[PATCH v4 3/3] ext4: each filesystem creates and uses its own mc_cache

From: T Makphaibulchoke
Date: Fri Jan 24 2014 - 13:32:53 EST

Next message: Shuah Khan: "[PATCH v2] staging/usbip: userspace to use linux header for usb_device_speed enum, missing speeds to speed_strings array"
Previous message: T Makphaibulchoke: "[PATCH v4 2/3] mbcache: decoupling the locking of local from global data"
In reply to: T Makphaibulchoke: "[PATCH v4 2/3] mbcache: decoupling the locking of local from global data"
Next in thread: Andi Kleen: "Re: [PATCH v4 0/3] ext4: increase mbcache scalability"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

This patch adds new interfaces to create and destory cache,
ext4_xattr_create_cache() and ext4_xattr_destroy_cache(), and remove the cache
creation and destory calls from ex4_init_xattr() and ext4_exitxattr() in
fs/ext4/xattr.c. It also increases the size of the mbcache's hash tables for
ext4.

fs/ext4/super.c has been changed so that when a filesystem is mounted a
cache is allocated and attched to its ext4_sb_info structure.

fs/mbcache.c has been changed so that only one slab allocator is allocated
and used by all mbcache structures.

Signed-off-by: T. Makphaibulchoke <tmac@xxxxxx>
---
fs/ext4/ext4.h | 1 +
fs/ext4/super.c | 24 ++++++++++++++++--------
fs/ext4/xattr.c | 51 ++++++++++++++++++++++++++++-----------------------
fs/ext4/xattr.h | 6 +++---
fs/mbcache.c | 19 +++++++++++++------
5 files changed, 61 insertions(+), 40 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index e618503..fa20aab 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1314,6 +1314,7 @@ struct ext4_sb_info {
struct list_head s_es_lru;
unsigned long s_es_last_sorted;
struct percpu_counter s_extent_cache_cnt;
+ struct mb_cache *s_mb_cache;
spinlock_t s_es_lru_lock ____cacheline_aligned_in_smp;

/* Ratelimit ext4 messages. */
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index c977f4e..176048e 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -59,6 +59,7 @@ static struct kset *ext4_kset;
static struct ext4_lazy_init *ext4_li_info;
static struct mutex ext4_li_mtx;
static struct ext4_features *ext4_feat;
+static int ext4_mballoc_ready;

static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
unsigned long journal_devnum);
@@ -845,6 +846,10 @@ static void ext4_put_super(struct super_block *sb)
invalidate_bdev(sbi->journal_bdev);
ext4_blkdev_remove(sbi);
}
+ if (sbi->s_mb_cache) {
+ ext4_xattr_destroy_cache(sbi->s_mb_cache);
+ sbi->s_mb_cache = NULL;
+ }
if (sbi->s_mmp_tsk)
kthread_stop(sbi->s_mmp_tsk);
sb->s_fs_info = NULL;
@@ -3981,6 +3986,13 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);

sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
+ if (ext4_mballoc_ready) {
+ sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
+ if (!sbi->s_mb_cache) {
+ ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
+ goto failed_mount_wq;
+ }
+ }

/*
* The journal may have updated the bg summary counts, so we
@@ -5501,11 +5513,9 @@ static int __init ext4_init_fs(void)

err = ext4_init_mballoc();
if (err)
- goto out3;
-
- err = ext4_init_xattr();
- if (err)
goto out2;
+ else
+ ext4_mballoc_ready = 1;
err = init_inodecache();
if (err)
goto out1;
@@ -5521,10 +5531,9 @@ out:
unregister_as_ext3();
destroy_inodecache();
out1:
- ext4_exit_xattr();
-out2:
+ ext4_mballoc_ready = 0;
ext4_exit_mballoc();
-out3:
+out2:
ext4_exit_feat_adverts();
out4:
if (ext4_proc_root)
@@ -5547,7 +5556,6 @@ static void __exit ext4_exit_fs(void)
unregister_as_ext3();
unregister_filesystem(&ext4_fs_type);
destroy_inodecache();
- ext4_exit_xattr();
ext4_exit_mballoc();
ext4_exit_feat_adverts();
remove_proc_entry("fs/ext4", NULL);
diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
index 1423c48..6e2788f 100644
--- a/fs/ext4/xattr.c
+++ b/fs/ext4/xattr.c
@@ -81,7 +81,7 @@
# define ea_bdebug(bh, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
#endif

