[PATCHSET percpu/for-3.18] add @gfp to init functions of percpu data structures

[PATCHSET percpu/for-3.18] add @gfp to init functions of percpu data structures

[Tejun Heo]

There's now a pending patchset[1] which implements atomic percpu
allocation.  This patchset propagates @gfp to percpu data structures
so that they can be allocated and initialized from !GFP_KERNEL
contexts too.  This will be used for opportunistic allocations of data
structures embedding percpu constructs in IO path.

This patchset adds @gfp to alloc/init functions of percpu_counter,
[flex_]proportions and percpu-refcount.  We could add separate
alloc/init functions which take @gfp but there aren't too many users
yet, so let's just add it to the existing ones.

This patchset contains the following patches

 0001-percpu_counter-add-gfp-to-percpu_counter_init.patch
 0002-proportions-add-gfp-to-init-functions.patch
 0003-percpu-refcount-add-gfp-to-percpu_ref_init.patch

and is on top of

 [1] [PATCHSET REPOST percpu/for-3.18] percpu: implement atomic allocation support

and avaliable in the following git branch.

 git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu.git/review-add-gfps

diffstat follows.  Thanks.

 arch/x86/kvm/mmu.c               |    2 +-
 drivers/target/target_core_tpg.c |    3 ++-
 fs/aio.c                         |    4 ++--
 fs/btrfs/disk-io.c               |    8 ++++----
 fs/btrfs/extent-tree.c           |    2 +-
 fs/ext2/super.c                  |    6 +++---
 fs/ext3/super.c                  |    6 +++---
 fs/ext4/super.c                  |   14 +++++++++-----
 fs/file_table.c                  |    2 +-
 fs/quota/dquot.c                 |    2 +-
 fs/super.c                       |    3 ++-
 include/linux/flex_proportions.h |    5 +++--
 include/linux/percpu-refcount.h  |    3 ++-
 include/linux/percpu_counter.h   |   10 ++++++----
 include/linux/proportions.h      |    5 +++--
 include/net/dst_ops.h            |    2 +-
 include/net/inet_frag.h          |    2 +-
 kernel/cgroup.c                  |    6 +++---
 lib/flex_proportions.c           |    8 ++++----
 lib/percpu-refcount.c            |    6 ++++--
 lib/percpu_counter.c             |    4 ++--
 lib/proportions.c                |   10 +++++-----
 mm/backing-dev.c                 |    4 ++--
 mm/mmap.c                        |    2 +-
 mm/nommu.c                       |    2 +-
 mm/page-writeback.c              |    2 +-
 mm/shmem.c                       |    2 +-
 net/dccp/proto.c                 |    2 +-
 net/ipv4/tcp.c                   |    4 ++--
 net/ipv4/tcp_memcontrol.c        |    2 +-
 net/sctp/protocol.c              |    2 +-
 31 files changed, 74 insertions(+), 61 deletions(-)

--
tejun

[1] http://lkml.kernel.org/g/1408726399-4436-1-git-send-email-tj@kernel.org

[PATCH 1/3] percpu_counter: add @gfp to percpu_counter_init()

[Tejun Heo]

Percpu allocator now supports allocation mask.  Add @gfp to
percpu_counter_init() so that !GFP_KERNEL allocation masks can be used
with percpu_counters too.

We could have left percpu_counter_init() alone and added
percpu_counter_init_gfp(); however, the number of users isn't that
high and introducing _gfp variants to all percpu data structures would
be quite ugly, so let's just do the conversion.  This is the one with
the most users.  Other percpu data structures are a lot easier to
convert.

This patch doesn't make any functional difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: x86@kernel.org
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jan Kara <jack@suse.cz>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 arch/x86/kvm/mmu.c             |  2 +-
 fs/btrfs/disk-io.c             |  8 ++++----
 fs/btrfs/extent-tree.c         |  2 +-
 fs/ext2/super.c                |  6 +++---
 fs/ext3/super.c                |  6 +++---
 fs/ext4/super.c                | 14 +++++++++-----
 fs/file_table.c                |  2 +-
 fs/quota/dquot.c               |  2 +-
 fs/super.c                     |  3 ++-
 include/linux/percpu_counter.h | 10 ++++++----
 include/net/dst_ops.h          |  2 +-
 include/net/inet_frag.h        |  2 +-
 lib/flex_proportions.c         |  4 ++--
 lib/percpu_counter.c           |  4 ++--
 lib/proportions.c              |  6 +++---
 mm/backing-dev.c               |  2 +-
 mm/mmap.c                      |  2 +-
 mm/nommu.c                     |  2 +-
 mm/shmem.c                     |  2 +-
 net/dccp/proto.c               |  2 +-
 net/ipv4/tcp.c                 |  4 ++--
 net/ipv4/tcp_memcontrol.c      |  2 +-
 net/sctp/protocol.c            |  2 +-
 23 files changed, 49 insertions(+), 42 deletions(-)

diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 9314678..5bd53f2 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -4534,7 +4534,7 @@ int kvm_mmu_module_init(void)
 	if (!mmu_page_header_cache)
 		goto nomem;
 
-	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0))
+	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0, GFP_KERNEL))
 		goto nomem;
 
 	register_shrinker(&mmu_shrinker);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 08e65e9..61dae01 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1180,7 +1180,7 @@ static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
 	if (!writers)
 		return ERR_PTR(-ENOMEM);
 
-	ret = percpu_counter_init(&writers->counter, 0);
+	ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
 	if (ret < 0) {
 		kfree(writers);
 		return ERR_PTR(ret);
@@ -2185,7 +2185,7 @@ int open_ctree(struct super_block *sb,
 		goto fail_srcu;
 	}
 
-	ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
+	ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
 	if (ret) {
 		err = ret;
 		goto fail_bdi;
@@ -2193,13 +2193,13 @@ int open_ctree(struct super_block *sb,
 	fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
 					(1 + ilog2(nr_cpu_ids));
 
-	ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
+	ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
 	if (ret) {
 		err = ret;
 		goto fail_dirty_metadata_bytes;
 	}
 
-	ret = percpu_counter_init(&fs_info->bio_counter, 0);
+	ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
 	if (ret) {
 		err = ret;
 		goto fail_delalloc_bytes;
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 813537f..94ec71e 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3493,7 +3493,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags,
 	if (!found)
 		return -ENOMEM;
 
-	ret = percpu_counter_init(&found->total_bytes_pinned, 0);
+	ret = percpu_counter_init(&found->total_bytes_pinned, 0, GFP_KERNEL);
 	if (ret) {
 		kfree(found);
 		return ret;
diff --git a/fs/ext2/super.c b/fs/ext2/super.c
index b88edc0..170dc41 100644
--- a/fs/ext2/super.c
+++ b/fs/ext2/super.c
@@ -1067,14 +1067,14 @@ static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
 
 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
-				ext2_count_free_blocks(sb));
+				ext2_count_free_blocks(sb), GFP_KERNEL);
 	if (!err) {
 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
-				ext2_count_free_inodes(sb));
+				ext2_count_free_inodes(sb), GFP_KERNEL);
 	}
 	if (!err) {
 		err = percpu_counter_init(&sbi->s_dirs_counter,
-				ext2_count_dirs(sb));
+				ext2_count_dirs(sb), GFP_KERNEL);
 	}
 	if (err) {
 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
diff --git a/fs/ext3/super.c b/fs/ext3/super.c
index 08cdfe5..eba021b 100644
--- a/fs/ext3/super.c
+++ b/fs/ext3/super.c
@@ -2039,14 +2039,14 @@ static int ext3_fill_super (struct super_block *sb, void *data, int silent)
 		goto failed_mount2;
 	}
 	err = percpu_counter_init(&sbi->s_freeblocks_counter,
-			ext3_count_free_blocks(sb));
+			ext3_count_free_blocks(sb), GFP_KERNEL);
 	if (!err) {
 		err = percpu_counter_init(&sbi->s_freeinodes_counter,
-				ext3_count_free_inodes(sb));
+				ext3_count_free_inodes(sb), GFP_KERNEL);
 	}
 	if (!err) {
 		err = percpu_counter_init(&sbi->s_dirs_counter,
-				ext3_count_dirs(sb));
+				ext3_count_dirs(sb), GFP_KERNEL);
 	}
 	if (err) {
 		ext3_msg(sb, KERN_ERR, "error: insufficient memory");
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 32b43ad..e25ca8f 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -3891,7 +3891,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	/* Register extent status tree shrinker */
 	ext4_es_register_shrinker(sbi);
 
