I have updated patches based on the review feedback from Andrew.
I have tested this on
128(64p) ppc64 sles
4(2p) ppc64 debian
4(2p) x86_64 ubuntu-gutsy
Updated patches are at
http://www.radian.org/~kvaneesh/ext4/jan-24-2008/
http://www.radian.org/~kvaneesh/ext4/jan-24-2008/patches.tar
Diff for reference
diff --git a/Documentation/filesystems/ext4.txt b/Documentation/filesystems/ext4.txt
index 4f329af..ec7d349 100644
--- a/Documentation/filesystems/ext4.txt
+++ b/Documentation/filesystems/ext4.txt
@@ -89,6 +89,8 @@ When mounting an ext4 filesystem, the following option are accepted:
extents ext4 will use extents to address file data. The
file system will no longer be mountable by ext3.
+noextents ext4 will not use extents for new files created.
+
journal_checksum Enable checksumming of the journal transactions.
This will allow the recovery code in e2fsck and the
kernel to detect corruption in the kernel. It is a
@@ -206,6 +208,10 @@ nobh (a) cache disk block mapping information
"nobh" option tries to avoid associating buffer
heads (supported only for "writeback" mode).
+mballoc (*) Use the mutliblock allocator for block allocation
+nomballoc disabled multiblock allocator for block allocation.
+stripe=n filesystem blocks per stripe for a RAID configuration.
+
Data Mode
---------
diff --git a/Documentation/filesystems/proc.txt b/Documentation/filesystems/proc.txt
index dec9945..4413a2d 100644
--- a/Documentation/filesystems/proc.txt
+++ b/Documentation/filesystems/proc.txt
@@ -857,6 +857,45 @@ CPUs.
The "procs_blocked" line gives the number of processes currently blocked,
waiting for I/O to complete.
+1.9 Ext4 file system parameters
+------------------------------
+Ext4 file system have one directory per partition under /proc/fs/ext4/
+# ls /proc/fs/ext4/hdc/
+group_prealloc max_to_scan mb_groups mb_history min_to_scan order2_req
+stats stream_req
+
+mb_groups:
+This file gives the details of mutiblock allocator buddy cache of free blocks
+
+mb_history:
+Multiblock allocation history.
+
+stats:
+This file indicate whether the multiblock allocator should start collecting
+statistics. The statistics are shown during unmount
+
+group_prealloc:
+The multiblock allocator normalize the block allocation request to
+group_prealloc filesystem blocks if we don't have strip value set.
+The stripe value can be specified at mount time or during mke2fs.
+
+max_to_scan:
+How long multiblock allocator can look for a best extent (in found extents)
+
+min_to_scan:
+How long multiblock allocator must look for a best extent
+
+order2_req:
+Multiblock allocator use 2^N search using buddies only for requests greater
+than or equal to order2_req. The request size is specfied in file system
+blocks. A value of 2 indicate only if the requests are greater than or equal
+to 4 blocks.
+
+stream_req:
+Files smaller than stream_req are served by the stream allocator, whose
+purpose is to pack requests as close each to other as possible to
+produce smooth I/O traffic. Avalue of 16 indicate that file smaller than 16
+filesystem block size will use group based preallocation.
------------------------------------------------------------------------------
Summary
diff --git a/fs/buffer.c b/fs/buffer.c
index 982cf1a..921eeec 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -3232,19 +3232,21 @@ int bh_uptodate_or_lock(struct buffer_head *bh)
return 1;
}
EXPORT_SYMBOL(bh_uptodate_or_lock);
+
/**
* bh_submit_read: Submit a locked buffer for reading
* @bh: struct buffer_head
*
- * Returns a negative error
+ * Returns zero on success and -EIO on error.
*/
int bh_submit_read(struct buffer_head *bh)
{
- if (!buffer_locked(bh))
- lock_buffer(bh);
+ BUG_ON(!buffer_locked(bh));
- if (buffer_uptodate(bh))
+ if (buffer_uptodate(bh)) {
+ unlock_buffer(bh);
return 0;
+ }
get_bh(bh);
bh->b_end_io = end_buffer_read_sync;
@@ -3255,6 +3257,7 @@ int bh_submit_read(struct buffer_head *bh)
return -EIO;
}
EXPORT_SYMBOL(bh_submit_read);
+
void __init buffer_init(void)
{
int nrpages;
diff --git a/fs/ext4/defrag.c b/fs/ext4/defrag.c
index 4ef3dc0..0d76c74 100644
--- a/fs/ext4/defrag.c
+++ b/fs/ext4/defrag.c
@@ -30,14 +30,6 @@ ext4_fsblk_t idx_pblock(struct ext4_extent_idx *ix)
return block;
}
-/* Will go away */
-static void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
-{
- ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
- ex->ee_start_hi =
- cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
-}
-
/*
* this structure is used to gather extents from the tree via ioctl
*/
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index cbda084..c2caf97 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -75,7 +75,7 @@ static ext4_fsblk_t idx_pblock(struct ext4_extent_idx *ix)
* stores a large physical block number into an extent struct,
* breaking it into parts
*/
-static void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
+void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
{
ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index e348ceb..bec699a 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -435,7 +435,7 @@ struct ext4_free_metadata {
struct ext4_group_info {
unsigned long bb_state;
- unsigned long bb_tid;
+ unsigned long bb_tid;
struct ext4_free_metadata *bb_md_cur;
unsigned short bb_first_free;
unsigned short bb_free;
@@ -489,7 +489,7 @@ struct ext4_free_extent {
*/
struct ext4_locality_group {
/* for allocator */
- struct semaphore lg_sem; /* to serialize allocates */
+ struct mutex lg_mutex; /* to serialize allocates */
struct list_head lg_prealloc_list;/* list of preallocations */
spinlock_t lg_prealloc_lock;
};
@@ -564,7 +564,10 @@ struct ext4_buddy {
#define EXT4_MB_BUDDY(e4b) ((e4b)->bd_buddy)
#ifndef EXT4_MB_HISTORY
-#define ext4_mb_store_history(ac)
+static inline void ext4_mb_store_history(struct ext4_allocation_context *ac)
+{
+ return;
+}
#else
static void ext4_mb_store_history(struct ext4_allocation_context *ac);
#endif
@@ -642,6 +645,10 @@ static ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb,
static inline int mb_test_bit(int bit, void *addr)
{
+ /*
+ * ext4_test_bit on architecture like powerpc
+ * needs unsigned long aligned address
+ */
mb_correct_addr_and_bit(bit, addr);
return ext4_test_bit(bit, addr);
}
@@ -670,7 +677,7 @@ static inline void mb_clear_bit_atomic(spinlock_t *lock, int bit, void *addr)
ext4_clear_bit_atomic(lock, bit, addr);
}
-static inline void *mb_find_buddy(struct ext4_buddy *e4b, int order, int *max)
+static void *mb_find_buddy(struct ext4_buddy *e4b, int order, int *max)
{
char *bb;
@@ -753,9 +760,20 @@ static void mb_cmp_bitmaps(struct ext4_buddy *e4b, void *bitmap)
}
#else
-#define mb_free_blocks_double(a, b, c, d)
-#define mb_mark_used_double(a, b, c)
-#define mb_cmp_bitmaps(a, b)
+static inline void mb_free_blocks_double(struct inode *inode,
+ struct ext4_buddy *e4b, int first, int count)
+{
+ return;
+}
+static inline void mb_mark_used_double(struct ext4_buddy *e4b,
+ int first, int count)
+{
+ return;
+}
+static inline void mb_cmp_bitmaps(struct ext4_buddy *e4b, void *bitmap)
+{
+ return;
+}
#endif
#ifdef AGGRESSIVE_CHECK
@@ -878,26 +896,6 @@ static int __mb_check_buddy(struct ext4_buddy *e4b, char *file,
#define mb_check_buddy(e4b)
#endif
-/* find most significant bit */
-static int fmsb(unsigned short word)
-{
- int order;
-
- if (word > 255) {
- order = 7;
- word >>= 8;
- } else {
- order = -1;
- }
-
- do {
- order++;
- word >>= 1;
- } while (word != 0);
-
- return order;
-}