-static void ext4_xattr_cache_insert(struct buffer_head *);
+static void ext4_xattr_cache_insert(struct mb_cache *, struct buffer_head *);
static struct buffer_head *ext4_xattr_cache_find(struct inode *,
struct ext4_xattr_header *,
struct mb_cache_entry **);
@@ -90,8 +90,6 @@ static void ext4_xattr_rehash(struct ext4_xattr_header *,
static int ext4_xattr_list(struct dentry *dentry, char *buffer,
size_t buffer_size);

-static struct mb_cache *ext4_xattr_cache;
-
static const struct xattr_handler *ext4_xattr_handler_map[] = {
[EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
#ifdef CONFIG_EXT4_FS_POSIX_ACL
@@ -117,6 +115,9 @@ const struct xattr_handler *ext4_xattr_handlers[] = {
NULL
};

+#define EXT4_GET_MB_CACHE(inode) (((struct ext4_sb_info *) \
+ inode->i_sb->s_fs_info)->s_mb_cache)
+
static __le32 ext4_xattr_block_csum(struct inode *inode,
sector_t block_nr,
struct ext4_xattr_header *hdr)
@@ -265,6 +266,7 @@ ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
struct ext4_xattr_entry *entry;
size_t size;
int error;
+ struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);

ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
name_index, name, buffer, (long)buffer_size);
@@ -286,7 +288,7 @@ bad_block:
error = -EIO;
goto cleanup;
}
- ext4_xattr_cache_insert(bh);
+ ext4_xattr_cache_insert(ext4_mb_cache, bh);
entry = BFIRST(bh);
error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
if (error == -EIO)
@@ -409,6 +411,7 @@ ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
struct inode *inode = dentry->d_inode;
struct buffer_head *bh = NULL;
int error;
+ struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);

ea_idebug(inode, "buffer=%p, buffer_size=%ld",
buffer, (long)buffer_size);
@@ -430,7 +433,7 @@ ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
error = -EIO;
goto cleanup;
}
- ext4_xattr_cache_insert(bh);
+ ext4_xattr_cache_insert(ext4_mb_cache, bh);
error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);

cleanup:
@@ -526,8 +529,9 @@ ext4_xattr_release_block(handle_t *handle, struct inode *inode,
{
struct mb_cache_entry *ce = NULL;
int error = 0;
+ struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);

- ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
+ ce = mb_cache_entry_get(ext4_mb_cache, bh->b_bdev, bh->b_blocknr);
error = ext4_journal_get_write_access(handle, bh);
if (error)
goto out;
@@ -745,13 +749,14 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
struct ext4_xattr_search *s = &bs->s;
struct mb_cache_entry *ce = NULL;
int error = 0;
+ struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);

#define header(x) ((struct ext4_xattr_header *)(x))

if (i->value && i->value_len > sb->s_blocksize)
return -ENOSPC;
if (s->base) {
- ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
+ ce = mb_cache_entry_get(ext4_mb_cache, bs->bh->b_bdev,
bs->bh->b_blocknr);
error = ext4_journal_get_write_access(handle, bs->bh);
if (error)
@@ -769,7 +774,8 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
if (!IS_LAST_ENTRY(s->first))
ext4_xattr_rehash(header(s->base),
s->here);
- ext4_xattr_cache_insert(bs->bh);
+ ext4_xattr_cache_insert(ext4_mb_cache,
+ bs->bh);
}
unlock_buffer(bs->bh);
if (error == -EIO)
@@ -905,7 +911,7 @@ getblk_failed:
memcpy(new_bh->b_data, s->base, new_bh->b_size);
set_buffer_uptodate(new_bh);
unlock_buffer(new_bh);
- ext4_xattr_cache_insert(new_bh);
+ ext4_xattr_cache_insert(ext4_mb_cache, new_bh);
error = ext4_handle_dirty_xattr_block(handle,
inode, new_bh);
if (error)
@@ -1495,13 +1501,13 @@ ext4_xattr_put_super(struct super_block *sb)
* Returns 0, or a negative error number on failure.
*/
static void
-ext4_xattr_cache_insert(struct buffer_head *bh)
+ext4_xattr_cache_insert(struct mb_cache *ext4_mb_cache, struct buffer_head *bh)
{
__u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
struct mb_cache_entry *ce;
int error;

- ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
+ ce = mb_cache_entry_alloc(ext4_mb_cache, GFP_NOFS);
if (!ce) {
ea_bdebug(bh, "out of memory");
return;
@@ -1573,12 +1579,13 @@ ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
{
__u32 hash = le32_to_cpu(header->h_hash);
struct mb_cache_entry *ce;
+ struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);

if (!header->h_hash)
return NULL; /* never share */
ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
again:
- ce = mb_cache_entry_find_first(ext4_xattr_cache, inode->i_sb->s_bdev,
+ ce = mb_cache_entry_find_first(ext4_mb_cache, inode->i_sb->s_bdev,
hash);
while (ce) {
struct buffer_head *bh;
@@ -1676,19 +1683,17 @@ static void ext4_xattr_rehash(struct ext4_xattr_header *header,

#undef BLOCK_HASH_SHIFT

-int __init
-ext4_init_xattr(void)
+#define HASH_BUCKET_BITS 10
+
+struct mb_cache *
+ext4_xattr_create_cache(char *name)
{
- ext4_xattr_cache = mb_cache_create("ext4_xattr", 6);
- if (!ext4_xattr_cache)
- return -ENOMEM;
- return 0;
+ return mb_cache_create(name, HASH_BUCKET_BITS);
}

-void
-ext4_exit_xattr(void)
+void ext4_xattr_destroy_cache(struct mb_cache *cache)
{
- if (ext4_xattr_cache)
- mb_cache_destroy(ext4_xattr_cache);
- ext4_xattr_cache = NULL;
+ if (cache)
+ mb_cache_destroy(cache);
}
+
diff --git a/fs/ext4/xattr.h b/fs/ext4/xattr.h
index c767dbd..b930320 100644
--- a/fs/ext4/xattr.h
+++ b/fs/ext4/xattr.h
@@ -112,9 +112,6 @@ extern void ext4_xattr_put_super(struct super_block *);
extern int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
struct ext4_inode *raw_inode, handle_t *handle);

-extern int __init ext4_init_xattr(void);
-extern void ext4_exit_xattr(void);
-
extern const struct xattr_handler *ext4_xattr_handlers[];

extern int ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
@@ -126,6 +123,9 @@ extern int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
struct ext4_xattr_info *i,
struct ext4_xattr_ibody_find *is);

+extern struct mb_cache *ext4_xattr_create_cache(char *name);
+extern void ext4_xattr_destroy_cache(struct mb_cache *);
+
#ifdef CONFIG_EXT4_FS_SECURITY
extern int ext4_init_security(handle_t *handle, struct inode *inode,
struct inode *dir, const struct qstr *qstr);
diff --git a/fs/mbcache.c b/fs/mbcache.c
index 0c4cec2..d620b7f 100644
--- a/fs/mbcache.c
+++ b/fs/mbcache.c
@@ -92,6 +92,7 @@
#define MB_CACHE_WRITER ((unsigned short)~0U >> 1)

static DECLARE_WAIT_QUEUE_HEAD(mb_cache_queue);
+static struct kmem_cache *mb_cache_kmem_cache;

MODULE_AUTHOR("Andreas Gruenbacher <a.gruenbacher@xxxxxxxxxxxx>");
MODULE_DESCRIPTION("Meta block cache (for extended attributes)");
@@ -320,11 +321,14 @@ mb_cache_create(const char *name, int bucket_bits)
goto fail;
for (n=0; n<bucket_count; 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);
- if (!cache->c_entry_cache)
- goto fail2;
+ if (!mb_cache_kmem_cache) {
+ mb_cache_kmem_cache = kmem_cache_create(name,
+ sizeof(struct mb_cache_entry), 0,
+ SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD, NULL);
+ if (!mb_cache_kmem_cache)
+ goto fail2;
+ }
+ cache->c_entry_cache = mb_cache_kmem_cache;

/*
* Set an upper limit on the number of cache entries so that the hash
@@ -438,7 +442,10 @@ mb_cache_destroy(struct mb_cache *cache)
atomic_read(&cache->c_entry_count));
}

- kmem_cache_destroy(cache->c_entry_cache);
+ if (list_empty(&mb_cache_list)) {
+ kmem_cache_destroy(mb_cache_kmem_cache);
+ mb_cache_kmem_cache = NULL;
+ }

kfree(cache->c_index_hash);
kfree(cache->c_block_hash);
--
1.7.11.3

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Next message: Shuah Khan: "[PATCH v2] staging/usbip: userspace to use linux header for usb_device_speed enum, missing speeds to speed_strings array"
Previous message: T Makphaibulchoke: "[PATCH v4 2/3] mbcache: decoupling the locking of local from global data"
In reply to: T Makphaibulchoke: "[PATCH v4 2/3] mbcache: decoupling the locking of local from global data"
Next in thread: Andi Kleen: "Re: [PATCH v4 0/3] ext4: increase mbcache scalability"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]