-	if ((err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0)) != 0) {
+	err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0, GFP_KERNEL);
+	if (err) {
 		ext4_msg(sb, KERN_ERR, "insufficient memory");
 		goto failed_mount3;
 	}
@@ -4105,17 +4106,20 @@ no_journal:
 	block = ext4_count_free_clusters(sb);
 	ext4_free_blocks_count_set(sbi->s_es, 
 				   EXT4_C2B(sbi, block));
-	err = percpu_counter_init(&sbi->s_freeclusters_counter, block);
+	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
+				  GFP_KERNEL);
 	if (!err) {
 		unsigned long freei = ext4_count_free_inodes(sb);
 		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
-		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei);
+		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
+					  GFP_KERNEL);
 	}
 	if (!err)
 		err = percpu_counter_init(&sbi->s_dirs_counter,
-					  ext4_count_dirs(sb));
+					  ext4_count_dirs(sb), GFP_KERNEL);
 	if (!err)
-		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
+		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
+					  GFP_KERNEL);
 	if (err) {
 		ext4_msg(sb, KERN_ERR, "insufficient memory");
 		goto failed_mount6;
diff --git a/fs/file_table.c b/fs/file_table.c
index 385bfd3..0bab12b 100644
--- a/fs/file_table.c
+++ b/fs/file_table.c
@@ -331,5 +331,5 @@ void __init files_init(unsigned long mempages)
 
 	n = (mempages * (PAGE_SIZE / 1024)) / 10;
 	files_stat.max_files = max_t(unsigned long, n, NR_FILE);
-	percpu_counter_init(&nr_files, 0);
+	percpu_counter_init(&nr_files, 0, GFP_KERNEL);
 } 
diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c
index f2d0eee..8b663b2 100644
--- a/fs/quota/dquot.c
+++ b/fs/quota/dquot.c
@@ -2725,7 +2725,7 @@ static int __init dquot_init(void)
 		panic("Cannot create dquot hash table");
 
 	for (i = 0; i < _DQST_DQSTAT_LAST; i++) {
-		ret = percpu_counter_init(&dqstats.counter[i], 0);
+		ret = percpu_counter_init(&dqstats.counter[i], 0, GFP_KERNEL);
 		if (ret)
 			panic("Cannot create dquot stat counters");
 	}
diff --git a/fs/super.c b/fs/super.c
index b9a214d..1b83610 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -175,7 +175,8 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags)
 		goto fail;
 
 	for (i = 0; i < SB_FREEZE_LEVELS; i++) {
-		if (percpu_counter_init(&s->s_writers.counter[i], 0) < 0)
+		if (percpu_counter_init(&s->s_writers.counter[i], 0,
+					GFP_KERNEL) < 0)
 			goto fail;
 		lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i],
 				 &type->s_writers_key[i], 0);
diff --git a/include/linux/percpu_counter.h b/include/linux/percpu_counter.h
index d5dd465..50e5009 100644
--- a/include/linux/percpu_counter.h
+++ b/include/linux/percpu_counter.h
@@ -12,6 +12,7 @@
 #include <linux/threads.h>
 #include <linux/percpu.h>
 #include <linux/types.h>
+#include <linux/gfp.h>
 
 #ifdef CONFIG_SMP
 
@@ -26,14 +27,14 @@ struct percpu_counter {
 
 extern int percpu_counter_batch;
 
-int __percpu_counter_init(struct percpu_counter *fbc, s64 amount,
+int __percpu_counter_init(struct percpu_counter *fbc, s64 amount, gfp_t gfp,
 			  struct lock_class_key *key);
 
-#define percpu_counter_init(fbc, value)					\
+#define percpu_counter_init(fbc, value, gfp)				\
 	({								\
 		static struct lock_class_key __key;			\
 									\
-		__percpu_counter_init(fbc, value, &__key);		\
+		__percpu_counter_init(fbc, value, gfp, &__key);		\
 	})
 
 void percpu_counter_destroy(struct percpu_counter *fbc);
@@ -89,7 +90,8 @@ struct percpu_counter {
 	s64 count;
 };
 
-static inline int percpu_counter_init(struct percpu_counter *fbc, s64 amount)
+static inline int percpu_counter_init(struct percpu_counter *fbc, s64 amount,
+				      gfp_t gfp)
 {
 	fbc->count = amount;
 	return 0;
diff --git a/include/net/dst_ops.h b/include/net/dst_ops.h
index 2f26dfb..1f99a1d 100644
--- a/include/net/dst_ops.h
+++ b/include/net/dst_ops.h
@@ -63,7 +63,7 @@ static inline void dst_entries_add(struct dst_ops *dst, int val)
 
 static inline int dst_entries_init(struct dst_ops *dst)
 {
-	return percpu_counter_init(&dst->pcpuc_entries, 0);
+	return percpu_counter_init(&dst->pcpuc_entries, 0, GFP_KERNEL);
 }
 
 static inline void dst_entries_destroy(struct dst_ops *dst)
diff --git a/include/net/inet_frag.h b/include/net/inet_frag.h
index 65a8855..8d17655 100644
--- a/include/net/inet_frag.h
+++ b/include/net/inet_frag.h
@@ -151,7 +151,7 @@ static inline void add_frag_mem_limit(struct inet_frag_queue *q, int i)
 
 static inline void init_frag_mem_limit(struct netns_frags *nf)
 {
-	percpu_counter_init(&nf->mem, 0);
+	percpu_counter_init(&nf->mem, 0, GFP_KERNEL);
 }
 
 static inline unsigned int sum_frag_mem_limit(struct netns_frags *nf)
diff --git a/lib/flex_proportions.c b/lib/flex_proportions.c
index ebf3bac..b9d026b 100644
--- a/lib/flex_proportions.c
+++ b/lib/flex_proportions.c
@@ -40,7 +40,7 @@ int fprop_global_init(struct fprop_global *p)
 
 	p->period = 0;
 	/* Use 1 to avoid dealing with periods with 0 events... */
-	err = percpu_counter_init(&p->events, 1);
+	err = percpu_counter_init(&p->events, 1, GFP_KERNEL);
 	if (err)
 		return err;
 	seqcount_init(&p->sequence);
@@ -172,7 +172,7 @@ int fprop_local_init_percpu(struct fprop_local_percpu *pl)
 {
 	int err;
 
-	err = percpu_counter_init(&pl->events, 0);
+	err = percpu_counter_init(&pl->events, 0, GFP_KERNEL);
 	if (err)
 		return err;
 	pl->period = 0;
diff --git a/lib/percpu_counter.c b/lib/percpu_counter.c
index 7dd33577..1a28f0f 100644
--- a/lib/percpu_counter.c
+++ b/lib/percpu_counter.c
@@ -112,13 +112,13 @@ s64 __percpu_counter_sum(struct percpu_counter *fbc)
 }
 EXPORT_SYMBOL(__percpu_counter_sum);
 
-int __percpu_counter_init(struct percpu_counter *fbc, s64 amount,
+int __percpu_counter_init(struct percpu_counter *fbc, s64 amount, gfp_t gfp,
 			  struct lock_class_key *key)
 {
 	raw_spin_lock_init(&fbc->lock);
 	lockdep_set_class(&fbc->lock, key);
 	fbc->count = amount;
-	fbc->counters = alloc_percpu(s32);
+	fbc->counters = alloc_percpu_gfp(s32, gfp);
 	if (!fbc->counters)
 		return -ENOMEM;
 
diff --git a/lib/proportions.c b/lib/proportions.c
index 05df848..ca95f8d 100644
--- a/lib/proportions.c
+++ b/lib/proportions.c
@@ -83,11 +83,11 @@ int prop_descriptor_init(struct prop_descriptor *pd, int shift)
 	pd->index = 0;
 	pd->pg[0].shift = shift;
 	mutex_init(&pd->mutex);
-	err = percpu_counter_init(&pd->pg[0].events, 0);
+	err = percpu_counter_init(&pd->pg[0].events, 0, GFP_KERNEL);
 	if (err)
 		goto out;
 
-	err = percpu_counter_init(&pd->pg[1].events, 0);
+	err = percpu_counter_init(&pd->pg[1].events, 0, GFP_KERNEL);
 	if (err)
 		percpu_counter_destroy(&pd->pg[0].events);
 
@@ -193,7 +193,7 @@ int prop_local_init_percpu(struct prop_local_percpu *pl)
 	raw_spin_lock_init(&pl->lock);
 	pl->shift = 0;
 	pl->period = 0;
-	return percpu_counter_init(&pl->events, 0);
+	return percpu_counter_init(&pl->events, 0, GFP_KERNEL);
 }
 
 void prop_local_destroy_percpu(struct prop_local_percpu *pl)
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 1706cbb..f19a818 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -455,7 +455,7 @@ int bdi_init(struct backing_dev_info *bdi)
 	bdi_wb_init(&bdi->wb, bdi);
 
 	for (i = 0; i < NR_BDI_STAT_ITEMS; i++) {
-		err = percpu_counter_init(&bdi->bdi_stat[i], 0);
+		err = percpu_counter_init(&bdi->bdi_stat[i], 0, GFP_KERNEL);
 		if (err)
 			goto err;
 	}
diff --git a/mm/mmap.c b/mm/mmap.c
index c1f2ea4..d7ec93e 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -3196,7 +3196,7 @@ void __init mmap_init(void)
 {
 	int ret;
 
-	ret = percpu_counter_init(&vm_committed_as, 0);
+	ret = percpu_counter_init(&vm_committed_as, 0, GFP_KERNEL);
 	VM_BUG_ON(ret);
 }
 
diff --git a/mm/nommu.c b/mm/nommu.c
index a881d96..bd1808e 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -539,7 +539,7 @@ void __init mmap_init(void)
 {
 	int ret;
 
-	ret = percpu_counter_init(&vm_committed_as, 0);
+	ret = percpu_counter_init(&vm_committed_as, 0, GFP_KERNEL);
 	VM_BUG_ON(ret);
 	vm_region_jar = KMEM_CACHE(vm_region, SLAB_PANIC);
 }
diff --git a/mm/shmem.c b/mm/shmem.c
index 0e5fb22..d4bc55d 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -2993,7 +2993,7 @@ int shmem_fill_super(struct super_block *sb, void *data, int silent)
 #endif
 
 	spin_lock_init(&sbinfo->stat_lock);
-	if (percpu_counter_init(&sbinfo->used_blocks, 0))
+	if (percpu_counter_init(&sbinfo->used_blocks, 0, GFP_KERNEL))
 		goto failed;
 	sbinfo->free_inodes = sbinfo->max_inodes;
 
diff --git a/net/dccp/proto.c b/net/dccp/proto.c
index de2c1e7..e421edd 100644
--- a/net/dccp/proto.c
+++ b/net/dccp/proto.c
@@ -1115,7 +1115,7 @@ static int __init dccp_init(void)
 