-
/* FIXME!! need more doc */
static void ext4_mb_mark_free_simple(struct super_block *sb,
void *buddy, unsigned first, int len,
@@ -918,7 +916,7 @@ static void ext4_mb_mark_free_simple(struct super_block *sb,
max = ffs(first | border) - 1;
/* find how many blocks of power 2 we need to mark */
- min = fmsb(len);
+ min = fls(len) - 1;
if (max < min)
min = max;
@@ -1030,10 +1028,9 @@ static int ext4_mb_init_cache(struct page *page, char *incore)
if (groups_per_page > 1) {
err = -ENOMEM;
i = sizeof(struct buffer_head *) * groups_per_page;
- bh = kmalloc(i, GFP_NOFS);
+ bh = kzalloc(i, GFP_NOFS);
if (bh == NULL)
goto out;
- memset(bh, 0, i);
} else
bh = &bhs;
@@ -1056,15 +1053,9 @@ static int ext4_mb_init_cache(struct page *page, char *incore)
if (bh[i] == NULL)
goto out;
- if (buffer_uptodate(bh[i]))
+ if (bh_uptodate_or_lock(bh[i]))
continue;
- lock_buffer(bh[i]);
- if (buffer_uptodate(bh[i])) {
- unlock_buffer(bh[i]);
- continue;
- }
-
if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
ext4_init_block_bitmap(sb, bh[i],
first_group + i, desc);
@@ -1303,7 +1294,7 @@ static void mb_set_bits(spinlock_t *lock, void *bm, int cur, int len)
len = cur + len;
while (cur < len) {
if ((cur & 31) == 0 && (len - cur) >= 32) {
- /* fast path: clear whole word at once */
+ /* fast path: set whole word at once */
addr = bm + (cur >> 3);
*addr = 0xffffffff;
cur += 32;
@@ -2681,7 +2672,7 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
for (i = 0; i < NR_CPUS; i++) {
struct ext4_locality_group *lg;
lg = &sbi->s_locality_groups[i];
- sema_init(&lg->lg_sem, 1);
+ mutex_init(&lg->lg_mutex);
INIT_LIST_HEAD(&lg->lg_prealloc_list);
spin_lock_init(&lg->lg_prealloc_lock);
}
@@ -2693,6 +2684,7 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
return 0;
}
+/* need to called with ext4 group lock (ext4_lock_group) */
static void ext4_mb_cleanup_pa(struct ext4_group_info *grp)
{
struct ext4_prealloc_space *pa;
@@ -2701,7 +2693,7 @@ static void ext4_mb_cleanup_pa(struct ext4_group_info *grp)
list_for_each_safe(cur, tmp, &grp->bb_prealloc_list) {
pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list);
- list_del_rcu(&pa->pa_group_list);
+ list_del(&pa->pa_group_list);
count++;
kfree(pa);
}
@@ -2735,7 +2727,9 @@ int ext4_mb_release(struct super_block *sb)
#ifdef DOUBLE_CHECK
kfree(grinfo->bb_bitmap);
#endif
+ ext4_lock_group(sb,i);
ext4_mb_cleanup_pa(grinfo);
+ ext4_lock_group(sb,i);
kfree(grinfo);
}
num_meta_group_infos = (sbi->s_groups_count +
@@ -3447,6 +3441,7 @@ static int ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
/*
* the function goes through all preallocation in this group and marks them
* used in in-core bitmap. buddy must be generated from this bitmap
+ * Need to be called with ext4 group lock (ext4_lock_group)
*/
static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
ext4_group_t group)
@@ -3468,7 +3463,7 @@ static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
* allocation in buddy when concurrent ext4_mb_put_pa()
* is dropping preallocation
*/
- list_for_each_rcu(cur, &grp->bb_prealloc_list) {
+ list_for_each(cur, &grp->bb_prealloc_list) {
pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list);
spin_lock(&pa->pa_lock);
ext4_get_group_no_and_offset(sb, pa->pa_pstart,
@@ -3492,7 +3487,6 @@ static void ext4_mb_pa_callback(struct rcu_head *head)
pa = container_of(head, struct ext4_prealloc_space, u.pa_rcu);
kmem_cache_free(ext4_pspace_cachep, pa);
}
-#define mb_call_rcu(__pa) call_rcu(&(__pa)->u.pa_rcu, ext4_mb_pa_callback)
/*
* drops a reference to preallocated space descriptor
@@ -3534,14 +3528,14 @@ static void ext4_mb_put_pa(struct ext4_allocation_context *ac,
* against that pair
*/
ext4_lock_group(sb, grp);
- list_del_rcu(&pa->pa_group_list);
+ list_del(&pa->pa_group_list);
ext4_unlock_group(sb, grp);
spin_lock(pa->pa_obj_lock);
list_del_rcu(&pa->pa_inode_list);
spin_unlock(pa->pa_obj_lock);
- mb_call_rcu(pa);
+ call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
}
/*
@@ -3621,7 +3615,7 @@ static int ext4_mb_new_inode_pa(struct ext4_allocation_context *ac)
pa->pa_inode = ac->ac_inode;
ext4_lock_group(sb, ac->ac_b_ex.fe_group);
- list_add_rcu(&pa->pa_group_list, &grp->bb_prealloc_list);
+ list_add(&pa->pa_group_list, &grp->bb_prealloc_list);
ext4_unlock_group(sb, ac->ac_b_ex.fe_group);
spin_lock(pa->pa_obj_lock);
@@ -3678,7 +3672,7 @@ static int ext4_mb_new_group_pa(struct ext4_allocation_context *ac)
pa->pa_inode = NULL;
ext4_lock_group(sb, ac->ac_b_ex.fe_group);
- list_add_rcu(&pa->pa_group_list, &grp->bb_prealloc_list);
+ list_add(&pa->pa_group_list, &grp->bb_prealloc_list);
ext4_unlock_group(sb, ac->ac_b_ex.fe_group);
spin_lock(pa->pa_obj_lock);
@@ -3859,7 +3853,7 @@ repeat:
spin_unlock(&pa->pa_lock);
- list_del_rcu(&pa->pa_group_list);
+ list_del(&pa->pa_group_list);
list_add(&pa->u.pa_tmp_list, &list);
}
@@ -3895,7 +3889,7 @@ repeat:
ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa);
list_del(&pa->u.pa_tmp_list);
- mb_call_rcu(pa);
+ call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
}
out:
@@ -3948,9 +3942,8 @@ repeat:
spin_unlock(&pa->pa_lock);
spin_unlock(&ei->i_prealloc_lock);
printk(KERN_ERR "uh-oh! used pa while discarding\n");
- dump_stack();
- current->state = TASK_UNINTERRUPTIBLE;
- schedule_timeout(HZ);
+ WARN_ON(1);
+ schedule_timeout_uninterruptible(HZ);
goto repeat;
}
@@ -3978,8 +3971,7 @@ repeat:
* add a flag to force wait only in case
* of ->clear_inode(), but not in case of
* regular truncate */
- current->state = TASK_UNINTERRUPTIBLE;
- schedule_timeout(HZ);
+ schedule_timeout_uninterruptible(HZ);
goto repeat;
}
spin_unlock(&ei->i_prealloc_lock);
@@ -3999,7 +3991,7 @@ repeat:
}
ext4_lock_group(sb, group);
- list_del_rcu(&pa->pa_group_list);
+ list_del(&pa->pa_group_list);
ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa);
ext4_unlock_group(sb, group);
@@ -4007,7 +3999,7 @@ repeat:
brelse(bitmap_bh);
list_del(&pa->u.pa_tmp_list);
- mb_call_rcu(pa);
+ call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
}
}
@@ -4057,7 +4049,8 @@ static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
struct ext4_prealloc_space *pa;
ext4_grpblk_t start;
struct list_head *cur;
- list_for_each_rcu(cur, &grp->bb_prealloc_list) {
+ ext4_lock_group(sb, i);
+ list_for_each(cur, &grp->bb_prealloc_list) {
pa = list_entry(cur, struct ext4_prealloc_space,
pa_group_list);
spin_lock(&pa->pa_lock);
@@ -4067,6 +4060,7 @@ static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
printk(KERN_ERR "PA:%lu:%d:%u \n", i,
start, pa->pa_len);
}
+ ext4_lock_group(sb, i);
if (grp->bb_free == 0)
continue;
@@ -4076,7 +4070,10 @@ static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
printk(KERN_ERR "\n");
}
#else
-#define ext4_mb_show_ac(x)
+static inline void ext4_mb_show_ac(struct ext4_allocation_context *ac)
+{
+ return;
+}
#endif
/*
@@ -4097,8 +4094,7 @@ static void ext4_mb_group_or_file(struct ext4_allocation_context *ac)
size = ac->ac_o_ex.fe_logical + ac->ac_o_ex.fe_len;
isize = i_size_read(ac->ac_inode) >> bsbits;
- if (size < isize)
- size = isize;
+ size = max(size, isize);
/* don't use group allocation for large files */
if (size >= sbi->s_mb_stream_request)
@@ -4108,6 +4104,11 @@ static void ext4_mb_group_or_file(struct ext4_allocation_context *ac)
return;
BUG_ON(ac->ac_lg != NULL);
+ /*
+ * locality group prealloc space are per cpu. The reason for having
+ * per cpu locality group is to reduce the contention between block
+ * request from multiple CPUs.
