[PATCH 00/29] Improve radix tree for 4.10

[PATCH 00/29] Improve radix tree for 4.10

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

Hi Andrew,

Please include these patches in the -mm tree for 4.10.  Mostly these are
improvements; the only bug fixes in here relate to multiorder entries
(which as far as I'm aware remain unused).  The IDR rewrite has the
highest potential for causing mayhem as the test suite is quite spartan.
We have an Outreachy intern scheduled to work on the test suite for the
2016 winter season, so hopefully it will improve soon.

I posted a series of patches to improve the radix tree code two months
ago, and Konstantin had the quite reasonable objection that they slowed
down the iterators.  He helpfully sent out the benchmark he used.
With this revision of the patchset, I have slightly improved performance
on his benchmark.

Noteworthy changes from the last revision of the patchset:

 - Fix the existing radix tree iterator design instead of rewriting it
 - Avoid adding the new radix_tree_for_each_tagged_max iterator; just
   check in the one caller whether we've exceeded the limit.
 - Change the name of radix_tree_iter_save() back to
   radix_tree_iter_next().  I went back and forth on this one, and could
   easily be convinced to rename it as I think radix_tree_iter_save()
   makes more sense to the callers.
 - Add radix_tree_iter_delete() instead of poking around the internals
   of the radix tree in ida_destroy()
 - Add accessors for the idr cyclic allocation cursor

Patches 1-9: Test suite improvements, including Konstantin's benchmark
Patches 10-12: join/split (Kirill wants these for huge page cache)
Patches 13-17: General improvements to the radix tree
Patches 18-22: Radix tree iterator improvements
Patches 23-29: Replace the IDR with the radix tree

Konstantin Khlebnikov (1):
  radix tree test suite: benchmark for iterator

Matthew Wilcox (28):
  tools: Add WARN_ON_ONCE
  radix tree test suite: Allow GFP_ATOMIC allocations to fail
  radix tree test suite: Track preempt_count
  radix tree test suite: Free preallocated nodes
  radix tree test suite: Make runs more reproducible
  radix tree test suite: Use rcu_barrier
  tools: Add more bitmap functions
  radix tree test suite: Use common find-bit code
  radix-tree: Add radix_tree_join
  radix-tree: Add radix_tree_split
  radix-tree: Add radix_tree_split_preload()
  radix-tree: Fix typo
  radix-tree: Move rcu_head into a union with private_list
  radix-tree: Create node_tag_set()
  radix-tree: Make radix_tree_find_next_bit more useful
  radix-tree: Improve dump output
  btrfs: Fix race in btrfs_free_dummy_fs_info()
  radix tree test suite: iteration test misuses RCU
  radix tree: Improve multiorder iterators
  radix-tree: Delete radix_tree_locate_item()
  radix-tree: Delete radix_tree_range_tag_if_tagged()
  idr: Add ida_is_empty
  tpm: Use idr_find(), not idr_find_slowpath()
  rxrpc: Abstract away knowledge of IDR internals
  idr: Reduce the number of bits per level from 8 to 6
  radix tree test suite: Add some more functionality
  radix-tree: Create all_tag_set
  Reimplement IDR and IDA using the radix tree

 drivers/char/tpm/tpm-chip.c                        |    4 +-
 drivers/usb/gadget/function/f_hid.c                |    6 +-
 drivers/usb/gadget/function/f_printer.c            |    6 +-
 fs/btrfs/tests/btrfs-tests.c                       |    1 +
 include/linux/idr.h                                |  156 +--
 include/linux/radix-tree.h                         |  161 ++-
 init/main.c                                        |    3 +-
 lib/idr.c                                          | 1075 ----------------
 lib/radix-tree.c                                   | 1328 +++++++++++++++-----
 mm/khugepaged.c                                    |    2 +-
 mm/page-writeback.c                                |   28 +-
 mm/shmem.c                                         |   32 +-
 net/rxrpc/af_rxrpc.c                               |   11 +-
 net/rxrpc/conn_client.c                            |    4 +-
 tools/include/asm/bug.h                            |   11 +
 tools/include/linux/bitmap.h                       |   26 +
 tools/lib/find_bit.c                               |    8 +
 tools/testing/radix-tree/Makefile                  |   18 +-
 tools/testing/radix-tree/benchmark.c               |   98 ++
 tools/testing/radix-tree/find_next_bit.c           |   57 -
 tools/testing/radix-tree/idr.c                     |  148 +++
 tools/testing/radix-tree/iteration_check.c         |   39 +-
 tools/testing/radix-tree/linux.c                   |   23 +-
 tools/testing/radix-tree/linux/bitops.h            |   40 +-
 tools/testing/radix-tree/linux/bitops/non-atomic.h |   13 +-
 tools/testing/radix-tree/linux/bug.h               |    2 +-
 tools/testing/radix-tree/linux/gfp.h               |   22 +-
 tools/testing/radix-tree/linux/idr.h               |    1 +
 tools/testing/radix-tree/linux/kernel.h            |   20 +
 tools/testing/radix-tree/linux/preempt.h           |    6 +-
 tools/testing/radix-tree/linux/slab.h              |    6 +-
 tools/testing/radix-tree/linux/types.h             |    2 -
 tools/testing/radix-tree/main.c                    |   80 +-
 tools/testing/radix-tree/multiorder.c              |  111 +-
 tools/testing/radix-tree/regression2.c             |    3 +-
 tools/testing/radix-tree/regression3.c             |    6 +-
 tools/testing/radix-tree/tag_check.c               |    9 +-
 tools/testing/radix-tree/test.c                    |   47 +
 tools/testing/radix-tree/test.h                    |   18 +
 39 files changed, 1884 insertions(+), 1747 deletions(-)
 create mode 100644 tools/testing/radix-tree/benchmark.c
 delete mode 100644 tools/testing/radix-tree/find_next_bit.c
 create mode 100644 tools/testing/radix-tree/idr.c
 create mode 100644 tools/testing/radix-tree/linux/idr.h

-- 
2.10.2

[PATCH 01/29] tools: Add WARN_ON_ONCE

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

The radix tree uses its own buggy WARN_ON_ONCE.  Replace it with the
definition from asm-generic/bug.h

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 tools/include/asm/bug.h                | 11 +++++++++++
 tools/testing/radix-tree/Makefile      |  2 +-
 tools/testing/radix-tree/linux/bug.h   |  2 +-
 tools/testing/radix-tree/linux/types.h |  2 --
 4 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/tools/include/asm/bug.h b/tools/include/asm/bug.h
index 9e5f484..beda1a8 100644
--- a/tools/include/asm/bug.h
+++ b/tools/include/asm/bug.h
@@ -12,6 +12,17 @@
 	unlikely(__ret_warn_on);		\
 })
 
+#define WARN_ON_ONCE(condition) ({			\
+	static int __warned;				\
+	int __ret_warn_once = !!(condition);		\
+							\
+	if (unlikely(__ret_warn_once && !__warned)) {	\
+		__warned = true;			\
+		WARN_ON(1);				\
+	}						\
+	unlikely(__ret_warn_once);			\
+})
+
 #define WARN_ONCE(condition, format...)	({	\
 	static int __warned;			\
 	int __ret_warn_once = !!(condition);	\
diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
index f2e07f2..3c338dc 100644
--- a/tools/testing/radix-tree/Makefile
+++ b/tools/testing/radix-tree/Makefile
@@ -1,5 +1,5 @@
 
-CFLAGS += -I. -g -O2 -Wall -D_LGPL_SOURCE
+CFLAGS += -I. -I../../include -g -O2 -Wall -D_LGPL_SOURCE
 LDFLAGS += -lpthread -lurcu
 TARGETS = main
 OFILES = main.o radix-tree.o linux.o test.o tag_check.o find_next_bit.o \
diff --git a/tools/testing/radix-tree/linux/bug.h b/tools/testing/radix-tree/linux/bug.h
index ccbe444..23b8ed5 100644
--- a/tools/testing/radix-tree/linux/bug.h
+++ b/tools/testing/radix-tree/linux/bug.h
@@ -1 +1 @@
-#define WARN_ON_ONCE(x)		assert(x)
+#include "asm/bug.h"
diff --git a/tools/testing/radix-tree/linux/types.h b/tools/testing/radix-tree/linux/types.h
index faa0b6f..8491d89 100644
--- a/tools/testing/radix-tree/linux/types.h
+++ b/tools/testing/radix-tree/linux/types.h
@@ -6,8 +6,6 @@
 #define __rcu
 #define __read_mostly
 
-#define BITS_PER_LONG (sizeof(long) * 8)
-
 static inline void INIT_LIST_HEAD(struct list_head *list)
 {
 	list->next = list;
-- 
2.10.2

[PATCH 02/29] radix tree test suite: Allow GFP_ATOMIC allocations to fail

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

In order to test the preload code, it is necessary to fail GFP_ATOMIC
allocations, which requires defining GFP_KERNEL and GFP_ATOMIC properly.
Remove the obsolete __GFP_WAIT and copy the definitions of the __GFP
flags which are used from the kernel include files.  We also need the
real definition of gfpflags_allow_blocking() to persuade the radix tree
to actually use its preallocated nodes.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 tools/testing/radix-tree/linux.c      |  7 ++++++-
 tools/testing/radix-tree/linux/gfp.h  | 22 +++++++++++++++++++---
 tools/testing/radix-tree/linux/slab.h |  5 -----
 3 files changed, 25 insertions(+), 9 deletions(-)

diff --git a/tools/testing/radix-tree/linux.c b/tools/testing/radix-tree/linux.c
index 1548237..3cfb04e 100644
--- a/tools/testing/radix-tree/linux.c
+++ b/tools/testing/radix-tree/linux.c
@@ -33,7 +33,12 @@ mempool_t *mempool_create(int min_nr, mempool_alloc_t *alloc_fn,
 
 void *kmem_cache_alloc(struct kmem_cache *cachep, int flags)
 {
-	void *ret = malloc(cachep->size);
+	void *ret;
+
+	if (flags & __GFP_NOWARN)
+		return NULL;
+
+	ret = malloc(cachep->size);
 	if (cachep->ctor)
 		cachep->ctor(ret);
 	uatomic_inc(&nr_allocated);
diff --git a/tools/testing/radix-tree/linux/gfp.h b/tools/testing/radix-tree/linux/gfp.h
index 5201b91..5b09b2c 100644
--- a/tools/testing/radix-tree/linux/gfp.h
+++ b/tools/testing/radix-tree/linux/gfp.h
@@ -3,8 +3,24 @@
 
 #define __GFP_BITS_SHIFT 26
 #define __GFP_BITS_MASK ((gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
-#define __GFP_WAIT 1
-#define __GFP_ACCOUNT 0
-#define __GFP_NOWARN 0
+
+#define __GFP_HIGH		0x20u
+#define __GFP_IO		0x40u
+#define __GFP_FS		0x80u
+#define __GFP_NOWARN		0x200u
+#define __GFP_ATOMIC		0x80000u
+#define __GFP_ACCOUNT		0x100000u
+#define __GFP_DIRECT_RECLAIM	0x400000u
+#define __GFP_KSWAPD_RECLAIM	0x2000000u
+
+#define __GFP_RECLAIM		(__GFP_DIRECT_RECLAIM|__GFP_KSWAPD_RECLAIM)
+
+#define GFP_ATOMIC		(__GFP_HIGH|__GFP_ATOMIC|__GFP_KSWAPD_RECLAIM)
+#define GFP_KERNEL		(__GFP_RECLAIM | __GFP_IO | __GFP_FS)
+
+static inline bool gfpflags_allow_blocking(const gfp_t gfp_flags)
+{
+	return !!(gfp_flags & __GFP_DIRECT_RECLAIM);
+}
 
 #endif
diff --git a/tools/testing/radix-tree/linux/slab.h b/tools/testing/radix-tree/linux/slab.h
index 6d5a347..452e2bf 100644
--- a/tools/testing/radix-tree/linux/slab.h
+++ b/tools/testing/radix-tree/linux/slab.h
@@ -7,11 +7,6 @@
 #define SLAB_PANIC 2
 #define SLAB_RECLAIM_ACCOUNT    0x00020000UL            /* Objects are reclaimable */
 
-static inline int gfpflags_allow_blocking(gfp_t mask)
-{
-	return 1;
-}
-
 struct kmem_cache {
 	int size;
 	void (*ctor)(void *);
-- 
2.10.2

[PATCH 03/29] radix tree test suite: Track preempt_count

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

Rather than simply NOP out preempt_enable() and preempt_disable(),
keep track of preempt_count and display it regularly in case either
the test suite or the code under test is forgetting to balance the
enables & disables.  Only found a test-case that was forgetting to
re-enable preemption, but it's a possibility worth checking.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 tools/testing/radix-tree/linux.c         |  1 +
 tools/testing/radix-tree/linux/preempt.h |  6 +++---
 tools/testing/radix-tree/main.c          | 30 ++++++++++++++++++++----------
 3 files changed, 24 insertions(+), 13 deletions(-)

diff --git a/tools/testing/radix-tree/linux.c b/tools/testing/radix-tree/linux.c
index 3cfb04e..1f32a16 100644
--- a/tools/testing/radix-tree/linux.c
+++ b/tools/testing/radix-tree/linux.c
@@ -9,6 +9,7 @@
 #include <urcu/uatomic.h>
 
 int nr_allocated;
+int preempt_count;
 
 void *mempool_alloc(mempool_t *pool, int gfp_mask)
 {
diff --git a/tools/testing/radix-tree/linux/preempt.h b/tools/testing/radix-tree/linux/preempt.h
index 6210672..65c04c2 100644
--- a/tools/testing/radix-tree/linux/preempt.h
+++ b/tools/testing/radix-tree/linux/preempt.h
@@ -1,4 +1,4 @@
-/* */
+extern int preempt_count;
 
-#define preempt_disable() do { } while (0)
-#define preempt_enable() do { } while (0)
+#define preempt_disable()	uatomic_inc(&preempt_count)
+#define preempt_enable()	uatomic_dec(&preempt_count)
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index daa9010..64ffe67 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -293,27 +293,36 @@ static void single_thread_tests(bool long_run)
 {
 	int i;
 
-	printf("starting single_thread_tests: %d allocated\n", nr_allocated);
+	printf("starting single_thread_tests: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	multiorder_checks();
-	printf("after multiorder_check: %d allocated\n", nr_allocated);
+	printf("after multiorder_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	locate_check();
-	printf("after locate_check: %d allocated\n", nr_allocated);
+	printf("after locate_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	tag_check();
-	printf("after tag_check: %d allocated\n", nr_allocated);
+	printf("after tag_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	gang_check();
-	printf("after gang_check: %d allocated\n", nr_allocated);
+	printf("after gang_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	add_and_check();
-	printf("after add_and_check: %d allocated\n", nr_allocated);
+	printf("after add_and_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	dynamic_height_check();
-	printf("after dynamic_height_check: %d allocated\n", nr_allocated);
+	printf("after dynamic_height_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	big_gang_check(long_run);
-	printf("after big_gang_check: %d allocated\n", nr_allocated);
+	printf("after big_gang_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	for (i = 0; i < (long_run ? 2000 : 3); i++) {
 		copy_tag_check();
 		printf("%d ", i);
 		fflush(stdout);
 	}
-	printf("after copy_tag_check: %d allocated\n", nr_allocated);
+	printf("after copy_tag_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 }
 
 int main(int argc, char **argv)
@@ -336,7 +345,8 @@ int main(int argc, char **argv)
 	single_thread_tests(long_run);
 
 	sleep(1);
-	printf("after sleep(1): %d allocated\n", nr_allocated);
+	printf("after sleep(1): %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	rcu_unregister_thread();
 
 	exit(0);
-- 
2.10.2

[PATCH 04/29] radix tree test suite: Free preallocated nodes

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

It can be a source of mild concern when the test suite shows that we're
leaking nodes.  While poring over the source code looking for leaks
can lead to some fascinating bugs being discovered, sometimes the leak
is simply that these nodes were preallocated and are sitting on the
per-CPU list.  Free them by faking a CPU_DEAD event.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 tools/testing/radix-tree/main.c | 3 +++
 tools/testing/radix-tree/test.h | 4 ++++
 2 files changed, 7 insertions(+)

diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 64ffe67..2930560 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -344,6 +344,9 @@ int main(int argc, char **argv)
 	iteration_test();
 	single_thread_tests(long_run);
 
+	/* Free any remaining preallocated nodes */
+	radix_tree_callback(NULL, CPU_DEAD, NULL);
+
 	sleep(1);
 	printf("after sleep(1): %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index 217fb24..8cd666a 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -2,6 +2,8 @@
 #include <linux/types.h>
 #include <linux/radix-tree.h>
 #include <linux/rcupdate.h>
+#include <linux/notifier.h>
+#include <linux/cpu.h>
 
 struct item {
 	unsigned long index;
@@ -44,3 +46,5 @@ void radix_tree_dump(struct radix_tree_root *root);
 int root_tag_get(struct radix_tree_root *root, unsigned int tag);
 unsigned long node_maxindex(struct radix_tree_node *);
 unsigned long shift_maxindex(unsigned int shift);
+int radix_tree_callback(struct notifier_block *nfb,
+			unsigned long action, void *hcpu);
-- 
2.10.2

[PATCH 05/29] radix tree test suite: Make runs more reproducible

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

Instead of reseeding the random number generator every time around the
loop in big_gang_check(), seed it at the beginning of execution.  Use
rand_r() and an independent base seed for each thread in iteration_test()
so they don't stomp all over each others state.  Since this particular
test depends on the kernel scheduler, the iteration test can't be
reproduced based purely on the random seed, but at least it won't pollute
the other tests.

Print the seed, and allow the seed to be specified so that a run which
hits a problem can be reproduced.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 tools/testing/radix-tree/iteration_check.c | 11 +++++++----
 tools/testing/radix-tree/main.c            |  9 +++++++--
 2 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/tools/testing/radix-tree/iteration_check.c b/tools/testing/radix-tree/iteration_check.c
index 9adb8e7..11d570c 100644
--- a/tools/testing/radix-tree/iteration_check.c
+++ b/tools/testing/radix-tree/iteration_check.c
@@ -20,6 +20,7 @@
 #define TAG 0
 static pthread_mutex_t tree_lock = PTHREAD_MUTEX_INITIALIZER;
 static pthread_t threads[NUM_THREADS];
+static unsigned int seeds[3];
 RADIX_TREE(tree, GFP_KERNEL);
 bool test_complete;
 
@@ -71,7 +72,7 @@ static void *tagged_iteration_fn(void *arg)
 				continue;
 			}
 
-			if (rand() % 50 == 0)
+			if (rand_r(&seeds[0]) % 50 == 0)
 				slot = radix_tree_iter_next(&iter);
 		}
 		rcu_read_unlock();
@@ -111,7 +112,7 @@ static void *untagged_iteration_fn(void *arg)
 				continue;
 			}
 
-			if (rand() % 50 == 0)
+			if (rand_r(&seeds[1]) % 50 == 0)
 				slot = radix_tree_iter_next(&iter);
 		}
 		rcu_read_unlock();
@@ -129,7 +130,7 @@ static void *remove_entries_fn(void *arg)
 	while (!test_complete) {
 		int pgoff;
 
-		pgoff = rand() % 100;
+		pgoff = rand_r(&seeds[2]) % 100;
 
 		pthread_mutex_lock(&tree_lock);
 		item_delete(&tree, pgoff);
@@ -146,9 +147,11 @@ void iteration_test(void)
 
 	printf("Running iteration tests for 10 seconds\n");
 
-	srand(time(0));
 	test_complete = false;
 
+	for (i = 0; i < 3; i++)
+		seeds[i] = rand();
+
 	if (pthread_create(&threads[0], NULL, tagged_iteration_fn, NULL)) {
 		perror("pthread_create");
 		exit(1);
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 2930560..5ddec8e 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -67,7 +67,6 @@ void big_gang_check(bool long_run)
 
 	for (i = 0; i < (long_run ? 1000 : 3); i++) {
 		__big_gang_check();
-		srand(time(0));
 		printf("%d ", i);
 		fflush(stdout);
 	}
@@ -329,12 +328,18 @@ int main(int argc, char **argv)
 {
 	bool long_run = false;
 	int opt;
+	unsigned int seed = time(NULL);
 
-	while ((opt = getopt(argc, argv, "l")) != -1) {
+	while ((opt = getopt(argc, argv, "ls:")) != -1) {
 		if (opt == 'l')
 			long_run = true;
+		else if (opt == 's')
+			seed = strtoul(optarg, NULL, 0);
 	}
 
+	printf("random seed %u\n", seed);
+	srand(seed);
+
 	rcu_register_thread();
 	radix_tree_init();
 
-- 
2.10.2

[PATCH 06/29] radix tree test suite: benchmark for iterator

[Matthew Wilcox]

From: Konstantin Khlebnikov <koct9i@gmail.com>

This adds simple benchmark for iterator similar to one I've used for
commit 78c1d78 ("radix-tree: introduce bit-optimized iterator")

Building with make BENCHMARK=1 set radix tree order to 6, this allows
to get performance comparable to in kernel performance.

Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com>
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 tools/testing/radix-tree/Makefile       |  6 +-
 tools/testing/radix-tree/benchmark.c    | 98 +++++++++++++++++++++++++++++++++
 tools/testing/radix-tree/linux/kernel.h |  4 ++
 tools/testing/radix-tree/main.c         |  2 +
 tools/testing/radix-tree/test.h         |  1 +
 5 files changed, 110 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/radix-tree/benchmark.c

diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
index 3c338dc..08283a8 100644
--- a/tools/testing/radix-tree/Makefile
+++ b/tools/testing/radix-tree/Makefile
@@ -4,7 +4,11 @@ LDFLAGS += -lpthread -lurcu
 TARGETS = main
 OFILES = main.o radix-tree.o linux.o test.o tag_check.o find_next_bit.o \
 	 regression1.o regression2.o regression3.o multiorder.o \
-	 iteration_check.o
+	 iteration_check.o benchmark.o
+
+ifdef BENCHMARK
+	CFLAGS += -DBENCHMARK=1
+endif
 
 targets: $(TARGETS)
 
diff --git a/tools/testing/radix-tree/benchmark.c b/tools/testing/radix-tree/benchmark.c
new file mode 100644
index 0000000..215ca86
--- /dev/null
+++ b/tools/testing/radix-tree/benchmark.c
@@ -0,0 +1,98 @@
+/*
+ * benchmark.c:
+ * Author: Konstantin Khlebnikov <koct9i@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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.
+ */
+#include <linux/radix-tree.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <time.h>
+#include "test.h"
+
+#define NSEC_PER_SEC	1000000000L
+
+static long long benchmark_iter(struct radix_tree_root *root, bool tagged)
+{
+	volatile unsigned long sink = 0;
+	struct radix_tree_iter iter;
+	struct timespec start, finish;
+	long long nsec;
+	int l, loops = 1;
+	void **slot;
+
+#ifdef BENCHMARK
+again:
+#endif
+	clock_gettime(CLOCK_MONOTONIC, &start);
+	for (l = 0; l < loops; l++) {
+		if (tagged) {
+			radix_tree_for_each_tagged(slot, root, &iter, 0, 0)
+				sink ^= (unsigned long)slot;
+		} else {
+			radix_tree_for_each_slot(slot, root, &iter, 0)
+				sink ^= (unsigned long)slot;
+		}
+	}
+	clock_gettime(CLOCK_MONOTONIC, &finish);
+
+	nsec = (finish.tv_sec - start.tv_sec) * NSEC_PER_SEC +
+	       (finish.tv_nsec - start.tv_nsec);
+
+#ifdef BENCHMARK
+	if (loops == 1 && nsec * 5 < NSEC_PER_SEC) {
+		loops = NSEC_PER_SEC / nsec / 4 + 1;
+		goto again;
+	}
+#endif
+
+	nsec /= loops;
+	return nsec;
+}
+
+static void benchmark_size(unsigned long size, unsigned long step, int order)
+{
+	RADIX_TREE(tree, GFP_KERNEL);
+	long long normal, tagged;
+	unsigned long index;
+
+	for (index = 0 ; index < size ; index += step) {
+		item_insert_order(&tree, index, order);
+		radix_tree_tag_set(&tree, index, 0);
+	}
+
+	tagged = benchmark_iter(&tree, true);
+	normal = benchmark_iter(&tree, false);
+
+	printf("Size %ld, step %6ld, order %d tagged %10lld ns, normal %10lld ns\n",
+		size, step, order, tagged, normal);
+
+	item_kill_tree(&tree);
+	rcu_barrier();
+}
+
+void benchmark(void)
+{
+	unsigned long size[] = {1 << 10, 1 << 20, 0};
+	unsigned long step[] = {1, 2, 7, 15, 63, 64, 65,
+				128, 256, 512, 12345, 0};
+	int c, s;
+
+	printf("starting benchmarks\n");
+	printf("RADIX_TREE_MAP_SHIFT = %d\n", RADIX_TREE_MAP_SHIFT);
+
+	for (c = 0; size[c]; c++)
+		for (s = 0; step[s]; s++)
+			benchmark_size(size[c], step[s], 0);
+
+	for (c = 0; size[c]; c++)
+		for (s = 0; step[s]; s++)
+			benchmark_size(size[c], step[s] << 9, 9);
+}
diff --git a/tools/testing/radix-tree/linux/kernel.h b/tools/testing/radix-tree/linux/kernel.h
index be98a47..dbe4b92 100644
--- a/tools/testing/radix-tree/linux/kernel.h
+++ b/tools/testing/radix-tree/linux/kernel.h
@@ -10,7 +10,11 @@
 #include "../../include/linux/compiler.h"
 #include "../../../include/linux/kconfig.h"
 
+#ifdef BENCHMARK
+#define RADIX_TREE_MAP_SHIFT	6
+#else
 #define RADIX_TREE_MAP_SHIFT	3
+#endif
 
 #ifndef NULL
 #define NULL	0
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 5ddec8e..f43706c 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -352,6 +352,8 @@ int main(int argc, char **argv)
 	/* Free any remaining preallocated nodes */
 	radix_tree_callback(NULL, CPU_DEAD, NULL);
 
+	benchmark();
+
 	sleep(1);
 	printf("after sleep(1): %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index 8cd666a..f2dc35f 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -30,6 +30,7 @@ void item_kill_tree(struct radix_tree_root *root);
 void tag_check(void);
 void multiorder_checks(void);
 void iteration_test(void);
+void benchmark(void);
 
 struct item *
 item_tag_set(struct radix_tree_root *root, unsigned long index, int tag);
-- 
2.10.2

[PATCH 07/29] radix tree test suite: Use rcu_barrier

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

Calling rcu_barrier() allows all of the rcu-freed memory to be actually
returned to the pool, and allows nr_allocated to return to 0.  As well
as allowing diffs between runs to be more useful, it also lets us
pinpoint leaks more effectively.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 tools/testing/radix-tree/main.c      | 12 ++++++++++--
 tools/testing/radix-tree/tag_check.c |  5 +++++
 2 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index f43706c..8621542 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -295,24 +295,31 @@ static void single_thread_tests(bool long_run)
 	printf("starting single_thread_tests: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	multiorder_checks();
+	rcu_barrier();
 	printf("after multiorder_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	locate_check();
+	rcu_barrier();
 	printf("after locate_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	tag_check();
+	rcu_barrier();
 	printf("after tag_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	gang_check();
+	rcu_barrier();
 	printf("after gang_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	add_and_check();
+	rcu_barrier();
 	printf("after add_and_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	dynamic_height_check();
+	rcu_barrier();
 	printf("after dynamic_height_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	big_gang_check(long_run);
+	rcu_barrier();
 	printf("after big_gang_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	for (i = 0; i < (long_run ? 2000 : 3); i++) {
@@ -320,6 +327,7 @@ static void single_thread_tests(bool long_run)
 		printf("%d ", i);
 		fflush(stdout);
 	}
+	rcu_barrier();
 	printf("after copy_tag_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 }
@@ -354,8 +362,8 @@ int main(int argc, char **argv)
 
 	benchmark();
 
-	sleep(1);
-	printf("after sleep(1): %d allocated, preempt %d\n",
+	rcu_barrier();
+	printf("after rcu_barrier: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	rcu_unregister_thread();
 
diff --git a/tools/testing/radix-tree/tag_check.c b/tools/testing/radix-tree/tag_check.c
index b0ac057..186f6e4 100644
--- a/tools/testing/radix-tree/tag_check.c
+++ b/tools/testing/radix-tree/tag_check.c
@@ -51,6 +51,7 @@ void simple_checks(void)
 	verify_tag_consistency(&tree, 1);
 	printf("before item_kill_tree: %d allocated\n", nr_allocated);
 	item_kill_tree(&tree);
+	rcu_barrier();
 	printf("after item_kill_tree: %d allocated\n", nr_allocated);
 }
 
@@ -331,12 +332,16 @@ void tag_check(void)
 	single_check();
 	extend_checks();
 	contract_checks();
+	rcu_barrier();
 	printf("after extend_checks: %d allocated\n", nr_allocated);
 	__leak_check();
 	leak_check();
+	rcu_barrier();
 	printf("after leak_check: %d allocated\n", nr_allocated);
 	simple_checks();
+	rcu_barrier();
 	printf("after simple_checks: %d allocated\n", nr_allocated);
 	thrash_tags();
+	rcu_barrier();
 	printf("after thrash_tags: %d allocated\n", nr_allocated);
 }
-- 
2.10.2

[PATCH 08/29] tools: Add more bitmap functions

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

I need the following functions for the radix tree:

bitmap_fill
find_next_zero_bit
bitmap_empty
bitmap_full

Copy the implementations from include/linux/bitmap.h and lib/find_bit.c
---
 tools/include/linux/bitmap.h | 26 ++++++++++++++++++++++++++
 tools/lib/find_bit.c         |  8 ++++++++
 2 files changed, 34 insertions(+)

diff --git a/tools/include/linux/bitmap.h b/tools/include/linux/bitmap.h
index 43c1c50..eef41d5 100644
--- a/tools/include/linux/bitmap.h
+++ b/tools/include/linux/bitmap.h
@@ -35,6 +35,32 @@ static inline void bitmap_zero(unsigned long *dst, int nbits)
 	}
 }
 
+static inline void bitmap_fill(unsigned long *dst, unsigned int nbits)
+{
+	unsigned int nlongs = BITS_TO_LONGS(nbits);
+	if (!small_const_nbits(nbits)) {
+		unsigned int len = (nlongs - 1) * sizeof(unsigned long);
+		memset(dst, 0xff,  len);
+	}
+	dst[nlongs - 1] = BITMAP_LAST_WORD_MASK(nbits);
+}
+
+static inline int bitmap_empty(const unsigned long *src, unsigned nbits)
+{
+	if (small_const_nbits(nbits))
+		return ! (*src & BITMAP_LAST_WORD_MASK(nbits));
+
+	return find_first_bit(src, nbits) == nbits;
+}
+
+static inline int bitmap_full(const unsigned long *src, unsigned int nbits)
+{
+	if (small_const_nbits(nbits))
+		return ! (~(*src) & BITMAP_LAST_WORD_MASK(nbits));
+
+	return find_first_zero_bit(src, nbits) == nbits;
+}
+
 static inline int bitmap_weight(const unsigned long *src, int nbits)
 {
 	if (small_const_nbits(nbits))
diff --git a/tools/lib/find_bit.c b/tools/lib/find_bit.c
index 9122a9e..de26b6f 100644
--- a/tools/lib/find_bit.c
+++ b/tools/lib/find_bit.c
@@ -66,6 +66,14 @@ unsigned long find_next_bit(const unsigned long *addr, unsigned long size,
 }
 #endif
 
+#ifndef find_next_zero_bit
+unsigned long find_next_zero_bit(const unsigned long *addr, unsigned long size,
+				unsigned long offset)
+{
+	return _find_next_bit(addr, size, offset, ~0UL);
+}
+#endif
+
 #ifndef find_first_bit
 /*
  * Find the first set bit in a memory region.
-- 
2.10.2

[PATCH 09/29] radix tree test suite: Use common find-bit code

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Remove the old find_next_bit code in favour of linking in the find_bit
code from tools/lib.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 tools/testing/radix-tree/Makefile                  |  7 ++-
 tools/testing/radix-tree/find_next_bit.c           | 57 ----------------------
 tools/testing/radix-tree/linux/bitops.h            | 40 +++++++++------
 tools/testing/radix-tree/linux/bitops/non-atomic.h | 13 +++--
 tools/testing/radix-tree/linux/kernel.h            | 11 +++++
 5 files changed, 48 insertions(+), 80 deletions(-)
 delete mode 100644 tools/testing/radix-tree/find_next_bit.c

diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
index 08283a8..3635e4d 100644
--- a/tools/testing/radix-tree/Makefile
+++ b/tools/testing/radix-tree/Makefile
@@ -18,7 +18,12 @@ main:	$(OFILES)
 clean:
 	$(RM) -f $(TARGETS) *.o radix-tree.c
 
-$(OFILES): *.h */*.h ../../../include/linux/radix-tree.h ../../include/linux/*.h
+find_next_bit.o: ../../lib/find_bit.c
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(OFILES): *.h */*.h \
+	../../include/linux/*.h \
+	../../../include/linux/radix-tree.h
 
 radix-tree.c: ../../../lib/radix-tree.c
 	sed -e 's/^static //' -e 's/__always_inline //' -e 's/inline //' < $< > $@
diff --git a/tools/testing/radix-tree/find_next_bit.c b/tools/testing/radix-tree/find_next_bit.c
deleted file mode 100644
index d1c2178..0000000
--- a/tools/testing/radix-tree/find_next_bit.c
+++ /dev/null
@@ -1,57 +0,0 @@
-/* find_next_bit.c: fallback find next bit implementation
- *
- * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * 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.
- */
-
-#include <linux/types.h>
-#include <linux/bitops.h>
-
-#define BITOP_WORD(nr)		((nr) / BITS_PER_LONG)
-
-/*
- * Find the next set bit in a memory region.
- */
-unsigned long find_next_bit(const unsigned long *addr, unsigned long size,
-			    unsigned long offset)
-{
-	const unsigned long *p = addr + BITOP_WORD(offset);
-	unsigned long result = offset & ~(BITS_PER_LONG-1);
-	unsigned long tmp;
-
-	if (offset >= size)
-		return size;
-	size -= result;
-	offset %= BITS_PER_LONG;
-	if (offset) {
-		tmp = *(p++);
-		tmp &= (~0UL << offset);
-		if (size < BITS_PER_LONG)
-			goto found_first;
-		if (tmp)
-			goto found_middle;
-		size -= BITS_PER_LONG;
-		result += BITS_PER_LONG;
-	}
-	while (size & ~(BITS_PER_LONG-1)) {
-		if ((tmp = *(p++)))
-			goto found_middle;
-		result += BITS_PER_LONG;
-		size -= BITS_PER_LONG;
-	}
-	if (!size)
-		return result;
-	tmp = *p;
-
-found_first:
-	tmp &= (~0UL >> (BITS_PER_LONG - size));
-	if (tmp == 0UL)		/* Are any bits set? */
-		return result + size;	/* Nope. */
-found_middle:
-	return result + __ffs(tmp);
-}
diff --git a/tools/testing/radix-tree/linux/bitops.h b/tools/testing/radix-tree/linux/bitops.h
index 71d58427..a13e9bc 100644
--- a/tools/testing/radix-tree/linux/bitops.h
+++ b/tools/testing/radix-tree/linux/bitops.h
@@ -2,9 +2,14 @@
 #define _ASM_GENERIC_BITOPS_NON_ATOMIC_H_
 
 #include <linux/types.h>
+#include <linux/bitops/find.h>
+#include <linux/bitops/hweight.h>
+#include <linux/kernel.h>
 
-#define BITOP_MASK(nr)		(1UL << ((nr) % BITS_PER_LONG))
-#define BITOP_WORD(nr)		((nr) / BITS_PER_LONG)
+#define BIT_MASK(nr)		(1UL << ((nr) % BITS_PER_LONG))
+#define BIT_WORD(nr)		((nr) / BITS_PER_LONG)
+#define BITS_PER_BYTE		8
+#define BITS_TO_LONGS(nr)	DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
 
 /**
  * __set_bit - Set a bit in memory
@@ -17,16 +22,16 @@
  */
 static inline void __set_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 
 	*p  |= mask;
 }
 
 static inline void __clear_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 
 	*p &= ~mask;
 }
@@ -42,8 +47,8 @@ static inline void __clear_bit(int nr, volatile unsigned long *addr)
  */
 static inline void __change_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 
 	*p ^= mask;
 }
@@ -59,8 +64,8 @@ static inline void __change_bit(int nr, volatile unsigned long *addr)
  */
 static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 	unsigned long old = *p;
 
 	*p = old | mask;
@@ -78,8 +83,8 @@ static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
  */
 static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 	unsigned long old = *p;
 
 	*p = old & ~mask;
@@ -90,8 +95,8 @@ static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
 static inline int __test_and_change_bit(int nr,
 					    volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 	unsigned long old = *p;
 
 	*p = old ^ mask;
@@ -105,7 +110,7 @@ static inline int __test_and_change_bit(int nr,
  */
 static inline int test_bit(int nr, const volatile unsigned long *addr)
 {
-	return 1UL & (addr[BITOP_WORD(nr)] >> (nr & (BITS_PER_LONG-1)));
+	return 1UL & (addr[BIT_WORD(nr)] >> (nr & (BITS_PER_LONG-1)));
 }
 
 /**
@@ -147,4 +152,9 @@ unsigned long find_next_bit(const unsigned long *addr,
 			    unsigned long size,
 			    unsigned long offset);
 
+static inline unsigned long hweight_long(unsigned long w)
+{
+	return sizeof(w) == 4 ? hweight32(w) : hweight64(w);
+}
+
 #endif /* _ASM_GENERIC_BITOPS_NON_ATOMIC_H_ */
diff --git a/tools/testing/radix-tree/linux/bitops/non-atomic.h b/tools/testing/radix-tree/linux/bitops/non-atomic.h
index 46a825c..6a1bcb9 100644
--- a/tools/testing/radix-tree/linux/bitops/non-atomic.h
+++ b/tools/testing/radix-tree/linux/bitops/non-atomic.h
@@ -3,7 +3,6 @@
 
 #include <asm/types.h>
 
-#define BITOP_MASK(nr)		(1UL << ((nr) % BITS_PER_LONG))
 #define BITOP_WORD(nr)		((nr) / BITS_PER_LONG)
 
 /**
@@ -17,7 +16,7 @@
  */
 static inline void __set_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 
 	*p  |= mask;
@@ -25,7 +24,7 @@ static inline void __set_bit(int nr, volatile unsigned long *addr)
 
 static inline void __clear_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 
 	*p &= ~mask;
@@ -42,7 +41,7 @@ static inline void __clear_bit(int nr, volatile unsigned long *addr)
  */
 static inline void __change_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 
 	*p ^= mask;
@@ -59,7 +58,7 @@ static inline void __change_bit(int nr, volatile unsigned long *addr)
  */
 static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 	unsigned long old = *p;
 
@@ -78,7 +77,7 @@ static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
  */
 static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 	unsigned long old = *p;
 
@@ -90,7 +89,7 @@ static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
 static inline int __test_and_change_bit(int nr,
 					    volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 	unsigned long old = *p;
 
diff --git a/tools/testing/radix-tree/linux/kernel.h b/tools/testing/radix-tree/linux/kernel.h
index dbe4b92..23e77f5 100644
--- a/tools/testing/radix-tree/linux/kernel.h
+++ b/tools/testing/radix-tree/linux/kernel.h
@@ -47,4 +47,15 @@ static inline int in_interrupt(void)
 {
 	return 0;
 }
+
+/*
+ * This looks more complex than it should be. But we need to
+ * get the type for the ~ right in round_down (it needs to be
+ * as wide as the result!), and we want to evaluate the macro
+ * arguments just once each.
+ */
+#define __round_mask(x, y) ((__typeof__(x))((y)-1))
+#define round_up(x, y) ((((x)-1) | __round_mask(x, y))+1)
+#define round_down(x, y) ((x) & ~__round_mask(x, y))
+
 #endif /* _KERNEL_H */
-- 
2.10.2

[PATCH 10/29] radix-tree: Add radix_tree_join

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

This new function allows for the replacement of many smaller entries in
the radix tree with one larger multiorder entry.  From the point of view
of an RCU walker, they may see a mixture of the smaller entries and the
large entry during the same walk, but they will never see NULL for an
index which was populated before the join.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 include/linux/radix-tree.h            |   2 +
 lib/radix-tree.c                      | 159 +++++++++++++++++++++++++++-------
 tools/testing/radix-tree/multiorder.c |  32 +++++++
 3 files changed, 163 insertions(+), 30 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index af3581b..1efd81f 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -319,6 +319,8 @@ static inline void radix_tree_preload_end(void)
 	preempt_enable();
 }
 
+int radix_tree_join(struct radix_tree_root *, unsigned long index,
+			unsigned new_order, void *);
 /**
  * struct radix_tree_iter - radix tree iterator state
  *
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 8e6d552..6a76252 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -314,18 +314,14 @@ static void radix_tree_node_rcu_free(struct rcu_head *head)
 {
 	struct radix_tree_node *node =
 			container_of(head, struct radix_tree_node, rcu_head);
-	int i;
 
 	/*
-	 * must only free zeroed nodes into the slab. radix_tree_shrink
-	 * can leave us with a non-NULL entry in the first slot, so clear
-	 * that here to make sure.
+	 * Must only free zeroed nodes into the slab.  We can be left with
+	 * non-NULL entries by radix_tree_free_nodes, so clear the entries
+	 * and tags here.
 	 */
-	for (i = 0; i < RADIX_TREE_MAX_TAGS; i++)
-		tag_clear(node, i, 0);
-
-	node->slots[0] = NULL;
-	node->count = 0;
+	memset(node->slots, 0, sizeof(node->slots));
+	memset(node->tags, 0, sizeof(node->tags));
 
 	kmem_cache_free(radix_tree_node_cachep, node);
 }
@@ -563,14 +559,14 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 	shift = radix_tree_load_root(root, &child, &maxindex);
 
 	/* Make sure the tree is high enough.  */
+	if (order > 0 && max == ((1UL << order) - 1))
+		max++;
 	if (max > maxindex) {
 		int error = radix_tree_extend(root, max, shift);
 		if (error < 0)
 			return error;
 		shift = error;
 		child = root->rnode;
-		if (order == shift)
-			shift += RADIX_TREE_MAP_SHIFT;
 	}
 
 	while (shift > order) {
@@ -582,6 +578,7 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 				return -ENOMEM;
 			child->shift = shift;
 			child->offset = offset;
+			child->count = 0;
 			child->parent = node;
 			rcu_assign_pointer(*slot, node_to_entry(child));
 			if (node)
@@ -595,31 +592,113 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 		slot = &node->slots[offset];
 	}
 
+	if (nodep)
+		*nodep = node;
+	if (slotp)
+		*slotp = slot;
+	return 0;
+}
+
 #ifdef CONFIG_RADIX_TREE_MULTIORDER
-	/* Insert pointers to the canonical entry */
-	if (order > shift) {
-		unsigned i, n = 1 << (order - shift);
+/*
+ * Free any nodes below this node.  The tree is presumed to not need
+ * shrinking, and any user data in the tree is presumed to not need a
+ * destructor called on it.  If we need to add a destructor, we can
+ * add that functionality later.  Note that we may not clear tags or
+ * slots from the tree as an RCU walker may still have a pointer into
+ * this subtree.  We could replace the entries with RADIX_TREE_RETRY,
+ * but we'll still have to clear those in rcu_free.
+ */
+static void radix_tree_free_nodes(struct radix_tree_node *node)
+{
+	unsigned offset = 0;
+	struct radix_tree_node *child = entry_to_node(node);
+
+	for (;;) {
+		void *entry = child->slots[offset];
+		if (radix_tree_is_internal_node(entry) &&
+					!is_sibling_entry(child, entry)) {
+			child = entry_to_node(entry);
+			offset = 0;
+			continue;
+		}
+		offset++;
+		while (offset == RADIX_TREE_MAP_SIZE) {
+			struct radix_tree_node *old = child;
+			offset = child->offset + 1;
+			child = child->parent;
+			radix_tree_node_free(old);
+			if (old == entry_to_node(node))
+				return;
+		}
+	}
+}
+
+static inline int insert_entries(struct radix_tree_node *node, void **slot,
+				void *ptr, unsigned order, bool replace)
+{
+	struct radix_tree_node *child;
+	unsigned i, n, tag, offset, tags = 0;
+
+	if (node) {
+		n = 1 << (order - node->shift);
+		offset = get_slot_offset(node, slot);
+	} else {
+		n = 1;
+		offset = 0;
+	}
+
+	if (n > 1) {
 		offset = offset & ~(n - 1);
 		slot = &node->slots[offset];
-		child = node_to_entry(slot);
-		for (i = 0; i < n; i++) {
-			if (slot[i])
+	}
+	child = node_to_entry(slot);
+
+	for (i = 0; i < n; i++) {
+		if (slot[i]) {
+			if (replace) {
+				node->count--;
+				for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+					if (tag_get(node, tag, offset + i))
+						tags |= 1 << tag;
+			} else
 				return -EEXIST;
 		}
+	}
 
-		for (i = 1; i < n; i++) {
+	for (i = 0; i < n; i++) {
+		struct radix_tree_node *old = slot[i];
+		if (i) {
 			rcu_assign_pointer(slot[i], child);
-			node->count++;
+			for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+				if (tags & (1 << tag))
+					tag_clear(node, tag, offset + i);
+		} else {
+			rcu_assign_pointer(slot[i], ptr);
+			for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+				if (tags & (1 << tag))
+					tag_set(node, tag, offset);
 		}
+		if (radix_tree_is_internal_node(old) &&
+					!is_sibling_entry(node, old))
+			radix_tree_free_nodes(old);
 	}
-#endif
-
-	if (nodep)
-		*nodep = node;
-	if (slotp)
-		*slotp = slot;
-	return 0;
+	if (node)
+		node->count += n;
+	return n;
+}
+#else
+static inline int insert_entries(struct radix_tree_node *node, void **slot,
+				void *ptr, unsigned order, bool replace)
+{
+	if (*slot)
+		return -EEXIST;
+	rcu_assign_pointer(*slot, ptr);
+	if (node)
+		node->count++;
+	return 1;
 }
+#endif
 
 /**
  *	__radix_tree_insert    -    insert into a radix tree
@@ -642,13 +721,13 @@ int __radix_tree_insert(struct radix_tree_root *root, unsigned long index,
 	error = __radix_tree_create(root, index, order, &node, &slot);
 	if (error)
 		return error;
-	if (*slot != NULL)
-		return -EEXIST;
-	rcu_assign_pointer(*slot, item);
+
+	error = insert_entries(node, slot, item, order, false);
+	if (error < 0)
+		return error;
 
 	if (node) {
 		unsigned offset = get_slot_offset(node, slot);
-		node->count++;
 		BUG_ON(tag_get(node, 0, offset));
 		BUG_ON(tag_get(node, 1, offset));
 		BUG_ON(tag_get(node, 2, offset));
@@ -746,6 +825,26 @@ void *radix_tree_lookup(struct radix_tree_root *root, unsigned long index)
 }
 EXPORT_SYMBOL(radix_tree_lookup);
 
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+int radix_tree_join(struct radix_tree_root *root, unsigned long index,
+			unsigned order, void *item)
+{
+	struct radix_tree_node *node;
+	void **slot;
+	int error;
+
+	BUG_ON(radix_tree_is_internal_node(item));
+
+	error = __radix_tree_create(root, index, order, &node, &slot);
+	if (!error)
+		error = insert_entries(node, slot, item, order, true);
+	if (error > 0)
+		error = 0;
+
+	return error;
+}
+#endif
+
 /**
  *	radix_tree_tag_set - set a tag on a radix tree node
  *	@root:		radix tree root
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 05d7bc4..4c66acc 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -325,6 +325,37 @@ void multiorder_tagged_iteration(void)
 	item_kill_tree(&tree);
 }
 
+static void __multiorder_join(unsigned long index,
+				unsigned order1, unsigned order2)
+{
+	unsigned long loc;
+	void *item, *item2 = item_create(index + 1);
+	RADIX_TREE(tree, GFP_KERNEL);
+
+	item_insert_order(&tree, index, order2);
+	item = radix_tree_lookup(&tree, index);
+	radix_tree_join(&tree, index + 1, order1, item2);
+	loc = radix_tree_locate_item(&tree, item);
+	if (loc == -1)
+		free(item);
+	item = radix_tree_lookup(&tree, index + 1);
+	assert(item == item2);
+	item_kill_tree(&tree);
+}
+
+static void multiorder_join(void)
+{
+	int i, j, idx;
+
+	for (idx = 0; idx < 1024; idx = idx * 2 + 3) {
+		for (i = 1; i < 15; i++) {
+			for (j = 0; j < i; j++) {
+				__multiorder_join(idx, i, j);
+			}
+		}
+	}
+}
+
 void multiorder_checks(void)
 {
 	int i;
@@ -342,4 +373,5 @@ void multiorder_checks(void)
 	multiorder_tag_tests();
 	multiorder_iteration();
 	multiorder_tagged_iteration();
+	multiorder_join();
 }
-- 
2.10.2

[PATCH 11/29] radix-tree: Add radix_tree_split

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

This new function splits a larger multiorder entry into smaller entries
(potentially multi-order entries).  These entries are initialised to
RADIX_TREE_RETRY to ensure that RCU walkers who see this state aren't
confused.  The caller should then call radix_tree_for_each_slot() and
radix_tree_replace_slot() in order to turn these retry entries into the
intended new entries.  Tags are replicated from the original multiorder
entry into each new entry.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 include/linux/radix-tree.h            |   3 +
 lib/radix-tree.c                      | 109 ++++++++++++++++++++++++++++++++--
 tools/testing/radix-tree/multiorder.c |  26 ++++++++
 3 files changed, 134 insertions(+), 4 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 1efd81f..f5518f1 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -319,8 +319,11 @@ static inline void radix_tree_preload_end(void)
 	preempt_enable();
 }
 
+int radix_tree_split(struct radix_tree_root *, unsigned long index,
+			unsigned new_order);
 int radix_tree_join(struct radix_tree_root *, unsigned long index,
 			unsigned new_order, void *);
+
 /**
  * struct radix_tree_iter - radix tree iterator state
  *
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 6a76252..eaf0f353 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -231,7 +231,10 @@ static void dump_node(struct radix_tree_node *node, unsigned long index)
 		void *entry = node->slots[i];
 		if (!entry)
 			continue;
-		if (is_sibling_entry(node, entry)) {
+		if (entry == RADIX_TREE_RETRY) {
+			pr_debug("radix retry offset %ld indices %ld-%ld\n",
+					i, first, last);
+		} else if (is_sibling_entry(node, entry)) {
 			pr_debug("radix sblng %p offset %ld val %p indices %ld-%ld\n",
 					entry, i,
 					*(void **)entry_to_node(entry),
@@ -641,7 +644,10 @@ static inline int insert_entries(struct radix_tree_node *node, void **slot,
 	unsigned i, n, tag, offset, tags = 0;
 
 	if (node) {
-		n = 1 << (order - node->shift);
+		if (order > node->shift)
+			n = 1 << (order - node->shift);
+		else
+			n = 1;
 		offset = get_slot_offset(node, slot);
 	} else {
 		n = 1;
@@ -680,7 +686,8 @@ static inline int insert_entries(struct radix_tree_node *node, void **slot,
 					tag_set(node, tag, offset);
 		}
 		if (radix_tree_is_internal_node(old) &&
-					!is_sibling_entry(node, old))
+					!is_sibling_entry(node, old) &&
+					(old != RADIX_TREE_RETRY))
 			radix_tree_free_nodes(old);
 	}
 	if (node)
@@ -843,6 +850,98 @@ int radix_tree_join(struct radix_tree_root *root, unsigned long index,
 
 	return error;
 }
+
+int radix_tree_split(struct radix_tree_root *root, unsigned long index,
+				unsigned order)
+{
+	struct radix_tree_node *parent, *node, *child;
+	void **slot;
+	unsigned int offset, end;
+	unsigned n, tag, tags = 0;
+
+	if (!__radix_tree_lookup(root, index, &parent, &slot))
+		return -ENOENT;
+	if (!parent)
+		return -ENOENT;
+
+	offset = get_slot_offset(parent, slot);
+
+	for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+		if (tag_get(parent, tag, offset))
+			tags |= 1 << tag;
+
+	for (end = offset + 1; end < RADIX_TREE_MAP_SIZE; end++) {
+		if (!is_sibling_entry(parent, parent->slots[end]))
+			break;
+		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+			if (tags & (1 << tag))
+				tag_set(parent, tag, end);
+		/* rcu_assign_pointer ensures tags are set before RETRY */
+		rcu_assign_pointer(parent->slots[end], RADIX_TREE_RETRY);
+	}
+
+	if (order == parent->shift)
+		return 0;
+	if (order > parent->shift) {
+		while (offset < end)
+			offset += insert_entries(parent, &parent->slots[offset],
+					RADIX_TREE_RETRY, order, true);
+		return 0;
+	}
+
+	node = parent;
+
+	for (;;) {
+		if (node->shift > order) {
+			child = radix_tree_node_alloc(root);
+			if (!child)
+				goto nomem;
+			child->shift = node->shift - RADIX_TREE_MAP_SHIFT;
+			child->offset = offset;
+			child->count = 0;
+			child->parent = node;
+			if (node != parent) {
+				node->count++;
+				node->slots[offset] = node_to_entry(child);
+				for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+					if (tags & (1 << tag))
+						tag_set(node, tag, offset);
+			}
+
+			node = child;
+			offset = 0;
+			continue;
+		}
+
+		n = insert_entries(node, &node->slots[offset],
+					RADIX_TREE_RETRY, order, false);
+		BUG_ON(n > RADIX_TREE_MAP_SIZE);
+
+		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+			if (tags & (1 << tag))
+				tag_set(node, tag, offset);
+		offset += n;
+
+		while (offset == RADIX_TREE_MAP_SIZE) {
+			if (node == parent)
+				break;
+			offset = node->offset;
+			child = node;
+			node = node->parent;
+			rcu_assign_pointer(node->slots[offset],
+						node_to_entry(child));
+			offset++;
+		}
+		if ((node == parent) && (offset == end))
+			return 0;
+	}
+
+ nomem:
+	/* Shouldn't happen; did user forget to preload? */
+	/* TODO: free all the allocated nodes */
+	WARN_ON(1);
+	return -ENOMEM;
+}
 #endif
 
 /**
@@ -1081,8 +1180,10 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 			child = rcu_dereference_raw(node->slots[offset]);
 		}
 
-		if ((child == NULL) || (child == RADIX_TREE_RETRY))
+		if (!child)
 			goto restart;
+		if (child == RADIX_TREE_RETRY)
+			break;
 	} while (radix_tree_is_internal_node(child));
 
 	/* Update the iterator state */
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 4c66acc..d9e8155 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -356,6 +356,31 @@ static void multiorder_join(void)
 	}
 }
 
+static void __multiorder_split(int old_order, int new_order)
+{
+	RADIX_TREE(tree, GFP_KERNEL);
+	void **slot;
+	struct radix_tree_iter iter;
+
+	item_insert_order(&tree, 0, old_order);
+	radix_tree_tag_set(&tree, 0, 2);
+	radix_tree_split(&tree, 0, new_order);
+	radix_tree_for_each_slot(slot, &tree, &iter, 0) {
+		radix_tree_replace_slot(slot, item_create(iter.index));
+	}
+
+	item_kill_tree(&tree);
+}
+
+static void multiorder_split(void)
+{
+	int i, j;
+
+	for (i = 9; i < 19; i++)
+		for (j = 0; j < i; j++)
+			__multiorder_split(i, j);
+}
+
 void multiorder_checks(void)
 {
 	int i;
@@ -374,4 +399,5 @@ void multiorder_checks(void)
 	multiorder_iteration();
 	multiorder_tagged_iteration();
 	multiorder_join();
+	multiorder_split();
 }
-- 
2.10.2

[PATCH 12/29] radix-tree: Add radix_tree_split_preload()

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Calculate how many nodes we need to allocate to split an old_order entry
into multiple entries, each of size new_order.  The test suite checks that
we allocated exactly the right number of nodes; neither too many (checked
by rtp->nr == 0), nor too few (checked by comparing nr_allocated before
and after the call to radix_tree_split()).

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 include/linux/radix-tree.h            |  1 +
 lib/radix-tree.c                      | 24 +++++++++++++++++++++++-
 tools/testing/radix-tree/multiorder.c | 32 ++++++++++++++++++++++++++++++--
 tools/testing/radix-tree/test.h       |  5 +++++
 4 files changed, 59 insertions(+), 3 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index f5518f1..8ffb051 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -319,6 +319,7 @@ static inline void radix_tree_preload_end(void)
 	preempt_enable();
 }
 
+int radix_tree_split_preload(unsigned old_order, unsigned new_order, gfp_t);
 int radix_tree_split(struct radix_tree_root *, unsigned long index,
 			unsigned new_order);
 int radix_tree_join(struct radix_tree_root *, unsigned long index,
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index eaf0f353..6d73575 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -344,7 +344,7 @@ radix_tree_node_free(struct radix_tree_node *node)
  * To make use of this facility, the radix tree must be initialised without
  * __GFP_DIRECT_RECLAIM being passed to INIT_RADIX_TREE().
  */
-static int __radix_tree_preload(gfp_t gfp_mask, int nr)
+static int __radix_tree_preload(gfp_t gfp_mask, unsigned nr)
 {
 	struct radix_tree_preload *rtp;
 	struct radix_tree_node *node;
@@ -410,6 +410,28 @@ int radix_tree_maybe_preload(gfp_t gfp_mask)
 }
 EXPORT_SYMBOL(radix_tree_maybe_preload);
 
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+/*
+ * Preload with enough objects to ensure that we can split a single entry
+ * of order @old_order into many entries of size @new_order
+ */
+int radix_tree_split_preload(unsigned int old_order, unsigned int new_order,
+							gfp_t gfp_mask)
+{
+	unsigned top = 1 << (old_order % RADIX_TREE_MAP_SHIFT);
+	unsigned layers = (old_order / RADIX_TREE_MAP_SHIFT) -
+				(new_order / RADIX_TREE_MAP_SHIFT);
+	unsigned nr = 0;
+
+	WARN_ON_ONCE(!gfpflags_allow_blocking(gfp_mask));
+	BUG_ON(new_order >= old_order);
+
+	while (layers--)
+		nr = nr * RADIX_TREE_MAP_SIZE + 1;
+	return __radix_tree_preload(gfp_mask, top * nr);
+}
+#endif
+
 /*
  * The same as function above, but preload number of nodes required to insert
  * (1 << order) continuous naturally-aligned elements.
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index d9e8155..25e0463 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -356,18 +356,46 @@ static void multiorder_join(void)
 	}
 }
 
+static void check_mem(unsigned old_order, unsigned new_order, unsigned alloc)
+{
+	struct radix_tree_preload *rtp = &radix_tree_preloads;
+	if (rtp->nr != 0)
+		printf("split(%u %u) remaining %u\n", old_order, new_order,
+							rtp->nr);
+	/*
+	 * Can't check for equality here as some nodes may have been
+	 * RCU-freed while we ran.  But we should never finish with more
+	 * nodes allocated since they should have all been preloaded.
+	 */
+	if (nr_allocated > alloc)
+		printf("split(%u %u) allocated %u %u\n", old_order, new_order,
+							alloc, nr_allocated);
+}
+
 static void __multiorder_split(int old_order, int new_order)
 {
-	RADIX_TREE(tree, GFP_KERNEL);
+	RADIX_TREE(tree, GFP_ATOMIC);
 	void **slot;
 	struct radix_tree_iter iter;
+	unsigned alloc;
+
+	radix_tree_preload(GFP_KERNEL);
+	assert(item_insert_order(&tree, 0, old_order) == 0);
+	radix_tree_preload_end();
+
+	/* Wipe out the preloaded cache or it'll confuse check_mem() */
+	radix_tree_callback(NULL, CPU_DEAD, NULL);
 
-	item_insert_order(&tree, 0, old_order);
 	radix_tree_tag_set(&tree, 0, 2);
+
+	radix_tree_split_preload(old_order, new_order, GFP_KERNEL);
+	alloc = nr_allocated;
 	radix_tree_split(&tree, 0, new_order);
+	check_mem(old_order, new_order, alloc);
 	radix_tree_for_each_slot(slot, &tree, &iter, 0) {
 		radix_tree_replace_slot(slot, item_create(iter.index));
 	}
+	radix_tree_preload_end();
 
 	item_kill_tree(&tree);
 }
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index f2dc35f..33d2b6b 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -49,3 +49,8 @@ unsigned long node_maxindex(struct radix_tree_node *);
 unsigned long shift_maxindex(unsigned int shift);
 int radix_tree_callback(struct notifier_block *nfb,
 			unsigned long action, void *hcpu);
+struct radix_tree_preload {
+	unsigned nr;
+	struct radix_tree_node *nodes;
+};
+extern struct radix_tree_preload radix_tree_preloads;
-- 
2.10.2

[PATCH 13/29] radix-tree: Fix typo

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

---
 lib/radix-tree.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 6d73575..e917c56 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1147,7 +1147,7 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 	 * because RADIX_TREE_MAP_SHIFT < BITS_PER_LONG.
 	 *
 	 * This condition also used by radix_tree_next_slot() to stop
-	 * contiguous iterating, and forbid swithing to the next chunk.
+	 * contiguous iterating, and forbid switching to the next chunk.
 	 */
 	index = iter->next_index;
 	if (!index && iter->index)
-- 
2.10.2

[PATCH 14/29] radix-tree: Move rcu_head into a union with private_list

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

I want to be able to reference node->parent after freeing node.
Currently node->parent is in a union with rcu_head, so it is overwritten
when the node is put on the RCU list.  We know that private_list is not
referenced after the node is freed, so it is safe for these two members
to share space.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 include/linux/radix-tree.h | 14 ++++----------
 lib/radix-tree.c           |  1 +
 2 files changed, 5 insertions(+), 10 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 8ffb051..66fb8c0 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -88,18 +88,12 @@ struct radix_tree_node {
 	unsigned char	shift;	/* Bits remaining in each slot */
 	unsigned char	offset;	/* Slot offset in parent */
 	unsigned int	count;
+	struct radix_tree_node *parent;		/* Used when ascending tree */
+	void *private_data;			/* For tree user */
 	union {
-		struct {
-			/* Used when ascending tree */
-			struct radix_tree_node *parent;
-			/* For tree user */
-			void *private_data;
-		};
-		/* Used when freeing node */
-		struct rcu_head	rcu_head;
+		struct list_head private_list;	/* For tree user */
+		struct rcu_head	rcu_head;	/* Used when freeing node */
 	};
-	/* For tree user */
-	struct list_head private_list;
 	void __rcu	*slots[RADIX_TREE_MAP_SIZE];
 	unsigned long	tags[RADIX_TREE_MAX_TAGS][RADIX_TREE_TAG_LONGS];
 };
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index e917c56..baf4ba1 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -325,6 +325,7 @@ static void radix_tree_node_rcu_free(struct rcu_head *head)
 	 */
 	memset(node->slots, 0, sizeof(node->slots));
 	memset(node->tags, 0, sizeof(node->tags));
+	INIT_LIST_HEAD(&node->private_list);
 
 	kmem_cache_free(radix_tree_node_cachep, node);
 }
-- 
2.10.2

[PATCH 15/29] radix-tree: Create node_tag_set()

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Similar to node_tag_clear(), factor node_tag_set() out of
radix_tree_range_tag_if_tagged().

