From: T Makphaibulchoke <tmac@hp.com>
To: viro@zeniv.linux.org.uk, tytso@mit.edu, adilger.kernel@dilger.ca,
linux-ext4@vger.kernel.org, linux-kernel@vger.kernel.org,
linux-fsdevel@vger.kernel.org
Cc: aswin@hp.com, T Makphaibulchoke <tmac@hp.com>
Subject: [PATCH v4 1/3] fs/mbcache.c change block and index hash chain to hlist_bl_node
Date: Fri, 24 Jan 2014 11:31:26 -0700 [thread overview]
Message-ID: <1390588288-66930-2-git-send-email-tmac@hp.com> (raw)
In-Reply-To: <1390588288-66930-1-git-send-email-tmac@hp.com>
This patch changes each mb_cache's both block and index hash chains from
regular linked list to hlist_bl_node, which contains a built-in lock. This is
the first step in decoupling of locks serializing accesses to mb_cache global
data and each mb_cache_entry local data.
Signed-off-by: T. Makphaibulchoke <tmac@hp.com>
---
fs/mbcache.c | 117 ++++++++++++++++++++++++++++++++----------------
include/linux/mbcache.h | 11 +++--
2 files changed, 85 insertions(+), 43 deletions(-)
diff --git a/fs/mbcache.c b/fs/mbcache.c
index e519e45..55db0da 100644
--- a/fs/mbcache.c
+++ b/fs/mbcache.c
@@ -34,9 +34,9 @@
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/sched.h>
-#include <linux/init.h>
+#include <linux/list_bl.h>
#include <linux/mbcache.h>
-
+#include <linux/init.h>
#ifdef MB_CACHE_DEBUG
# define mb_debug(f...) do { \
@@ -87,21 +87,38 @@ static LIST_HEAD(mb_cache_lru_list);
static DEFINE_SPINLOCK(mb_cache_spinlock);
static inline int
-__mb_cache_entry_is_hashed(struct mb_cache_entry *ce)
+__mb_cache_entry_is_block_hashed(struct mb_cache_entry *ce)
{
- return !list_empty(&ce->e_block_list);
+ return !hlist_bl_unhashed(&ce->e_block_list);
}
-static void
-__mb_cache_entry_unhash(struct mb_cache_entry *ce)
+static inline void
+__mb_cache_entry_unhash_block(struct mb_cache_entry *ce)
{
- if (__mb_cache_entry_is_hashed(ce)) {
- list_del_init(&ce->e_block_list);
- list_del(&ce->e_index.o_list);
- }
+ if (__mb_cache_entry_is_block_hashed(ce))
+ hlist_bl_del_init(&ce->e_block_list);
+}
+
+static inline int
+__mb_cache_entry_is_index_hashed(struct mb_cache_entry *ce)
+{
+ return !hlist_bl_unhashed(&ce->e_index.o_list);
}
+static inline void
+__mb_cache_entry_unhash_index(struct mb_cache_entry *ce)
+{
+ if (__mb_cache_entry_is_index_hashed(ce))
+ hlist_bl_del_init(&ce->e_index.o_list);
+}
+
+static inline void
+__mb_cache_entry_unhash(struct mb_cache_entry *ce)
+{
+ __mb_cache_entry_unhash_index(ce);
+ __mb_cache_entry_unhash_block(ce);
+}
static void
__mb_cache_entry_forget(struct mb_cache_entry *ce, gfp_t gfp_mask)
@@ -125,7 +142,7 @@ __mb_cache_entry_release_unlock(struct mb_cache_entry *ce)
ce->e_used -= MB_CACHE_WRITER;
ce->e_used--;
if (!(ce->e_used || ce->e_queued)) {
- if (!__mb_cache_entry_is_hashed(ce))
+ if (!__mb_cache_entry_is_block_hashed(ce))
goto forget;