 	BUILD_BUG_ON(sizeof(struct dccp_skb_cb) >
 		     FIELD_SIZEOF(struct sk_buff, cb));
-	rc = percpu_counter_init(&dccp_orphan_count, 0);
+	rc = percpu_counter_init(&dccp_orphan_count, 0, GFP_KERNEL);
 	if (rc)
 		goto out_fail;
 	rc = -ENOBUFS;
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 541f26a..d59c260 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -3188,8 +3188,8 @@ void __init tcp_init(void)
 
 	BUILD_BUG_ON(sizeof(struct tcp_skb_cb) > sizeof(skb->cb));
 
-	percpu_counter_init(&tcp_sockets_allocated, 0);
-	percpu_counter_init(&tcp_orphan_count, 0);
+	percpu_counter_init(&tcp_sockets_allocated, 0, GFP_KERNEL);
+	percpu_counter_init(&tcp_orphan_count, 0, GFP_KERNEL);
 	tcp_hashinfo.bind_bucket_cachep =
 		kmem_cache_create("tcp_bind_bucket",
 				  sizeof(struct inet_bind_bucket), 0,
diff --git a/net/ipv4/tcp_memcontrol.c b/net/ipv4/tcp_memcontrol.c
index 3af5226..1d19135 100644
--- a/net/ipv4/tcp_memcontrol.c
+++ b/net/ipv4/tcp_memcontrol.c
@@ -32,7 +32,7 @@ int tcp_init_cgroup(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
 		res_parent = &parent_cg->memory_allocated;
 
 	res_counter_init(&cg_proto->memory_allocated, res_parent);
-	percpu_counter_init(&cg_proto->sockets_allocated, 0);
+	percpu_counter_init(&cg_proto->sockets_allocated, 0, GFP_KERNEL);
 
 	return 0;
 }
diff --git a/net/sctp/protocol.c b/net/sctp/protocol.c
index 6240834..f00a85a 100644
--- a/net/sctp/protocol.c
+++ b/net/sctp/protocol.c
@@ -1341,7 +1341,7 @@ static __init int sctp_init(void)
 	if (!sctp_chunk_cachep)
 		goto err_chunk_cachep;
 
-	status = percpu_counter_init(&sctp_sockets_allocated, 0);
+	status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
 	if (status)
 		goto err_percpu_counter_init;
 
-- 
1.9.3

Re: [PATCH 1/3] percpu_counter: add @gfp to percpu_counter_init()

[David Miller]

From: Tejun Heo <tj@kernel.org>
Date: Mon, 25 Aug 2014 20:43:30 -0400

> Percpu allocator now supports allocation mask.  Add @gfp to
> percpu_counter_init() so that !GFP_KERNEL allocation masks can be used
> with percpu_counters too.
> 
> We could have left percpu_counter_init() alone and added
> percpu_counter_init_gfp(); however, the number of users isn't that
> high and introducing _gfp variants to all percpu data structures would
> be quite ugly, so let's just do the conversion.  This is the one with
> the most users.  Other percpu data structures are a lot easier to
> convert.
> 
> This patch doesn't make any functional difference.
> 
> Signed-off-by: Tejun Heo <tj@kernel.org>

For networking bits:

Acked-by: David S. Miller <davem@davemloft.net>

Re: [PATCH 1/3] percpu_counter: add @gfp to percpu_counter_init()

[Jan Kara]

On Mon 25-08-14 20:43:30, Tejun Heo wrote:
> Percpu allocator now supports allocation mask.  Add @gfp to
> percpu_counter_init() so that !GFP_KERNEL allocation masks can be used
> with percpu_counters too.
> 
> We could have left percpu_counter_init() alone and added
> percpu_counter_init_gfp(); however, the number of users isn't that
> high and introducing _gfp variants to all percpu data structures would
> be quite ugly, so let's just do the conversion.  This is the one with
> the most users.  Other percpu data structures are a lot easier to
> convert.
> 
> This patch doesn't make any functional difference.
> 
> Signed-off-by: Tejun Heo <tj@kernel.org>
> Cc: x86@kernel.org
> Cc: Jens Axboe <axboe@kernel.dk>
> Cc: Jan Kara <jack@suse.cz>
> Cc: "Theodore Ts'o" <tytso@mit.edu>
> Cc: Alexander Viro <viro@zeniv.linux.org.uk>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Andrew Morton <akpm@linux-foundation.org>
  For the ext?, quota, lib, bdi parts:
Acked-by: Jan Kara <jack@suse.cz>