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 lib/radix-tree.c | 41 +++++++++++++++++++----------------------
 1 file changed, 19 insertions(+), 22 deletions(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index baf4ba1..bf1303f 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1028,6 +1028,22 @@ static void node_tag_clear(struct radix_tree_root *root,
 		root_tag_clear(root, tag);
 }
 
+static void node_tag_set(struct radix_tree_root *root,
+				struct radix_tree_node *node,
+				unsigned int tag, unsigned int offset)
+{
+	while (node) {
+		if (tag_get(node, tag, offset))
+			return;
+		tag_set(node, tag, offset);
+		offset = node->offset;
+		node = node->parent;
+	}
+
+	if (!root_tag_get(root, tag))
+		root_tag_set(root, tag);
+}
+
 /**
  *	radix_tree_tag_clear - clear a tag on a radix tree node
  *	@root:		radix tree root
@@ -1268,7 +1284,7 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
 		unsigned long nr_to_tag,
 		unsigned int iftag, unsigned int settag)
 {
-	struct radix_tree_node *parent, *node, *child;
+	struct radix_tree_node *node, *child;
 	unsigned long maxindex;
 	unsigned long tagged = 0;
 	unsigned long index = *first_indexp;
@@ -1303,22 +1319,8 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
 			continue;
 		}
 
-		/* tag the leaf */
 		tagged++;
-		tag_set(node, settag, offset);
-
-		/* walk back up the path tagging interior nodes */
-		parent = node;
-		for (;;) {
-			offset = parent->offset;
-			parent = parent->parent;
-			if (!parent)
-				break;
-			/* stop if we find a node with the tag already set */
-			if (tag_get(parent, settag, offset))
-				break;
-			tag_set(parent, settag, offset);
-		}
+		node_tag_set(root, node, settag, offset);
  next:
 		/* Go to next entry in node */
 		index = ((index >> node->shift) + 1) << node->shift;
@@ -1340,12 +1342,7 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
 		if (tagged >= nr_to_tag)
 			break;
 	}
-	/*
-	 * We need not to tag the root tag if there is no tag which is set with
-	 * settag within the range from *first_indexp to last_index.
-	 */
-	if (tagged > 0)
-		root_tag_set(root, settag);
+
 	*first_indexp = index;
 
 	return tagged;
-- 
2.10.2

[PATCH 16/29] radix-tree: Make radix_tree_find_next_bit more useful

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Since this function is specialised to the radix tree, pass in the node
and tag to calculate the address of the bitmap in radix_tree_find_next_bit()
instead of the caller.  Likewise, there is no need to pass in the size of
the bitmap.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 lib/radix-tree.c | 17 +++++++----------
 1 file changed, 7 insertions(+), 10 deletions(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index bf1303f..2c3fac4 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -191,13 +191,12 @@ static inline int any_tag_set(struct radix_tree_node *node, unsigned int tag)
  * Returns next bit offset, or size if nothing found.
  */
 static __always_inline unsigned long
-radix_tree_find_next_bit(const unsigned long *addr,
-			 unsigned long size, unsigned long offset)
+radix_tree_find_next_bit(struct radix_tree_node *node, unsigned int tag,
+			 unsigned long offset)
 {
-	if (!__builtin_constant_p(size))
-		return find_next_bit(addr, size, offset);
+	const unsigned long *addr = node->tags[tag];
 
-	if (offset < size) {
+	if (offset < RADIX_TREE_MAP_SIZE) {
 		unsigned long tmp;
 
 		addr += offset / BITS_PER_LONG;
@@ -205,14 +204,14 @@ radix_tree_find_next_bit(const unsigned long *addr,
 		if (tmp)
 			return __ffs(tmp) + offset;
 		offset = (offset + BITS_PER_LONG) & ~(BITS_PER_LONG - 1);
-		while (offset < size) {
+		while (offset < RADIX_TREE_MAP_SIZE) {
 			tmp = *++addr;
 			if (tmp)
 				return __ffs(tmp) + offset;
 			offset += BITS_PER_LONG;
 		}
 	}
-	return size;
+	return RADIX_TREE_MAP_SIZE;
 }
 
 #ifndef __KERNEL__
@@ -1197,9 +1196,7 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 				return NULL;
 
 			if (flags & RADIX_TREE_ITER_TAGGED)
-				offset = radix_tree_find_next_bit(
-						node->tags[tag],
-						RADIX_TREE_MAP_SIZE,
+				offset = radix_tree_find_next_bit(node, tag,
 						offset + 1);
 			else
 				while (++offset	< RADIX_TREE_MAP_SIZE) {
-- 
2.10.2

[PATCH 17/29] radix-tree: Improve dump output

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

Print the indices of the entries as unsigned (instead of signed) integers
and print the parent node of each entry to help navigate around larger
trees where the layout is not quite so obvious.  Print the indices
covered by a node.  Rearrange the order of fields printed so the indices
and parents line up for each type of entry.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 lib/radix-tree.c | 50 +++++++++++++++++++++++++-------------------------
 1 file changed, 25 insertions(+), 25 deletions(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 2c3fac4..09c5f1d 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -214,15 +214,29 @@ radix_tree_find_next_bit(struct radix_tree_node *node, unsigned int tag,
 	return RADIX_TREE_MAP_SIZE;
 }
 
+/*
+ * The maximum index which can be stored in a radix tree
+ */
+static inline unsigned long shift_maxindex(unsigned int shift)
+{
+	return (RADIX_TREE_MAP_SIZE << shift) - 1;
+}
+
+static inline unsigned long node_maxindex(struct radix_tree_node *node)
+{
+	return shift_maxindex(node->shift);
+}
+
 #ifndef __KERNEL__
 static void dump_node(struct radix_tree_node *node, unsigned long index)
 {
 	unsigned long i;
 
-	pr_debug("radix node: %p offset %d tags %lx %lx %lx shift %d count %d parent %p\n",
-		node, node->offset,
+	pr_debug("radix node: %p offset %d indices %lu-%lu parent %p tags %lx %lx %lx shift %d count %d\n",
+		node, node->offset, index, index | node_maxindex(node),
+		node->parent,
 		node->tags[0][0], node->tags[1][0], node->tags[2][0],
-		node->shift, node->count, node->parent);
+		node->shift, node->count);
 
 	for (i = 0; i < RADIX_TREE_MAP_SIZE; i++) {
 		unsigned long first = index | (i << node->shift);
@@ -231,16 +245,15 @@ static void dump_node(struct radix_tree_node *node, unsigned long index)
 		if (!entry)
 			continue;
 		if (entry == RADIX_TREE_RETRY) {
-			pr_debug("radix retry offset %ld indices %ld-%ld\n",
-					i, first, last);
-		} else if (is_sibling_entry(node, entry)) {
-			pr_debug("radix sblng %p offset %ld val %p indices %ld-%ld\n",
-					entry, i,
-					*(void **)entry_to_node(entry),
-					first, last);
+			pr_debug("radix retry offset %ld indices %lu-%lu parent %p\n",
+					i, first, last, node);
 		} else if (!radix_tree_is_internal_node(entry)) {
-			pr_debug("radix entry %p offset %ld indices %ld-%ld\n",
-					entry, i, first, last);
+			pr_debug("radix entry %p offset %ld indices %lu-%lu parent %p\n",
+					entry, i, first, last, node);
+		} else if (is_sibling_entry(node, entry)) {
+			pr_debug("radix sblng %p offset %ld indices %lu-%lu parent %p val %p\n",
+					entry, i, first, last, node,
+					*(void **)entry_to_node(entry));
 		} else {
 			dump_node(entry_to_node(entry), first);
 		}
@@ -477,19 +490,6 @@ int radix_tree_maybe_preload_order(gfp_t gfp_mask, int order)
 	return __radix_tree_preload(gfp_mask, nr_nodes);
 }
 
-/*
- * The maximum index which can be stored in a radix tree
- */
-static inline unsigned long shift_maxindex(unsigned int shift)
-{
-	return (RADIX_TREE_MAP_SIZE << shift) - 1;
-}
-
-static inline unsigned long node_maxindex(struct radix_tree_node *node)
-{
-	return shift_maxindex(node->shift);
-}
-
 static unsigned radix_tree_load_root(struct radix_tree_root *root,
 		struct radix_tree_node **nodep, unsigned long *maxindex)
 {
-- 
2.10.2

[PATCH 18/29] btrfs: Fix race in btrfs_free_dummy_fs_info()

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

We drop the lock which protects the radix tree, so we must call
radix_tree_iter_next() in order to avoid a modification to the tree
invalidating the iterator state.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 fs/btrfs/tests/btrfs-tests.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c
index bf62ad9..73076a0 100644
--- a/fs/btrfs/tests/btrfs-tests.c
+++ b/fs/btrfs/tests/btrfs-tests.c
@@ -162,6 +162,7 @@ void btrfs_free_dummy_fs_info(struct btrfs_fs_info *fs_info)
 				slot = radix_tree_iter_retry(&iter);
 			continue;
 		}
+		slot = radix_tree_iter_next(&iter);
 		spin_unlock(&fs_info->buffer_lock);
 		free_extent_buffer_stale(eb);
 		spin_lock(&fs_info->buffer_lock);
-- 
2.10.2

[PATCH 19/29] radix tree test suite: iteration test misuses RCU

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

Each thread needs to register itself with RCU, otherwise the reading
thread's read lock has no effect and the freeing thread will free the
memory in the tree without waiting for the read lock to be dropped.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 tools/testing/radix-tree/iteration_check.c | 28 ++++++++++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)

diff --git a/tools/testing/radix-tree/iteration_check.c b/tools/testing/radix-tree/iteration_check.c
index 11d570c..df71cb8 100644
--- a/tools/testing/radix-tree/iteration_check.c
+++ b/tools/testing/radix-tree/iteration_check.c
@@ -29,6 +29,8 @@ static void *add_entries_fn(void *arg)
 {
 	int pgoff;
 
+	rcu_register_thread();
+
 	while (!test_complete) {
 		for (pgoff = 0; pgoff < 100; pgoff++) {
 			pthread_mutex_lock(&tree_lock);
@@ -38,6 +40,8 @@ static void *add_entries_fn(void *arg)
 		}
 	}
 
+	rcu_unregister_thread();
+
 	return NULL;
 }
 
@@ -53,6 +57,8 @@ static void *tagged_iteration_fn(void *arg)
 	struct radix_tree_iter iter;
 	void **slot;
 
+	rcu_register_thread();
+
 	while (!test_complete) {
 		rcu_read_lock();
 		radix_tree_for_each_tagged(slot, &tree, &iter, 0, TAG) {
@@ -72,12 +78,18 @@ static void *tagged_iteration_fn(void *arg)
 				continue;
 			}
 
-			if (rand_r(&seeds[0]) % 50 == 0)
+			if (rand_r(&seeds[0]) % 50 == 0) {
 				slot = radix_tree_iter_next(&iter);
+				rcu_read_unlock();
+				rcu_barrier();
+				rcu_read_lock();
+			}
 		}
 		rcu_read_unlock();
 	}
 
+	rcu_unregister_thread();
+
 	return NULL;
 }
 
@@ -93,6 +105,8 @@ static void *untagged_iteration_fn(void *arg)
 	struct radix_tree_iter iter;
 	void **slot;
 
+	rcu_register_thread();
+
 	while (!test_complete) {
 		rcu_read_lock();
 		radix_tree_for_each_slot(slot, &tree, &iter, 0) {
@@ -112,12 +126,18 @@ static void *untagged_iteration_fn(void *arg)
 				continue;
 			}
 
-			if (rand_r(&seeds[1]) % 50 == 0)
+			if (rand_r(&seeds[1]) % 50 == 0) {
 				slot = radix_tree_iter_next(&iter);
+				rcu_read_unlock();
+				rcu_barrier();
+				rcu_read_lock();
+			}
 		}
 		rcu_read_unlock();
 	}
 
+	rcu_unregister_thread();
+
 	return NULL;
 }
 
@@ -127,6 +147,8 @@ static void *untagged_iteration_fn(void *arg)
  */
 static void *remove_entries_fn(void *arg)
 {
+	rcu_register_thread();
+
 	while (!test_complete) {
 		int pgoff;
 
@@ -137,6 +159,8 @@ static void *remove_entries_fn(void *arg)
 		pthread_mutex_unlock(&tree_lock);
 	}
 
+	rcu_unregister_thread();
+
 	return NULL;
 }
 
-- 
2.10.2

[PATCH 20/29] radix tree: Improve multiorder iterators

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

This fixes several interlinked problems with the iterators in the
presence of multiorder entries.

1. radix_tree_iter_next() would only advance by one slot, which would
result in the iterators returning the same entry more than once if there
were sibling entries.

2. radix_tree_next_slot() could return an internal pointer instead of
a user pointer if a tagged multiorder entry was immediately followed by
an entry of lower order.

3. radix_tree_next_slot() expanded to a lot more code than it used to
when multiorder support was compiled in.  And I wasn't comfortable with
entry_to_node() being in a header file.

Fixing radix_tree_iter_next() for the presence of sibling entries
necessarily involves examining the contents of the radix tree, so we now
need to pass 'slot' to radix_tree_iter_next(), and we need to change the
calling convention so it is called *before* dropping the lock which
protects the tree.  Fortunately, unconverted code won't compile.

radix_tree_next_slot() becomes closer to how it looked before multiorder
support was introduced.  It only checks to see if the next entry in the
chunk is a sibling entry or a pointer to a node; this should be rare
enough that handling this case out of line is not a performance impact
(and such impact is amortised by the fact that the entry we just
processed was a multiorder entry).  Also, radix_tree_next_slot() used
to force a new chunk lookup for untagged entries, which is more
expensive than the out of line sibling entry skipping.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 fs/btrfs/tests/btrfs-tests.c               |  2 +-
 include/linux/radix-tree.h                 | 63 ++++++++------------
 lib/radix-tree.c                           | 94 ++++++++++++++++++++++++++++++
 mm/khugepaged.c                            |  2 +-
 mm/shmem.c                                 |  6 +-
 tools/testing/radix-tree/iteration_check.c |  4 +-
 tools/testing/radix-tree/multiorder.c      | 12 ++++
 tools/testing/radix-tree/regression3.c     |  6 +-
 tools/testing/radix-tree/test.h            |  1 +
 9 files changed, 142 insertions(+), 48 deletions(-)

diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c
index 73076a0..6d3457a 100644
--- a/fs/btrfs/tests/btrfs-tests.c
+++ b/fs/btrfs/tests/btrfs-tests.c
@@ -162,7 +162,7 @@ void btrfs_free_dummy_fs_info(struct btrfs_fs_info *fs_info)
 				slot = radix_tree_iter_retry(&iter);
 			continue;
 		}
-		slot = radix_tree_iter_next(&iter);
+		slot = radix_tree_iter_next(slot, &iter);
 		spin_unlock(&fs_info->buffer_lock);
 		free_extent_buffer_stale(eb);
 		spin_lock(&fs_info->buffer_lock);
diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 66fb8c0..36c6175 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -424,15 +424,10 @@ __radix_tree_iter_add(struct radix_tree_iter *iter, unsigned long slots)
  *
  * If the iterator needs to release then reacquire a lock, the chunk may
  * have been invalidated by an insertion or deletion.  Call this function
- * to continue the iteration from the next index.
+ * before releasing the lock to continue the iteration from the next index.
  */
-static inline __must_check
-void **radix_tree_iter_next(struct radix_tree_iter *iter)
-{
-	iter->next_index = __radix_tree_iter_add(iter, 1);
-	iter->tags = 0;
-	return NULL;
-}
+void **__must_check radix_tree_iter_next(void **slot,
+					struct radix_tree_iter *iter);
 
 /**
  * radix_tree_chunk_size - get current chunk size
@@ -446,10 +441,17 @@ radix_tree_chunk_size(struct radix_tree_iter *iter)
 	return (iter->next_index - iter->index) >> iter_shift(iter);
 }
 
-static inline struct radix_tree_node *entry_to_node(void *ptr)
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
+				unsigned flags);
+#else
+/* Can't happen without sibling entries, but the compiler can't tell that */
+static inline void ** __radix_tree_next_slot(void **slot,
+				struct radix_tree_iter *iter, unsigned flags)
 {
-	return (void *)((unsigned long)ptr & ~RADIX_TREE_INTERNAL_NODE);
+	return slot;
 }
+#endif
 
 /**
  * radix_tree_next_slot - find next slot in chunk
@@ -474,51 +476,31 @@ static __always_inline void **
 radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
 {
 	if (flags & RADIX_TREE_ITER_TAGGED) {
-		void *canon = slot;
-
 		iter->tags >>= 1;
 		if (unlikely(!iter->tags))
 			return NULL;
-		while (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
-					radix_tree_is_internal_node(slot[1])) {
-			if (entry_to_node(slot[1]) == canon) {
-				iter->tags >>= 1;
-				iter->index = __radix_tree_iter_add(iter, 1);
-				slot++;
-				continue;
-			}
-			iter->next_index = __radix_tree_iter_add(iter, 1);
-			return NULL;
-		}
 		if (likely(iter->tags & 1ul)) {
 			iter->index = __radix_tree_iter_add(iter, 1);
-			return slot + 1;
+			slot++;
+			goto found;
 		}
 		if (!(flags & RADIX_TREE_ITER_CONTIG)) {
 			unsigned offset = __ffs(iter->tags);
 
-			iter->tags >>= offset;
-			iter->index = __radix_tree_iter_add(iter, offset + 1);
-			return slot + offset + 1;
+			iter->tags >>= offset++;
+			iter->index = __radix_tree_iter_add(iter, offset);
+			slot += offset;
+			goto found;
 		}
 	} else {
 		long count = radix_tree_chunk_size(iter);
-		void *canon = slot;
 
 		while (--count > 0) {
 			slot++;
 			iter->index = __radix_tree_iter_add(iter, 1);
 
-			if (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
-			    radix_tree_is_internal_node(*slot)) {
-				if (entry_to_node(*slot) == canon)
-					continue;
-				iter->next_index = iter->index;
-				break;
-			}
-
 			if (likely(*slot))
-				return slot;
+				goto found;
 			if (flags & RADIX_TREE_ITER_CONTIG) {
 				/* forbid switching to the next chunk */
 				iter->next_index = 0;
@@ -527,6 +509,11 @@ radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
 		}
 	}
 	return NULL;
+
+ found:
+	if (unlikely(radix_tree_is_internal_node(*slot)))
+		return __radix_tree_next_slot(slot, iter, flags);
+	return slot;
 }
 
 /**
@@ -577,6 +564,6 @@ radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
 	     slot || (slot = radix_tree_next_chunk(root, iter,		\
 			      RADIX_TREE_ITER_TAGGED | tag)) ;		\
 	     slot = radix_tree_next_slot(slot, iter,			\
-				RADIX_TREE_ITER_TAGGED))
+				RADIX_TREE_ITER_TAGGED | tag))
 
 #endif /* _LINUX_RADIX_TREE_H */
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 09c5f1d..27b53ef 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -69,6 +69,11 @@ struct radix_tree_preload {
 };
 static DEFINE_PER_CPU(struct radix_tree_preload, radix_tree_preloads) = { 0, };
 
+static inline struct radix_tree_node *entry_to_node(void *ptr)
+{
+	return (void *)((unsigned long)ptr & ~RADIX_TREE_INTERNAL_NODE);
+}
+
 static inline void *node_to_entry(void *ptr)
 {
 	return (void *)((unsigned long)ptr | RADIX_TREE_INTERNAL_NODE);
@@ -1138,6 +1143,95 @@ static inline void __set_iter_shift(struct radix_tree_iter *iter,
 #endif
 }
 
+static void ** __radix_tree_iter_next(struct radix_tree_node **nodep,
+			void **slot, struct radix_tree_iter *iter)
+{
+	void *sib = node_to_entry(slot - 1);
+
+	while (iter->index < iter->next_index) {
+		*nodep = rcu_dereference_raw(*slot);
+		if (*nodep && *nodep != sib)
+			return slot;
+		slot++;
+		iter->index = __radix_tree_iter_add(iter, 1);
+		iter->tags >>= 1;
+	}
+
+	*nodep = NULL;
+	return NULL;
+}
+
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
+					unsigned flags)
+{
+	unsigned tag = flags & RADIX_TREE_ITER_TAG_MASK;
+	struct radix_tree_node *node = rcu_dereference_raw(*slot);
+
+	slot = __radix_tree_iter_next(&node, slot, iter);
+
+	while (radix_tree_is_internal_node(node)) {
+		unsigned offset;
+
+		if (node == RADIX_TREE_RETRY)
+			return slot;
+		node = entry_to_node(node);
+
+		if (flags & RADIX_TREE_ITER_TAGGED) {
+			unsigned tag_long, tag_bit;
+			offset = radix_tree_find_next_bit(node, tag, 0);
+			if (offset == RADIX_TREE_MAP_SIZE)
+				return NULL;
+			slot = &node->slots[offset];
+
+			tag_long = offset / BITS_PER_LONG;
+			tag_bit  = offset % BITS_PER_LONG;
+			iter->tags = node->tags[tag][tag_long] >> tag_bit;
+			BUG_ON(iter->tags >= (RADIX_TREE_MAP_SIZE * 2));
+			node = rcu_dereference_raw(*slot);
+		} else {
+			offset = 0;
+			slot = &node->slots[0];
+			for (;;) {
+				node = rcu_dereference_raw(*slot);
+				if (node)
+					break;
+				slot++;
+				offset++;
+				if (offset == RADIX_TREE_MAP_SIZE)
+					return NULL;
+			}
+		}
+		if ((flags & RADIX_TREE_ITER_CONTIG) && (offset > 0))
+			goto none;
+		iter->shift -= RADIX_TREE_MAP_SHIFT;
+		iter->index = __radix_tree_iter_add(iter, offset);
+		iter->next_index = (iter->index | shift_maxindex(iter->shift)) +
+					1;
+	}
+
+	return slot;
+ none:
+	iter->next_index = 0;
+	return NULL;
+}
+EXPORT_SYMBOL(__radix_tree_next_slot);
+#endif
+
+void **radix_tree_iter_next(void **slot, struct radix_tree_iter *iter)
+{
+	struct radix_tree_node *node;
+
+	slot++;
+	iter->index = __radix_tree_iter_add(iter, 1);
+	node = rcu_dereference_raw(*slot);
+	__radix_tree_iter_next(&node, slot, iter);
+	iter->next_index = iter->index;
+	iter->tags = 0;
+	return NULL;
+}
+EXPORT_SYMBOL(radix_tree_iter_next);
+
 /**
  * radix_tree_next_chunk - find next chunk of slots for iteration
  *
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 728d779..46155d1 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -1614,8 +1614,8 @@ static void khugepaged_scan_shmem(struct mm_struct *mm,
 		present++;
 
 		if (need_resched()) {
+			slot = radix_tree_iter_next(slot, &iter);
 			cond_resched_rcu();
-			slot = radix_tree_iter_next(&iter);
 		}
 	}
 	rcu_read_unlock();
diff --git a/mm/shmem.c b/mm/shmem.c
index 166ebf5..0b3fe33 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -658,8 +658,8 @@ unsigned long shmem_partial_swap_usage(struct address_space *mapping,
 			swapped++;
 
 		if (need_resched()) {
+			slot = radix_tree_iter_next(slot, &iter);
 			cond_resched_rcu();
-			slot = radix_tree_iter_next(&iter);
 		}
 	}
 
@@ -2434,8 +2434,8 @@ static void shmem_tag_pins(struct address_space *mapping)
 		}
 
 		if (need_resched()) {
+			slot = radix_tree_iter_next(slot, &iter);
 			cond_resched_rcu();
-			slot = radix_tree_iter_next(&iter);
 		}
 	}
 	rcu_read_unlock();
@@ -2504,8 +2504,8 @@ static int shmem_wait_for_pins(struct address_space *mapping)
 			spin_unlock_irq(&mapping->tree_lock);
 continue_resched:
 			if (need_resched()) {
+				slot = radix_tree_iter_next(slot, &iter);
 				cond_resched_rcu();
-				slot = radix_tree_iter_next(&iter);
 			}
 		}
 		rcu_read_unlock();
diff --git a/tools/testing/radix-tree/iteration_check.c b/tools/testing/radix-tree/iteration_check.c
index df71cb8..6eca4c2 100644
--- a/tools/testing/radix-tree/iteration_check.c
+++ b/tools/testing/radix-tree/iteration_check.c
@@ -79,7 +79,7 @@ static void *tagged_iteration_fn(void *arg)
 			}
 
 			if (rand_r(&seeds[0]) % 50 == 0) {
-				slot = radix_tree_iter_next(&iter);
+				slot = radix_tree_iter_next(slot, &iter);
 				rcu_read_unlock();
 				rcu_barrier();
 				rcu_read_lock();
@@ -127,7 +127,7 @@ static void *untagged_iteration_fn(void *arg)
 			}
 
 			if (rand_r(&seeds[1]) % 50 == 0) {
-				slot = radix_tree_iter_next(&iter);
+				slot = radix_tree_iter_next(slot, &iter);
 				rcu_read_unlock();
 				rcu_barrier();
 				rcu_read_lock();
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 25e0463..588209a 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -232,10 +232,14 @@ void multiorder_iteration(void)
 			int height = order[i] / RADIX_TREE_MAP_SHIFT;
 			int shift = height * RADIX_TREE_MAP_SHIFT;
 			int mask = (1 << order[i]) - 1;
+			struct item *item = *slot;
 
 			assert(iter.index >= (index[i] &~ mask));
 			assert(iter.index <= (index[i] | mask));
 			assert(iter.shift == shift);
+			assert(!radix_tree_is_internal_node(item));
+			assert(item->index >= (index[i] &~ mask));
+			assert(item->index <= (index[i] | mask));
 			i++;
 		}
 	}
@@ -279,12 +283,16 @@ void multiorder_tagged_iteration(void)
 		}
 
 		radix_tree_for_each_tagged(slot, &tree, &iter, j, 1) {
+			struct item *item = *slot;
 			for (k = i; index[k] < tag_index[i]; k++)
 				;
 			mask = (1 << order[k]) - 1;
 
 			assert(iter.index >= (tag_index[i] &~ mask));
 			assert(iter.index <= (tag_index[i] | mask));
+			assert(!radix_tree_is_internal_node(item));
+			assert(item->index >= (tag_index[i] &~ mask));
+			assert(item->index <= (tag_index[i] | mask));
 			i++;
 		}
 	}
@@ -303,12 +311,16 @@ void multiorder_tagged_iteration(void)
 		}
 
 		radix_tree_for_each_tagged(slot, &tree, &iter, j, 2) {
+			struct item *item = *slot;
 			for (k = i; index[k] < tag_index[i]; k++)
 				;
 			mask = (1 << order[k]) - 1;
 
 			assert(iter.index >= (tag_index[i] &~ mask));
 			assert(iter.index <= (tag_index[i] | mask));
+			assert(!radix_tree_is_internal_node(item));
+			assert(item->index >= (tag_index[i] &~ mask));
+			assert(item->index <= (tag_index[i] | mask));
 			i++;
 		}
 	}
diff --git a/tools/testing/radix-tree/regression3.c b/tools/testing/radix-tree/regression3.c
index 1f06ed7..4d28eeb 100644
--- a/tools/testing/radix-tree/regression3.c
+++ b/tools/testing/radix-tree/regression3.c
@@ -88,7 +88,7 @@ void regression3_test(void)
 		printf("slot %ld %p\n", iter.index, *slot);
 		if (!iter.index) {
 			printf("next at %ld\n", iter.index);
-			slot = radix_tree_iter_next(&iter);
+			slot = radix_tree_iter_next(slot, &iter);
 		}
 	}
 
@@ -96,7 +96,7 @@ void regression3_test(void)
 		printf("contig %ld %p\n", iter.index, *slot);
 		if (!iter.index) {
 			printf("next at %ld\n", iter.index);
-			slot = radix_tree_iter_next(&iter);
+			slot = radix_tree_iter_next(slot, &iter);
 		}
 	}
 
@@ -106,7 +106,7 @@ void regression3_test(void)
 		printf("tagged %ld %p\n", iter.index, *slot);
 		if (!iter.index) {
 			printf("next at %ld\n", iter.index);
-			slot = radix_tree_iter_next(&iter);
+			slot = radix_tree_iter_next(slot, &iter);
 		}
 	}
 
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index 33d2b6b..b678f13 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -43,6 +43,7 @@ void verify_tag_consistency(struct radix_tree_root *root, unsigned int tag);
 extern int nr_allocated;
 
 /* Normally private parts of lib/radix-tree.c */
+struct radix_tree_node *entry_to_node(void *ptr);
 void radix_tree_dump(struct radix_tree_root *root);
 int root_tag_get(struct radix_tree_root *root, unsigned int tag);
 unsigned long node_maxindex(struct radix_tree_node *);
-- 
2.10.2

[PATCH 21/29] radix-tree: Delete radix_tree_locate_item()

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

This rather complicated function can be better implemented as an iterator.
It has only one caller, so move the functionality to the only place that
needs it.  Update the test suite to follow the same pattern.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 include/linux/radix-tree.h            |  1 -
 lib/radix-tree.c                      | 99 -----------------------------------
 mm/shmem.c                            | 26 ++++++++-
 tools/testing/radix-tree/main.c       |  8 +--
 tools/testing/radix-tree/multiorder.c |  2 +-
 tools/testing/radix-tree/test.c       | 22 ++++++++
 tools/testing/radix-tree/test.h       |  2 +
 7 files changed, 54 insertions(+), 106 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 36c6175..57bf635 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -306,7 +306,6 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
 		unsigned long nr_to_tag,
 		unsigned int fromtag, unsigned int totag);
 int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
-unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item);
 
 static inline void radix_tree_preload_end(void)
 {
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 27b53ef..7e70ac9 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1605,105 +1605,6 @@ radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
 }
 EXPORT_SYMBOL(radix_tree_gang_lookup_tag_slot);
 
-#if defined(CONFIG_SHMEM) && defined(CONFIG_SWAP)
-#include <linux/sched.h> /* for cond_resched() */
-
-struct locate_info {
-	unsigned long found_index;
-	bool stop;
-};
-
-/*
- * This linear search is at present only useful to shmem_unuse_inode().
- */
-static unsigned long __locate(struct radix_tree_node *slot, void *item,
-			      unsigned long index, struct locate_info *info)
-{
-	unsigned long i;
-
-	do {
-		unsigned int shift = slot->shift;
-
-		for (i = (index >> shift) & RADIX_TREE_MAP_MASK;
-		     i < RADIX_TREE_MAP_SIZE;
-		     i++, index += (1UL << shift)) {
-			struct radix_tree_node *node =
-					rcu_dereference_raw(slot->slots[i]);
-			if (node == RADIX_TREE_RETRY)
-				goto out;
-			if (!radix_tree_is_internal_node(node)) {
-				if (node == item) {
-					info->found_index = index;
-					info->stop = true;
-					goto out;
-				}
-				continue;
-			}
-			node = entry_to_node(node);
-			if (is_sibling_entry(slot, node))
-				continue;
-			slot = node;
-			break;
-		}
-	} while (i < RADIX_TREE_MAP_SIZE);
-
-out:
-	if ((index == 0) && (i == RADIX_TREE_MAP_SIZE))
-		info->stop = true;
-	return index;
-}
-
-/**
- *	radix_tree_locate_item - search through radix tree for item
- *	@root:		radix tree root
- *	@item:		item to be found
- *
- *	Returns index where item was found, or -1 if not found.
- *	Caller must hold no lock (since this time-consuming function needs
- *	to be preemptible), and must check afterwards if item is still there.
- */
-unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item)
-{
-	struct radix_tree_node *node;
-	unsigned long max_index;
-	unsigned long cur_index = 0;
-	struct locate_info info = {
-		.found_index = -1,
-		.stop = false,
-	};
-
-	do {
-		rcu_read_lock();
-		node = rcu_dereference_raw(root->rnode);
-		if (!radix_tree_is_internal_node(node)) {
-			rcu_read_unlock();
-			if (node == item)
-				info.found_index = 0;
-			break;
-		}
-
-		node = entry_to_node(node);
-
-		max_index = node_maxindex(node);
-		if (cur_index > max_index) {
-			rcu_read_unlock();
-			break;
-		}
-
-		cur_index = __locate(node, item, cur_index, &info);
-		rcu_read_unlock();
-		cond_resched();
-	} while (!info.stop && cur_index <= max_index);
-
-	return info.found_index;
-}
-#else
-unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item)
-{
-	return -1;
-}
-#endif /* CONFIG_SHMEM && CONFIG_SWAP */
-
 /**
  *	radix_tree_shrink    -    shrink radix tree to minimum height
  *	@root		radix tree root
diff --git a/mm/shmem.c b/mm/shmem.c
index 0b3fe33..8f9c9aa 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -1046,6 +1046,30 @@ static void shmem_evict_inode(struct inode *inode)
 	clear_inode(inode);
 }
 
+static unsigned long find_swap_entry(struct radix_tree_root *root, void *item)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+	unsigned long found = -1;
+	unsigned int checked = 0;
+
+	rcu_read_lock();
+	radix_tree_for_each_slot(slot, root, &iter, 0) {
+		if (*slot == item) {
+			found = iter.index;
+			break;
+		}
+		checked++;
+		if ((checked % 4096) != 0)
+			continue;
+		slot = radix_tree_iter_next(slot, &iter);
+		cond_resched_rcu();
+	}
+
+	rcu_read_unlock();
+	return found;
+}
+
 /*
  * If swap found in inode, free it and move page from swapcache to filecache.
  */
@@ -1059,7 +1083,7 @@ static int shmem_unuse_inode(struct shmem_inode_info *info,
 	int error = 0;
 
 	radswap = swp_to_radix_entry(swap);
-	index = radix_tree_locate_item(&mapping->page_tree, radswap);
+	index = find_swap_entry(&mapping->page_tree, radswap);
 	if (index == -1)
 		return -EAGAIN;	/* tell shmem_unuse we found nothing */
 
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 8621542..93a77f9 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -239,7 +239,7 @@ static void __locate_check(struct radix_tree_root *tree, unsigned long index,
 
 	item_insert_order(tree, index, order);
 	item = item_lookup(tree, index);
-	index2 = radix_tree_locate_item(tree, item);
+	index2 = find_item(tree, item);
 	if (index != index2) {
 		printf("index %ld order %d inserted; found %ld\n",
 			index, order, index2);
@@ -273,17 +273,17 @@ static void locate_check(void)
 			     index += (1UL << order)) {
 				__locate_check(&tree, index + offset, order);
 			}
-			if (radix_tree_locate_item(&tree, &tree) != -1)
+			if (find_item(&tree, &tree) != -1)
 				abort();
 
 			item_kill_tree(&tree);
 		}
 	}
 
-	if (radix_tree_locate_item(&tree, &tree) != -1)
+	if (find_item(&tree, &tree) != -1)
 		abort();
 	__locate_check(&tree, -1, 0);
-	if (radix_tree_locate_item(&tree, &tree) != -1)
+	if (find_item(&tree, &tree) != -1)
 		abort();
 	item_kill_tree(&tree);
 }
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 588209a..400de5c 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -347,7 +347,7 @@ static void __multiorder_join(unsigned long index,
 	item_insert_order(&tree, index, order2);
 	item = radix_tree_lookup(&tree, index);
 	radix_tree_join(&tree, index + 1, order1, item2);
-	loc = radix_tree_locate_item(&tree, item);
+	loc = find_item(&tree, item);
 	if (loc == -1)
 		free(item);
 	item = radix_tree_lookup(&tree, index + 1);
diff --git a/tools/testing/radix-tree/test.c b/tools/testing/radix-tree/test.c
index a6e8099..a68ed3b 100644
--- a/tools/testing/radix-tree/test.c
+++ b/tools/testing/radix-tree/test.c
@@ -142,6 +142,28 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
 	assert(nfound == 0);
 }
 
+/* Use the same pattern as find_swap_entry() in mm/shmem.c */
+unsigned long find_item(struct radix_tree_root *root, void *item)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+	unsigned long found = -1;
+	unsigned long checked = 0;
+
+	radix_tree_for_each_slot(slot, root, &iter, 0) {
+		if (*slot == item) {
+			found = iter.index;
+			break;
+		}
+		checked++;
+		if ((checked % 4) != 0)
+			continue;
+		slot = radix_tree_iter_next(slot, &iter);
+	}
+
+	return found;
+}
+
 static int verify_node(struct radix_tree_node *slot, unsigned int tag,
 			int tagged)
 {
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index b678f13..ccdd3c1 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -27,6 +27,8 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
 			unsigned long nr, int chunk);
 void item_kill_tree(struct radix_tree_root *root);
 
+unsigned long find_item(struct radix_tree_root *, void *item);
+
 void tag_check(void);
 void multiorder_checks(void);
 void iteration_test(void);
-- 
2.10.2

[PATCH 22/29] radix-tree: Delete radix_tree_range_tag_if_tagged()

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

This is an exceptionally complicated function with just one caller
(tag_pages_for_writeback).  We devote a large portion of the runtime
of the test suite to testing this one function which has one caller.
By introducing the new function radix_tree_iter_tag_set(), we
can eliminate all of the complexity while keeping the performance.
The caller can now use a fairly standard radix_tree_for_each() loop,
and it doesn't need to worry about tricksy things like 'start' wrapping.

The test suite continues to spend a large amount of time investigating
this function, but now it's testing the underlying primitives such as
radix_tree_iter_next() and the radix_tree_for_each_tagged() iterator
which are also used by other parts of the kernel.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 include/linux/radix-tree.h             |  74 ++++++++++-----------
 lib/radix-tree.c                       | 114 +++++----------------------------
 mm/page-writeback.c                    |  28 +++++---
 tools/testing/radix-tree/main.c        |  12 ++--
 tools/testing/radix-tree/multiorder.c  |  13 ++--
 tools/testing/radix-tree/regression2.c |   3 +-
 tools/testing/radix-tree/tag_check.c   |   4 +-
 tools/testing/radix-tree/test.c        |  25 ++++++++
 tools/testing/radix-tree/test.h        |   3 +
 9 files changed, 113 insertions(+), 163 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 57bf635..ca4eea1 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -124,6 +124,41 @@ static inline bool radix_tree_empty(struct radix_tree_root *root)
 }
 
 /**
+ * struct radix_tree_iter - radix tree iterator state
+ *
+ * @index:	index of current slot
+ * @next_index:	one beyond the last index for this chunk
+ * @tags:	bit-mask for tag-iterating
+ * @node:	node that contains current slot
+ * @shift:	shift for the node that holds our slots
+ *
+ * This radix tree iterator works in terms of "chunks" of slots.  A chunk is a
+ * subinterval of slots contained within one radix tree leaf node.  It is
+ * described by a pointer to its first slot and a struct radix_tree_iter
+ * which holds the chunk's position in the tree and its size.  For tagged
+ * iteration radix_tree_iter also holds the slots' bit-mask for one chosen
+ * radix tree tag.
+ */
+struct radix_tree_iter {
+	unsigned long	index;
+	unsigned long	next_index;
+	unsigned long	tags;
+	struct radix_tree_node *node;
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+	unsigned int	shift;
+#endif
+};
+
+static inline unsigned int iter_shift(struct radix_tree_iter *iter)
+{
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+	return iter->shift;
+#else
+	return 0;
+#endif
+}
+
+/**
  * Radix-tree synchronization
  *
  * The radix-tree API requires that users provide all synchronisation (with
@@ -293,6 +328,8 @@ void *radix_tree_tag_clear(struct radix_tree_root *root,
 			unsigned long index, unsigned int tag);
 int radix_tree_tag_get(struct radix_tree_root *root,
 			unsigned long index, unsigned int tag);
+void radix_tree_iter_tag_set(struct radix_tree_root *root,
+		const struct radix_tree_iter *iter, unsigned int tag);
 unsigned int
 radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
 		unsigned long first_index, unsigned int max_items,
@@ -301,10 +338,6 @@ unsigned int
 radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
 		unsigned long first_index, unsigned int max_items,
 		unsigned int tag);
-unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
-		unsigned long *first_indexp, unsigned long last_index,
-		unsigned long nr_to_tag,
-		unsigned int fromtag, unsigned int totag);
 int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
 
 static inline void radix_tree_preload_end(void)
@@ -318,39 +351,6 @@ int radix_tree_split(struct radix_tree_root *, unsigned long index,
 int radix_tree_join(struct radix_tree_root *, unsigned long index,
 			unsigned new_order, void *);
 
-/**
- * struct radix_tree_iter - radix tree iterator state
- *
- * @index:	index of current slot
- * @next_index:	one beyond the last index for this chunk
- * @tags:	bit-mask for tag-iterating
- * @shift:	shift for the node that holds our slots
- *
- * This radix tree iterator works in terms of "chunks" of slots.  A chunk is a
- * subinterval of slots contained within one radix tree leaf node.  It is
- * described by a pointer to its first slot and a struct radix_tree_iter
- * which holds the chunk's position in the tree and its size.  For tagged
- * iteration radix_tree_iter also holds the slots' bit-mask for one chosen
- * radix tree tag.
- */
-struct radix_tree_iter {
-	unsigned long	index;
-	unsigned long	next_index;
-	unsigned long	tags;
-#ifdef CONFIG_RADIX_TREE_MULTIORDER
-	unsigned int	shift;
-#endif
-};
-
-static inline unsigned int iter_shift(struct radix_tree_iter *iter)
-{
-#ifdef CONFIG_RADIX_TREE_MULTIORDER
-	return iter->shift;
-#else
-	return 0;
-#endif
-}
-
 #define RADIX_TREE_ITER_TAG_MASK	0x00FF	/* tag index in lower byte */
 #define RADIX_TREE_ITER_TAGGED		0x0100	/* lookup tagged slots */
 #define RADIX_TREE_ITER_CONTIG		0x0200	/* stop at first hole */
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 7e70ac9..c8ef657 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1049,6 +1049,20 @@ static void node_tag_set(struct radix_tree_root *root,
 }
 
 /**
+ * radix_tree_iter_tag_set - set a tag on the current iterator entry
+ *
+ * @root:	radix tree root
+ * @iter:	iterator state
+ * @tag:	tag to set
+ */
+void radix_tree_iter_tag_set(struct radix_tree_root *root,
+			const struct radix_tree_iter *iter, unsigned int tag)
+{
+	unsigned offset = (iter->index >> iter->shift) & RADIX_TREE_MAP_MASK;
+	node_tag_set(root, iter->node, tag, offset);
+}
+
+/**
  *	radix_tree_tag_clear - clear a tag on a radix tree node
  *	@root:		radix tree root
  *	@index:		index key
@@ -1176,6 +1190,7 @@ void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
 		if (node == RADIX_TREE_RETRY)
 			return slot;
 		node = entry_to_node(node);
+		iter->node = node;
 
 		if (flags & RADIX_TREE_ITER_TAGGED) {
 			unsigned tag_long, tag_bit;
@@ -1275,6 +1290,7 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 		iter->index = index;
 		iter->next_index = maxindex + 1;
 		iter->tags = 1;
+		iter->node = NULL;
 		__set_iter_shift(iter, 0);
 		return (void **)&root->rnode;
 	}
@@ -1319,6 +1335,7 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 	/* Update the iterator state */
 	iter->index = (index &~ node_maxindex(node)) | (offset << node->shift);
 	iter->next_index = (index | node_maxindex(node)) + 1;
+	iter->node = node;
 	__set_iter_shift(iter, node->shift);
 
 	/* Construct iter->tags bit-mask from node->tags[tag] array */
@@ -1344,103 +1361,6 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 EXPORT_SYMBOL(radix_tree_next_chunk);
 
 /**
- * radix_tree_range_tag_if_tagged - for each item in given range set given
- *				   tag if item has another tag set
- * @root:		radix tree root
- * @first_indexp:	pointer to a starting index of a range to scan
- * @last_index:		last index of a range to scan
- * @nr_to_tag:		maximum number items to tag
- * @iftag:		tag index to test
- * @settag:		tag index to set if tested tag is set
- *
- * This function scans range of radix tree from first_index to last_index
- * (inclusive).  For each item in the range if iftag is set, the function sets
- * also settag. The function stops either after tagging nr_to_tag items or
- * after reaching last_index.
- *
- * The tags must be set from the leaf level only and propagated back up the
- * path to the root. We must do this so that we resolve the full path before
- * setting any tags on intermediate nodes. If we set tags as we descend, then
- * we can get to the leaf node and find that the index that has the iftag
- * set is outside the range we are scanning. This reults in dangling tags and
- * can lead to problems with later tag operations (e.g. livelocks on lookups).
- *
- * The function returns the number of leaves where the tag was set and sets
- * *first_indexp to the first unscanned index.
- * WARNING! *first_indexp can wrap if last_index is ULONG_MAX. Caller must
- * be prepared to handle that.
- */
-unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
-		unsigned long *first_indexp, unsigned long last_index,
-		unsigned long nr_to_tag,
-		unsigned int iftag, unsigned int settag)
-{
-	struct radix_tree_node *node, *child;
-	unsigned long maxindex;
-	unsigned long tagged = 0;
-	unsigned long index = *first_indexp;
-
-	radix_tree_load_root(root, &child, &maxindex);
-	last_index = min(last_index, maxindex);
-	if (index > last_index)
-		return 0;
-	if (!nr_to_tag)
-		return 0;
-	if (!root_tag_get(root, iftag)) {
-		*first_indexp = last_index + 1;
-		return 0;
-	}
-	if (!radix_tree_is_internal_node(child)) {
-		*first_indexp = last_index + 1;
-		root_tag_set(root, settag);
-		return 1;
-	}
-
-	node = entry_to_node(child);
-
-	for (;;) {
-		unsigned offset = radix_tree_descend(node, &child, index);
-		if (!child)
-			goto next;
-		if (!tag_get(node, iftag, offset))
-			goto next;
-		/* Sibling slots never have tags set on them */
-		if (radix_tree_is_internal_node(child)) {
-			node = entry_to_node(child);
-			continue;
-		}
-
-		tagged++;
-		node_tag_set(root, node, settag, offset);
- next:
-		/* Go to next entry in node */
-		index = ((index >> node->shift) + 1) << node->shift;
-		/* Overflow can happen when last_index is ~0UL... */
-		if (index > last_index || !index)
-			break;
-		offset = (index >> node->shift) & RADIX_TREE_MAP_MASK;
-		while (offset == 0) {
-			/*
-			 * We've fully scanned this node. Go up. Because
-			 * last_index is guaranteed to be in the tree, what
-			 * we do below cannot wander astray.
-			 */
-			node = node->parent;
-			offset = (index >> node->shift) & RADIX_TREE_MAP_MASK;
-		}
-		if (is_sibling_entry(node, node->slots[offset]))
-			goto next;
-		if (tagged >= nr_to_tag)
-			break;
-	}
-
-	*first_indexp = index;
-
-	return tagged;
-}
-EXPORT_SYMBOL(radix_tree_range_tag_if_tagged);
-
-/**
  *	radix_tree_gang_lookup - perform multiple lookup on a radix tree
  *	@root:		radix tree root
  *	@results:	where the results of the lookup are placed
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 439cc63..c593715 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -2105,18 +2105,26 @@ void tag_pages_for_writeback(struct address_space *mapping,
 			     pgoff_t start, pgoff_t end)
 {
 #define WRITEBACK_TAG_BATCH 4096
-	unsigned long tagged;
-
-	do {
-		spin_lock_irq(&mapping->tree_lock);
-		tagged = radix_tree_range_tag_if_tagged(&mapping->page_tree,
-				&start, end, WRITEBACK_TAG_BATCH,
-				PAGECACHE_TAG_DIRTY, PAGECACHE_TAG_TOWRITE);
+	unsigned long tagged = 0;
+	struct radix_tree_iter iter;
+	void **slot;
+
+	spin_lock_irq(&mapping->tree_lock);
+	radix_tree_for_each_tagged(slot, &mapping->page_tree, &iter, start,
+							PAGECACHE_TAG_DIRTY) {
+		if (iter.index > end)
+			break;
+		radix_tree_iter_tag_set(&mapping->page_tree, &iter,
+							PAGECACHE_TAG_TOWRITE);
+		tagged++;
+		if ((tagged % WRITEBACK_TAG_BATCH) != 0)
+			continue;
+		slot = radix_tree_iter_next(slot, &iter);
 		spin_unlock_irq(&mapping->tree_lock);
-		WARN_ON_ONCE(tagged > WRITEBACK_TAG_BATCH);
 		cond_resched();
-		/* We check 'start' to handle wrapping when end == ~0UL */
-	} while (tagged >= WRITEBACK_TAG_BATCH && start);
+		spin_lock_irq(&mapping->tree_lock);
+	}
+	spin_unlock_irq(&mapping->tree_lock);
 }
 EXPORT_SYMBOL(tag_pages_for_writeback);
 
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 93a77f9..7d14812 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -205,8 +205,7 @@ void copy_tag_check(void)
 	}
 
 //	printf("\ncopying tags...\n");
-	cur = start;
-	tagged = radix_tree_range_tag_if_tagged(&tree, &cur, end, ITEMS, 0, 1);
+	tagged = tag_tagged_items(&tree, start, end, ITEMS, 0, 1);
 
 //	printf("checking copied tags\n");
 	assert(tagged == count);
@@ -214,16 +213,13 @@ void copy_tag_check(void)
 
 	/* Copy tags in several rounds */
 //	printf("\ncopying tags...\n");
-	cur = start;
-	do {
-		tmp = rand() % (count/10+2);
-		tagged = radix_tree_range_tag_if_tagged(&tree, &cur, end, tmp, 0, 2);
-	} while (tmp == tagged);
+	tmp = rand() % (count / 10 + 2);
+	tagged = tag_tagged_items(&tree, start, end, tmp, 0, 2);
+	assert(tagged == count);
 
 //	printf("%lu %lu %lu\n", tagged, tmp, count);
 //	printf("checking copied tags\n");
 	check_copied_tags(&tree, start, end, idx, ITEMS, 0, 2);
-	assert(tagged < tmp);
 	verify_tag_consistency(&tree, 0);
 	verify_tag_consistency(&tree, 1);
 	verify_tag_consistency(&tree, 2);
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 400de5c..6a0287d 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -26,7 +26,6 @@ static void __multiorder_tag_test(int index, int order)
 {
 	RADIX_TREE(tree, GFP_KERNEL);
 	int base, err, i;
-	unsigned long first = 0;
 
 	/* our canonical entry */
 	base = index & ~((1 << order) - 1);
@@ -60,7 +59,7 @@ static void __multiorder_tag_test(int index, int order)
 		assert(!radix_tree_tag_get(&tree, i, 1));
 	}
 
-	assert(radix_tree_range_tag_if_tagged(&tree, &first, ~0UL, 10, 0, 1) == 1);
+	assert(tag_tagged_items(&tree, 0, ~0UL, 10, 0, 1) == 1);
 	assert(radix_tree_tag_clear(&tree, index, 0));
 
 	for_each_index(i, base, order) {
@@ -252,7 +251,6 @@ void multiorder_tagged_iteration(void)
 	RADIX_TREE(tree, GFP_KERNEL);
 	struct radix_tree_iter iter;
 	void **slot;
-	unsigned long first = 0;
 	int i, j;
 
 	printf("Multiorder tagged iteration test\n");
@@ -297,8 +295,8 @@ void multiorder_tagged_iteration(void)
 		}
 	}
 
-	radix_tree_range_tag_if_tagged(&tree, &first, ~0UL,
-					MT_NUM_ENTRIES, 1, 2);
+	assert(tag_tagged_items(&tree, 0, ~0UL, TAG_ENTRIES, 1, 2) ==
+				TAG_ENTRIES);
 
 	for (j = 0; j < 256; j++) {
 		int mask, k;
@@ -325,9 +323,8 @@ void multiorder_tagged_iteration(void)
 		}
 	}
 
-	first = 1;
-	radix_tree_range_tag_if_tagged(&tree, &first, ~0UL,
-					MT_NUM_ENTRIES, 1, 0);
+	assert(tag_tagged_items(&tree, 1, ~0UL, MT_NUM_ENTRIES * 2, 1, 0)
+			== TAG_ENTRIES);
 	i = 0;
 	radix_tree_for_each_tagged(slot, &tree, &iter, 0, 0) {
 		assert(iter.index == tag_index[i]);
diff --git a/tools/testing/radix-tree/regression2.c b/tools/testing/radix-tree/regression2.c
index 63bf347..5cd2c0f 100644
--- a/tools/testing/radix-tree/regression2.c
+++ b/tools/testing/radix-tree/regression2.c
@@ -50,6 +50,7 @@
 #include <stdio.h>
 
 #include "regression.h"
+#include "test.h"
 
 #define PAGECACHE_TAG_DIRTY     0
 #define PAGECACHE_TAG_WRITEBACK 1
@@ -90,7 +91,7 @@ void regression2_test(void)
 	/* 1. */
 	start = 0;
 	end = max_slots - 2;
-	radix_tree_range_tag_if_tagged(&mt_tree, &start, end, 1,
+	tag_tagged_items(&mt_tree, start, end, 1,
 				PAGECACHE_TAG_DIRTY, PAGECACHE_TAG_TOWRITE);
 
 	/* 2. */
diff --git a/tools/testing/radix-tree/tag_check.c b/tools/testing/radix-tree/tag_check.c
index 186f6e4..b047b2d 100644
--- a/tools/testing/radix-tree/tag_check.c
+++ b/tools/testing/radix-tree/tag_check.c
@@ -23,7 +23,7 @@ __simple_checks(struct radix_tree_root *tree, unsigned long index, int tag)
 	item_tag_set(tree, index, tag);
 	ret = item_tag_get(tree, index, tag);
 	assert(ret != 0);
-	ret = radix_tree_range_tag_if_tagged(tree, &first, ~0UL, 10, tag, !tag);
+	ret = tag_tagged_items(tree, first, ~0UL, 10, tag, !tag);
 	assert(ret == 1);
 	ret = item_tag_get(tree, index, !tag);
 	assert(ret != 0);
@@ -320,7 +320,7 @@ static void single_check(void)
 	assert(ret == 0);
 	verify_tag_consistency(&tree, 0);
 	verify_tag_consistency(&tree, 1);
-	ret = radix_tree_range_tag_if_tagged(&tree, &first, 10, 10, 0, 1);
+	ret = tag_tagged_items(&tree, first, 10, 10, 0, 1);
 	assert(ret == 1);
 	ret = radix_tree_gang_lookup_tag(&tree, (void **)items, 0, BATCH, 1);
 	assert(ret == 1);
diff --git a/tools/testing/radix-tree/test.c b/tools/testing/radix-tree/test.c
index a68ed3b..322c34b 100644
--- a/tools/testing/radix-tree/test.c
+++ b/tools/testing/radix-tree/test.c
@@ -142,6 +142,31 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
 	assert(nfound == 0);
 }
 
+/* Use the same pattern as tag_pages_for_writeback() in mm/page-writeback.c */
+int tag_tagged_items(struct radix_tree_root *root,
+			unsigned long start, unsigned long end, unsigned batch,
+			unsigned iftag, unsigned thentag)
+{
+	unsigned long tagged = 0;
+	struct radix_tree_iter iter;
+	void **slot;
+
+	if (batch == 0)
+		batch = 1;
+
+	radix_tree_for_each_tagged(slot, root, &iter, start, iftag) {
+		if (iter.index > end)
+			break;
+		radix_tree_iter_tag_set(root, &iter, thentag);
+		tagged++;
+		if ((tagged % batch) != 0)
+			continue;
+		slot = radix_tree_iter_next(slot, &iter);
+	}
+
+	return tagged;
+}
+
 /* Use the same pattern as find_swap_entry() in mm/shmem.c */
 unsigned long find_item(struct radix_tree_root *root, void *item)
 {
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index ccdd3c1..1faf0a3 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -27,6 +27,9 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
 			unsigned long nr, int chunk);
 void item_kill_tree(struct radix_tree_root *root);
 
+int tag_tagged_items(struct radix_tree_root *, unsigned long start,
+			unsigned long end, unsigned batch,
+			unsigned iftag, unsigned thentag);
 unsigned long find_item(struct radix_tree_root *, void *item);
 
 void tag_check(void);
-- 
2.10.2

[PATCH 23/29] idr: Add ida_is_empty

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Two of the USB Gadgets were poking around in the internals of struct ida
in order to determine if it is empty.  Add the appropriate abstraction.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 drivers/usb/gadget/function/f_hid.c     | 6 +++---
 drivers/usb/gadget/function/f_printer.c | 6 +++---
 include/linux/idr.h                     | 5 +++++
 3 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/drivers/usb/gadget/function/f_hid.c b/drivers/usb/gadget/function/f_hid.c
index e2966f8..b9b04fc 100644
--- a/drivers/usb/gadget/function/f_hid.c
+++ b/drivers/usb/gadget/function/f_hid.c
@@ -840,7 +840,7 @@ static void hidg_free_inst(struct usb_function_instance *f)
 	mutex_lock(&hidg_ida_lock);
 
 	hidg_put_minor(opts->minor);
-	if (idr_is_empty(&hidg_ida.idr))
+	if (ida_is_empty(&hidg_ida))
 		ghid_cleanup();
 
 	mutex_unlock(&hidg_ida_lock);
@@ -866,7 +866,7 @@ static struct usb_function_instance *hidg_alloc_inst(void)
 
 	mutex_lock(&hidg_ida_lock);
 
-	if (idr_is_empty(&hidg_ida.idr)) {
+	if (ida_is_empty(&hidg_ida)) {
 		status = ghid_setup(NULL, HIDG_MINORS);
 		if (status)  {
 			ret = ERR_PTR(status);
@@ -879,7 +879,7 @@ static struct usb_function_instance *hidg_alloc_inst(void)
 	if (opts->minor < 0) {
 		ret = ERR_PTR(opts->minor);
 		kfree(opts);
-		if (idr_is_empty(&hidg_ida.idr))
+		if (ida_is_empty(&hidg_ida))
 			ghid_cleanup();
 		goto unlock;
 	}
diff --git a/drivers/usb/gadget/function/f_printer.c b/drivers/usb/gadget/function/f_printer.c
index 0de36cd..8054da9 100644
--- a/drivers/usb/gadget/function/f_printer.c
+++ b/drivers/usb/gadget/function/f_printer.c
@@ -1265,7 +1265,7 @@ static void gprinter_free_inst(struct usb_function_instance *f)
 	mutex_lock(&printer_ida_lock);
 
 	gprinter_put_minor(opts->minor);
-	if (idr_is_empty(&printer_ida.idr))
+	if (ida_is_empty(&printer_ida))
 		gprinter_cleanup();
 
 	mutex_unlock(&printer_ida_lock);
@@ -1289,7 +1289,7 @@ static struct usb_function_instance *gprinter_alloc_inst(void)
 
 	mutex_lock(&printer_ida_lock);
 
-	if (idr_is_empty(&printer_ida.idr)) {
+	if (ida_is_empty(&printer_ida)) {
 		status = gprinter_setup(PRINTER_MINORS);
 		if (status) {
 			ret = ERR_PTR(status);
@@ -1302,7 +1302,7 @@ static struct usb_function_instance *gprinter_alloc_inst(void)
 	if (opts->minor < 0) {
 		ret = ERR_PTR(opts->minor);
 		kfree(opts);
-		if (idr_is_empty(&printer_ida.idr))
+		if (ida_is_empty(&printer_ida))
 			gprinter_cleanup();
 		goto unlock;
 	}
diff --git a/include/linux/idr.h b/include/linux/idr.h
index 083d61e..3639a28 100644
--- a/include/linux/idr.h
+++ b/include/linux/idr.h
@@ -195,6 +195,11 @@ static inline int ida_get_new(struct ida *ida, int *p_id)
 	return ida_get_new_above(ida, 0, p_id);
 }
 
+static inline bool ida_is_empty(struct ida *ida)
+{
+	return idr_is_empty(&ida->idr);
+}
+
 void __init idr_init_cache(void);
 
 #endif /* __IDR_H__ */
-- 
2.10.2

[PATCH 24/29] tpm: Use idr_find(), not idr_find_slowpath()

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

idr_find_slowpath() is not intended to be part of the public API, it's
an implementation detail.  There's no reason to skip straight to the
slowpath here.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 drivers/char/tpm/tpm-chip.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
index e595013..447b6ca 100644
--- a/drivers/char/tpm/tpm-chip.c
+++ b/drivers/char/tpm/tpm-chip.c
@@ -84,7 +84,7 @@ EXPORT_SYMBOL_GPL(tpm_put_ops);
  *
  * The return'd chip has been tpm_try_get_ops'd and must be released via
  * tpm_put_ops
-  */
+ */
 struct tpm_chip *tpm_chip_find_get(int chip_num)
 {
 	struct tpm_chip *chip, *res = NULL;
@@ -103,7 +103,7 @@ struct tpm_chip *tpm_chip_find_get(int chip_num)
 			}
 		} while (chip_prev != chip_num);
 	} else {
-		chip = idr_find_slowpath(&dev_nums_idr, chip_num);
+		chip = idr_find(&dev_nums_idr, chip_num);
 		if (chip && !tpm_try_get_ops(chip))
 			res = chip;
 	}
-- 
2.10.2

[PATCH 25/29] rxrpc: Abstract away knowledge of IDR internals

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

Add idr_get_cursor() / idr_set_cursor() APIs, and remove the rounding
up to IDR_SIZE.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 include/linux/idr.h     | 26 ++++++++++++++++++++++++++
 net/rxrpc/af_rxrpc.c    | 11 ++++++-----
 net/rxrpc/conn_client.c |  4 ++--
 3 files changed, 34 insertions(+), 7 deletions(-)

diff --git a/include/linux/idr.h b/include/linux/idr.h
index 3639a28..1eb755f 100644
--- a/include/linux/idr.h
+++ b/include/linux/idr.h
@@ -56,6 +56,32 @@ struct idr {
 #define DEFINE_IDR(name)	struct idr name = IDR_INIT(name)
 
 /**
+ * idr_get_cursor - Return the current position of the cyclic allocator
+ * @idr: idr handle
+ *
+ * The value returned is the value that will be next returned from
+ * idr_alloc_cyclic() if it is free (otherwise the search will start from
+ * this position).
+ */
+static inline unsigned int idr_get_cursor(struct idr *idr)
+{
+	return READ_ONCE(idr->cur);
+}
+
+/**
+ * idr_set_cursor - Set the current position of the cyclic allocator
+ * @idr: idr handle
+ * @val: new position
+ *
+ * The next call to idr_alloc_cyclic() will return @val if it is free
+ * (otherwise the search will start from this position).
+ */
+static inline void idr_set_cursor(struct idr *idr, unsigned int val)
+{
+	WRITE_ONCE(idr->cur, val);
+}
+
+/**
  * DOC: idr sync
  * idr synchronization (stolen from radix-tree.h)
  *
diff --git a/net/rxrpc/af_rxrpc.c b/net/rxrpc/af_rxrpc.c
index 2d59c9b..5f63f6d 100644
--- a/net/rxrpc/af_rxrpc.c
+++ b/net/rxrpc/af_rxrpc.c
@@ -762,16 +762,17 @@ static const struct net_proto_family rxrpc_family_ops = {
 static int __init af_rxrpc_init(void)
 {
 	int ret = -1;
+	unsigned int tmp;
 
 	BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
 
 	get_random_bytes(&rxrpc_epoch, sizeof(rxrpc_epoch));
 	rxrpc_epoch |= RXRPC_RANDOM_EPOCH;
-	get_random_bytes(&rxrpc_client_conn_ids.cur,
-			 sizeof(rxrpc_client_conn_ids.cur));
-	rxrpc_client_conn_ids.cur &= 0x3fffffff;
-	if (rxrpc_client_conn_ids.cur == 0)
-		rxrpc_client_conn_ids.cur = 1;
+	get_random_bytes(&tmp, sizeof(tmp));
+	tmp &= 0x3fffffff;
+	if (tmp == 0)
+		tmp = 1;
+	idr_set_cursor(&rxrpc_client_conn_ids, tmp);
 
 	ret = -ENOMEM;
 	rxrpc_call_jar = kmem_cache_create(
diff --git a/net/rxrpc/conn_client.c b/net/rxrpc/conn_client.c
index 60ef960..9706f60 100644
--- a/net/rxrpc/conn_client.c
+++ b/net/rxrpc/conn_client.c
@@ -263,12 +263,12 @@ static bool rxrpc_may_reuse_conn(struct rxrpc_connection *conn)
 	 * times the maximum number of client conns away from the current
 	 * allocation point to try and keep the IDs concentrated.
 	 */
-	id_cursor = READ_ONCE(rxrpc_client_conn_ids.cur);
+	id_cursor = idr_get_cursor(&rxrpc_client_conn_ids);
 	id = conn->proto.cid >> RXRPC_CIDSHIFT;
 	distance = id - id_cursor;
 	if (distance < 0)
 		distance = -distance;
-	limit = round_up(rxrpc_max_client_connections, IDR_SIZE) * 4;
+	limit = rxrpc_max_client_connections * 4;
 	if (distance > limit)
 		goto mark_dont_reuse;
 
-- 
2.10.2

[PATCH 26/29] idr: Reduce the number of bits per level from 8 to 6

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

In preparation for merging the IDR and radix tree, reduce the fanout at
each level from 256 to 64.  If this causes a performance problem then
a bisect will point to this commit, and we'll have a better idea about
what we might do to fix it.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 include/linux/idr.h | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/include/linux/idr.h b/include/linux/idr.h
index 1eb755f..3c01b89 100644
--- a/include/linux/idr.h
+++ b/include/linux/idr.h
@@ -18,12 +18,11 @@
 #include <linux/rcupdate.h>
 
 /*
- * We want shallower trees and thus more bits covered at each layer.  8
- * bits gives us large enough first layer for most use cases and maximum
- * tree depth of 4.  Each idr_layer is slightly larger than 2k on 64bit and
- * 1k on 32bit.
+ * Using 6 bits at each layer allows us to allocate 7 layers out of each page.
+ * 8 bits only gave us 3 layers out of every pair of pages, which is less
+ * efficient except for trees with a largest element between 192-255 inclusive.
  */
-#define IDR_BITS 8
+#define IDR_BITS 6
 #define IDR_SIZE (1 << IDR_BITS)
 #define IDR_MASK ((1 << IDR_BITS)-1)
 
-- 
2.10.2

[PATCH 27/29] radix tree test suite: Add some more functionality

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

IDR needs more functionality from the kernel: kmalloc()/kfree(), and xchg().

