(sb);
>  
> -		if (!list_empty(&sb->s_inodes)) {
> +		if (!pcpu_list_empty(sb->s_inodes)) {
>  			printk("VFS: Busy inodes after unmount of %s. "
>  			   "Self-destruct in 5 seconds.  Have a nice day...\n",
>  			   sb->s_id);
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index ae68100..b533bda 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -28,6 +28,7 @@
>  #include <linux/uidgid.h>
>  #include <linux/lockdep.h>
>  #include <linux/percpu-rwsem.h>
> +#include <linux/percpu-list.h>
>  #include <linux/blk_types.h>
>  #include <linux/workqueue.h>
>  #include <linux/percpu-rwsem.h>
> @@ -648,7 +649,7 @@ struct inode {
>  	u16			i_wb_frn_history;
>  #endif
>  	struct list_head	i_lru;		/* inode LRU list */
> -	struct list_head	i_sb_list;
> +	struct pcpu_list_node	i_sb_list;
>  	union {
>  		struct hlist_head	i_dentry;
>  		struct rcu_head		i_rcu;
> @@ -1397,9 +1398,8 @@ struct super_block {
>  	 */
>  	int s_stack_depth;
>  
> -	/* s_inode_list_lock protects s_inodes */
> -	spinlock_t		s_inode_list_lock ____cacheline_aligned_in_smp;
> -	struct list_head	s_inodes;	/* all inodes */
> +	/* The percpu locks protect s_inodes */
> +	struct pcpu_list_head __percpu *s_inodes;	/* all inodes */
>  };
>  
>  extern struct timespec current_fs_time(struct super_block *sb);
> -- 
> 1.7.1
> 
>

Comments

Ingo Molnar Feb. 24, 2016, 8:36 a.m. UTC | #1
* Jan Kara <jack@suse.cz> wrote:

> On Tue 23-02-16 14:04:32, Waiman Long wrote:
> > When many threads are trying to add or delete inode to or from
> > a superblock's s_inodes list, spinlock contention on the list can
> > become a performance bottleneck.
> > 
> > This patch changes the s_inodes field to become a per-cpu list with
> > per-cpu spinlocks. As a result, the following superblock inode list
> > (sb->s_inodes) iteration functions in vfs are also being modified:
> > 
> >  1. iterate_bdevs()
> >  2. drop_pagecache_sb()
> >  3. wait_sb_inodes()
> >  4. evict_inodes()
> >  5. invalidate_inodes()
> >  6. fsnotify_unmount_inodes()
> >  7. add_dquot_ref()
> >  8. remove_dquot_ref()
> > 
> > With an exit microbenchmark that creates a large number of threads,
> > attachs many inodes to them and then exits. The runtimes of that
> > microbenchmark with 1000 threads before and after the patch on a
> > 4-socket Intel E7-4820 v3 system (40 cores, 80 threads) were as
> > follows:
> > 
> >   Kernel            Elapsed Time    System Time
> >   ------            ------------    -----------
> >   Vanilla 4.5-rc4      65.29s         82m14s
> >   Patched 4.5-rc4      22.81s         23m03s
> > 
> > Before the patch, spinlock contention at the inode_sb_list_add()
> > function at the startup phase and the inode_sb_list_del() function at
> > the exit phase were about 79% and 93% of total CPU time respectively
> > (as measured by perf). After the patch, the percpu_list_add()
> > function consumed only about 0.04% of CPU time at startup phase. The
> > percpu_list_del() function consumed about 0.4% of CPU time at exit
> > phase. There were still some spinlock contention, but they happened
> > elsewhere.
> 
> While looking through this patch, I have noticed that the
> list_for_each_entry_safe() iterations in evict_inodes() and
> invalidate_inodes() are actually unnecessary. So if you first apply the
> attached patch, you don't have to implement safe iteration variants at all.
> 
> As a second comment, I'd note that this patch grows struct inode by 1 pointer. 
> It is probably acceptable for large machines given the speedup but it should be 
> noted in the changelog. Furthermore for UP or even small SMP systems this is 
> IMHO undesired bloat since the speedup won't be noticeable.
> 
> So for these small systems it would be good if per-cpu list magic would just 
> fall back to single linked list with a spinlock. Do you think that is reasonably 
> doable?