+ */
ac->ac_lg = &sbi->s_locality_groups[get_cpu()];
put_cpu();
@@ -4115,7 +4116,7 @@ static void ext4_mb_group_or_file(struct ext4_allocation_context *ac)
ac->ac_flags |= EXT4_MB_HINT_GROUP_ALLOC;
/* serialize all allocations in the group */
- down(&ac->ac_lg->lg_sem);
+ mutex_lock(&ac->ac_lg->lg_mutex);
}
static int ext4_mb_initialize_context(struct ext4_allocation_context *ac,
@@ -4209,7 +4210,7 @@ static int ext4_mb_release_context(struct ext4_allocation_context *ac)
if (ac->ac_buddy_page)
page_cache_release(ac->ac_buddy_page);
if (ac->ac_flags & EXT4_MB_HINT_GROUP_ALLOC)
- up(&ac->ac_lg->lg_sem);
+ mutex_unlock(&ac->ac_lg->lg_mutex);
ext4_mb_collect_stats(ac);
return 0;
}
diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c
index 6b40f55..5e9c7e8 100644
--- a/fs/ext4/migrate.c
+++ b/fs/ext4/migrate.c
@@ -16,19 +16,15 @@
#include <linux/ext4_jbd2.h>
#include <linux/ext4_fs_extents.h>
+/*
+ * The contiguous blocks details which can be
+ * represented by a single extent
+ */
struct list_blocks_struct {
ext4_lblk_t first_block, last_block;
ext4_fsblk_t first_pblock, last_pblock;
};
-/* will go away */
-static void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
-{
- ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
- ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1)
- & 0xffff);
-}
-
static int finish_range(handle_t *handle, struct inode *inode,
struct list_blocks_struct *lb)
@@ -61,15 +57,11 @@ static int finish_range(handle_t *handle, struct inode *inode,
/*
* Make sure the credit we accumalated is not really high
*/
-
if (needed && handle->h_buffer_credits >= EXT4_RESERVE_TRANS_BLOCKS) {
-
retval = ext4_journal_restart(handle, needed);
if (retval)
goto err_out;
-
}
-
if (needed) {
retval = ext4_journal_extend(handle, needed);
if (retval != 0) {
@@ -81,19 +73,17 @@ static int finish_range(handle_t *handle, struct inode *inode,
goto err_out;
}
}
-
retval = ext4_ext_insert_extent(handle, inode, path, &newext);
-
err_out:
lb->first_pblock = 0;
return retval;
}
+
static int update_extent_range(handle_t *handle, struct inode *inode,
ext4_fsblk_t pblock, ext4_lblk_t blk_num,
struct list_blocks_struct *lb)
{
int retval;
-
/*
* See if we can add on to the existing range (if it exists)
*/
@@ -112,7 +102,6 @@ static int update_extent_range(handle_t *handle, struct inode *inode,
lb->first_block = lb->last_block = blk_num;
return retval;
-
}
static int update_ind_extent_range(handle_t *handle, struct inode *inode,
@@ -136,7 +125,6 @@ static int update_ind_extent_range(handle_t *handle, struct inode *inode,
return -EIO;
i_data = (__le32 *)bh->b_data;
-
for (i = 0; i < max_entries; i++, blk_count++) {
if (i_data[i]) {
retval = update_extent_range(handle, inode,
@@ -153,6 +141,7 @@ static int update_ind_extent_range(handle_t *handle, struct inode *inode,
return retval;
}
+
static int update_dind_extent_range(handle_t *handle, struct inode *inode,
ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
struct list_blocks_struct *lb)
@@ -168,13 +157,11 @@ static int update_dind_extent_range(handle_t *handle, struct inode *inode,
*blk_nump += max_entries * max_entries;
return 0;
}
-
bh = sb_bread(inode->i_sb, pblock);
if (!bh)
return -EIO;
i_data = (__le32 *)bh->b_data;
-
for (i = 0; i < max_entries; i++) {
if (i_data[i]) {
retval = update_ind_extent_range(handle, inode,
@@ -194,6 +181,7 @@ static int update_dind_extent_range(handle_t *handle, struct inode *inode,
return retval;
}
+
static int update_tind_extent_range(handle_t *handle, struct inode *inode,
ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
struct list_blocks_struct *lb)
@@ -209,13 +197,11 @@ static int update_tind_extent_range(handle_t *handle, struct inode *inode,
*blk_nump += max_entries * max_entries * max_entries;
return 0;
}
-
bh = sb_bread(inode->i_sb, pblock);
if (!bh)
return -EIO;
i_data = (__le32 *)bh->b_data;
-
for (i = 0; i < max_entries; i++) {
if (i_data[i]) {
retval = update_dind_extent_range(handle, inode,
@@ -228,7 +214,6 @@ static int update_tind_extent_range(handle_t *handle, struct inode *inode,
blk_count += max_entries * max_entries;
}
}
-
/* Update the file block number */
*blk_nump = blk_count;
brelse(bh);
@@ -236,7 +221,6 @@ static int update_tind_extent_range(handle_t *handle, struct inode *inode,
}
-
static int free_dind_blocks(handle_t *handle,
struct inode *inode, __le32 i_data)
{
@@ -258,10 +242,7 @@ static int free_dind_blocks(handle_t *handle,
}
brelse(bh);
ext4_free_blocks(handle, inode, le32_to_cpu(i_data), 1, 1);
-
return 0;
-
-
}
static int free_tind_blocks(handle_t *handle,
@@ -277,7 +258,6 @@ static int free_tind_blocks(handle_t *handle,
return -EIO;
tmp_idata = (__le32 *)bh->b_data;
-
for (i = 0; i < max_entries; i++) {
if (tmp_idata[i]) {
retval = free_dind_blocks(handle,
@@ -290,10 +270,7 @@ static int free_tind_blocks(handle_t *handle,
}
brelse(bh);
ext4_free_blocks(handle, inode, le32_to_cpu(i_data), 1, 1);
-
return 0;
-
-
}
static int free_ind_block(handle_t *handle, struct inode *inode)
@@ -302,10 +279,8 @@ static int free_ind_block(handle_t *handle, struct inode *inode)
struct ext4_inode_info *ei = EXT4_I(inode);
if (ei->i_data[EXT4_IND_BLOCK]) {
-
ext4_free_blocks(handle, inode,
le32_to_cpu(ei->i_data[EXT4_IND_BLOCK]), 1, 1);
-
}
if (ei->i_data[EXT4_DIND_BLOCK]) {
@@ -321,17 +296,15 @@ static int free_ind_block(handle_t *handle, struct inode *inode)
if (retval)
return retval;
}
-
-
return 0;
}
+
static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
struct inode *tmp_inode, int retval)
{
struct ext4_inode_info *ei = EXT4_I(inode);
struct ext4_inode_info *tmp_ei = EXT4_I(tmp_inode);
-
retval = free_ind_block(handle, inode);
if (retval)
goto err_out;
@@ -368,9 +341,7 @@ static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
spin_unlock(&inode->i_lock);
ext4_mark_inode_dirty(handle, inode);
-