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 tools/testing/radix-tree/linux.c        | 15 +++++++++++++++
 tools/testing/radix-tree/linux/kernel.h |  3 +++
 tools/testing/radix-tree/linux/slab.h   |  3 +++
 3 files changed, 21 insertions(+)

diff --git a/tools/testing/radix-tree/linux.c b/tools/testing/radix-tree/linux.c
index 1f32a16..ff0452e 100644
--- a/tools/testing/radix-tree/linux.c
+++ b/tools/testing/radix-tree/linux.c
@@ -54,6 +54,21 @@ void kmem_cache_free(struct kmem_cache *cachep, void *objp)
 	free(objp);
 }
 
+void *kmalloc(size_t size, gfp_t gfp)
+{
+	void *ret = malloc(size);
+	uatomic_inc(&nr_allocated);
+	return ret;
+}
+
+void kfree(void *p)
+{
+	if (!p)
+		return;
+	uatomic_dec(&nr_allocated);
+	free(p);
+}
+
 struct kmem_cache *
 kmem_cache_create(const char *name, size_t size, size_t offset,
 	unsigned long flags, void (*ctor)(void *))
diff --git a/tools/testing/radix-tree/linux/kernel.h b/tools/testing/radix-tree/linux/kernel.h
index 23e77f5..9b43b49 100644
--- a/tools/testing/radix-tree/linux/kernel.h
+++ b/tools/testing/radix-tree/linux/kernel.h
@@ -8,6 +8,7 @@
 #include <limits.h>
 
 #include "../../include/linux/compiler.h"
+#include "../../include/linux/err.h"
 #include "../../../include/linux/kconfig.h"
 
 #ifdef BENCHMARK
@@ -58,4 +59,6 @@ static inline int in_interrupt(void)
 #define round_up(x, y) ((((x)-1) | __round_mask(x, y))+1)
 #define round_down(x, y) ((x) & ~__round_mask(x, y))
 
+#define xchg(ptr, x)	uatomic_xchg(ptr, x)
+
 #endif /* _KERNEL_H */
diff --git a/tools/testing/radix-tree/linux/slab.h b/tools/testing/radix-tree/linux/slab.h
index 452e2bf..446639f 100644
--- a/tools/testing/radix-tree/linux/slab.h
+++ b/tools/testing/radix-tree/linux/slab.h
@@ -7,6 +7,9 @@
 #define SLAB_PANIC 2
 #define SLAB_RECLAIM_ACCOUNT    0x00020000UL            /* Objects are reclaimable */
 
+void *kmalloc(size_t size, gfp_t);
+void kfree(void *);
+
 struct kmem_cache {
 	int size;
 	void (*ctor)(void *);
-- 
2.10.2

[PATCH 28/29] radix-tree: Create all_tag_set

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

all_tag_set() sets every tag on a node.  This is useful for the IDR code
when we're creating new nodes which contain only free slots.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 lib/radix-tree.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index c8ef657..e063ca2 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -22,6 +22,8 @@
  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#include <linux/bitmap.h>
+#include <linux/bitops.h>
 #include <linux/errno.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
@@ -33,7 +35,6 @@
 #include <linux/notifier.h>
 #include <linux/cpu.h>
 #include <linux/string.h>
-#include <linux/bitops.h>
 #include <linux/rcupdate.h>
 #include <linux/preempt.h>		/* in_interrupt() */
 
@@ -184,6 +185,11 @@ static inline int any_tag_set(struct radix_tree_node *node, unsigned int tag)
 	return 0;
 }
 
+static inline void all_tag_set(struct radix_tree_node *node, unsigned int tag)
+{
+	bitmap_fill(node->tags[tag], RADIX_TREE_MAP_SIZE);
+}
+
 /**
  * radix_tree_find_next_bit - find the next set bit in a memory region
  *
-- 
2.10.2

[PATCH 29/29] Reimplement IDR and IDA using the radix tree

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

The IDR is very similar to the radix tree.  It has some functionality
that the radix tree did not have (alloc next free, cyclic allocation,
a callback-based for_each, destroy tree), which is readily implementable
on top of the radix tree.  A few small changes were needed in order to
use a tag to represent nodes with free space below them.

The IDA is reimplemented as a client of the newly enhanced radix tree.
As in the current implementation, it uses a bitmap at the last level of
the tree.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 include/linux/idr.h                     |  132 ++--
 include/linux/radix-tree.h              |    5 +-
 init/main.c                             |    3 +-
 lib/idr.c                               | 1075 -------------------------------
 lib/radix-tree.c                        |  624 ++++++++++++++++--
 tools/testing/radix-tree/Makefile       |    5 +-
 tools/testing/radix-tree/idr.c          |  148 +++++
 tools/testing/radix-tree/linux/idr.h    |    1 +
 tools/testing/radix-tree/linux/kernel.h |    2 +
 tools/testing/radix-tree/main.c         |    6 +
 tools/testing/radix-tree/test.h         |    2 +
 11 files changed, 791 insertions(+), 1212 deletions(-)
 create mode 100644 tools/testing/radix-tree/idr.c
 create mode 100644 tools/testing/radix-tree/linux/idr.h

diff --git a/include/linux/idr.h b/include/linux/idr.h
index 3c01b89..0035799 100644
--- a/include/linux/idr.h
+++ b/include/linux/idr.h
@@ -12,47 +12,22 @@
 #ifndef __IDR_H__
 #define __IDR_H__
 
-#include <linux/types.h>
-#include <linux/bitops.h>
-#include <linux/init.h>
-#include <linux/rcupdate.h>
-
-/*
- * Using 6 bits at each layer allows us to allocate 7 layers out of each page.
- * 8 bits only gave us 3 layers out of every pair of pages, which is less
- * efficient except for trees with a largest element between 192-255 inclusive.
- */
-#define IDR_BITS 6
-#define IDR_SIZE (1 << IDR_BITS)
-#define IDR_MASK ((1 << IDR_BITS)-1)
-
-struct idr_layer {
-	int			prefix;	/* the ID prefix of this idr_layer */
-	int			layer;	/* distance from leaf */
-	struct idr_layer __rcu	*ary[1<<IDR_BITS];
-	int			count;	/* When zero, we can release it */
-	union {
-		/* A zero bit means "space here" */
-		DECLARE_BITMAP(bitmap, IDR_SIZE);
-		struct rcu_head		rcu_head;
-	};
-};
+#include <linux/radix-tree.h>
+#include <linux/gfp.h>
 
 struct idr {
-	struct idr_layer __rcu	*hint;	/* the last layer allocated from */
-	struct idr_layer __rcu	*top;
-	int			layers;	/* only valid w/o concurrent changes */
-	int			cur;	/* current pos for cyclic allocation */
-	spinlock_t		lock;
-	int			id_free_cnt;
-	struct idr_layer	*id_free;
+	struct radix_tree_root	idr_rt;
+	unsigned int		idr_next;
 };
 
-#define IDR_INIT(name)							\
+/* Set the IDR flag and the IDR_FREE tag */
+#define IDR_RT_MARKER		((__force gfp_t)(3 << __GFP_BITS_SHIFT))
+
+#define IDR_INIT							\
 {									\
-	.lock			= __SPIN_LOCK_UNLOCKED(name.lock),	\
+	.idr_rt = RADIX_TREE_INIT(IDR_RT_MARKER)			\
 }
-#define DEFINE_IDR(name)	struct idr name = IDR_INIT(name)
+#define DEFINE_IDR(name)	struct idr name = IDR_INIT
 
 /**
  * idr_get_cursor - Return the current position of the cyclic allocator
@@ -64,7 +39,7 @@ struct idr {
  */
 static inline unsigned int idr_get_cursor(struct idr *idr)
 {
-	return READ_ONCE(idr->cur);
+	return READ_ONCE(idr->idr_next);
 }
 
 /**
@@ -77,7 +52,7 @@ static inline unsigned int idr_get_cursor(struct idr *idr)
  */
 static inline void idr_set_cursor(struct idr *idr, unsigned int val)
 {
-	WRITE_ONCE(idr->cur, val);
+	WRITE_ONCE(idr->idr_next, val);
 }
 
 /**
@@ -97,22 +72,30 @@ static inline void idr_set_cursor(struct idr *idr, unsigned int val)
  * period).
  */
 
-/*
- * This is what we export.
- */
-
-void *idr_find_slowpath(struct idr *idp, int id);
 void idr_preload(gfp_t gfp_mask);
-int idr_alloc(struct idr *idp, void *ptr, int start, int end, gfp_t gfp_mask);
-int idr_alloc_cyclic(struct idr *idr, void *ptr, int start, int end, gfp_t gfp_mask);
-int idr_for_each(struct idr *idp,
+int idr_alloc(struct idr *, void *, int start, int end, gfp_t gfp_mask);
+int idr_alloc_cyclic(struct idr *, void *, int start, int end, gfp_t gfp_mask);
+int idr_for_each(struct idr *,
 		 int (*fn)(int id, void *p, void *data), void *data);
-void *idr_get_next(struct idr *idp, int *nextid);
-void *idr_replace(struct idr *idp, void *ptr, int id);
-void idr_remove(struct idr *idp, int id);
-void idr_destroy(struct idr *idp);
-void idr_init(struct idr *idp);
-bool idr_is_empty(struct idr *idp);
+void *idr_get_next(struct idr *, int *nextid);
+void *idr_replace(struct idr *, void *, int id);
+void idr_destroy(struct idr *);
+
+static inline void idr_remove(struct idr *idp, int id)
+{
+	radix_tree_delete(&idp->idr_rt, id);
+}
+
+static inline void idr_init(struct idr *idp)
+{
+	memset(idp, 0, sizeof(*idp));
+	idp->idr_rt.gfp_mask = IDR_RT_MARKER;
+}
+
+static inline bool idr_is_empty(struct idr *idp)
+{
+	return radix_tree_empty(&idp->idr_rt);
+}
 
 /**
  * idr_preload_end - end preload section started with idr_preload()
@@ -139,17 +122,12 @@ static inline void idr_preload_end(void)
  */
 static inline void *idr_find(struct idr *idr, int id)
 {
-	struct idr_layer *hint = rcu_dereference_raw(idr->hint);
-
-	if (hint && (id & ~IDR_MASK) == hint->prefix)
-		return rcu_dereference_raw(hint->ary[id & IDR_MASK]);
-
-	return idr_find_slowpath(idr, id);
+	return radix_tree_lookup(&idr->idr_rt, id);
 }
 
 /**
  * idr_for_each_entry - iterate over an idr's elements of a given type
- * @idp:     idr handle
+ * @idr:     idr handle
  * @entry:   the type * to use as cursor
  * @id:      id entry's key
  *
@@ -157,57 +135,58 @@ static inline void *idr_find(struct idr *idr, int id)
  * after normal terminatinon @entry is left with the value NULL.  This
  * is convenient for a "not found" value.
  */
-#define idr_for_each_entry(idp, entry, id)			\
-	for (id = 0; ((entry) = idr_get_next(idp, &(id))) != NULL; ++id)
+#define idr_for_each_entry(idr, entry, id)			\
+	for (id = 0; ((entry) = idr_get_next(idr, &(id))) != NULL; ++id)
 
 /**
- * idr_for_each_entry - continue iteration over an idr's elements of a given type
- * @idp:     idr handle
+ * idr_for_each_entry_continue - continue iteration over an idr's elements of a given type
+ * @idr:     idr handle
  * @entry:   the type * to use as cursor
  * @id:      id entry's key
  *
  * Continue to iterate over list of given type, continuing after
  * the current position.
  */
-#define idr_for_each_entry_continue(idp, entry, id)			\
-	for ((entry) = idr_get_next((idp), &(id));			\
+#define idr_for_each_entry_continue(idr, entry, id)			\
+	for ((entry) = idr_get_next((idr), &(id));			\
 	     entry;							\
-	     ++id, (entry) = idr_get_next((idp), &(id)))
+	     ++id, (entry) = idr_get_next((idr), &(id)))
 
 /*
  * IDA - IDR based id allocator, use when translation from id to
  * pointer isn't necessary.
- *
- * IDA_BITMAP_LONGS is calculated to be one less to accommodate
- * ida_bitmap->nr_busy so that the whole struct fits in 128 bytes.
  */
 #define IDA_CHUNK_SIZE		128	/* 128 bytes per chunk */
-#define IDA_BITMAP_LONGS	(IDA_CHUNK_SIZE / sizeof(long) - 1)
+#define IDA_BITMAP_LONGS	(IDA_CHUNK_SIZE / sizeof(long))
 #define IDA_BITMAP_BITS 	(IDA_BITMAP_LONGS * sizeof(long) * 8)
 
 struct ida_bitmap {
-	long			nr_busy;
 	unsigned long		bitmap[IDA_BITMAP_LONGS];
 };
 
 struct ida {
-	struct idr		idr;
+	struct radix_tree_root	ida_rt;
 	struct ida_bitmap	*free_bitmap;
 };
 
-#define IDA_INIT(name)		{ .idr = IDR_INIT((name).idr), .free_bitmap = NULL, }
-#define DEFINE_IDA(name)	struct ida name = IDA_INIT(name)
+#define IDA_INIT		{ .ida_rt = RADIX_TREE_INIT(IDR_RT_MARKER), }
+#define DEFINE_IDA(name)	struct ida name = IDA_INIT
 
 int ida_pre_get(struct ida *ida, gfp_t gfp_mask);
 int ida_get_new_above(struct ida *ida, int starting_id, int *p_id);
 void ida_remove(struct ida *ida, int id);
 void ida_destroy(struct ida *ida);
-void ida_init(struct ida *ida);
 
 int ida_simple_get(struct ida *ida, unsigned int start, unsigned int end,
 		   gfp_t gfp_mask);
 void ida_simple_remove(struct ida *ida, unsigned int id);
 
+static inline void ida_init(struct ida *ida)
+{
+	memset(ida, 0, sizeof(*ida));
+	ida->ida_rt.gfp_mask = IDR_RT_MARKER;
+}
+
 /**
  * ida_get_new - allocate new ID
  * @ida:	idr handle
@@ -222,9 +201,6 @@ static inline int ida_get_new(struct ida *ida, int *p_id)
 
 static inline bool ida_is_empty(struct ida *ida)
 {
-	return idr_is_empty(&ida->idr);
+	return radix_tree_empty(&ida->ida_rt);
 }
-
-void __init idr_init_cache(void);
-
 #endif /* __IDR_H__ */
diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index ca4eea1..6483c73c 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -98,7 +98,10 @@ struct radix_tree_node {
 	unsigned long	tags[RADIX_TREE_MAX_TAGS][RADIX_TREE_TAG_LONGS];
 };
 
-/* root tags are stored in gfp_mask, shifted by __GFP_BITS_SHIFT */
+/* The top bits of gfp_mask are used to store the root tags and the IDR flag */
+#define ROOT_IS_IDR	(1 << __GFP_BITS_SHIFT)
+#define ROOT_TAG_SHIFT	(__GFP_BITS_SHIFT + 1)
+
 struct radix_tree_root {
 	gfp_t			gfp_mask;
 	struct radix_tree_node	__rcu *rnode;
diff --git a/init/main.c b/init/main.c
index 2858be7..2757b84 100644
--- a/init/main.c
+++ b/init/main.c
@@ -550,14 +550,13 @@ asmlinkage __visible void __init start_kernel(void)
 	if (WARN(!irqs_disabled(),
 		 "Interrupts were enabled *very* early, fixing it\n"))
 		local_irq_disable();
-	idr_init_cache();
+	radix_tree_init();
 	rcu_init();
 
 	/* trace_printk() and trace points may be used after this */
 	trace_init();
 
 	context_tracking_init();
-	radix_tree_init();
 	/* init some links before init_ISA_irqs() */
 	early_irq_init();
 	init_IRQ();
diff --git a/lib/idr.c b/lib/idr.c
index 6098336..3c88abb7 100644
--- a/lib/idr.c
+++ b/lib/idr.c
@@ -1,1068 +1,8 @@
-/*
- * 2002-10-18  written by Jim Houston jim.houston@ccur.com
- *	Copyright (C) 2002 by Concurrent Computer Corporation
- *	Distributed under the GNU GPL license version 2.
- *
- * Modified by George Anzinger to reuse immediately and to use
- * find bit instructions.  Also removed _irq on spinlocks.
- *
- * Modified by Nadia Derbey to make it RCU safe.
- *
- * Small id to pointer translation service.
- *
- * It uses a radix tree like structure as a sparse array indexed
- * by the id to obtain the pointer.  The bitmap makes allocating
- * a new id quick.
- *
- * You call it to allocate an id (an int) an associate with that id a
- * pointer or what ever, we treat it as a (void *).  You can pass this
- * id to a user for him to pass back at a later time.  You then pass
- * that id to this code and it returns your pointer.
- */
-
-#ifndef TEST                        // to test in user space...
-#include <linux/slab.h>
-#include <linux/init.h>
-#include <linux/export.h>
-#endif
-#include <linux/err.h>
-#include <linux/string.h>
 #include <linux/idr.h>
 #include <linux/spinlock.h>
-#include <linux/percpu.h>
-
-#define MAX_IDR_SHIFT		(sizeof(int) * 8 - 1)
-#define MAX_IDR_BIT		(1U << MAX_IDR_SHIFT)
-
-/* Leave the possibility of an incomplete final layer */
-#define MAX_IDR_LEVEL ((MAX_IDR_SHIFT + IDR_BITS - 1) / IDR_BITS)
-
-/* Number of id_layer structs to leave in free list */
-#define MAX_IDR_FREE (MAX_IDR_LEVEL * 2)
 
-static struct kmem_cache *idr_layer_cache;
-static DEFINE_PER_CPU(struct idr_layer *, idr_preload_head);
-static DEFINE_PER_CPU(int, idr_preload_cnt);
 static DEFINE_SPINLOCK(simple_ida_lock);
 
-/* the maximum ID which can be allocated given idr->layers */
-static int idr_max(int layers)
-{
-	int bits = min_t(int, layers * IDR_BITS, MAX_IDR_SHIFT);
-
-	return (1 << bits) - 1;
-}
-
-/*
- * Prefix mask for an idr_layer at @layer.  For layer 0, the prefix mask is
- * all bits except for the lower IDR_BITS.  For layer 1, 2 * IDR_BITS, and
- * so on.
- */
-static int idr_layer_prefix_mask(int layer)
-{
-	return ~idr_max(layer + 1);
-}
-
-static struct idr_layer *get_from_free_list(struct idr *idp)
-{
-	struct idr_layer *p;
-	unsigned long flags;
-
-	spin_lock_irqsave(&idp->lock, flags);
-	if ((p = idp->id_free)) {
-		idp->id_free = p->ary[0];
-		idp->id_free_cnt--;
-		p->ary[0] = NULL;
-	}
-	spin_unlock_irqrestore(&idp->lock, flags);
-	return(p);
-}
-
-/**
- * idr_layer_alloc - allocate a new idr_layer
- * @gfp_mask: allocation mask
- * @layer_idr: optional idr to allocate from
- *
- * If @layer_idr is %NULL, directly allocate one using @gfp_mask or fetch
- * one from the per-cpu preload buffer.  If @layer_idr is not %NULL, fetch
- * an idr_layer from @idr->id_free.
- *
- * @layer_idr is to maintain backward compatibility with the old alloc
- * interface - idr_pre_get() and idr_get_new*() - and will be removed
- * together with per-pool preload buffer.
- */
-static struct idr_layer *idr_layer_alloc(gfp_t gfp_mask, struct idr *layer_idr)
-{
-	struct idr_layer *new;
-
-	/* this is the old path, bypass to get_from_free_list() */
-	if (layer_idr)
-		return get_from_free_list(layer_idr);
-
-	/*
-	 * Try to allocate directly from kmem_cache.  We want to try this
-	 * before preload buffer; otherwise, non-preloading idr_alloc()
-	 * users will end up taking advantage of preloading ones.  As the
-	 * following is allowed to fail for preloaded cases, suppress
-	 * warning this time.
-	 */
-	new = kmem_cache_zalloc(idr_layer_cache, gfp_mask | __GFP_NOWARN);
-	if (new)
-		return new;
-
-	/*
-	 * Try to fetch one from the per-cpu preload buffer if in process
-	 * context.  See idr_preload() for details.
-	 */
-	if (!in_interrupt()) {
-		preempt_disable();
-		new = __this_cpu_read(idr_preload_head);
-		if (new) {
-			__this_cpu_write(idr_preload_head, new->ary[0]);
-			__this_cpu_dec(idr_preload_cnt);
-			new->ary[0] = NULL;
-		}
-		preempt_enable();
-		if (new)
-			return new;
-	}
-
-	/*
-	 * Both failed.  Try kmem_cache again w/o adding __GFP_NOWARN so
-	 * that memory allocation failure warning is printed as intended.
-	 */
-	return kmem_cache_zalloc(idr_layer_cache, gfp_mask);
-}
-
-static void idr_layer_rcu_free(struct rcu_head *head)
-{
-	struct idr_layer *layer;
-
-	layer = container_of(head, struct idr_layer, rcu_head);
-	kmem_cache_free(idr_layer_cache, layer);
-}
-
-static inline void free_layer(struct idr *idr, struct idr_layer *p)
-{
-	if (idr->hint == p)
-		RCU_INIT_POINTER(idr->hint, NULL);
-	call_rcu(&p->rcu_head, idr_layer_rcu_free);
-}
-
-/* only called when idp->lock is held */
-static void __move_to_free_list(struct idr *idp, struct idr_layer *p)
-{
-	p->ary[0] = idp->id_free;
-	idp->id_free = p;
-	idp->id_free_cnt++;
-}
-
-static void move_to_free_list(struct idr *idp, struct idr_layer *p)
-{
-	unsigned long flags;
-
-	/*
-	 * Depends on the return element being zeroed.
-	 */
-	spin_lock_irqsave(&idp->lock, flags);
-	__move_to_free_list(idp, p);
-	spin_unlock_irqrestore(&idp->lock, flags);
-}
-
-static void idr_mark_full(struct idr_layer **pa, int id)
-{
-	struct idr_layer *p = pa[0];
-	int l = 0;
-
-	__set_bit(id & IDR_MASK, p->bitmap);
-	/*
-	 * If this layer is full mark the bit in the layer above to
-	 * show that this part of the radix tree is full.  This may
-	 * complete the layer above and require walking up the radix
-	 * tree.
-	 */
-	while (bitmap_full(p->bitmap, IDR_SIZE)) {
-		if (!(p = pa[++l]))
-			break;
-		id = id >> IDR_BITS;
-		__set_bit((id & IDR_MASK), p->bitmap);
-	}
-}
-
-static int __idr_pre_get(struct idr *idp, gfp_t gfp_mask)
-{
-	while (idp->id_free_cnt < MAX_IDR_FREE) {
-		struct idr_layer *new;
-		new = kmem_cache_zalloc(idr_layer_cache, gfp_mask);
-		if (new == NULL)
-			return (0);
-		move_to_free_list(idp, new);
-	}
-	return 1;
-}
-
-/**
- * sub_alloc - try to allocate an id without growing the tree depth
- * @idp: idr handle
- * @starting_id: id to start search at
- * @pa: idr_layer[MAX_IDR_LEVEL] used as backtrack buffer
- * @gfp_mask: allocation mask for idr_layer_alloc()
- * @layer_idr: optional idr passed to idr_layer_alloc()
- *
- * Allocate an id in range [@starting_id, INT_MAX] from @idp without
- * growing its depth.  Returns
- *
- *  the allocated id >= 0 if successful,
- *  -EAGAIN if the tree needs to grow for allocation to succeed,
- *  -ENOSPC if the id space is exhausted,
- *  -ENOMEM if more idr_layers need to be allocated.
- */
-static int sub_alloc(struct idr *idp, int *starting_id, struct idr_layer **pa,
-		     gfp_t gfp_mask, struct idr *layer_idr)
-{
-	int n, m, sh;
-	struct idr_layer *p, *new;
-	int l, id, oid;
-
-	id = *starting_id;
- restart:
-	p = idp->top;
-	l = idp->layers;
-	pa[l--] = NULL;
-	while (1) {
-		/*
-		 * We run around this while until we reach the leaf node...
-		 */
-		n = (id >> (IDR_BITS*l)) & IDR_MASK;
-		m = find_next_zero_bit(p->bitmap, IDR_SIZE, n);
-		if (m == IDR_SIZE) {
-			/* no space available go back to previous layer. */
-			l++;
-			oid = id;
-			id = (id | ((1 << (IDR_BITS * l)) - 1)) + 1;
-
-			/* if already at the top layer, we need to grow */
-			if (id > idr_max(idp->layers)) {
-				*starting_id = id;
-				return -EAGAIN;
-			}
-			p = pa[l];
-			BUG_ON(!p);
-
-			/* If we need to go up one layer, continue the
-			 * loop; otherwise, restart from the top.
-			 */
-			sh = IDR_BITS * (l + 1);
-			if (oid >> sh == id >> sh)
-				continue;
-			else
-				goto restart;
-		}
-		if (m != n) {
-			sh = IDR_BITS*l;
-			id = ((id >> sh) ^ n ^ m) << sh;
-		}
-		if ((id >= MAX_IDR_BIT) || (id < 0))
-			return -ENOSPC;
-		if (l == 0)
-			break;
-		/*
-		 * Create the layer below if it is missing.
-		 */
-		if (!p->ary[m]) {
-			new = idr_layer_alloc(gfp_mask, layer_idr);
-			if (!new)
-				return -ENOMEM;
-			new->layer = l-1;
-			new->prefix = id & idr_layer_prefix_mask(new->layer);
-			rcu_assign_pointer(p->ary[m], new);
-			p->count++;
-		}
-		pa[l--] = p;
-		p = p->ary[m];
-	}
-
-	pa[l] = p;
-	return id;
-}
-
-static int idr_get_empty_slot(struct idr *idp, int starting_id,
-			      struct idr_layer **pa, gfp_t gfp_mask,
-			      struct idr *layer_idr)
-{
-	struct idr_layer *p, *new;
-	int layers, v, id;
-	unsigned long flags;
-
-	id = starting_id;
-build_up:
-	p = idp->top;
-	layers = idp->layers;
-	if (unlikely(!p)) {
-		if (!(p = idr_layer_alloc(gfp_mask, layer_idr)))
-			return -ENOMEM;
-		p->layer = 0;
-		layers = 1;
-	}
-	/*
-	 * Add a new layer to the top of the tree if the requested
-	 * id is larger than the currently allocated space.
-	 */
-	while (id > idr_max(layers)) {
-		layers++;
-		if (!p->count) {
-			/* special case: if the tree is currently empty,
-			 * then we grow the tree by moving the top node
-			 * upwards.
-			 */
-			p->layer++;
-			WARN_ON_ONCE(p->prefix);
-			continue;
-		}
-		if (!(new = idr_layer_alloc(gfp_mask, layer_idr))) {
-			/*
-			 * The allocation failed.  If we built part of
-			 * the structure tear it down.
-			 */
-			spin_lock_irqsave(&idp->lock, flags);
-			for (new = p; p && p != idp->top; new = p) {
-				p = p->ary[0];
-				new->ary[0] = NULL;
-				new->count = 0;
-				bitmap_clear(new->bitmap, 0, IDR_SIZE);
-				__move_to_free_list(idp, new);
-			}
-			spin_unlock_irqrestore(&idp->lock, flags);
-			return -ENOMEM;
-		}
-		new->ary[0] = p;
-		new->count = 1;
-		new->layer = layers-1;
-		new->prefix = id & idr_layer_prefix_mask(new->layer);
-		if (bitmap_full(p->bitmap, IDR_SIZE))
-			__set_bit(0, new->bitmap);
-		p = new;
-	}
-	rcu_assign_pointer(idp->top, p);
-	idp->layers = layers;
-	v = sub_alloc(idp, &id, pa, gfp_mask, layer_idr);
-	if (v == -EAGAIN)
-		goto build_up;
-	return(v);
-}
-
-/*
- * @id and @pa are from a successful allocation from idr_get_empty_slot().
- * Install the user pointer @ptr and mark the slot full.
- */
-static void idr_fill_slot(struct idr *idr, void *ptr, int id,
-			  struct idr_layer **pa)
-{
-	/* update hint used for lookup, cleared from free_layer() */
-	rcu_assign_pointer(idr->hint, pa[0]);
-
-	rcu_assign_pointer(pa[0]->ary[id & IDR_MASK], (struct idr_layer *)ptr);
-	pa[0]->count++;
-	idr_mark_full(pa, id);
-}
-
-
-/**
- * idr_preload - preload for idr_alloc()
- * @gfp_mask: allocation mask to use for preloading
- *
- * Preload per-cpu layer buffer for idr_alloc().  Can only be used from
- * process context and each idr_preload() invocation should be matched with
- * idr_preload_end().  Note that preemption is disabled while preloaded.
- *
- * The first idr_alloc() in the preloaded section can be treated as if it
- * were invoked with @gfp_mask used for preloading.  This allows using more
- * permissive allocation masks for idrs protected by spinlocks.
- *
- * For example, if idr_alloc() below fails, the failure can be treated as
- * if idr_alloc() were called with GFP_KERNEL rather than GFP_NOWAIT.
- *
- *	idr_preload(GFP_KERNEL);
- *	spin_lock(lock);
- *
- *	id = idr_alloc(idr, ptr, start, end, GFP_NOWAIT);
- *
- *	spin_unlock(lock);
- *	idr_preload_end();
- *	if (id < 0)
- *		error;
- */
-void idr_preload(gfp_t gfp_mask)
-{
-	/*
-	 * Consuming preload buffer from non-process context breaks preload
-	 * allocation guarantee.  Disallow usage from those contexts.
-	 */
-	WARN_ON_ONCE(in_interrupt());
-	might_sleep_if(gfpflags_allow_blocking(gfp_mask));
-
-	preempt_disable();
-
-	/*
-	 * idr_alloc() is likely to succeed w/o full idr_layer buffer and
-	 * return value from idr_alloc() needs to be checked for failure
-	 * anyway.  Silently give up if allocation fails.  The caller can
-	 * treat failures from idr_alloc() as if idr_alloc() were called
-	 * with @gfp_mask which should be enough.
-	 */
-	while (__this_cpu_read(idr_preload_cnt) < MAX_IDR_FREE) {
-		struct idr_layer *new;
-
-		preempt_enable();
-		new = kmem_cache_zalloc(idr_layer_cache, gfp_mask);
-		preempt_disable();
-		if (!new)
-			break;
-
-		/* link the new one to per-cpu preload list */
-		new->ary[0] = __this_cpu_read(idr_preload_head);
-		__this_cpu_write(idr_preload_head, new);
-		__this_cpu_inc(idr_preload_cnt);
-	}
-}
-EXPORT_SYMBOL(idr_preload);
-
-/**
- * idr_alloc - allocate new idr entry
- * @idr: the (initialized) idr
- * @ptr: pointer to be associated with the new id
- * @start: the minimum id (inclusive)
- * @end: the maximum id (exclusive, <= 0 for max)
- * @gfp_mask: memory allocation flags
- *
- * Allocate an id in [start, end) and associate it with @ptr.  If no ID is
- * available in the specified range, returns -ENOSPC.  On memory allocation
- * failure, returns -ENOMEM.
- *
- * Note that @end is treated as max when <= 0.  This is to always allow
- * using @start + N as @end as long as N is inside integer range.
- *
- * The user is responsible for exclusively synchronizing all operations
- * which may modify @idr.  However, read-only accesses such as idr_find()
- * or iteration can be performed under RCU read lock provided the user
- * destroys @ptr in RCU-safe way after removal from idr.
- */
-int idr_alloc(struct idr *idr, void *ptr, int start, int end, gfp_t gfp_mask)
-{
-	int max = end > 0 ? end - 1 : INT_MAX;	/* inclusive upper limit */
-	struct idr_layer *pa[MAX_IDR_LEVEL + 1];
-	int id;
-
-	might_sleep_if(gfpflags_allow_blocking(gfp_mask));
-
-	/* sanity checks */
-	if (WARN_ON_ONCE(start < 0))
-		return -EINVAL;
-	if (unlikely(max < start))
-		return -ENOSPC;
-
-	/* allocate id */
-	id = idr_get_empty_slot(idr, start, pa, gfp_mask, NULL);
-	if (unlikely(id < 0))
-		return id;
-	if (unlikely(id > max))
-		return -ENOSPC;
-
-	idr_fill_slot(idr, ptr, id, pa);
-	return id;
-}
-EXPORT_SYMBOL_GPL(idr_alloc);
-
-/**
- * idr_alloc_cyclic - allocate new idr entry in a cyclical fashion
- * @idr: the (initialized) idr
- * @ptr: pointer to be associated with the new id
- * @start: the minimum id (inclusive)
- * @end: the maximum id (exclusive, <= 0 for max)
- * @gfp_mask: memory allocation flags
- *
- * Essentially the same as idr_alloc, but prefers to allocate progressively
- * higher ids if it can. If the "cur" counter wraps, then it will start again
- * at the "start" end of the range and allocate one that has already been used.
- */
-int idr_alloc_cyclic(struct idr *idr, void *ptr, int start, int end,
-			gfp_t gfp_mask)
-{
-	int id;
-
-	id = idr_alloc(idr, ptr, max(start, idr->cur), end, gfp_mask);
-	if (id == -ENOSPC)
-		id = idr_alloc(idr, ptr, start, end, gfp_mask);
-
-	if (likely(id >= 0))
-		idr->cur = id + 1;
-	return id;
-}
-EXPORT_SYMBOL(idr_alloc_cyclic);
-
-static void idr_remove_warning(int id)
-{
-	WARN(1, "idr_remove called for id=%d which is not allocated.\n", id);
-}
-
-static void sub_remove(struct idr *idp, int shift, int id)
-{
-	struct idr_layer *p = idp->top;
-	struct idr_layer **pa[MAX_IDR_LEVEL + 1];
-	struct idr_layer ***paa = &pa[0];
-	struct idr_layer *to_free;
-	int n;
-
-	*paa = NULL;
-	*++paa = &idp->top;
-
-	while ((shift > 0) && p) {
-		n = (id >> shift) & IDR_MASK;
-		__clear_bit(n, p->bitmap);
-		*++paa = &p->ary[n];
-		p = p->ary[n];
-		shift -= IDR_BITS;
-	}
-	n = id & IDR_MASK;
-	if (likely(p != NULL && test_bit(n, p->bitmap))) {
-		__clear_bit(n, p->bitmap);
-		RCU_INIT_POINTER(p->ary[n], NULL);
-		to_free = NULL;
-		while(*paa && ! --((**paa)->count)){
-			if (to_free)
-				free_layer(idp, to_free);
-			to_free = **paa;
-			**paa-- = NULL;
-		}
-		if (!*paa)
-			idp->layers = 0;
-		if (to_free)
-			free_layer(idp, to_free);
-	} else
-		idr_remove_warning(id);
-}
-
-/**
- * idr_remove - remove the given id and free its slot
- * @idp: idr handle
- * @id: unique key
- */
-void idr_remove(struct idr *idp, int id)
-{
-	struct idr_layer *p;
-	struct idr_layer *to_free;
-
-	if (id < 0)
-		return;
-
-	if (id > idr_max(idp->layers)) {
-		idr_remove_warning(id);
-		return;
-	}
-
-	sub_remove(idp, (idp->layers - 1) * IDR_BITS, id);
-	if (idp->top && idp->top->count == 1 && (idp->layers > 1) &&
-	    idp->top->ary[0]) {
-		/*
-		 * Single child at leftmost slot: we can shrink the tree.
-		 * This level is not needed anymore since when layers are
-		 * inserted, they are inserted at the top of the existing
-		 * tree.
-		 */
-		to_free = idp->top;
-		p = idp->top->ary[0];
-		rcu_assign_pointer(idp->top, p);
-		--idp->layers;
-		to_free->count = 0;
-		bitmap_clear(to_free->bitmap, 0, IDR_SIZE);
-		free_layer(idp, to_free);
-	}
-}
-EXPORT_SYMBOL(idr_remove);
-
-static void __idr_remove_all(struct idr *idp)
-{
-	int n, id, max;
-	int bt_mask;
-	struct idr_layer *p;
-	struct idr_layer *pa[MAX_IDR_LEVEL + 1];
-	struct idr_layer **paa = &pa[0];
-
-	n = idp->layers * IDR_BITS;
-	*paa = idp->top;
-	RCU_INIT_POINTER(idp->top, NULL);
-	max = idr_max(idp->layers);
-
-	id = 0;
-	while (id >= 0 && id <= max) {
-		p = *paa;
-		while (n > IDR_BITS && p) {
-			n -= IDR_BITS;
-			p = p->ary[(id >> n) & IDR_MASK];
-			*++paa = p;
-		}
-
-		bt_mask = id;
-		id += 1 << n;
-		/* Get the highest bit that the above add changed from 0->1. */
-		while (n < fls(id ^ bt_mask)) {
-			if (*paa)
-				free_layer(idp, *paa);
-			n += IDR_BITS;
-			--paa;
-		}
-	}
-	idp->layers = 0;
-}
-
-/**
- * idr_destroy - release all cached layers within an idr tree
- * @idp: idr handle
- *
- * Free all id mappings and all idp_layers.  After this function, @idp is
- * completely unused and can be freed / recycled.  The caller is
- * responsible for ensuring that no one else accesses @idp during or after
- * idr_destroy().
- *
- * A typical clean-up sequence for objects stored in an idr tree will use
- * idr_for_each() to free all objects, if necessary, then idr_destroy() to
- * free up the id mappings and cached idr_layers.
- */
-void idr_destroy(struct idr *idp)
-{
-	__idr_remove_all(idp);
-
-	while (idp->id_free_cnt) {
-		struct idr_layer *p = get_from_free_list(idp);
-		kmem_cache_free(idr_layer_cache, p);
-	}
-}
-EXPORT_SYMBOL(idr_destroy);
-
-void *idr_find_slowpath(struct idr *idp, int id)
-{
-	int n;
-	struct idr_layer *p;
-
-	if (id < 0)
-		return NULL;
-
-	p = rcu_dereference_raw(idp->top);
-	if (!p)
-		return NULL;
-	n = (p->layer+1) * IDR_BITS;
-
-	if (id > idr_max(p->layer + 1))
-		return NULL;
-	BUG_ON(n == 0);
-
-	while (n > 0 && p) {
-		n -= IDR_BITS;
-		BUG_ON(n != p->layer*IDR_BITS);
-		p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
-	}
-	return((void *)p);
-}
-EXPORT_SYMBOL(idr_find_slowpath);
-
-/**
- * idr_for_each - iterate through all stored pointers
- * @idp: idr handle
- * @fn: function to be called for each pointer
- * @data: data passed back to callback function
- *
- * Iterate over the pointers registered with the given idr.  The
- * callback function will be called for each pointer currently
- * registered, passing the id, the pointer and the data pointer passed
- * to this function.  It is not safe to modify the idr tree while in
- * the callback, so functions such as idr_get_new and idr_remove are
- * not allowed.
- *
- * We check the return of @fn each time. If it returns anything other
- * than %0, we break out and return that value.
- *
- * The caller must serialize idr_for_each() vs idr_get_new() and idr_remove().
- */
-int idr_for_each(struct idr *idp,
-		 int (*fn)(int id, void *p, void *data), void *data)
-{
-	int n, id, max, error = 0;
-	struct idr_layer *p;
-	struct idr_layer *pa[MAX_IDR_LEVEL + 1];
-	struct idr_layer **paa = &pa[0];
-
-	n = idp->layers * IDR_BITS;
-	*paa = rcu_dereference_raw(idp->top);
-	max = idr_max(idp->layers);
-
-	id = 0;
-	while (id >= 0 && id <= max) {
-		p = *paa;
-		while (n > 0 && p) {
-			n -= IDR_BITS;
-			p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
-			*++paa = p;
-		}
-
-		if (p) {
-			error = fn(id, (void *)p, data);
-			if (error)
-				break;
-		}
-
-		id += 1 << n;
-		while (n < fls(id)) {
-			n += IDR_BITS;
-			--paa;
-		}
-	}
-
-	return error;
-}
-EXPORT_SYMBOL(idr_for_each);
-
-/**
- * idr_get_next - lookup next object of id to given id.
- * @idp: idr handle
- * @nextidp:  pointer to lookup key
- *
- * Returns pointer to registered object with id, which is next number to
- * given id. After being looked up, *@nextidp will be updated for the next
- * iteration.
- *
- * This function can be called under rcu_read_lock(), given that the leaf
- * pointers lifetimes are correctly managed.
- */
-void *idr_get_next(struct idr *idp, int *nextidp)
-{
-	struct idr_layer *p, *pa[MAX_IDR_LEVEL + 1];
-	struct idr_layer **paa = &pa[0];
-	int id = *nextidp;
-	int n, max;
-
-	/* find first ent */
-	p = *paa = rcu_dereference_raw(idp->top);
-	if (!p)
-		return NULL;
-	n = (p->layer + 1) * IDR_BITS;
-	max = idr_max(p->layer + 1);
-
-	while (id >= 0 && id <= max) {
-		p = *paa;
-		while (n > 0 && p) {
-			n -= IDR_BITS;
-			p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
-			*++paa = p;
-		}
-
-		if (p) {
-			*nextidp = id;
-			return p;
-		}
-
-		/*
-		 * Proceed to the next layer at the current level.  Unlike
-		 * idr_for_each(), @id isn't guaranteed to be aligned to
-		 * layer boundary at this point and adding 1 << n may
-		 * incorrectly skip IDs.  Make sure we jump to the
-		 * beginning of the next layer using round_up().
-		 */
-		id = round_up(id + 1, 1 << n);
-		while (n < fls(id)) {
-			n += IDR_BITS;
-			--paa;
-		}
-	}
-	return NULL;
-}
-EXPORT_SYMBOL(idr_get_next);
-
-
-/**
- * idr_replace - replace pointer for given id
- * @idp: idr handle
- * @ptr: pointer you want associated with the id
- * @id: lookup key
- *
- * Replace the pointer registered with an id and return the old value.
- * A %-ENOENT return indicates that @id was not found.
- * A %-EINVAL return indicates that @id was not within valid constraints.
- *
- * The caller must serialize with writers.
- */
-void *idr_replace(struct idr *idp, void *ptr, int id)
-{
-	int n;
-	struct idr_layer *p, *old_p;
-
-	if (id < 0)
-		return ERR_PTR(-EINVAL);
-
-	p = idp->top;
-	if (!p)
-		return ERR_PTR(-ENOENT);
-
-	if (id > idr_max(p->layer + 1))
-		return ERR_PTR(-ENOENT);
-
-	n = p->layer * IDR_BITS;
-	while ((n > 0) && p) {
-		p = p->ary[(id >> n) & IDR_MASK];
-		n -= IDR_BITS;
-	}
-
-	n = id & IDR_MASK;
-	if (unlikely(p == NULL || !test_bit(n, p->bitmap)))
-		return ERR_PTR(-ENOENT);
-
-	old_p = p->ary[n];
-	rcu_assign_pointer(p->ary[n], ptr);
-
-	return old_p;
-}
-EXPORT_SYMBOL(idr_replace);
-
-void __init idr_init_cache(void)
-{
-	idr_layer_cache = kmem_cache_create("idr_layer_cache",
-				sizeof(struct idr_layer), 0, SLAB_PANIC, NULL);
-}
-
-/**
- * idr_init - initialize idr handle
- * @idp:	idr handle
- *
- * This function is use to set up the handle (@idp) that you will pass
- * to the rest of the functions.
- */
-void idr_init(struct idr *idp)
-{
-	memset(idp, 0, sizeof(struct idr));
-	spin_lock_init(&idp->lock);
-}
-EXPORT_SYMBOL(idr_init);
-
-static int idr_has_entry(int id, void *p, void *data)
-{
-	return 1;
-}
-
-bool idr_is_empty(struct idr *idp)
-{
-	return !idr_for_each(idp, idr_has_entry, NULL);
-}
-EXPORT_SYMBOL(idr_is_empty);
-
-/**
- * DOC: IDA description
- * IDA - IDR based ID allocator
- *
- * This is id allocator without id -> pointer translation.  Memory
- * usage is much lower than full blown idr because each id only
- * occupies a bit.  ida uses a custom leaf node which contains
- * IDA_BITMAP_BITS slots.
- *
- * 2007-04-25  written by Tejun Heo <htejun@gmail.com>
- */
-
-static void free_bitmap(struct ida *ida, struct ida_bitmap *bitmap)
-{
-	unsigned long flags;
-
-	if (!ida->free_bitmap) {
-		spin_lock_irqsave(&ida->idr.lock, flags);
-		if (!ida->free_bitmap) {
-			ida->free_bitmap = bitmap;
-			bitmap = NULL;
-		}
-		spin_unlock_irqrestore(&ida->idr.lock, flags);
-	}
-
-	kfree(bitmap);
-}
-
-/**
- * ida_pre_get - reserve resources for ida allocation
- * @ida:	ida handle
- * @gfp_mask:	memory allocation flag
- *
- * This function should be called prior to locking and calling the
- * following function.  It preallocates enough memory to satisfy the
- * worst possible allocation.
- *
- * If the system is REALLY out of memory this function returns %0,
- * otherwise %1.
- */
-int ida_pre_get(struct ida *ida, gfp_t gfp_mask)
-{
-	/* allocate idr_layers */
-	if (!__idr_pre_get(&ida->idr, gfp_mask))
-		return 0;
-
-	/* allocate free_bitmap */
-	if (!ida->free_bitmap) {
-		struct ida_bitmap *bitmap;
-
-		bitmap = kmalloc(sizeof(struct ida_bitmap), gfp_mask);
-		if (!bitmap)
-			return 0;
-
-		free_bitmap(ida, bitmap);
-	}
-
-	return 1;
-}
-EXPORT_SYMBOL(ida_pre_get);
-
-/**
- * ida_get_new_above - allocate new ID above or equal to a start id
- * @ida:	ida handle
- * @starting_id: id to start search at
- * @p_id:	pointer to the allocated handle
- *
- * Allocate new ID above or equal to @starting_id.  It should be called
- * with any required locks.
- *
- * If memory is required, it will return %-EAGAIN, you should unlock
- * and go back to the ida_pre_get() call.  If the ida is full, it will
- * return %-ENOSPC.
- *
- * @p_id returns a value in the range @starting_id ... %0x7fffffff.
- */
-int ida_get_new_above(struct ida *ida, int starting_id, int *p_id)
-{
-	struct idr_layer *pa[MAX_IDR_LEVEL + 1];
-	struct ida_bitmap *bitmap;
-	unsigned long flags;
-	int idr_id = starting_id / IDA_BITMAP_BITS;
-	int offset = starting_id % IDA_BITMAP_BITS;
-	int t, id;
-
- restart:
-	/* get vacant slot */
-	t = idr_get_empty_slot(&ida->idr, idr_id, pa, 0, &ida->idr);
-	if (t < 0)
-		return t == -ENOMEM ? -EAGAIN : t;
-
-	if (t * IDA_BITMAP_BITS >= MAX_IDR_BIT)
-		return -ENOSPC;
-
-	if (t != idr_id)
-		offset = 0;
-	idr_id = t;
-
-	/* if bitmap isn't there, create a new one */
-	bitmap = (void *)pa[0]->ary[idr_id & IDR_MASK];
-	if (!bitmap) {
-		spin_lock_irqsave(&ida->idr.lock, flags);
-		bitmap = ida->free_bitmap;
-		ida->free_bitmap = NULL;
-		spin_unlock_irqrestore(&ida->idr.lock, flags);
-
-		if (!bitmap)
-			return -EAGAIN;
-
-		memset(bitmap, 0, sizeof(struct ida_bitmap));
-		rcu_assign_pointer(pa[0]->ary[idr_id & IDR_MASK],
-				(void *)bitmap);
-		pa[0]->count++;
-	}
-
-	/* lookup for empty slot */
-	t = find_next_zero_bit(bitmap->bitmap, IDA_BITMAP_BITS, offset);
-	if (t == IDA_BITMAP_BITS) {
-		/* no empty slot after offset, continue to the next chunk */
-		idr_id++;
-		offset = 0;
-		goto restart;
-	}
-
-	id = idr_id * IDA_BITMAP_BITS + t;
-	if (id >= MAX_IDR_BIT)
-		return -ENOSPC;
-
-	__set_bit(t, bitmap->bitmap);
-	if (++bitmap->nr_busy == IDA_BITMAP_BITS)
-		idr_mark_full(pa, idr_id);
-
-	*p_id = id;
-
-	/* Each leaf node can handle nearly a thousand slots and the
-	 * whole idea of ida is to have small memory foot print.
-	 * Throw away extra resources one by one after each successful
-	 * allocation.
-	 */
-	if (ida->idr.id_free_cnt || ida->free_bitmap) {
-		struct idr_layer *p = get_from_free_list(&ida->idr);
-		if (p)
-			kmem_cache_free(idr_layer_cache, p);
-	}
-
-	return 0;
-}
-EXPORT_SYMBOL(ida_get_new_above);
-
-/**
- * ida_remove - remove the given ID
- * @ida:	ida handle
- * @id:		ID to free
- */
-void ida_remove(struct ida *ida, int id)
-{
-	struct idr_layer *p = ida->idr.top;
-	int shift = (ida->idr.layers - 1) * IDR_BITS;
-	int idr_id = id / IDA_BITMAP_BITS;
-	int offset = id % IDA_BITMAP_BITS;
-	int n;
-	struct ida_bitmap *bitmap;
-
-	if (idr_id > idr_max(ida->idr.layers))
-		goto err;
-
-	/* clear full bits while looking up the leaf idr_layer */
-	while ((shift > 0) && p) {
-		n = (idr_id >> shift) & IDR_MASK;
-		__clear_bit(n, p->bitmap);
-		p = p->ary[n];
-		shift -= IDR_BITS;
-	}
-
-	if (p == NULL)
-		goto err;
-
-	n = idr_id & IDR_MASK;
-	__clear_bit(n, p->bitmap);
-
-	bitmap = (void *)p->ary[n];
-	if (!bitmap || !test_bit(offset, bitmap->bitmap))
-		goto err;
-
-	/* update bitmap and remove it if empty */
-	__clear_bit(offset, bitmap->bitmap);
-	if (--bitmap->nr_busy == 0) {
-		__set_bit(n, p->bitmap);	/* to please idr_remove() */
-		idr_remove(&ida->idr, idr_id);
-		free_bitmap(ida, bitmap);
-	}
-
-	return;
-
- err:
-	WARN(1, "ida_remove called for id=%d which is not allocated.\n", id);
-}
-EXPORT_SYMBOL(ida_remove);
-
-/**
- * ida_destroy - release all cached layers within an ida tree
- * @ida:		ida handle
- */
-void ida_destroy(struct ida *ida)
-{
-	idr_destroy(&ida->idr);
-	kfree(ida->free_bitmap);
-}
-EXPORT_SYMBOL(ida_destroy);
-
 /**
  * ida_simple_get - get a new id.
  * @ida: the (initialized) ida.
@@ -1130,18 +70,3 @@ void ida_simple_remove(struct ida *ida, unsigned int id)
 	spin_unlock_irqrestore(&simple_ida_lock, flags);
 }
 EXPORT_SYMBOL(ida_simple_remove);
-
-/**
- * ida_init - initialize ida handle
- * @ida:	ida handle
- *
- * This function is use to set up the handle (@ida) that you will pass
- * to the rest of the functions.
- */
-void ida_init(struct ida *ida)
-{
-	memset(ida, 0, sizeof(struct ida));
-	idr_init(&ida->idr);
-
-}
-EXPORT_SYMBOL(ida_init);
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index e063ca2..3160de3 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -24,19 +24,20 @@
 
 #include <linux/bitmap.h>
 #include <linux/bitops.h>
+#include <linux/cpu.h>
 #include <linux/errno.h>
+#include <linux/export.h>
+#include <linux/idr.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
-#include <linux/export.h>
-#include <linux/radix-tree.h>
-#include <linux/percpu.h>
-#include <linux/slab.h>
 #include <linux/kmemleak.h>
 #include <linux/notifier.h>
-#include <linux/cpu.h>
-#include <linux/string.h>
-#include <linux/rcupdate.h>
+#include <linux/percpu.h>
 #include <linux/preempt.h>		/* in_interrupt() */
+#include <linux/radix-tree.h>
+#include <linux/rcupdate.h>
+#include <linux/slab.h>
+#include <linux/string.h>
 
 
 /* Number of nodes in fully populated tree of given height */
@@ -61,6 +62,15 @@ static struct kmem_cache *radix_tree_node_cachep;
 #define RADIX_TREE_PRELOAD_SIZE (RADIX_TREE_MAX_PATH * 2 - 1)
 
 /*
+ * The IDR does not have to be as high since it can only store a 31-bit integer
+ * at its maximum height
+ */
+#define IDR_INDEX_BITS		(8 /* CHAR_BIT */ * sizeof(int) - 1)
+#define IDR_MAX_PATH		(DIV_ROUND_UP(IDR_INDEX_BITS, \
+						RADIX_TREE_MAP_SHIFT))
+#define IDR_PRELOAD_SIZE	(IDR_MAX_PATH * 2 - 1)
+
+/*
  * Per-cpu pool of preloaded nodes
  */
 struct radix_tree_preload {
@@ -148,27 +158,38 @@ static inline int tag_get(struct radix_tree_node *node, unsigned int tag,
 
 static inline void root_tag_set(struct radix_tree_root *root, unsigned int tag)
 {
-	root->gfp_mask |= (__force gfp_t)(1 << (tag + __GFP_BITS_SHIFT));
+	root->gfp_mask |= (__force gfp_t)(1 << (tag + ROOT_TAG_SHIFT));
 }
 
 static inline void root_tag_clear(struct radix_tree_root *root, unsigned tag)
 {
-	root->gfp_mask &= (__force gfp_t)~(1 << (tag + __GFP_BITS_SHIFT));
+	root->gfp_mask &= (__force gfp_t)~(1 << (tag + ROOT_TAG_SHIFT));
 }
 
 static inline void root_tag_clear_all(struct radix_tree_root *root)
 {
-	root->gfp_mask &= __GFP_BITS_MASK;
+	root->gfp_mask &= (1 << ROOT_TAG_SHIFT) - 1;
 }
 
 static inline int root_tag_get(struct radix_tree_root *root, unsigned int tag)
 {
-	return (__force int)root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT));
+	return (__force int)root->gfp_mask & (1 << (tag + ROOT_TAG_SHIFT));
 }
 
 static inline unsigned root_tags_get(struct radix_tree_root *root)
 {
-	return (__force unsigned)root->gfp_mask >> __GFP_BITS_SHIFT;
+	return (__force unsigned)root->gfp_mask >> ROOT_TAG_SHIFT;
+}
+
+/*
+ * IDRs do not expose the tagging functionality of the radix tree to their
+ * users.  Reuse tag 0 to track whether a node has free space below it.
+ */
+#define IDR_FREE	0
+
+static inline bool is_idr(struct radix_tree_root *root)
+{
+	return (__force unsigned)root->gfp_mask & ROOT_IS_IDR;
 }
 
 /*
@@ -238,6 +259,13 @@ static inline unsigned long node_maxindex(struct radix_tree_node *node)
 	return shift_maxindex(node->shift);
 }
 
+static unsigned long next_index(unsigned long index,
+				struct radix_tree_node *node,
+				unsigned long offset)
+{
+	return (index & ~node_maxindex(node)) + (offset << node->shift);
+}
+
 #ifndef __KERNEL__
 static void dump_node(struct radix_tree_node *node, unsigned long index)
 {
@@ -276,11 +304,47 @@ static void radix_tree_dump(struct radix_tree_root *root)
 {
 	pr_debug("radix root: %p rnode %p tags %x\n",
 			root, root->rnode,
-			root->gfp_mask >> __GFP_BITS_SHIFT);
+			root->gfp_mask >> ROOT_TAG_SHIFT);
 	if (!radix_tree_is_internal_node(root->rnode))
 		return;
 	dump_node(entry_to_node(root->rnode), 0);
 }
+
+static void dump_ida_node(void *entry, unsigned long index)
+{
+	unsigned long i;
+
+	if (!entry)
+		return;
+
+	if (radix_tree_is_internal_node(entry)) {
+		struct radix_tree_node *node = entry_to_node(entry);
+
+		pr_debug("ida node: %p offset %d indices %lu-%lu parent %p free %lx shift %d count %d\n",
+			node, node->offset, index, index | node_maxindex(node),
+			node->parent, node->tags[0][0], node->shift,
+			node->count);
+		for (i = 0; i < RADIX_TREE_MAP_SIZE; i++)
+			dump_ida_node(node->slots[i],
+					index | (i << node->shift));
+	} else {
+		struct ida_bitmap *bitmap = entry;
+
+		pr_debug("ida btmp: %p index %lu data", bitmap, index);
+		for (i = 0; i < IDA_BITMAP_LONGS; i++)
+			pr_cont(" %lx", bitmap->bitmap[i]);
+		pr_cont("\n");
+	}
+}
+
+static void ida_dump(struct ida *ida)
+{
+	struct radix_tree_root *root = &ida->ida_rt;
+	pr_debug("ida: %p %p free %d bitmap %p\n", ida, root->rnode,
+				root->gfp_mask >> ROOT_TAG_SHIFT,
+				ida->free_bitmap);
+	dump_ida_node(root->rnode, 0);
+}
 #endif
 
 /*
@@ -288,10 +352,9 @@ static void radix_tree_dump(struct radix_tree_root *root)
  * that the caller has pinned this thread of control to the current CPU.
  */
 static struct radix_tree_node *
-radix_tree_node_alloc(struct radix_tree_root *root)
+radix_tree_node_alloc(gfp_t gfp_mask)
 {
 	struct radix_tree_node *ret = NULL;
-	gfp_t gfp_mask = root_gfp_mask(root);
 