Even many 'small' systems tend to be SMP these days.

If you do this then please keep it a separate add-on patch, so that the 'UP cost' 
becomes apparent. Uniprocessor #ifdeffery is really painful in places and we might 
be better off with a single extra pointer. Forthermore UP kernels are tested a lot 
less stringently than SMP kernels. It's just 4 bytes for a truly small 32-bit 
system.

Thanks,

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Jan Kara Feb. 24, 2016, 8:58 a.m. UTC | #2
On Wed 24-02-16 09:36:30, Ingo Molnar wrote:
> 
> * Jan Kara <jack@suse.cz> wrote:
> 
> > On Tue 23-02-16 14:04:32, Waiman Long wrote:
> > > When many threads are trying to add or delete inode to or from
> > > a superblock's s_inodes list, spinlock contention on the list can
> > > become a performance bottleneck.
> > > 
> > > This patch changes the s_inodes field to become a per-cpu list with
> > > per-cpu spinlocks. As a result, the following superblock inode list
> > > (sb->s_inodes) iteration functions in vfs are also being modified:
> > > 
> > >  1. iterate_bdevs()
> > >  2. drop_pagecache_sb()
> > >  3. wait_sb_inodes()
> > >  4. evict_inodes()
> > >  5. invalidate_inodes()
> > >  6. fsnotify_unmount_inodes()
> > >  7. add_dquot_ref()
> > >  8. remove_dquot_ref()
> > > 
> > > With an exit microbenchmark that creates a large number of threads,
> > > attachs many inodes to them and then exits. The runtimes of that
> > > microbenchmark with 1000 threads before and after the patch on a
> > > 4-socket Intel E7-4820 v3 system (40 cores, 80 threads) were as
> > > follows:
> > > 
> > >   Kernel            Elapsed Time    System Time
> > >   ------            ------------    -----------
> > >   Vanilla 4.5-rc4      65.29s         82m14s
> > >   Patched 4.5-rc4      22.81s         23m03s
> > > 
> > > Before the patch, spinlock contention at the inode_sb_list_add()
> > > function at the startup phase and the inode_sb_list_del() function at
> > > the exit phase were about 79% and 93% of total CPU time respectively
> > > (as measured by perf). After the patch, the percpu_list_add()
> > > function consumed only about 0.04% of CPU time at startup phase. The
> > > percpu_list_del() function consumed about 0.4% of CPU time at exit
> > > phase. There were still some spinlock contention, but they happened
> > > elsewhere.
> > 
> > While looking through this patch, I have noticed that the
> > list_for_each_entry_safe() iterations in evict_inodes() and
> > invalidate_inodes() are actually unnecessary. So if you first apply the
> > attached patch, you don't have to implement safe iteration variants at all.
> > 
> > As a second comment, I'd note that this patch grows struct inode by 1 pointer. 
> > It is probably acceptable for large machines given the speedup but it should be 
> > noted in the changelog. Furthermore for UP or even small SMP systems this is 
> > IMHO undesired bloat since the speedup won't be noticeable.
> > 
> > So for these small systems it would be good if per-cpu list magic would just 
> > fall back to single linked list with a spinlock. Do you think that is reasonably 
> > doable?
> 
> Even many 'small' systems tend to be SMP these days.

Yes, I know. But my tablet with 4 ARM cores is unlikely to benefit from
this change either. It will just have to pay the memory cost. And frankly I
don't care that much myself but if there is some reasonably easy way to
avoid the cost, it would be welcome.

> If you do this then please keep it a separate add-on patch, so that the
> 'UP cost' becomes apparent. Uniprocessor #ifdeffery is really painful in
> places and we might be better off with a single extra pointer.
> Forthermore UP kernels are tested a lot less stringently than SMP
> kernels. It's just 4 bytes for a truly small 32-bit system.