> ---
>  arch/x86/kvm/mmu.c             |  2 +-
>  fs/btrfs/disk-io.c             |  8 ++++----
>  fs/btrfs/extent-tree.c         |  2 +-
>  fs/ext2/super.c                |  6 +++---
>  fs/ext3/super.c                |  6 +++---
>  fs/ext4/super.c                | 14 +++++++++-----
>  fs/file_table.c                |  2 +-
>  fs/quota/dquot.c               |  2 +-
>  fs/super.c                     |  3 ++-
>  include/linux/percpu_counter.h | 10 ++++++----
>  include/net/dst_ops.h          |  2 +-
>  include/net/inet_frag.h        |  2 +-
>  lib/flex_proportions.c         |  4 ++--
>  lib/percpu_counter.c           |  4 ++--
>  lib/proportions.c              |  6 +++---
>  mm/backing-dev.c               |  2 +-
>  mm/mmap.c                      |  2 +-
>  mm/nommu.c                     |  2 +-
>  mm/shmem.c                     |  2 +-
>  net/dccp/proto.c               |  2 +-
>  net/ipv4/tcp.c                 |  4 ++--
>  net/ipv4/tcp_memcontrol.c      |  2 +-
>  net/sctp/protocol.c            |  2 +-
>  23 files changed, 49 insertions(+), 42 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index 9314678..5bd53f2 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -4534,7 +4534,7 @@ int kvm_mmu_module_init(void)
>  	if (!mmu_page_header_cache)
>  		goto nomem;
>  
> -	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0))
> +	if (percpu_counter_init(&kvm_total_used_mmu_pages, 0, GFP_KERNEL))
>  		goto nomem;
>  
>  	register_shrinker(&mmu_shrinker);
> diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
> index 08e65e9..61dae01 100644
> --- a/fs/btrfs/disk-io.c
> +++ b/fs/btrfs/disk-io.c
> @@ -1180,7 +1180,7 @@ static struct btrfs_subvolume_writers *btrfs_alloc_subvolume_writers(void)
>  	if (!writers)
>  		return ERR_PTR(-ENOMEM);
>  
> -	ret = percpu_counter_init(&writers->counter, 0);
> +	ret = percpu_counter_init(&writers->counter, 0, GFP_KERNEL);
>  	if (ret < 0) {
>  		kfree(writers);
>  		return ERR_PTR(ret);
> @@ -2185,7 +2185,7 @@ int open_ctree(struct super_block *sb,
>  		goto fail_srcu;
>  	}
>  
> -	ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0);
> +	ret = percpu_counter_init(&fs_info->dirty_metadata_bytes, 0, GFP_KERNEL);
>  	if (ret) {
>  		err = ret;
>  		goto fail_bdi;
> @@ -2193,13 +2193,13 @@ int open_ctree(struct super_block *sb,
>  	fs_info->dirty_metadata_batch = PAGE_CACHE_SIZE *
>  					(1 + ilog2(nr_cpu_ids));
>  
> -	ret = percpu_counter_init(&fs_info->delalloc_bytes, 0);
> +	ret = percpu_counter_init(&fs_info->delalloc_bytes, 0, GFP_KERNEL);
>  	if (ret) {
>  		err = ret;
>  		goto fail_dirty_metadata_bytes;
>  	}
>  
> -	ret = percpu_counter_init(&fs_info->bio_counter, 0);
> +	ret = percpu_counter_init(&fs_info->bio_counter, 0, GFP_KERNEL);
>  	if (ret) {
>  		err = ret;
>  		goto fail_delalloc_bytes;
> diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
> index 813537f..94ec71e 100644
> --- a/fs/btrfs/extent-tree.c
> +++ b/fs/btrfs/extent-tree.c
> @@ -3493,7 +3493,7 @@ static int update_space_info(struct btrfs_fs_info *info, u64 flags,
>  	if (!found)
>  		return -ENOMEM;
>  
> -	ret = percpu_counter_init(&found->total_bytes_pinned, 0);
> +	ret = percpu_counter_init(&found->total_bytes_pinned, 0, GFP_KERNEL);
>  	if (ret) {
>  		kfree(found);
>  		return ret;
> diff --git a/fs/ext2/super.c b/fs/ext2/super.c
> index b88edc0..170dc41 100644
> --- a/fs/ext2/super.c
> +++ b/fs/ext2/super.c
> @@ -1067,14 +1067,14 @@ static int ext2_fill_super(struct super_block *sb, void *data, int silent)
>  	ext2_rsv_window_add(sb, &sbi->s_rsv_window_head);
>  
>  	err = percpu_counter_init(&sbi->s_freeblocks_counter,
> -				ext2_count_free_blocks(sb));
> +				ext2_count_free_blocks(sb), GFP_KERNEL);
>  	if (!err) {
>  		err = percpu_counter_init(&sbi->s_freeinodes_counter,
> -				ext2_count_free_inodes(sb));
> +				ext2_count_free_inodes(sb), GFP_KERNEL);
>  	}
>  	if (!err) {
>  		err = percpu_counter_init(&sbi->s_dirs_counter,
> -				ext2_count_dirs(sb));
> +				ext2_count_dirs(sb), GFP_KERNEL);
>  	}
>  	if (err) {
>  		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
> diff --git a/fs/ext3/super.c b/fs/ext3/super.c
> index 08cdfe5..eba021b 100644
> --- a/fs/ext3/super.c
> +++ b/fs/ext3/super.c
> @@ -2039,14 +2039,14 @@ static int ext3_fill_super (struct super_block *sb, void *data, int silent)
>  		goto failed_mount2;
>  	}
>  	err = percpu_counter_init(&sbi->s_freeblocks_counter,
> -			ext3_count_free_blocks(sb));
> +			ext3_count_free_blocks(sb), GFP_KERNEL);
>  	if (!err) {
>  		err = percpu_counter_init(&sbi->s_freeinodes_counter,
> -				ext3_count_free_inodes(sb));
> +				ext3_count_free_inodes(sb), GFP_KERNEL);
>  	}
>  	if (!err) {
>  		err = percpu_counter_init(&sbi->s_dirs_counter,
> -				ext3_count_dirs(sb));
> +				ext3_count_dirs(sb), GFP_KERNEL);
>  	}
>  	if (err) {
>  		ext3_msg(sb, KERN_ERR, "error: insufficient memory");
> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
> index 32b43ad..e25ca8f 100644
> --- a/fs/ext4/super.c
> +++ b/fs/ext4/super.c
> @@ -3891,7 +3891,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
>  	/* Register extent status tree shrinker */
>  	ext4_es_register_shrinker(sbi);
>  
> -	if ((err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0)) != 0) {
> +	err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0, GFP_KERNEL);
> +	if (err) {
>  		ext4_msg(sb, KERN_ERR, "insufficient memory");
>  		goto failed_mount3;
>  	}
> @@ -4105,17 +4106,20 @@ no_journal:
>  	block = ext4_count_free_clusters(sb);
>  	ext4_free_blocks_count_set(sbi->s_es, 
>  				   EXT4_C2B(sbi, block));
> -	err = percpu_counter_init(&sbi->s_freeclusters_counter, block);
> +	err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
> +				  GFP_KERNEL);
>  	if (!err) {
>  		unsigned long freei = ext4_count_free_inodes(sb);
>  		sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
> -		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei);
> +		err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
> +					  GFP_KERNEL);
>  	}
>  	if (!err)
>  		err = percpu_counter_init(&sbi->s_dirs_counter,
> -					  ext4_count_dirs(sb));
> +					  ext4_count_dirs(sb), GFP_KERNEL);
>  	if (!err)
> -		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
> +		err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
> +					  GFP_KERNEL);
>  	if (err) {
>  		ext4_msg(sb, KERN_ERR, "insufficient memory");
>  		goto failed_mount6;
> diff --git a/fs/file_table.c b/fs/file_table.c
> index 385bfd3..0bab12b 100644
> --- a/fs/file_table.c
> +++ b/fs/file_table.c
> @@ -331,5 +331,5 @@ void __init files_init(unsigned long mempages)
>  
>  	n = (mempages * (PAGE_SIZE / 1024)) / 10;
>  	files_stat.max_files = max_t(unsigned long, n, NR_FILE);
> -	percpu_counter_init(&nr_files, 0);
> +	percpu_counter_init(&nr_files, 0, GFP_KERNEL);
>  } 
> diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c
> index f2d0eee..8b663b2 100644
> --- a/fs/quota/dquot.c
> +++ b/fs/quota/dquot.c
> @@ -2725,7 +2725,7 @@ static int __init dquot_init(void)
>  		panic("Cannot create dquot hash table");
>  
>  	for (i = 0; i < _DQST_DQSTAT_LAST; i++) {
> -		ret = percpu_counter_init(&dqstats.counter[i], 0);
> +		ret = percpu_counter_init(&dqstats.counter[i], 0, GFP_KERNEL);
>  		if (ret)
>  			panic("Cannot create dquot stat counters");
>  	}
> diff --git a/fs/super.c b/fs/super.c
> index b9a214d..1b83610 100644
> --- a/fs/super.c
> +++ b/fs/super.c
> @@ -175,7 +175,8 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags)
>  		goto fail;
>  
>  	for (i = 0; i < SB_FREEZE_LEVELS; i++) {
> -		if (percpu_counter_init(&s->s_writers.counter[i], 0) < 0)
> +		if (percpu_counter_init(&s->s_writers.counter[i], 0,
> +					GFP_KERNEL) < 0)
>  			goto fail;
>  		lockdep_init_map(&s->s_writers.lock_map[i], sb_writers_name[i],
>  				 &type->s_writers_key[i], 0);
> diff --git a/include/linux/percpu_counter.h b/include/linux/percpu_counter.h
> index d5dd465..50e5009 100644
> --- a/include/linux/percpu_counter.h
> +++ b/include/linux/percpu_counter.h
> @@ -12,6 +12,7 @@
>  #include <linux/threads.h>
>  #include <linux/percpu.h>
>  #include <linux/types.h>
> +#include <linux/gfp.h>
>  
>  #ifdef CONFIG_SMP
>  
> @@ -26,14 +27,14 @@ struct percpu_counter {
>  
>  extern int percpu_counter_batch;
>  
> -int __percpu_counter_init(struct percpu_counter *fbc, s64 amount,
> +int __percpu_counter_init(struct percpu_counter *fbc, s64 amount, gfp_t gfp,
>  			  struct lock_class_key *key);
>  
> -#define percpu_counter_init(fbc, value)					\
> +#define percpu_counter_init(fbc, value, gfp)				\
>  	({								\
>  		static struct lock_class_key __key;			\
>  									\
> -		__percpu_counter_init(fbc, value, &__key);		\
> +		__percpu_counter_init(fbc, value, gfp, &__key);		\
>  	})
>  
>  void percpu_counter_destroy(struct percpu_counter *fbc);
> @@ -89,7 +90,8 @@ struct percpu_counter {
>  	s64 count;
>  };
>  
> -static inline int percpu_counter_init(struct percpu_counter *fbc, s64 amount)
> +static inline int percpu_counter_init(struct percpu_counter *fbc, s64 amount,
> +				      gfp_t gfp)
>  {
>  	fbc->count = amount;
>  	return 0;
> diff --git a/include/net/dst_ops.h b/include/net/dst_ops.h
> index 2f26dfb..1f99a1d 100644
> --- a/include/net/dst_ops.h
> +++ b/include/net/dst_ops.h
> @@ -63,7 +63,7 @@ static inline void dst_entries_add(struct dst_ops *dst, int val)
>  
>  static inline int dst_entries_init(struct dst_ops *dst)
>  {
> -	return percpu_counter_init(&dst->pcpuc_entries, 0);
> +	return percpu_counter_init(&dst->pcpuc_entries, 0, GFP_KERNEL);
>  }
>  
>  static inline void dst_entries_destroy(struct dst_ops *dst)
> diff --git a/include/net/inet_frag.h b/include/net/inet_frag.h
> index 65a8855..8d17655 100644
> --- a/include/net/inet_frag.h
> +++ b/include/net/inet_frag.h
> @@ -151,7 +151,7 @@ static inline void add_frag_mem_limit(struct inet_frag_queue *q, int i)
>  
>  static inline void init_frag_mem_limit(struct netns_frags *nf)
>  {
> -	percpu_counter_init(&nf->mem, 0);
> +	percpu_counter_init(&nf->mem, 0, GFP_KERNEL);
>  }
>  
>  static inline unsigned int sum_frag_mem_limit(struct netns_frags *nf)
> diff --git a/lib/flex_proportions.c b/lib/flex_proportions.c
> index ebf3bac..b9d026b 100644
> --- a/lib/flex_proportions.c
> +++ b/lib/flex_proportions.c
> @@ -40,7 +40,7 @@ int fprop_global_init(struct fprop_global *p)
>  
>  	p->period = 0;
>  	/* Use 1 to avoid dealing with periods with 0 events... */
> -	err = percpu_counter_init(&p->events, 1);
> +	err = percpu_counter_init(&p->events, 1, GFP_KERNEL);
>  	if (err)
>  		return err;
>  	seqcount_init(&p->sequence);
> @@ -172,7 +172,7 @@ int fprop_local_init_percpu(struct fprop_local_percpu *pl)
>  {
>  	int err;
>  
> -	err = percpu_counter_init(&pl->events, 0);
> +	err = percpu_counter_init(&pl->events, 0, GFP_KERNEL);
>  	if (err)
>  		return err;
>  	pl->period = 0;
> diff --git a/lib/percpu_counter.c b/lib/percpu_counter.c
> index 7dd33577..1a28f0f 100644
> --- a/lib/percpu_counter.c
> +++ b/lib/percpu_counter.c
> @@ -112,13 +112,13 @@ s64 __percpu_counter_sum(struct percpu_counter *fbc)
>  }
>  EXPORT_SYMBOL(__percpu_counter_sum);
>  
> -int __percpu_counter_init(struct percpu_counter *fbc, s64 amount,
> +int __percpu_counter_init(struct percpu_counter *fbc, s64 amount, gfp_t gfp,
>  			  struct lock_class_key *key)
>  {
>  	raw_spin_lock_init(&fbc->lock);
>  	lockdep_set_class(&fbc->lock, key);
>  	fbc->count = amount;
> -	fbc->counters = alloc_percpu(s32);
> +	fbc->counters = alloc_percpu_gfp(s32, gfp);
>  	if (!fbc->counters)
>  		return -ENOMEM;
>  
> diff --git a/lib/proportions.c b/lib/proportions.c
> index 05df848..ca95f8d 100644
> --- a/lib/proportions.c
> +++ b/lib/proportions.c
> @@ -83,11 +83,11 @@ int prop_descriptor_init(struct prop_descriptor *pd, int shift)
>  	pd->index = 0;
>  	pd->pg[0].shift = shift;
>  	mutex_init(&pd->mutex);
> -	err = percpu_counter_init(&pd->pg[0].events, 0);
> +	err = percpu_counter_init(&pd->pg[0].events, 0, GFP_KERNEL);
>  	if (err)
>  		goto out;
>  
> -	err = percpu_counter_init(&pd->pg[1].events, 0);
> +	err = percpu_counter_init(&pd->pg[1].events, 0, GFP_KERNEL);
>  	if (err)
>  		percpu_counter_destroy(&pd->pg[0].events);
>  
> @@ -193,7 +193,7 @@ int prop_local_init_percpu(struct prop_local_percpu *pl)
>  	raw_spin_lock_init(&pl->lock);
>  	pl->shift = 0;
>  	pl->period = 0;
> -	return percpu_counter_init(&pl->events, 0);
> +	return percpu_counter_init(&pl->events, 0, GFP_KERNEL);
>  }
>  
>  void prop_local_destroy_percpu(struct prop_local_percpu *pl)
> diff --git a/mm/backing-dev.c b/mm/backing-dev.c
> index 1706cbb..f19a818 100644
> --- a/mm/backing-dev.c
> +++ b/mm/backing-dev.c
> @@ -455,7 +455,7 @@ int bdi_init(struct backing_dev_info *bdi)
>  	bdi_wb_init(&bdi->wb, bdi);
>  
>  	for (i = 0; i < NR_BDI_STAT_ITEMS; i++) {
> -		err = percpu_counter_init(&bdi->bdi_stat[i], 0);
> +		err = percpu_counter_init(&bdi->bdi_stat[i], 0, GFP_KERNEL);
>  		if (err)
>  			goto err;
>  	}
> diff --git a/mm/mmap.c b/mm/mmap.c
> index c1f2ea4..d7ec93e 100644
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -3196,7 +3196,7 @@ void __init mmap_init(void)
>  {
>  	int ret;
>  
> -	ret = percpu_counter_init(&vm_committed_as, 0);
> +	ret = percpu_counter_init(&vm_committed_as, 0, GFP_KERNEL);
>  	VM_BUG_ON(ret);
>  }
>  
> diff --git a/mm/nommu.c b/mm/nommu.c
> index a881d96..bd1808e 100644
> --- a/mm/nommu.c
> +++ b/mm/nommu.c
> @@ -539,7 +539,7 @@ void __init mmap_init(void)
>  {
>  	int ret;
>  
> -	ret = percpu_counter_init(&vm_committed_as, 0);
> +	ret = percpu_counter_init(&vm_committed_as, 0, GFP_KERNEL);
>  	VM_BUG_ON(ret);
>  	vm_region_jar = KMEM_CACHE(vm_region, SLAB_PANIC);
>  }
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 0e5fb22..d4bc55d 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -2993,7 +2993,7 @@ int shmem_fill_super(struct super_block *sb, void *data, int silent)
>  #endif
>  
>  	spin_lock_init(&sbinfo->stat_lock);
> -	if (percpu_counter_init(&sbinfo->used_blocks, 0))
> +	if (percpu_counter_init(&sbinfo->used_blocks, 0, GFP_KERNEL))
>  		goto failed;
>  	sbinfo->free_inodes = sbinfo->max_inodes;
>  
> diff --git a/net/dccp/proto.c b/net/dccp/proto.c
> index de2c1e7..e421edd 100644
> --- a/net/dccp/proto.c
> +++ b/net/dccp/proto.c
> @@ -1115,7 +1115,7 @@ static int __init dccp_init(void)
>  
>  	BUILD_BUG_ON(sizeof(struct dccp_skb_cb) >
>  		     FIELD_SIZEOF(struct sk_buff, cb));
> -	rc = percpu_counter_init(&dccp_orphan_count, 0);
> +	rc = percpu_counter_init(&dccp_orphan_count, 0, GFP_KERNEL);
>  	if (rc)
>  		goto out_fail;
>  	rc = -ENOBUFS;
> diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
> index 541f26a..d59c260 100644
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -3188,8 +3188,8 @@ void __init tcp_init(void)
>  
>  	BUILD_BUG_ON(sizeof(struct tcp_skb_cb) > sizeof(skb->cb));
>  
> -	percpu_counter_init(&tcp_sockets_allocated, 0);
> -	percpu_counter_init(&tcp_orphan_count, 0);
> +	percpu_counter_init(&tcp_sockets_allocated, 0, GFP_KERNEL);
> +	percpu_counter_init(&tcp_orphan_count, 0, GFP_KERNEL);
>  	tcp_hashinfo.bind_bucket_cachep =
>  		kmem_cache_create("tcp_bind_bucket",
>  				  sizeof(struct inet_bind_bucket), 0,
> diff --git a/net/ipv4/tcp_memcontrol.c b/net/ipv4/tcp_memcontrol.c
> index 3af5226..1d19135 100644
> --- a/net/ipv4/tcp_memcontrol.c
> +++ b/net/ipv4/tcp_memcontrol.c
> @@ -32,7 +32,7 @@ int tcp_init_cgroup(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
>  		res_parent = &parent_cg->memory_allocated;
>  
>  	res_counter_init(&cg_proto->memory_allocated, res_parent);
> -	percpu_counter_init(&cg_proto->sockets_allocated, 0);
> +	percpu_counter_init(&cg_proto->sockets_allocated, 0, GFP_KERNEL);
>  
>  	return 0;
>  }
> diff --git a/net/sctp/protocol.c b/net/sctp/protocol.c
> index 6240834..f00a85a 100644
> --- a/net/sctp/protocol.c
> +++ b/net/sctp/protocol.c
> @@ -1341,7 +1341,7 @@ static __init int sctp_init(void)
>  	if (!sctp_chunk_cachep)
>  		goto err_chunk_cachep;
>  
> -	status = percpu_counter_init(&sctp_sockets_allocated, 0);
> +	status = percpu_counter_init(&sctp_sockets_allocated, 0, GFP_KERNEL);
>  	if (status)
>  		goto err_percpu_counter_init;
>  
> -- 
> 1.9.3
> 
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR

[PATCH 2/3] proportions: add @gfp to init functions

[Tejun Heo]

Percpu allocator now supports allocation mask.  Add @gfp to
[flex_]proportions init functions so that !GFP_KERNEL allocation masks
can be used with them too.

This patch doesn't make any functional difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Peter Zijlstra <peterz@infradead.org>
---
 include/linux/flex_proportions.h |  5 +++--
 include/linux/proportions.h      |  5 +++--
 lib/flex_proportions.c           |  8 ++++----
 lib/proportions.c                | 10 +++++-----
 mm/backing-dev.c                 |  2 +-
 mm/page-writeback.c              |  2 +-
 6 files changed, 17 insertions(+), 15 deletions(-)

diff --git a/include/linux/flex_proportions.h b/include/linux/flex_proportions.h
index 4ebc49f..0d348e0 100644
--- a/include/linux/flex_proportions.h
+++ b/include/linux/flex_proportions.h
@@ -10,6 +10,7 @@
 #include <linux/percpu_counter.h>
 #include <linux/spinlock.h>
 #include <linux/seqlock.h>
+#include <linux/gfp.h>
 
 /*
  * When maximum proportion of some event type is specified, this is the
@@ -32,7 +33,7 @@ struct fprop_global {
 	seqcount_t sequence;
 };
 
-int fprop_global_init(struct fprop_global *p);
+int fprop_global_init(struct fprop_global *p, gfp_t gfp);
 void fprop_global_destroy(struct fprop_global *p);
 bool fprop_new_period(struct fprop_global *p, int periods);
 
@@ -79,7 +80,7 @@ struct fprop_local_percpu {
 	raw_spinlock_t lock;	/* Protect period and numerator */
 };
 
-int fprop_local_init_percpu(struct fprop_local_percpu *pl);
+int fprop_local_init_percpu(struct fprop_local_percpu *pl, gfp_t gfp);
 void fprop_local_destroy_percpu(struct fprop_local_percpu *pl);
 void __fprop_inc_percpu(struct fprop_global *p, struct fprop_local_percpu *pl);
 void __fprop_inc_percpu_max(struct fprop_global *p, struct fprop_local_percpu *pl,
diff --git a/include/linux/proportions.h b/include/linux/proportions.h
index 26a8a4e..00e8e8f 100644
--- a/include/linux/proportions.h
+++ b/include/linux/proportions.h
@@ -12,6 +12,7 @@
 #include <linux/percpu_counter.h>
 #include <linux/spinlock.h>
 #include <linux/mutex.h>
+#include <linux/gfp.h>
 
 struct prop_global {
 	/*
@@ -40,7 +41,7 @@ struct prop_descriptor {
 	struct mutex mutex;		/* serialize the prop_global switch */
 };
 
-int prop_descriptor_init(struct prop_descriptor *pd, int shift);
+int prop_descriptor_init(struct prop_descriptor *pd, int shift, gfp_t gfp);
 void prop_change_shift(struct prop_descriptor *pd, int new_shift);
 
 /*
@@ -61,7 +62,7 @@ struct prop_local_percpu {
 	raw_spinlock_t lock;		/* protect the snapshot state */
 };
 