err_out:
-
return retval;
}
@@ -392,7 +363,6 @@ static int free_ext_idx(handle_t *handle, struct inode *inode,
struct buffer_head *bh;
struct ext4_extent_header *eh;
-
block = idx_pblock(ix);
bh = sb_bread(inode->i_sb, block);
if (!bh)
@@ -400,24 +370,19 @@ static int free_ext_idx(handle_t *handle, struct inode *inode,
eh = (struct ext4_extent_header *)bh->b_data;
if (eh->eh_depth == 0) {
-
brelse(bh);
ext4_free_blocks(handle, inode, block, 1, 1);
-
} else {
-
ix = EXT_FIRST_INDEX(eh);
for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ix++) {
retval = free_ext_idx(handle, inode, ix);
if (retval)
return retval;
}
-
}
-
return retval;
-
}
+
/*
* Free the extent meta data blocks only
*/
@@ -439,10 +404,10 @@ static int free_ext_block(handle_t *handle, struct inode *inode)
if (retval)
return retval;
}
-
return retval;
}
+
int ext4_ext_migrate(struct inode *inode, struct file *filp,
unsigned int cmd, unsigned long arg)
{
@@ -455,7 +420,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
struct list_blocks_struct lb;
unsigned long max_entries;
-
if (!test_opt(inode->i_sb, EXTENTS)) {
/*
* if mounted with noextents
@@ -468,8 +432,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
return -EINVAL;
down_write(&EXT4_I(inode)->i_data_sem);
-
-
handle = ext4_journal_start(inode,
EXT4_DATA_TRANS_BLOCKS(inode->i_sb) +
EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
@@ -479,18 +441,15 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
retval = PTR_ERR(handle);
goto err_out;
}
-
tmp_inode = ext4_new_inode(handle,
inode->i_sb->s_root->d_inode,
S_IFREG);
-
if (IS_ERR(tmp_inode)) {
retval = -ENOMEM;
ext4_journal_stop(handle);
tmp_inode = NULL;
goto err_out;
}
-
i_size_write(tmp_inode, i_size_read(inode));
/*
* We don't want the inode to be reclaimed
@@ -523,7 +482,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
*/
handle = ext4_journal_start(inode, 1);
for (i = 0; i < EXT4_NDIR_BLOCKS; i++, blk_count++) {
-
if (i_data[i]) {
retval = update_extent_range(handle, tmp_inode,
le32_to_cpu(i_data[i]),
@@ -532,7 +490,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
goto err_out;
}
}
-
if (i_data[EXT4_IND_BLOCK]) {
retval = update_ind_extent_range(handle, tmp_inode,
le32_to_cpu(i_data[EXT4_IND_BLOCK]),
@@ -542,7 +499,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
} else {
blk_count += max_entries;
}
-
if (i_data[EXT4_DIND_BLOCK]) {
retval = update_dind_extent_range(handle, tmp_inode,
le32_to_cpu(i_data[EXT4_DIND_BLOCK]),
@@ -552,8 +508,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
} else {
blk_count += max_entries * max_entries;
}
-
-
if (i_data[EXT4_TIND_BLOCK]) {
retval = update_tind_extent_range(handle, tmp_inode,
le32_to_cpu(i_data[EXT4_TIND_BLOCK]),
@@ -561,12 +515,10 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
if (retval)
goto err_out;
}
-
/*
* Build the last extent
*/
retval = finish_range(handle, tmp_inode, &lb);
-
err_out:
/*
* We are either freeing extent information or indirect
@@ -577,14 +529,12 @@ err_out:
*
* FIXME!! we may be touching bitmaps in different block groups.
*/
-
if (ext4_journal_extend(handle,
4 + 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb)) != 0) {
ext4_journal_restart(handle,
4 + 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb));
}
-
if (retval) {
/*
* Failure case delete the extent information with the
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index cf2f612..416d919 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -1841,13 +1841,14 @@ static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
unsigned long stripe_width =
le32_to_cpu(sbi->s_es->s_raid_stripe_width);
- if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) {
+ if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
return sbi->s_stripe;
- } else if (stripe_width <= sbi->s_blocks_per_group) {
+
+ if (stripe_width <= sbi->s_blocks_per_group)
return stripe_width;
- } else if (stride <= sbi->s_blocks_per_group) {
+
+ if (stride <= sbi->s_blocks_per_group)
return stride;
- }
return 0;
}
diff --git a/include/linux/ext4_fs_extents.h b/include/linux/ext4_fs_extents.h
index be4ada4..8bece0e 100644
--- a/include/linux/ext4_fs_extents.h
+++ b/include/linux/ext4_fs_extents.h
@@ -225,6 +225,7 @@ static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
(le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
}
+extern void ext4_ext_store_pblock(struct ext4_extent *, ext4_fsblk_t);
extern int ext4_extent_tree_init(handle_t *, struct inode *);
extern int ext4_ext_calc_credits_for_insert(struct inode *, struct ext4_ext_path *);
extern int ext4_ext_try_to_merge(struct inode *inode,
On Jan 24, 2008 20:20 +0530, Aneesh Kumar K.V wrote:
> @@ -89,6 +89,8 @@ When mounting an ext4 filesystem, the following option are accepted:
> extents ext4 will use extents to address file data. The
> file system will no longer be mountable by ext3.
>
> +noextents ext4 will not use extents for new files created.
> +
s/new files created/newly created files/
> journal_checksum Enable checksumming of the journal transactions.
> This will allow the recovery code in e2fsck and the
> kernel to detect corruption in the kernel. It is a
> @@ -206,6 +208,10 @@ nobh (a) cache disk block mapping information
> "nobh" option tries to avoid associating buffer
> heads (supported only for "writeback" mode).
>
> +mballoc (*) Use the mutliblock allocator for block allocation
> +nomballoc disabled multiblock allocator for block allocation.
> +stripe=n filesystem blocks per stripe for a RAID configuration.
Please provide a more verbose description of what a "stripe" is, since the
RAID terminology is sadly vague. Something like "number of filesystem blocks
that mballoc will try to use for allocation size and alignment. For RAID5/6
systems this should be the number of data disks * number of filesystem blocks
per data disk."