								Honza
Waiman Long Feb. 24, 2016, 8:23 p.m. UTC | #3
On 02/24/2016 03:28 AM, Jan Kara wrote:
> On Tue 23-02-16 14:04:32, Waiman Long wrote:
>> When many threads are trying to add or delete inode to or from
>> a superblock's s_inodes list, spinlock contention on the list can
>> become a performance bottleneck.
>>
>> This patch changes the s_inodes field to become a per-cpu list with
>> per-cpu spinlocks. As a result, the following superblock inode list
>> (sb->s_inodes) iteration functions in vfs are also being modified:
>>
>>   1. iterate_bdevs()
>>   2. drop_pagecache_sb()
>>   3. wait_sb_inodes()
>>   4. evict_inodes()
>>   5. invalidate_inodes()
>>   6. fsnotify_unmount_inodes()
>>   7. add_dquot_ref()
>>   8. remove_dquot_ref()
>>
>> With an exit microbenchmark that creates a large number of threads,
>> attachs many inodes to them and then exits. The runtimes of that
>> microbenchmark with 1000 threads before and after the patch on a
>> 4-socket Intel E7-4820 v3 system (40 cores, 80 threads) were as
>> follows:
>>
>>    Kernel            Elapsed Time    System Time
>>    ------            ------------    -----------
>>    Vanilla 4.5-rc4      65.29s         82m14s
>>    Patched 4.5-rc4      22.81s         23m03s
>>
>> Before the patch, spinlock contention at the inode_sb_list_add()
>> function at the startup phase and the inode_sb_list_del() function at
>> the exit phase were about 79% and 93% of total CPU time respectively
>> (as measured by perf). After the patch, the percpu_list_add()
>> function consumed only about 0.04% of CPU time at startup phase. The
>> percpu_list_del() function consumed about 0.4% of CPU time at exit
>> phase. There were still some spinlock contention, but they happened
>> elsewhere.
> While looking through this patch, I have noticed that the
> list_for_each_entry_safe() iterations in evict_inodes() and
> invalidate_inodes() are actually unnecessary. So if you first apply the
> attached patch, you don't have to implement safe iteration variants at all.

Thank for the patch. I will apply that in my next update. As for the 
safe iteration variant, I think I will keep it since I had implemented 
that already just in case it may be needed in some other places.

> As a second comment, I'd note that this patch grows struct inode by 1
> pointer. It is probably acceptable for large machines given the speedup but
> it should be noted in the changelog. Furthermore for UP or even small SMP
> systems this is IMHO undesired bloat since the speedup won't be noticeable.
>
> So for these small systems it would be good if per-cpu list magic would just
> fall back to single linked list with a spinlock. Do you think that is
> reasonably doable?
>

I already have a somewhat separate code path for UP. So I can remove the 
lock pointer for that. For small SMP system, however, the only way to 
avoid the extra pointer is to add a config parameter to turn this 
feature off. That can be added as a separate patch, if necessary.

BTW, I think the current inode structure is already pretty big, adding 
one more pointer will have too much impact on its overall size.

Cheers,
Longman


--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Ingo Molnar Feb. 25, 2016, 8:06 a.m. UTC | #4
* Jan Kara <jack@suse.cz> wrote:

> > > > With an exit microbenchmark that creates a large number of threads, 
> > > > attachs many inodes to them and then exits. The runtimes of that 
> > > > microbenchmark with 1000 threads before and after the patch on a 4-socket 
> > > > Intel E7-4820 v3 system (40 cores, 80 threads) were as follows:
> > > > 
> > > >   Kernel            Elapsed Time    System Time
> > > >   ------            ------------    -----------
> > > >   Vanilla 4.5-rc4      65.29s         82m14s
> > > >   Patched 4.5-rc4      22.81s         23m03s
> > > > 
> > > > Before the patch, spinlock contention at the inode_sb_list_add() function 
> > > > at the startup phase and the inode_sb_list_del() function at the exit 
> > > > phase were about 79% and 93% of total CPU time respectively (as measured 
> > > > by perf). After the patch, the percpu_list_add() function consumed only 
> > > > about 0.04% of CPU time at startup phase. The percpu_list_del() function 
> > > > consumed about 0.4% of CPU time at exit phase. There were still some 
> > > > spinlock contention, but they happened elsewhere.
> > > 
> > > While looking through this patch, I have noticed that the 
> > > list_for_each_entry_safe() iterations in evict_inodes() and 
> > > invalidate_inodes() are actually unnecessary. So if you first apply the 
> > > attached patch, you don't have to implement safe iteration variants at all.
> > > 
> > > As a second comment, I'd note that this patch grows struct inode by 1 
> > > pointer. It is probably acceptable for large machines given the speedup but 
> > > it should be noted in the changelog. Furthermore for UP or even small SMP 
> > > systems this is IMHO undesired bloat since the speedup won't be noticeable.
> > > 
> > > So for these small systems it would be good if per-cpu list magic would just 
> > > fall back to single linked list with a spinlock. Do you think that is 
> > > reasonably doable?
> > 
> > Even many 'small' systems tend to be SMP these days.
> 
> Yes, I know. But my tablet with 4 ARM cores is unlikely to benefit from this 
> change either. [...]