 	/*
 	 * Preload code isn't irq safe and it doesn't make sense to use
@@ -521,7 +584,7 @@ static unsigned radix_tree_load_root(struct radix_tree_root *root,
 /*
  *	Extend a radix tree so it can store key @index.
  */
-static int radix_tree_extend(struct radix_tree_root *root,
+static int radix_tree_extend(struct radix_tree_root *root, gfp_t gfp_mask,
 				unsigned long index, unsigned int shift)
 {
 	struct radix_tree_node *slot;
@@ -538,15 +601,22 @@ static int radix_tree_extend(struct radix_tree_root *root,
 		goto out;
 
 	do {
-		struct radix_tree_node *node = radix_tree_node_alloc(root);
+		struct radix_tree_node *node = radix_tree_node_alloc(gfp_mask);
 
 		if (!node)
 			return -ENOMEM;
 
-		/* Propagate the aggregated tag info into the new root */
-		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++) {
-			if (root_tag_get(root, tag))
-				tag_set(node, tag, 0);
+		if (is_idr(root)) {
+			all_tag_set(node, IDR_FREE);
+			if (!root_tag_get(root, IDR_FREE))
+				tag_clear(node, IDR_FREE, 0);
+			root_tag_set(root, IDR_FREE);
+		} else {
+			/* Propagate the aggregated tag info to the new child */
+			for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++) {
+				if (root_tag_get(root, tag))
+					tag_set(node, tag, 0);
+			}
 		}
 
 		BUG_ON(shift > BITS_PER_LONG);
@@ -565,26 +635,9 @@ static int radix_tree_extend(struct radix_tree_root *root,
 	return maxshift + RADIX_TREE_MAP_SHIFT;
 }
 
-/**
- *	__radix_tree_create	-	create a slot in a radix tree
- *	@root:		radix tree root
- *	@index:		index key
- *	@order:		index occupies 2^order aligned slots
- *	@nodep:		returns node
- *	@slotp:		returns slot
- *
- *	Create, if necessary, and return the node and slot for an item
- *	at position @index in the radix tree @root.
- *
- *	Until there is more than one item in the tree, no nodes are
- *	allocated and @root->rnode is used as a direct slot instead of
- *	pointing to a node, in which case *@nodep will be NULL.
- *
- *	Returns -ENOMEM, or 0 for success.
- */
-int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
-			unsigned order, struct radix_tree_node **nodep,
-			void ***slotp)
+static int _radix_tree_create(struct radix_tree_root *root, gfp_t gfp_mask,
+			unsigned long index, unsigned int order,
+			struct radix_tree_node **nodep, void ***slotp)
 {
 	struct radix_tree_node *node = NULL, *child;
 	void **slot = (void **)&root->rnode;
@@ -598,7 +651,7 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 	if (order > 0 && max == ((1UL << order) - 1))
 		max++;
 	if (max > maxindex) {
-		int error = radix_tree_extend(root, max, shift);
+		int error = radix_tree_extend(root, gfp_mask, max, shift);
 		if (error < 0)
 			return error;
 		shift = error;
@@ -609,7 +662,7 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 		shift -= RADIX_TREE_MAP_SHIFT;
 		if (child == NULL) {
 			/* Have to add a child node.  */
-			child = radix_tree_node_alloc(root);
+			child = radix_tree_node_alloc(gfp_mask);
 			if (!child)
 				return -ENOMEM;
 			child->shift = shift;
@@ -635,7 +688,6 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 	return 0;
 }
 
-#ifdef CONFIG_RADIX_TREE_MULTIORDER
 /*
  * Free any nodes below this node.  The tree is presumed to not need
  * shrinking, and any user data in the tree is presumed to not need a
@@ -670,6 +722,7 @@ static void radix_tree_free_nodes(struct radix_tree_node *node)
 	}
 }
 
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
 static inline int insert_entries(struct radix_tree_node *node, void **slot,
 				void *ptr, unsigned order, bool replace)
 {
@@ -741,6 +794,31 @@ static inline int insert_entries(struct radix_tree_node *node, void **slot,
 #endif
 
 /**
+ *	__radix_tree_create	-	create a slot in a radix tree
+ *	@root:		radix tree root
+ *	@index:		index key
+ *	@order:		index occupies 2^order aligned slots
+ *	@nodep:		returns node
+ *	@slotp:		returns slot
+ *
+ *	Create, if necessary, and return the node and slot for an item
+ *	at position @index in the radix tree @root.
+ *
+ *	Until there is more than one item in the tree, no nodes are
+ *	allocated and @root->rnode is used as a direct slot instead of
+ *	pointing to a node, in which case *@nodep will be NULL.
+ *
+ *	Returns -ENOMEM, or 0 for success.
+ */
+int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
+			unsigned order, struct radix_tree_node **nodep,
+			void ***slotp)
+{
+	return _radix_tree_create(root, root_gfp_mask(root), index, order,
+					nodep, slotp);
+}
+
+/**
  *	__radix_tree_insert    -    insert into a radix tree
  *	@root:		radix tree root
  *	@index:		index key
@@ -891,6 +969,7 @@ int radix_tree_split(struct radix_tree_root *root, unsigned long index,
 	void **slot;
 	unsigned int offset, end;
 	unsigned n, tag, tags = 0;
+	gfp_t gfp = root_gfp_mask(root);
 
 	if (!__radix_tree_lookup(root, index, &parent, &slot))
 		return -ENOENT;
@@ -926,7 +1005,7 @@ int radix_tree_split(struct radix_tree_root *root, unsigned long index,
 
 	for (;;) {
 		if (node->shift > order) {
-			child = radix_tree_node_alloc(root);
+			child = radix_tree_node_alloc(gfp);
 			if (!child)
 				goto nomem;
 			child->shift = node->shift - RADIX_TREE_MAP_SHIFT;
@@ -1571,6 +1650,8 @@ static inline bool radix_tree_shrink(struct radix_tree_root *root)
 		 * one (root->rnode) as far as dependent read barriers go.
 		 */
 		root->rnode = child;
+		if (is_idr(root) && !tag_get(node, IDR_FREE, 0))
+			root_tag_clear(root, IDR_FREE);
 
 		/*
 		 * We have a dilemma here. The node's slot[0] must not be
@@ -1630,7 +1711,12 @@ bool __radix_tree_delete_node(struct radix_tree_root *root,
 			parent->slots[node->offset] = NULL;
 			parent->count--;
 		} else {
-			root_tag_clear_all(root);
+			/*
+			 * Shouldn't the tags already have all been cleared
+			 * by the caller?
+			 */
+			if (!is_idr(root))
+				root_tag_clear_all(root);
 			root->rnode = NULL;
 		}
 
@@ -1643,6 +1729,17 @@ bool __radix_tree_delete_node(struct radix_tree_root *root,
 	return deleted;
 }
 
+static void radix_tree_iter_delete(struct radix_tree_root *root,
+					struct radix_tree_iter *iter)
+{
+	unsigned offset = (iter->index >> iter->shift) & RADIX_TREE_MAP_MASK;
+	struct radix_tree_node *node = iter->node;
+
+	node->slots[offset] = NULL;
+	node->count--;
+	__radix_tree_delete_node(root, node);
+}
+
 static inline void delete_sibling_entries(struct radix_tree_node *node,
 					void *ptr, unsigned offset)
 {
@@ -1665,7 +1762,7 @@ static inline void delete_sibling_entries(struct radix_tree_node *node,
  *
  *	Remove @item at @index from the radix tree rooted at @root.
  *
- *	Returns the address of the deleted item, or NULL if it was not present
+ *	Returns the value of the deleted item, or NULL if it was not present
  *	or the entry at the given @index was not @item.
  */
 void *radix_tree_delete_item(struct radix_tree_root *root,
@@ -1685,16 +1782,21 @@ void *radix_tree_delete_item(struct radix_tree_root *root,
 		return NULL;
 
 	if (!node) {
-		root_tag_clear_all(root);
+		if (is_idr(root))
+			root_tag_set(root, IDR_FREE);
+		else
+			root_tag_clear_all(root);
 		root->rnode = NULL;
 		return entry;
 	}
 
 	offset = get_slot_offset(node, slot);
 
-	/* Clear all tags associated with the item to be deleted.  */
-	for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
-		node_tag_clear(root, node, tag, offset);
+	if (is_idr(root))
+		node_tag_set(root, node, IDR_FREE, offset);
+	else
+		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+			node_tag_clear(root, node, tag, offset);
 
 	delete_sibling_entries(node, node_to_entry(slot), offset);
 	node->slots[offset] = NULL;
@@ -1713,7 +1815,7 @@ EXPORT_SYMBOL(radix_tree_delete_item);
  *
  *	Remove the item at @index from the radix tree rooted at @root.
  *
- *	Returns the address of the deleted item, or NULL if it was not present.
+ *	Returns the value of the deleted item, or NULL if it was not present.
  */
 void *radix_tree_delete(struct radix_tree_root *root, unsigned long index)
 {
@@ -1730,8 +1832,7 @@ void radix_tree_clear_tags(struct radix_tree_root *root,
 		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
 			node_tag_clear(root, node, tag, offset);
 	} else {
-		/* Clear root node tags */
-		root->gfp_mask &= __GFP_BITS_MASK;
+		root_tag_clear_all(root);
 	}
 }
 
@@ -1746,6 +1847,421 @@ int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag)
 }
 EXPORT_SYMBOL(radix_tree_tagged);
 
+/**
+ * idr_preload - preload for idr_alloc()
+ * @gfp_mask: allocation mask to use for preloading
+ *
+ * Preallocate memory to use for the next call to idr_alloc().  This function
+ * returns with preemption disabled.  It will be enabled by idr_preload_end().
+ */
+void idr_preload(gfp_t gfp_mask)
+{
+	__radix_tree_preload(gfp_mask, IDR_PRELOAD_SIZE);
+}
+EXPORT_SYMBOL(idr_preload);
+
+static int __idr_get_empty(struct radix_tree_root *root, gfp_t gfp,
+				unsigned long start, int end,
+				struct radix_tree_node **nodep, void ***slotp)
+{
+	struct radix_tree_node *node = NULL, *child;
+	void **slot = (void **)&root->rnode;
+	unsigned long maxindex;
+	unsigned long max = end > 0 ? end - 1 : INT_MAX;
+	unsigned int shift, offset = 0;
+
+ grow:
+	shift = radix_tree_load_root(root, &child, &maxindex);
+	if (!radix_tree_tagged(root, IDR_FREE))
+		start = max(start, maxindex + 1);
+	if (start > max)
+		return -ENOSPC;
+
+	if (start > maxindex) {
+		int error = radix_tree_extend(root, gfp, start, shift);
+		if (error < 0)
+			return error;
+		shift = error;
+		child = root->rnode;
+	}
+
+	while (shift) {
+		shift -= RADIX_TREE_MAP_SHIFT;
+		if (child == NULL) {
+			/* Have to add a child node.  */
+			child = radix_tree_node_alloc(gfp);
+			if (!child)
+				return -ENOMEM;
+			child->shift = shift;
+			child->offset = offset;
+			child->parent = node;
+			all_tag_set(child, IDR_FREE);
+			rcu_assign_pointer(*slot, node_to_entry(child));
+			if (node)
+				node->count++;
+		} else if (!radix_tree_is_internal_node(child))
+			break;
+
+		node = entry_to_node(child);
+		offset = radix_tree_descend(node, &child, start);
+		if (!tag_get(node, IDR_FREE, offset)) {
+			offset = radix_tree_find_next_bit(node, IDR_FREE,
+							offset + 1);
+			start = next_index(start, node, offset);
+			if (start > max)
+				return -ENOSPC;
+			while (offset == RADIX_TREE_MAP_SIZE) {
+				offset = node->offset + 1;
+				node = node->parent;
+				if (!node)
+					goto grow;
+				shift = node->shift;
+			}
+			child = node->slots[offset];
+		}
+		slot = &node->slots[offset];
+	}
+
+	*nodep = node;
+	*slotp = slot;
+	return start;
+}
+
+/**
+ * idr_alloc - allocate an id
+ * @idr: idr handle
+ * @ptr: pointer to be associated with the new id
+ * @start: the minimum id (inclusive)
+ * @end: the maximum id (exclusive)
+ * @gfp: memory allocation flags
+ *
+ * Allocates an unused ID in the range [start, end).  Returns -ENOSPC
+ * if there are no unused IDs in that range.
+ */
+int idr_alloc(struct idr *idr, void *ptr, int start, int end, gfp_t gfp)
+{
+	struct radix_tree_node *node;
+	void **slot;
+	int id;
+
+	if (WARN_ON_ONCE(start < 0))
+		return -EINVAL;
+
+	id = __idr_get_empty(&idr->idr_rt, gfp, start, end, &node, &slot);
+	if (id < 0)
+		return id;
+
+	BUG_ON(radix_tree_is_internal_node(ptr));
+
+	rcu_assign_pointer(*slot, ptr);
+	if (node)
+		node->count++;
+	node_tag_clear(&idr->idr_rt, node, IDR_FREE,
+						get_slot_offset(node, slot));
+	return id;
+}
+EXPORT_SYMBOL(idr_alloc);
+
+/**
+ * idr_alloc_cyclic - allocate new idr entry in a cyclical fashion
+ * @idr: idr handle
+ * @ptr: pointer to be associated with the new id
+ * @start: the minimum id (inclusive)
+ * @end: the maximum id (exclusive)
+ * @gfp: memory allocation flags
+ *
+ * Allocates an ID larger than the last ID allocated if one is available.
+ * If not, it will attempt to allocate the smallest ID that is larger or
+ * equal to @start.
+ */
+int idr_alloc_cyclic(struct idr *idr, void *ptr, int start, int end, gfp_t gfp)
+{
+	int id, curr = idr->idr_next;
+
+	if (curr < start)
+		curr = start;
+
+	id = idr_alloc(idr, ptr, curr, end, gfp);
+	if ((id == -ENOSPC) && (curr > start))
+		id = idr_alloc(idr, ptr, start, curr, gfp);
+
+	if (id >= 0)
+		idr->idr_next = id + 1U;
+
+	return id;
+}
+EXPORT_SYMBOL(idr_alloc_cyclic);
+
+/**
+ * idr_for_each - iterate through all stored pointers
+ * @idr: idr handle
+ * @fn: function to be called for each pointer
+ * @data: data passed to callback function
+ *
+ * The callback function will be called for each entry in @idr, passing
+ * the id, the pointer and the data pointer passed to this function.
+ *
+ * If @fn returns anything other than %0, the iteration stops and that
+ * value is returned from this function.
+ *
+ * idr_for_each() can be called concurrently with idr_get_new() and
+ * idr_remove() if protected by RCU.  Newly added entries may not be
+ * seen and deleted entries may be seen, but adding and removing entries
+ * will not cause other entries to be skipped, nor spurious ones to be seen.
+ */
+int idr_for_each(struct idr *idr,
+		int (*fn)(int id, void *p, void *data), void *data)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+
+	radix_tree_for_each_slot(slot, &idr->idr_rt, &iter, 0) {
+		int ret = fn(iter.index, *slot, data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(idr_for_each);
+
+/**
+ * idr_get_next - Find next populated entry
+ * @idr: idr handle
+ * @nextid: Pointer to lowest possible ID to return
+ *
+ * Returns the next populated entry in the tree with an ID greater than
+ * or equal to the value pointed to by @nextid.  On exit, @nextid is updated
+ * to the ID of the found value.  To use in a loop, the value pointed to by
+ * nextid must be incremented by the user.
+ */
+void *idr_get_next(struct idr *idr, int *nextid)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+
+	radix_tree_for_each_slot(slot, &idr->idr_rt, &iter, *nextid) {
+		*nextid = iter.index;
+		return *slot;
+	}
+
+	return NULL;
+}
+EXPORT_SYMBOL(idr_get_next);
+
+/**
+ * idr_replace - replace pointer for given id
+ * @idr: idr handle
+ * @ptr: New pointer to associate with the ID
+ * @id: Lookup key
+ *
+ * Replace the pointer registered with an id and return the old value.
+ * A %-ENOENT return indicates that @id was not found.
+ * A %-EINVAL return indicates that @id was not within valid constraints.
+ *
+ * This function can be called under the RCU read lock concurrently with
+ * idr_remove().
+ */
+void *idr_replace(struct idr *idr, void *ptr, int id)
+{
+	void **slot;
+	void *entry;
+
+	if (id < 0)
+		return ERR_PTR(-EINVAL);
+	if (!ptr || radix_tree_is_internal_node(ptr))
+		return ERR_PTR(-EINVAL);
+
+	entry = __radix_tree_lookup(&idr->idr_rt, id, NULL, &slot);
+
+	if (!entry)
+		return ERR_PTR(-ENOENT);
+
+	radix_tree_replace_slot(slot, ptr);
+
+	return entry;
+}
+EXPORT_SYMBOL(idr_replace);
+
+/**
+ * idr_destroy - release all internal memory from an IDR
+ * @idr: idr handle
+ *
+ * After this function is called, the IDR is empty, and may be reused or
+ * the data structure containing it may be freed.
+ *
+ * A typical clean-up sequence for objects stored in an idr tree will use
+ * idr_for_each() to free all objects, if necessary, then idr_destroy() to
+ * free the memory used to keep track of those objects.
+ */
+void idr_destroy(struct idr *idr)
+{
+	struct radix_tree_node **slot = &idr->idr_rt.rnode;
+	if (radix_tree_is_internal_node(*slot))
+		radix_tree_free_nodes(*slot);
+	*slot = NULL;
+	root_tag_set(&idr->idr_rt, IDR_FREE);
+}
+EXPORT_SYMBOL(idr_destroy);
+
+/**
+ * ida_pre_get - reserve resources for ida allocation
+ * @ida: ida handle
+ * @gfp: memory allocation flags
+ *
+ * This function should be called before calling ida_get_new_above().  If it
+ * is unable to allocate memory, it will return %0.  On success, it returns %1.
+ */
+int ida_pre_get(struct ida *ida, gfp_t gfp)
+{
+	struct ida_bitmap *bitmap;
+
+	idr_preload(gfp);
+	idr_preload_end();
+
+	if (!ida->free_bitmap) {
+		bitmap = kmalloc(sizeof(struct ida_bitmap), gfp);
+		if (!bitmap)
+			return 0;
+		bitmap = xchg(&ida->free_bitmap, bitmap);
+		kfree(bitmap);
+	}
+
+	return 1;
+}
+EXPORT_SYMBOL(ida_pre_get);
+
+/**
+ * ida_get_new_above - allocate new ID above or equal to a start id
+ * @ida: ida handle
+ * @starting_id: id to start search at
+ * @p_id: pointer to the allocated handle
+ *
+ * Allocate new ID above or equal to @starting_id.  It should be called
+ * with any required locks.
+ *
+ * If memory is required, it will return %-EAGAIN, you should unlock
+ * and go back to the ida_pre_get() call.  If the ida is full, it will
+ * return %-ENOSPC.
+ *
+ * @p_id returns a value in the range @starting_id ... %0x7fffffff.
+ */
+int ida_get_new_above(struct ida *ida, int start, int *id)
+{
+	struct radix_tree_root *root = &ida->ida_rt;
+	void **slot = (void **)&root->rnode;
+	struct radix_tree_node *node;
+	struct ida_bitmap *bitmap = NULL;
+	unsigned long index;
+	unsigned bit, offset = 0;
+
+	index = start / IDA_BITMAP_BITS;
+	bit = start % IDA_BITMAP_BITS;
+
+ restart:
+	index = __idr_get_empty(root, GFP_ATOMIC, index, INT_MAX, &node, &slot);
+	if (index > INT_MAX)
+		return index;
+
+	index *= IDA_BITMAP_BITS;
+	if (index > INT_MAX)
+		return -ENOSPC;
+
+	if (index > start)
+		bit = 0;
+	offset = get_slot_offset(node, slot);
+
+	bitmap = *slot;
+	if (bitmap) {
+		bit = find_next_zero_bit(bitmap->bitmap, IDA_BITMAP_BITS, bit);
+		index += bit;
+		if (index > INT_MAX)
+			return -ENOSPC;
+		if (bit == IDA_BITMAP_BITS) {
+			index /= IDA_BITMAP_BITS;
+			goto restart;
+		}
+		__set_bit(bit, bitmap->bitmap);
+		if (bitmap_full(bitmap->bitmap, IDA_BITMAP_BITS))
+			node_tag_clear(root, node, IDR_FREE, offset);
+		bitmap = xchg(&ida->free_bitmap, NULL);
+		kfree(bitmap);
+	} else {
+		index += bit;
+		bitmap = xchg(&ida->free_bitmap, NULL);
+		if (!bitmap)
+			return -EAGAIN;
+		memset(bitmap, 0, sizeof(*bitmap));
+		__set_bit(bit, bitmap->bitmap);
+		rcu_assign_pointer(*slot, bitmap);
+		if (node)
+			node->count++;
+	}
+
+	*id = index;
+	return 0;
+}
+EXPORT_SYMBOL(ida_get_new_above);
+
+/**
+ * ida_remove - Free the given ID
+ * @ida: ida handle
+ * @id: ID to free
+ *
+ * This function should not be called at the same time as ida_get_new_above().
+ */
+void ida_remove(struct ida *ida, int id)
+{
+	unsigned long index = id / IDA_BITMAP_BITS;
+	unsigned offset = id % IDA_BITMAP_BITS;
+	struct ida_bitmap *bitmap;
+	struct radix_tree_node *node;
+	void **slot;
+
+	bitmap = __radix_tree_lookup(&ida->ida_rt, index, &node, &slot);
+	if (!bitmap || !test_bit(offset, bitmap->bitmap))
+		goto err;
+
+	__clear_bit(offset, bitmap->bitmap);
+	node_tag_set(&ida->ida_rt, node, IDR_FREE, get_slot_offset(node, slot));
+	if (bitmap_empty(bitmap->bitmap, IDA_BITMAP_BITS)) {
+		*slot = NULL;
+		kfree(bitmap);
+		if (node) {
+			node->count--;
+			__radix_tree_delete_node(&ida->ida_rt, node);
+		}
+	}
+	return;
+ err:
+	WARN(1, "ida_remove called for id=%d which is not allocated.\n", id);
+}
+EXPORT_SYMBOL(ida_remove);
+
+/**
+ * ida_destroy - Free the contents of an ida
+ * @ida: ida handle
+ *
+ * Calling this function releases all resources associated with an IDA.  When
+ * this call returns, the IDA is empty and can be reused or freed.  The caller
+ * should not allow ida_remove() or ida_get_new_above() to be called at the
+ * same time.
+ */
+void ida_destroy(struct ida *ida)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+
+	radix_tree_for_each_slot(slot, &ida->ida_rt, &iter, 0) {
+		struct ida_bitmap *bitmap = *slot;
+		kfree(bitmap);
+		radix_tree_iter_delete(&ida->ida_rt, &iter);
+	}
+
+	kfree(ida->free_bitmap);
+}
+EXPORT_SYMBOL(ida_destroy);
+
 static void
 radix_tree_node_ctor(void *arg)
 {
diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
index 3635e4d..5a616a3 100644
--- a/tools/testing/radix-tree/Makefile
+++ b/tools/testing/radix-tree/Makefile
@@ -3,7 +3,7 @@ CFLAGS += -I. -I../../include -g -O2 -Wall -D_LGPL_SOURCE
 LDFLAGS += -lpthread -lurcu
 TARGETS = main
 OFILES = main.o radix-tree.o linux.o test.o tag_check.o find_next_bit.o \
-	 regression1.o regression2.o regression3.o multiorder.o \
+	 regression1.o regression2.o regression3.o multiorder.o idr.o \
 	 iteration_check.o benchmark.o
 
 ifdef BENCHMARK
@@ -23,7 +23,8 @@ find_next_bit.o: ../../lib/find_bit.c
 
 $(OFILES): *.h */*.h \
 	../../include/linux/*.h \
-	../../../include/linux/radix-tree.h
+	../../../include/linux/radix-tree.h \
+	../../../include/linux/idr.h
 
 radix-tree.c: ../../../lib/radix-tree.c
 	sed -e 's/^static //' -e 's/__always_inline //' -e 's/inline //' < $< > $@
diff --git a/tools/testing/radix-tree/idr.c b/tools/testing/radix-tree/idr.c
new file mode 100644
index 0000000..0f9e7b1
--- /dev/null
+++ b/tools/testing/radix-tree/idr.c
@@ -0,0 +1,148 @@
+/*
+ * idr.c: Test the IDR API
+ * Copyright (c) 2016 Matthew Wilcox <willy@infradead.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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.
+ */
+#include <linux/idr.h>
+#include <linux/slab.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+
+#include "test.h"
+
+#define DUMMY_PTR	((void *)0x12)
+
+int item_idr_free(int id, void *p, void *data)
+{
+	struct item *item = p;
+	assert(item->index == id);
+	idr_remove(data, id);
+	free(p);
+
+	return 0;
+}
+
+void item_idr_remove(struct idr *idr, int id)
+{
+	struct item *item = idr_find(idr, id);
+	assert(item->index == id);
+	idr_remove(idr, id);
+	free(item);
+}
+
+void idr_alloc_test(void)
+{
+	unsigned long i;
+	DEFINE_IDR(idr);
+
+	assert(idr_alloc_cyclic(&idr, DUMMY_PTR, 0, 0x4000, GFP_KERNEL) == 0);
+	assert(idr_alloc_cyclic(&idr, DUMMY_PTR, 0x3ffd, 0x4000, GFP_KERNEL) == 0x3ffd);
+	idr_remove(&idr, 0x3ffd);
+	idr_remove(&idr, 0);
+
+	for (i = 0x3ffe; i < 0x4003; i++) {
+		int id;
+		struct item *item;
+
+		if (i < 0x4000)
+			item = item_create(i);
+		else
+			item = item_create(i - 0x3fff);
+
+		id = idr_alloc_cyclic(&idr, item, 1, 0x4000, GFP_KERNEL);
+		assert(id == item->index);
+	}
+
+	idr_for_each(&idr, item_idr_free, &idr);
+}
+
+void idr_checks(void)
+{
+	unsigned long i;
+	DEFINE_IDR(idr);
+
+	for (i = 0; i < 10000; i++) {
+		struct item *item = item_create(i);
+		assert(idr_alloc(&idr, item, 0, 20000, GFP_KERNEL) == i);
+	}
+
+	assert(idr_alloc(&idr, DUMMY_PTR, 5, 30, GFP_KERNEL) < 0);
+
+	for (i = 0; i < 5000; i++)
+		item_idr_remove(&idr, i);
+
+	idr_for_each(&idr, item_idr_free, &idr);
+
+	assert(idr_is_empty(&idr));
+
+	for (i = INT_MAX - 3UL; i < INT_MAX + 1UL; i++) {
+		struct item *item = item_create(i);
+		assert(idr_alloc(&idr, item, i, i + 10, GFP_KERNEL) == i);
+	}
+	assert(idr_alloc(&idr, DUMMY_PTR, i - 2, i, GFP_KERNEL) == -ENOSPC);
+
+	idr_destroy(&idr);
+	idr_destroy(&idr);
+
+	assert(idr_is_empty(&idr));
+
+	for (i = 1; i < 10000; i++) {
+		struct item *item = item_create(i);
+		assert(idr_alloc(&idr, item, 1, 20000, GFP_KERNEL) == i);
+	}
+
+	idr_destroy(&idr);
+
+	idr_alloc_test();
+}
+
+void ida_checks(void)
+{
+	DEFINE_IDA(ida);
+
+	unsigned long i;
+	int id;
+
+	for (i = 0; i < 10000; i++) {
+		ida_pre_get(&ida, GFP_KERNEL);
+		ida_get_new(&ida, &id);
+		assert(id == i);
+	}
+
+	ida_remove(&ida, 20);
+	ida_remove(&ida, 21);
+	for (i = 0; i < 3; i++) {
+		ida_pre_get(&ida, GFP_KERNEL);
+		ida_get_new(&ida, &id);
+		if (i == 2)
+			assert(id == 10000);
+	}
+
+	for (i = 0; i < 5000; i++)
+		ida_remove(&ida, i);
+
+	ida_pre_get(&ida, GFP_KERNEL);
+	ida_get_new_above(&ida, 5000, &id);
+	assert(id == 10001);
+
+	ida_destroy(&ida);
+
+	assert(ida_is_empty(&ida));
+
+	ida_pre_get(&ida, GFP_KERNEL);
+	ida_get_new_above(&ida, 1, &id);
+	assert(id == 1);
+
+	ida_remove(&ida, id);
+	ida_destroy(&ida);
+
+	radix_tree_callback(NULL, CPU_DEAD, NULL);
+}
diff --git a/tools/testing/radix-tree/linux/idr.h b/tools/testing/radix-tree/linux/idr.h
new file mode 100644
index 0000000..4e342f2
--- /dev/null
+++ b/tools/testing/radix-tree/linux/idr.h
@@ -0,0 +1 @@
+#include "../../../../include/linux/idr.h"
diff --git a/tools/testing/radix-tree/linux/kernel.h b/tools/testing/radix-tree/linux/kernel.h
index 9b43b49..7d214e9 100644
--- a/tools/testing/radix-tree/linux/kernel.h
+++ b/tools/testing/radix-tree/linux/kernel.h
@@ -30,6 +30,7 @@
 #define __force
 #define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
 #define pr_debug printk
+#define pr_cont printk
 
 #define smp_rmb()	barrier()
 #define smp_wmb()	barrier()
@@ -41,6 +42,7 @@
 	const typeof( ((type *)0)->member ) *__mptr = (ptr);    \
 	(type *)( (char *)__mptr - offsetof(type, member) );})
 #define min(a, b) ((a) < (b) ? (a) : (b))
+#define max(a, b) ((a) < (b) ? (b) : (a))
 
 #define cond_resched()	sched_yield()
 
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 7d14812..e267d5f 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -3,6 +3,7 @@
 #include <unistd.h>
 #include <time.h>
 #include <assert.h>
+#include <limits.h>
 
 #include <linux/slab.h>
 #include <linux/radix-tree.h>
@@ -314,6 +315,11 @@ static void single_thread_tests(bool long_run)
 	rcu_barrier();
 	printf("after dynamic_height_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
+	idr_checks();
+	ida_checks();
+	rcu_barrier();
+	printf("after idr_checks: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	big_gang_check(long_run);
 	rcu_barrier();
 	printf("after big_gang_check: %d allocated, preempt %d\n",
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index 1faf0a3..4baec38 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -36,6 +36,8 @@ void tag_check(void);
 void multiorder_checks(void);
 void iteration_test(void);
 void benchmark(void);
+void idr_checks(void);
+void ida_checks(void);
 
 struct item *
 item_tag_set(struct radix_tree_root *root, unsigned long index, int tag);
-- 
2.10.2

[PATCH 00/29] Improve radix tree for 4.10

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

Hi Andrew,

Please include these patches in the -mm tree for 4.10.  Mostly these are
improvements; the only bug fixes in here relate to multiorder entries
(which as far as I'm aware remain unused).  The IDR rewrite has the
highest potential for causing mayhem as the test suite is quite spartan.
We have an Outreachy intern scheduled to work on the test suite for the
2016 winter season, so hopefully it will improve soon.

I posted a series of patches to improve the radix tree code two months
ago, and Konstantin had the quite reasonable objection that they slowed
down the iterators.  He helpfully sent out the benchmark he used.
With this revision of the patchset, I have slightly improved performance
on his benchmark.

Noteworthy changes from the last revision of the patchset:

 - Fix the existing radix tree iterator design instead of rewriting it
 - Avoid adding the new radix_tree_for_each_tagged_max iterator; just
   check in the one caller whether we've exceeded the limit.
 - Change the name of radix_tree_iter_save() back to
   radix_tree_iter_next().  I went back and forth on this one, and could
   easily be convinced to rename it as I think radix_tree_iter_save()
   makes more sense to the callers.
 - Add radix_tree_iter_delete() instead of poking around the internals
   of the radix tree in ida_destroy()
 - Add accessors for the idr cyclic allocation cursor

Patches 1-9: Test suite improvements, including Konstantin's benchmark
Patches 10-12: join/split (Kirill wants these for huge page cache)
Patches 13-17: General improvements to the radix tree
Patches 18-22: Radix tree iterator improvements
Patches 23-29: Replace the IDR with the radix tree

Konstantin Khlebnikov (1):
  radix tree test suite: benchmark for iterator

Matthew Wilcox (28):
  tools: Add WARN_ON_ONCE
  radix tree test suite: Allow GFP_ATOMIC allocations to fail
  radix tree test suite: Track preempt_count
  radix tree test suite: Free preallocated nodes
  radix tree test suite: Make runs more reproducible
  radix tree test suite: Use rcu_barrier
  tools: Add more bitmap functions
  radix tree test suite: Use common find-bit code
  radix-tree: Add radix_tree_join
  radix-tree: Add radix_tree_split
  radix-tree: Add radix_tree_split_preload()
  radix-tree: Fix typo
  radix-tree: Move rcu_head into a union with private_list
  radix-tree: Create node_tag_set()
  radix-tree: Make radix_tree_find_next_bit more useful
  radix-tree: Improve dump output
  btrfs: Fix race in btrfs_free_dummy_fs_info()
  radix tree test suite: iteration test misuses RCU
  radix tree: Improve multiorder iterators
  radix-tree: Delete radix_tree_locate_item()
  radix-tree: Delete radix_tree_range_tag_if_tagged()
  idr: Add ida_is_empty
  tpm: Use idr_find(), not idr_find_slowpath()
  rxrpc: Abstract away knowledge of IDR internals
  idr: Reduce the number of bits per level from 8 to 6
  radix tree test suite: Add some more functionality
  radix-tree: Create all_tag_set
  Reimplement IDR and IDA using the radix tree

 drivers/char/tpm/tpm-chip.c                        |    4 +-
 drivers/usb/gadget/function/f_hid.c                |    6 +-
 drivers/usb/gadget/function/f_printer.c            |    6 +-
 fs/btrfs/tests/btrfs-tests.c                       |    1 +
 include/linux/idr.h                                |  156 +--
 include/linux/radix-tree.h                         |  161 ++-
 init/main.c                                        |    3 +-
 lib/idr.c                                          | 1075 ----------------
 lib/radix-tree.c                                   | 1328 +++++++++++++++-----
 mm/khugepaged.c                                    |    2 +-
 mm/page-writeback.c                                |   28 +-
 mm/shmem.c                                         |   32 +-
 net/rxrpc/af_rxrpc.c                               |   11 +-
 net/rxrpc/conn_client.c                            |    4 +-
 tools/include/asm/bug.h                            |   11 +
 tools/include/linux/bitmap.h                       |   26 +
 tools/lib/find_bit.c                               |    8 +
 tools/testing/radix-tree/Makefile                  |   18 +-
 tools/testing/radix-tree/benchmark.c               |   98 ++
 tools/testing/radix-tree/find_next_bit.c           |   57 -
 tools/testing/radix-tree/idr.c                     |  148 +++
 tools/testing/radix-tree/iteration_check.c         |   39 +-
 tools/testing/radix-tree/linux.c                   |   23 +-
 tools/testing/radix-tree/linux/bitops.h            |   40 +-
 tools/testing/radix-tree/linux/bitops/non-atomic.h |   13 +-
 tools/testing/radix-tree/linux/bug.h               |    2 +-
 tools/testing/radix-tree/linux/gfp.h               |   22 +-
 tools/testing/radix-tree/linux/idr.h               |    1 +
 tools/testing/radix-tree/linux/kernel.h            |   20 +
 tools/testing/radix-tree/linux/preempt.h           |    6 +-
 tools/testing/radix-tree/linux/slab.h              |    6 +-
 tools/testing/radix-tree/linux/types.h             |    2 -
 tools/testing/radix-tree/main.c                    |   80 +-
 tools/testing/radix-tree/multiorder.c              |  111 +-
 tools/testing/radix-tree/regression2.c             |    3 +-
 tools/testing/radix-tree/regression3.c             |    6 +-
 tools/testing/radix-tree/tag_check.c               |    9 +-
 tools/testing/radix-tree/test.c                    |   47 +
 tools/testing/radix-tree/test.h                    |   18 +
 39 files changed, 1884 insertions(+), 1747 deletions(-)
 create mode 100644 tools/testing/radix-tree/benchmark.c
 delete mode 100644 tools/testing/radix-tree/find_next_bit.c
 create mode 100644 tools/testing/radix-tree/idr.c
 create mode 100644 tools/testing/radix-tree/linux/idr.h

-- 
2.10.2

[PATCH 01/29] tools: Add WARN_ON_ONCE

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

The radix tree uses its own buggy WARN_ON_ONCE.  Replace it with the
definition from asm-generic/bug.h

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 tools/include/asm/bug.h                | 11 +++++++++++
 tools/testing/radix-tree/Makefile      |  2 +-
 tools/testing/radix-tree/linux/bug.h   |  2 +-
 tools/testing/radix-tree/linux/types.h |  2 --
 4 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/tools/include/asm/bug.h b/tools/include/asm/bug.h
index 9e5f484..beda1a8 100644
--- a/tools/include/asm/bug.h
+++ b/tools/include/asm/bug.h
@@ -12,6 +12,17 @@
 	unlikely(__ret_warn_on);		\
 })
 
+#define WARN_ON_ONCE(condition) ({			\
+	static int __warned;				\
+	int __ret_warn_once = !!(condition);		\
+							\
+	if (unlikely(__ret_warn_once && !__warned)) {	\
+		__warned = true;			\
+		WARN_ON(1);				\
+	}						\
+	unlikely(__ret_warn_once);			\
+})
+
 #define WARN_ONCE(condition, format...)	({	\
 	static int __warned;			\
 	int __ret_warn_once = !!(condition);	\
diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
index f2e07f2..3c338dc 100644
--- a/tools/testing/radix-tree/Makefile
+++ b/tools/testing/radix-tree/Makefile
@@ -1,5 +1,5 @@
 
-CFLAGS += -I. -g -O2 -Wall -D_LGPL_SOURCE
+CFLAGS += -I. -I../../include -g -O2 -Wall -D_LGPL_SOURCE
 LDFLAGS += -lpthread -lurcu
 TARGETS = main
 OFILES = main.o radix-tree.o linux.o test.o tag_check.o find_next_bit.o \
diff --git a/tools/testing/radix-tree/linux/bug.h b/tools/testing/radix-tree/linux/bug.h
index ccbe444..23b8ed5 100644
--- a/tools/testing/radix-tree/linux/bug.h
+++ b/tools/testing/radix-tree/linux/bug.h
@@ -1 +1 @@
-#define WARN_ON_ONCE(x)		assert(x)
+#include "asm/bug.h"
diff --git a/tools/testing/radix-tree/linux/types.h b/tools/testing/radix-tree/linux/types.h
index faa0b6f..8491d89 100644
--- a/tools/testing/radix-tree/linux/types.h
+++ b/tools/testing/radix-tree/linux/types.h
@@ -6,8 +6,6 @@
 #define __rcu
 #define __read_mostly
 
-#define BITS_PER_LONG (sizeof(long) * 8)
-
 static inline void INIT_LIST_HEAD(struct list_head *list)
 {
 	list->next = list;
-- 
2.10.2

[PATCH 02/29] radix tree test suite: Allow GFP_ATOMIC allocations to fail

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

In order to test the preload code, it is necessary to fail GFP_ATOMIC
allocations, which requires defining GFP_KERNEL and GFP_ATOMIC properly.
Remove the obsolete __GFP_WAIT and copy the definitions of the __GFP
flags which are used from the kernel include files.  We also need the
real definition of gfpflags_allow_blocking() to persuade the radix tree
to actually use its preallocated nodes.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 tools/testing/radix-tree/linux.c      |  7 ++++++-
 tools/testing/radix-tree/linux/gfp.h  | 22 +++++++++++++++++++---
 tools/testing/radix-tree/linux/slab.h |  5 -----
 3 files changed, 25 insertions(+), 9 deletions(-)

diff --git a/tools/testing/radix-tree/linux.c b/tools/testing/radix-tree/linux.c
index 1548237..3cfb04e 100644
--- a/tools/testing/radix-tree/linux.c
+++ b/tools/testing/radix-tree/linux.c
@@ -33,7 +33,12 @@ mempool_t *mempool_create(int min_nr, mempool_alloc_t *alloc_fn,
 
 void *kmem_cache_alloc(struct kmem_cache *cachep, int flags)
 {
-	void *ret = malloc(cachep->size);
+	void *ret;
+
+	if (flags & __GFP_NOWARN)
+		return NULL;
+
+	ret = malloc(cachep->size);
 	if (cachep->ctor)
 		cachep->ctor(ret);
 	uatomic_inc(&nr_allocated);
diff --git a/tools/testing/radix-tree/linux/gfp.h b/tools/testing/radix-tree/linux/gfp.h
index 5201b91..5b09b2c 100644
--- a/tools/testing/radix-tree/linux/gfp.h
+++ b/tools/testing/radix-tree/linux/gfp.h
@@ -3,8 +3,24 @@
 
 #define __GFP_BITS_SHIFT 26
 #define __GFP_BITS_MASK ((gfp_t)((1 << __GFP_BITS_SHIFT) - 1))
-#define __GFP_WAIT 1
-#define __GFP_ACCOUNT 0
-#define __GFP_NOWARN 0
+
+#define __GFP_HIGH		0x20u
+#define __GFP_IO		0x40u
+#define __GFP_FS		0x80u
+#define __GFP_NOWARN		0x200u
+#define __GFP_ATOMIC		0x80000u
+#define __GFP_ACCOUNT		0x100000u
+#define __GFP_DIRECT_RECLAIM	0x400000u
+#define __GFP_KSWAPD_RECLAIM	0x2000000u
+
+#define __GFP_RECLAIM		(__GFP_DIRECT_RECLAIM|__GFP_KSWAPD_RECLAIM)
+
+#define GFP_ATOMIC		(__GFP_HIGH|__GFP_ATOMIC|__GFP_KSWAPD_RECLAIM)
+#define GFP_KERNEL		(__GFP_RECLAIM | __GFP_IO | __GFP_FS)
+
+static inline bool gfpflags_allow_blocking(const gfp_t gfp_flags)
+{
+	return !!(gfp_flags & __GFP_DIRECT_RECLAIM);
+}
 
 #endif
diff --git a/tools/testing/radix-tree/linux/slab.h b/tools/testing/radix-tree/linux/slab.h
index 6d5a347..452e2bf 100644
--- a/tools/testing/radix-tree/linux/slab.h
+++ b/tools/testing/radix-tree/linux/slab.h
@@ -7,11 +7,6 @@
 #define SLAB_PANIC 2
 #define SLAB_RECLAIM_ACCOUNT    0x00020000UL            /* Objects are reclaimable */
 
-static inline int gfpflags_allow_blocking(gfp_t mask)
-{
-	return 1;
-}
-
 struct kmem_cache {
 	int size;
 	void (*ctor)(void *);
-- 
2.10.2

[PATCH 03/29] radix tree test suite: Track preempt_count

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

Rather than simply NOP out preempt_enable() and preempt_disable(),
keep track of preempt_count and display it regularly in case either
the test suite or the code under test is forgetting to balance the
enables & disables.  Only found a test-case that was forgetting to
re-enable preemption, but it's a possibility worth checking.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 tools/testing/radix-tree/linux.c         |  1 +
 tools/testing/radix-tree/linux/preempt.h |  6 +++---
 tools/testing/radix-tree/main.c          | 30 ++++++++++++++++++++----------
 3 files changed, 24 insertions(+), 13 deletions(-)

diff --git a/tools/testing/radix-tree/linux.c b/tools/testing/radix-tree/linux.c
index 3cfb04e..1f32a16 100644
--- a/tools/testing/radix-tree/linux.c
+++ b/tools/testing/radix-tree/linux.c
@@ -9,6 +9,7 @@
 #include <urcu/uatomic.h>
 
 int nr_allocated;
+int preempt_count;
 
 void *mempool_alloc(mempool_t *pool, int gfp_mask)
 {
diff --git a/tools/testing/radix-tree/linux/preempt.h b/tools/testing/radix-tree/linux/preempt.h
index 6210672..65c04c2 100644
--- a/tools/testing/radix-tree/linux/preempt.h
+++ b/tools/testing/radix-tree/linux/preempt.h
@@ -1,4 +1,4 @@
-/* */
+extern int preempt_count;
 
-#define preempt_disable() do { } while (0)
-#define preempt_enable() do { } while (0)
+#define preempt_disable()	uatomic_inc(&preempt_count)
+#define preempt_enable()	uatomic_dec(&preempt_count)
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index daa9010..64ffe67 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -293,27 +293,36 @@ static void single_thread_tests(bool long_run)
 {
 	int i;
 
-	printf("starting single_thread_tests: %d allocated\n", nr_allocated);
+	printf("starting single_thread_tests: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	multiorder_checks();
-	printf("after multiorder_check: %d allocated\n", nr_allocated);
+	printf("after multiorder_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	locate_check();
-	printf("after locate_check: %d allocated\n", nr_allocated);
+	printf("after locate_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	tag_check();
-	printf("after tag_check: %d allocated\n", nr_allocated);
+	printf("after tag_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	gang_check();
-	printf("after gang_check: %d allocated\n", nr_allocated);
+	printf("after gang_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	add_and_check();
-	printf("after add_and_check: %d allocated\n", nr_allocated);
+	printf("after add_and_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	dynamic_height_check();
-	printf("after dynamic_height_check: %d allocated\n", nr_allocated);
+	printf("after dynamic_height_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	big_gang_check(long_run);
-	printf("after big_gang_check: %d allocated\n", nr_allocated);
+	printf("after big_gang_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	for (i = 0; i < (long_run ? 2000 : 3); i++) {
 		copy_tag_check();
 		printf("%d ", i);
 		fflush(stdout);
 	}
-	printf("after copy_tag_check: %d allocated\n", nr_allocated);
+	printf("after copy_tag_check: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 }
 
 int main(int argc, char **argv)
@@ -336,7 +345,8 @@ int main(int argc, char **argv)
 	single_thread_tests(long_run);
 
 	sleep(1);
-	printf("after sleep(1): %d allocated\n", nr_allocated);
+	printf("after sleep(1): %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	rcu_unregister_thread();
 
 	exit(0);
-- 
2.10.2

[PATCH 04/29] radix tree test suite: Free preallocated nodes

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

It can be a source of mild concern when the test suite shows that we're
leaking nodes.  While poring over the source code looking for leaks
can lead to some fascinating bugs being discovered, sometimes the leak
is simply that these nodes were preallocated and are sitting on the
per-CPU list.  Free them by faking a CPU_DEAD event.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 tools/testing/radix-tree/main.c | 3 +++
 tools/testing/radix-tree/test.h | 4 ++++
 2 files changed, 7 insertions(+)

diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 64ffe67..2930560 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -344,6 +344,9 @@ int main(int argc, char **argv)
 	iteration_test();
 	single_thread_tests(long_run);
 
+	/* Free any remaining preallocated nodes */
+	radix_tree_callback(NULL, CPU_DEAD, NULL);
+
 	sleep(1);
 	printf("after sleep(1): %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index 217fb24..8cd666a 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -2,6 +2,8 @@
 #include <linux/types.h>
 #include <linux/radix-tree.h>
 #include <linux/rcupdate.h>
+#include <linux/notifier.h>
+#include <linux/cpu.h>
 
 struct item {
 	unsigned long index;
@@ -44,3 +46,5 @@ void radix_tree_dump(struct radix_tree_root *root);
 int root_tag_get(struct radix_tree_root *root, unsigned int tag);
 unsigned long node_maxindex(struct radix_tree_node *);
 unsigned long shift_maxindex(unsigned int shift);
+int radix_tree_callback(struct notifier_block *nfb,
+			unsigned long action, void *hcpu);
-- 
2.10.2

[PATCH 05/29] radix tree test suite: Make runs more reproducible

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

Instead of reseeding the random number generator every time around the
loop in big_gang_check(), seed it at the beginning of execution.  Use
rand_r() and an independent base seed for each thread in iteration_test()
so they don't stomp all over each others state.  Since this particular
test depends on the kernel scheduler, the iteration test can't be
reproduced based purely on the random seed, but at least it won't pollute
the other tests.

Print the seed, and allow the seed to be specified so that a run which
hits a problem can be reproduced.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 tools/testing/radix-tree/iteration_check.c | 11 +++++++----
 tools/testing/radix-tree/main.c            |  9 +++++++--
 2 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/tools/testing/radix-tree/iteration_check.c b/tools/testing/radix-tree/iteration_check.c
index 9adb8e7..11d570c 100644
--- a/tools/testing/radix-tree/iteration_check.c
+++ b/tools/testing/radix-tree/iteration_check.c
@@ -20,6 +20,7 @@
 #define TAG 0
 static pthread_mutex_t tree_lock = PTHREAD_MUTEX_INITIALIZER;
 static pthread_t threads[NUM_THREADS];
+static unsigned int seeds[3];
 RADIX_TREE(tree, GFP_KERNEL);
 bool test_complete;
 
@@ -71,7 +72,7 @@ static void *tagged_iteration_fn(void *arg)
 				continue;
 			}
 
-			if (rand() % 50 == 0)
+			if (rand_r(&seeds[0]) % 50 == 0)
 				slot = radix_tree_iter_next(&iter);
 		}
 		rcu_read_unlock();
@@ -111,7 +112,7 @@ static void *untagged_iteration_fn(void *arg)
 				continue;
 			}
 
-			if (rand() % 50 == 0)
+			if (rand_r(&seeds[1]) % 50 == 0)
 				slot = radix_tree_iter_next(&iter);
 		}
 		rcu_read_unlock();
@@ -129,7 +130,7 @@ static void *remove_entries_fn(void *arg)
 	while (!test_complete) {
 		int pgoff;
 
-		pgoff = rand() % 100;
+		pgoff = rand_r(&seeds[2]) % 100;
 
 		pthread_mutex_lock(&tree_lock);
 		item_delete(&tree, pgoff);
@@ -146,9 +147,11 @@ void iteration_test(void)
 
 	printf("Running iteration tests for 10 seconds\n");
 
-	srand(time(0));
 	test_complete = false;
 
+	for (i = 0; i < 3; i++)
+		seeds[i] = rand();
+
 	if (pthread_create(&threads[0], NULL, tagged_iteration_fn, NULL)) {
 		perror("pthread_create");
 		exit(1);
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 2930560..5ddec8e 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -67,7 +67,6 @@ void big_gang_check(bool long_run)
 
 	for (i = 0; i < (long_run ? 1000 : 3); i++) {
 		__big_gang_check();
-		srand(time(0));
 		printf("%d ", i);
 		fflush(stdout);
 	}
@@ -329,12 +328,18 @@ int main(int argc, char **argv)
 {
 	bool long_run = false;
 	int opt;
+	unsigned int seed = time(NULL);
 
-	while ((opt = getopt(argc, argv, "l")) != -1) {
+	while ((opt = getopt(argc, argv, "ls:")) != -1) {
 		if (opt == 'l')
 			long_run = true;
+		else if (opt == 's')
+			seed = strtoul(optarg, NULL, 0);
 	}
 
+	printf("random seed %u\n", seed);
+	srand(seed);
+
 	rcu_register_thread();
 	radix_tree_init();
 
-- 
2.10.2

[PATCH 06/29] radix tree test suite: benchmark for iterator

[Matthew Wilcox]

From: Konstantin Khlebnikov <koct9i@gmail.com>

This adds simple benchmark for iterator similar to one I've used for
commit 78c1d78 ("radix-tree: introduce bit-optimized iterator")

Building with make BENCHMARK=1 set radix tree order to 6, this allows
to get performance comparable to in kernel performance.

Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com>
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 tools/testing/radix-tree/Makefile       |  6 +-
 tools/testing/radix-tree/benchmark.c    | 98 +++++++++++++++++++++++++++++++++
 tools/testing/radix-tree/linux/kernel.h |  4 ++
 tools/testing/radix-tree/main.c         |  2 +
 tools/testing/radix-tree/test.h         |  1 +
 5 files changed, 110 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/radix-tree/benchmark.c

diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
index 3c338dc..08283a8 100644
--- a/tools/testing/radix-tree/Makefile
+++ b/tools/testing/radix-tree/Makefile
@@ -4,7 +4,11 @@ LDFLAGS += -lpthread -lurcu
 TARGETS = main
 OFILES = main.o radix-tree.o linux.o test.o tag_check.o find_next_bit.o \
 	 regression1.o regression2.o regression3.o multiorder.o \
-	 iteration_check.o
+	 iteration_check.o benchmark.o
+
+ifdef BENCHMARK
+	CFLAGS += -DBENCHMARK=1
+endif
 
 targets: $(TARGETS)
 
diff --git a/tools/testing/radix-tree/benchmark.c b/tools/testing/radix-tree/benchmark.c
new file mode 100644
index 0000000..215ca86
--- /dev/null
+++ b/tools/testing/radix-tree/benchmark.c
@@ -0,0 +1,98 @@
+/*
+ * benchmark.c:
+ * Author: Konstantin Khlebnikov <koct9i@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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.
+ */
+#include <linux/radix-tree.h>
+#include <linux/slab.h>
+#include <linux/errno.h>
+#include <time.h>
+#include "test.h"
+
+#define NSEC_PER_SEC	1000000000L
+
+static long long benchmark_iter(struct radix_tree_root *root, bool tagged)
+{
+	volatile unsigned long sink = 0;
+	struct radix_tree_iter iter;
+	struct timespec start, finish;
+	long long nsec;
+	int l, loops = 1;
+	void **slot;
+
+#ifdef BENCHMARK
+again:
+#endif
+	clock_gettime(CLOCK_MONOTONIC, &start);
+	for (l = 0; l < loops; l++) {
+		if (tagged) {
+			radix_tree_for_each_tagged(slot, root, &iter, 0, 0)
+				sink ^= (unsigned long)slot;
+		} else {
+			radix_tree_for_each_slot(slot, root, &iter, 0)
+				sink ^= (unsigned long)slot;
+		}
+	}
+	clock_gettime(CLOCK_MONOTONIC, &finish);
+
+	nsec = (finish.tv_sec - start.tv_sec) * NSEC_PER_SEC +
+	       (finish.tv_nsec - start.tv_nsec);
+
+#ifdef BENCHMARK
+	if (loops == 1 && nsec * 5 < NSEC_PER_SEC) {
+		loops = NSEC_PER_SEC / nsec / 4 + 1;
+		goto again;
+	}
+#endif
+
+	nsec /= loops;
+	return nsec;
+}
+
+static void benchmark_size(unsigned long size, unsigned long step, int order)
+{
+	RADIX_TREE(tree, GFP_KERNEL);
+	long long normal, tagged;
+	unsigned long index;
+
+	for (index = 0 ; index < size ; index += step) {
+		item_insert_order(&tree, index, order);
+		radix_tree_tag_set(&tree, index, 0);
+	}
+
+	tagged = benchmark_iter(&tree, true);
+	normal = benchmark_iter(&tree, false);
+
+	printf("Size %ld, step %6ld, order %d tagged %10lld ns, normal %10lld ns\n",
+		size, step, order, tagged, normal);
+
+	item_kill_tree(&tree);
+	rcu_barrier();
+}
+
+void benchmark(void)
+{
+	unsigned long size[] = {1 << 10, 1 << 20, 0};
+	unsigned long step[] = {1, 2, 7, 15, 63, 64, 65,
+				128, 256, 512, 12345, 0};
+	int c, s;
+
+	printf("starting benchmarks\n");
+	printf("RADIX_TREE_MAP_SHIFT = %d\n", RADIX_TREE_MAP_SHIFT);
+
+	for (c = 0; size[c]; c++)
+		for (s = 0; step[s]; s++)
+			benchmark_size(size[c], step[s], 0);
+
+	for (c = 0; size[c]; c++)
+		for (s = 0; step[s]; s++)
+			benchmark_size(size[c], step[s] << 9, 9);
+}
diff --git a/tools/testing/radix-tree/linux/kernel.h b/tools/testing/radix-tree/linux/kernel.h
index be98a47..dbe4b92 100644
--- a/tools/testing/radix-tree/linux/kernel.h
+++ b/tools/testing/radix-tree/linux/kernel.h
@@ -10,7 +10,11 @@
 #include "../../include/linux/compiler.h"
 #include "../../../include/linux/kconfig.h"
 
+#ifdef BENCHMARK
+#define RADIX_TREE_MAP_SHIFT	6
+#else
 #define RADIX_TREE_MAP_SHIFT	3
+#endif
 
 #ifndef NULL
 #define NULL	0
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 5ddec8e..f43706c 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -352,6 +352,8 @@ int main(int argc, char **argv)
 	/* Free any remaining preallocated nodes */
 	radix_tree_callback(NULL, CPU_DEAD, NULL);
 
+	benchmark();
+
 	sleep(1);
 	printf("after sleep(1): %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index 8cd666a..f2dc35f 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -30,6 +30,7 @@ void item_kill_tree(struct radix_tree_root *root);
 void tag_check(void);
 void multiorder_checks(void);
 void iteration_test(void);
+void benchmark(void);
 
 struct item *
 item_tag_set(struct radix_tree_root *root, unsigned long index, int tag);
-- 
2.10.2

[PATCH 07/29] radix tree test suite: Use rcu_barrier

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

Calling rcu_barrier() allows all of the rcu-freed memory to be actually
returned to the pool, and allows nr_allocated to return to 0.  As well
as allowing diffs between runs to be more useful, it also lets us
pinpoint leaks more effectively.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 tools/testing/radix-tree/main.c      | 12 ++++++++++--
 tools/testing/radix-tree/tag_check.c |  5 +++++
 2 files changed, 15 insertions(+), 2 deletions(-)

diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index f43706c..8621542 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -295,24 +295,31 @@ static void single_thread_tests(bool long_run)
 	printf("starting single_thread_tests: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	multiorder_checks();
+	rcu_barrier();
 	printf("after multiorder_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	locate_check();
+	rcu_barrier();
 	printf("after locate_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	tag_check();
+	rcu_barrier();
 	printf("after tag_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	gang_check();
+	rcu_barrier();
 	printf("after gang_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	add_and_check();
+	rcu_barrier();
 	printf("after add_and_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	dynamic_height_check();
+	rcu_barrier();
 	printf("after dynamic_height_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	big_gang_check(long_run);
+	rcu_barrier();
 	printf("after big_gang_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	for (i = 0; i < (long_run ? 2000 : 3); i++) {
@@ -320,6 +327,7 @@ static void single_thread_tests(bool long_run)
 		printf("%d ", i);
 		fflush(stdout);
 	}
+	rcu_barrier();
 	printf("after copy_tag_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 }
@@ -354,8 +362,8 @@ int main(int argc, char **argv)
 
 	benchmark();
 
-	sleep(1);
-	printf("after sleep(1): %d allocated, preempt %d\n",
+	rcu_barrier();
+	printf("after rcu_barrier: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
 	rcu_unregister_thread();
 
diff --git a/tools/testing/radix-tree/tag_check.c b/tools/testing/radix-tree/tag_check.c
index b0ac057..186f6e4 100644
--- a/tools/testing/radix-tree/tag_check.c
+++ b/tools/testing/radix-tree/tag_check.c
@@ -51,6 +51,7 @@ void simple_checks(void)
 	verify_tag_consistency(&tree, 1);
 	printf("before item_kill_tree: %d allocated\n", nr_allocated);
 	item_kill_tree(&tree);
+	rcu_barrier();
 	printf("after item_kill_tree: %d allocated\n", nr_allocated);
 }
 
@@ -331,12 +332,16 @@ void tag_check(void)
 	single_check();
 	extend_checks();
 	contract_checks();
+	rcu_barrier();
 	printf("after extend_checks: %d allocated\n", nr_allocated);
 	__leak_check();
 	leak_check();
+	rcu_barrier();
 	printf("after leak_check: %d allocated\n", nr_allocated);
 	simple_checks();
+	rcu_barrier();
 	printf("after simple_checks: %d allocated\n", nr_allocated);
 	thrash_tags();
+	rcu_barrier();
 	printf("after thrash_tags: %d allocated\n", nr_allocated);
 }
-- 
2.10.2

[PATCH 08/29] tools: Add more bitmap functions

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

I need the following functions for the radix tree:

bitmap_fill
find_next_zero_bit
bitmap_empty
bitmap_full

Copy the implementations from include/linux/bitmap.h and lib/find_bit.c
---
 tools/include/linux/bitmap.h | 26 ++++++++++++++++++++++++++
 tools/lib/find_bit.c         |  8 ++++++++
 2 files changed, 34 insertions(+)

diff --git a/tools/include/linux/bitmap.h b/tools/include/linux/bitmap.h
index 43c1c50..eef41d5 100644
--- a/tools/include/linux/bitmap.h
+++ b/tools/include/linux/bitmap.h
@@ -35,6 +35,32 @@ static inline void bitmap_zero(unsigned long *dst, int nbits)
 	}
 }
 
+static inline void bitmap_fill(unsigned long *dst, unsigned int nbits)
+{
+	unsigned int nlongs = BITS_TO_LONGS(nbits);
+	if (!small_const_nbits(nbits)) {
+		unsigned int len = (nlongs - 1) * sizeof(unsigned long);
+		memset(dst, 0xff,  len);
+	}
+	dst[nlongs - 1] = BITMAP_LAST_WORD_MASK(nbits);
+}
+
+static inline int bitmap_empty(const unsigned long *src, unsigned nbits)
+{
+	if (small_const_nbits(nbits))
+		return ! (*src & BITMAP_LAST_WORD_MASK(nbits));
+
+	return find_first_bit(src, nbits) == nbits;
+}
+
+static inline int bitmap_full(const unsigned long *src, unsigned int nbits)
+{
+	if (small_const_nbits(nbits))
+		return ! (~(*src) & BITMAP_LAST_WORD_MASK(nbits));
+
+	return find_first_zero_bit(src, nbits) == nbits;
+}
+
 static inline int bitmap_weight(const unsigned long *src, int nbits)
 {
 	if (small_const_nbits(nbits))
diff --git a/tools/lib/find_bit.c b/tools/lib/find_bit.c
index 9122a9e..de26b6f 100644
--- a/tools/lib/find_bit.c
+++ b/tools/lib/find_bit.c
@@ -66,6 +66,14 @@ unsigned long find_next_bit(const unsigned long *addr, unsigned long size,
 }
 #endif
 