-int prop_local_init_percpu(struct prop_local_percpu *pl);
+int prop_local_init_percpu(struct prop_local_percpu *pl, gfp_t gfp);
 void prop_local_destroy_percpu(struct prop_local_percpu *pl);
 void __prop_inc_percpu(struct prop_descriptor *pd, struct prop_local_percpu *pl);
 void prop_fraction_percpu(struct prop_descriptor *pd, struct prop_local_percpu *pl,
diff --git a/lib/flex_proportions.c b/lib/flex_proportions.c
index b9d026b..8f25652 100644
--- a/lib/flex_proportions.c
+++ b/lib/flex_proportions.c
@@ -34,13 +34,13 @@
  */
 #include <linux/flex_proportions.h>
 
-int fprop_global_init(struct fprop_global *p)
+int fprop_global_init(struct fprop_global *p, gfp_t gfp)
 {
 	int err;
 
 	p->period = 0;
 	/* Use 1 to avoid dealing with periods with 0 events... */
-	err = percpu_counter_init(&p->events, 1, GFP_KERNEL);
+	err = percpu_counter_init(&p->events, 1, gfp);
 	if (err)
 		return err;
 	seqcount_init(&p->sequence);
@@ -168,11 +168,11 @@ void fprop_fraction_single(struct fprop_global *p,
  */
 #define PROP_BATCH (8*(1+ilog2(nr_cpu_ids)))
 
-int fprop_local_init_percpu(struct fprop_local_percpu *pl)
+int fprop_local_init_percpu(struct fprop_local_percpu *pl, gfp_t gfp)
 {
 	int err;
 
-	err = percpu_counter_init(&pl->events, 0, GFP_KERNEL);
+	err = percpu_counter_init(&pl->events, 0, gfp);
 	if (err)
 		return err;
 	pl->period = 0;
diff --git a/lib/proportions.c b/lib/proportions.c
index ca95f8d..6f72429 100644
--- a/lib/proportions.c
+++ b/lib/proportions.c
@@ -73,7 +73,7 @@
 #include <linux/proportions.h>
 #include <linux/rcupdate.h>
 
-int prop_descriptor_init(struct prop_descriptor *pd, int shift)
+int prop_descriptor_init(struct prop_descriptor *pd, int shift, gfp_t gfp)
 {
 	int err;
 
@@ -83,11 +83,11 @@ int prop_descriptor_init(struct prop_descriptor *pd, int shift)
 	pd->index = 0;
 	pd->pg[0].shift = shift;
 	mutex_init(&pd->mutex);
-	err = percpu_counter_init(&pd->pg[0].events, 0, GFP_KERNEL);
+	err = percpu_counter_init(&pd->pg[0].events, 0, gfp);
 	if (err)
 		goto out;
 
-	err = percpu_counter_init(&pd->pg[1].events, 0, GFP_KERNEL);
+	err = percpu_counter_init(&pd->pg[1].events, 0, gfp);
 	if (err)
 		percpu_counter_destroy(&pd->pg[0].events);
 
@@ -188,12 +188,12 @@ prop_adjust_shift(int *pl_shift, unsigned long *pl_period, int new_shift)
 
 #define PROP_BATCH (8*(1+ilog2(nr_cpu_ids)))
 
-int prop_local_init_percpu(struct prop_local_percpu *pl)
+int prop_local_init_percpu(struct prop_local_percpu *pl, gfp_t gfp)
 {
 	raw_spin_lock_init(&pl->lock);
 	pl->shift = 0;
 	pl->period = 0;
-	return percpu_counter_init(&pl->events, 0, GFP_KERNEL);
+	return percpu_counter_init(&pl->events, 0, gfp);
 }
 
 void prop_local_destroy_percpu(struct prop_local_percpu *pl)
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index f19a818..64ec49d 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -470,7 +470,7 @@ int bdi_init(struct backing_dev_info *bdi)
 	bdi->write_bandwidth = INIT_BW;
 	bdi->avg_write_bandwidth = INIT_BW;
 
-	err = fprop_local_init_percpu(&bdi->completions);
+	err = fprop_local_init_percpu(&bdi->completions, GFP_KERNEL);
 
 	if (err) {
 err:
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 91d73ef..5085994 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -1777,7 +1777,7 @@ void __init page_writeback_init(void)
 	writeback_set_ratelimit();
 	register_cpu_notifier(&ratelimit_nb);
 
-	fprop_global_init(&writeout_completions);
+	fprop_global_init(&writeout_completions, GFP_KERNEL);
 }
 
 /**
-- 
1.9.3

Re: [PATCH 2/3] proportions: add @gfp to init functions

[Jan Kara]

On Mon 25-08-14 20:43:31, Tejun Heo wrote:
> Percpu allocator now supports allocation mask.  Add @gfp to
> [flex_]proportions init functions so that !GFP_KERNEL allocation masks
> can be used with them too.
> 
> This patch doesn't make any functional difference.
  Looks good to me. You can add:
Reviewed-by: Jan Kara <jack@suse.cz>

								Honza

> 
> Signed-off-by: Tejun Heo <tj@kernel.org>
> Cc: Jan Kara <jack@suse.cz>
> Cc: Peter Zijlstra <peterz@infradead.org>
> ---
>  include/linux/flex_proportions.h |  5 +++--
>  include/linux/proportions.h      |  5 +++--
>  lib/flex_proportions.c           |  8 ++++----
>  lib/proportions.c                | 10 +++++-----
>  mm/backing-dev.c                 |  2 +-
>  mm/page-writeback.c              |  2 +-
>  6 files changed, 17 insertions(+), 15 deletions(-)
> 
> diff --git a/include/linux/flex_proportions.h b/include/linux/flex_proportions.h
> index 4ebc49f..0d348e0 100644
> --- a/include/linux/flex_proportions.h
> +++ b/include/linux/flex_proportions.h
> @@ -10,6 +10,7 @@
>  #include <linux/percpu_counter.h>
>  #include <linux/spinlock.h>
>  #include <linux/seqlock.h>
> +#include <linux/gfp.h>
>  
>  /*
>   * When maximum proportion of some event type is specified, this is the
> @@ -32,7 +33,7 @@ struct fprop_global {
>  	seqcount_t sequence;
>  };
>  
> -int fprop_global_init(struct fprop_global *p);
> +int fprop_global_init(struct fprop_global *p, gfp_t gfp);
>  void fprop_global_destroy(struct fprop_global *p);
>  bool fprop_new_period(struct fprop_global *p, int periods);
>  
> @@ -79,7 +80,7 @@ struct fprop_local_percpu {
>  	raw_spinlock_t lock;	/* Protect period and numerator */
>  };
>  
> -int fprop_local_init_percpu(struct fprop_local_percpu *pl);
> +int fprop_local_init_percpu(struct fprop_local_percpu *pl, gfp_t gfp);
>  void fprop_local_destroy_percpu(struct fprop_local_percpu *pl);
>  void __fprop_inc_percpu(struct fprop_global *p, struct fprop_local_percpu *pl);
>  void __fprop_inc_percpu_max(struct fprop_global *p, struct fprop_local_percpu *pl,
> diff --git a/include/linux/proportions.h b/include/linux/proportions.h
> index 26a8a4e..00e8e8f 100644
> --- a/include/linux/proportions.h
> +++ b/include/linux/proportions.h
> @@ -12,6 +12,7 @@
>  #include <linux/percpu_counter.h>
>  #include <linux/spinlock.h>
>  #include <linux/mutex.h>
> +#include <linux/gfp.h>
>  
>  struct prop_global {
>  	/*
> @@ -40,7 +41,7 @@ struct prop_descriptor {
>  	struct mutex mutex;		/* serialize the prop_global switch */
>  };
>  
> -int prop_descriptor_init(struct prop_descriptor *pd, int shift);
> +int prop_descriptor_init(struct prop_descriptor *pd, int shift, gfp_t gfp);
>  void prop_change_shift(struct prop_descriptor *pd, int new_shift);
>  
>  /*
> @@ -61,7 +62,7 @@ struct prop_local_percpu {
>  	raw_spinlock_t lock;		/* protect the snapshot state */
>  };
>  
> -int prop_local_init_percpu(struct prop_local_percpu *pl);
> +int prop_local_init_percpu(struct prop_local_percpu *pl, gfp_t gfp);
>  void prop_local_destroy_percpu(struct prop_local_percpu *pl);
>  void __prop_inc_percpu(struct prop_descriptor *pd, struct prop_local_percpu *pl);
>  void prop_fraction_percpu(struct prop_descriptor *pd, struct prop_local_percpu *pl,
> diff --git a/lib/flex_proportions.c b/lib/flex_proportions.c
> index b9d026b..8f25652 100644
> --- a/lib/flex_proportions.c
> +++ b/lib/flex_proportions.c
> @@ -34,13 +34,13 @@
>   */
>  #include <linux/flex_proportions.h>
>  
> -int fprop_global_init(struct fprop_global *p)
> +int fprop_global_init(struct fprop_global *p, gfp_t gfp)
>  {
>  	int err;
>  
>  	p->period = 0;
>  	/* Use 1 to avoid dealing with periods with 0 events... */
> -	err = percpu_counter_init(&p->events, 1, GFP_KERNEL);
> +	err = percpu_counter_init(&p->events, 1, gfp);
>  	if (err)
>  		return err;
>  	seqcount_init(&p->sequence);
> @@ -168,11 +168,11 @@ void fprop_fraction_single(struct fprop_global *p,
>   */
>  #define PROP_BATCH (8*(1+ilog2(nr_cpu_ids)))
>  
> -int fprop_local_init_percpu(struct fprop_local_percpu *pl)
> +int fprop_local_init_percpu(struct fprop_local_percpu *pl, gfp_t gfp)
>  {
>  	int err;
>  
> -	err = percpu_counter_init(&pl->events, 0, GFP_KERNEL);
> +	err = percpu_counter_init(&pl->events, 0, gfp);
>  	if (err)
>  		return err;
>  	pl->period = 0;
> diff --git a/lib/proportions.c b/lib/proportions.c
> index ca95f8d..6f72429 100644
> --- a/lib/proportions.c
> +++ b/lib/proportions.c
> @@ -73,7 +73,7 @@
>  #include <linux/proportions.h>
>  #include <linux/rcupdate.h>
>  
> -int prop_descriptor_init(struct prop_descriptor *pd, int shift)
> +int prop_descriptor_init(struct prop_descriptor *pd, int shift, gfp_t gfp)
>  {
>  	int err;
>  
> @@ -83,11 +83,11 @@ int prop_descriptor_init(struct prop_descriptor *pd, int shift)
>  	pd->index = 0;
>  	pd->pg[0].shift = shift;
>  	mutex_init(&pd->mutex);
> -	err = percpu_counter_init(&pd->pg[0].events, 0, GFP_KERNEL);
> +	err = percpu_counter_init(&pd->pg[0].events, 0, gfp);
>  	if (err)
>  		goto out;
>  
> -	err = percpu_counter_init(&pd->pg[1].events, 0, GFP_KERNEL);
> +	err = percpu_counter_init(&pd->pg[1].events, 0, gfp);
>  	if (err)
>  		percpu_counter_destroy(&pd->pg[0].events);
>  
> @@ -188,12 +188,12 @@ prop_adjust_shift(int *pl_shift, unsigned long *pl_period, int new_shift)
>  
>  #define PROP_BATCH (8*(1+ilog2(nr_cpu_ids)))
>  
> -int prop_local_init_percpu(struct prop_local_percpu *pl)
> +int prop_local_init_percpu(struct prop_local_percpu *pl, gfp_t gfp)
>  {
>  	raw_spin_lock_init(&pl->lock);
>  	pl->shift = 0;
>  	pl->period = 0;
> -	return percpu_counter_init(&pl->events, 0, GFP_KERNEL);
> +	return percpu_counter_init(&pl->events, 0, gfp);
>  }
>  
>  void prop_local_destroy_percpu(struct prop_local_percpu *pl)
> diff --git a/mm/backing-dev.c b/mm/backing-dev.c
> index f19a818..64ec49d 100644
> --- a/mm/backing-dev.c
> +++ b/mm/backing-dev.c
> @@ -470,7 +470,7 @@ int bdi_init(struct backing_dev_info *bdi)
>  	bdi->write_bandwidth = INIT_BW;
>  	bdi->avg_write_bandwidth = INIT_BW;
>  
> -	err = fprop_local_init_percpu(&bdi->completions);
> +	err = fprop_local_init_percpu(&bdi->completions, GFP_KERNEL);
>  
>  	if (err) {
>  err:
> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
> index 91d73ef..5085994 100644
> --- a/mm/page-writeback.c
> +++ b/mm/page-writeback.c
> @@ -1777,7 +1777,7 @@ void __init page_writeback_init(void)
>  	writeback_set_ratelimit();
>  	register_cpu_notifier(&ratelimit_nb);
>  
> -	fprop_global_init(&writeout_completions);
> +	fprop_global_init(&writeout_completions, GFP_KERNEL);
>  }
>  
>  /**
> -- 
> 1.9.3
> 
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR

[PATCH 3/3] percpu-refcount: add @gfp to percpu_ref_init()

[Tejun Heo]

Percpu allocator now supports allocation mask.  Add @gfp to
percpu_ref_init() so that !GFP_KERNEL allocation masks can be used
with percpu_refs too.

This patch doesn't make any functional difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Kent Overstreet <koverstreet@google.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Nicholas A. Bellinger <nab@linux-iscsi.org>
---
 drivers/target/target_core_tpg.c | 3 ++-
 fs/aio.c                         | 4 ++--
 include/linux/percpu-refcount.h  | 3 ++-
 kernel/cgroup.c                  | 6 +++---
 lib/percpu-refcount.c            | 6 ++++--
 5 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/drivers/target/target_core_tpg.c b/drivers/target/target_core_tpg.c
index fddfae6..4ab6da3 100644
--- a/drivers/target/target_core_tpg.c
+++ b/drivers/target/target_core_tpg.c
@@ -819,7 +819,8 @@ int core_tpg_add_lun(
 {
 	int ret;
 
-	ret = percpu_ref_init(&lun->lun_ref, core_tpg_lun_ref_release);
+	ret = percpu_ref_init(&lun->lun_ref, core_tpg_lun_ref_release,
+			      GFP_KERNEL);
 	if (ret < 0)
 		return ret;
 
diff --git a/fs/aio.c b/fs/aio.c
index bd7ec2c..93fbcc0f 100644
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -666,10 +666,10 @@ static struct kioctx *ioctx_alloc(unsigned nr_events)
 
 	INIT_LIST_HEAD(&ctx->active_reqs);
 
-	if (percpu_ref_init(&ctx->users, free_ioctx_users))
+	if (percpu_ref_init(&ctx->users, free_ioctx_users, GFP_KERNEL))
 		goto err;
 
-	if (percpu_ref_init(&ctx->reqs, free_ioctx_reqs))
+	if (percpu_ref_init(&ctx->reqs, free_ioctx_reqs, GFP_KERNEL))
 		goto err;
 
 	ctx->cpu = alloc_percpu(struct kioctx_cpu);
diff --git a/include/linux/percpu-refcount.h b/include/linux/percpu-refcount.h
index 3dfbf23..ee83251 100644
--- a/include/linux/percpu-refcount.h
+++ b/include/linux/percpu-refcount.h
@@ -49,6 +49,7 @@
 #include <linux/kernel.h>
 #include <linux/percpu.h>
 #include <linux/rcupdate.h>
+#include <linux/gfp.h>
 
 struct percpu_ref;
 typedef void (percpu_ref_func_t)(struct percpu_ref *);
@@ -66,7 +67,7 @@ struct percpu_ref {
 };
 
 int __must_check percpu_ref_init(struct percpu_ref *ref,
-				 percpu_ref_func_t *release);
+				 percpu_ref_func_t *release, gfp_t gfp);
 void percpu_ref_reinit(struct percpu_ref *ref);
 void percpu_ref_exit(struct percpu_ref *ref);
 void percpu_ref_kill_and_confirm(struct percpu_ref *ref,
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 7dc8788..589b4d8 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -1628,7 +1628,7 @@ static int cgroup_setup_root(struct cgroup_root *root, unsigned int ss_mask)
 		goto out;
 	root_cgrp->id = ret;
 
-	ret = percpu_ref_init(&root_cgrp->self.refcnt, css_release);
+	ret = percpu_ref_init(&root_cgrp->self.refcnt, css_release, GFP_KERNEL);
 	if (ret)
 		goto out;
 
@@ -4487,7 +4487,7 @@ static int create_css(struct cgroup *cgrp, struct cgroup_subsys *ss,
 
 	init_and_link_css(css, ss, cgrp);
 
-	err = percpu_ref_init(&css->refcnt, css_release);
+	err = percpu_ref_init(&css->refcnt, css_release, GFP_KERNEL);
 	if (err)
 		goto err_free_css;
 
@@ -4555,7 +4555,7 @@ static int cgroup_mkdir(struct kernfs_node *parent_kn, const char *name,
 		goto out_unlock;
 	}
 
-	ret = percpu_ref_init(&cgrp->self.refcnt, css_release);
+	ret = percpu_ref_init(&cgrp->self.refcnt, css_release, GFP_KERNEL);
 	if (ret)
 		goto out_free_cgrp;
 
diff --git a/lib/percpu-refcount.c b/lib/percpu-refcount.c
index fe5a334..ff99032 100644
--- a/lib/percpu-refcount.c
+++ b/lib/percpu-refcount.c
@@ -40,6 +40,7 @@ static unsigned __percpu *pcpu_count_ptr(struct percpu_ref *ref)
  * percpu_ref_init - initialize a percpu refcount
  * @ref: percpu_ref to initialize
  * @release: function which will be called when refcount hits 0
+ * @gfp: allocation mask to use
  *
  * Initializes the refcount in single atomic counter mode with a refcount of 1;
  * analagous to atomic_set(ref, 1).
@@ -47,11 +48,12 @@ static unsigned __percpu *pcpu_count_ptr(struct percpu_ref *ref)
  * Note that @release must not sleep - it may potentially be called from RCU
  * callback context by percpu_ref_kill().
  */
-int percpu_ref_init(struct percpu_ref *ref, percpu_ref_func_t *release)
+int percpu_ref_init(struct percpu_ref *ref, percpu_ref_func_t *release,
+		    gfp_t gfp)
 {
 	atomic_set(&ref->count, 1 + PCPU_COUNT_BIAS);
 
-	ref->pcpu_count_ptr = (unsigned long)alloc_percpu(unsigned);
+	ref->pcpu_count_ptr = (unsigned long)alloc_percpu_gfp(unsigned, gfp);
 	if (!ref->pcpu_count_ptr)
 		return -ENOMEM;
 
-- 
1.9.3

[PATCH v2 3/3] percpu-refcount: add @gfp to percpu_ref_init()

[Tejun Heo]

>From ad5328cfb8c0d828e07ac428fdf3fbcb4f8698ad Mon Sep 17 00:00:00 2001
From: Tejun Heo <tj@kernel.org>
Date: Tue, 26 Aug 2014 11:33:17 -0400

Percpu allocator now supports allocation mask.  Add @gfp to
percpu_ref_init() so that !GFP_KERNEL allocation masks can be used
with percpu_refs too.