I'm not sure about that at all, the above numbers are showing a 3x-4x speedup in 
system time, which ought to be noticeable on smaller SMP systems as well.

Waiman, could you please post the microbenchmark?

Thanks,

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Waiman Long Feb. 25, 2016, 2:43 p.m. UTC | #5
On 02/25/2016 03:06 AM, Ingo Molnar wrote:
> * Jan Kara<jack@suse.cz>  wrote:
>
>>>>> With an exit microbenchmark that creates a large number of threads,
>>>>> attachs many inodes to them and then exits. The runtimes of that
>>>>> microbenchmark with 1000 threads before and after the patch on a 4-socket
>>>>> Intel E7-4820 v3 system (40 cores, 80 threads) were as follows:
>>>>>
>>>>>    Kernel            Elapsed Time    System Time
>>>>>    ------            ------------    -----------
>>>>>    Vanilla 4.5-rc4      65.29s         82m14s
>>>>>    Patched 4.5-rc4      22.81s         23m03s
>>>>>
>>>>> Before the patch, spinlock contention at the inode_sb_list_add() function
>>>>> at the startup phase and the inode_sb_list_del() function at the exit
>>>>> phase were about 79% and 93% of total CPU time respectively (as measured
>>>>> by perf). After the patch, the percpu_list_add() function consumed only
>>>>> about 0.04% of CPU time at startup phase. The percpu_list_del() function
>>>>> consumed about 0.4% of CPU time at exit phase. There were still some
>>>>> spinlock contention, but they happened elsewhere.
>>>> While looking through this patch, I have noticed that the
>>>> list_for_each_entry_safe() iterations in evict_inodes() and
>>>> invalidate_inodes() are actually unnecessary. So if you first apply the
>>>> attached patch, you don't have to implement safe iteration variants at all.
>>>>
>>>> As a second comment, I'd note that this patch grows struct inode by 1
>>>> pointer. It is probably acceptable for large machines given the speedup but
>>>> it should be noted in the changelog. Furthermore for UP or even small SMP
>>>> systems this is IMHO undesired bloat since the speedup won't be noticeable.
>>>>
>>>> So for these small systems it would be good if per-cpu list magic would just
>>>> fall back to single linked list with a spinlock. Do you think that is
>>>> reasonably doable?
>>> Even many 'small' systems tend to be SMP these days.
>> Yes, I know. But my tablet with 4 ARM cores is unlikely to benefit from this
>> change either. [...]
> I'm not sure about that at all, the above numbers are showing a 3x-4x speedup in
> system time, which ought to be noticeable on smaller SMP systems as well.
>
> Waiman, could you please post the microbenchmark?
>
> Thanks,
>
> 	Ingo

The microbenchmark that I used is attached.

I do agree that performance benefit will decrease as the number of CPUs 
get smaller. The system that I used for testing have 4 sockets with 40 
cores (80 threads). Dave Chinner had run his fstests on a 16-core system 
(probably 2-socket) which showed modest improvement in performance 
(~4m40s vs 4m30s in runtime).

This patch enables parallel insertion and deletion to/from the inode 
list which used to be a serialized operation. So if that list operation 
is a bottleneck, you will see significant improvement. If it is not, we 
may not notice that much of a difference. For a single-socket 4-core 
system, I agree that the performance benefit, if any, will be limited.