+#ifndef find_next_zero_bit
+unsigned long find_next_zero_bit(const unsigned long *addr, unsigned long size,
+				unsigned long offset)
+{
+	return _find_next_bit(addr, size, offset, ~0UL);
+}
+#endif
+
 #ifndef find_first_bit
 /*
  * Find the first set bit in a memory region.
-- 
2.10.2

[PATCH 09/29] radix tree test suite: Use common find-bit code

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Remove the old find_next_bit code in favour of linking in the find_bit
code from tools/lib.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 tools/testing/radix-tree/Makefile                  |  7 ++-
 tools/testing/radix-tree/find_next_bit.c           | 57 ----------------------
 tools/testing/radix-tree/linux/bitops.h            | 40 +++++++++------
 tools/testing/radix-tree/linux/bitops/non-atomic.h | 13 +++--
 tools/testing/radix-tree/linux/kernel.h            | 11 +++++
 5 files changed, 48 insertions(+), 80 deletions(-)
 delete mode 100644 tools/testing/radix-tree/find_next_bit.c

diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
index 08283a8..3635e4d 100644
--- a/tools/testing/radix-tree/Makefile
+++ b/tools/testing/radix-tree/Makefile
@@ -18,7 +18,12 @@ main:	$(OFILES)
 clean:
 	$(RM) -f $(TARGETS) *.o radix-tree.c
 
-$(OFILES): *.h */*.h ../../../include/linux/radix-tree.h ../../include/linux/*.h
+find_next_bit.o: ../../lib/find_bit.c
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+$(OFILES): *.h */*.h \
+	../../include/linux/*.h \
+	../../../include/linux/radix-tree.h
 
 radix-tree.c: ../../../lib/radix-tree.c
 	sed -e 's/^static //' -e 's/__always_inline //' -e 's/inline //' < $< > $@
diff --git a/tools/testing/radix-tree/find_next_bit.c b/tools/testing/radix-tree/find_next_bit.c
deleted file mode 100644
index d1c2178..0000000
--- a/tools/testing/radix-tree/find_next_bit.c
+++ /dev/null
@@ -1,57 +0,0 @@
-/* find_next_bit.c: fallback find next bit implementation
- *
- * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * 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.
- */
-
-#include <linux/types.h>
-#include <linux/bitops.h>
-
-#define BITOP_WORD(nr)		((nr) / BITS_PER_LONG)
-
-/*
- * Find the next set bit in a memory region.
- */
-unsigned long find_next_bit(const unsigned long *addr, unsigned long size,
-			    unsigned long offset)
-{
-	const unsigned long *p = addr + BITOP_WORD(offset);
-	unsigned long result = offset & ~(BITS_PER_LONG-1);
-	unsigned long tmp;
-
-	if (offset >= size)
-		return size;
-	size -= result;
-	offset %= BITS_PER_LONG;
-	if (offset) {
-		tmp = *(p++);
-		tmp &= (~0UL << offset);
-		if (size < BITS_PER_LONG)
-			goto found_first;
-		if (tmp)
-			goto found_middle;
-		size -= BITS_PER_LONG;
-		result += BITS_PER_LONG;
-	}
-	while (size & ~(BITS_PER_LONG-1)) {
-		if ((tmp = *(p++)))
-			goto found_middle;
-		result += BITS_PER_LONG;
-		size -= BITS_PER_LONG;
-	}
-	if (!size)
-		return result;
-	tmp = *p;
-
-found_first:
-	tmp &= (~0UL >> (BITS_PER_LONG - size));
-	if (tmp == 0UL)		/* Are any bits set? */
-		return result + size;	/* Nope. */
-found_middle:
-	return result + __ffs(tmp);
-}
diff --git a/tools/testing/radix-tree/linux/bitops.h b/tools/testing/radix-tree/linux/bitops.h
index 71d58427..a13e9bc 100644
--- a/tools/testing/radix-tree/linux/bitops.h
+++ b/tools/testing/radix-tree/linux/bitops.h
@@ -2,9 +2,14 @@
 #define _ASM_GENERIC_BITOPS_NON_ATOMIC_H_
 
 #include <linux/types.h>
+#include <linux/bitops/find.h>
+#include <linux/bitops/hweight.h>
+#include <linux/kernel.h>
 
-#define BITOP_MASK(nr)		(1UL << ((nr) % BITS_PER_LONG))
-#define BITOP_WORD(nr)		((nr) / BITS_PER_LONG)
+#define BIT_MASK(nr)		(1UL << ((nr) % BITS_PER_LONG))
+#define BIT_WORD(nr)		((nr) / BITS_PER_LONG)
+#define BITS_PER_BYTE		8
+#define BITS_TO_LONGS(nr)	DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
 
 /**
  * __set_bit - Set a bit in memory
@@ -17,16 +22,16 @@
  */
 static inline void __set_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 
 	*p  |= mask;
 }
 
 static inline void __clear_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 
 	*p &= ~mask;
 }
@@ -42,8 +47,8 @@ static inline void __clear_bit(int nr, volatile unsigned long *addr)
  */
 static inline void __change_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 
 	*p ^= mask;
 }
@@ -59,8 +64,8 @@ static inline void __change_bit(int nr, volatile unsigned long *addr)
  */
 static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 	unsigned long old = *p;
 
 	*p = old | mask;
@@ -78,8 +83,8 @@ static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
  */
 static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 	unsigned long old = *p;
 
 	*p = old & ~mask;
@@ -90,8 +95,8 @@ static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
 static inline int __test_and_change_bit(int nr,
 					    volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
-	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
+	unsigned long mask = BIT_MASK(nr);
+	unsigned long *p = ((unsigned long *)addr) + BIT_WORD(nr);
 	unsigned long old = *p;
 
 	*p = old ^ mask;
@@ -105,7 +110,7 @@ static inline int __test_and_change_bit(int nr,
  */
 static inline int test_bit(int nr, const volatile unsigned long *addr)
 {
-	return 1UL & (addr[BITOP_WORD(nr)] >> (nr & (BITS_PER_LONG-1)));
+	return 1UL & (addr[BIT_WORD(nr)] >> (nr & (BITS_PER_LONG-1)));
 }
 
 /**
@@ -147,4 +152,9 @@ unsigned long find_next_bit(const unsigned long *addr,
 			    unsigned long size,
 			    unsigned long offset);
 
+static inline unsigned long hweight_long(unsigned long w)
+{
+	return sizeof(w) == 4 ? hweight32(w) : hweight64(w);
+}
+
 #endif /* _ASM_GENERIC_BITOPS_NON_ATOMIC_H_ */
diff --git a/tools/testing/radix-tree/linux/bitops/non-atomic.h b/tools/testing/radix-tree/linux/bitops/non-atomic.h
index 46a825c..6a1bcb9 100644
--- a/tools/testing/radix-tree/linux/bitops/non-atomic.h
+++ b/tools/testing/radix-tree/linux/bitops/non-atomic.h
@@ -3,7 +3,6 @@
 
 #include <asm/types.h>
 
-#define BITOP_MASK(nr)		(1UL << ((nr) % BITS_PER_LONG))
 #define BITOP_WORD(nr)		((nr) / BITS_PER_LONG)
 
 /**
@@ -17,7 +16,7 @@
  */
 static inline void __set_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 
 	*p  |= mask;
@@ -25,7 +24,7 @@ static inline void __set_bit(int nr, volatile unsigned long *addr)
 
 static inline void __clear_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 
 	*p &= ~mask;
@@ -42,7 +41,7 @@ static inline void __clear_bit(int nr, volatile unsigned long *addr)
  */
 static inline void __change_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 
 	*p ^= mask;
@@ -59,7 +58,7 @@ static inline void __change_bit(int nr, volatile unsigned long *addr)
  */
 static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 	unsigned long old = *p;
 
@@ -78,7 +77,7 @@ static inline int __test_and_set_bit(int nr, volatile unsigned long *addr)
  */
 static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 	unsigned long old = *p;
 
@@ -90,7 +89,7 @@ static inline int __test_and_clear_bit(int nr, volatile unsigned long *addr)
 static inline int __test_and_change_bit(int nr,
 					    volatile unsigned long *addr)
 {
-	unsigned long mask = BITOP_MASK(nr);
+	unsigned long mask = BIT_MASK(nr);
 	unsigned long *p = ((unsigned long *)addr) + BITOP_WORD(nr);
 	unsigned long old = *p;
 
diff --git a/tools/testing/radix-tree/linux/kernel.h b/tools/testing/radix-tree/linux/kernel.h
index dbe4b92..23e77f5 100644
--- a/tools/testing/radix-tree/linux/kernel.h
+++ b/tools/testing/radix-tree/linux/kernel.h
@@ -47,4 +47,15 @@ static inline int in_interrupt(void)
 {
 	return 0;
 }
+
+/*
+ * This looks more complex than it should be. But we need to
+ * get the type for the ~ right in round_down (it needs to be
+ * as wide as the result!), and we want to evaluate the macro
+ * arguments just once each.
+ */
+#define __round_mask(x, y) ((__typeof__(x))((y)-1))
+#define round_up(x, y) ((((x)-1) | __round_mask(x, y))+1)
+#define round_down(x, y) ((x) & ~__round_mask(x, y))
+
 #endif /* _KERNEL_H */
-- 
2.10.2

[PATCH 10/29] radix-tree: Add radix_tree_join

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

This new function allows for the replacement of many smaller entries in
the radix tree with one larger multiorder entry.  From the point of view
of an RCU walker, they may see a mixture of the smaller entries and the
large entry during the same walk, but they will never see NULL for an
index which was populated before the join.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 include/linux/radix-tree.h            |   2 +
 lib/radix-tree.c                      | 159 +++++++++++++++++++++++++++-------
 tools/testing/radix-tree/multiorder.c |  32 +++++++
 3 files changed, 163 insertions(+), 30 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index af3581b..1efd81f 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -319,6 +319,8 @@ static inline void radix_tree_preload_end(void)
 	preempt_enable();
 }
 
+int radix_tree_join(struct radix_tree_root *, unsigned long index,
+			unsigned new_order, void *);
 /**
  * struct radix_tree_iter - radix tree iterator state
  *
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 8e6d552..6a76252 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -314,18 +314,14 @@ static void radix_tree_node_rcu_free(struct rcu_head *head)
 {
 	struct radix_tree_node *node =
 			container_of(head, struct radix_tree_node, rcu_head);
-	int i;
 
 	/*
-	 * must only free zeroed nodes into the slab. radix_tree_shrink
-	 * can leave us with a non-NULL entry in the first slot, so clear
-	 * that here to make sure.
+	 * Must only free zeroed nodes into the slab.  We can be left with
+	 * non-NULL entries by radix_tree_free_nodes, so clear the entries
+	 * and tags here.
 	 */
-	for (i = 0; i < RADIX_TREE_MAX_TAGS; i++)
-		tag_clear(node, i, 0);
-
-	node->slots[0] = NULL;
-	node->count = 0;
+	memset(node->slots, 0, sizeof(node->slots));
+	memset(node->tags, 0, sizeof(node->tags));
 
 	kmem_cache_free(radix_tree_node_cachep, node);
 }
@@ -563,14 +559,14 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 	shift = radix_tree_load_root(root, &child, &maxindex);
 
 	/* Make sure the tree is high enough.  */
+	if (order > 0 && max == ((1UL << order) - 1))
+		max++;
 	if (max > maxindex) {
 		int error = radix_tree_extend(root, max, shift);
 		if (error < 0)
 			return error;
 		shift = error;
 		child = root->rnode;
-		if (order == shift)
-			shift += RADIX_TREE_MAP_SHIFT;
 	}
 
 	while (shift > order) {
@@ -582,6 +578,7 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 				return -ENOMEM;
 			child->shift = shift;
 			child->offset = offset;
+			child->count = 0;
 			child->parent = node;
 			rcu_assign_pointer(*slot, node_to_entry(child));
 			if (node)
@@ -595,31 +592,113 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 		slot = &node->slots[offset];
 	}
 
+	if (nodep)
+		*nodep = node;
+	if (slotp)
+		*slotp = slot;
+	return 0;
+}
+
 #ifdef CONFIG_RADIX_TREE_MULTIORDER
-	/* Insert pointers to the canonical entry */
-	if (order > shift) {
-		unsigned i, n = 1 << (order - shift);
+/*
+ * Free any nodes below this node.  The tree is presumed to not need
+ * shrinking, and any user data in the tree is presumed to not need a
+ * destructor called on it.  If we need to add a destructor, we can
+ * add that functionality later.  Note that we may not clear tags or
+ * slots from the tree as an RCU walker may still have a pointer into
+ * this subtree.  We could replace the entries with RADIX_TREE_RETRY,
+ * but we'll still have to clear those in rcu_free.
+ */
+static void radix_tree_free_nodes(struct radix_tree_node *node)
+{
+	unsigned offset = 0;
+	struct radix_tree_node *child = entry_to_node(node);
+
+	for (;;) {
+		void *entry = child->slots[offset];
+		if (radix_tree_is_internal_node(entry) &&
+					!is_sibling_entry(child, entry)) {
+			child = entry_to_node(entry);
+			offset = 0;
+			continue;
+		}
+		offset++;
+		while (offset == RADIX_TREE_MAP_SIZE) {
+			struct radix_tree_node *old = child;
+			offset = child->offset + 1;
+			child = child->parent;
+			radix_tree_node_free(old);
+			if (old == entry_to_node(node))
+				return;
+		}
+	}
+}
+
+static inline int insert_entries(struct radix_tree_node *node, void **slot,
+				void *ptr, unsigned order, bool replace)
+{
+	struct radix_tree_node *child;
+	unsigned i, n, tag, offset, tags = 0;
+
+	if (node) {
+		n = 1 << (order - node->shift);
+		offset = get_slot_offset(node, slot);
+	} else {
+		n = 1;
+		offset = 0;
+	}
+
+	if (n > 1) {
 		offset = offset & ~(n - 1);
 		slot = &node->slots[offset];
-		child = node_to_entry(slot);
-		for (i = 0; i < n; i++) {
-			if (slot[i])
+	}
+	child = node_to_entry(slot);
+
+	for (i = 0; i < n; i++) {
+		if (slot[i]) {
+			if (replace) {
+				node->count--;
+				for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+					if (tag_get(node, tag, offset + i))
+						tags |= 1 << tag;
+			} else
 				return -EEXIST;
 		}
+	}
 
-		for (i = 1; i < n; i++) {
+	for (i = 0; i < n; i++) {
+		struct radix_tree_node *old = slot[i];
+		if (i) {
 			rcu_assign_pointer(slot[i], child);
-			node->count++;
+			for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+				if (tags & (1 << tag))
+					tag_clear(node, tag, offset + i);
+		} else {
+			rcu_assign_pointer(slot[i], ptr);
+			for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+				if (tags & (1 << tag))
+					tag_set(node, tag, offset);
 		}
+		if (radix_tree_is_internal_node(old) &&
+					!is_sibling_entry(node, old))
+			radix_tree_free_nodes(old);
 	}
-#endif
-
-	if (nodep)
-		*nodep = node;
-	if (slotp)
-		*slotp = slot;
-	return 0;
+	if (node)
+		node->count += n;
+	return n;
+}
+#else
+static inline int insert_entries(struct radix_tree_node *node, void **slot,
+				void *ptr, unsigned order, bool replace)
+{
+	if (*slot)
+		return -EEXIST;
+	rcu_assign_pointer(*slot, ptr);
+	if (node)
+		node->count++;
+	return 1;
 }
+#endif
 
 /**
  *	__radix_tree_insert    -    insert into a radix tree
@@ -642,13 +721,13 @@ int __radix_tree_insert(struct radix_tree_root *root, unsigned long index,
 	error = __radix_tree_create(root, index, order, &node, &slot);
 	if (error)
 		return error;
-	if (*slot != NULL)
-		return -EEXIST;
-	rcu_assign_pointer(*slot, item);
+
+	error = insert_entries(node, slot, item, order, false);
+	if (error < 0)
+		return error;
 
 	if (node) {
 		unsigned offset = get_slot_offset(node, slot);
-		node->count++;
 		BUG_ON(tag_get(node, 0, offset));
 		BUG_ON(tag_get(node, 1, offset));
 		BUG_ON(tag_get(node, 2, offset));
@@ -746,6 +825,26 @@ void *radix_tree_lookup(struct radix_tree_root *root, unsigned long index)
 }
 EXPORT_SYMBOL(radix_tree_lookup);
 
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+int radix_tree_join(struct radix_tree_root *root, unsigned long index,
+			unsigned order, void *item)
+{
+	struct radix_tree_node *node;
+	void **slot;
+	int error;
+
+	BUG_ON(radix_tree_is_internal_node(item));
+
+	error = __radix_tree_create(root, index, order, &node, &slot);
+	if (!error)
+		error = insert_entries(node, slot, item, order, true);
+	if (error > 0)
+		error = 0;
+
+	return error;
+}
+#endif
+
 /**
  *	radix_tree_tag_set - set a tag on a radix tree node
  *	@root:		radix tree root
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 05d7bc4..4c66acc 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -325,6 +325,37 @@ void multiorder_tagged_iteration(void)
 	item_kill_tree(&tree);
 }
 
+static void __multiorder_join(unsigned long index,
+				unsigned order1, unsigned order2)
+{
+	unsigned long loc;
+	void *item, *item2 = item_create(index + 1);
+	RADIX_TREE(tree, GFP_KERNEL);
+
+	item_insert_order(&tree, index, order2);
+	item = radix_tree_lookup(&tree, index);
+	radix_tree_join(&tree, index + 1, order1, item2);
+	loc = radix_tree_locate_item(&tree, item);
+	if (loc == -1)
+		free(item);
+	item = radix_tree_lookup(&tree, index + 1);
+	assert(item == item2);
+	item_kill_tree(&tree);
+}
+
+static void multiorder_join(void)
+{
+	int i, j, idx;
+
+	for (idx = 0; idx < 1024; idx = idx * 2 + 3) {
+		for (i = 1; i < 15; i++) {
+			for (j = 0; j < i; j++) {
+				__multiorder_join(idx, i, j);
+			}
+		}
+	}
+}
+
 void multiorder_checks(void)
 {
 	int i;
@@ -342,4 +373,5 @@ void multiorder_checks(void)
 	multiorder_tag_tests();
 	multiorder_iteration();
 	multiorder_tagged_iteration();
+	multiorder_join();
 }
-- 
2.10.2

[PATCH 11/29] radix-tree: Add radix_tree_split

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

This new function splits a larger multiorder entry into smaller entries
(potentially multi-order entries).  These entries are initialised to
RADIX_TREE_RETRY to ensure that RCU walkers who see this state aren't
confused.  The caller should then call radix_tree_for_each_slot() and
radix_tree_replace_slot() in order to turn these retry entries into the
intended new entries.  Tags are replicated from the original multiorder
entry into each new entry.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 include/linux/radix-tree.h            |   3 +
 lib/radix-tree.c                      | 109 ++++++++++++++++++++++++++++++++--
 tools/testing/radix-tree/multiorder.c |  26 ++++++++
 3 files changed, 134 insertions(+), 4 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 1efd81f..f5518f1 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -319,8 +319,11 @@ static inline void radix_tree_preload_end(void)
 	preempt_enable();
 }
 
+int radix_tree_split(struct radix_tree_root *, unsigned long index,
+			unsigned new_order);
 int radix_tree_join(struct radix_tree_root *, unsigned long index,
 			unsigned new_order, void *);
+
 /**
  * struct radix_tree_iter - radix tree iterator state
  *
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 6a76252..eaf0f353 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -231,7 +231,10 @@ static void dump_node(struct radix_tree_node *node, unsigned long index)
 		void *entry = node->slots[i];
 		if (!entry)
 			continue;
-		if (is_sibling_entry(node, entry)) {
+		if (entry == RADIX_TREE_RETRY) {
+			pr_debug("radix retry offset %ld indices %ld-%ld\n",
+					i, first, last);
+		} else if (is_sibling_entry(node, entry)) {
 			pr_debug("radix sblng %p offset %ld val %p indices %ld-%ld\n",
 					entry, i,
 					*(void **)entry_to_node(entry),
@@ -641,7 +644,10 @@ static inline int insert_entries(struct radix_tree_node *node, void **slot,
 	unsigned i, n, tag, offset, tags = 0;
 
 	if (node) {
-		n = 1 << (order - node->shift);
+		if (order > node->shift)
+			n = 1 << (order - node->shift);
+		else
+			n = 1;
 		offset = get_slot_offset(node, slot);
 	} else {
 		n = 1;
@@ -680,7 +686,8 @@ static inline int insert_entries(struct radix_tree_node *node, void **slot,
 					tag_set(node, tag, offset);
 		}
 		if (radix_tree_is_internal_node(old) &&
-					!is_sibling_entry(node, old))
+					!is_sibling_entry(node, old) &&
+					(old != RADIX_TREE_RETRY))
 			radix_tree_free_nodes(old);
 	}
 	if (node)
@@ -843,6 +850,98 @@ int radix_tree_join(struct radix_tree_root *root, unsigned long index,
 
 	return error;
 }
+
+int radix_tree_split(struct radix_tree_root *root, unsigned long index,
+				unsigned order)
+{
+	struct radix_tree_node *parent, *node, *child;
+	void **slot;
+	unsigned int offset, end;
+	unsigned n, tag, tags = 0;
+
+	if (!__radix_tree_lookup(root, index, &parent, &slot))
+		return -ENOENT;
+	if (!parent)
+		return -ENOENT;
+
+	offset = get_slot_offset(parent, slot);
+
+	for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+		if (tag_get(parent, tag, offset))
+			tags |= 1 << tag;
+
+	for (end = offset + 1; end < RADIX_TREE_MAP_SIZE; end++) {
+		if (!is_sibling_entry(parent, parent->slots[end]))
+			break;
+		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+			if (tags & (1 << tag))
+				tag_set(parent, tag, end);
+		/* rcu_assign_pointer ensures tags are set before RETRY */
+		rcu_assign_pointer(parent->slots[end], RADIX_TREE_RETRY);
+	}
+
+	if (order == parent->shift)
+		return 0;
+	if (order > parent->shift) {
+		while (offset < end)
+			offset += insert_entries(parent, &parent->slots[offset],
+					RADIX_TREE_RETRY, order, true);
+		return 0;
+	}
+
+	node = parent;
+
+	for (;;) {
+		if (node->shift > order) {
+			child = radix_tree_node_alloc(root);
+			if (!child)
+				goto nomem;
+			child->shift = node->shift - RADIX_TREE_MAP_SHIFT;
+			child->offset = offset;
+			child->count = 0;
+			child->parent = node;
+			if (node != parent) {
+				node->count++;
+				node->slots[offset] = node_to_entry(child);
+				for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+					if (tags & (1 << tag))
+						tag_set(node, tag, offset);
+			}
+
+			node = child;
+			offset = 0;
+			continue;
+		}
+
+		n = insert_entries(node, &node->slots[offset],
+					RADIX_TREE_RETRY, order, false);
+		BUG_ON(n > RADIX_TREE_MAP_SIZE);
+
+		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+			if (tags & (1 << tag))
+				tag_set(node, tag, offset);
+		offset += n;
+
+		while (offset == RADIX_TREE_MAP_SIZE) {
+			if (node == parent)
+				break;
+			offset = node->offset;
+			child = node;
+			node = node->parent;
+			rcu_assign_pointer(node->slots[offset],
+						node_to_entry(child));
+			offset++;
+		}
+		if ((node == parent) && (offset == end))
+			return 0;
+	}
+
+ nomem:
+	/* Shouldn't happen; did user forget to preload? */
+	/* TODO: free all the allocated nodes */
+	WARN_ON(1);
+	return -ENOMEM;
+}
 #endif
 
 /**
@@ -1081,8 +1180,10 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 			child = rcu_dereference_raw(node->slots[offset]);
 		}
 
-		if ((child == NULL) || (child == RADIX_TREE_RETRY))
+		if (!child)
 			goto restart;
+		if (child == RADIX_TREE_RETRY)
+			break;
 	} while (radix_tree_is_internal_node(child));
 
 	/* Update the iterator state */
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 4c66acc..d9e8155 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -356,6 +356,31 @@ static void multiorder_join(void)
 	}
 }
 
+static void __multiorder_split(int old_order, int new_order)
+{
+	RADIX_TREE(tree, GFP_KERNEL);
+	void **slot;
+	struct radix_tree_iter iter;
+
+	item_insert_order(&tree, 0, old_order);
+	radix_tree_tag_set(&tree, 0, 2);
+	radix_tree_split(&tree, 0, new_order);
+	radix_tree_for_each_slot(slot, &tree, &iter, 0) {
+		radix_tree_replace_slot(slot, item_create(iter.index));
+	}
+
+	item_kill_tree(&tree);
+}
+
+static void multiorder_split(void)
+{
+	int i, j;
+
+	for (i = 9; i < 19; i++)
+		for (j = 0; j < i; j++)
+			__multiorder_split(i, j);
+}
+
 void multiorder_checks(void)
 {
 	int i;
@@ -374,4 +399,5 @@ void multiorder_checks(void)
 	multiorder_iteration();
 	multiorder_tagged_iteration();
 	multiorder_join();
+	multiorder_split();
 }
-- 
2.10.2

[PATCH 12/29] radix-tree: Add radix_tree_split_preload()

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Calculate how many nodes we need to allocate to split an old_order entry
into multiple entries, each of size new_order.  The test suite checks that
we allocated exactly the right number of nodes; neither too many (checked
by rtp->nr == 0), nor too few (checked by comparing nr_allocated before
and after the call to radix_tree_split()).

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 include/linux/radix-tree.h            |  1 +
 lib/radix-tree.c                      | 24 +++++++++++++++++++++++-
 tools/testing/radix-tree/multiorder.c | 32 ++++++++++++++++++++++++++++++--
 tools/testing/radix-tree/test.h       |  5 +++++
 4 files changed, 59 insertions(+), 3 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index f5518f1..8ffb051 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -319,6 +319,7 @@ static inline void radix_tree_preload_end(void)
 	preempt_enable();
 }
 
+int radix_tree_split_preload(unsigned old_order, unsigned new_order, gfp_t);
 int radix_tree_split(struct radix_tree_root *, unsigned long index,
 			unsigned new_order);
 int radix_tree_join(struct radix_tree_root *, unsigned long index,
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index eaf0f353..6d73575 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -344,7 +344,7 @@ radix_tree_node_free(struct radix_tree_node *node)
  * To make use of this facility, the radix tree must be initialised without
  * __GFP_DIRECT_RECLAIM being passed to INIT_RADIX_TREE().
  */
-static int __radix_tree_preload(gfp_t gfp_mask, int nr)
+static int __radix_tree_preload(gfp_t gfp_mask, unsigned nr)
 {
 	struct radix_tree_preload *rtp;
 	struct radix_tree_node *node;
@@ -410,6 +410,28 @@ int radix_tree_maybe_preload(gfp_t gfp_mask)
 }
 EXPORT_SYMBOL(radix_tree_maybe_preload);
 
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+/*
+ * Preload with enough objects to ensure that we can split a single entry
+ * of order @old_order into many entries of size @new_order
+ */
+int radix_tree_split_preload(unsigned int old_order, unsigned int new_order,
+							gfp_t gfp_mask)
+{
+	unsigned top = 1 << (old_order % RADIX_TREE_MAP_SHIFT);
+	unsigned layers = (old_order / RADIX_TREE_MAP_SHIFT) -
+				(new_order / RADIX_TREE_MAP_SHIFT);
+	unsigned nr = 0;
+
+	WARN_ON_ONCE(!gfpflags_allow_blocking(gfp_mask));
+	BUG_ON(new_order >= old_order);
+
+	while (layers--)
+		nr = nr * RADIX_TREE_MAP_SIZE + 1;
+	return __radix_tree_preload(gfp_mask, top * nr);
+}
+#endif
+
 /*
  * The same as function above, but preload number of nodes required to insert
  * (1 << order) continuous naturally-aligned elements.
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index d9e8155..25e0463 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -356,18 +356,46 @@ static void multiorder_join(void)
 	}
 }
 
+static void check_mem(unsigned old_order, unsigned new_order, unsigned alloc)
+{
+	struct radix_tree_preload *rtp = &radix_tree_preloads;
+	if (rtp->nr != 0)
+		printf("split(%u %u) remaining %u\n", old_order, new_order,
+							rtp->nr);
+	/*
+	 * Can't check for equality here as some nodes may have been
+	 * RCU-freed while we ran.  But we should never finish with more
+	 * nodes allocated since they should have all been preloaded.
+	 */
+	if (nr_allocated > alloc)
+		printf("split(%u %u) allocated %u %u\n", old_order, new_order,
+							alloc, nr_allocated);
+}
+
 static void __multiorder_split(int old_order, int new_order)
 {
-	RADIX_TREE(tree, GFP_KERNEL);
+	RADIX_TREE(tree, GFP_ATOMIC);
 	void **slot;
 	struct radix_tree_iter iter;
+	unsigned alloc;
+
+	radix_tree_preload(GFP_KERNEL);
+	assert(item_insert_order(&tree, 0, old_order) == 0);
+	radix_tree_preload_end();
+
+	/* Wipe out the preloaded cache or it'll confuse check_mem() */
+	radix_tree_callback(NULL, CPU_DEAD, NULL);
 
-	item_insert_order(&tree, 0, old_order);
 	radix_tree_tag_set(&tree, 0, 2);
+
+	radix_tree_split_preload(old_order, new_order, GFP_KERNEL);
+	alloc = nr_allocated;
 	radix_tree_split(&tree, 0, new_order);
+	check_mem(old_order, new_order, alloc);
 	radix_tree_for_each_slot(slot, &tree, &iter, 0) {
 		radix_tree_replace_slot(slot, item_create(iter.index));
 	}
+	radix_tree_preload_end();
 
 	item_kill_tree(&tree);
 }
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index f2dc35f..33d2b6b 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -49,3 +49,8 @@ unsigned long node_maxindex(struct radix_tree_node *);
 unsigned long shift_maxindex(unsigned int shift);
 int radix_tree_callback(struct notifier_block *nfb,
 			unsigned long action, void *hcpu);
+struct radix_tree_preload {
+	unsigned nr;
+	struct radix_tree_node *nodes;
+};
+extern struct radix_tree_preload radix_tree_preloads;
-- 
2.10.2

[PATCH 13/29] radix-tree: Fix typo

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

---
 lib/radix-tree.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 6d73575..e917c56 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1147,7 +1147,7 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 	 * because RADIX_TREE_MAP_SHIFT < BITS_PER_LONG.
 	 *
 	 * This condition also used by radix_tree_next_slot() to stop
-	 * contiguous iterating, and forbid swithing to the next chunk.
+	 * contiguous iterating, and forbid switching to the next chunk.
 	 */
 	index = iter->next_index;
 	if (!index && iter->index)
-- 
2.10.2

[PATCH 14/29] radix-tree: Move rcu_head into a union with private_list

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

I want to be able to reference node->parent after freeing node.
Currently node->parent is in a union with rcu_head, so it is overwritten
when the node is put on the RCU list.  We know that private_list is not
referenced after the node is freed, so it is safe for these two members
to share space.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 include/linux/radix-tree.h | 14 ++++----------
 lib/radix-tree.c           |  1 +
 2 files changed, 5 insertions(+), 10 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 8ffb051..66fb8c0 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -88,18 +88,12 @@ struct radix_tree_node {
 	unsigned char	shift;	/* Bits remaining in each slot */
 	unsigned char	offset;	/* Slot offset in parent */
 	unsigned int	count;
+	struct radix_tree_node *parent;		/* Used when ascending tree */
+	void *private_data;			/* For tree user */
 	union {
-		struct {
-			/* Used when ascending tree */
-			struct radix_tree_node *parent;
-			/* For tree user */
-			void *private_data;
-		};
-		/* Used when freeing node */
-		struct rcu_head	rcu_head;
+		struct list_head private_list;	/* For tree user */
+		struct rcu_head	rcu_head;	/* Used when freeing node */
 	};
-	/* For tree user */
-	struct list_head private_list;
 	void __rcu	*slots[RADIX_TREE_MAP_SIZE];
 	unsigned long	tags[RADIX_TREE_MAX_TAGS][RADIX_TREE_TAG_LONGS];
 };
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index e917c56..baf4ba1 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -325,6 +325,7 @@ static void radix_tree_node_rcu_free(struct rcu_head *head)
 	 */
 	memset(node->slots, 0, sizeof(node->slots));
 	memset(node->tags, 0, sizeof(node->tags));
+	INIT_LIST_HEAD(&node->private_list);
 
 	kmem_cache_free(radix_tree_node_cachep, node);
 }
-- 
2.10.2

[PATCH 15/29] radix-tree: Create node_tag_set()

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Similar to node_tag_clear(), factor node_tag_set() out of
radix_tree_range_tag_if_tagged().

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 lib/radix-tree.c | 41 +++++++++++++++++++----------------------
 1 file changed, 19 insertions(+), 22 deletions(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index baf4ba1..bf1303f 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1028,6 +1028,22 @@ static void node_tag_clear(struct radix_tree_root *root,
 		root_tag_clear(root, tag);
 }
 
+static void node_tag_set(struct radix_tree_root *root,
+				struct radix_tree_node *node,
+				unsigned int tag, unsigned int offset)
+{
+	while (node) {
+		if (tag_get(node, tag, offset))
+			return;
+		tag_set(node, tag, offset);
+		offset = node->offset;
+		node = node->parent;
+	}
+
+	if (!root_tag_get(root, tag))
+		root_tag_set(root, tag);
+}
+
 /**
  *	radix_tree_tag_clear - clear a tag on a radix tree node
  *	@root:		radix tree root
@@ -1268,7 +1284,7 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
 		unsigned long nr_to_tag,
 		unsigned int iftag, unsigned int settag)
 {
-	struct radix_tree_node *parent, *node, *child;
+	struct radix_tree_node *node, *child;
 	unsigned long maxindex;
 	unsigned long tagged = 0;
 	unsigned long index = *first_indexp;
@@ -1303,22 +1319,8 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
 			continue;
 		}
 
-		/* tag the leaf */
 		tagged++;
-		tag_set(node, settag, offset);
-
-		/* walk back up the path tagging interior nodes */
-		parent = node;
-		for (;;) {
-			offset = parent->offset;
-			parent = parent->parent;
-			if (!parent)
-				break;
-			/* stop if we find a node with the tag already set */
-			if (tag_get(parent, settag, offset))
-				break;
-			tag_set(parent, settag, offset);
-		}
+		node_tag_set(root, node, settag, offset);
  next:
 		/* Go to next entry in node */
 		index = ((index >> node->shift) + 1) << node->shift;
@@ -1340,12 +1342,7 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
 		if (tagged >= nr_to_tag)
 			break;
 	}
-	/*
-	 * We need not to tag the root tag if there is no tag which is set with
-	 * settag within the range from *first_indexp to last_index.
-	 */
-	if (tagged > 0)
-		root_tag_set(root, settag);
+
 	*first_indexp = index;
 
 	return tagged;
-- 
2.10.2

[PATCH 16/29] radix-tree: Make radix_tree_find_next_bit more useful

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Since this function is specialised to the radix tree, pass in the node
and tag to calculate the address of the bitmap in radix_tree_find_next_bit()
instead of the caller.  Likewise, there is no need to pass in the size of
the bitmap.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 lib/radix-tree.c | 17 +++++++----------
 1 file changed, 7 insertions(+), 10 deletions(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index bf1303f..2c3fac4 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -191,13 +191,12 @@ static inline int any_tag_set(struct radix_tree_node *node, unsigned int tag)
  * Returns next bit offset, or size if nothing found.
  */
 static __always_inline unsigned long
-radix_tree_find_next_bit(const unsigned long *addr,
-			 unsigned long size, unsigned long offset)
+radix_tree_find_next_bit(struct radix_tree_node *node, unsigned int tag,
+			 unsigned long offset)
 {
-	if (!__builtin_constant_p(size))
-		return find_next_bit(addr, size, offset);
+	const unsigned long *addr = node->tags[tag];
 
-	if (offset < size) {
+	if (offset < RADIX_TREE_MAP_SIZE) {
 		unsigned long tmp;
 
 		addr += offset / BITS_PER_LONG;
@@ -205,14 +204,14 @@ radix_tree_find_next_bit(const unsigned long *addr,
 		if (tmp)
 			return __ffs(tmp) + offset;
 		offset = (offset + BITS_PER_LONG) & ~(BITS_PER_LONG - 1);
-		while (offset < size) {
+		while (offset < RADIX_TREE_MAP_SIZE) {
 			tmp = *++addr;
 			if (tmp)
 				return __ffs(tmp) + offset;
 			offset += BITS_PER_LONG;
 		}
 	}
-	return size;
+	return RADIX_TREE_MAP_SIZE;
 }
 
 #ifndef __KERNEL__
@@ -1197,9 +1196,7 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 				return NULL;
 
 			if (flags & RADIX_TREE_ITER_TAGGED)
-				offset = radix_tree_find_next_bit(
-						node->tags[tag],
-						RADIX_TREE_MAP_SIZE,
+				offset = radix_tree_find_next_bit(node, tag,
 						offset + 1);
 			else
 				while (++offset	< RADIX_TREE_MAP_SIZE) {
-- 
2.10.2

[PATCH 17/29] radix-tree: Improve dump output

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

Print the indices of the entries as unsigned (instead of signed) integers
and print the parent node of each entry to help navigate around larger
trees where the layout is not quite so obvious.  Print the indices
covered by a node.  Rearrange the order of fields printed so the indices
and parents line up for each type of entry.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 lib/radix-tree.c | 50 +++++++++++++++++++++++++-------------------------
 1 file changed, 25 insertions(+), 25 deletions(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 2c3fac4..09c5f1d 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -214,15 +214,29 @@ radix_tree_find_next_bit(struct radix_tree_node *node, unsigned int tag,
 	return RADIX_TREE_MAP_SIZE;
 }
 
+/*
+ * The maximum index which can be stored in a radix tree
+ */
+static inline unsigned long shift_maxindex(unsigned int shift)
+{
+	return (RADIX_TREE_MAP_SIZE << shift) - 1;
+}
+
+static inline unsigned long node_maxindex(struct radix_tree_node *node)
+{
+	return shift_maxindex(node->shift);
+}
+
 #ifndef __KERNEL__
 static void dump_node(struct radix_tree_node *node, unsigned long index)
 {
 	unsigned long i;
 
-	pr_debug("radix node: %p offset %d tags %lx %lx %lx shift %d count %d parent %p\n",
-		node, node->offset,
+	pr_debug("radix node: %p offset %d indices %lu-%lu parent %p tags %lx %lx %lx shift %d count %d\n",
+		node, node->offset, index, index | node_maxindex(node),
+		node->parent,
 		node->tags[0][0], node->tags[1][0], node->tags[2][0],
-		node->shift, node->count, node->parent);
+		node->shift, node->count);
 
 	for (i = 0; i < RADIX_TREE_MAP_SIZE; i++) {
 		unsigned long first = index | (i << node->shift);
@@ -231,16 +245,15 @@ static void dump_node(struct radix_tree_node *node, unsigned long index)
 		if (!entry)
 			continue;
 		if (entry == RADIX_TREE_RETRY) {
-			pr_debug("radix retry offset %ld indices %ld-%ld\n",
-					i, first, last);
-		} else if (is_sibling_entry(node, entry)) {
-			pr_debug("radix sblng %p offset %ld val %p indices %ld-%ld\n",
-					entry, i,
-					*(void **)entry_to_node(entry),
-					first, last);
+			pr_debug("radix retry offset %ld indices %lu-%lu parent %p\n",
+					i, first, last, node);
 		} else if (!radix_tree_is_internal_node(entry)) {
-			pr_debug("radix entry %p offset %ld indices %ld-%ld\n",
-					entry, i, first, last);
+			pr_debug("radix entry %p offset %ld indices %lu-%lu parent %p\n",
+					entry, i, first, last, node);
+		} else if (is_sibling_entry(node, entry)) {
+			pr_debug("radix sblng %p offset %ld indices %lu-%lu parent %p val %p\n",
+					entry, i, first, last, node,
+					*(void **)entry_to_node(entry));
 		} else {
 			dump_node(entry_to_node(entry), first);
 		}
@@ -477,19 +490,6 @@ int radix_tree_maybe_preload_order(gfp_t gfp_mask, int order)
 	return __radix_tree_preload(gfp_mask, nr_nodes);
 }
 
-/*
- * The maximum index which can be stored in a radix tree
- */
-static inline unsigned long shift_maxindex(unsigned int shift)
-{
-	return (RADIX_TREE_MAP_SIZE << shift) - 1;
-}
-
-static inline unsigned long node_maxindex(struct radix_tree_node *node)
-{
-	return shift_maxindex(node->shift);
-}
-
 static unsigned radix_tree_load_root(struct radix_tree_root *root,
 		struct radix_tree_node **nodep, unsigned long *maxindex)
 {
-- 
2.10.2

[PATCH 18/29] btrfs: Fix race in btrfs_free_dummy_fs_info()

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

We drop the lock which protects the radix tree, so we must call
radix_tree_iter_next() in order to avoid a modification to the tree
invalidating the iterator state.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 fs/btrfs/tests/btrfs-tests.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c
index bf62ad9..73076a0 100644
--- a/fs/btrfs/tests/btrfs-tests.c
+++ b/fs/btrfs/tests/btrfs-tests.c
@@ -162,6 +162,7 @@ void btrfs_free_dummy_fs_info(struct btrfs_fs_info *fs_info)
 				slot = radix_tree_iter_retry(&iter);
 			continue;
 		}
+		slot = radix_tree_iter_next(&iter);
 		spin_unlock(&fs_info->buffer_lock);
 		free_extent_buffer_stale(eb);
 		spin_lock(&fs_info->buffer_lock);
-- 
2.10.2

[PATCH 19/29] radix tree test suite: iteration test misuses RCU

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

Each thread needs to register itself with RCU, otherwise the reading
thread's read lock has no effect and the freeing thread will free the
memory in the tree without waiting for the read lock to be dropped.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 tools/testing/radix-tree/iteration_check.c | 28 ++++++++++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)

diff --git a/tools/testing/radix-tree/iteration_check.c b/tools/testing/radix-tree/iteration_check.c
index 11d570c..df71cb8 100644
--- a/tools/testing/radix-tree/iteration_check.c
+++ b/tools/testing/radix-tree/iteration_check.c
@@ -29,6 +29,8 @@ static void *add_entries_fn(void *arg)
 {
 	int pgoff;
 
+	rcu_register_thread();
+
 	while (!test_complete) {
 		for (pgoff = 0; pgoff < 100; pgoff++) {
 			pthread_mutex_lock(&tree_lock);
@@ -38,6 +40,8 @@ static void *add_entries_fn(void *arg)
 		}
 	}
 
+	rcu_unregister_thread();
+
 	return NULL;
 }
 
@@ -53,6 +57,8 @@ static void *tagged_iteration_fn(void *arg)
 	struct radix_tree_iter iter;
 	void **slot;
 
+	rcu_register_thread();
+
 	while (!test_complete) {
 		rcu_read_lock();
 		radix_tree_for_each_tagged(slot, &tree, &iter, 0, TAG) {
@@ -72,12 +78,18 @@ static void *tagged_iteration_fn(void *arg)
 				continue;
 			}
 
-			if (rand_r(&seeds[0]) % 50 == 0)
+			if (rand_r(&seeds[0]) % 50 == 0) {
 				slot = radix_tree_iter_next(&iter);
+				rcu_read_unlock();
+				rcu_barrier();
+				rcu_read_lock();
+			}
 		}
 		rcu_read_unlock();
 	}
 
+	rcu_unregister_thread();
+
 	return NULL;
 }
 
@@ -93,6 +105,8 @@ static void *untagged_iteration_fn(void *arg)
 	struct radix_tree_iter iter;
 	void **slot;
 
+	rcu_register_thread();
+
 	while (!test_complete) {
 		rcu_read_lock();
 		radix_tree_for_each_slot(slot, &tree, &iter, 0) {
@@ -112,12 +126,18 @@ static void *untagged_iteration_fn(void *arg)
 				continue;
 			}
 
-			if (rand_r(&seeds[1]) % 50 == 0)
+			if (rand_r(&seeds[1]) % 50 == 0) {
 				slot = radix_tree_iter_next(&iter);
+				rcu_read_unlock();
+				rcu_barrier();
+				rcu_read_lock();
+			}
 		}
 		rcu_read_unlock();
 	}
 
+	rcu_unregister_thread();
+
 	return NULL;
 }
 
@@ -127,6 +147,8 @@ static void *untagged_iteration_fn(void *arg)
  */
 static void *remove_entries_fn(void *arg)
 {
+	rcu_register_thread();
+
 	while (!test_complete) {
 		int pgoff;
 
@@ -137,6 +159,8 @@ static void *remove_entries_fn(void *arg)
 		pthread_mutex_unlock(&tree_lock);
 	}
 
+	rcu_unregister_thread();
+
 	return NULL;
 }
 
-- 
2.10.2

[PATCH 20/29] radix tree: Improve multiorder iterators

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

This fixes several interlinked problems with the iterators in the
presence of multiorder entries.

1. radix_tree_iter_next() would only advance by one slot, which would
result in the iterators returning the same entry more than once if there
were sibling entries.

2. radix_tree_next_slot() could return an internal pointer instead of
a user pointer if a tagged multiorder entry was immediately followed by
an entry of lower order.

3. radix_tree_next_slot() expanded to a lot more code than it used to
when multiorder support was compiled in.  And I wasn't comfortable with
entry_to_node() being in a header file.

Fixing radix_tree_iter_next() for the presence of sibling entries
necessarily involves examining the contents of the radix tree, so we now
need to pass 'slot' to radix_tree_iter_next(), and we need to change the
calling convention so it is called *before* dropping the lock which
protects the tree.  Fortunately, unconverted code won't compile.

radix_tree_next_slot() becomes closer to how it looked before multiorder
support was introduced.  It only checks to see if the next entry in the
chunk is a sibling entry or a pointer to a node; this should be rare
enough that handling this case out of line is not a performance impact
(and such impact is amortised by the fact that the entry we just
processed was a multiorder entry).  Also, radix_tree_next_slot() used
to force a new chunk lookup for untagged entries, which is more
expensive than the out of line sibling entry skipping.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 fs/btrfs/tests/btrfs-tests.c               |  2 +-
 include/linux/radix-tree.h                 | 63 ++++++++------------
 lib/radix-tree.c                           | 94 ++++++++++++++++++++++++++++++
 mm/khugepaged.c                            |  2 +-
 mm/shmem.c                                 |  6 +-
 tools/testing/radix-tree/iteration_check.c |  4 +-
 tools/testing/radix-tree/multiorder.c      | 12 ++++
 tools/testing/radix-tree/regression3.c     |  6 +-
 tools/testing/radix-tree/test.h            |  1 +
 9 files changed, 142 insertions(+), 48 deletions(-)

diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c
index 73076a0..6d3457a 100644
--- a/fs/btrfs/tests/btrfs-tests.c
+++ b/fs/btrfs/tests/btrfs-tests.c
@@ -162,7 +162,7 @@ void btrfs_free_dummy_fs_info(struct btrfs_fs_info *fs_info)
 				slot = radix_tree_iter_retry(&iter);
 			continue;
 		}
-		slot = radix_tree_iter_next(&iter);
+		slot = radix_tree_iter_next(slot, &iter);
 		spin_unlock(&fs_info->buffer_lock);
 		free_extent_buffer_stale(eb);
 		spin_lock(&fs_info->buffer_lock);
diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 66fb8c0..36c6175 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -424,15 +424,10 @@ __radix_tree_iter_add(struct radix_tree_iter *iter, unsigned long slots)
  *
  * If the iterator needs to release then reacquire a lock, the chunk may
  * have been invalidated by an insertion or deletion.  Call this function
- * to continue the iteration from the next index.
+ * before releasing the lock to continue the iteration from the next index.
  */
-static inline __must_check
-void **radix_tree_iter_next(struct radix_tree_iter *iter)
-{
-	iter->next_index = __radix_tree_iter_add(iter, 1);
-	iter->tags = 0;
-	return NULL;
-}
+void **__must_check radix_tree_iter_next(void **slot,
+					struct radix_tree_iter *iter);
 
 /**
  * radix_tree_chunk_size - get current chunk size
@@ -446,10 +441,17 @@ radix_tree_chunk_size(struct radix_tree_iter *iter)
 	return (iter->next_index - iter->index) >> iter_shift(iter);
 }
 
-static inline struct radix_tree_node *entry_to_node(void *ptr)
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
+				unsigned flags);
+#else
+/* Can't happen without sibling entries, but the compiler can't tell that */
+static inline void ** __radix_tree_next_slot(void **slot,
+				struct radix_tree_iter *iter, unsigned flags)
 {
-	return (void *)((unsigned long)ptr & ~RADIX_TREE_INTERNAL_NODE);
+	return slot;
 }
+#endif
 
 /**
  * radix_tree_next_slot - find next slot in chunk
@@ -474,51 +476,31 @@ static __always_inline void **
 radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
 {
 	if (flags & RADIX_TREE_ITER_TAGGED) {
-		void *canon = slot;
-
 		iter->tags >>= 1;
 		if (unlikely(!iter->tags))
 			return NULL;
-		while (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
-					radix_tree_is_internal_node(slot[1])) {
-			if (entry_to_node(slot[1]) == canon) {
-				iter->tags >>= 1;
-				iter->index = __radix_tree_iter_add(iter, 1);
-				slot++;
-				continue;
-			}
-			iter->next_index = __radix_tree_iter_add(iter, 1);
-			return NULL;
-		}
 		if (likely(iter->tags & 1ul)) {
 			iter->index = __radix_tree_iter_add(iter, 1);
-			return slot + 1;
+			slot++;
+			goto found;
 		}
 		if (!(flags & RADIX_TREE_ITER_CONTIG)) {
 			unsigned offset = __ffs(iter->tags);
 
-			iter->tags >>= offset;
-			iter->index = __radix_tree_iter_add(iter, offset + 1);
-			return slot + offset + 1;
+			iter->tags >>= offset++;
+			iter->index = __radix_tree_iter_add(iter, offset);
+			slot += offset;
+			goto found;
 		}
 	} else {
 		long count = radix_tree_chunk_size(iter);
-		void *canon = slot;
 
 		while (--count > 0) {
 			slot++;
 			iter->index = __radix_tree_iter_add(iter, 1);
 
-			if (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
-			    radix_tree_is_internal_node(*slot)) {
-				if (entry_to_node(*slot) == canon)
-					continue;
-				iter->next_index = iter->index;
-				break;
-			}
-
 			if (likely(*slot))
-				return slot;
+				goto found;
 			if (flags & RADIX_TREE_ITER_CONTIG) {
 				/* forbid switching to the next chunk */
 				iter->next_index = 0;
@@ -527,6 +509,11 @@ radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
 		}
 	}
 	return NULL;
+
+ found:
+	if (unlikely(radix_tree_is_internal_node(*slot)))
+		return __radix_tree_next_slot(slot, iter, flags);
+	return slot;
 }
 
 /**
@@ -577,6 +564,6 @@ radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
 	     slot || (slot = radix_tree_next_chunk(root, iter,		\
 			      RADIX_TREE_ITER_TAGGED | tag)) ;		\
 	     slot = radix_tree_next_slot(slot, iter,			\
-				RADIX_TREE_ITER_TAGGED))
+				RADIX_TREE_ITER_TAGGED | tag))
 
 #endif /* _LINUX_RADIX_TREE_H */
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 09c5f1d..27b53ef 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -69,6 +69,11 @@ struct radix_tree_preload {
 };
 static DEFINE_PER_CPU(struct radix_tree_preload, radix_tree_preloads) = { 0, };
 
+static inline struct radix_tree_node *entry_to_node(void *ptr)
+{
+	return (void *)((unsigned long)ptr & ~RADIX_TREE_INTERNAL_NODE);
+}
+
 static inline void *node_to_entry(void *ptr)
 {
 	return (void *)((unsigned long)ptr | RADIX_TREE_INTERNAL_NODE);
@@ -1138,6 +1143,95 @@ static inline void __set_iter_shift(struct radix_tree_iter *iter,
 #endif
 }
 
+static void ** __radix_tree_iter_next(struct radix_tree_node **nodep,
+			void **slot, struct radix_tree_iter *iter)
+{
+	void *sib = node_to_entry(slot - 1);
+
+	while (iter->index < iter->next_index) {
+		*nodep = rcu_dereference_raw(*slot);
+		if (*nodep && *nodep != sib)
+			return slot;
+		slot++;
+		iter->index = __radix_tree_iter_add(iter, 1);
+		iter->tags >>= 1;
+	}
+
+	*nodep = NULL;
+	return NULL;
+}
+
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
+					unsigned flags)
+{
+	unsigned tag = flags & RADIX_TREE_ITER_TAG_MASK;
+	struct radix_tree_node *node = rcu_dereference_raw(*slot);
+
+	slot = __radix_tree_iter_next(&node, slot, iter);
+
+	while (radix_tree_is_internal_node(node)) {
+		unsigned offset;
+
+		if (node == RADIX_TREE_RETRY)
+			return slot;
+		node = entry_to_node(node);
+
+		if (flags & RADIX_TREE_ITER_TAGGED) {
+			unsigned tag_long, tag_bit;
+			offset = radix_tree_find_next_bit(node, tag, 0);
+			if (offset == RADIX_TREE_MAP_SIZE)
+				return NULL;
+			slot = &node->slots[offset];
+
+			tag_long = offset / BITS_PER_LONG;
+			tag_bit  = offset % BITS_PER_LONG;
+			iter->tags = node->tags[tag][tag_long] >> tag_bit;
+			BUG_ON(iter->tags >= (RADIX_TREE_MAP_SIZE * 2));
+			node = rcu_dereference_raw(*slot);
+		} else {
+			offset = 0;
+			slot = &node->slots[0];
+			for (;;) {
+				node = rcu_dereference_raw(*slot);
+				if (node)
+					break;
+				slot++;
+				offset++;
+				if (offset == RADIX_TREE_MAP_SIZE)
+					return NULL;
+			}
+		}
+		if ((flags & RADIX_TREE_ITER_CONTIG) && (offset > 0))
+			goto none;
+		iter->shift -= RADIX_TREE_MAP_SHIFT;
+		iter->index = __radix_tree_iter_add(iter, offset);
+		iter->next_index = (iter->index | shift_maxindex(iter->shift)) +
+					1;
+	}
+
+	return slot;
+ none:
+	iter->next_index = 0;
+	return NULL;
+}
+EXPORT_SYMBOL(__radix_tree_next_slot);
+#endif
+
+void **radix_tree_iter_next(void **slot, struct radix_tree_iter *iter)
+{
+	struct radix_tree_node *node;
+
+	slot++;
+	iter->index = __radix_tree_iter_add(iter, 1);
+	node = rcu_dereference_raw(*slot);
+	__radix_tree_iter_next(&node, slot, iter);
+	iter->next_index = iter->index;
+	iter->tags = 0;
+	return NULL;
+}
+EXPORT_SYMBOL(radix_tree_iter_next);
+
 /**
  * radix_tree_next_chunk - find next chunk of slots for iteration
  *
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 728d779..46155d1 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -1614,8 +1614,8 @@ static void khugepaged_scan_shmem(struct mm_struct *mm,
 		present++;
 
 		if (need_resched()) {
+			slot = radix_tree_iter_next(slot, &iter);
 			cond_resched_rcu();
-			slot = radix_tree_iter_next(&iter);
 		}
 	}
 	rcu_read_unlock();
diff --git a/mm/shmem.c b/mm/shmem.c
index 166ebf5..0b3fe33 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -658,8 +658,8 @@ unsigned long shmem_partial_swap_usage(struct address_space *mapping,
 			swapped++;
 
 		if (need_resched()) {
+			slot = radix_tree_iter_next(slot, &iter);
 			cond_resched_rcu();
-			slot = radix_tree_iter_next(&iter);
 		}
 	}
 
@@ -2434,8 +2434,8 @@ static void shmem_tag_pins(struct address_space *mapping)
 		}
 
 		if (need_resched()) {
+			slot = radix_tree_iter_next(slot, &iter);
 			cond_resched_rcu();
-			slot = radix_tree_iter_next(&iter);
 		}
 	}
 	rcu_read_unlock();
@@ -2504,8 +2504,8 @@ static int shmem_wait_for_pins(struct address_space *mapping)
 			spin_unlock_irq(&mapping->tree_lock);
 continue_resched:
 			if (need_resched()) {
+				slot = radix_tree_iter_next(slot, &iter);
 				cond_resched_rcu();
-				slot = radix_tree_iter_next(&iter);
 			}
 		}
 		rcu_read_unlock();
diff --git a/tools/testing/radix-tree/iteration_check.c b/tools/testing/radix-tree/iteration_check.c
index df71cb8..6eca4c2 100644
--- a/tools/testing/radix-tree/iteration_check.c
+++ b/tools/testing/radix-tree/iteration_check.c
@@ -79,7 +79,7 @@ static void *tagged_iteration_fn(void *arg)
 			}
 
 			if (rand_r(&seeds[0]) % 50 == 0) {
-				slot = radix_tree_iter_next(&iter);
+				slot = radix_tree_iter_next(slot, &iter);
 				rcu_read_unlock();
 				rcu_barrier();
 				rcu_read_lock();
@@ -127,7 +127,7 @@ static void *untagged_iteration_fn(void *arg)
 			}
 
 			if (rand_r(&seeds[1]) % 50 == 0) {
-				slot = radix_tree_iter_next(&iter);
+				slot = radix_tree_iter_next(slot, &iter);
 				rcu_read_unlock();
 				rcu_barrier();
 				rcu_read_lock();
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 25e0463..588209a 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -232,10 +232,14 @@ void multiorder_iteration(void)
 			int height = order[i] / RADIX_TREE_MAP_SHIFT;
 			int shift = height * RADIX_TREE_MAP_SHIFT;
 			int mask = (1 << order[i]) - 1;
+			struct item *item = *slot;
 
 			assert(iter.index >= (index[i] &~ mask));
 			assert(iter.index <= (index[i] | mask));
 			assert(iter.shift == shift);
+			assert(!radix_tree_is_internal_node(item));
+			assert(item->index >= (index[i] &~ mask));
+			assert(item->index <= (index[i] | mask));
 			i++;
 		}
 	}
@@ -279,12 +283,16 @@ void multiorder_tagged_iteration(void)
 		}
 
 		radix_tree_for_each_tagged(slot, &tree, &iter, j, 1) {
+			struct item *item = *slot;
 			for (k = i; index[k] < tag_index[i]; k++)
 				;
 			mask = (1 << order[k]) - 1;
 
 			assert(iter.index >= (tag_index[i] &~ mask));
 			assert(iter.index <= (tag_index[i] | mask));
+			assert(!radix_tree_is_internal_node(item));
+			assert(item->index >= (tag_index[i] &~ mask));
+			assert(item->index <= (tag_index[i] | mask));
 			i++;
 		}
 	}
@@ -303,12 +311,16 @@ void multiorder_tagged_iteration(void)
 		}
 
 		radix_tree_for_each_tagged(slot, &tree, &iter, j, 2) {
+			struct item *item = *slot;
 			for (k = i; index[k] < tag_index[i]; k++)
 				;
 			mask = (1 << order[k]) - 1;
 
 			assert(iter.index >= (tag_index[i] &~ mask));
 			assert(iter.index <= (tag_index[i] | mask));
+			assert(!radix_tree_is_internal_node(item));
+			assert(item->index >= (tag_index[i] &~ mask));
+			assert(item->index <= (tag_index[i] | mask));
 			i++;
 		}
 	}
diff --git a/tools/testing/radix-tree/regression3.c b/tools/testing/radix-tree/regression3.c
index 1f06ed7..4d28eeb 100644
--- a/tools/testing/radix-tree/regression3.c
+++ b/tools/testing/radix-tree/regression3.c
@@ -88,7 +88,7 @@ void regression3_test(void)
 		printf("slot %ld %p\n", iter.index, *slot);
 		if (!iter.index) {
 			printf("next at %ld\n", iter.index);
-			slot = radix_tree_iter_next(&iter);
+			slot = radix_tree_iter_next(slot, &iter);
 		}
 	}
 
@@ -96,7 +96,7 @@ void regression3_test(void)
 		printf("contig %ld %p\n", iter.index, *slot);
 		if (!iter.index) {
 			printf("next at %ld\n", iter.index);
-			slot = radix_tree_iter_next(&iter);
+			slot = radix_tree_iter_next(slot, &iter);
 		}
 	}
 
@@ -106,7 +106,7 @@ void regression3_test(void)
 		printf("tagged %ld %p\n", iter.index, *slot);
 		if (!iter.index) {
 			printf("next at %ld\n", iter.index);
-			slot = radix_tree_iter_next(&iter);
+			slot = radix_tree_iter_next(slot, &iter);
 		}
 	}
 
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index 33d2b6b..b678f13 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -43,6 +43,7 @@ void verify_tag_consistency(struct radix_tree_root *root, unsigned int tag);
 extern int nr_allocated;
 
 /* Normally private parts of lib/radix-tree.c */
+struct radix_tree_node *entry_to_node(void *ptr);
 void radix_tree_dump(struct radix_tree_root *root);
 int root_tag_get(struct radix_tree_root *root, unsigned int tag);
 unsigned long node_maxindex(struct radix_tree_node *);
-- 
2.10.2

[PATCH 21/29] radix-tree: Delete radix_tree_locate_item()

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

This rather complicated function can be better implemented as an iterator.
It has only one caller, so move the functionality to the only place that
needs it.  Update the test suite to follow the same pattern.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 include/linux/radix-tree.h            |  1 -
 lib/radix-tree.c                      | 99 -----------------------------------
 mm/shmem.c                            | 26 ++++++++-
 tools/testing/radix-tree/main.c       |  8 +--
 tools/testing/radix-tree/multiorder.c |  2 +-
 tools/testing/radix-tree/test.c       | 22 ++++++++
 tools/testing/radix-tree/test.h       |  2 +
 7 files changed, 54 insertions(+), 106 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 36c6175..57bf635 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -306,7 +306,6 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
 		unsigned long nr_to_tag,
 		unsigned int fromtag, unsigned int totag);
 int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
-unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item);
 
 static inline void radix_tree_preload_end(void)
 {
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 27b53ef..7e70ac9 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1605,105 +1605,6 @@ radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
 }
 EXPORT_SYMBOL(radix_tree_gang_lookup_tag_slot);
 
-#if defined(CONFIG_SHMEM) && defined(CONFIG_SWAP)
-#include <linux/sched.h> /* for cond_resched() */
-
-struct locate_info {
-	unsigned long found_index;
-	bool stop;
-};
-
-/*
- * This linear search is at present only useful to shmem_unuse_inode().
- */
-static unsigned long __locate(struct radix_tree_node *slot, void *item,
-			      unsigned long index, struct locate_info *info)
-{
-	unsigned long i;
-
-	do {
-		unsigned int shift = slot->shift;
-
-		for (i = (index >> shift) & RADIX_TREE_MAP_MASK;
-		     i < RADIX_TREE_MAP_SIZE;
-		     i++, index += (1UL << shift)) {
-			struct radix_tree_node *node =
-					rcu_dereference_raw(slot->slots[i]);
-			if (node == RADIX_TREE_RETRY)
-				goto out;
-			if (!radix_tree_is_internal_node(node)) {
-				if (node == item) {
-					info->found_index = index;
-					info->stop = true;
-					goto out;
-				}
-				continue;
-			}
-			node = entry_to_node(node);
-			if (is_sibling_entry(slot, node))
-				continue;
-			slot = node;
-			break;
-		}
-	} while (i < RADIX_TREE_MAP_SIZE);
-
-out:
-	if ((index == 0) && (i == RADIX_TREE_MAP_SIZE))
-		info->stop = true;
-	return index;
-}
-
-/**
- *	radix_tree_locate_item - search through radix tree for item
- *	@root:		radix tree root
- *	@item:		item to be found
- *
- *	Returns index where item was found, or -1 if not found.
- *	Caller must hold no lock (since this time-consuming function needs
- *	to be preemptible), and must check afterwards if item is still there.
- */
-unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item)
-{
-	struct radix_tree_node *node;
-	unsigned long max_index;
-	unsigned long cur_index = 0;
-	struct locate_info info = {
-		.found_index = -1,
-		.stop = false,
-	};
-
-	do {
-		rcu_read_lock();
-		node = rcu_dereference_raw(root->rnode);
-		if (!radix_tree_is_internal_node(node)) {
-			rcu_read_unlock();
-			if (node == item)
-				info.found_index = 0;
-			break;
-		}
-
-		node = entry_to_node(node);
-
-		max_index = node_maxindex(node);
-		if (cur_index > max_index) {
-			rcu_read_unlock();
-			break;
-		}
-
-		cur_index = __locate(node, item, cur_index, &info);
-		rcu_read_unlock();
-		cond_resched();
-	} while (!info.stop && cur_index <= max_index);
-
-	return info.found_index;
-}
-#else
-unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item)
-{
-	return -1;
-}
-#endif /* CONFIG_SHMEM && CONFIG_SWAP */
-
 /**
  *	radix_tree_shrink    -    shrink radix tree to minimum height
  *	@root		radix tree root
diff --git a/mm/shmem.c b/mm/shmem.c
index 0b3fe33..8f9c9aa 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -1046,6 +1046,30 @@ static void shmem_evict_inode(struct inode *inode)
 	clear_inode(inode);
 }
 
+static unsigned long find_swap_entry(struct radix_tree_root *root, void *item)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+	unsigned long found = -1;
+	unsigned int checked = 0;
+
+	rcu_read_lock();
+	radix_tree_for_each_slot(slot, root, &iter, 0) {
+		if (*slot == item) {
+			found = iter.index;
+			break;
+		}
+		checked++;
+		if ((checked % 4096) != 0)
+			continue;
+		slot = radix_tree_iter_next(slot, &iter);
+		cond_resched_rcu();
+	}
+
+	rcu_read_unlock();
+	return found;
+}
+
 /*
  * If swap found in inode, free it and move page from swapcache to filecache.
  */
@@ -1059,7 +1083,7 @@ static int shmem_unuse_inode(struct shmem_inode_info *info,
 	int error = 0;
 
 	radswap = swp_to_radix_entry(swap);
-	index = radix_tree_locate_item(&mapping->page_tree, radswap);
+	index = find_swap_entry(&mapping->page_tree, radswap);
 	if (index == -1)
 		return -EAGAIN;	/* tell shmem_unuse we found nothing */
 
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 8621542..93a77f9 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -239,7 +239,7 @@ static void __locate_check(struct radix_tree_root *tree, unsigned long index,
 
 	item_insert_order(tree, index, order);
 	item = item_lookup(tree, index);
-	index2 = radix_tree_locate_item(tree, item);
+	index2 = find_item(tree, item);
 	if (index != index2) {
 		printf("index %ld order %d inserted; found %ld\n",
 			index, order, index2);
@@ -273,17 +273,17 @@ static void locate_check(void)
 			     index += (1UL << order)) {
 				__locate_check(&tree, index + offset, order);
 			}
-			if (radix_tree_locate_item(&tree, &tree) != -1)
+			if (find_item(&tree, &tree) != -1)
 				abort();
 
 			item_kill_tree(&tree);
 		}
 	}
 
-	if (radix_tree_locate_item(&tree, &tree) != -1)
+	if (find_item(&tree, &tree) != -1)
 		abort();
 	__locate_check(&tree, -1, 0);
-	if (radix_tree_locate_item(&tree, &tree) != -1)
+	if (find_item(&tree, &tree) != -1)
 		abort();
 	item_kill_tree(&tree);
 }
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 588209a..400de5c 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -347,7 +347,7 @@ static void __multiorder_join(unsigned long index,
 	item_insert_order(&tree, index, order2);
 	item = radix_tree_lookup(&tree, index);
 	radix_tree_join(&tree, index + 1, order1, item2);
-	loc = radix_tree_locate_item(&tree, item);
+	loc = find_item(&tree, item);
 	if (loc == -1)
 		free(item);
 	item = radix_tree_lookup(&tree, index + 1);
diff --git a/tools/testing/radix-tree/test.c b/tools/testing/radix-tree/test.c
index a6e8099..a68ed3b 100644
--- a/tools/testing/radix-tree/test.c
+++ b/tools/testing/radix-tree/test.c
@@ -142,6 +142,28 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
 	assert(nfound == 0);
 }
 
+/* Use the same pattern as find_swap_entry() in mm/shmem.c */
+unsigned long find_item(struct radix_tree_root *root, void *item)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+	unsigned long found = -1;
+	unsigned long checked = 0;
+
+	radix_tree_for_each_slot(slot, root, &iter, 0) {
+		if (*slot == item) {
+			found = iter.index;
+			break;
+		}
+		checked++;
+		if ((checked % 4) != 0)
+			continue;
+		slot = radix_tree_iter_next(slot, &iter);
+	}
+
+	return found;
+}
+
 static int verify_node(struct radix_tree_node *slot, unsigned int tag,
 			int tagged)
 {
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index b678f13..ccdd3c1 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -27,6 +27,8 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
 			unsigned long nr, int chunk);
 void item_kill_tree(struct radix_tree_root *root);
 
+unsigned long find_item(struct radix_tree_root *, void *item);
+
 void tag_check(void);
 void multiorder_checks(void);
 void iteration_test(void);
-- 
2.10.2

[PATCH 22/29] radix-tree: Delete radix_tree_range_tag_if_tagged()

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

This is an exceptionally complicated function with just one caller
(tag_pages_for_writeback).  We devote a large portion of the runtime
of the test suite to testing this one function which has one caller.
By introducing the new function radix_tree_iter_tag_set(), we
can eliminate all of the complexity while keeping the performance.
The caller can now use a fairly standard radix_tree_for_each() loop,
and it doesn't need to worry about tricksy things like 'start' wrapping.