This patch doesn't make any functional difference.

v2: blk-mq conversion was missing.  Updated.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Kent Overstreet <koverstreet@google.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: Li Zefan <lizefan@huawei.com>
Cc: Nicholas A. Bellinger <nab@linux-iscsi.org>
Cc: Jens Axboe <axboe@kernel.dk>
---
Missed blk-mq conversion while forward porting to v3.17-rc1.  Updated.

Thanks.

 block/blk-mq.c                   | 3 ++-
 drivers/target/target_core_tpg.c | 3 ++-
 fs/aio.c                         | 4 ++--
 include/linux/percpu-refcount.h  | 3 ++-
 kernel/cgroup.c                  | 6 +++---
 lib/percpu-refcount.c            | 6 ++++--
 6 files changed, 15 insertions(+), 10 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 5189cb1..702df07 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1776,7 +1776,8 @@ struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *set)
 	if (!q)
 		goto err_hctxs;
 
-	if (percpu_ref_init(&q->mq_usage_counter, blk_mq_usage_counter_release))
+	if (percpu_ref_init(&q->mq_usage_counter, blk_mq_usage_counter_release,
+			    GFP_KERNEL))
 		goto err_map;
 
 	setup_timer(&q->timeout, blk_mq_rq_timer, (unsigned long) q);
diff --git a/drivers/target/target_core_tpg.c b/drivers/target/target_core_tpg.c
index fddfae6..4ab6da3 100644
--- a/drivers/target/target_core_tpg.c
+++ b/drivers/target/target_core_tpg.c
@@ -819,7 +819,8 @@ int core_tpg_add_lun(
 {
 	int ret;
 
-	ret = percpu_ref_init(&lun->lun_ref, core_tpg_lun_ref_release);
+	ret = percpu_ref_init(&lun->lun_ref, core_tpg_lun_ref_release,
+			      GFP_KERNEL);
 	if (ret < 0)
 		return ret;
 
diff --git a/fs/aio.c b/fs/aio.c
index bd7ec2c..93fbcc0f 100644
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -666,10 +666,10 @@ static struct kioctx *ioctx_alloc(unsigned nr_events)
 
 	INIT_LIST_HEAD(&ctx->active_reqs);
 
-	if (percpu_ref_init(&ctx->users, free_ioctx_users))
+	if (percpu_ref_init(&ctx->users, free_ioctx_users, GFP_KERNEL))
 		goto err;
 
-	if (percpu_ref_init(&ctx->reqs, free_ioctx_reqs))
+	if (percpu_ref_init(&ctx->reqs, free_ioctx_reqs, GFP_KERNEL))
 		goto err;
 
 	ctx->cpu = alloc_percpu(struct kioctx_cpu);
diff --git a/include/linux/percpu-refcount.h b/include/linux/percpu-refcount.h
index 3dfbf23..ee83251 100644
--- a/include/linux/percpu-refcount.h
+++ b/include/linux/percpu-refcount.h
@@ -49,6 +49,7 @@
 #include <linux/kernel.h>
 #include <linux/percpu.h>
 #include <linux/rcupdate.h>
+#include <linux/gfp.h>
 
 struct percpu_ref;
 typedef void (percpu_ref_func_t)(struct percpu_ref *);
@@ -66,7 +67,7 @@ struct percpu_ref {
 };
 
 int __must_check percpu_ref_init(struct percpu_ref *ref,
-				 percpu_ref_func_t *release);
+				 percpu_ref_func_t *release, gfp_t gfp);
 void percpu_ref_reinit(struct percpu_ref *ref);
 void percpu_ref_exit(struct percpu_ref *ref);
 void percpu_ref_kill_and_confirm(struct percpu_ref *ref,
diff --git a/kernel/cgroup.c b/kernel/cgroup.c
index 7dc8788..589b4d8 100644
--- a/kernel/cgroup.c
+++ b/kernel/cgroup.c
@@ -1628,7 +1628,7 @@ static int cgroup_setup_root(struct cgroup_root *root, unsigned int ss_mask)
 		goto out;
 	root_cgrp->id = ret;
 
-	ret = percpu_ref_init(&root_cgrp->self.refcnt, css_release);
+	ret = percpu_ref_init(&root_cgrp->self.refcnt, css_release, GFP_KERNEL);
 	if (ret)
 		goto out;
 
@@ -4487,7 +4487,7 @@ static int create_css(struct cgroup *cgrp, struct cgroup_subsys *ss,
 
 	init_and_link_css(css, ss, cgrp);
 
-	err = percpu_ref_init(&css->refcnt, css_release);
+	err = percpu_ref_init(&css->refcnt, css_release, GFP_KERNEL);
 	if (err)
 		goto err_free_css;
 
@@ -4555,7 +4555,7 @@ static int cgroup_mkdir(struct kernfs_node *parent_kn, const char *name,
 		goto out_unlock;
 	}
 
-	ret = percpu_ref_init(&cgrp->self.refcnt, css_release);
+	ret = percpu_ref_init(&cgrp->self.refcnt, css_release, GFP_KERNEL);
 	if (ret)
 		goto out_free_cgrp;
 
diff --git a/lib/percpu-refcount.c b/lib/percpu-refcount.c
index fe5a334..ff99032 100644
--- a/lib/percpu-refcount.c
+++ b/lib/percpu-refcount.c
@@ -40,6 +40,7 @@ static unsigned __percpu *pcpu_count_ptr(struct percpu_ref *ref)
  * percpu_ref_init - initialize a percpu refcount
  * @ref: percpu_ref to initialize
  * @release: function which will be called when refcount hits 0
+ * @gfp: allocation mask to use
  *
  * Initializes the refcount in single atomic counter mode with a refcount of 1;
  * analagous to atomic_set(ref, 1).
@@ -47,11 +48,12 @@ static unsigned __percpu *pcpu_count_ptr(struct percpu_ref *ref)
  * Note that @release must not sleep - it may potentially be called from RCU
  * callback context by percpu_ref_kill().
  */
-int percpu_ref_init(struct percpu_ref *ref, percpu_ref_func_t *release)
+int percpu_ref_init(struct percpu_ref *ref, percpu_ref_func_t *release,
+		    gfp_t gfp)
 {
 	atomic_set(&ref->count, 1 + PCPU_COUNT_BIAS);
 
-	ref->pcpu_count_ptr = (unsigned long)alloc_percpu(unsigned);
+	ref->pcpu_count_ptr = (unsigned long)alloc_percpu_gfp(unsigned, gfp);
 	if (!ref->pcpu_count_ptr)
 		return -ENOMEM;
 
-- 
1.9.3

[PATCH 0.5/3] percpu_counter: make percpu_counters_lock irq-safe

[Tejun Heo]

percpu_counter is scheduled to grow @gfp support to allow atomic
initialization.  This patch makes percpu_counters_lock irq-safe so
that it can be safely used from atomic contexts.

Signed-off-by: Tejun Heo <tj@kernel.org>
---
 lib/percpu_counter.c |   16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

--- a/lib/percpu_counter.c
+++ b/lib/percpu_counter.c
@@ -115,6 +115,8 @@ EXPORT_SYMBOL(__percpu_counter_sum);
 int __percpu_counter_init(struct percpu_counter *fbc, s64 amount, gfp_t gfp,
 			  struct lock_class_key *key)
 {
+	unsigned long flags __maybe_unused;
+
 	raw_spin_lock_init(&fbc->lock);
 	lockdep_set_class(&fbc->lock, key);
 	fbc->count = amount;
@@ -126,9 +128,9 @@ int __percpu_counter_init(struct percpu_
 
 #ifdef CONFIG_HOTPLUG_CPU
 	INIT_LIST_HEAD(&fbc->list);
-	spin_lock(&percpu_counters_lock);
+	spin_lock_irqsave(&percpu_counters_lock, flags);
 	list_add(&fbc->list, &percpu_counters);
-	spin_unlock(&percpu_counters_lock);
+	spin_unlock_irqrestore(&percpu_counters_lock, flags);
 #endif
 	return 0;
 }
@@ -136,15 +138,17 @@ EXPORT_SYMBOL(__percpu_counter_init);
 
 void percpu_counter_destroy(struct percpu_counter *fbc)
 {
+	unsigned long flags __maybe_unused;
+
 	if (!fbc->counters)
 		return;
 
 	debug_percpu_counter_deactivate(fbc);
 
 #ifdef CONFIG_HOTPLUG_CPU
-	spin_lock(&percpu_counters_lock);
+	spin_lock_irqsave(&percpu_counters_lock, flags);
 	list_del(&fbc->list);
-	spin_unlock(&percpu_counters_lock);
+	spin_unlock_irqrestore(&percpu_counters_lock, flags);
 #endif
 	free_percpu(fbc->counters);
 	fbc->counters = NULL;
@@ -173,7 +177,7 @@ static int percpu_counter_hotcpu_callbac
 		return NOTIFY_OK;
 
 	cpu = (unsigned long)hcpu;
-	spin_lock(&percpu_counters_lock);
+	spin_lock_irq(&percpu_counters_lock);
 	list_for_each_entry(fbc, &percpu_counters, list) {
 		s32 *pcount;
 		unsigned long flags;
@@ -184,7 +188,7 @@ static int percpu_counter_hotcpu_callbac
 		*pcount = 0;
 		raw_spin_unlock_irqrestore(&fbc->lock, flags);
 	}
-	spin_unlock(&percpu_counters_lock);
+	spin_unlock_irq(&percpu_counters_lock);
 #endif
 	return NOTIFY_OK;
 }

Re: [PATCHSET percpu/for-3.18] add @gfp to init functions of percpu data structures

[Tejun Heo]

Applied to percpu/for-3.18.

Thanks.

-- 
tejun