> @@ -3948,9 +3942,8 @@ repeat:
> spin_unlock(&pa->pa_lock);
> spin_unlock(&ei->i_prealloc_lock);
> printk(KERN_ERR "uh-oh! used pa while discarding\n");
> - dump_stack();
> - current->state = TASK_UNINTERRUPTIBLE;
> - schedule_timeout(HZ);
> + WARN_ON(1);
This printk and dump stack can just go away, we have removed it from our
mballoc patch as well because it was only needed for determining how often
this condition is hit and is otherwise useless.
> @@ -577,14 +529,12 @@ err_out:
> *
> * FIXME!! we may be touching bitmaps in different block groups.
> */
> -
> if (ext4_journal_extend(handle,
> 4 + 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb)) != 0) {
>
> ext4_journal_restart(handle,
> 4 + 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb));
> }
> -
There don't actually need to be braces here either.
Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.
Andreas Dilger wrote:
> On Jan 24, 2008 20:20 +0530, Aneesh Kumar K.V wrote:
>> @@ -89,6 +89,8 @@ When mounting an ext4 filesystem, the following option are accepted:
>> extents ext4 will use extents to address file data. The
>> file system will no longer be mountable by ext3.
>>
>> +noextents ext4 will not use extents for new files created.
>> +
>
> s/new files created/newly created files/
Would a blurb about keeping ext3 disk-format compatibility be worthwhile
here?
>> journal_checksum Enable checksumming of the journal transactions.
>> This will allow the recovery code in e2fsck and the
>> kernel to detect corruption in the kernel. It is a
>> @@ -206,6 +208,10 @@ nobh (a) cache disk block mapping information
>> "nobh" option tries to avoid associating buffer
>> heads (supported only for "writeback" mode).
>>
>> +mballoc (*) Use the mutliblock allocator for block allocation
speeeeling on "mutliblock" too :)
-Eric
On Thu, 2008-01-24 at 20:20 +0530, Aneesh Kumar K.V wrote:
> I have updated patches based on the review feedback from Andrew.
>
> I have tested this on
> 128(64p) ppc64 sles
> 4(2p) ppc64 debian
> 4(2p) x86_64 ubuntu-gutsy
>
> Updated patches are at
> http://www.radian.org/~kvaneesh/ext4/jan-24-2008/
> http://www.radian.org/~kvaneesh/ext4/jan-24-2008/patches.tar
>
Thanks, updated ext4 patch queue with your changes. And fixed the
checkpatch warnings with mballoc-core.patch.
Mingming
> Diff for reference
>
> diff --git a/Documentation/filesystems/ext4.txt b/Documentation/filesystems/ext4.txt
> index 4f329af..ec7d349 100644
> --- a/Documentation/filesystems/ext4.txt
> +++ b/Documentation/filesystems/ext4.txt
> @@ -89,6 +89,8 @@ When mounting an ext4 filesystem, the following option are accepted:
> extents ext4 will use extents to address file data. The
> file system will no longer be mountable by ext3.
>
> +noextents ext4 will not use extents for new files created.
> +
> journal_checksum Enable checksumming of the journal transactions.
> This will allow the recovery code in e2fsck and the
> kernel to detect corruption in the kernel. It is a
> @@ -206,6 +208,10 @@ nobh (a) cache disk block mapping information
> "nobh" option tries to avoid associating buffer
> heads (supported only for "writeback" mode).
>
> +mballoc (*) Use the mutliblock allocator for block allocation
> +nomballoc disabled multiblock allocator for block allocation.
> +stripe=n filesystem blocks per stripe for a RAID configuration.
> +
>
> Data Mode
> ---------
> diff --git a/Documentation/filesystems/proc.txt b/Documentation/filesystems/proc.txt
> index dec9945..4413a2d 100644
> --- a/Documentation/filesystems/proc.txt
> +++ b/Documentation/filesystems/proc.txt
> @@ -857,6 +857,45 @@ CPUs.
> The "procs_blocked" line gives the number of processes currently blocked,
> waiting for I/O to complete.
>
> +1.9 Ext4 file system parameters
> +------------------------------
> +Ext4 file system have one directory per partition under /proc/fs/ext4/
> +# ls /proc/fs/ext4/hdc/
> +group_prealloc max_to_scan mb_groups mb_history min_to_scan order2_req
> +stats stream_req
> +
> +mb_groups:
> +This file gives the details of mutiblock allocator buddy cache of free blocks
> +
> +mb_history:
> +Multiblock allocation history.
> +
> +stats:
> +This file indicate whether the multiblock allocator should start collecting
> +statistics. The statistics are shown during unmount
> +
> +group_prealloc:
> +The multiblock allocator normalize the block allocation request to
> +group_prealloc filesystem blocks if we don't have strip value set.
> +The stripe value can be specified at mount time or during mke2fs.
> +
> +max_to_scan:
> +How long multiblock allocator can look for a best extent (in found extents)
> +
> +min_to_scan:
> +How long multiblock allocator must look for a best extent
> +
> +order2_req:
> +Multiblock allocator use 2^N search using buddies only for requests greater
> +than or equal to order2_req. The request size is specfied in file system
> +blocks. A value of 2 indicate only if the requests are greater than or equal
> +to 4 blocks.
> +
> +stream_req:
> +Files smaller than stream_req are served by the stream allocator, whose
> +purpose is to pack requests as close each to other as possible to
> +produce smooth I/O traffic. Avalue of 16 indicate that file smaller than 16
> +filesystem block size will use group based preallocation.
>
> ------------------------------------------------------------------------------
> Summary
> diff --git a/fs/buffer.c b/fs/buffer.c
> index 982cf1a..921eeec 100644
> --- a/fs/buffer.c
> +++ b/fs/buffer.c
> @@ -3232,19 +3232,21 @@ int bh_uptodate_or_lock(struct buffer_head *bh)
> return 1;
> }
> EXPORT_SYMBOL(bh_uptodate_or_lock);
> +
> /**
> * bh_submit_read: Submit a locked buffer for reading
> * @bh: struct buffer_head
> *
> - * Returns a negative error
> + * Returns zero on success and -EIO on error.