mb_assert(list_empty(&ce->e_lru_list));
list_add_tail(&ce->e_lru_list, &mb_cache_lru_list);
@@ -221,18 +238,18 @@ mb_cache_create(const char *name, int bucket_bits)
cache->c_name = name;
atomic_set(&cache->c_entry_count, 0);
cache->c_bucket_bits = bucket_bits;
- cache->c_block_hash = kmalloc(bucket_count * sizeof(struct list_head),
- GFP_KERNEL);
+ cache->c_block_hash = kmalloc(bucket_count *
+ sizeof(struct hlist_bl_head), GFP_KERNEL);
if (!cache->c_block_hash)
goto fail;
for (n=0; n<bucket_count; n++)
- INIT_LIST_HEAD(&cache->c_block_hash[n]);
- cache->c_index_hash = kmalloc(bucket_count * sizeof(struct list_head),
- GFP_KERNEL);
+ INIT_HLIST_BL_HEAD(&cache->c_block_hash[n]);
+ cache->c_index_hash = kmalloc(bucket_count *
+ sizeof(struct hlist_bl_head), GFP_KERNEL);
if (!cache->c_index_hash)
goto fail;
for (n=0; n<bucket_count; n++)
- INIT_LIST_HEAD(&cache->c_index_hash[n]);
+ INIT_HLIST_BL_HEAD(&cache->c_index_hash[n]);
cache->c_entry_cache = kmem_cache_create(name,
sizeof(struct mb_cache_entry), 0,
SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD, NULL);
@@ -364,10 +381,13 @@ mb_cache_entry_alloc(struct mb_cache *cache, gfp_t gfp_flags)
return NULL;
atomic_inc(&cache->c_entry_count);
INIT_LIST_HEAD(&ce->e_lru_list);
- INIT_LIST_HEAD(&ce->e_block_list);
+ INIT_HLIST_BL_NODE(&ce->e_block_list);
+ INIT_HLIST_BL_NODE(&ce->e_index.o_list);
ce->e_cache = cache;
ce->e_queued = 0;
}
+ ce->e_block_hash_p = &cache->c_block_hash[0];
+ ce->e_index_hash_p = &cache->c_index_hash[0];
ce->e_used = 1 + MB_CACHE_WRITER;
return ce;
}
@@ -393,25 +413,32 @@ mb_cache_entry_insert(struct mb_cache_entry *ce, struct block_device *bdev,
{
struct mb_cache *cache = ce->e_cache;
unsigned int bucket;
- struct list_head *l;
+ struct hlist_bl_node *l;
int error = -EBUSY;
+ struct hlist_bl_head *block_hash_p;
+ struct hlist_bl_head *index_hash_p;
+ struct mb_cache_entry *lce;
+ mb_assert(ce);
bucket = hash_long((unsigned long)bdev + (block & 0xffffffff),
cache->c_bucket_bits);
+ block_hash_p = &cache->c_block_hash[bucket];
spin_lock(&mb_cache_spinlock);
- list_for_each_prev(l, &cache->c_block_hash[bucket]) {
- struct mb_cache_entry *ce =
- list_entry(l, struct mb_cache_entry, e_block_list);
- if (ce->e_bdev == bdev && ce->e_block == block)
+ hlist_bl_for_each_entry(lce, l, block_hash_p, e_block_list) {
+ if (lce->e_bdev == bdev && lce->e_block == block)
goto out;
}
+ mb_assert(!__mb_cache_entry_is_block_hashed(ce));
__mb_cache_entry_unhash(ce);
ce->e_bdev = bdev;
ce->e_block = block;
- list_add(&ce->e_block_list, &cache->c_block_hash[bucket]);
+ ce->e_block_hash_p = block_hash_p;
ce->e_index.o_key = key;
bucket = hash_long(key, cache->c_bucket_bits);
- list_add(&ce->e_index.o_list, &cache->c_index_hash[bucket]);
+ index_hash_p = &cache->c_index_hash[bucket];
+ ce->e_index_hash_p = index_hash_p;