The test suite continues to spend a large amount of time investigating
this function, but now it's testing the underlying primitives such as
radix_tree_iter_next() and the radix_tree_for_each_tagged() iterator
which are also used by other parts of the kernel.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 include/linux/radix-tree.h             |  74 ++++++++++-----------
 lib/radix-tree.c                       | 114 +++++----------------------------
 mm/page-writeback.c                    |  28 +++++---
 tools/testing/radix-tree/main.c        |  12 ++--
 tools/testing/radix-tree/multiorder.c  |  13 ++--
 tools/testing/radix-tree/regression2.c |   3 +-
 tools/testing/radix-tree/tag_check.c   |   4 +-
 tools/testing/radix-tree/test.c        |  25 ++++++++
 tools/testing/radix-tree/test.h        |   3 +
 9 files changed, 113 insertions(+), 163 deletions(-)

diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 57bf635..ca4eea1 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -124,6 +124,41 @@ static inline bool radix_tree_empty(struct radix_tree_root *root)
 }
 
 /**
+ * struct radix_tree_iter - radix tree iterator state
+ *
+ * @index:	index of current slot
+ * @next_index:	one beyond the last index for this chunk
+ * @tags:	bit-mask for tag-iterating
+ * @node:	node that contains current slot
+ * @shift:	shift for the node that holds our slots
+ *
+ * This radix tree iterator works in terms of "chunks" of slots.  A chunk is a
+ * subinterval of slots contained within one radix tree leaf node.  It is
+ * described by a pointer to its first slot and a struct radix_tree_iter
+ * which holds the chunk's position in the tree and its size.  For tagged
+ * iteration radix_tree_iter also holds the slots' bit-mask for one chosen
+ * radix tree tag.
+ */
+struct radix_tree_iter {
+	unsigned long	index;
+	unsigned long	next_index;
+	unsigned long	tags;
+	struct radix_tree_node *node;
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+	unsigned int	shift;
+#endif
+};
+
+static inline unsigned int iter_shift(struct radix_tree_iter *iter)
+{
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
+	return iter->shift;
+#else
+	return 0;
+#endif
+}
+
+/**
  * Radix-tree synchronization
  *
  * The radix-tree API requires that users provide all synchronisation (with
@@ -293,6 +328,8 @@ void *radix_tree_tag_clear(struct radix_tree_root *root,
 			unsigned long index, unsigned int tag);
 int radix_tree_tag_get(struct radix_tree_root *root,
 			unsigned long index, unsigned int tag);
+void radix_tree_iter_tag_set(struct radix_tree_root *root,
+		const struct radix_tree_iter *iter, unsigned int tag);
 unsigned int
 radix_tree_gang_lookup_tag(struct radix_tree_root *root, void **results,
 		unsigned long first_index, unsigned int max_items,
@@ -301,10 +338,6 @@ unsigned int
 radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
 		unsigned long first_index, unsigned int max_items,
 		unsigned int tag);
-unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
-		unsigned long *first_indexp, unsigned long last_index,
-		unsigned long nr_to_tag,
-		unsigned int fromtag, unsigned int totag);
 int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
 
 static inline void radix_tree_preload_end(void)
@@ -318,39 +351,6 @@ int radix_tree_split(struct radix_tree_root *, unsigned long index,
 int radix_tree_join(struct radix_tree_root *, unsigned long index,
 			unsigned new_order, void *);
 
-/**
- * struct radix_tree_iter - radix tree iterator state
- *
- * @index:	index of current slot
- * @next_index:	one beyond the last index for this chunk
- * @tags:	bit-mask for tag-iterating
- * @shift:	shift for the node that holds our slots
- *
- * This radix tree iterator works in terms of "chunks" of slots.  A chunk is a
- * subinterval of slots contained within one radix tree leaf node.  It is
- * described by a pointer to its first slot and a struct radix_tree_iter
- * which holds the chunk's position in the tree and its size.  For tagged
- * iteration radix_tree_iter also holds the slots' bit-mask for one chosen
- * radix tree tag.
- */
-struct radix_tree_iter {
-	unsigned long	index;
-	unsigned long	next_index;
-	unsigned long	tags;
-#ifdef CONFIG_RADIX_TREE_MULTIORDER
-	unsigned int	shift;
-#endif
-};
-
-static inline unsigned int iter_shift(struct radix_tree_iter *iter)
-{
-#ifdef CONFIG_RADIX_TREE_MULTIORDER
-	return iter->shift;
-#else
-	return 0;
-#endif
-}
-
 #define RADIX_TREE_ITER_TAG_MASK	0x00FF	/* tag index in lower byte */
 #define RADIX_TREE_ITER_TAGGED		0x0100	/* lookup tagged slots */
 #define RADIX_TREE_ITER_CONTIG		0x0200	/* stop at first hole */
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 7e70ac9..c8ef657 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1049,6 +1049,20 @@ static void node_tag_set(struct radix_tree_root *root,
 }
 
 /**
+ * radix_tree_iter_tag_set - set a tag on the current iterator entry
+ *
+ * @root:	radix tree root
+ * @iter:	iterator state
+ * @tag:	tag to set
+ */
+void radix_tree_iter_tag_set(struct radix_tree_root *root,
+			const struct radix_tree_iter *iter, unsigned int tag)
+{
+	unsigned offset = (iter->index >> iter->shift) & RADIX_TREE_MAP_MASK;
+	node_tag_set(root, iter->node, tag, offset);
+}
+
+/**
  *	radix_tree_tag_clear - clear a tag on a radix tree node
  *	@root:		radix tree root
  *	@index:		index key
@@ -1176,6 +1190,7 @@ void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
 		if (node == RADIX_TREE_RETRY)
 			return slot;
 		node = entry_to_node(node);
+		iter->node = node;
 
 		if (flags & RADIX_TREE_ITER_TAGGED) {
 			unsigned tag_long, tag_bit;
@@ -1275,6 +1290,7 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 		iter->index = index;
 		iter->next_index = maxindex + 1;
 		iter->tags = 1;
+		iter->node = NULL;
 		__set_iter_shift(iter, 0);
 		return (void **)&root->rnode;
 	}
@@ -1319,6 +1335,7 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 	/* Update the iterator state */
 	iter->index = (index &~ node_maxindex(node)) | (offset << node->shift);
 	iter->next_index = (index | node_maxindex(node)) + 1;
+	iter->node = node;
 	__set_iter_shift(iter, node->shift);
 
 	/* Construct iter->tags bit-mask from node->tags[tag] array */
@@ -1344,103 +1361,6 @@ void **radix_tree_next_chunk(struct radix_tree_root *root,
 EXPORT_SYMBOL(radix_tree_next_chunk);
 
 /**
- * radix_tree_range_tag_if_tagged - for each item in given range set given
- *				   tag if item has another tag set
- * @root:		radix tree root
- * @first_indexp:	pointer to a starting index of a range to scan
- * @last_index:		last index of a range to scan
- * @nr_to_tag:		maximum number items to tag
- * @iftag:		tag index to test
- * @settag:		tag index to set if tested tag is set
- *
- * This function scans range of radix tree from first_index to last_index
- * (inclusive).  For each item in the range if iftag is set, the function sets
- * also settag. The function stops either after tagging nr_to_tag items or
- * after reaching last_index.
- *
- * The tags must be set from the leaf level only and propagated back up the
- * path to the root. We must do this so that we resolve the full path before
- * setting any tags on intermediate nodes. If we set tags as we descend, then
- * we can get to the leaf node and find that the index that has the iftag
- * set is outside the range we are scanning. This reults in dangling tags and
- * can lead to problems with later tag operations (e.g. livelocks on lookups).
- *
- * The function returns the number of leaves where the tag was set and sets
- * *first_indexp to the first unscanned index.
- * WARNING! *first_indexp can wrap if last_index is ULONG_MAX. Caller must
- * be prepared to handle that.
- */
-unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
-		unsigned long *first_indexp, unsigned long last_index,
-		unsigned long nr_to_tag,
-		unsigned int iftag, unsigned int settag)
-{
-	struct radix_tree_node *node, *child;
-	unsigned long maxindex;
-	unsigned long tagged = 0;
-	unsigned long index = *first_indexp;
-
-	radix_tree_load_root(root, &child, &maxindex);
-	last_index = min(last_index, maxindex);
-	if (index > last_index)
-		return 0;
-	if (!nr_to_tag)
-		return 0;
-	if (!root_tag_get(root, iftag)) {
-		*first_indexp = last_index + 1;
-		return 0;
-	}
-	if (!radix_tree_is_internal_node(child)) {
-		*first_indexp = last_index + 1;
-		root_tag_set(root, settag);
-		return 1;
-	}
-
-	node = entry_to_node(child);
-
-	for (;;) {
-		unsigned offset = radix_tree_descend(node, &child, index);
-		if (!child)
-			goto next;
-		if (!tag_get(node, iftag, offset))
-			goto next;
-		/* Sibling slots never have tags set on them */
-		if (radix_tree_is_internal_node(child)) {
-			node = entry_to_node(child);
-			continue;
-		}
-
-		tagged++;
-		node_tag_set(root, node, settag, offset);
- next:
-		/* Go to next entry in node */
-		index = ((index >> node->shift) + 1) << node->shift;
-		/* Overflow can happen when last_index is ~0UL... */
-		if (index > last_index || !index)
-			break;
-		offset = (index >> node->shift) & RADIX_TREE_MAP_MASK;
-		while (offset == 0) {
-			/*
-			 * We've fully scanned this node. Go up. Because
-			 * last_index is guaranteed to be in the tree, what
-			 * we do below cannot wander astray.
-			 */
-			node = node->parent;
-			offset = (index >> node->shift) & RADIX_TREE_MAP_MASK;
-		}
-		if (is_sibling_entry(node, node->slots[offset]))
-			goto next;
-		if (tagged >= nr_to_tag)
-			break;
-	}
-
-	*first_indexp = index;
-
-	return tagged;
-}
-EXPORT_SYMBOL(radix_tree_range_tag_if_tagged);
-
-/**
  *	radix_tree_gang_lookup - perform multiple lookup on a radix tree
  *	@root:		radix tree root
  *	@results:	where the results of the lookup are placed
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 439cc63..c593715 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -2105,18 +2105,26 @@ void tag_pages_for_writeback(struct address_space *mapping,
 			     pgoff_t start, pgoff_t end)
 {
 #define WRITEBACK_TAG_BATCH 4096
-	unsigned long tagged;
-
-	do {
-		spin_lock_irq(&mapping->tree_lock);
-		tagged = radix_tree_range_tag_if_tagged(&mapping->page_tree,
-				&start, end, WRITEBACK_TAG_BATCH,
-				PAGECACHE_TAG_DIRTY, PAGECACHE_TAG_TOWRITE);
+	unsigned long tagged = 0;
+	struct radix_tree_iter iter;
+	void **slot;
+
+	spin_lock_irq(&mapping->tree_lock);
+	radix_tree_for_each_tagged(slot, &mapping->page_tree, &iter, start,
+							PAGECACHE_TAG_DIRTY) {
+		if (iter.index > end)
+			break;
+		radix_tree_iter_tag_set(&mapping->page_tree, &iter,
+							PAGECACHE_TAG_TOWRITE);
+		tagged++;
+		if ((tagged % WRITEBACK_TAG_BATCH) != 0)
+			continue;
+		slot = radix_tree_iter_next(slot, &iter);
 		spin_unlock_irq(&mapping->tree_lock);
-		WARN_ON_ONCE(tagged > WRITEBACK_TAG_BATCH);
 		cond_resched();
-		/* We check 'start' to handle wrapping when end == ~0UL */
-	} while (tagged >= WRITEBACK_TAG_BATCH && start);
+		spin_lock_irq(&mapping->tree_lock);
+	}
+	spin_unlock_irq(&mapping->tree_lock);
 }
 EXPORT_SYMBOL(tag_pages_for_writeback);
 
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 93a77f9..7d14812 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -205,8 +205,7 @@ void copy_tag_check(void)
 	}
 
 //	printf("\ncopying tags...\n");
-	cur = start;
-	tagged = radix_tree_range_tag_if_tagged(&tree, &cur, end, ITEMS, 0, 1);
+	tagged = tag_tagged_items(&tree, start, end, ITEMS, 0, 1);
 
 //	printf("checking copied tags\n");
 	assert(tagged == count);
@@ -214,16 +213,13 @@ void copy_tag_check(void)
 
 	/* Copy tags in several rounds */
 //	printf("\ncopying tags...\n");
-	cur = start;
-	do {
-		tmp = rand() % (count/10+2);
-		tagged = radix_tree_range_tag_if_tagged(&tree, &cur, end, tmp, 0, 2);
-	} while (tmp == tagged);
+	tmp = rand() % (count / 10 + 2);
+	tagged = tag_tagged_items(&tree, start, end, tmp, 0, 2);
+	assert(tagged == count);
 
 //	printf("%lu %lu %lu\n", tagged, tmp, count);
 //	printf("checking copied tags\n");
 	check_copied_tags(&tree, start, end, idx, ITEMS, 0, 2);
-	assert(tagged < tmp);
 	verify_tag_consistency(&tree, 0);
 	verify_tag_consistency(&tree, 1);
 	verify_tag_consistency(&tree, 2);
diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
index 400de5c..6a0287d 100644
--- a/tools/testing/radix-tree/multiorder.c
+++ b/tools/testing/radix-tree/multiorder.c
@@ -26,7 +26,6 @@ static void __multiorder_tag_test(int index, int order)
 {
 	RADIX_TREE(tree, GFP_KERNEL);
 	int base, err, i;
-	unsigned long first = 0;
 
 	/* our canonical entry */
 	base = index & ~((1 << order) - 1);
@@ -60,7 +59,7 @@ static void __multiorder_tag_test(int index, int order)
 		assert(!radix_tree_tag_get(&tree, i, 1));
 	}
 
-	assert(radix_tree_range_tag_if_tagged(&tree, &first, ~0UL, 10, 0, 1) == 1);
+	assert(tag_tagged_items(&tree, 0, ~0UL, 10, 0, 1) == 1);
 	assert(radix_tree_tag_clear(&tree, index, 0));
 
 	for_each_index(i, base, order) {
@@ -252,7 +251,6 @@ void multiorder_tagged_iteration(void)
 	RADIX_TREE(tree, GFP_KERNEL);
 	struct radix_tree_iter iter;
 	void **slot;
-	unsigned long first = 0;
 	int i, j;
 
 	printf("Multiorder tagged iteration test\n");
@@ -297,8 +295,8 @@ void multiorder_tagged_iteration(void)
 		}
 	}
 
-	radix_tree_range_tag_if_tagged(&tree, &first, ~0UL,
-					MT_NUM_ENTRIES, 1, 2);
+	assert(tag_tagged_items(&tree, 0, ~0UL, TAG_ENTRIES, 1, 2) ==
+				TAG_ENTRIES);
 
 	for (j = 0; j < 256; j++) {
 		int mask, k;
@@ -325,9 +323,8 @@ void multiorder_tagged_iteration(void)
 		}
 	}
 
-	first = 1;
-	radix_tree_range_tag_if_tagged(&tree, &first, ~0UL,
-					MT_NUM_ENTRIES, 1, 0);
+	assert(tag_tagged_items(&tree, 1, ~0UL, MT_NUM_ENTRIES * 2, 1, 0)
+			== TAG_ENTRIES);
 	i = 0;
 	radix_tree_for_each_tagged(slot, &tree, &iter, 0, 0) {
 		assert(iter.index == tag_index[i]);
diff --git a/tools/testing/radix-tree/regression2.c b/tools/testing/radix-tree/regression2.c
index 63bf347..5cd2c0f 100644
--- a/tools/testing/radix-tree/regression2.c
+++ b/tools/testing/radix-tree/regression2.c
@@ -50,6 +50,7 @@
 #include <stdio.h>
 
 #include "regression.h"
+#include "test.h"
 
 #define PAGECACHE_TAG_DIRTY     0
 #define PAGECACHE_TAG_WRITEBACK 1
@@ -90,7 +91,7 @@ void regression2_test(void)
 	/* 1. */
 	start = 0;
 	end = max_slots - 2;
-	radix_tree_range_tag_if_tagged(&mt_tree, &start, end, 1,
+	tag_tagged_items(&mt_tree, start, end, 1,
 				PAGECACHE_TAG_DIRTY, PAGECACHE_TAG_TOWRITE);
 
 	/* 2. */
diff --git a/tools/testing/radix-tree/tag_check.c b/tools/testing/radix-tree/tag_check.c
index 186f6e4..b047b2d 100644
--- a/tools/testing/radix-tree/tag_check.c
+++ b/tools/testing/radix-tree/tag_check.c
@@ -23,7 +23,7 @@ __simple_checks(struct radix_tree_root *tree, unsigned long index, int tag)
 	item_tag_set(tree, index, tag);
 	ret = item_tag_get(tree, index, tag);
 	assert(ret != 0);
-	ret = radix_tree_range_tag_if_tagged(tree, &first, ~0UL, 10, tag, !tag);
+	ret = tag_tagged_items(tree, first, ~0UL, 10, tag, !tag);
 	assert(ret == 1);
 	ret = item_tag_get(tree, index, !tag);
 	assert(ret != 0);
@@ -320,7 +320,7 @@ static void single_check(void)
 	assert(ret == 0);
 	verify_tag_consistency(&tree, 0);
 	verify_tag_consistency(&tree, 1);
-	ret = radix_tree_range_tag_if_tagged(&tree, &first, 10, 10, 0, 1);
+	ret = tag_tagged_items(&tree, first, 10, 10, 0, 1);
 	assert(ret == 1);
 	ret = radix_tree_gang_lookup_tag(&tree, (void **)items, 0, BATCH, 1);
 	assert(ret == 1);
diff --git a/tools/testing/radix-tree/test.c b/tools/testing/radix-tree/test.c
index a68ed3b..322c34b 100644
--- a/tools/testing/radix-tree/test.c
+++ b/tools/testing/radix-tree/test.c
@@ -142,6 +142,31 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
 	assert(nfound == 0);
 }
 
+/* Use the same pattern as tag_pages_for_writeback() in mm/page-writeback.c */
+int tag_tagged_items(struct radix_tree_root *root,
+			unsigned long start, unsigned long end, unsigned batch,
+			unsigned iftag, unsigned thentag)
+{
+	unsigned long tagged = 0;
+	struct radix_tree_iter iter;
+	void **slot;
+
+	if (batch == 0)
+		batch = 1;
+
+	radix_tree_for_each_tagged(slot, root, &iter, start, iftag) {
+		if (iter.index > end)
+			break;
+		radix_tree_iter_tag_set(root, &iter, thentag);
+		tagged++;
+		if ((tagged % batch) != 0)
+			continue;
+		slot = radix_tree_iter_next(slot, &iter);
+	}
+
+	return tagged;
+}
+
 /* Use the same pattern as find_swap_entry() in mm/shmem.c */
 unsigned long find_item(struct radix_tree_root *root, void *item)
 {
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index ccdd3c1..1faf0a3 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -27,6 +27,9 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
 			unsigned long nr, int chunk);
 void item_kill_tree(struct radix_tree_root *root);
 
+int tag_tagged_items(struct radix_tree_root *, unsigned long start,
+			unsigned long end, unsigned batch,
+			unsigned iftag, unsigned thentag);
 unsigned long find_item(struct radix_tree_root *, void *item);
 
 void tag_check(void);
-- 
2.10.2

[PATCH 23/29] idr: Add ida_is_empty

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

Two of the USB Gadgets were poking around in the internals of struct ida
in order to determine if it is empty.  Add the appropriate abstraction.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 drivers/usb/gadget/function/f_hid.c     | 6 +++---
 drivers/usb/gadget/function/f_printer.c | 6 +++---
 include/linux/idr.h                     | 5 +++++
 3 files changed, 11 insertions(+), 6 deletions(-)

diff --git a/drivers/usb/gadget/function/f_hid.c b/drivers/usb/gadget/function/f_hid.c
index e2966f8..b9b04fc 100644
--- a/drivers/usb/gadget/function/f_hid.c
+++ b/drivers/usb/gadget/function/f_hid.c
@@ -840,7 +840,7 @@ static void hidg_free_inst(struct usb_function_instance *f)
 	mutex_lock(&hidg_ida_lock);
 
 	hidg_put_minor(opts->minor);
-	if (idr_is_empty(&hidg_ida.idr))
+	if (ida_is_empty(&hidg_ida))
 		ghid_cleanup();
 
 	mutex_unlock(&hidg_ida_lock);
@@ -866,7 +866,7 @@ static struct usb_function_instance *hidg_alloc_inst(void)
 
 	mutex_lock(&hidg_ida_lock);
 
-	if (idr_is_empty(&hidg_ida.idr)) {
+	if (ida_is_empty(&hidg_ida)) {
 		status = ghid_setup(NULL, HIDG_MINORS);
 		if (status)  {
 			ret = ERR_PTR(status);
@@ -879,7 +879,7 @@ static struct usb_function_instance *hidg_alloc_inst(void)
 	if (opts->minor < 0) {
 		ret = ERR_PTR(opts->minor);
 		kfree(opts);
-		if (idr_is_empty(&hidg_ida.idr))
+		if (ida_is_empty(&hidg_ida))
 			ghid_cleanup();
 		goto unlock;
 	}
diff --git a/drivers/usb/gadget/function/f_printer.c b/drivers/usb/gadget/function/f_printer.c
index 0de36cd..8054da9 100644
--- a/drivers/usb/gadget/function/f_printer.c
+++ b/drivers/usb/gadget/function/f_printer.c
@@ -1265,7 +1265,7 @@ static void gprinter_free_inst(struct usb_function_instance *f)
 	mutex_lock(&printer_ida_lock);
 
 	gprinter_put_minor(opts->minor);
-	if (idr_is_empty(&printer_ida.idr))
+	if (ida_is_empty(&printer_ida))
 		gprinter_cleanup();
 
 	mutex_unlock(&printer_ida_lock);
@@ -1289,7 +1289,7 @@ static struct usb_function_instance *gprinter_alloc_inst(void)
 
 	mutex_lock(&printer_ida_lock);
 
-	if (idr_is_empty(&printer_ida.idr)) {
+	if (ida_is_empty(&printer_ida)) {
 		status = gprinter_setup(PRINTER_MINORS);
 		if (status) {
 			ret = ERR_PTR(status);
@@ -1302,7 +1302,7 @@ static struct usb_function_instance *gprinter_alloc_inst(void)
 	if (opts->minor < 0) {
 		ret = ERR_PTR(opts->minor);
 		kfree(opts);
-		if (idr_is_empty(&printer_ida.idr))
+		if (ida_is_empty(&printer_ida))
 			gprinter_cleanup();
 		goto unlock;
 	}
diff --git a/include/linux/idr.h b/include/linux/idr.h
index 083d61e..3639a28 100644
--- a/include/linux/idr.h
+++ b/include/linux/idr.h
@@ -195,6 +195,11 @@ static inline int ida_get_new(struct ida *ida, int *p_id)
 	return ida_get_new_above(ida, 0, p_id);
 }
 
+static inline bool ida_is_empty(struct ida *ida)
+{
+	return idr_is_empty(&ida->idr);
+}
+
 void __init idr_init_cache(void);
 
 #endif /* __IDR_H__ */
-- 
2.10.2

[PATCH 24/29] tpm: Use idr_find(), not idr_find_slowpath()

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

idr_find_slowpath() is not intended to be part of the public API, it's
an implementation detail.  There's no reason to skip straight to the
slowpath here.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 drivers/char/tpm/tpm-chip.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/char/tpm/tpm-chip.c b/drivers/char/tpm/tpm-chip.c
index e595013..447b6ca 100644
--- a/drivers/char/tpm/tpm-chip.c
+++ b/drivers/char/tpm/tpm-chip.c
@@ -84,7 +84,7 @@ EXPORT_SYMBOL_GPL(tpm_put_ops);
  *
  * The return'd chip has been tpm_try_get_ops'd and must be released via
  * tpm_put_ops
-  */
+ */
 struct tpm_chip *tpm_chip_find_get(int chip_num)
 {
 	struct tpm_chip *chip, *res = NULL;
@@ -103,7 +103,7 @@ struct tpm_chip *tpm_chip_find_get(int chip_num)
 			}
 		} while (chip_prev != chip_num);
 	} else {
-		chip = idr_find_slowpath(&dev_nums_idr, chip_num);
+		chip = idr_find(&dev_nums_idr, chip_num);
 		if (chip && !tpm_try_get_ops(chip))
 			res = chip;
 	}
-- 
2.10.2

[PATCH 25/29] rxrpc: Abstract away knowledge of IDR internals

[Matthew Wilcox]

From: Matthew Wilcox <mawilcox@microsoft.com>

Add idr_get_cursor() / idr_set_cursor() APIs, and remove the rounding
up to IDR_SIZE.

Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
---
 include/linux/idr.h     | 26 ++++++++++++++++++++++++++
 net/rxrpc/af_rxrpc.c    | 11 ++++++-----
 net/rxrpc/conn_client.c |  4 ++--
 3 files changed, 34 insertions(+), 7 deletions(-)

diff --git a/include/linux/idr.h b/include/linux/idr.h
index 3639a28..1eb755f 100644
--- a/include/linux/idr.h
+++ b/include/linux/idr.h
@@ -56,6 +56,32 @@ struct idr {
 #define DEFINE_IDR(name)	struct idr name = IDR_INIT(name)
 
 /**
+ * idr_get_cursor - Return the current position of the cyclic allocator
+ * @idr: idr handle
+ *
+ * The value returned is the value that will be next returned from
+ * idr_alloc_cyclic() if it is free (otherwise the search will start from
+ * this position).
+ */
+static inline unsigned int idr_get_cursor(struct idr *idr)
+{
+	return READ_ONCE(idr->cur);
+}
+
+/**
+ * idr_set_cursor - Set the current position of the cyclic allocator
+ * @idr: idr handle
+ * @val: new position
+ *
+ * The next call to idr_alloc_cyclic() will return @val if it is free
+ * (otherwise the search will start from this position).
+ */
+static inline void idr_set_cursor(struct idr *idr, unsigned int val)
+{
+	WRITE_ONCE(idr->cur, val);
+}
+
+/**
  * DOC: idr sync
  * idr synchronization (stolen from radix-tree.h)
  *
diff --git a/net/rxrpc/af_rxrpc.c b/net/rxrpc/af_rxrpc.c
index 2d59c9b..5f63f6d 100644
--- a/net/rxrpc/af_rxrpc.c
+++ b/net/rxrpc/af_rxrpc.c
@@ -762,16 +762,17 @@ static const struct net_proto_family rxrpc_family_ops = {
 static int __init af_rxrpc_init(void)
 {
 	int ret = -1;
+	unsigned int tmp;
 
 	BUILD_BUG_ON(sizeof(struct rxrpc_skb_priv) > FIELD_SIZEOF(struct sk_buff, cb));
 
 	get_random_bytes(&rxrpc_epoch, sizeof(rxrpc_epoch));
 	rxrpc_epoch |= RXRPC_RANDOM_EPOCH;
-	get_random_bytes(&rxrpc_client_conn_ids.cur,
-			 sizeof(rxrpc_client_conn_ids.cur));
-	rxrpc_client_conn_ids.cur &= 0x3fffffff;
-	if (rxrpc_client_conn_ids.cur == 0)
-		rxrpc_client_conn_ids.cur = 1;
+	get_random_bytes(&tmp, sizeof(tmp));
+	tmp &= 0x3fffffff;
+	if (tmp == 0)
+		tmp = 1;
+	idr_set_cursor(&rxrpc_client_conn_ids, tmp);
 
 	ret = -ENOMEM;
 	rxrpc_call_jar = kmem_cache_create(
diff --git a/net/rxrpc/conn_client.c b/net/rxrpc/conn_client.c
index 60ef960..9706f60 100644
--- a/net/rxrpc/conn_client.c
+++ b/net/rxrpc/conn_client.c
@@ -263,12 +263,12 @@ static bool rxrpc_may_reuse_conn(struct rxrpc_connection *conn)
 	 * times the maximum number of client conns away from the current
 	 * allocation point to try and keep the IDs concentrated.
 	 */
-	id_cursor = READ_ONCE(rxrpc_client_conn_ids.cur);
+	id_cursor = idr_get_cursor(&rxrpc_client_conn_ids);
 	id = conn->proto.cid >> RXRPC_CIDSHIFT;
 	distance = id - id_cursor;
 	if (distance < 0)
 		distance = -distance;
-	limit = round_up(rxrpc_max_client_connections, IDR_SIZE) * 4;
+	limit = rxrpc_max_client_connections * 4;
 	if (distance > limit)
 		goto mark_dont_reuse;
 
-- 
2.10.2

[PATCH 26/29] idr: Reduce the number of bits per level from 8 to 6

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

In preparation for merging the IDR and radix tree, reduce the fanout at
each level from 256 to 64.  If this causes a performance problem then
a bisect will point to this commit, and we'll have a better idea about
what we might do to fix it.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 include/linux/idr.h | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/include/linux/idr.h b/include/linux/idr.h
index 1eb755f..3c01b89 100644
--- a/include/linux/idr.h
+++ b/include/linux/idr.h
@@ -18,12 +18,11 @@
 #include <linux/rcupdate.h>
 
 /*
- * We want shallower trees and thus more bits covered at each layer.  8
- * bits gives us large enough first layer for most use cases and maximum
- * tree depth of 4.  Each idr_layer is slightly larger than 2k on 64bit and
- * 1k on 32bit.
+ * Using 6 bits at each layer allows us to allocate 7 layers out of each page.
+ * 8 bits only gave us 3 layers out of every pair of pages, which is less
+ * efficient except for trees with a largest element between 192-255 inclusive.
  */
-#define IDR_BITS 8
+#define IDR_BITS 6
 #define IDR_SIZE (1 << IDR_BITS)
 #define IDR_MASK ((1 << IDR_BITS)-1)
 
-- 
2.10.2

[PATCH 27/29] radix tree test suite: Add some more functionality

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

IDR needs more functionality from the kernel: kmalloc()/kfree(), and xchg().

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 tools/testing/radix-tree/linux.c        | 15 +++++++++++++++
 tools/testing/radix-tree/linux/kernel.h |  3 +++
 tools/testing/radix-tree/linux/slab.h   |  3 +++
 3 files changed, 21 insertions(+)

diff --git a/tools/testing/radix-tree/linux.c b/tools/testing/radix-tree/linux.c
index 1f32a16..ff0452e 100644
--- a/tools/testing/radix-tree/linux.c
+++ b/tools/testing/radix-tree/linux.c
@@ -54,6 +54,21 @@ void kmem_cache_free(struct kmem_cache *cachep, void *objp)
 	free(objp);
 }
 
+void *kmalloc(size_t size, gfp_t gfp)
+{
+	void *ret = malloc(size);
+	uatomic_inc(&nr_allocated);
+	return ret;
+}
+
+void kfree(void *p)
+{
+	if (!p)
+		return;
+	uatomic_dec(&nr_allocated);
+	free(p);
+}
+
 struct kmem_cache *
 kmem_cache_create(const char *name, size_t size, size_t offset,
 	unsigned long flags, void (*ctor)(void *))
diff --git a/tools/testing/radix-tree/linux/kernel.h b/tools/testing/radix-tree/linux/kernel.h
index 23e77f5..9b43b49 100644
--- a/tools/testing/radix-tree/linux/kernel.h
+++ b/tools/testing/radix-tree/linux/kernel.h
@@ -8,6 +8,7 @@
 #include <limits.h>
 
 #include "../../include/linux/compiler.h"
+#include "../../include/linux/err.h"
 #include "../../../include/linux/kconfig.h"
 
 #ifdef BENCHMARK
@@ -58,4 +59,6 @@ static inline int in_interrupt(void)
 #define round_up(x, y) ((((x)-1) | __round_mask(x, y))+1)
 #define round_down(x, y) ((x) & ~__round_mask(x, y))
 
+#define xchg(ptr, x)	uatomic_xchg(ptr, x)
+
 #endif /* _KERNEL_H */
diff --git a/tools/testing/radix-tree/linux/slab.h b/tools/testing/radix-tree/linux/slab.h
index 452e2bf..446639f 100644
--- a/tools/testing/radix-tree/linux/slab.h
+++ b/tools/testing/radix-tree/linux/slab.h
@@ -7,6 +7,9 @@
 #define SLAB_PANIC 2
 #define SLAB_RECLAIM_ACCOUNT    0x00020000UL            /* Objects are reclaimable */
 
+void *kmalloc(size_t size, gfp_t);
+void kfree(void *);
+
 struct kmem_cache {
 	int size;
 	void (*ctor)(void *);
-- 
2.10.2

[PATCH 28/29] radix-tree: Create all_tag_set

[Matthew Wilcox]

From: Matthew Wilcox <willy@linux.intel.com>

all_tag_set() sets every tag on a node.  This is useful for the IDR code
when we're creating new nodes which contain only free slots.

Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
---
 lib/radix-tree.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index c8ef657..e063ca2 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -22,6 +22,8 @@
  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#include <linux/bitmap.h>
+#include <linux/bitops.h>
 #include <linux/errno.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
@@ -33,7 +35,6 @@
 #include <linux/notifier.h>
 #include <linux/cpu.h>
 #include <linux/string.h>
-#include <linux/bitops.h>
 #include <linux/rcupdate.h>
 #include <linux/preempt.h>		/* in_interrupt() */
 
@@ -184,6 +185,11 @@ static inline int any_tag_set(struct radix_tree_node *node, unsigned int tag)
 	return 0;
 }
 
+static inline void all_tag_set(struct radix_tree_node *node, unsigned int tag)
+{
+	bitmap_fill(node->tags[tag], RADIX_TREE_MAP_SIZE);
+}
+
 /**
  * radix_tree_find_next_bit - find the next set bit in a memory region
  *
-- 
2.10.2

[PATCH 29/29] Reimplement IDR and IDA using the radix tree

[Matthew Wilcox]

From: Matthew Wilcox <willy@infradead.org>

The IDR is very similar to the radix tree.  It has some functionality
that the radix tree did not have (alloc next free, cyclic allocation,
a callback-based for_each, destroy tree), which is readily implementable
on top of the radix tree.  A few small changes were needed in order to
use a tag to represent nodes with free space below them.

The IDA is reimplemented as a client of the newly enhanced radix tree.
As in the current implementation, it uses a bitmap at the last level of
the tree.

Signed-off-by: Matthew Wilcox <willy@infradead.org>
---
 include/linux/idr.h                     |  132 ++--
 include/linux/radix-tree.h              |    5 +-
 init/main.c                             |    3 +-
 lib/idr.c                               | 1075 -------------------------------
 lib/radix-tree.c                        |  624 ++++++++++++++++--
 tools/testing/radix-tree/Makefile       |    5 +-
 tools/testing/radix-tree/idr.c          |  148 +++++
 tools/testing/radix-tree/linux/idr.h    |    1 +
 tools/testing/radix-tree/linux/kernel.h |    2 +
 tools/testing/radix-tree/main.c         |    6 +
 tools/testing/radix-tree/test.h         |    2 +
 11 files changed, 791 insertions(+), 1212 deletions(-)
 create mode 100644 tools/testing/radix-tree/idr.c
 create mode 100644 tools/testing/radix-tree/linux/idr.h

diff --git a/include/linux/idr.h b/include/linux/idr.h
index 3c01b89..0035799 100644
--- a/include/linux/idr.h
+++ b/include/linux/idr.h
@@ -12,47 +12,22 @@
 #ifndef __IDR_H__
 #define __IDR_H__
 
-#include <linux/types.h>
-#include <linux/bitops.h>
-#include <linux/init.h>
-#include <linux/rcupdate.h>
-
-/*
- * Using 6 bits at each layer allows us to allocate 7 layers out of each page.
- * 8 bits only gave us 3 layers out of every pair of pages, which is less
- * efficient except for trees with a largest element between 192-255 inclusive.
- */
-#define IDR_BITS 6
-#define IDR_SIZE (1 << IDR_BITS)
-#define IDR_MASK ((1 << IDR_BITS)-1)
-
-struct idr_layer {
-	int			prefix;	/* the ID prefix of this idr_layer */
-	int			layer;	/* distance from leaf */
-	struct idr_layer __rcu	*ary[1<<IDR_BITS];
-	int			count;	/* When zero, we can release it */
-	union {
-		/* A zero bit means "space here" */
-		DECLARE_BITMAP(bitmap, IDR_SIZE);
-		struct rcu_head		rcu_head;
-	};
-};
+#include <linux/radix-tree.h>
+#include <linux/gfp.h>
 
 struct idr {
-	struct idr_layer __rcu	*hint;	/* the last layer allocated from */
-	struct idr_layer __rcu	*top;
-	int			layers;	/* only valid w/o concurrent changes */
-	int			cur;	/* current pos for cyclic allocation */
-	spinlock_t		lock;
-	int			id_free_cnt;
-	struct idr_layer	*id_free;
+	struct radix_tree_root	idr_rt;
+	unsigned int		idr_next;
 };
 
-#define IDR_INIT(name)							\
+/* Set the IDR flag and the IDR_FREE tag */
+#define IDR_RT_MARKER		((__force gfp_t)(3 << __GFP_BITS_SHIFT))
+
+#define IDR_INIT							\
 {									\
-	.lock			= __SPIN_LOCK_UNLOCKED(name.lock),	\
+	.idr_rt = RADIX_TREE_INIT(IDR_RT_MARKER)			\
 }
-#define DEFINE_IDR(name)	struct idr name = IDR_INIT(name)
+#define DEFINE_IDR(name)	struct idr name = IDR_INIT
 
 /**
  * idr_get_cursor - Return the current position of the cyclic allocator
@@ -64,7 +39,7 @@ struct idr {
  */
 static inline unsigned int idr_get_cursor(struct idr *idr)
 {
-	return READ_ONCE(idr->cur);
+	return READ_ONCE(idr->idr_next);
 }
 
 /**
@@ -77,7 +52,7 @@ static inline unsigned int idr_get_cursor(struct idr *idr)
  */
 static inline void idr_set_cursor(struct idr *idr, unsigned int val)
 {
-	WRITE_ONCE(idr->cur, val);
+	WRITE_ONCE(idr->idr_next, val);
 }
 
 /**
@@ -97,22 +72,30 @@ static inline void idr_set_cursor(struct idr *idr, unsigned int val)
  * period).
  */
 
-/*
- * This is what we export.
- */
-
-void *idr_find_slowpath(struct idr *idp, int id);
 void idr_preload(gfp_t gfp_mask);
-int idr_alloc(struct idr *idp, void *ptr, int start, int end, gfp_t gfp_mask);
-int idr_alloc_cyclic(struct idr *idr, void *ptr, int start, int end, gfp_t gfp_mask);
-int idr_for_each(struct idr *idp,
+int idr_alloc(struct idr *, void *, int start, int end, gfp_t gfp_mask);
+int idr_alloc_cyclic(struct idr *, void *, int start, int end, gfp_t gfp_mask);
+int idr_for_each(struct idr *,
 		 int (*fn)(int id, void *p, void *data), void *data);
-void *idr_get_next(struct idr *idp, int *nextid);
-void *idr_replace(struct idr *idp, void *ptr, int id);
-void idr_remove(struct idr *idp, int id);
-void idr_destroy(struct idr *idp);
-void idr_init(struct idr *idp);
-bool idr_is_empty(struct idr *idp);
+void *idr_get_next(struct idr *, int *nextid);
+void *idr_replace(struct idr *, void *, int id);
+void idr_destroy(struct idr *);
+
+static inline void idr_remove(struct idr *idp, int id)
+{
+	radix_tree_delete(&idp->idr_rt, id);
+}
+
+static inline void idr_init(struct idr *idp)
+{
+	memset(idp, 0, sizeof(*idp));
+	idp->idr_rt.gfp_mask = IDR_RT_MARKER;
+}
+
+static inline bool idr_is_empty(struct idr *idp)
+{
+	return radix_tree_empty(&idp->idr_rt);
+}
 
 /**
  * idr_preload_end - end preload section started with idr_preload()
@@ -139,17 +122,12 @@ static inline void idr_preload_end(void)
  */
 static inline void *idr_find(struct idr *idr, int id)
 {
-	struct idr_layer *hint = rcu_dereference_raw(idr->hint);
-
-	if (hint && (id & ~IDR_MASK) == hint->prefix)
-		return rcu_dereference_raw(hint->ary[id & IDR_MASK]);
-
-	return idr_find_slowpath(idr, id);
+	return radix_tree_lookup(&idr->idr_rt, id);
 }
 
 /**
  * idr_for_each_entry - iterate over an idr's elements of a given type
- * @idp:     idr handle
+ * @idr:     idr handle
  * @entry:   the type * to use as cursor
  * @id:      id entry's key
  *
@@ -157,57 +135,58 @@ static inline void *idr_find(struct idr *idr, int id)
  * after normal terminatinon @entry is left with the value NULL.  This
  * is convenient for a "not found" value.
  */
-#define idr_for_each_entry(idp, entry, id)			\
-	for (id = 0; ((entry) = idr_get_next(idp, &(id))) != NULL; ++id)
+#define idr_for_each_entry(idr, entry, id)			\
+	for (id = 0; ((entry) = idr_get_next(idr, &(id))) != NULL; ++id)
 
 /**
- * idr_for_each_entry - continue iteration over an idr's elements of a given type
- * @idp:     idr handle
+ * idr_for_each_entry_continue - continue iteration over an idr's elements of a given type
+ * @idr:     idr handle
  * @entry:   the type * to use as cursor
  * @id:      id entry's key
  *
  * Continue to iterate over list of given type, continuing after
  * the current position.
  */
-#define idr_for_each_entry_continue(idp, entry, id)			\
-	for ((entry) = idr_get_next((idp), &(id));			\
+#define idr_for_each_entry_continue(idr, entry, id)			\
+	for ((entry) = idr_get_next((idr), &(id));			\
 	     entry;							\
-	     ++id, (entry) = idr_get_next((idp), &(id)))
+	     ++id, (entry) = idr_get_next((idr), &(id)))
 
 /*
  * IDA - IDR based id allocator, use when translation from id to
  * pointer isn't necessary.
- *
- * IDA_BITMAP_LONGS is calculated to be one less to accommodate
- * ida_bitmap->nr_busy so that the whole struct fits in 128 bytes.
  */
 #define IDA_CHUNK_SIZE		128	/* 128 bytes per chunk */
-#define IDA_BITMAP_LONGS	(IDA_CHUNK_SIZE / sizeof(long) - 1)
+#define IDA_BITMAP_LONGS	(IDA_CHUNK_SIZE / sizeof(long))
 #define IDA_BITMAP_BITS 	(IDA_BITMAP_LONGS * sizeof(long) * 8)
 
 struct ida_bitmap {
-	long			nr_busy;
 	unsigned long		bitmap[IDA_BITMAP_LONGS];
 };
 
 struct ida {
-	struct idr		idr;
+	struct radix_tree_root	ida_rt;
 	struct ida_bitmap	*free_bitmap;
 };
 
-#define IDA_INIT(name)		{ .idr = IDR_INIT((name).idr), .free_bitmap = NULL, }
-#define DEFINE_IDA(name)	struct ida name = IDA_INIT(name)
+#define IDA_INIT		{ .ida_rt = RADIX_TREE_INIT(IDR_RT_MARKER), }
+#define DEFINE_IDA(name)	struct ida name = IDA_INIT
 
 int ida_pre_get(struct ida *ida, gfp_t gfp_mask);
 int ida_get_new_above(struct ida *ida, int starting_id, int *p_id);
 void ida_remove(struct ida *ida, int id);
 void ida_destroy(struct ida *ida);
-void ida_init(struct ida *ida);
 
 int ida_simple_get(struct ida *ida, unsigned int start, unsigned int end,
 		   gfp_t gfp_mask);
 void ida_simple_remove(struct ida *ida, unsigned int id);
 
+static inline void ida_init(struct ida *ida)
+{
+	memset(ida, 0, sizeof(*ida));
+	ida->ida_rt.gfp_mask = IDR_RT_MARKER;
+}
+
 /**
  * ida_get_new - allocate new ID
  * @ida:	idr handle
@@ -222,9 +201,6 @@ static inline int ida_get_new(struct ida *ida, int *p_id)
 
 static inline bool ida_is_empty(struct ida *ida)
 {
-	return idr_is_empty(&ida->idr);
+	return radix_tree_empty(&ida->ida_rt);
 }
-
-void __init idr_init_cache(void);
-
 #endif /* __IDR_H__ */
diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index ca4eea1..6483c73c 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -98,7 +98,10 @@ struct radix_tree_node {
 	unsigned long	tags[RADIX_TREE_MAX_TAGS][RADIX_TREE_TAG_LONGS];
 };
 
-/* root tags are stored in gfp_mask, shifted by __GFP_BITS_SHIFT */
+/* The top bits of gfp_mask are used to store the root tags and the IDR flag */
+#define ROOT_IS_IDR	(1 << __GFP_BITS_SHIFT)
+#define ROOT_TAG_SHIFT	(__GFP_BITS_SHIFT + 1)
+
 struct radix_tree_root {
 	gfp_t			gfp_mask;
 	struct radix_tree_node	__rcu *rnode;
diff --git a/init/main.c b/init/main.c
index 2858be7..2757b84 100644
--- a/init/main.c
+++ b/init/main.c
@@ -550,14 +550,13 @@ asmlinkage __visible void __init start_kernel(void)
 	if (WARN(!irqs_disabled(),
 		 "Interrupts were enabled *very* early, fixing it\n"))
 		local_irq_disable();
-	idr_init_cache();
+	radix_tree_init();
 	rcu_init();
 
 	/* trace_printk() and trace points may be used after this */
 	trace_init();
 
 	context_tracking_init();
-	radix_tree_init();
 	/* init some links before init_ISA_irqs() */
 	early_irq_init();
 	init_IRQ();
diff --git a/lib/idr.c b/lib/idr.c
index 6098336..3c88abb7 100644
--- a/lib/idr.c
+++ b/lib/idr.c
@@ -1,1068 +1,8 @@
-/*
- * 2002-10-18  written by Jim Houston jim.houston@ccur.com
- *	Copyright (C) 2002 by Concurrent Computer Corporation
- *	Distributed under the GNU GPL license version 2.
- *
- * Modified by George Anzinger to reuse immediately and to use
- * find bit instructions.  Also removed _irq on spinlocks.
- *
- * Modified by Nadia Derbey to make it RCU safe.
- *
- * Small id to pointer translation service.
- *
- * It uses a radix tree like structure as a sparse array indexed
- * by the id to obtain the pointer.  The bitmap makes allocating
- * a new id quick.
- *
- * You call it to allocate an id (an int) an associate with that id a
- * pointer or what ever, we treat it as a (void *).  You can pass this
- * id to a user for him to pass back at a later time.  You then pass
- * that id to this code and it returns your pointer.
- */
-
-#ifndef TEST                        // to test in user space...
-#include <linux/slab.h>
-#include <linux/init.h>
-#include <linux/export.h>
-#endif
-#include <linux/err.h>
-#include <linux/string.h>
 #include <linux/idr.h>
 #include <linux/spinlock.h>
-#include <linux/percpu.h>
-
-#define MAX_IDR_SHIFT		(sizeof(int) * 8 - 1)
-#define MAX_IDR_BIT		(1U << MAX_IDR_SHIFT)
-
-/* Leave the possibility of an incomplete final layer */
-#define MAX_IDR_LEVEL ((MAX_IDR_SHIFT + IDR_BITS - 1) / IDR_BITS)
-
-/* Number of id_layer structs to leave in free list */
-#define MAX_IDR_FREE (MAX_IDR_LEVEL * 2)
 
-static struct kmem_cache *idr_layer_cache;
-static DEFINE_PER_CPU(struct idr_layer *, idr_preload_head);
-static DEFINE_PER_CPU(int, idr_preload_cnt);
 static DEFINE_SPINLOCK(simple_ida_lock);
 