> */
> int bh_submit_read(struct buffer_head *bh)
> {
> - if (!buffer_locked(bh))
> - lock_buffer(bh);
> + BUG_ON(!buffer_locked(bh));
>
> - if (buffer_uptodate(bh))
> + if (buffer_uptodate(bh)) {
> + unlock_buffer(bh);
> return 0;
> + }
>
> get_bh(bh);
> bh->b_end_io = end_buffer_read_sync;
> @@ -3255,6 +3257,7 @@ int bh_submit_read(struct buffer_head *bh)
> return -EIO;
> }
> EXPORT_SYMBOL(bh_submit_read);
> +
> void __init buffer_init(void)
> {
> int nrpages;
> diff --git a/fs/ext4/defrag.c b/fs/ext4/defrag.c
> index 4ef3dc0..0d76c74 100644
> --- a/fs/ext4/defrag.c
> +++ b/fs/ext4/defrag.c
> @@ -30,14 +30,6 @@ ext4_fsblk_t idx_pblock(struct ext4_extent_idx *ix)
> return block;
> }
>
> -/* Will go away */
> -static void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
> -{
> - ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
> - ex->ee_start_hi =
> - cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
> -}
> -
> /*
> * this structure is used to gather extents from the tree via ioctl
> */
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index cbda084..c2caf97 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -75,7 +75,7 @@ static ext4_fsblk_t idx_pblock(struct ext4_extent_idx *ix)
> * stores a large physical block number into an extent struct,
> * breaking it into parts
> */
> -static void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
> +void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
> {
> ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
> ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
> index e348ceb..bec699a 100644
> --- a/fs/ext4/mballoc.c
> +++ b/fs/ext4/mballoc.c
> @@ -435,7 +435,7 @@ struct ext4_free_metadata {
>
> struct ext4_group_info {
> unsigned long bb_state;
> - unsigned long bb_tid;
> + unsigned long bb_tid;
> struct ext4_free_metadata *bb_md_cur;
> unsigned short bb_first_free;
> unsigned short bb_free;
> @@ -489,7 +489,7 @@ struct ext4_free_extent {
> */
> struct ext4_locality_group {
> /* for allocator */
> - struct semaphore lg_sem; /* to serialize allocates */
> + struct mutex lg_mutex; /* to serialize allocates */
> struct list_head lg_prealloc_list;/* list of preallocations */
> spinlock_t lg_prealloc_lock;
> };
> @@ -564,7 +564,10 @@ struct ext4_buddy {
> #define EXT4_MB_BUDDY(e4b) ((e4b)->bd_buddy)
>
> #ifndef EXT4_MB_HISTORY
> -#define ext4_mb_store_history(ac)
> +static inline void ext4_mb_store_history(struct ext4_allocation_context *ac)
> +{
> + return;
> +}
> #else
> static void ext4_mb_store_history(struct ext4_allocation_context *ac);
> #endif
> @@ -642,6 +645,10 @@ static ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb,
>
> static inline int mb_test_bit(int bit, void *addr)
> {
> + /*
> + * ext4_test_bit on architecture like powerpc
> + * needs unsigned long aligned address
> + */
> mb_correct_addr_and_bit(bit, addr);
> return ext4_test_bit(bit, addr);
> }
> @@ -670,7 +677,7 @@ static inline void mb_clear_bit_atomic(spinlock_t *lock, int bit, void *addr)
> ext4_clear_bit_atomic(lock, bit, addr);
> }
>
> -static inline void *mb_find_buddy(struct ext4_buddy *e4b, int order, int *max)
> +static void *mb_find_buddy(struct ext4_buddy *e4b, int order, int *max)
> {
> char *bb;
>
> @@ -753,9 +760,20 @@ static void mb_cmp_bitmaps(struct ext4_buddy *e4b, void *bitmap)
> }
>
> #else
> -#define mb_free_blocks_double(a, b, c, d)
> -#define mb_mark_used_double(a, b, c)
> -#define mb_cmp_bitmaps(a, b)
> +static inline void mb_free_blocks_double(struct inode *inode,
> + struct ext4_buddy *e4b, int first, int count)
> +{
> + return;
> +}
> +static inline void mb_mark_used_double(struct ext4_buddy *e4b,
> + int first, int count)
> +{
> + return;
> +}
> +static inline void mb_cmp_bitmaps(struct ext4_buddy *e4b, void *bitmap)
> +{
> + return;
> +}
> #endif
>
> #ifdef AGGRESSIVE_CHECK
> @@ -878,26 +896,6 @@ static int __mb_check_buddy(struct ext4_buddy *e4b, char *file,
> #define mb_check_buddy(e4b)
> #endif
>
> -/* find most significant bit */
> -static int fmsb(unsigned short word)
> -{
> - int order;
> -
> - if (word > 255) {
> - order = 7;
> - word >>= 8;
> - } else {
> - order = -1;
> - }
> -
> - do {
> - order++;
> - word >>= 1;
> - } while (word != 0);
> -
> - return order;
> -}
> -
> /* FIXME!! need more doc */
> static void ext4_mb_mark_free_simple(struct super_block *sb,
> void *buddy, unsigned first, int len,
> @@ -918,7 +916,7 @@ static void ext4_mb_mark_free_simple(struct super_block *sb,
> max = ffs(first | border) - 1;
>
> /* find how many blocks of power 2 we need to mark */
> - min = fmsb(len);
> + min = fls(len) - 1;
>
> if (max < min)
> min = max;
> @@ -1030,10 +1028,9 @@ static int ext4_mb_init_cache(struct page *page, char *incore)
> if (groups_per_page > 1) {
> err = -ENOMEM;
> i = sizeof(struct buffer_head *) * groups_per_page;
> - bh = kmalloc(i, GFP_NOFS);
> + bh = kzalloc(i, GFP_NOFS);
> if (bh == NULL)
> goto out;
> - memset(bh, 0, i);
> } else
> bh = &bhs;
>
> @@ -1056,15 +1053,9 @@ static int ext4_mb_init_cache(struct page *page, char *incore)
> if (bh[i] == NULL)
> goto out;
>
> - if (buffer_uptodate(bh[i]))
> + if (bh_uptodate_or_lock(bh[i]))
> continue;
>
> - lock_buffer(bh[i]);
> - if (buffer_uptodate(bh[i])) {
> - unlock_buffer(bh[i]);
> - continue;
> - }
> -
> if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
> ext4_init_block_bitmap(sb, bh[i],
> first_group + i, desc);
> @@ -1303,7 +1294,7 @@ static void mb_set_bits(spinlock_t *lock, void *bm, int cur, int len)
> len = cur + len;
> while (cur < len) {
> if ((cur & 31) == 0 && (len - cur) >= 32) {
> - /* fast path: clear whole word at once */
> + /* fast path: set whole word at once */
> addr = bm + (cur >> 3);
> *addr = 0xffffffff;
> cur += 32;
> @@ -2681,7 +2672,7 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
> for (i = 0; i < NR_CPUS; i++) {
> struct ext4_locality_group *lg;
> lg = &sbi->s_locality_groups[i];
> - sema_init(&lg->lg_sem, 1);
> + mutex_init(&lg->lg_mutex);
> INIT_LIST_HEAD(&lg->lg_prealloc_list);
> spin_lock_init(&lg->lg_prealloc_lock);
> }
> @@ -2693,6 +2684,7 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