+ hlist_bl_add_head(&ce->e_index.o_list, index_hash_p);
+ hlist_bl_add_head(&ce->e_block_list, block_hash_p);
error = 0;
out:
spin_unlock(&mb_cache_spinlock);
@@ -463,14 +490,16 @@ mb_cache_entry_get(struct mb_cache *cache, struct block_device *bdev,
sector_t block)
{
unsigned int bucket;
- struct list_head *l;
+ struct hlist_bl_node *l;
struct mb_cache_entry *ce;
+ struct hlist_bl_head *block_hash_p;
bucket = hash_long((unsigned long)bdev + (block & 0xffffffff),
cache->c_bucket_bits);
+ block_hash_p = &cache->c_block_hash[bucket];
spin_lock(&mb_cache_spinlock);
- list_for_each(l, &cache->c_block_hash[bucket]) {
- ce = list_entry(l, struct mb_cache_entry, e_block_list);
+ hlist_bl_for_each_entry(ce, l, block_hash_p, e_block_list) {
+ mb_assert(ce->e_block_hash_p == block_hash_p);
if (ce->e_bdev == bdev && ce->e_block == block) {
DEFINE_WAIT(wait);
@@ -489,7 +518,7 @@ mb_cache_entry_get(struct mb_cache *cache, struct block_device *bdev,
finish_wait(&mb_cache_queue, &wait);
ce->e_used += 1 + MB_CACHE_WRITER;
- if (!__mb_cache_entry_is_hashed(ce)) {
+ if (!__mb_cache_entry_is_block_hashed(ce)) {
__mb_cache_entry_release_unlock(ce);
return NULL;
}
@@ -506,12 +535,14 @@ cleanup:
#if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0)
static struct mb_cache_entry *
-__mb_cache_entry_find(struct list_head *l, struct list_head *head,
+__mb_cache_entry_find(struct hlist_bl_node *l, struct hlist_bl_head *head,
struct block_device *bdev, unsigned int key)
{
- while (l != head) {
+ while (l != NULL) {
struct mb_cache_entry *ce =
- list_entry(l, struct mb_cache_entry, e_index.o_list);
+ hlist_bl_entry(l, struct mb_cache_entry,
+ e_index.o_list);
+ mb_assert(ce->e_index_hash_p == head);
if (ce->e_bdev == bdev && ce->e_index.o_key == key) {
DEFINE_WAIT(wait);
@@ -532,7 +563,7 @@ __mb_cache_entry_find(struct list_head *l, struct list_head *head,
}
finish_wait(&mb_cache_queue, &wait);
- if (!__mb_cache_entry_is_hashed(ce)) {
+ if (!__mb_cache_entry_is_block_hashed(ce)) {
__mb_cache_entry_release_unlock(ce);
spin_lock(&mb_cache_spinlock);
return ERR_PTR(-EAGAIN);
@@ -562,12 +593,16 @@ mb_cache_entry_find_first(struct mb_cache *cache, struct block_device *bdev,
unsigned int key)
{
unsigned int bucket = hash_long(key, cache->c_bucket_bits);
- struct list_head *l;
- struct mb_cache_entry *ce;
+ struct hlist_bl_node *l;
+ struct mb_cache_entry *ce = NULL;
+ struct hlist_bl_head *index_hash_p;
+ index_hash_p = &cache->c_index_hash[bucket];
spin_lock(&mb_cache_spinlock);
- l = cache->c_index_hash[bucket].next;
- ce = __mb_cache_entry_find(l, &cache->c_index_hash[bucket], bdev, key);
+ if (!hlist_bl_empty(index_hash_p)) {
+ l = hlist_bl_first(index_hash_p);
+ ce = __mb_cache_entry_find(l, index_hash_p, bdev, key);
+ }
spin_unlock(&mb_cache_spinlock);
return ce;
}