Cheers,
Longman
/*
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * Authors: Waiman Long <waiman.long@hp.com>
 */
/*
 * This is an exit test
 */
#include <ctype.h>
#include <errno.h>
#include <pthread.h>
#include <stdlib.h>
#include <stdio.h>
#include <time.h>
#include <unistd.h>
#include <signal.h>
#include <sys/types.h>
#include <sys/syscall.h>


#define do_exit()	syscall(SYS_exit)
#define	gettid()	syscall(SYS_gettid)
#define	MAX_THREADS	2048

static inline void cpu_relax(void)
{
        __asm__ __volatile__("rep;nop": : :"memory");
}

static inline void atomic_inc(volatile int *v)
{
	__asm__ __volatile__("lock incl %0": "+m" (*v));
}

static volatile int exit_now  = 0;
static volatile int threadcnt = 0;

/*
 * Walk the /proc/<pid> filesystem to make them fill the dentry cache
 */
static void walk_procfs(void)
{
	char cmdbuf[256];
	pid_t tid = gettid();

	snprintf(cmdbuf, sizeof(cmdbuf), "find /proc/%d > /dev/null 2>&1", tid);
	if (system(cmdbuf) < 0)
		perror("system() failed!");
}

static void *exit_thread(void *dummy)
{
	long tid = (long)dummy;

	walk_procfs();
	atomic_inc(&threadcnt);
	/*
	 * Busy wait until the do_exit flag is set and then call exit
	 */
	while (!exit_now)
		sleep(1);
	do_exit();
}

static void exit_test(int threads)
{
	pthread_t thread[threads];
	long i = 0, finish;
	time_t start = time(NULL);

	while (i++ < threads) {
		if (pthread_create(thread + i - 1, NULL, exit_thread,
				  (void *)i)) {
			perror("pthread_create");
			exit(1);
		}
#if 0
		/*
		 * Pipelining to reduce contention & improve speed
		 */
		if ((i & 0xf) == 0)
			 while (i - threadcnt > 12)
				usleep(1);
#endif
	}
	while (threadcnt != threads)
		usleep(1);
	walk_procfs();
	printf("Setup time = %lus\n", time(NULL) - start);
	printf("Process ready to exit!\n");
	kill(0, SIGKILL);
	exit(0);
}

int main(int argc, char *argv[])
{
	int   tcnt;	/* Thread counts */
	char *cmd = argv[0];