> return 0;
> }
>
> +/* need to called with ext4 group lock (ext4_lock_group) */
> static void ext4_mb_cleanup_pa(struct ext4_group_info *grp)
> {
> struct ext4_prealloc_space *pa;
> @@ -2701,7 +2693,7 @@ static void ext4_mb_cleanup_pa(struct ext4_group_info *grp)
>
> list_for_each_safe(cur, tmp, &grp->bb_prealloc_list) {
> pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list);
> - list_del_rcu(&pa->pa_group_list);
> + list_del(&pa->pa_group_list);
> count++;
> kfree(pa);
> }
> @@ -2735,7 +2727,9 @@ int ext4_mb_release(struct super_block *sb)
> #ifdef DOUBLE_CHECK
> kfree(grinfo->bb_bitmap);
> #endif
> + ext4_lock_group(sb,i);
> ext4_mb_cleanup_pa(grinfo);
> + ext4_lock_group(sb,i);
> kfree(grinfo);
> }
> num_meta_group_infos = (sbi->s_groups_count +
> @@ -3447,6 +3441,7 @@ static int ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
> /*
> * the function goes through all preallocation in this group and marks them
> * used in in-core bitmap. buddy must be generated from this bitmap
> + * Need to be called with ext4 group lock (ext4_lock_group)
> */
> static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
> ext4_group_t group)
> @@ -3468,7 +3463,7 @@ static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
> * allocation in buddy when concurrent ext4_mb_put_pa()
> * is dropping preallocation
> */
> - list_for_each_rcu(cur, &grp->bb_prealloc_list) {
> + list_for_each(cur, &grp->bb_prealloc_list) {
> pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list);
> spin_lock(&pa->pa_lock);
> ext4_get_group_no_and_offset(sb, pa->pa_pstart,
> @@ -3492,7 +3487,6 @@ static void ext4_mb_pa_callback(struct rcu_head *head)
> pa = container_of(head, struct ext4_prealloc_space, u.pa_rcu);
> kmem_cache_free(ext4_pspace_cachep, pa);
> }
> -#define mb_call_rcu(__pa) call_rcu(&(__pa)->u.pa_rcu, ext4_mb_pa_callback)
>
> /*
> * drops a reference to preallocated space descriptor
> @@ -3534,14 +3528,14 @@ static void ext4_mb_put_pa(struct ext4_allocation_context *ac,
> * against that pair
> */
> ext4_lock_group(sb, grp);
> - list_del_rcu(&pa->pa_group_list);
> + list_del(&pa->pa_group_list);
> ext4_unlock_group(sb, grp);
>
> spin_lock(pa->pa_obj_lock);
> list_del_rcu(&pa->pa_inode_list);
> spin_unlock(pa->pa_obj_lock);
>
> - mb_call_rcu(pa);
> + call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
> }
>
> /*
> @@ -3621,7 +3615,7 @@ static int ext4_mb_new_inode_pa(struct ext4_allocation_context *ac)
> pa->pa_inode = ac->ac_inode;
>
> ext4_lock_group(sb, ac->ac_b_ex.fe_group);
> - list_add_rcu(&pa->pa_group_list, &grp->bb_prealloc_list);
> + list_add(&pa->pa_group_list, &grp->bb_prealloc_list);
> ext4_unlock_group(sb, ac->ac_b_ex.fe_group);
>
> spin_lock(pa->pa_obj_lock);
> @@ -3678,7 +3672,7 @@ static int ext4_mb_new_group_pa(struct ext4_allocation_context *ac)
> pa->pa_inode = NULL;
>
> ext4_lock_group(sb, ac->ac_b_ex.fe_group);
> - list_add_rcu(&pa->pa_group_list, &grp->bb_prealloc_list);
> + list_add(&pa->pa_group_list, &grp->bb_prealloc_list);
> ext4_unlock_group(sb, ac->ac_b_ex.fe_group);
>
> spin_lock(pa->pa_obj_lock);
> @@ -3859,7 +3853,7 @@ repeat:
>
> spin_unlock(&pa->pa_lock);
>
> - list_del_rcu(&pa->pa_group_list);
> + list_del(&pa->pa_group_list);
> list_add(&pa->u.pa_tmp_list, &list);
> }
>
> @@ -3895,7 +3889,7 @@ repeat:
> ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa);
>
> list_del(&pa->u.pa_tmp_list);
> - mb_call_rcu(pa);
> + call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
> }
>
> out:
> @@ -3948,9 +3942,8 @@ repeat:
> spin_unlock(&pa->pa_lock);
> spin_unlock(&ei->i_prealloc_lock);
> printk(KERN_ERR "uh-oh! used pa while discarding\n");
> - dump_stack();
> - current->state = TASK_UNINTERRUPTIBLE;
> - schedule_timeout(HZ);
> + WARN_ON(1);
> + schedule_timeout_uninterruptible(HZ);
> goto repeat;
>
> }
> @@ -3978,8 +3971,7 @@ repeat:
> * add a flag to force wait only in case
> * of ->clear_inode(), but not in case of
> * regular truncate */
> - current->state = TASK_UNINTERRUPTIBLE;
> - schedule_timeout(HZ);
> + schedule_timeout_uninterruptible(HZ);
> goto repeat;
> }
> spin_unlock(&ei->i_prealloc_lock);
> @@ -3999,7 +3991,7 @@ repeat:
> }
>
> ext4_lock_group(sb, group);
> - list_del_rcu(&pa->pa_group_list);
> + list_del(&pa->pa_group_list);
> ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa);
> ext4_unlock_group(sb, group);
>
> @@ -4007,7 +3999,7 @@ repeat:
> brelse(bitmap_bh);
>
> list_del(&pa->u.pa_tmp_list);
> - mb_call_rcu(pa);
> + call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
> }
> }
>
> @@ -4057,7 +4049,8 @@ static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
> struct ext4_prealloc_space *pa;
> ext4_grpblk_t start;
> struct list_head *cur;
> - list_for_each_rcu(cur, &grp->bb_prealloc_list) {
> + ext4_lock_group(sb, i);
> + list_for_each(cur, &grp->bb_prealloc_list) {
> pa = list_entry(cur, struct ext4_prealloc_space,
> pa_group_list);
> spin_lock(&pa->pa_lock);
> @@ -4067,6 +4060,7 @@ static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
> printk(KERN_ERR "PA:%lu:%d:%u \n", i,
> start, pa->pa_len);
> }
> + ext4_lock_group(sb, i);
>
> if (grp->bb_free == 0)
> continue;
> @@ -4076,7 +4070,10 @@ static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
> printk(KERN_ERR "\n");
> }
> #else
> -#define ext4_mb_show_ac(x)
> +static inline void ext4_mb_show_ac(struct ext4_allocation_context *ac)
> +{
> + return;
> +}
> #endif
>
> /*
> @@ -4097,8 +4094,7 @@ static void ext4_mb_group_or_file(struct ext4_allocation_context *ac)
>
> size = ac->ac_o_ex.fe_logical + ac->ac_o_ex.fe_len;
> isize = i_size_read(ac->ac_inode) >> bsbits;
> - if (size < isize)
> - size = isize;
> + size = max(size, isize);
>
> /* don't use group allocation for large files */
> if (size >= sbi->s_mb_stream_request)
> @@ -4108,6 +4104,11 @@ static void ext4_mb_group_or_file(struct ext4_allocation_context *ac)
> return;
>
> BUG_ON(ac->ac_lg != NULL);
> + /*
> + * locality group prealloc space are per cpu. The reason for having
> + * per cpu locality group is to reduce the contention between block
> + * request from multiple CPUs.