@@ -597,12 +632,16 @@ mb_cache_entry_find_next(struct mb_cache_entry *prev,
{
struct mb_cache *cache = prev->e_cache;
unsigned int bucket = hash_long(key, cache->c_bucket_bits);
- struct list_head *l;
+ struct hlist_bl_node *l;
struct mb_cache_entry *ce;
+ struct hlist_bl_head *index_hash_p;
+ index_hash_p = &cache->c_index_hash[bucket];
+ mb_assert(prev->e_index_hash_p == index_hash_p);
spin_lock(&mb_cache_spinlock);
+ mb_assert(!hlist_bl_empty(index_hash_p));
l = prev->e_index.o_list.next;
- ce = __mb_cache_entry_find(l, &cache->c_index_hash[bucket], bdev, key);
+ ce = __mb_cache_entry_find(l, index_hash_p, bdev, key);
__mb_cache_entry_release_unlock(prev);
return ce;
}
diff --git a/include/linux/mbcache.h b/include/linux/mbcache.h
index 5525d37..e76cc0e 100644
--- a/include/linux/mbcache.h
+++ b/include/linux/mbcache.h
@@ -11,11 +11,14 @@ struct mb_cache_entry {
unsigned short e_queued;
struct block_device *e_bdev;
sector_t e_block;
- struct list_head e_block_list;
+ struct hlist_bl_node e_block_list;
struct {
- struct list_head o_list;
+ struct hlist_bl_node o_list;
unsigned int o_key;
} e_index;
+ struct hlist_bl_head *e_block_hash_p;
+ struct hlist_bl_head *e_index_hash_p;
+ spinlock_t e_entry_lock;
};
struct mb_cache {
@@ -25,8 +28,8 @@ struct mb_cache {
int c_max_entries;
int c_bucket_bits;
struct kmem_cache *c_entry_cache;
- struct list_head *c_block_hash;
- struct list_head *c_index_hash;
+ struct hlist_bl_head *c_block_hash;
+ struct hlist_bl_head *c_index_hash;
};
/* Functions on caches */
--
1.7.11.3
next prev parent reply other threads:[~2014-01-24 18:31 UTC|newest]
Thread overview: 63+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-07-18 0:55 [PATCH 0/2] ext4: increase mbcache scalability T Makphaibulchoke
2013-08-22 15:54 ` [PATCH v2 " T Makphaibulchoke
2013-08-22 15:54 ` [PATCH v2 1/2] mbcache: decoupling the locking of local from global data T Makphaibulchoke
2013-08-22 16:53 ` Linus Torvalds
2013-08-22 15:33 ` Thavatchai Makphaibulchoke
2013-08-22 15:54 ` [PATCH v2 2/2] ext4: each filesystem creates and uses its own mc_cache T Makphaibulchoke
2013-08-22 15:54 ` T Makphaibulchoke
2014-01-24 18:31 ` [PATCH v4 0/3] ext4: increase mbcache scalability T Makphaibulchoke
2014-01-24 18:31 ` T Makphaibulchoke [this message]
2014-01-24 18:31 ` [PATCH v4 2/3] mbcache: decoupling the locking of local from global data T Makphaibulchoke
2014-01-24 18:31 ` [PATCH v4 3/3] ext4: each filesystem creates and uses its own mc_cache T Makphaibulchoke
2014-01-24 21:38 ` [PATCH v4 0/3] ext4: increase mbcache scalability Andi Kleen
2014-01-25 1:13 ` Thavatchai Makphaibulchoke
2014-01-25 6:09 ` Andreas Dilger
2014-01-27 12:27 ` Thavatchai Makphaibulchoke
2014-02-09 19:46 ` Thavatchai Makphaibulchoke
2014-02-11 19:58 ` Thavatchai Makphaibulchoke
2014-02-13 2:01 ` Andreas Dilger