	if ((argc != 2) || !isdigit(argv[1][0])) {
		fprintf(stderr, "Usage: %s <thread count>\n", cmd);
		exit(1);
	}
	tcnt = strtoul(argv[1], NULL, 10);
	if (tcnt > MAX_THREADS) {
		fprintf(stderr, "Error: thread count should be <= %d\n",
			MAX_THREADS);
		exit(1);
	}
	exit_test(tcnt);
	return 0;	/* Not reaachable */
}
Waiman Long Feb. 25, 2016, 2:50 p.m. UTC | #6
On 02/24/2016 03:23 PM, Waiman Long wrote:
> On 02/24/2016 03:28 AM, Jan Kara wrote:
>> On Tue 23-02-16 14:04:32, Waiman Long wrote:
>>> When many threads are trying to add or delete inode to or from
>>> a superblock's s_inodes list, spinlock contention on the list can
>>> become a performance bottleneck.
>>>
>>> This patch changes the s_inodes field to become a per-cpu list with
>>> per-cpu spinlocks. As a result, the following superblock inode list
>>> (sb->s_inodes) iteration functions in vfs are also being modified:
>>>
>>>   1. iterate_bdevs()
>>>   2. drop_pagecache_sb()
>>>   3. wait_sb_inodes()
>>>   4. evict_inodes()
>>>   5. invalidate_inodes()
>>>   6. fsnotify_unmount_inodes()
>>>   7. add_dquot_ref()
>>>   8. remove_dquot_ref()
>>>
>>> With an exit microbenchmark that creates a large number of threads,
>>> attachs many inodes to them and then exits. The runtimes of that
>>> microbenchmark with 1000 threads before and after the patch on a
>>> 4-socket Intel E7-4820 v3 system (40 cores, 80 threads) were as
>>> follows:
>>>
>>>    Kernel            Elapsed Time    System Time
>>>    ------            ------------    -----------
>>>    Vanilla 4.5-rc4      65.29s         82m14s
>>>    Patched 4.5-rc4      22.81s         23m03s
>>>
>>> Before the patch, spinlock contention at the inode_sb_list_add()
>>> function at the startup phase and the inode_sb_list_del() function at
>>> the exit phase were about 79% and 93% of total CPU time respectively
>>> (as measured by perf). After the patch, the percpu_list_add()
>>> function consumed only about 0.04% of CPU time at startup phase. The
>>> percpu_list_del() function consumed about 0.4% of CPU time at exit
>>> phase. There were still some spinlock contention, but they happened
>>> elsewhere.
>> While looking through this patch, I have noticed that the
>> list_for_each_entry_safe() iterations in evict_inodes() and
>> invalidate_inodes() are actually unnecessary. So if you first apply the
>> attached patch, you don't have to implement safe iteration variants 
>> at all.
>
> Thank for the patch. I will apply that in my next update. As for the 
> safe iteration variant, I think I will keep it since I had implemented 
> that already just in case it may be needed in some other places.
>
>> As a second comment, I'd note that this patch grows struct inode by 1
>> pointer. It is probably acceptable for large machines given the 
>> speedup but
>> it should be noted in the changelog. Furthermore for UP or even small 
>> SMP
>> systems this is IMHO undesired bloat since the speedup won't be 
>> noticeable.
>>
>> So for these small systems it would be good if per-cpu list magic 
>> would just
>> fall back to single linked list with a spinlock. Do you think that is
>> reasonably doable?
>>
>
> I already have a somewhat separate code path for UP. So I can remove 
> the lock pointer for that. For small SMP system, however, the only way 
> to avoid the extra pointer is to add a config parameter to turn this 
> feature off. That can be added as a separate patch, if necessary.

I am sorry that I need to retreat from this promise for UP. Removing the 
lock pointer will require change in the list deletion API to pass in the 
lock information. So I am not going to change it for the time being.

Cheers,
Longman

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch
diff mbox

From ede070a2159d4c49c6a29be601594f7119872417 Mon Sep 17 00:00:00 2001
From: Jan Kara <jack@suse.cz>
Date: Wed, 24 Feb 2016 09:05:32 +0100
Subject: [PATCH] vfs: Remove unnecessary list_for_each_entry_safe() variants

evict_inodes() and invalidate_inodes() use list_for_each_entry_safe()
to iterate sb->s_inodes list. However, since we use i_lru list entry for
our local temporary list of inodes to destroy, the inode is guaranteed
to stay in sb->s_inodes list while we hold sb->s_inode_list_lock. So
there is no real need for safe iteration variant and we can use
list_for_each_entry() just fine.

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/inode.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/inode.c b/fs/inode.c
index 9f62db3bcc3e..efb1bea53695 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -590,12 +590,12 @@  static void dispose_list(struct list_head *head)
  */
 void evict_inodes(struct super_block *sb)
 {
-	struct inode *inode, *next;
+	struct inode *inode;
 	LIST_HEAD(dispose);
 
 again:
 	spin_lock(&sb->s_inode_list_lock);
-	list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
+	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
 		if (atomic_read(&inode->i_count))
 			continue;
 
@@ -640,11 +640,11 @@  again:
 int invalidate_inodes(struct super_block *sb, bool kill_dirty)
 {
 	int busy = 0;
-	struct inode *inode, *next;
+	struct inode *inode;
 	LIST_HEAD(dispose);
 
 	spin_lock(&sb->s_inode_list_lock);
-	list_for_each_entry_safe(inode, next, &sb->s_inodes, i_sb_list) {
+	list_for_each_entry(inode, &sb->s_inodes, i_sb_list) {
 		spin_lock(&inode->i_lock);
 		if (inode->i_state & (I_NEW | I_FREEING | I_WILL_FREE)) {
 			spin_unlock(&inode->i_lock);
-- 
2.6.2