-/* the maximum ID which can be allocated given idr->layers */
-static int idr_max(int layers)
-{
-	int bits = min_t(int, layers * IDR_BITS, MAX_IDR_SHIFT);
-
-	return (1 << bits) - 1;
-}
-
-/*
- * Prefix mask for an idr_layer at @layer.  For layer 0, the prefix mask is
- * all bits except for the lower IDR_BITS.  For layer 1, 2 * IDR_BITS, and
- * so on.
- */
-static int idr_layer_prefix_mask(int layer)
-{
-	return ~idr_max(layer + 1);
-}
-
-static struct idr_layer *get_from_free_list(struct idr *idp)
-{
-	struct idr_layer *p;
-	unsigned long flags;
-
-	spin_lock_irqsave(&idp->lock, flags);
-	if ((p = idp->id_free)) {
-		idp->id_free = p->ary[0];
-		idp->id_free_cnt--;
-		p->ary[0] = NULL;
-	}
-	spin_unlock_irqrestore(&idp->lock, flags);
-	return(p);
-}
-
-/**
- * idr_layer_alloc - allocate a new idr_layer
- * @gfp_mask: allocation mask
- * @layer_idr: optional idr to allocate from
- *
- * If @layer_idr is %NULL, directly allocate one using @gfp_mask or fetch
- * one from the per-cpu preload buffer.  If @layer_idr is not %NULL, fetch
- * an idr_layer from @idr->id_free.
- *
- * @layer_idr is to maintain backward compatibility with the old alloc
- * interface - idr_pre_get() and idr_get_new*() - and will be removed
- * together with per-pool preload buffer.
- */
-static struct idr_layer *idr_layer_alloc(gfp_t gfp_mask, struct idr *layer_idr)
-{
-	struct idr_layer *new;
-
-	/* this is the old path, bypass to get_from_free_list() */
-	if (layer_idr)
-		return get_from_free_list(layer_idr);
-
-	/*
-	 * Try to allocate directly from kmem_cache.  We want to try this
-	 * before preload buffer; otherwise, non-preloading idr_alloc()
-	 * users will end up taking advantage of preloading ones.  As the
-	 * following is allowed to fail for preloaded cases, suppress
-	 * warning this time.
-	 */
-	new = kmem_cache_zalloc(idr_layer_cache, gfp_mask | __GFP_NOWARN);
-	if (new)
-		return new;
-
-	/*
-	 * Try to fetch one from the per-cpu preload buffer if in process
-	 * context.  See idr_preload() for details.
-	 */
-	if (!in_interrupt()) {
-		preempt_disable();
-		new = __this_cpu_read(idr_preload_head);
-		if (new) {
-			__this_cpu_write(idr_preload_head, new->ary[0]);
-			__this_cpu_dec(idr_preload_cnt);
-			new->ary[0] = NULL;
-		}
-		preempt_enable();
-		if (new)
-			return new;
-	}
-
-	/*
-	 * Both failed.  Try kmem_cache again w/o adding __GFP_NOWARN so
-	 * that memory allocation failure warning is printed as intended.
-	 */
-	return kmem_cache_zalloc(idr_layer_cache, gfp_mask);
-}
-
-static void idr_layer_rcu_free(struct rcu_head *head)
-{
-	struct idr_layer *layer;
-
-	layer = container_of(head, struct idr_layer, rcu_head);
-	kmem_cache_free(idr_layer_cache, layer);
-}
-
-static inline void free_layer(struct idr *idr, struct idr_layer *p)
-{
-	if (idr->hint == p)
-		RCU_INIT_POINTER(idr->hint, NULL);
-	call_rcu(&p->rcu_head, idr_layer_rcu_free);
-}
-
-/* only called when idp->lock is held */
-static void __move_to_free_list(struct idr *idp, struct idr_layer *p)
-{
-	p->ary[0] = idp->id_free;
-	idp->id_free = p;
-	idp->id_free_cnt++;
-}
-
-static void move_to_free_list(struct idr *idp, struct idr_layer *p)
-{
-	unsigned long flags;
-
-	/*
-	 * Depends on the return element being zeroed.
-	 */
-	spin_lock_irqsave(&idp->lock, flags);
-	__move_to_free_list(idp, p);
-	spin_unlock_irqrestore(&idp->lock, flags);
-}
-
-static void idr_mark_full(struct idr_layer **pa, int id)
-{
-	struct idr_layer *p = pa[0];
-	int l = 0;
-
-	__set_bit(id & IDR_MASK, p->bitmap);
-	/*
-	 * If this layer is full mark the bit in the layer above to
-	 * show that this part of the radix tree is full.  This may
-	 * complete the layer above and require walking up the radix
-	 * tree.
-	 */
-	while (bitmap_full(p->bitmap, IDR_SIZE)) {
-		if (!(p = pa[++l]))
-			break;
-		id = id >> IDR_BITS;
-		__set_bit((id & IDR_MASK), p->bitmap);
-	}
-}
-
-static int __idr_pre_get(struct idr *idp, gfp_t gfp_mask)
-{
-	while (idp->id_free_cnt < MAX_IDR_FREE) {
-		struct idr_layer *new;
-		new = kmem_cache_zalloc(idr_layer_cache, gfp_mask);
-		if (new == NULL)
-			return (0);
-		move_to_free_list(idp, new);
-	}
-	return 1;
-}
-
-/**
- * sub_alloc - try to allocate an id without growing the tree depth
- * @idp: idr handle
- * @starting_id: id to start search at
- * @pa: idr_layer[MAX_IDR_LEVEL] used as backtrack buffer
- * @gfp_mask: allocation mask for idr_layer_alloc()
- * @layer_idr: optional idr passed to idr_layer_alloc()
- *
- * Allocate an id in range [@starting_id, INT_MAX] from @idp without
- * growing its depth.  Returns
- *
- *  the allocated id >= 0 if successful,
- *  -EAGAIN if the tree needs to grow for allocation to succeed,
- *  -ENOSPC if the id space is exhausted,
- *  -ENOMEM if more idr_layers need to be allocated.
- */
-static int sub_alloc(struct idr *idp, int *starting_id, struct idr_layer **pa,
-		     gfp_t gfp_mask, struct idr *layer_idr)
-{
-	int n, m, sh;
-	struct idr_layer *p, *new;
-	int l, id, oid;
-
-	id = *starting_id;
- restart:
-	p = idp->top;
-	l = idp->layers;
-	pa[l--] = NULL;
-	while (1) {
-		/*
-		 * We run around this while until we reach the leaf node...
-		 */
-		n = (id >> (IDR_BITS*l)) & IDR_MASK;
-		m = find_next_zero_bit(p->bitmap, IDR_SIZE, n);
-		if (m == IDR_SIZE) {
-			/* no space available go back to previous layer. */
-			l++;
-			oid = id;
-			id = (id | ((1 << (IDR_BITS * l)) - 1)) + 1;
-
-			/* if already at the top layer, we need to grow */
-			if (id > idr_max(idp->layers)) {
-				*starting_id = id;
-				return -EAGAIN;
-			}
-			p = pa[l];
-			BUG_ON(!p);
-
-			/* If we need to go up one layer, continue the
-			 * loop; otherwise, restart from the top.
-			 */
-			sh = IDR_BITS * (l + 1);
-			if (oid >> sh == id >> sh)
-				continue;
-			else
-				goto restart;
-		}
-		if (m != n) {
-			sh = IDR_BITS*l;
-			id = ((id >> sh) ^ n ^ m) << sh;
-		}
-		if ((id >= MAX_IDR_BIT) || (id < 0))
-			return -ENOSPC;
-		if (l == 0)
-			break;
-		/*
-		 * Create the layer below if it is missing.
-		 */
-		if (!p->ary[m]) {
-			new = idr_layer_alloc(gfp_mask, layer_idr);
-			if (!new)
-				return -ENOMEM;
-			new->layer = l-1;
-			new->prefix = id & idr_layer_prefix_mask(new->layer);
-			rcu_assign_pointer(p->ary[m], new);
-			p->count++;
-		}
-		pa[l--] = p;
-		p = p->ary[m];
-	}
-
-	pa[l] = p;
-	return id;
-}
-
-static int idr_get_empty_slot(struct idr *idp, int starting_id,
-			      struct idr_layer **pa, gfp_t gfp_mask,
-			      struct idr *layer_idr)
-{
-	struct idr_layer *p, *new;
-	int layers, v, id;
-	unsigned long flags;
-
-	id = starting_id;
-build_up:
-	p = idp->top;
-	layers = idp->layers;
-	if (unlikely(!p)) {
-		if (!(p = idr_layer_alloc(gfp_mask, layer_idr)))
-			return -ENOMEM;
-		p->layer = 0;
-		layers = 1;
-	}
-	/*
-	 * Add a new layer to the top of the tree if the requested
-	 * id is larger than the currently allocated space.
-	 */
-	while (id > idr_max(layers)) {
-		layers++;
-		if (!p->count) {
-			/* special case: if the tree is currently empty,
-			 * then we grow the tree by moving the top node
-			 * upwards.
-			 */
-			p->layer++;
-			WARN_ON_ONCE(p->prefix);
-			continue;
-		}
-		if (!(new = idr_layer_alloc(gfp_mask, layer_idr))) {
-			/*
-			 * The allocation failed.  If we built part of
-			 * the structure tear it down.
-			 */
-			spin_lock_irqsave(&idp->lock, flags);
-			for (new = p; p && p != idp->top; new = p) {
-				p = p->ary[0];
-				new->ary[0] = NULL;
-				new->count = 0;
-				bitmap_clear(new->bitmap, 0, IDR_SIZE);
-				__move_to_free_list(idp, new);
-			}
-			spin_unlock_irqrestore(&idp->lock, flags);
-			return -ENOMEM;
-		}
-		new->ary[0] = p;
-		new->count = 1;
-		new->layer = layers-1;
-		new->prefix = id & idr_layer_prefix_mask(new->layer);
-		if (bitmap_full(p->bitmap, IDR_SIZE))
-			__set_bit(0, new->bitmap);
-		p = new;
-	}
-	rcu_assign_pointer(idp->top, p);
-	idp->layers = layers;
-	v = sub_alloc(idp, &id, pa, gfp_mask, layer_idr);
-	if (v == -EAGAIN)
-		goto build_up;
-	return(v);
-}
-
-/*
- * @id and @pa are from a successful allocation from idr_get_empty_slot().
- * Install the user pointer @ptr and mark the slot full.
- */
-static void idr_fill_slot(struct idr *idr, void *ptr, int id,
-			  struct idr_layer **pa)
-{
-	/* update hint used for lookup, cleared from free_layer() */
-	rcu_assign_pointer(idr->hint, pa[0]);
-
-	rcu_assign_pointer(pa[0]->ary[id & IDR_MASK], (struct idr_layer *)ptr);
-	pa[0]->count++;
-	idr_mark_full(pa, id);
-}
-
-
-/**
- * idr_preload - preload for idr_alloc()
- * @gfp_mask: allocation mask to use for preloading
- *
- * Preload per-cpu layer buffer for idr_alloc().  Can only be used from
- * process context and each idr_preload() invocation should be matched with
- * idr_preload_end().  Note that preemption is disabled while preloaded.
- *
- * The first idr_alloc() in the preloaded section can be treated as if it
- * were invoked with @gfp_mask used for preloading.  This allows using more
- * permissive allocation masks for idrs protected by spinlocks.
- *
- * For example, if idr_alloc() below fails, the failure can be treated as
- * if idr_alloc() were called with GFP_KERNEL rather than GFP_NOWAIT.
- *
- *	idr_preload(GFP_KERNEL);
- *	spin_lock(lock);
- *
- *	id = idr_alloc(idr, ptr, start, end, GFP_NOWAIT);
- *
- *	spin_unlock(lock);
- *	idr_preload_end();
- *	if (id < 0)
- *		error;
- */
-void idr_preload(gfp_t gfp_mask)
-{
-	/*
-	 * Consuming preload buffer from non-process context breaks preload
-	 * allocation guarantee.  Disallow usage from those contexts.
-	 */
-	WARN_ON_ONCE(in_interrupt());
-	might_sleep_if(gfpflags_allow_blocking(gfp_mask));
-
-	preempt_disable();
-
-	/*
-	 * idr_alloc() is likely to succeed w/o full idr_layer buffer and
-	 * return value from idr_alloc() needs to be checked for failure
-	 * anyway.  Silently give up if allocation fails.  The caller can
-	 * treat failures from idr_alloc() as if idr_alloc() were called
-	 * with @gfp_mask which should be enough.
-	 */
-	while (__this_cpu_read(idr_preload_cnt) < MAX_IDR_FREE) {
-		struct idr_layer *new;
-
-		preempt_enable();
-		new = kmem_cache_zalloc(idr_layer_cache, gfp_mask);
-		preempt_disable();
-		if (!new)
-			break;
-
-		/* link the new one to per-cpu preload list */
-		new->ary[0] = __this_cpu_read(idr_preload_head);
-		__this_cpu_write(idr_preload_head, new);
-		__this_cpu_inc(idr_preload_cnt);
-	}
-}
-EXPORT_SYMBOL(idr_preload);
-
-/**
- * idr_alloc - allocate new idr entry
- * @idr: the (initialized) idr
- * @ptr: pointer to be associated with the new id
- * @start: the minimum id (inclusive)
- * @end: the maximum id (exclusive, <= 0 for max)
- * @gfp_mask: memory allocation flags
- *
- * Allocate an id in [start, end) and associate it with @ptr.  If no ID is
- * available in the specified range, returns -ENOSPC.  On memory allocation
- * failure, returns -ENOMEM.
- *
- * Note that @end is treated as max when <= 0.  This is to always allow
- * using @start + N as @end as long as N is inside integer range.
- *
- * The user is responsible for exclusively synchronizing all operations
- * which may modify @idr.  However, read-only accesses such as idr_find()
- * or iteration can be performed under RCU read lock provided the user
- * destroys @ptr in RCU-safe way after removal from idr.
- */
-int idr_alloc(struct idr *idr, void *ptr, int start, int end, gfp_t gfp_mask)
-{
-	int max = end > 0 ? end - 1 : INT_MAX;	/* inclusive upper limit */
-	struct idr_layer *pa[MAX_IDR_LEVEL + 1];
-	int id;
-
-	might_sleep_if(gfpflags_allow_blocking(gfp_mask));
-
-	/* sanity checks */
-	if (WARN_ON_ONCE(start < 0))
-		return -EINVAL;
-	if (unlikely(max < start))
-		return -ENOSPC;
-
-	/* allocate id */
-	id = idr_get_empty_slot(idr, start, pa, gfp_mask, NULL);
-	if (unlikely(id < 0))
-		return id;
-	if (unlikely(id > max))
-		return -ENOSPC;
-
-	idr_fill_slot(idr, ptr, id, pa);
-	return id;
-}
-EXPORT_SYMBOL_GPL(idr_alloc);
-
-/**
- * idr_alloc_cyclic - allocate new idr entry in a cyclical fashion
- * @idr: the (initialized) idr
- * @ptr: pointer to be associated with the new id
- * @start: the minimum id (inclusive)
- * @end: the maximum id (exclusive, <= 0 for max)
- * @gfp_mask: memory allocation flags
- *
- * Essentially the same as idr_alloc, but prefers to allocate progressively
- * higher ids if it can. If the "cur" counter wraps, then it will start again
- * at the "start" end of the range and allocate one that has already been used.
- */
-int idr_alloc_cyclic(struct idr *idr, void *ptr, int start, int end,
-			gfp_t gfp_mask)
-{
-	int id;
-
-	id = idr_alloc(idr, ptr, max(start, idr->cur), end, gfp_mask);
-	if (id == -ENOSPC)
-		id = idr_alloc(idr, ptr, start, end, gfp_mask);
-
-	if (likely(id >= 0))
-		idr->cur = id + 1;
-	return id;
-}
-EXPORT_SYMBOL(idr_alloc_cyclic);
-
-static void idr_remove_warning(int id)
-{
-	WARN(1, "idr_remove called for id=%d which is not allocated.\n", id);
-}
-
-static void sub_remove(struct idr *idp, int shift, int id)
-{
-	struct idr_layer *p = idp->top;
-	struct idr_layer **pa[MAX_IDR_LEVEL + 1];
-	struct idr_layer ***paa = &pa[0];
-	struct idr_layer *to_free;
-	int n;
-
-	*paa = NULL;
-	*++paa = &idp->top;
-
-	while ((shift > 0) && p) {
-		n = (id >> shift) & IDR_MASK;
-		__clear_bit(n, p->bitmap);
-		*++paa = &p->ary[n];
-		p = p->ary[n];
-		shift -= IDR_BITS;
-	}
-	n = id & IDR_MASK;
-	if (likely(p != NULL && test_bit(n, p->bitmap))) {
-		__clear_bit(n, p->bitmap);
-		RCU_INIT_POINTER(p->ary[n], NULL);
-		to_free = NULL;
-		while(*paa && ! --((**paa)->count)){
-			if (to_free)
-				free_layer(idp, to_free);
-			to_free = **paa;
-			**paa-- = NULL;
-		}
-		if (!*paa)
-			idp->layers = 0;
-		if (to_free)
-			free_layer(idp, to_free);
-	} else
-		idr_remove_warning(id);
-}
-
-/**
- * idr_remove - remove the given id and free its slot
- * @idp: idr handle
- * @id: unique key
- */
-void idr_remove(struct idr *idp, int id)
-{
-	struct idr_layer *p;
-	struct idr_layer *to_free;
-
-	if (id < 0)
-		return;
-
-	if (id > idr_max(idp->layers)) {
-		idr_remove_warning(id);
-		return;
-	}
-
-	sub_remove(idp, (idp->layers - 1) * IDR_BITS, id);
-	if (idp->top && idp->top->count == 1 && (idp->layers > 1) &&
-	    idp->top->ary[0]) {
-		/*
-		 * Single child at leftmost slot: we can shrink the tree.
-		 * This level is not needed anymore since when layers are
-		 * inserted, they are inserted at the top of the existing
-		 * tree.
-		 */
-		to_free = idp->top;
-		p = idp->top->ary[0];
-		rcu_assign_pointer(idp->top, p);
-		--idp->layers;
-		to_free->count = 0;
-		bitmap_clear(to_free->bitmap, 0, IDR_SIZE);
-		free_layer(idp, to_free);
-	}
-}
-EXPORT_SYMBOL(idr_remove);
-
-static void __idr_remove_all(struct idr *idp)
-{
-	int n, id, max;
-	int bt_mask;
-	struct idr_layer *p;
-	struct idr_layer *pa[MAX_IDR_LEVEL + 1];
-	struct idr_layer **paa = &pa[0];
-
-	n = idp->layers * IDR_BITS;
-	*paa = idp->top;
-	RCU_INIT_POINTER(idp->top, NULL);
-	max = idr_max(idp->layers);
-
-	id = 0;
-	while (id >= 0 && id <= max) {
-		p = *paa;
-		while (n > IDR_BITS && p) {
-			n -= IDR_BITS;
-			p = p->ary[(id >> n) & IDR_MASK];
-			*++paa = p;
-		}
-
-		bt_mask = id;
-		id += 1 << n;
-		/* Get the highest bit that the above add changed from 0->1. */
-		while (n < fls(id ^ bt_mask)) {
-			if (*paa)
-				free_layer(idp, *paa);
-			n += IDR_BITS;
-			--paa;
-		}
-	}
-	idp->layers = 0;
-}
-
-/**
- * idr_destroy - release all cached layers within an idr tree
- * @idp: idr handle
- *
- * Free all id mappings and all idp_layers.  After this function, @idp is
- * completely unused and can be freed / recycled.  The caller is
- * responsible for ensuring that no one else accesses @idp during or after
- * idr_destroy().
- *
- * A typical clean-up sequence for objects stored in an idr tree will use
- * idr_for_each() to free all objects, if necessary, then idr_destroy() to
- * free up the id mappings and cached idr_layers.
- */
-void idr_destroy(struct idr *idp)
-{
-	__idr_remove_all(idp);
-
-	while (idp->id_free_cnt) {
-		struct idr_layer *p = get_from_free_list(idp);
-		kmem_cache_free(idr_layer_cache, p);
-	}
-}
-EXPORT_SYMBOL(idr_destroy);
-
-void *idr_find_slowpath(struct idr *idp, int id)
-{
-	int n;
-	struct idr_layer *p;
-
-	if (id < 0)
-		return NULL;
-
-	p = rcu_dereference_raw(idp->top);
-	if (!p)
-		return NULL;
-	n = (p->layer+1) * IDR_BITS;
-
-	if (id > idr_max(p->layer + 1))
-		return NULL;
-	BUG_ON(n == 0);
-
-	while (n > 0 && p) {
-		n -= IDR_BITS;
-		BUG_ON(n != p->layer*IDR_BITS);
-		p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
-	}
-	return((void *)p);
-}
-EXPORT_SYMBOL(idr_find_slowpath);
-
-/**
- * idr_for_each - iterate through all stored pointers
- * @idp: idr handle
- * @fn: function to be called for each pointer
- * @data: data passed back to callback function
- *
- * Iterate over the pointers registered with the given idr.  The
- * callback function will be called for each pointer currently
- * registered, passing the id, the pointer and the data pointer passed
- * to this function.  It is not safe to modify the idr tree while in
- * the callback, so functions such as idr_get_new and idr_remove are
- * not allowed.
- *
- * We check the return of @fn each time. If it returns anything other
- * than %0, we break out and return that value.
- *
- * The caller must serialize idr_for_each() vs idr_get_new() and idr_remove().
- */
-int idr_for_each(struct idr *idp,
-		 int (*fn)(int id, void *p, void *data), void *data)
-{
-	int n, id, max, error = 0;
-	struct idr_layer *p;
-	struct idr_layer *pa[MAX_IDR_LEVEL + 1];
-	struct idr_layer **paa = &pa[0];
-
-	n = idp->layers * IDR_BITS;
-	*paa = rcu_dereference_raw(idp->top);
-	max = idr_max(idp->layers);
-
-	id = 0;
-	while (id >= 0 && id <= max) {
-		p = *paa;
-		while (n > 0 && p) {
-			n -= IDR_BITS;
-			p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
-			*++paa = p;
-		}
-
-		if (p) {
-			error = fn(id, (void *)p, data);
-			if (error)
-				break;
-		}
-
-		id += 1 << n;
-		while (n < fls(id)) {
-			n += IDR_BITS;
-			--paa;
-		}
-	}
-
-	return error;
-}
-EXPORT_SYMBOL(idr_for_each);
-
-/**
- * idr_get_next - lookup next object of id to given id.
- * @idp: idr handle
- * @nextidp:  pointer to lookup key
- *
- * Returns pointer to registered object with id, which is next number to
- * given id. After being looked up, *@nextidp will be updated for the next
- * iteration.
- *
- * This function can be called under rcu_read_lock(), given that the leaf
- * pointers lifetimes are correctly managed.
- */
-void *idr_get_next(struct idr *idp, int *nextidp)
-{
-	struct idr_layer *p, *pa[MAX_IDR_LEVEL + 1];
-	struct idr_layer **paa = &pa[0];
-	int id = *nextidp;
-	int n, max;
-
-	/* find first ent */
-	p = *paa = rcu_dereference_raw(idp->top);
-	if (!p)
-		return NULL;
-	n = (p->layer + 1) * IDR_BITS;
-	max = idr_max(p->layer + 1);
-
-	while (id >= 0 && id <= max) {
-		p = *paa;
-		while (n > 0 && p) {
-			n -= IDR_BITS;
-			p = rcu_dereference_raw(p->ary[(id >> n) & IDR_MASK]);
-			*++paa = p;
-		}
-
-		if (p) {
-			*nextidp = id;
-			return p;
-		}
-
-		/*
-		 * Proceed to the next layer at the current level.  Unlike
-		 * idr_for_each(), @id isn't guaranteed to be aligned to
-		 * layer boundary at this point and adding 1 << n may
-		 * incorrectly skip IDs.  Make sure we jump to the
-		 * beginning of the next layer using round_up().
-		 */
-		id = round_up(id + 1, 1 << n);
-		while (n < fls(id)) {
-			n += IDR_BITS;
-			--paa;
-		}
-	}
-	return NULL;
-}
-EXPORT_SYMBOL(idr_get_next);
-
-
-/**
- * idr_replace - replace pointer for given id
- * @idp: idr handle
- * @ptr: pointer you want associated with the id
- * @id: lookup key
- *
- * Replace the pointer registered with an id and return the old value.
- * A %-ENOENT return indicates that @id was not found.
- * A %-EINVAL return indicates that @id was not within valid constraints.
- *
- * The caller must serialize with writers.
- */
-void *idr_replace(struct idr *idp, void *ptr, int id)
-{
-	int n;
-	struct idr_layer *p, *old_p;
-
-	if (id < 0)
-		return ERR_PTR(-EINVAL);
-
-	p = idp->top;
-	if (!p)
-		return ERR_PTR(-ENOENT);
-
-	if (id > idr_max(p->layer + 1))
-		return ERR_PTR(-ENOENT);
-
-	n = p->layer * IDR_BITS;
-	while ((n > 0) && p) {
-		p = p->ary[(id >> n) & IDR_MASK];
-		n -= IDR_BITS;
-	}
-
-	n = id & IDR_MASK;
-	if (unlikely(p == NULL || !test_bit(n, p->bitmap)))
-		return ERR_PTR(-ENOENT);
-
-	old_p = p->ary[n];
-	rcu_assign_pointer(p->ary[n], ptr);
-
-	return old_p;
-}
-EXPORT_SYMBOL(idr_replace);
-
-void __init idr_init_cache(void)
-{
-	idr_layer_cache = kmem_cache_create("idr_layer_cache",
-				sizeof(struct idr_layer), 0, SLAB_PANIC, NULL);
-}
-
-/**
- * idr_init - initialize idr handle
- * @idp:	idr handle
- *
- * This function is use to set up the handle (@idp) that you will pass
- * to the rest of the functions.
- */
-void idr_init(struct idr *idp)
-{
-	memset(idp, 0, sizeof(struct idr));
-	spin_lock_init(&idp->lock);
-}
-EXPORT_SYMBOL(idr_init);
-
-static int idr_has_entry(int id, void *p, void *data)
-{
-	return 1;
-}
-
-bool idr_is_empty(struct idr *idp)
-{
-	return !idr_for_each(idp, idr_has_entry, NULL);
-}
-EXPORT_SYMBOL(idr_is_empty);
-
-/**
- * DOC: IDA description
- * IDA - IDR based ID allocator
- *
- * This is id allocator without id -> pointer translation.  Memory
- * usage is much lower than full blown idr because each id only
- * occupies a bit.  ida uses a custom leaf node which contains
- * IDA_BITMAP_BITS slots.
- *
- * 2007-04-25  written by Tejun Heo <htejun@gmail.com>
- */
-
-static void free_bitmap(struct ida *ida, struct ida_bitmap *bitmap)
-{
-	unsigned long flags;
-
-	if (!ida->free_bitmap) {
-		spin_lock_irqsave(&ida->idr.lock, flags);
-		if (!ida->free_bitmap) {
-			ida->free_bitmap = bitmap;
-			bitmap = NULL;
-		}
-		spin_unlock_irqrestore(&ida->idr.lock, flags);
-	}
-
-	kfree(bitmap);
-}
-
-/**
- * ida_pre_get - reserve resources for ida allocation
- * @ida:	ida handle
- * @gfp_mask:	memory allocation flag
- *
- * This function should be called prior to locking and calling the
- * following function.  It preallocates enough memory to satisfy the
- * worst possible allocation.
- *
- * If the system is REALLY out of memory this function returns %0,
- * otherwise %1.
- */
-int ida_pre_get(struct ida *ida, gfp_t gfp_mask)
-{
-	/* allocate idr_layers */
-	if (!__idr_pre_get(&ida->idr, gfp_mask))
-		return 0;
-
-	/* allocate free_bitmap */
-	if (!ida->free_bitmap) {
-		struct ida_bitmap *bitmap;
-
-		bitmap = kmalloc(sizeof(struct ida_bitmap), gfp_mask);
-		if (!bitmap)
-			return 0;
-
-		free_bitmap(ida, bitmap);
-	}
-
-	return 1;
-}
-EXPORT_SYMBOL(ida_pre_get);
-
-/**
- * ida_get_new_above - allocate new ID above or equal to a start id
- * @ida:	ida handle
- * @starting_id: id to start search at
- * @p_id:	pointer to the allocated handle
- *
- * Allocate new ID above or equal to @starting_id.  It should be called
- * with any required locks.
- *
- * If memory is required, it will return %-EAGAIN, you should unlock
- * and go back to the ida_pre_get() call.  If the ida is full, it will
- * return %-ENOSPC.
- *
- * @p_id returns a value in the range @starting_id ... %0x7fffffff.
- */
-int ida_get_new_above(struct ida *ida, int starting_id, int *p_id)
-{
-	struct idr_layer *pa[MAX_IDR_LEVEL + 1];
-	struct ida_bitmap *bitmap;
-	unsigned long flags;
-	int idr_id = starting_id / IDA_BITMAP_BITS;
-	int offset = starting_id % IDA_BITMAP_BITS;
-	int t, id;
-
- restart:
-	/* get vacant slot */
-	t = idr_get_empty_slot(&ida->idr, idr_id, pa, 0, &ida->idr);
-	if (t < 0)
-		return t == -ENOMEM ? -EAGAIN : t;
-
-	if (t * IDA_BITMAP_BITS >= MAX_IDR_BIT)
-		return -ENOSPC;
-
-	if (t != idr_id)
-		offset = 0;
-	idr_id = t;
-
-	/* if bitmap isn't there, create a new one */
-	bitmap = (void *)pa[0]->ary[idr_id & IDR_MASK];
-	if (!bitmap) {
-		spin_lock_irqsave(&ida->idr.lock, flags);
-		bitmap = ida->free_bitmap;
-		ida->free_bitmap = NULL;
-		spin_unlock_irqrestore(&ida->idr.lock, flags);
-
-		if (!bitmap)
-			return -EAGAIN;
-
-		memset(bitmap, 0, sizeof(struct ida_bitmap));
-		rcu_assign_pointer(pa[0]->ary[idr_id & IDR_MASK],
-				(void *)bitmap);
-		pa[0]->count++;
-	}
-
-	/* lookup for empty slot */
-	t = find_next_zero_bit(bitmap->bitmap, IDA_BITMAP_BITS, offset);
-	if (t == IDA_BITMAP_BITS) {
-		/* no empty slot after offset, continue to the next chunk */
-		idr_id++;
-		offset = 0;
-		goto restart;
-	}
-
-	id = idr_id * IDA_BITMAP_BITS + t;
-	if (id >= MAX_IDR_BIT)
-		return -ENOSPC;
-
-	__set_bit(t, bitmap->bitmap);
-	if (++bitmap->nr_busy == IDA_BITMAP_BITS)
-		idr_mark_full(pa, idr_id);
-
-	*p_id = id;
-
-	/* Each leaf node can handle nearly a thousand slots and the
-	 * whole idea of ida is to have small memory foot print.
-	 * Throw away extra resources one by one after each successful
-	 * allocation.
-	 */
-	if (ida->idr.id_free_cnt || ida->free_bitmap) {
-		struct idr_layer *p = get_from_free_list(&ida->idr);
-		if (p)
-			kmem_cache_free(idr_layer_cache, p);
-	}
-
-	return 0;
-}
-EXPORT_SYMBOL(ida_get_new_above);
-
-/**
- * ida_remove - remove the given ID
- * @ida:	ida handle
- * @id:		ID to free
- */
-void ida_remove(struct ida *ida, int id)
-{
-	struct idr_layer *p = ida->idr.top;
-	int shift = (ida->idr.layers - 1) * IDR_BITS;
-	int idr_id = id / IDA_BITMAP_BITS;
-	int offset = id % IDA_BITMAP_BITS;
-	int n;
-	struct ida_bitmap *bitmap;
-
-	if (idr_id > idr_max(ida->idr.layers))
-		goto err;
-
-	/* clear full bits while looking up the leaf idr_layer */
-	while ((shift > 0) && p) {
-		n = (idr_id >> shift) & IDR_MASK;
-		__clear_bit(n, p->bitmap);
-		p = p->ary[n];
-		shift -= IDR_BITS;
-	}
-
-	if (p == NULL)
-		goto err;
-
-	n = idr_id & IDR_MASK;
-	__clear_bit(n, p->bitmap);
-
-	bitmap = (void *)p->ary[n];
-	if (!bitmap || !test_bit(offset, bitmap->bitmap))
-		goto err;
-
-	/* update bitmap and remove it if empty */
-	__clear_bit(offset, bitmap->bitmap);
-	if (--bitmap->nr_busy == 0) {
-		__set_bit(n, p->bitmap);	/* to please idr_remove() */
-		idr_remove(&ida->idr, idr_id);
-		free_bitmap(ida, bitmap);
-	}
-
-	return;
-
- err:
-	WARN(1, "ida_remove called for id=%d which is not allocated.\n", id);
-}
-EXPORT_SYMBOL(ida_remove);
-
-/**
- * ida_destroy - release all cached layers within an ida tree
- * @ida:		ida handle
- */
-void ida_destroy(struct ida *ida)
-{
-	idr_destroy(&ida->idr);
-	kfree(ida->free_bitmap);
-}
-EXPORT_SYMBOL(ida_destroy);
-
 /**
  * ida_simple_get - get a new id.
  * @ida: the (initialized) ida.
@@ -1130,18 +70,3 @@ void ida_simple_remove(struct ida *ida, unsigned int id)
 	spin_unlock_irqrestore(&simple_ida_lock, flags);
 }
 EXPORT_SYMBOL(ida_simple_remove);
-
-/**
- * ida_init - initialize ida handle
- * @ida:	ida handle
- *
- * This function is use to set up the handle (@ida) that you will pass
- * to the rest of the functions.
- */
-void ida_init(struct ida *ida)
-{
-	memset(ida, 0, sizeof(struct ida));
-	idr_init(&ida->idr);
-
-}
-EXPORT_SYMBOL(ida_init);
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index e063ca2..3160de3 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -24,19 +24,20 @@
 
 #include <linux/bitmap.h>
 #include <linux/bitops.h>
+#include <linux/cpu.h>
 #include <linux/errno.h>
+#include <linux/export.h>
+#include <linux/idr.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
-#include <linux/export.h>
-#include <linux/radix-tree.h>
-#include <linux/percpu.h>
-#include <linux/slab.h>
 #include <linux/kmemleak.h>
 #include <linux/notifier.h>
-#include <linux/cpu.h>
-#include <linux/string.h>
-#include <linux/rcupdate.h>
+#include <linux/percpu.h>
 #include <linux/preempt.h>		/* in_interrupt() */
+#include <linux/radix-tree.h>
+#include <linux/rcupdate.h>
+#include <linux/slab.h>
+#include <linux/string.h>
 
 
 /* Number of nodes in fully populated tree of given height */
@@ -61,6 +62,15 @@ static struct kmem_cache *radix_tree_node_cachep;
 #define RADIX_TREE_PRELOAD_SIZE (RADIX_TREE_MAX_PATH * 2 - 1)
 
 /*
+ * The IDR does not have to be as high since it can only store a 31-bit integer
+ * at its maximum height
+ */
+#define IDR_INDEX_BITS		(8 /* CHAR_BIT */ * sizeof(int) - 1)
+#define IDR_MAX_PATH		(DIV_ROUND_UP(IDR_INDEX_BITS, \
+						RADIX_TREE_MAP_SHIFT))
+#define IDR_PRELOAD_SIZE	(IDR_MAX_PATH * 2 - 1)
+
+/*
  * Per-cpu pool of preloaded nodes
  */
 struct radix_tree_preload {
@@ -148,27 +158,38 @@ static inline int tag_get(struct radix_tree_node *node, unsigned int tag,
 
 static inline void root_tag_set(struct radix_tree_root *root, unsigned int tag)
 {
-	root->gfp_mask |= (__force gfp_t)(1 << (tag + __GFP_BITS_SHIFT));
+	root->gfp_mask |= (__force gfp_t)(1 << (tag + ROOT_TAG_SHIFT));
 }
 
 static inline void root_tag_clear(struct radix_tree_root *root, unsigned tag)
 {
-	root->gfp_mask &= (__force gfp_t)~(1 << (tag + __GFP_BITS_SHIFT));
+	root->gfp_mask &= (__force gfp_t)~(1 << (tag + ROOT_TAG_SHIFT));
 }
 
 static inline void root_tag_clear_all(struct radix_tree_root *root)
 {
-	root->gfp_mask &= __GFP_BITS_MASK;
+	root->gfp_mask &= (1 << ROOT_TAG_SHIFT) - 1;
 }
 
 static inline int root_tag_get(struct radix_tree_root *root, unsigned int tag)
 {
-	return (__force int)root->gfp_mask & (1 << (tag + __GFP_BITS_SHIFT));
+	return (__force int)root->gfp_mask & (1 << (tag + ROOT_TAG_SHIFT));
 }
 
 static inline unsigned root_tags_get(struct radix_tree_root *root)
 {
-	return (__force unsigned)root->gfp_mask >> __GFP_BITS_SHIFT;
+	return (__force unsigned)root->gfp_mask >> ROOT_TAG_SHIFT;
+}
+
+/*
+ * IDRs do not expose the tagging functionality of the radix tree to their
+ * users.  Reuse tag 0 to track whether a node has free space below it.
+ */
+#define IDR_FREE	0
+
+static inline bool is_idr(struct radix_tree_root *root)
+{
+	return (__force unsigned)root->gfp_mask & ROOT_IS_IDR;
 }
 
 /*
@@ -238,6 +259,13 @@ static inline unsigned long node_maxindex(struct radix_tree_node *node)
 	return shift_maxindex(node->shift);
 }
 
+static unsigned long next_index(unsigned long index,
+				struct radix_tree_node *node,
+				unsigned long offset)
+{
+	return (index & ~node_maxindex(node)) + (offset << node->shift);
+}
+
 #ifndef __KERNEL__
 static void dump_node(struct radix_tree_node *node, unsigned long index)
 {
@@ -276,11 +304,47 @@ static void radix_tree_dump(struct radix_tree_root *root)
 {
 	pr_debug("radix root: %p rnode %p tags %x\n",
 			root, root->rnode,
-			root->gfp_mask >> __GFP_BITS_SHIFT);
+			root->gfp_mask >> ROOT_TAG_SHIFT);
 	if (!radix_tree_is_internal_node(root->rnode))
 		return;
 	dump_node(entry_to_node(root->rnode), 0);
 }
+
+static void dump_ida_node(void *entry, unsigned long index)
+{
+	unsigned long i;
+
+	if (!entry)
+		return;
+
+	if (radix_tree_is_internal_node(entry)) {
+		struct radix_tree_node *node = entry_to_node(entry);
+
+		pr_debug("ida node: %p offset %d indices %lu-%lu parent %p free %lx shift %d count %d\n",
+			node, node->offset, index, index | node_maxindex(node),
+			node->parent, node->tags[0][0], node->shift,
+			node->count);
+		for (i = 0; i < RADIX_TREE_MAP_SIZE; i++)
+			dump_ida_node(node->slots[i],
+					index | (i << node->shift));
+	} else {
+		struct ida_bitmap *bitmap = entry;
+
+		pr_debug("ida btmp: %p index %lu data", bitmap, index);
+		for (i = 0; i < IDA_BITMAP_LONGS; i++)
+			pr_cont(" %lx", bitmap->bitmap[i]);
+		pr_cont("\n");
+	}
+}
+
+static void ida_dump(struct ida *ida)
+{
+	struct radix_tree_root *root = &ida->ida_rt;
+	pr_debug("ida: %p %p free %d bitmap %p\n", ida, root->rnode,
+				root->gfp_mask >> ROOT_TAG_SHIFT,
+				ida->free_bitmap);
+	dump_ida_node(root->rnode, 0);
+}
 #endif
 
 /*
@@ -288,10 +352,9 @@ static void radix_tree_dump(struct radix_tree_root *root)
  * that the caller has pinned this thread of control to the current CPU.
  */
 static struct radix_tree_node *
-radix_tree_node_alloc(struct radix_tree_root *root)
+radix_tree_node_alloc(gfp_t gfp_mask)
 {
 	struct radix_tree_node *ret = NULL;
-	gfp_t gfp_mask = root_gfp_mask(root);
 
 	/*
 	 * Preload code isn't irq safe and it doesn't make sense to use
@@ -521,7 +584,7 @@ static unsigned radix_tree_load_root(struct radix_tree_root *root,
 /*
  *	Extend a radix tree so it can store key @index.
  */
-static int radix_tree_extend(struct radix_tree_root *root,
+static int radix_tree_extend(struct radix_tree_root *root, gfp_t gfp_mask,
 				unsigned long index, unsigned int shift)
 {
 	struct radix_tree_node *slot;
@@ -538,15 +601,22 @@ static int radix_tree_extend(struct radix_tree_root *root,
 		goto out;
 
 	do {
-		struct radix_tree_node *node = radix_tree_node_alloc(root);
+		struct radix_tree_node *node = radix_tree_node_alloc(gfp_mask);
 
 		if (!node)
 			return -ENOMEM;
 
-		/* Propagate the aggregated tag info into the new root */
-		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++) {
-			if (root_tag_get(root, tag))
-				tag_set(node, tag, 0);
+		if (is_idr(root)) {
+			all_tag_set(node, IDR_FREE);
+			if (!root_tag_get(root, IDR_FREE))
+				tag_clear(node, IDR_FREE, 0);
+			root_tag_set(root, IDR_FREE);
+		} else {
+			/* Propagate the aggregated tag info to the new child */
+			for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++) {
+				if (root_tag_get(root, tag))
+					tag_set(node, tag, 0);
+			}
 		}
 
 		BUG_ON(shift > BITS_PER_LONG);
@@ -565,26 +635,9 @@ static int radix_tree_extend(struct radix_tree_root *root,
 	return maxshift + RADIX_TREE_MAP_SHIFT;
 }
 
-/**
- *	__radix_tree_create	-	create a slot in a radix tree
- *	@root:		radix tree root
- *	@index:		index key
- *	@order:		index occupies 2^order aligned slots
- *	@nodep:		returns node
- *	@slotp:		returns slot
- *
- *	Create, if necessary, and return the node and slot for an item
- *	at position @index in the radix tree @root.
- *
- *	Until there is more than one item in the tree, no nodes are
- *	allocated and @root->rnode is used as a direct slot instead of
- *	pointing to a node, in which case *@nodep will be NULL.
- *
- *	Returns -ENOMEM, or 0 for success.
- */
-int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
-			unsigned order, struct radix_tree_node **nodep,
-			void ***slotp)
+static int _radix_tree_create(struct radix_tree_root *root, gfp_t gfp_mask,
+			unsigned long index, unsigned int order,
+			struct radix_tree_node **nodep, void ***slotp)
 {
 	struct radix_tree_node *node = NULL, *child;
 	void **slot = (void **)&root->rnode;
@@ -598,7 +651,7 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 	if (order > 0 && max == ((1UL << order) - 1))
 		max++;
 	if (max > maxindex) {
-		int error = radix_tree_extend(root, max, shift);
+		int error = radix_tree_extend(root, gfp_mask, max, shift);
 		if (error < 0)
 			return error;
 		shift = error;
@@ -609,7 +662,7 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 		shift -= RADIX_TREE_MAP_SHIFT;
 		if (child == NULL) {
 			/* Have to add a child node.  */
-			child = radix_tree_node_alloc(root);
+			child = radix_tree_node_alloc(gfp_mask);
 			if (!child)
 				return -ENOMEM;
 			child->shift = shift;
@@ -635,7 +688,6 @@ int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
 	return 0;
 }
 
-#ifdef CONFIG_RADIX_TREE_MULTIORDER
 /*
  * Free any nodes below this node.  The tree is presumed to not need
  * shrinking, and any user data in the tree is presumed to not need a
@@ -670,6 +722,7 @@ static void radix_tree_free_nodes(struct radix_tree_node *node)
 	}
 }
 
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
 static inline int insert_entries(struct radix_tree_node *node, void **slot,
 				void *ptr, unsigned order, bool replace)
 {
@@ -741,6 +794,31 @@ static inline int insert_entries(struct radix_tree_node *node, void **slot,
 #endif
 
 /**
+ *	__radix_tree_create	-	create a slot in a radix tree
+ *	@root:		radix tree root
+ *	@index:		index key
+ *	@order:		index occupies 2^order aligned slots
+ *	@nodep:		returns node
+ *	@slotp:		returns slot
+ *
+ *	Create, if necessary, and return the node and slot for an item
+ *	at position @index in the radix tree @root.
+ *
+ *	Until there is more than one item in the tree, no nodes are
+ *	allocated and @root->rnode is used as a direct slot instead of
+ *	pointing to a node, in which case *@nodep will be NULL.
+ *
+ *	Returns -ENOMEM, or 0 for success.
+ */
+int __radix_tree_create(struct radix_tree_root *root, unsigned long index,
+			unsigned order, struct radix_tree_node **nodep,
+			void ***slotp)
+{
+	return _radix_tree_create(root, root_gfp_mask(root), index, order,
+					nodep, slotp);
+}
+
+/**
  *	__radix_tree_insert    -    insert into a radix tree
  *	@root:		radix tree root
  *	@index:		index key
@@ -891,6 +969,7 @@ int radix_tree_split(struct radix_tree_root *root, unsigned long index,
 	void **slot;
 	unsigned int offset, end;
 	unsigned n, tag, tags = 0;
+	gfp_t gfp = root_gfp_mask(root);
 
 	if (!__radix_tree_lookup(root, index, &parent, &slot))
 		return -ENOENT;
@@ -926,7 +1005,7 @@ int radix_tree_split(struct radix_tree_root *root, unsigned long index,
 
 	for (;;) {
 		if (node->shift > order) {
-			child = radix_tree_node_alloc(root);
+			child = radix_tree_node_alloc(gfp);
 			if (!child)
 				goto nomem;
 			child->shift = node->shift - RADIX_TREE_MAP_SHIFT;
@@ -1571,6 +1650,8 @@ static inline bool radix_tree_shrink(struct radix_tree_root *root)
 		 * one (root->rnode) as far as dependent read barriers go.
 		 */
 		root->rnode = child;
+		if (is_idr(root) && !tag_get(node, IDR_FREE, 0))
+			root_tag_clear(root, IDR_FREE);
 
 		/*
 		 * We have a dilemma here. The node's slot[0] must not be
@@ -1630,7 +1711,12 @@ bool __radix_tree_delete_node(struct radix_tree_root *root,
 			parent->slots[node->offset] = NULL;
 			parent->count--;
 		} else {
-			root_tag_clear_all(root);
+			/*
+			 * Shouldn't the tags already have all been cleared
+			 * by the caller?
+			 */
+			if (!is_idr(root))
+				root_tag_clear_all(root);
 			root->rnode = NULL;
 		}
 
@@ -1643,6 +1729,17 @@ bool __radix_tree_delete_node(struct radix_tree_root *root,
 	return deleted;
 }
 
+static void radix_tree_iter_delete(struct radix_tree_root *root,
+					struct radix_tree_iter *iter)
+{
+	unsigned offset = (iter->index >> iter->shift) & RADIX_TREE_MAP_MASK;
+	struct radix_tree_node *node = iter->node;
+
+	node->slots[offset] = NULL;
+	node->count--;
+	__radix_tree_delete_node(root, node);
+}
+
 static inline void delete_sibling_entries(struct radix_tree_node *node,
 					void *ptr, unsigned offset)
 {
@@ -1665,7 +1762,7 @@ static inline void delete_sibling_entries(struct radix_tree_node *node,
  *
  *	Remove @item at @index from the radix tree rooted at @root.
  *
- *	Returns the address of the deleted item, or NULL if it was not present
+ *	Returns the value of the deleted item, or NULL if it was not present
  *	or the entry at the given @index was not @item.
  */
 void *radix_tree_delete_item(struct radix_tree_root *root,
@@ -1685,16 +1782,21 @@ void *radix_tree_delete_item(struct radix_tree_root *root,
 		return NULL;
 
 	if (!node) {
-		root_tag_clear_all(root);
+		if (is_idr(root))
+			root_tag_set(root, IDR_FREE);
+		else
+			root_tag_clear_all(root);
 		root->rnode = NULL;
 		return entry;
 	}
 
 	offset = get_slot_offset(node, slot);
 
-	/* Clear all tags associated with the item to be deleted.  */
-	for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
-		node_tag_clear(root, node, tag, offset);
+	if (is_idr(root))
+		node_tag_set(root, node, IDR_FREE, offset);
+	else
+		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
+			node_tag_clear(root, node, tag, offset);
 
 	delete_sibling_entries(node, node_to_entry(slot), offset);
 	node->slots[offset] = NULL;
@@ -1713,7 +1815,7 @@ EXPORT_SYMBOL(radix_tree_delete_item);
  *
  *	Remove the item at @index from the radix tree rooted at @root.
  *
- *	Returns the address of the deleted item, or NULL if it was not present.
+ *	Returns the value of the deleted item, or NULL if it was not present.
  */
 void *radix_tree_delete(struct radix_tree_root *root, unsigned long index)
 {
@@ -1730,8 +1832,7 @@ void radix_tree_clear_tags(struct radix_tree_root *root,
 		for (tag = 0; tag < RADIX_TREE_MAX_TAGS; tag++)
 			node_tag_clear(root, node, tag, offset);
 	} else {
-		/* Clear root node tags */
-		root->gfp_mask &= __GFP_BITS_MASK;
+		root_tag_clear_all(root);
 	}
 }
 
@@ -1746,6 +1847,421 @@ int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag)
 }
 EXPORT_SYMBOL(radix_tree_tagged);
 
+/**
+ * idr_preload - preload for idr_alloc()
+ * @gfp_mask: allocation mask to use for preloading
+ *
+ * Preallocate memory to use for the next call to idr_alloc().  This function
+ * returns with preemption disabled.  It will be enabled by idr_preload_end().
+ */
+void idr_preload(gfp_t gfp_mask)
+{
+	__radix_tree_preload(gfp_mask, IDR_PRELOAD_SIZE);
+}
+EXPORT_SYMBOL(idr_preload);
+
+static int __idr_get_empty(struct radix_tree_root *root, gfp_t gfp,
+				unsigned long start, int end,
+				struct radix_tree_node **nodep, void ***slotp)
+{
+	struct radix_tree_node *node = NULL, *child;
+	void **slot = (void **)&root->rnode;
+	unsigned long maxindex;
+	unsigned long max = end > 0 ? end - 1 : INT_MAX;
+	unsigned int shift, offset = 0;
+
+ grow:
+	shift = radix_tree_load_root(root, &child, &maxindex);
+	if (!radix_tree_tagged(root, IDR_FREE))
+		start = max(start, maxindex + 1);
+	if (start > max)
+		return -ENOSPC;
+
+	if (start > maxindex) {
+		int error = radix_tree_extend(root, gfp, start, shift);
+		if (error < 0)
+			return error;
+		shift = error;
+		child = root->rnode;
+	}
+
+	while (shift) {
+		shift -= RADIX_TREE_MAP_SHIFT;
+		if (child == NULL) {
+			/* Have to add a child node.  */
+			child = radix_tree_node_alloc(gfp);
+			if (!child)
+				return -ENOMEM;
+			child->shift = shift;
+			child->offset = offset;
+			child->parent = node;
+			all_tag_set(child, IDR_FREE);
+			rcu_assign_pointer(*slot, node_to_entry(child));
+			if (node)
+				node->count++;
+		} else if (!radix_tree_is_internal_node(child))
+			break;
+
+		node = entry_to_node(child);
+		offset = radix_tree_descend(node, &child, start);
+		if (!tag_get(node, IDR_FREE, offset)) {
+			offset = radix_tree_find_next_bit(node, IDR_FREE,
+							offset + 1);
+			start = next_index(start, node, offset);
+			if (start > max)
+				return -ENOSPC;
+			while (offset == RADIX_TREE_MAP_SIZE) {
+				offset = node->offset + 1;
+				node = node->parent;
+				if (!node)
+					goto grow;
+				shift = node->shift;
+			}
+			child = node->slots[offset];
+		}
+		slot = &node->slots[offset];
+	}
+
+	*nodep = node;
+	*slotp = slot;
+	return start;
+}
+
+/**
+ * idr_alloc - allocate an id
+ * @idr: idr handle
+ * @ptr: pointer to be associated with the new id
+ * @start: the minimum id (inclusive)
+ * @end: the maximum id (exclusive)
+ * @gfp: memory allocation flags
+ *
+ * Allocates an unused ID in the range [start, end).  Returns -ENOSPC
+ * if there are no unused IDs in that range.
+ */
+int idr_alloc(struct idr *idr, void *ptr, int start, int end, gfp_t gfp)
+{
+	struct radix_tree_node *node;
+	void **slot;
+	int id;
+
+	if (WARN_ON_ONCE(start < 0))
+		return -EINVAL;
+
+	id = __idr_get_empty(&idr->idr_rt, gfp, start, end, &node, &slot);
+	if (id < 0)
+		return id;
+
+	BUG_ON(radix_tree_is_internal_node(ptr));
+
+	rcu_assign_pointer(*slot, ptr);
+	if (node)
+		node->count++;
+	node_tag_clear(&idr->idr_rt, node, IDR_FREE,
+						get_slot_offset(node, slot));
+	return id;
+}
+EXPORT_SYMBOL(idr_alloc);
+
+/**
+ * idr_alloc_cyclic - allocate new idr entry in a cyclical fashion
+ * @idr: idr handle
+ * @ptr: pointer to be associated with the new id
+ * @start: the minimum id (inclusive)
+ * @end: the maximum id (exclusive)
+ * @gfp: memory allocation flags
+ *
+ * Allocates an ID larger than the last ID allocated if one is available.
+ * If not, it will attempt to allocate the smallest ID that is larger or
+ * equal to @start.
+ */
+int idr_alloc_cyclic(struct idr *idr, void *ptr, int start, int end, gfp_t gfp)
+{
+	int id, curr = idr->idr_next;
+
+	if (curr < start)
+		curr = start;
+
+	id = idr_alloc(idr, ptr, curr, end, gfp);
+	if ((id == -ENOSPC) && (curr > start))
+		id = idr_alloc(idr, ptr, start, curr, gfp);
+
+	if (id >= 0)
+		idr->idr_next = id + 1U;
+
+	return id;
+}
+EXPORT_SYMBOL(idr_alloc_cyclic);
+
+/**
+ * idr_for_each - iterate through all stored pointers
+ * @idr: idr handle
+ * @fn: function to be called for each pointer
+ * @data: data passed to callback function
+ *
+ * The callback function will be called for each entry in @idr, passing
+ * the id, the pointer and the data pointer passed to this function.
+ *
+ * If @fn returns anything other than %0, the iteration stops and that
+ * value is returned from this function.
+ *
+ * idr_for_each() can be called concurrently with idr_get_new() and
+ * idr_remove() if protected by RCU.  Newly added entries may not be
+ * seen and deleted entries may be seen, but adding and removing entries
+ * will not cause other entries to be skipped, nor spurious ones to be seen.
+ */
+int idr_for_each(struct idr *idr,
+		int (*fn)(int id, void *p, void *data), void *data)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+
+	radix_tree_for_each_slot(slot, &idr->idr_rt, &iter, 0) {
+		int ret = fn(iter.index, *slot, data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(idr_for_each);
+
+/**
+ * idr_get_next - Find next populated entry
+ * @idr: idr handle
+ * @nextid: Pointer to lowest possible ID to return
+ *
+ * Returns the next populated entry in the tree with an ID greater than
+ * or equal to the value pointed to by @nextid.  On exit, @nextid is updated
+ * to the ID of the found value.  To use in a loop, the value pointed to by
+ * nextid must be incremented by the user.
+ */
+void *idr_get_next(struct idr *idr, int *nextid)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+
+	radix_tree_for_each_slot(slot, &idr->idr_rt, &iter, *nextid) {
+		*nextid = iter.index;
+		return *slot;
+	}
+
+	return NULL;
+}
+EXPORT_SYMBOL(idr_get_next);
+
+/**
+ * idr_replace - replace pointer for given id
+ * @idr: idr handle
+ * @ptr: New pointer to associate with the ID
+ * @id: Lookup key
+ *
+ * Replace the pointer registered with an id and return the old value.
+ * A %-ENOENT return indicates that @id was not found.
+ * A %-EINVAL return indicates that @id was not within valid constraints.
+ *
+ * This function can be called under the RCU read lock concurrently with
+ * idr_remove().
+ */
+void *idr_replace(struct idr *idr, void *ptr, int id)
+{
+	void **slot;
+	void *entry;
+
+	if (id < 0)
+		return ERR_PTR(-EINVAL);
+	if (!ptr || radix_tree_is_internal_node(ptr))
+		return ERR_PTR(-EINVAL);
+
+	entry = __radix_tree_lookup(&idr->idr_rt, id, NULL, &slot);
+
+	if (!entry)
+		return ERR_PTR(-ENOENT);
+
+	radix_tree_replace_slot(slot, ptr);
+
+	return entry;
+}
+EXPORT_SYMBOL(idr_replace);
+
+/**
+ * idr_destroy - release all internal memory from an IDR
+ * @idr: idr handle
+ *
+ * After this function is called, the IDR is empty, and may be reused or
+ * the data structure containing it may be freed.
+ *
+ * A typical clean-up sequence for objects stored in an idr tree will use
+ * idr_for_each() to free all objects, if necessary, then idr_destroy() to
+ * free the memory used to keep track of those objects.
+ */
+void idr_destroy(struct idr *idr)
+{
+	struct radix_tree_node **slot = &idr->idr_rt.rnode;
+	if (radix_tree_is_internal_node(*slot))
+		radix_tree_free_nodes(*slot);
+	*slot = NULL;
+	root_tag_set(&idr->idr_rt, IDR_FREE);
+}
+EXPORT_SYMBOL(idr_destroy);
+
+/**
+ * ida_pre_get - reserve resources for ida allocation
+ * @ida: ida handle
+ * @gfp: memory allocation flags
+ *
+ * This function should be called before calling ida_get_new_above().  If it
+ * is unable to allocate memory, it will return %0.  On success, it returns %1.
+ */
+int ida_pre_get(struct ida *ida, gfp_t gfp)
+{
+	struct ida_bitmap *bitmap;
+
+	idr_preload(gfp);
+	idr_preload_end();
+
+	if (!ida->free_bitmap) {
+		bitmap = kmalloc(sizeof(struct ida_bitmap), gfp);
+		if (!bitmap)
+			return 0;
+		bitmap = xchg(&ida->free_bitmap, bitmap);
+		kfree(bitmap);
+	}
+
+	return 1;
+}
+EXPORT_SYMBOL(ida_pre_get);
+
+/**
+ * ida_get_new_above - allocate new ID above or equal to a start id
+ * @ida: ida handle
+ * @starting_id: id to start search at
+ * @p_id: pointer to the allocated handle
+ *
+ * Allocate new ID above or equal to @starting_id.  It should be called
+ * with any required locks.
+ *
+ * If memory is required, it will return %-EAGAIN, you should unlock
+ * and go back to the ida_pre_get() call.  If the ida is full, it will
+ * return %-ENOSPC.
+ *
+ * @p_id returns a value in the range @starting_id ... %0x7fffffff.
+ */
+int ida_get_new_above(struct ida *ida, int start, int *id)
+{
+	struct radix_tree_root *root = &ida->ida_rt;
+	void **slot = (void **)&root->rnode;
+	struct radix_tree_node *node;
+	struct ida_bitmap *bitmap = NULL;
+	unsigned long index;
+	unsigned bit, offset = 0;
+
+	index = start / IDA_BITMAP_BITS;
+	bit = start % IDA_BITMAP_BITS;
+
+ restart:
+	index = __idr_get_empty(root, GFP_ATOMIC, index, INT_MAX, &node, &slot);
+	if (index > INT_MAX)
+		return index;
+
+	index *= IDA_BITMAP_BITS;
+	if (index > INT_MAX)
+		return -ENOSPC;
+
+	if (index > start)
+		bit = 0;
+	offset = get_slot_offset(node, slot);
+
+	bitmap = *slot;
+	if (bitmap) {
+		bit = find_next_zero_bit(bitmap->bitmap, IDA_BITMAP_BITS, bit);
+		index += bit;
+		if (index > INT_MAX)
+			return -ENOSPC;
+		if (bit == IDA_BITMAP_BITS) {
+			index /= IDA_BITMAP_BITS;
+			goto restart;
+		}
+		__set_bit(bit, bitmap->bitmap);
+		if (bitmap_full(bitmap->bitmap, IDA_BITMAP_BITS))
+			node_tag_clear(root, node, IDR_FREE, offset);
+		bitmap = xchg(&ida->free_bitmap, NULL);
+		kfree(bitmap);
+	} else {
+		index += bit;
+		bitmap = xchg(&ida->free_bitmap, NULL);
+		if (!bitmap)
+			return -EAGAIN;
+		memset(bitmap, 0, sizeof(*bitmap));
+		__set_bit(bit, bitmap->bitmap);
+		rcu_assign_pointer(*slot, bitmap);
+		if (node)
+			node->count++;
+	}
+
+	*id = index;
+	return 0;
+}
+EXPORT_SYMBOL(ida_get_new_above);
+
+/**
+ * ida_remove - Free the given ID
+ * @ida: ida handle
+ * @id: ID to free
+ *
+ * This function should not be called at the same time as ida_get_new_above().
+ */
+void ida_remove(struct ida *ida, int id)
+{
+	unsigned long index = id / IDA_BITMAP_BITS;
+	unsigned offset = id % IDA_BITMAP_BITS;
+	struct ida_bitmap *bitmap;
+	struct radix_tree_node *node;
+	void **slot;
+
+	bitmap = __radix_tree_lookup(&ida->ida_rt, index, &node, &slot);
+	if (!bitmap || !test_bit(offset, bitmap->bitmap))
+		goto err;
+
+	__clear_bit(offset, bitmap->bitmap);
+	node_tag_set(&ida->ida_rt, node, IDR_FREE, get_slot_offset(node, slot));
+	if (bitmap_empty(bitmap->bitmap, IDA_BITMAP_BITS)) {
+		*slot = NULL;
+		kfree(bitmap);
+		if (node) {
+			node->count--;
+			__radix_tree_delete_node(&ida->ida_rt, node);
+		}
+	}
+	return;
+ err:
+	WARN(1, "ida_remove called for id=%d which is not allocated.\n", id);
+}
+EXPORT_SYMBOL(ida_remove);
+
+/**
+ * ida_destroy - Free the contents of an ida
+ * @ida: ida handle
+ *
+ * Calling this function releases all resources associated with an IDA.  When
+ * this call returns, the IDA is empty and can be reused or freed.  The caller
+ * should not allow ida_remove() or ida_get_new_above() to be called at the
+ * same time.
+ */
+void ida_destroy(struct ida *ida)
+{
+	struct radix_tree_iter iter;
+	void **slot;
+
+	radix_tree_for_each_slot(slot, &ida->ida_rt, &iter, 0) {
+		struct ida_bitmap *bitmap = *slot;
+		kfree(bitmap);
+		radix_tree_iter_delete(&ida->ida_rt, &iter);
+	}
+
+	kfree(ida->free_bitmap);
+}
+EXPORT_SYMBOL(ida_destroy);
+
 static void
 radix_tree_node_ctor(void *arg)
 {
diff --git a/tools/testing/radix-tree/Makefile b/tools/testing/radix-tree/Makefile
index 3635e4d..5a616a3 100644
--- a/tools/testing/radix-tree/Makefile
+++ b/tools/testing/radix-tree/Makefile
@@ -3,7 +3,7 @@ CFLAGS += -I. -I../../include -g -O2 -Wall -D_LGPL_SOURCE
 LDFLAGS += -lpthread -lurcu
 TARGETS = main
 OFILES = main.o radix-tree.o linux.o test.o tag_check.o find_next_bit.o \
-	 regression1.o regression2.o regression3.o multiorder.o \
+	 regression1.o regression2.o regression3.o multiorder.o idr.o \
 	 iteration_check.o benchmark.o
 
 ifdef BENCHMARK
@@ -23,7 +23,8 @@ find_next_bit.o: ../../lib/find_bit.c
 
 $(OFILES): *.h */*.h \
 	../../include/linux/*.h \
-	../../../include/linux/radix-tree.h
+	../../../include/linux/radix-tree.h \
+	../../../include/linux/idr.h
 
 radix-tree.c: ../../../lib/radix-tree.c
 	sed -e 's/^static //' -e 's/__always_inline //' -e 's/inline //' < $< > $@
diff --git a/tools/testing/radix-tree/idr.c b/tools/testing/radix-tree/idr.c
new file mode 100644
index 0000000..0f9e7b1
--- /dev/null
+++ b/tools/testing/radix-tree/idr.c
@@ -0,0 +1,148 @@
+/*
+ * idr.c: Test the IDR API
+ * Copyright (c) 2016 Matthew Wilcox <willy@infradead.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope 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.
+ */
+#include <linux/idr.h>
+#include <linux/slab.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+
+#include "test.h"
+
+#define DUMMY_PTR	((void *)0x12)
+
+int item_idr_free(int id, void *p, void *data)
+{
+	struct item *item = p;
+	assert(item->index == id);
+	idr_remove(data, id);
+	free(p);
+
+	return 0;
+}
+
+void item_idr_remove(struct idr *idr, int id)
+{
+	struct item *item = idr_find(idr, id);
+	assert(item->index == id);
+	idr_remove(idr, id);
+	free(item);
+}
+
+void idr_alloc_test(void)
+{
+	unsigned long i;
+	DEFINE_IDR(idr);
+
+	assert(idr_alloc_cyclic(&idr, DUMMY_PTR, 0, 0x4000, GFP_KERNEL) == 0);
+	assert(idr_alloc_cyclic(&idr, DUMMY_PTR, 0x3ffd, 0x4000, GFP_KERNEL) == 0x3ffd);
+	idr_remove(&idr, 0x3ffd);
+	idr_remove(&idr, 0);
+
+	for (i = 0x3ffe; i < 0x4003; i++) {
+		int id;
+		struct item *item;
+
+		if (i < 0x4000)
+			item = item_create(i);
+		else
+			item = item_create(i - 0x3fff);
+
+		id = idr_alloc_cyclic(&idr, item, 1, 0x4000, GFP_KERNEL);
+		assert(id == item->index);
+	}
+
+	idr_for_each(&idr, item_idr_free, &idr);
+}
+
+void idr_checks(void)
+{
+	unsigned long i;
+	DEFINE_IDR(idr);
+
+	for (i = 0; i < 10000; i++) {
+		struct item *item = item_create(i);
+		assert(idr_alloc(&idr, item, 0, 20000, GFP_KERNEL) == i);
+	}
+
+	assert(idr_alloc(&idr, DUMMY_PTR, 5, 30, GFP_KERNEL) < 0);
+
+	for (i = 0; i < 5000; i++)
+		item_idr_remove(&idr, i);
+
+	idr_for_each(&idr, item_idr_free, &idr);
+
+	assert(idr_is_empty(&idr));
+
+	for (i = INT_MAX - 3UL; i < INT_MAX + 1UL; i++) {
+		struct item *item = item_create(i);
+		assert(idr_alloc(&idr, item, i, i + 10, GFP_KERNEL) == i);
+	}
+	assert(idr_alloc(&idr, DUMMY_PTR, i - 2, i, GFP_KERNEL) == -ENOSPC);
+
+	idr_destroy(&idr);
+	idr_destroy(&idr);
+
+	assert(idr_is_empty(&idr));
+
+	for (i = 1; i < 10000; i++) {
+		struct item *item = item_create(i);
+		assert(idr_alloc(&idr, item, 1, 20000, GFP_KERNEL) == i);
+	}
+
+	idr_destroy(&idr);
+
+	idr_alloc_test();
+}
+
+void ida_checks(void)
+{
+	DEFINE_IDA(ida);
+
+	unsigned long i;
+	int id;
+
+	for (i = 0; i < 10000; i++) {
+		ida_pre_get(&ida, GFP_KERNEL);
+		ida_get_new(&ida, &id);
+		assert(id == i);
+	}
+
+	ida_remove(&ida, 20);
+	ida_remove(&ida, 21);
+	for (i = 0; i < 3; i++) {
+		ida_pre_get(&ida, GFP_KERNEL);
+		ida_get_new(&ida, &id);
+		if (i == 2)
+			assert(id == 10000);
+	}
+
+	for (i = 0; i < 5000; i++)
+		ida_remove(&ida, i);
+
+	ida_pre_get(&ida, GFP_KERNEL);
+	ida_get_new_above(&ida, 5000, &id);
+	assert(id == 10001);
+
+	ida_destroy(&ida);
+
+	assert(ida_is_empty(&ida));
+
+	ida_pre_get(&ida, GFP_KERNEL);
+	ida_get_new_above(&ida, 1, &id);
+	assert(id == 1);
+
+	ida_remove(&ida, id);
+	ida_destroy(&ida);
+
+	radix_tree_callback(NULL, CPU_DEAD, NULL);
+}
diff --git a/tools/testing/radix-tree/linux/idr.h b/tools/testing/radix-tree/linux/idr.h
new file mode 100644
index 0000000..4e342f2
--- /dev/null
+++ b/tools/testing/radix-tree/linux/idr.h
@@ -0,0 +1 @@
+#include "../../../../include/linux/idr.h"
diff --git a/tools/testing/radix-tree/linux/kernel.h b/tools/testing/radix-tree/linux/kernel.h
index 9b43b49..7d214e9 100644
--- a/tools/testing/radix-tree/linux/kernel.h
+++ b/tools/testing/radix-tree/linux/kernel.h
@@ -30,6 +30,7 @@
 #define __force
 #define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
 #define pr_debug printk
+#define pr_cont printk
 
 #define smp_rmb()	barrier()
 #define smp_wmb()	barrier()
@@ -41,6 +42,7 @@
 	const typeof( ((type *)0)->member ) *__mptr = (ptr);    \
 	(type *)( (char *)__mptr - offsetof(type, member) );})
 #define min(a, b) ((a) < (b) ? (a) : (b))
+#define max(a, b) ((a) < (b) ? (b) : (a))
 
 #define cond_resched()	sched_yield()
 
diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
index 7d14812..e267d5f 100644
--- a/tools/testing/radix-tree/main.c
+++ b/tools/testing/radix-tree/main.c
@@ -3,6 +3,7 @@
 #include <unistd.h>
 #include <time.h>
 #include <assert.h>
+#include <limits.h>
 
 #include <linux/slab.h>
 #include <linux/radix-tree.h>
@@ -314,6 +315,11 @@ static void single_thread_tests(bool long_run)
 	rcu_barrier();
 	printf("after dynamic_height_check: %d allocated, preempt %d\n",
 		nr_allocated, preempt_count);
+	idr_checks();
+	ida_checks();
+	rcu_barrier();
+	printf("after idr_checks: %d allocated, preempt %d\n",
+		nr_allocated, preempt_count);
 	big_gang_check(long_run);
 	rcu_barrier();
 	printf("after big_gang_check: %d allocated, preempt %d\n",
diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
index 1faf0a3..4baec38 100644
--- a/tools/testing/radix-tree/test.h
+++ b/tools/testing/radix-tree/test.h
@@ -36,6 +36,8 @@ void tag_check(void);
 void multiorder_checks(void);
 void iteration_test(void);
 void benchmark(void);
+void idr_checks(void);
+void ida_checks(void);
 
 struct item *
 item_tag_set(struct radix_tree_root *root, unsigned long index, int tag);
-- 
2.10.2

RE: [PATCH 00/29] Improve radix tree for 4.10

[Matthew Wilcox]

My apologies for the confusion here.  Some of these patches are indeed duplicates of each other.  I had run git format-patch, then noticed a mistake, edited the patch set and ran git format-patch again without emptying the directory of patches, so I ended up sending 36 emails instead of 29.