2013-09-04 16:39 ` [PATCH v3 0/2] " T Makphaibulchoke
2013-09-04 16:39 ` [PATCH v3 1/2] mbcache: decoupling the locking of local from global data T Makphaibulchoke
2013-10-30 13:27 ` Theodore Ts'o
2013-10-30 14:42 ` Theodore Ts'o
2013-10-30 17:32 ` Thavatchai Makphaibulchoke
2013-09-04 16:39 ` [PATCH v3 2/2] ext4: each filesystem creates and uses its own mb_cache T Makphaibulchoke
2013-10-30 14:30 ` Theodore Ts'o
2013-09-04 20:00 ` [PATCH v3 0/2] ext4: increase mbcache scalability Andreas Dilger
2013-09-04 15:33 ` Thavatchai Makphaibulchoke
2013-09-05 15:22 ` Thavatchai Makphaibulchoke
2013-09-05 2:35 ` Theodore Ts'o
2013-09-05 9:49 ` Thavatchai Makphaibulchoke
2013-09-06 5:10 ` Andreas Dilger
2013-09-06 12:23 ` Thavatchai Makphaibulchoke
2013-09-10 20:47 ` Andreas Dilger
2013-09-10 21:02 ` Theodore Ts'o
2013-09-10 17:10 ` Thavatchai Makphaibulchoke
2013-09-11 3:13 ` Eric Sandeen
2013-09-11 11:30 ` Theodore Ts'o
2013-09-11 16:49 ` Eric Sandeen
2013-09-11 19:33 ` Eric Sandeen
2013-09-11 20:32 ` David Lang
2013-09-11 20:48 ` Eric Sandeen
2013-09-11 21:25 ` Theodore Ts'o
2013-09-11 20:36 ` Thavatchai Makphaibulchoke
2013-09-12 3:42 ` Eric Sandeen
2013-09-12 12:23 ` Theodore Ts'o
2013-09-13 12:04 ` Thavatchai Makphaibulchoke
2013-09-13 18:59 ` Eric Sandeen
2013-09-17 16:43 ` Thavatchai Makphaibulchoke
2013-09-18 14:19 ` Eric Sandeen
2013-09-21 18:53 ` Theodore Ts'o
2013-09-23 15:35 ` Thavatchai Makphaibulchoke
[not found] ` <D59BF31F-934D-4DE6-9E5C-F92014F09C68@dilger.ca>
2013-09-25 18:43 ` Thavatchai Makphaibulchoke
2013-10-04 3:22 ` Andreas Dilger
2013-10-18 12:43 ` Thavatchai Makphaibulchoke
2013-10-19 23:06 ` Eric Sandeen
2014-02-20 18:37 ` [PATCH V5 0/3] " T Makphaibulchoke
2014-02-20 18:37 ` [PATCH V5 1/3] fs/mbcache.c change block and index hash chain to hlist_bl_node T Makphaibulchoke
2014-02-20 18:37 ` [PATCH V5 2/3] mbcache: decoupling the locking of local from global data T Makphaibulchoke
2014-02-20 18:37 ` [PATCH V5 3/3] ext4: each filesystem creates and uses its own mb_cache T Makphaibulchoke
2014-03-12 16:19 ` [PATCH v5 RESEND 0/3] ext4: increase mbcache scalability T Makphaibulchoke
2014-03-12 16:19 ` [PATCH v5 RESEND 1/3] fs/mbcache.c change block and index hash chain to hlist_bl_node T Makphaibulchoke
2014-03-12 16:19 ` [PATCH v5 RESEND 2/3] mbcache: decoupling the locking of local from global data T Makphaibulchoke
2014-03-12 16:19 ` [PATCH v5 RESEND 3/3] ext4: each filesystem creates and uses its own mb_cache T Makphaibulchoke
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1390588288-66930-2-git-send-email-tmac@hp.com \
--to=tmac@hp.com \
--cc=adilger.kernel@dilger.ca \
--cc=aswin@hp.com \
--cc=linux-ext4@vger.kernel.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=tytso@mit.edu \
--cc=viro@zeniv.linux.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.