There were also a couple of small bugs caught by 0day and an infelicity in the rxrpc code pointed out by Dave Howells.  Instead of spamming you all with another 29 emails, the latest version of the patchset can be found here:

http://git.infradead.org/users/willy/linux-dax.git/shortlog/refs/heads/idr-2016-11-17

Re: [PATCH 00/29] Improve radix tree for 4.10

[Ross Zwisler]

On Wed, Nov 16, 2016 at 04:17:01PM -0800, Matthew Wilcox wrote:
> From: Matthew Wilcox <mawilcox@microsoft.com>
> 
> Hi Andrew,
> 
> Please include these patches in the -mm tree for 4.10.  Mostly these are
> improvements; the only bug fixes in here relate to multiorder entries
> (which as far as I'm aware remain unused).  

My DAX PMD patches use multiorder entries, and are queued for v4.10 merge:

http://www.spinics.net/lists/linux-fsdevel/msg104041.html

RE: [PATCH 00/29] Improve radix tree for 4.10

[Matthew Wilcox]

From: Ross Zwisler [mailto:ross.zwisler@linux.intel.com]
> On Wed, Nov 16, 2016 at 04:17:01PM -0800, Matthew Wilcox wrote:
> > From: Matthew Wilcox <mawilcox@microsoft.com>
> >
> > Hi Andrew,
> >
> > Please include these patches in the -mm tree for 4.10.  Mostly these are
> > improvements; the only bug fixes in here relate to multiorder entries
> > (which as far as I'm aware remain unused).
> 
> My DAX PMD patches use multiorder entries, and are queued for v4.10 merge:

That's great; the point that I was trying to make was that while there are indeed bug fixes in the patches, they're latent bugs because there are no current users in the tree.  So there's no need to hurry up and get these fixes in, they can wait for 4.10.  I don't think your patches will run into these bugs either because you don't use tags.  I do think that Kirill's patches might hit them though.

Re: [PATCH 20/29] radix tree: Improve multiorder iterators

[Konstantin Khlebnikov]

On Thu, Nov 17, 2016 at 3:17 AM, Matthew Wilcox
<mawilcox@linuxonhyperv.com> wrote:
> From: Matthew Wilcox <mawilcox@microsoft.com>

This code still looks overengineered for me.

>
> This fixes several interlinked problems with the iterators in the
> presence of multiorder entries.
>
> 1. radix_tree_iter_next() would only advance by one slot, which would
> result in the iterators returning the same entry more than once if there
> were sibling entries.

Is this a problem? Do we have users who cannot evalate length of entry
by looking into it head?

>
> 2. radix_tree_next_slot() could return an internal pointer instead of
> a user pointer if a tagged multiorder entry was immediately followed by
> an entry of lower order.
>
> 3. radix_tree_next_slot() expanded to a lot more code than it used to
> when multiorder support was compiled in.  And I wasn't comfortable with
> entry_to_node() being in a header file.
>
> Fixing radix_tree_iter_next() for the presence of sibling entries
> necessarily involves examining the contents of the radix tree, so we now
> need to pass 'slot' to radix_tree_iter_next(), and we need to change the
> calling convention so it is called *before* dropping the lock which
> protects the tree.  Fortunately, unconverted code won't compile.
>
> radix_tree_next_slot() becomes closer to how it looked before multiorder
> support was introduced.  It only checks to see if the next entry in the
> chunk is a sibling entry or a pointer to a node; this should be rare
> enough that handling this case out of line is not a performance impact
> (and such impact is amortised by the fact that the entry we just
> processed was a multiorder entry).  Also, radix_tree_next_slot() used
> to force a new chunk lookup for untagged entries, which is more
> expensive than the out of line sibling entry
>
> Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
> ---
>  fs/btrfs/tests/btrfs-tests.c               |  2 +-
>  include/linux/radix-tree.h                 | 63 ++++++++------------
>  lib/radix-tree.c                           | 94 ++++++++++++++++++++++++++++++
>  mm/khugepaged.c                            |  2 +-
>  mm/shmem.c                                 |  6 +-
>  tools/testing/radix-tree/iteration_check.c |  4 +-
>  tools/testing/radix-tree/multiorder.c      | 12 ++++
>  tools/testing/radix-tree/regression3.c     |  6 +-
>  tools/testing/radix-tree/test.h            |  1 +
>  9 files changed, 142 insertions(+), 48 deletions(-)
>
> diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c
> index 73076a0..6d3457a 100644
> --- a/fs/btrfs/tests/btrfs-tests.c
> +++ b/fs/btrfs/tests/btrfs-tests.c
> @@ -162,7 +162,7 @@ void btrfs_free_dummy_fs_info(struct btrfs_fs_info *fs_info)
>                                 slot = radix_tree_iter_retry(&iter);
>                         continue;
>                 }
> -               slot = radix_tree_iter_next(&iter);
> +               slot = radix_tree_iter_next(slot, &iter);
>                 spin_unlock(&fs_info->buffer_lock);
>                 free_extent_buffer_stale(eb);
>                 spin_lock(&fs_info->buffer_lock);
> diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
> index 66fb8c0..36c6175 100644
> --- a/include/linux/radix-tree.h
> +++ b/include/linux/radix-tree.h
> @@ -424,15 +424,10 @@ __radix_tree_iter_add(struct radix_tree_iter *iter, unsigned long slots)
>   *
>   * If the iterator needs to release then reacquire a lock, the chunk may
>   * have been invalidated by an insertion or deletion.  Call this function
> - * to continue the iteration from the next index.
> + * before releasing the lock to continue the iteration from the next index.
>   */
> -static inline __must_check
> -void **radix_tree_iter_next(struct radix_tree_iter *iter)
> -{
> -       iter->next_index = __radix_tree_iter_add(iter, 1);
> -       iter->tags = 0;
> -       return NULL;
> -}
> +void **__must_check radix_tree_iter_next(void **slot,
> +                                       struct radix_tree_iter *iter);
>
>  /**
>   * radix_tree_chunk_size - get current chunk size
> @@ -446,10 +441,17 @@ radix_tree_chunk_size(struct radix_tree_iter *iter)
>         return (iter->next_index - iter->index) >> iter_shift(iter);
>  }
>
> -static inline struct radix_tree_node *entry_to_node(void *ptr)
> +#ifdef CONFIG_RADIX_TREE_MULTIORDER
> +void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
> +                               unsigned flags);
> +#else
> +/* Can't happen without sibling entries, but the compiler can't tell that */
> +static inline void ** __radix_tree_next_slot(void **slot,
> +                               struct radix_tree_iter *iter, unsigned flags)
>  {
> -       return (void *)((unsigned long)ptr & ~RADIX_TREE_INTERNAL_NODE);
> +       return slot;
>  }
> +#endif
>
>  /**
>   * radix_tree_next_slot - find next slot in chunk
> @@ -474,51 +476,31 @@ static __always_inline void **
>  radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
>  {
>         if (flags & RADIX_TREE_ITER_TAGGED) {
> -               void *canon = slot;
> -
>                 iter->tags >>= 1;
>                 if (unlikely(!iter->tags))
>                         return NULL;
> -               while (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
> -                                       radix_tree_is_internal_node(slot[1])) {
> -                       if (entry_to_node(slot[1]) == canon) {
> -                               iter->tags >>= 1;
> -                               iter->index = __radix_tree_iter_add(iter, 1);
> -                               slot++;
> -                               continue;
> -                       }
> -                       iter->next_index = __radix_tree_iter_add(iter, 1);
> -                       return NULL;
> -               }
>                 if (likely(iter->tags & 1ul)) {
>                         iter->index = __radix_tree_iter_add(iter, 1);
> -                       return slot + 1;
> +                       slot++;
> +                       goto found;
>                 }
>                 if (!(flags & RADIX_TREE_ITER_CONTIG)) {
>                         unsigned offset = __ffs(iter->tags);
>
> -                       iter->tags >>= offset;
> -                       iter->index = __radix_tree_iter_add(iter, offset + 1);
> -                       return slot + offset + 1;
> +                       iter->tags >>= offset++;
> +                       iter->index = __radix_tree_iter_add(iter, offset);
> +                       slot += offset;
> +                       goto found;
>                 }
>         } else {
>                 long count = radix_tree_chunk_size(iter);
> -               void *canon = slot;
>
>                 while (--count > 0) {
>                         slot++;
>                         iter->index = __radix_tree_iter_add(iter, 1);
>
> -                       if (IS_ENABLED(CONFIG_RADIX_TREE_MULTIORDER) &&
> -                           radix_tree_is_internal_node(*slot)) {
> -                               if (entry_to_node(*slot) == canon)
> -                                       continue;
> -                               iter->next_index = iter->index;
> -                               break;
> -                       }
> -
>                         if (likely(*slot))
> -                               return slot;
> +                               goto found;
>                         if (flags & RADIX_TREE_ITER_CONTIG) {
>                                 /* forbid switching to the next chunk */
>                                 iter->next_index = 0;
> @@ -527,6 +509,11 @@ radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
>                 }
>         }
>         return NULL;
> +
> + found:
> +       if (unlikely(radix_tree_is_internal_node(*slot)))
> +               return __radix_tree_next_slot(slot, iter, flags);
> +       return slot;
>  }
>
>  /**
> @@ -577,6 +564,6 @@ radix_tree_next_slot(void **slot, struct radix_tree_iter *iter, unsigned flags)
>              slot || (slot = radix_tree_next_chunk(root, iter,          \
>                               RADIX_TREE_ITER_TAGGED | tag)) ;          \
>              slot = radix_tree_next_slot(slot, iter,                    \
> -                               RADIX_TREE_ITER_TAGGED))
> +                               RADIX_TREE_ITER_TAGGED | tag))
>
>  #endif /* _LINUX_RADIX_TREE_H */
> diff --git a/lib/radix-tree.c b/lib/radix-tree.c
> index 09c5f1d..27b53ef 100644
> --- a/lib/radix-tree.c
> +++ b/lib/radix-tree.c
> @@ -69,6 +69,11 @@ struct radix_tree_preload {
>  };
>  static DEFINE_PER_CPU(struct radix_tree_preload, radix_tree_preloads) = { 0, };
>
> +static inline struct radix_tree_node *entry_to_node(void *ptr)
> +{
> +       return (void *)((unsigned long)ptr & ~RADIX_TREE_INTERNAL_NODE);
> +}
> +
>  static inline void *node_to_entry(void *ptr)
>  {
>         return (void *)((unsigned long)ptr | RADIX_TREE_INTERNAL_NODE);
> @@ -1138,6 +1143,95 @@ static inline void __set_iter_shift(struct radix_tree_iter *iter,
>  #endif
>  }
>
> +static void ** __radix_tree_iter_next(struct radix_tree_node **nodep,
> +                       void **slot, struct radix_tree_iter *iter)
> +{
> +       void *sib = node_to_entry(slot - 1);
> +
> +       while (iter->index < iter->next_index) {
> +               *nodep = rcu_dereference_raw(*slot);
> +               if (*nodep && *nodep != sib)
> +                       return slot;
> +               slot++;
> +               iter->index = __radix_tree_iter_add(iter, 1);
> +               iter->tags >>= 1;
> +       }
> +
> +       *nodep = NULL;
> +       return NULL;
> +}
> +
> +#ifdef CONFIG_RADIX_TREE_MULTIORDER
> +void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
> +                                       unsigned flags)
> +{
> +       unsigned tag = flags & RADIX_TREE_ITER_TAG_MASK;
> +       struct radix_tree_node *node = rcu_dereference_raw(*slot);
> +
> +       slot = __radix_tree_iter_next(&node, slot, iter);
> +
> +       while (radix_tree_is_internal_node(node)) {
> +               unsigned offset;
> +
> +               if (node == RADIX_TREE_RETRY)
> +                       return slot;
> +               node = entry_to_node(node);
> +
> +               if (flags & RADIX_TREE_ITER_TAGGED) {
> +                       unsigned tag_long, tag_bit;
> +                       offset = radix_tree_find_next_bit(node, tag, 0);
> +                       if (offset == RADIX_TREE_MAP_SIZE)
> +                               return NULL;
> +                       slot = &node->slots[offset];
> +
> +                       tag_long = offset / BITS_PER_LONG;
> +                       tag_bit  = offset % BITS_PER_LONG;
> +                       iter->tags = node->tags[tag][tag_long] >> tag_bit;
> +                       BUG_ON(iter->tags >= (RADIX_TREE_MAP_SIZE * 2));
> +                       node = rcu_dereference_raw(*slot);
> +               } else {
> +                       offset = 0;
> +                       slot = &node->slots[0];
> +                       for (;;) {
> +                               node = rcu_dereference_raw(*slot);
> +                               if (node)
> +                                       break;
> +                               slot++;
> +                               offset++;
> +                               if (offset == RADIX_TREE_MAP_SIZE)
> +                                       return NULL;
> +                       }
> +               }
> +               if ((flags & RADIX_TREE_ITER_CONTIG) && (offset > 0))
> +                       goto none;
> +               iter->shift -= RADIX_TREE_MAP_SHIFT;
> +               iter->index = __radix_tree_iter_add(iter, offset);
> +               iter->next_index = (iter->index | shift_maxindex(iter->shift)) +
> +                                       1;
> +       }
> +
> +       return slot;
> + none:
> +       iter->next_index = 0;
> +       return NULL;
> +}
> +EXPORT_SYMBOL(__radix_tree_next_slot);
> +#endif
> +
> +void **radix_tree_iter_next(void **slot, struct radix_tree_iter *iter)
> +{
> +       struct radix_tree_node *node;
> +
> +       slot++;
> +       iter->index = __radix_tree_iter_add(iter, 1);
> +       node = rcu_dereference_raw(*slot);
> +       __radix_tree_iter_next(&node, slot, iter);
> +       iter->next_index = iter->index;
> +       iter->tags = 0;
> +       return NULL;
> +}
> +EXPORT_SYMBOL(radix_tree_iter_next);
> +
>  /**
>   * radix_tree_next_chunk - find next chunk of slots for iteration
>   *
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 728d779..46155d1 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -1614,8 +1614,8 @@ static void khugepaged_scan_shmem(struct mm_struct *mm,
>                 present++;
>
>                 if (need_resched()) {
> +                       slot = radix_tree_iter_next(slot, &iter);
>                         cond_resched_rcu();
> -                       slot = radix_tree_iter_next(&iter);
>                 }
>         }
>         rcu_read_unlock();
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 166ebf5..0b3fe33 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -658,8 +658,8 @@ unsigned long shmem_partial_swap_usage(struct address_space *mapping,
>                         swapped++;
>
>                 if (need_resched()) {
> +                       slot = radix_tree_iter_next(slot, &iter);
>                         cond_resched_rcu();
> -                       slot = radix_tree_iter_next(&iter);
>                 }
>         }
>
> @@ -2434,8 +2434,8 @@ static void shmem_tag_pins(struct address_space *mapping)
>                 }
>
>                 if (need_resched()) {
> +                       slot = radix_tree_iter_next(slot, &iter);
>                         cond_resched_rcu();
> -                       slot = radix_tree_iter_next(&iter);
>                 }
>         }
>         rcu_read_unlock();
> @@ -2504,8 +2504,8 @@ static int shmem_wait_for_pins(struct address_space *mapping)
>                         spin_unlock_irq(&mapping->tree_lock);
>  continue_resched:
>                         if (need_resched()) {
> +                               slot = radix_tree_iter_next(slot, &iter);
>                                 cond_resched_rcu();
> -                               slot = radix_tree_iter_next(&iter);
>                         }
>                 }
>                 rcu_read_unlock();
> diff --git a/tools/testing/radix-tree/iteration_check.c b/tools/testing/radix-tree/iteration_check.c
> index df71cb8..6eca4c2 100644
> --- a/tools/testing/radix-tree/iteration_check.c
> +++ b/tools/testing/radix-tree/iteration_check.c
> @@ -79,7 +79,7 @@ static void *tagged_iteration_fn(void *arg)
>                         }
>
>                         if (rand_r(&seeds[0]) % 50 == 0) {
> -                               slot = radix_tree_iter_next(&iter);
> +                               slot = radix_tree_iter_next(slot, &iter);
>                                 rcu_read_unlock();
>                                 rcu_barrier();
>                                 rcu_read_lock();
> @@ -127,7 +127,7 @@ static void *untagged_iteration_fn(void *arg)
>                         }
>
>                         if (rand_r(&seeds[1]) % 50 == 0) {
> -                               slot = radix_tree_iter_next(&iter);
> +                               slot = radix_tree_iter_next(slot, &iter);
>                                 rcu_read_unlock();
>                                 rcu_barrier();
>                                 rcu_read_lock();
> diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
> index 25e0463..588209a 100644
> --- a/tools/testing/radix-tree/multiorder.c
> +++ b/tools/testing/radix-tree/multiorder.c
> @@ -232,10 +232,14 @@ void multiorder_iteration(void)
>                         int height = order[i] / RADIX_TREE_MAP_SHIFT;
>                         int shift = height * RADIX_TREE_MAP_SHIFT;
>                         int mask = (1 << order[i]) - 1;
> +                       struct item *item = *slot;
>
>                         assert(iter.index >= (index[i] &~ mask));
>                         assert(iter.index <= (index[i] | mask));
>                         assert(iter.shift == shift);
> +                       assert(!radix_tree_is_internal_node(item));
> +                       assert(item->index >= (index[i] &~ mask));
> +                       assert(item->index <= (index[i] | mask));
>                         i++;
>                 }
>         }
> @@ -279,12 +283,16 @@ void multiorder_tagged_iteration(void)
>                 }
>
>                 radix_tree_for_each_tagged(slot, &tree, &iter, j, 1) {
> +                       struct item *item = *slot;
>                         for (k = i; index[k] < tag_index[i]; k++)
>                                 ;
>                         mask = (1 << order[k]) - 1;
>
>                         assert(iter.index >= (tag_index[i] &~ mask));
>                         assert(iter.index <= (tag_index[i] | mask));
> +                       assert(!radix_tree_is_internal_node(item));
> +                       assert(item->index >= (tag_index[i] &~ mask));
> +                       assert(item->index <= (tag_index[i] | mask));
>                         i++;
>                 }
>         }
> @@ -303,12 +311,16 @@ void multiorder_tagged_iteration(void)
>                 }
>
>                 radix_tree_for_each_tagged(slot, &tree, &iter, j, 2) {
> +                       struct item *item = *slot;
>                         for (k = i; index[k] < tag_index[i]; k++)
>                                 ;
>                         mask = (1 << order[k]) - 1;
>
>                         assert(iter.index >= (tag_index[i] &~ mask));
>                         assert(iter.index <= (tag_index[i] | mask));
> +                       assert(!radix_tree_is_internal_node(item));
> +                       assert(item->index >= (tag_index[i] &~ mask));
> +                       assert(item->index <= (tag_index[i] | mask));
>                         i++;
>                 }
>         }
> diff --git a/tools/testing/radix-tree/regression3.c b/tools/testing/radix-tree/regression3.c
> index 1f06ed7..4d28eeb 100644
> --- a/tools/testing/radix-tree/regression3.c
> +++ b/tools/testing/radix-tree/regression3.c
> @@ -88,7 +88,7 @@ void regression3_test(void)
>                 printf("slot %ld %p\n", iter.index, *slot);
>                 if (!iter.index) {
>                         printf("next at %ld\n", iter.index);
> -                       slot = radix_tree_iter_next(&iter);
> +                       slot = radix_tree_iter_next(slot, &iter);
>                 }
>         }
>
> @@ -96,7 +96,7 @@ void regression3_test(void)
>                 printf("contig %ld %p\n", iter.index, *slot);
>                 if (!iter.index) {
>                         printf("next at %ld\n", iter.index);
> -                       slot = radix_tree_iter_next(&iter);
> +                       slot = radix_tree_iter_next(slot, &iter);
>                 }
>         }
>
> @@ -106,7 +106,7 @@ void regression3_test(void)
>                 printf("tagged %ld %p\n", iter.index, *slot);
>                 if (!iter.index) {
>                         printf("next at %ld\n", iter.index);
> -                       slot = radix_tree_iter_next(&iter);
> +                       slot = radix_tree_iter_next(slot, &iter);
>                 }
>         }
>
> diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
> index 33d2b6b..b678f13 100644
> --- a/tools/testing/radix-tree/test.h
> +++ b/tools/testing/radix-tree/test.h
> @@ -43,6 +43,7 @@ void verify_tag_consistency(struct radix_tree_root *root, unsigned int tag);
>  extern int nr_allocated;
>
>  /* Normally private parts of lib/radix-tree.c */
> +struct radix_tree_node *entry_to_node(void *ptr);
>  void radix_tree_dump(struct radix_tree_root *root);
>  int root_tag_get(struct radix_tree_root *root, unsigned int tag);
>  unsigned long node_maxindex(struct radix_tree_node *);
> --
> 2.10.2
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 21/29] radix-tree: Delete radix_tree_locate_item()

[Konstantin Khlebnikov]

On Thu, Nov 17, 2016 at 3:17 AM, Matthew Wilcox
<mawilcox@linuxonhyperv.com> wrote:
> From: Matthew Wilcox <mawilcox@microsoft.com>
>
> This rather complicated function can be better implemented as an iterator.
> It has only one caller, so move the functionality to the only place that
> needs it.  Update the test suite to follow the same pattern.

Looks good. I suppose this patch could be applied separately.

>
> Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
> ---
>  include/linux/radix-tree.h            |  1 -
>  lib/radix-tree.c                      | 99 -----------------------------------
>  mm/shmem.c                            | 26 ++++++++-
>  tools/testing/radix-tree/main.c       |  8 +--
>  tools/testing/radix-tree/multiorder.c |  2 +-
>  tools/testing/radix-tree/test.c       | 22 ++++++++
>  tools/testing/radix-tree/test.h       |  2 +
>  7 files changed, 54 insertions(+), 106 deletions(-)
>
> diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
> index 36c6175..57bf635 100644
> --- a/include/linux/radix-tree.h
> +++ b/include/linux/radix-tree.h
> @@ -306,7 +306,6 @@ unsigned long radix_tree_range_tag_if_tagged(struct radix_tree_root *root,
>                 unsigned long nr_to_tag,
>                 unsigned int fromtag, unsigned int totag);
>  int radix_tree_tagged(struct radix_tree_root *root, unsigned int tag);
> -unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item);
>
>  static inline void radix_tree_preload_end(void)
>  {
> diff --git a/lib/radix-tree.c b/lib/radix-tree.c
> index 27b53ef..7e70ac9 100644
> --- a/lib/radix-tree.c
> +++ b/lib/radix-tree.c
> @@ -1605,105 +1605,6 @@ radix_tree_gang_lookup_tag_slot(struct radix_tree_root *root, void ***results,
>  }
>  EXPORT_SYMBOL(radix_tree_gang_lookup_tag_slot);
>
> -#if defined(CONFIG_SHMEM) && defined(CONFIG_SWAP)
> -#include <linux/sched.h> /* for cond_resched() */
> -
> -struct locate_info {
> -       unsigned long found_index;
> -       bool stop;
> -};
> -
> -/*
> - * This linear search is at present only useful to shmem_unuse_inode().
> - */
> -static unsigned long __locate(struct radix_tree_node *slot, void *item,
> -                             unsigned long index, struct locate_info *info)
> -{
> -       unsigned long i;
> -
> -       do {
> -               unsigned int shift = slot->shift;
> -
> -               for (i = (index >> shift) & RADIX_TREE_MAP_MASK;
> -                    i < RADIX_TREE_MAP_SIZE;
> -                    i++, index += (1UL << shift)) {
> -                       struct radix_tree_node *node =
> -                                       rcu_dereference_raw(slot->slots[i]);
> -                       if (node == RADIX_TREE_RETRY)
> -                               goto out;
> -                       if (!radix_tree_is_internal_node(node)) {
> -                               if (node == item) {
> -                                       info->found_index = index;
> -                                       info->stop = true;
> -                                       goto out;
> -                               }
> -                               continue;
> -                       }
> -                       node = entry_to_node(node);
> -                       if (is_sibling_entry(slot, node))
> -                               continue;
> -                       slot = node;
> -                       break;
> -               }
> -       } while (i < RADIX_TREE_MAP_SIZE);
> -
> -out:
> -       if ((index == 0) && (i == RADIX_TREE_MAP_SIZE))
> -               info->stop = true;
> -       return index;
> -}
> -
> -/**
> - *     radix_tree_locate_item - search through radix tree for item
> - *     @root:          radix tree root
> - *     @item:          item to be found
> - *
> - *     Returns index where item was found, or -1 if not found.
> - *     Caller must hold no lock (since this time-consuming function needs
> - *     to be preemptible), and must check afterwards if item is still there.
> - */
> -unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item)
> -{
> -       struct radix_tree_node *node;
> -       unsigned long max_index;
> -       unsigned long cur_index = 0;
> -       struct locate_info info = {
> -               .found_index = -1,
> -               .stop = false,
> -       };
> -
> -       do {
> -               rcu_read_lock();
> -               node = rcu_dereference_raw(root->rnode);
> -               if (!radix_tree_is_internal_node(node)) {
> -                       rcu_read_unlock();
> -                       if (node == item)
> -                               info.found_index = 0;
> -                       break;
> -               }
> -
> -               node = entry_to_node(node);
> -
> -               max_index = node_maxindex(node);
> -               if (cur_index > max_index) {
> -                       rcu_read_unlock();
> -                       break;
> -               }
> -
> -               cur_index = __locate(node, item, cur_index, &info);
> -               rcu_read_unlock();
> -               cond_resched();
> -       } while (!info.stop && cur_index <= max_index);
> -
> -       return info.found_index;
> -}
> -#else
> -unsigned long radix_tree_locate_item(struct radix_tree_root *root, void *item)
> -{
> -       return -1;
> -}
> -#endif /* CONFIG_SHMEM && CONFIG_SWAP */
> -
>  /**
>   *     radix_tree_shrink    -    shrink radix tree to minimum height
>   *     @root           radix tree root
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 0b3fe33..8f9c9aa 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -1046,6 +1046,30 @@ static void shmem_evict_inode(struct inode *inode)
>         clear_inode(inode);
>  }
>
> +static unsigned long find_swap_entry(struct radix_tree_root *root, void *item)
> +{
> +       struct radix_tree_iter iter;
> +       void **slot;
> +       unsigned long found = -1;
> +       unsigned int checked = 0;
> +
> +       rcu_read_lock();
> +       radix_tree_for_each_slot(slot, root, &iter, 0) {
> +               if (*slot == item) {
> +                       found = iter.index;
> +                       break;
> +               }
> +               checked++;
> +               if ((checked % 4096) != 0)
> +                       continue;
> +               slot = radix_tree_iter_next(slot, &iter);
> +               cond_resched_rcu();
> +       }
> +
> +       rcu_read_unlock();
> +       return found;
> +}
> +
>  /*
>   * If swap found in inode, free it and move page from swapcache to filecache.
>   */
> @@ -1059,7 +1083,7 @@ static int shmem_unuse_inode(struct shmem_inode_info *info,
>         int error = 0;
>
>         radswap = swp_to_radix_entry(swap);
> -       index = radix_tree_locate_item(&mapping->page_tree, radswap);
> +       index = find_swap_entry(&mapping->page_tree, radswap);
>         if (index == -1)
>                 return -EAGAIN; /* tell shmem_unuse we found nothing */
>
> diff --git a/tools/testing/radix-tree/main.c b/tools/testing/radix-tree/main.c
> index 8621542..93a77f9 100644
> --- a/tools/testing/radix-tree/main.c
> +++ b/tools/testing/radix-tree/main.c
> @@ -239,7 +239,7 @@ static void __locate_check(struct radix_tree_root *tree, unsigned long index,
>
>         item_insert_order(tree, index, order);
>         item = item_lookup(tree, index);
> -       index2 = radix_tree_locate_item(tree, item);
> +       index2 = find_item(tree, item);
>         if (index != index2) {
>                 printf("index %ld order %d inserted; found %ld\n",
>                         index, order, index2);
> @@ -273,17 +273,17 @@ static void locate_check(void)
>                              index += (1UL << order)) {
>                                 __locate_check(&tree, index + offset, order);
>                         }
> -                       if (radix_tree_locate_item(&tree, &tree) != -1)
> +                       if (find_item(&tree, &tree) != -1)
>                                 abort();
>
>                         item_kill_tree(&tree);
>                 }
>         }
>
> -       if (radix_tree_locate_item(&tree, &tree) != -1)
> +       if (find_item(&tree, &tree) != -1)
>                 abort();
>         __locate_check(&tree, -1, 0);
> -       if (radix_tree_locate_item(&tree, &tree) != -1)
> +       if (find_item(&tree, &tree) != -1)
>                 abort();
>         item_kill_tree(&tree);
>  }
> diff --git a/tools/testing/radix-tree/multiorder.c b/tools/testing/radix-tree/multiorder.c
> index 588209a..400de5c 100644
> --- a/tools/testing/radix-tree/multiorder.c
> +++ b/tools/testing/radix-tree/multiorder.c
> @@ -347,7 +347,7 @@ static void __multiorder_join(unsigned long index,
>         item_insert_order(&tree, index, order2);
>         item = radix_tree_lookup(&tree, index);
>         radix_tree_join(&tree, index + 1, order1, item2);
> -       loc = radix_tree_locate_item(&tree, item);
> +       loc = find_item(&tree, item);
>         if (loc == -1)
>                 free(item);
>         item = radix_tree_lookup(&tree, index + 1);
> diff --git a/tools/testing/radix-tree/test.c b/tools/testing/radix-tree/test.c
> index a6e8099..a68ed3b 100644
> --- a/tools/testing/radix-tree/test.c
> +++ b/tools/testing/radix-tree/test.c
> @@ -142,6 +142,28 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
>         assert(nfound == 0);
>  }
>
> +/* Use the same pattern as find_swap_entry() in mm/shmem.c */
> +unsigned long find_item(struct radix_tree_root *root, void *item)
> +{
> +       struct radix_tree_iter iter;
> +       void **slot;
> +       unsigned long found = -1;
> +       unsigned long checked = 0;
> +
> +       radix_tree_for_each_slot(slot, root, &iter, 0) {
> +               if (*slot == item) {
> +                       found = iter.index;
> +                       break;
> +               }
> +               checked++;
> +               if ((checked % 4) != 0)
> +                       continue;
> +               slot = radix_tree_iter_next(slot, &iter);
> +       }
> +
> +       return found;
> +}
> +
>  static int verify_node(struct radix_tree_node *slot, unsigned int tag,
>                         int tagged)
>  {
> diff --git a/tools/testing/radix-tree/test.h b/tools/testing/radix-tree/test.h
> index b678f13..ccdd3c1 100644
> --- a/tools/testing/radix-tree/test.h
> +++ b/tools/testing/radix-tree/test.h
> @@ -27,6 +27,8 @@ void item_full_scan(struct radix_tree_root *root, unsigned long start,
>                         unsigned long nr, int chunk);
>  void item_kill_tree(struct radix_tree_root *root);
>
> +unsigned long find_item(struct radix_tree_root *, void *item);
> +
>  void tag_check(void);
>  void multiorder_checks(void);
>  void iteration_test(void);
> --
> 2.10.2
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 23/29] idr: Add ida_is_empty

[Konstantin Khlebnikov]

On Thu, Nov 17, 2016 at 3:16 AM, Matthew Wilcox
<mawilcox@linuxonhyperv.com> wrote:
> From: Matthew Wilcox <willy@linux.intel.com>
>
> Two of the USB Gadgets were poking around in the internals of struct ida
> in order to determine if it is empty.  Add the appropriate abstraction.

Looks good. This could be applied separately.

>
> Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
> ---
>  drivers/usb/gadget/function/f_hid.c     | 6 +++---
>  drivers/usb/gadget/function/f_printer.c | 6 +++---
>  include/linux/idr.h                     | 5 +++++
>  3 files changed, 11 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/usb/gadget/function/f_hid.c b/drivers/usb/gadget/function/f_hid.c
> index e2966f8..b9b04fc 100644
> --- a/drivers/usb/gadget/function/f_hid.c
> +++ b/drivers/usb/gadget/function/f_hid.c
> @@ -840,7 +840,7 @@ static void hidg_free_inst(struct usb_function_instance *f)
>         mutex_lock(&hidg_ida_lock);
>
>         hidg_put_minor(opts->minor);
> -       if (idr_is_empty(&hidg_ida.idr))
> +       if (ida_is_empty(&hidg_ida))
>                 ghid_cleanup();
>
>         mutex_unlock(&hidg_ida_lock);
> @@ -866,7 +866,7 @@ static struct usb_function_instance *hidg_alloc_inst(void)
>
>         mutex_lock(&hidg_ida_lock);
>
> -       if (idr_is_empty(&hidg_ida.idr)) {
> +       if (ida_is_empty(&hidg_ida)) {
>                 status = ghid_setup(NULL, HIDG_MINORS);
>                 if (status)  {
>                         ret = ERR_PTR(status);
> @@ -879,7 +879,7 @@ static struct usb_function_instance *hidg_alloc_inst(void)
>         if (opts->minor < 0) {
>                 ret = ERR_PTR(opts->minor);
>                 kfree(opts);
> -               if (idr_is_empty(&hidg_ida.idr))
> +               if (ida_is_empty(&hidg_ida))
>                         ghid_cleanup();
>                 goto unlock;
>         }
> diff --git a/drivers/usb/gadget/function/f_printer.c b/drivers/usb/gadget/function/f_printer.c
> index 0de36cd..8054da9 100644
> --- a/drivers/usb/gadget/function/f_printer.c
> +++ b/drivers/usb/gadget/function/f_printer.c
> @@ -1265,7 +1265,7 @@ static void gprinter_free_inst(struct usb_function_instance *f)
>         mutex_lock(&printer_ida_lock);
>
>         gprinter_put_minor(opts->minor);
> -       if (idr_is_empty(&printer_ida.idr))
> +       if (ida_is_empty(&printer_ida))
>                 gprinter_cleanup();
>
>         mutex_unlock(&printer_ida_lock);
> @@ -1289,7 +1289,7 @@ static struct usb_function_instance *gprinter_alloc_inst(void)
>
>         mutex_lock(&printer_ida_lock);
>
> -       if (idr_is_empty(&printer_ida.idr)) {
> +       if (ida_is_empty(&printer_ida)) {
>                 status = gprinter_setup(PRINTER_MINORS);
>                 if (status) {
>                         ret = ERR_PTR(status);
> @@ -1302,7 +1302,7 @@ static struct usb_function_instance *gprinter_alloc_inst(void)
>         if (opts->minor < 0) {
>                 ret = ERR_PTR(opts->minor);
>                 kfree(opts);
> -               if (idr_is_empty(&printer_ida.idr))
> +               if (ida_is_empty(&printer_ida))
>                         gprinter_cleanup();
>                 goto unlock;
>         }
> diff --git a/include/linux/idr.h b/include/linux/idr.h
> index 083d61e..3639a28 100644
> --- a/include/linux/idr.h
> +++ b/include/linux/idr.h
> @@ -195,6 +195,11 @@ static inline int ida_get_new(struct ida *ida, int *p_id)
>         return ida_get_new_above(ida, 0, p_id);
>  }
>
> +static inline bool ida_is_empty(struct ida *ida)
> +{
> +       return idr_is_empty(&ida->idr);
> +}
> +
>  void __init idr_init_cache(void);
>
>  #endif /* __IDR_H__ */
> --
> 2.10.2
>

RE: [PATCH 20/29] radix tree: Improve multiorder iterators

[Matthew Wilcox]

From: Konstantin Khlebnikov [mailto:koct9i@gmail.com]
> On Thu, Nov 17, 2016 at 3:17 AM, Matthew Wilcox
> <mawilcox@linuxonhyperv.com> wrote:
> > This fixes several interlinked problems with the iterators in the
> > presence of multiorder entries.
> >
> > 1. radix_tree_iter_next() would only advance by one slot, which would
> > result in the iterators returning the same entry more than once if there
> > were sibling entries.
> 
> Is this a problem? Do we have users who cannot evalate length of entry
> by looking into it head?

At the moment we have no users in tree :-)  The two users I know of are the page cache and DAX.  The page cache stores a pointer to a struct page, which has compound_order() to tell you the size.  DAX uses a couple of bits in the radix tree entry to describe whether this is a PTE/PMD/PUD and so also knows the size of the entry that it found.  We also store swap cache entries in the same radix tree (tagged exceptional).  These currently have no information in them to describe their size because each one represents only one page.  The latest patchset to support swapping huge pages inserts 512 entries into the radix tree instead of taking advantage of the multiorder entry infrastructure.  Hopefully that gets fixed soon, but it will require stealing a bit from either the number of swap files allowed or from the maximum size of each swap file (currently 32/128GB)

I think what you're suggesting is that we introduce a new API:

 slot = radix_tree_iter_save(&iter, order);

where the caller tells us the order of the entry it just consumed.  Or maybe you're suggesting

 slot = radix_tree_iter_advance(&iter, newindex)

which would allow us to skip to any index.  Although ... isn't that just radix_tree_iter_init()?

It does push a bit of complexity onto the callers.  We have 7 callers of radix_tree_iter_next() in my current tree (after applying this patch set, so range_tag_if_tagged and locate_item have been pushed into their callers): btrfs, kugepaged, page-writeback and shmem.  btrfs knows its objects occupy one slot.  khugepaged knows that its page is order 0 at the time it calls radix_tree_iter_next().  Page-writeback has a struct page and can simply use compound_order().  It's shmem where things get sticky, although it's all solvable with some temporary variables.

MM people, what do you think of this patch?  It's on top of my current IDR tree, although I'd fold it into patch 20 of the series if it's acceptable.

diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c
index 6d3457a..7f864ec 100644
--- a/fs/btrfs/tests/btrfs-tests.c
+++ b/fs/btrfs/tests/btrfs-tests.c
@@ -162,7 +162,7 @@ void btrfs_free_dummy_fs_info(struct btrfs_fs_info *fs_info)
 				slot = radix_tree_iter_retry(&iter);
 			continue;
 		}
-		slot = radix_tree_iter_next(slot, &iter);
+		slot = radix_tree_iter_save(&iter, 0);
 		spin_unlock(&fs_info->buffer_lock);
 		free_extent_buffer_stale(eb);
 		spin_lock(&fs_info->buffer_lock);
diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
index 4e42d4d..4419325 100644
--- a/include/linux/radix-tree.h
+++ b/include/linux/radix-tree.h
@@ -421,15 +421,22 @@ __radix_tree_iter_add(struct radix_tree_iter *iter, unsigned long slots)
 }
 
 /**
- * radix_tree_iter_next - resume iterating when the chunk may be invalid
- * @iter:	iterator state
+ * radix_tree_iter_save - resume iterating when the chunk may be invalid
+ * @iter: iterator state
+ * @order: order of the entry that was just processed
  *
- * If the iterator needs to release then reacquire a lock, the chunk may
- * have been invalidated by an insertion or deletion.  Call this function
+ * If the iteration needs to release then reacquire a lock, the chunk may
+ * be invalidated by an insertion or deletion.  Call this function
  * before releasing the lock to continue the iteration from the next index.
  */
-void **__must_check radix_tree_iter_next(void **slot,
-					struct radix_tree_iter *iter);
+static inline void **__must_check
+radix_tree_iter_save(struct radix_tree_iter *iter, unsigned order)
+{
+	iter->next_index = round_up(iter->index, 1 << order);
+	iter->index = 0;
+	iter->tags = 0;
+	return NULL;
+}
 
 /**
  * radix_tree_chunk_size - get current chunk size
@@ -467,7 +474,7 @@ static inline void ** __radix_tree_next_slot(void **slot,
  * For tagged lookup it also eats @iter->tags.
  *
  * There are several cases where 'slot' can be passed in as NULL to this
- * function.  These cases result from the use of radix_tree_iter_next() or
+ * function.  These cases result from the use of radix_tree_iter_save() or
  * radix_tree_iter_retry().  In these cases we don't end up dereferencing
  * 'slot' because either:
  * a) we are doing tagged iteration and iter->tags has been set to 0, or
diff --git a/lib/radix-tree.c b/lib/radix-tree.c
index 7e2469b..fcbadad 100644
--- a/lib/radix-tree.c
+++ b/lib/radix-tree.c
@@ -1245,6 +1245,7 @@ static inline void __set_iter_shift(struct radix_tree_iter *iter,
 #endif
 }
 
+#ifdef CONFIG_RADIX_TREE_MULTIORDER
 static void ** __radix_tree_iter_next(struct radix_tree_node **nodep,
 			void **slot, struct radix_tree_iter *iter)
 {
@@ -1263,7 +1264,6 @@ static void ** __radix_tree_iter_next(struct radix_tree_node **nodep,
 	return NULL;
 }
 
-#ifdef CONFIG_RADIX_TREE_MULTIORDER
 void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
 					unsigned flags)
 {
@@ -1321,20 +1321,6 @@ void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
 EXPORT_SYMBOL(__radix_tree_next_slot);
 #endif
 
-void **radix_tree_iter_next(void **slot, struct radix_tree_iter *iter)
-{
-	struct radix_tree_node *node;
-
-	slot++;
-	iter->index = __radix_tree_iter_add(iter, 1);
-	node = rcu_dereference_raw(*slot);
-	__radix_tree_iter_next(&node, slot, iter);
-	iter->next_index = iter->index;
-	iter->tags = 0;
-	return NULL;
-}
-EXPORT_SYMBOL(radix_tree_iter_next);
-
 /**
  * radix_tree_next_chunk - find next chunk of slots for iteration
  *
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 46155d1..54446e6 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -1614,7 +1614,7 @@ static void khugepaged_scan_shmem(struct mm_struct *mm,
 		present++;
 
 		if (need_resched()) {
-			slot = radix_tree_iter_next(slot, &iter);
+			slot = radix_tree_iter_save(&iter, 0);
 			cond_resched_rcu();
 		}
 	}
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index c593715..7d6b870 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -2119,7 +2119,7 @@ void tag_pages_for_writeback(struct address_space *mapping,
 		tagged++;
 		if ((tagged % WRITEBACK_TAG_BATCH) != 0)
 			continue;
-		slot = radix_tree_iter_next(slot, &iter);
+		slot = radix_tree_iter_save(&iter, compound_order(*slot));
 		spin_unlock_irq(&mapping->tree_lock);
 		cond_resched();
 		spin_lock_irq(&mapping->tree_lock);
diff --git a/mm/shmem.c b/mm/shmem.c
index 8f9c9aa..3f2d07a 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -644,6 +644,7 @@ unsigned long shmem_partial_swap_usage(struct address_space *mapping,
 	rcu_read_lock();
 
 	radix_tree_for_each_slot(slot, &mapping->page_tree, &iter, start) {
+		unsigned int order = 0;
 		if (iter.index >= end)
 			break;
 
@@ -656,9 +657,11 @@ unsigned long shmem_partial_swap_usage(struct address_space *mapping,
 
 		if (radix_tree_exceptional_entry(page))
 			swapped++;
+		else
+			order = compound_order(page);
 
 		if (need_resched()) {
-			slot = radix_tree_iter_next(slot, &iter);
+			slot = radix_tree_iter_save(&iter, order);
 			cond_resched_rcu();
 		}
 	}
@@ -1062,7 +1065,7 @@ static unsigned long find_swap_entry(struct radix_tree_root *root, void *item)
 		checked++;
 		if ((checked % 4096) != 0)
 			continue;
-		slot = radix_tree_iter_next(slot, &iter);
+		slot = radix_tree_iter_save(&iter, 0);
 		cond_resched_rcu();
 	}
 
@@ -2444,21 +2447,25 @@ static void shmem_tag_pins(struct address_space *mapping)
 	rcu_read_lock();
 
 	radix_tree_for_each_slot(slot, &mapping->page_tree, &iter, start) {
+		unsigned int order = 0;
 		page = radix_tree_deref_slot(slot);
 		if (!page || radix_tree_exception(page)) {
 			if (radix_tree_deref_retry(page)) {
 				slot = radix_tree_iter_retry(&iter);
 				continue;
 			}
-		} else if (page_count(page) - page_mapcount(page) > 1) {
-			spin_lock_irq(&mapping->tree_lock);
-			radix_tree_tag_set(&mapping->page_tree, iter.index,
-					   SHMEM_TAG_PINNED);
-			spin_unlock_irq(&mapping->tree_lock);
+		} else {
+			order = compound_order(page);
+			if (page_count(page) - page_mapcount(page) > 1) {
+				spin_lock_irq(&mapping->tree_lock);
+				radix_tree_tag_set(&mapping->page_tree,
+						iter.index, SHMEM_TAG_PINNED);
+				spin_unlock_irq(&mapping->tree_lock);
+			}
 		}
 
 		if (need_resched()) {
-			slot = radix_tree_iter_next(slot, &iter);
+			slot = radix_tree_iter_save(&iter, order);
 			cond_resched_rcu();
 		}
 	}
@@ -2528,7 +2535,10 @@ static int shmem_wait_for_pins(struct address_space *mapping)
 			spin_unlock_irq(&mapping->tree_lock);
 continue_resched:
 			if (need_resched()) {
-				slot = radix_tree_iter_next(slot, &iter);
+				unsigned int order = 0;
+				if (page)
+					order = compound_order(page);
+				slot = radix_tree_iter_save(&iter, order);
 				cond_resched_rcu();
 			}
 		}

RE: [PATCH 21/29] radix-tree: Delete radix_tree_locate_item()

[Matthew Wilcox]

From: Konstantin Khlebnikov [mailto:koct9i@gmail.com]
> On Thu, Nov 17, 2016 at 3:17 AM, Matthew Wilcox
> <mawilcox@linuxonhyperv.com> wrote:
> > This rather complicated function can be better implemented as an iterator.
> > It has only one caller, so move the functionality to the only place that
> > needs it.  Update the test suite to follow the same pattern.
> 
> Looks good. I suppose this patch could be applied separately.

Yes, a number of these patches stand alone from each other and could easily be separated out.
I'm going to put your Reviewed-by: on the patches you've said "Looks good" to.  Thanks!

Re: [PATCH 20/29] radix tree: Improve multiorder iterators

[Konstantin Khlebnikov]

On Fri, Nov 18, 2016 at 7:31 PM, Matthew Wilcox <mawilcox@microsoft.com> wrote:
> From: Konstantin Khlebnikov [mailto:koct9i@gmail.com]
>> On Thu, Nov 17, 2016 at 3:17 AM, Matthew Wilcox
>> <mawilcox@linuxonhyperv.com> wrote:
>> > This fixes several interlinked problems with the iterators in the
>> > presence of multiorder entries.
>> >
>> > 1. radix_tree_iter_next() would only advance by one slot, which would
>> > result in the iterators returning the same entry more than once if there
>> > were sibling entries.
>>
>> Is this a problem? Do we have users who cannot evalate length of entry
>> by looking into it head?
>
> At the moment we have no users in tree :-)  The two users I know of are the page cache and DAX.  The page cache stores a pointer to a struct page, which has compound_order() to tell you the size.  DAX uses a couple of bits in the radix tree entry to describe whether this is a PTE/PMD/PUD and so also knows the size of the entry that it found.  We also store swap cache entries in the same radix tree (tagged exceptional).  These currently have no information in them to describe their size because each one represents only one page.  The latest patchset to support swapping huge pages inserts 512 entries into the radix tree instead of taking advantage of the multiorder entry infrastructure.  Hopefully that gets fixed soon, but it will require stealing a bit from either the number of swap files allowed or from the maximum size of each swap file (currently 32/128GB)
>
> I think what you're suggesting is that we introduce a new API:
>
>  slot = radix_tree_iter_save(&iter, order);
>
> where the caller tells us the order of the entry it just consumed.  Or maybe you're suggesting
>
>  slot = radix_tree_iter_advance(&iter, newindex)

Yes, someting like that.

>
> which would allow us to skip to any index.  Although ... isn't that just radix_tree_iter_init()?

Iterator could keep pointer to current node and reuse it for next
iteration if possible.

>
> It does push a bit of complexity onto the callers.  We have 7 callers of radix_tree_iter_next() in my current tree (after applying this patch set, so range_tag_if_tagged and locate_item have been pushed into their callers): btrfs, kugepaged, page-writeback and shmem.  btrfs knows its objects occupy one slot.  khugepaged knows that its page is order 0 at the time it calls radix_tree_iter_next().  Page-writeback has a struct page and can simply use compound_order().  It's shmem where things get sticky, although it's all solvable with some temporary variables.
>

Users who work only with single slot enties don't get any complications,
all other already manage these multiorder entries somehow.

> MM people, what do you think of this patch?  It's on top of my current IDR tree, although I'd fold it into patch 20 of the series if it's acceptable.
>
> diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c
> index 6d3457a..7f864ec 100644
> --- a/fs/btrfs/tests/btrfs-tests.c
> +++ b/fs/btrfs/tests/btrfs-tests.c
> @@ -162,7 +162,7 @@ void btrfs_free_dummy_fs_info(struct btrfs_fs_info *fs_info)
>                                 slot = radix_tree_iter_retry(&iter);
>                         continue;
>                 }
> -               slot = radix_tree_iter_next(slot, &iter);
> +               slot = radix_tree_iter_save(&iter, 0);
>                 spin_unlock(&fs_info->buffer_lock);
>                 free_extent_buffer_stale(eb);
>                 spin_lock(&fs_info->buffer_lock);
> diff --git a/include/linux/radix-tree.h b/include/linux/radix-tree.h
> index 4e42d4d..4419325 100644
> --- a/include/linux/radix-tree.h
> +++ b/include/linux/radix-tree.h
> @@ -421,15 +421,22 @@ __radix_tree_iter_add(struct radix_tree_iter *iter, unsigned long slots)
>  }
>
>  /**
> - * radix_tree_iter_next - resume iterating when the chunk may be invalid
> - * @iter:      iterator state
> + * radix_tree_iter_save - resume iterating when the chunk may be invalid
> + * @iter: iterator state
> + * @order: order of the entry that was just processed
>   *
> - * If the iterator needs to release then reacquire a lock, the chunk may
> - * have been invalidated by an insertion or deletion.  Call this function
> + * If the iteration needs to release then reacquire a lock, the chunk may
> + * be invalidated by an insertion or deletion.  Call this function
>   * before releasing the lock to continue the iteration from the next index.
>   */
> -void **__must_check radix_tree_iter_next(void **slot,
> -                                       struct radix_tree_iter *iter);
> +static inline void **__must_check
> +radix_tree_iter_save(struct radix_tree_iter *iter, unsigned order)
> +{
> +       iter->next_index = round_up(iter->index, 1 << order);
> +       iter->index = 0;
> +       iter->tags = 0;
> +       return NULL;
> +}
>
>  /**
>   * radix_tree_chunk_size - get current chunk size
> @@ -467,7 +474,7 @@ static inline void ** __radix_tree_next_slot(void **slot,
>   * For tagged lookup it also eats @iter->tags.
>   *
>   * There are several cases where 'slot' can be passed in as NULL to this
> - * function.  These cases result from the use of radix_tree_iter_next() or
> + * function.  These cases result from the use of radix_tree_iter_save() or
>   * radix_tree_iter_retry().  In these cases we don't end up dereferencing
>   * 'slot' because either:
>   * a) we are doing tagged iteration and iter->tags has been set to 0, or
> diff --git a/lib/radix-tree.c b/lib/radix-tree.c
> index 7e2469b..fcbadad 100644
> --- a/lib/radix-tree.c
> +++ b/lib/radix-tree.c
> @@ -1245,6 +1245,7 @@ static inline void __set_iter_shift(struct radix_tree_iter *iter,
>  #endif
>  }
>
> +#ifdef CONFIG_RADIX_TREE_MULTIORDER
>  static void ** __radix_tree_iter_next(struct radix_tree_node **nodep,
>                         void **slot, struct radix_tree_iter *iter)
>  {
> @@ -1263,7 +1264,6 @@ static void ** __radix_tree_iter_next(struct radix_tree_node **nodep,
>         return NULL;
>  }
>
> -#ifdef CONFIG_RADIX_TREE_MULTIORDER
>  void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
>                                         unsigned flags)
>  {
> @@ -1321,20 +1321,6 @@ void ** __radix_tree_next_slot(void **slot, struct radix_tree_iter *iter,
>  EXPORT_SYMBOL(__radix_tree_next_slot);
>  #endif
>
> -void **radix_tree_iter_next(void **slot, struct radix_tree_iter *iter)
> -{
> -       struct radix_tree_node *node;
> -
> -       slot++;
> -       iter->index = __radix_tree_iter_add(iter, 1);
> -       node = rcu_dereference_raw(*slot);
> -       __radix_tree_iter_next(&node, slot, iter);
> -       iter->next_index = iter->index;
> -       iter->tags = 0;
> -       return NULL;
> -}
> -EXPORT_SYMBOL(radix_tree_iter_next);
> -
>  /**
>   * radix_tree_next_chunk - find next chunk of slots for iteration
>   *
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 46155d1..54446e6 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -1614,7 +1614,7 @@ static void khugepaged_scan_shmem(struct mm_struct *mm,
>                 present++;
>
>                 if (need_resched()) {
> -                       slot = radix_tree_iter_next(slot, &iter);
> +                       slot = radix_tree_iter_save(&iter, 0);
>                         cond_resched_rcu();
>                 }
>         }
> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
> index c593715..7d6b870 100644
> --- a/mm/page-writeback.c
> +++ b/mm/page-writeback.c
> @@ -2119,7 +2119,7 @@ void tag_pages_for_writeback(struct address_space *mapping,
>                 tagged++;
>                 if ((tagged % WRITEBACK_TAG_BATCH) != 0)
>                         continue;
> -               slot = radix_tree_iter_next(slot, &iter);
> +               slot = radix_tree_iter_save(&iter, compound_order(*slot));
>                 spin_unlock_irq(&mapping->tree_lock);
>                 cond_resched();
>                 spin_lock_irq(&mapping->tree_lock);
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 8f9c9aa..3f2d07a 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -644,6 +644,7 @@ unsigned long shmem_partial_swap_usage(struct address_space *mapping,
>         rcu_read_lock();
>
>         radix_tree_for_each_slot(slot, &mapping->page_tree, &iter, start) {
> +               unsigned int order = 0;
>                 if (iter.index >= end)
>                         break;
>
> @@ -656,9 +657,11 @@ unsigned long shmem_partial_swap_usage(struct address_space *mapping,
>
>                 if (radix_tree_exceptional_entry(page))
>                         swapped++;
> +               else
> +                       order = compound_order(page);
>
>                 if (need_resched()) {
> -                       slot = radix_tree_iter_next(slot, &iter);
> +                       slot = radix_tree_iter_save(&iter, order);
>                         cond_resched_rcu();
>                 }
>         }
> @@ -1062,7 +1065,7 @@ static unsigned long find_swap_entry(struct radix_tree_root *root, void *item)
>                 checked++;
>                 if ((checked % 4096) != 0)
>                         continue;
> -               slot = radix_tree_iter_next(slot, &iter);
> +               slot = radix_tree_iter_save(&iter, 0);
>                 cond_resched_rcu();
>         }
>
> @@ -2444,21 +2447,25 @@ static void shmem_tag_pins(struct address_space *mapping)
>         rcu_read_lock();
>
>         radix_tree_for_each_slot(slot, &mapping->page_tree, &iter, start) {
> +               unsigned int order = 0;
>                 page = radix_tree_deref_slot(slot);
>                 if (!page || radix_tree_exception(page)) {
>                         if (radix_tree_deref_retry(page)) {
>                                 slot = radix_tree_iter_retry(&iter);
>                                 continue;
>                         }
> -               } else if (page_count(page) - page_mapcount(page) > 1) {
> -                       spin_lock_irq(&mapping->tree_lock);
> -                       radix_tree_tag_set(&mapping->page_tree, iter.index,
> -                                          SHMEM_TAG_PINNED);
> -                       spin_unlock_irq(&mapping->tree_lock);
> +               } else {
> +                       order = compound_order(page);
> +                       if (page_count(page) - page_mapcount(page) > 1) {
> +                               spin_lock_irq(&mapping->tree_lock);
> +                               radix_tree_tag_set(&mapping->page_tree,
> +                                               iter.index, SHMEM_TAG_PINNED);
> +                               spin_unlock_irq(&mapping->tree_lock);
> +                       }
>                 }
>
>                 if (need_resched()) {
> -                       slot = radix_tree_iter_next(slot, &iter);
> +                       slot = radix_tree_iter_save(&iter, order);
>                         cond_resched_rcu();
>                 }
>         }
> @@ -2528,7 +2535,10 @@ static int shmem_wait_for_pins(struct address_space *mapping)
>                         spin_unlock_irq(&mapping->tree_lock);
>  continue_resched:
>                         if (need_resched()) {
> -                               slot = radix_tree_iter_next(slot, &iter);
> +                               unsigned int order = 0;
> +                               if (page)
> +                                       order = compound_order(page);
> +                               slot = radix_tree_iter_save(&iter, order);
>                                 cond_resched_rcu();
>                         }
>                 }

RE: [PATCH 20/29] radix tree: Improve multiorder iterators

[Matthew Wilcox]

From: Konstantin Khlebnikov [mailto:koct9i@gmail.com]
> On Fri, Nov 18, 2016 at 7:31 PM, Matthew Wilcox <mawilcox@microsoft.com>
> wrote:
> > I think what you're suggesting is that we introduce a new API:
> >
> >  slot = radix_tree_iter_save(&iter, order);
> >
> > where the caller tells us the order of the entry it just consumed.  Or maybe
> you're suggesting
> >
> >  slot = radix_tree_iter_advance(&iter, newindex)
> 
> Yes, someting like that.
> 
> >
> > which would allow us to skip to any index.  Although ... isn't that just
> radix_tree_iter_init()?
> 
> Iterator could keep pointer to current node and reuse it for next
> iteration if possible.

The point of this API is that it's never possible, because we're about to drop the lock and allow other users to modify the tree.  Actually, it is different from radix_tree_iter_init() because it has to set ->tags to 0 and ->index == ->next_index in order to get through a call to radix_tree_next_slot().

> > It does push a bit of complexity onto the callers.  We have 7 callers of
> > radix_tree_iter_next() in my current tree (after applying this patch set, so
> > range_tag_if_tagged and locate_item have been pushed into their callers):
> > btrfs, kugepaged, page-writeback and shmem.  btrfs knows its objects occupy
> > one slot.  khugepaged knows that its page is order 0 at the time it calls
> > radix_tree_iter_next().  Page-writeback has a struct page and can simply use
> > compound_order().  It's shmem where things get sticky, although it's all
> > solvable with some temporary variables.
> 
> Users who work only with single slot enties don't get any complications,
> all other already manage these multiorder entries somehow.

If you read the patch below, you'll see that most callers don't need to be concerned with the size of the entry they're looking at.  I'll trim away the trivial ones so it's easier to see my point.

It's not a huge amount of code in each caller, but is this a burden we really want to push onto the callers when we could handle it behind the interface?

diff --git a/mm/shmem.c b/mm/shmem.c
index 8f9c9aa..90dd18d9 100644
--- a/mm/shmem.c
+++ b/mm/shmem.c
@@ -658,7 +658,10 @@ unsigned long shmem_partial_swap_usage(struct address_space *mapping,
 			swapped++;
 
 		if (need_resched()) {
-			slot = radix_tree_iter_next(slot, &iter);
+			unsigned int order = 0;
+			if (!radix_tree_exceptional_entry(page))
+				order = compound_order(page);
+			slot = radix_tree_iter_save(&iter, order);
 			cond_resched_rcu();
 		}
 	}
@@ -2450,6 +2453,7 @@ static void shmem_tag_pins(struct address_space *mapping)
 				slot = radix_tree_iter_retry(&iter);
 				continue;
 			}
+			page = NULL;
 		} else if (page_count(page) - page_mapcount(page) > 1) {
 			spin_lock_irq(&mapping->tree_lock);
 			radix_tree_tag_set(&mapping->page_tree, iter.index,
@@ -2458,7 +2462,8 @@ static void shmem_tag_pins(struct address_space *mapping)
 		}
 
 		if (need_resched()) {
-			slot = radix_tree_iter_next(slot, &iter);
+			unsigned int order = page ? compound_order(page) : 0;
+			slot = radix_tree_iter_save(&iter, order);
 			cond_resched_rcu();
 		}
 	}
@@ -2528,7 +2533,10 @@ static int shmem_wait_for_pins(struct address_space *mapping)
 			spin_unlock_irq(&mapping->tree_lock);
 continue_resched:
 			if (need_resched()) {
-				slot = radix_tree_iter_next(slot, &iter);
+				unsigned int order = 0;
+				if (page)
+					order = compound_order(page);
+				slot = radix_tree_iter_save(&iter, order);
 				cond_resched_rcu();
 			}
 		}