> + */
> ac->ac_lg = &sbi->s_locality_groups[get_cpu()];
> put_cpu();
>
> @@ -4115,7 +4116,7 @@ static void ext4_mb_group_or_file(struct ext4_allocation_context *ac)
> ac->ac_flags |= EXT4_MB_HINT_GROUP_ALLOC;
>
> /* serialize all allocations in the group */
> - down(&ac->ac_lg->lg_sem);
> + mutex_lock(&ac->ac_lg->lg_mutex);
> }
>
> static int ext4_mb_initialize_context(struct ext4_allocation_context *ac,
> @@ -4209,7 +4210,7 @@ static int ext4_mb_release_context(struct ext4_allocation_context *ac)
> if (ac->ac_buddy_page)
> page_cache_release(ac->ac_buddy_page);
> if (ac->ac_flags & EXT4_MB_HINT_GROUP_ALLOC)
> - up(&ac->ac_lg->lg_sem);
> + mutex_unlock(&ac->ac_lg->lg_mutex);
> ext4_mb_collect_stats(ac);
> return 0;
> }
> diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c
> index 6b40f55..5e9c7e8 100644
> --- a/fs/ext4/migrate.c
> +++ b/fs/ext4/migrate.c
> @@ -16,19 +16,15 @@
> #include <linux/ext4_jbd2.h>
> #include <linux/ext4_fs_extents.h>
>
> +/*
> + * The contiguous blocks details which can be
> + * represented by a single extent
> + */
> struct list_blocks_struct {
> ext4_lblk_t first_block, last_block;
> ext4_fsblk_t first_pblock, last_pblock;
> };
>
> -/* will go away */
> -static void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
> -{
> - ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
> - ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1)
> - & 0xffff);
> -}
> -
> static int finish_range(handle_t *handle, struct inode *inode,
> struct list_blocks_struct *lb)
>
> @@ -61,15 +57,11 @@ static int finish_range(handle_t *handle, struct inode *inode,
> /*
> * Make sure the credit we accumalated is not really high
> */
> -
> if (needed && handle->h_buffer_credits >= EXT4_RESERVE_TRANS_BLOCKS) {
> -
> retval = ext4_journal_restart(handle, needed);
> if (retval)
> goto err_out;
> -
> }
> -
> if (needed) {
> retval = ext4_journal_extend(handle, needed);
> if (retval != 0) {
> @@ -81,19 +73,17 @@ static int finish_range(handle_t *handle, struct inode *inode,
> goto err_out;
> }
> }
> -
> retval = ext4_ext_insert_extent(handle, inode, path, &newext);
> -
> err_out:
> lb->first_pblock = 0;
> return retval;
> }
> +
> static int update_extent_range(handle_t *handle, struct inode *inode,
> ext4_fsblk_t pblock, ext4_lblk_t blk_num,
> struct list_blocks_struct *lb)
> {
> int retval;
> -
> /*
> * See if we can add on to the existing range (if it exists)
> */
> @@ -112,7 +102,6 @@ static int update_extent_range(handle_t *handle, struct inode *inode,
> lb->first_block = lb->last_block = blk_num;
>
> return retval;
> -
> }
>
> static int update_ind_extent_range(handle_t *handle, struct inode *inode,
> @@ -136,7 +125,6 @@ static int update_ind_extent_range(handle_t *handle, struct inode *inode,
> return -EIO;
>
> i_data = (__le32 *)bh->b_data;
> -
> for (i = 0; i < max_entries; i++, blk_count++) {
> if (i_data[i]) {
> retval = update_extent_range(handle, inode,
> @@ -153,6 +141,7 @@ static int update_ind_extent_range(handle_t *handle, struct inode *inode,
> return retval;
>
> }
> +
> static int update_dind_extent_range(handle_t *handle, struct inode *inode,
> ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
> struct list_blocks_struct *lb)
> @@ -168,13 +157,11 @@ static int update_dind_extent_range(handle_t *handle, struct inode *inode,
> *blk_nump += max_entries * max_entries;
> return 0;
> }
> -
> bh = sb_bread(inode->i_sb, pblock);
> if (!bh)
> return -EIO;
>
> i_data = (__le32 *)bh->b_data;
> -
> for (i = 0; i < max_entries; i++) {
> if (i_data[i]) {
> retval = update_ind_extent_range(handle, inode,
> @@ -194,6 +181,7 @@ static int update_dind_extent_range(handle_t *handle, struct inode *inode,
> return retval;
>
> }
> +
> static int update_tind_extent_range(handle_t *handle, struct inode *inode,
> ext4_fsblk_t pblock, ext4_lblk_t *blk_nump,
> struct list_blocks_struct *lb)
> @@ -209,13 +197,11 @@ static int update_tind_extent_range(handle_t *handle, struct inode *inode,
> *blk_nump += max_entries * max_entries * max_entries;
> return 0;
> }
> -
> bh = sb_bread(inode->i_sb, pblock);
> if (!bh)
> return -EIO;
>
> i_data = (__le32 *)bh->b_data;
> -
> for (i = 0; i < max_entries; i++) {
> if (i_data[i]) {
> retval = update_dind_extent_range(handle, inode,
> @@ -228,7 +214,6 @@ static int update_tind_extent_range(handle_t *handle, struct inode *inode,
> blk_count += max_entries * max_entries;
> }
> }
> -
> /* Update the file block number */
> *blk_nump = blk_count;
> brelse(bh);
> @@ -236,7 +221,6 @@ static int update_tind_extent_range(handle_t *handle, struct inode *inode,
>
> }
>
> -
> static int free_dind_blocks(handle_t *handle,
> struct inode *inode, __le32 i_data)
> {
> @@ -258,10 +242,7 @@ static int free_dind_blocks(handle_t *handle,
> }
> brelse(bh);
> ext4_free_blocks(handle, inode, le32_to_cpu(i_data), 1, 1);
> -
> return 0;
> -
> -
> }
>
> static int free_tind_blocks(handle_t *handle,
> @@ -277,7 +258,6 @@ static int free_tind_blocks(handle_t *handle,
> return -EIO;
>
> tmp_idata = (__le32 *)bh->b_data;
> -
> for (i = 0; i < max_entries; i++) {
> if (tmp_idata[i]) {
> retval = free_dind_blocks(handle,
> @@ -290,10 +270,7 @@ static int free_tind_blocks(handle_t *handle,
> }
> brelse(bh);
> ext4_free_blocks(handle, inode, le32_to_cpu(i_data), 1, 1);
> -
> return 0;
> -
> -
> }
>
> static int free_ind_block(handle_t *handle, struct inode *inode)
> @@ -302,10 +279,8 @@ static int free_ind_block(handle_t *handle, struct inode *inode)
> struct ext4_inode_info *ei = EXT4_I(inode);
>
> if (ei->i_data[EXT4_IND_BLOCK]) {
> -
> ext4_free_blocks(handle, inode,
> le32_to_cpu(ei->i_data[EXT4_IND_BLOCK]), 1, 1);
> -
> }
>
> if (ei->i_data[EXT4_DIND_BLOCK]) {
> @@ -321,17 +296,15 @@ static int free_ind_block(handle_t *handle, struct inode *inode)
> if (retval)
> return retval;
> }
> -
> -
> return 0;
> }
> +
> static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
> struct inode *tmp_inode, int retval)
> {
> struct ext4_inode_info *ei = EXT4_I(inode);
> struct ext4_inode_info *tmp_ei = EXT4_I(tmp_inode);
>
> -
> retval = free_ind_block(handle, inode);
> if (retval)
> goto err_out;
> @@ -368,9 +341,7 @@ static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
> spin_unlock(&inode->i_lock);
>
> ext4_mark_inode_dirty(handle, inode);
> -
> err_out:
> -
> return retval;
> }
>
> @@ -392,7 +363,6 @@ static int free_ext_idx(handle_t *handle, struct inode *inode,
> struct buffer_head *bh;
> struct ext4_extent_header *eh;
>
> -
> block = idx_pblock(ix);
> bh = sb_bread(inode->i_sb, block);
> if (!bh)
> @@ -400,24 +370,19 @@ static int free_ext_idx(handle_t *handle, struct inode *inode,
>
> eh = (struct ext4_extent_header *)bh->b_data;
> if (eh->eh_depth == 0) {
> -
> brelse(bh);
> ext4_free_blocks(handle, inode, block, 1, 1);
> -
> } else {
> -
> ix = EXT_FIRST_INDEX(eh);
> for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ix++) {
> retval = free_ext_idx(handle, inode, ix);
> if (retval)
> return retval;
> }
> -
> }
> -
> return retval;
> -
> }
> +
> /*
> * Free the extent meta data blocks only
> */
> @@ -439,10 +404,10 @@ static int free_ext_block(handle_t *handle, struct inode *inode)
> if (retval)
> return retval;
> }
> -
> return retval;
>
> }
> +
> int ext4_ext_migrate(struct inode *inode, struct file *filp,
> unsigned int cmd, unsigned long arg)
> {
> @@ -455,7 +420,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
> struct list_blocks_struct lb;
> unsigned long max_entries;
>
> -
> if (!test_opt(inode->i_sb, EXTENTS)) {
> /*
> * if mounted with noextents
> @@ -468,8 +432,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
> return -EINVAL;
>
> down_write(&EXT4_I(inode)->i_data_sem);
> -
> -
> handle = ext4_journal_start(inode,
> EXT4_DATA_TRANS_BLOCKS(inode->i_sb) +
> EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
> @@ -479,18 +441,15 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
> retval = PTR_ERR(handle);
> goto err_out;
> }
> -
> tmp_inode = ext4_new_inode(handle,
> inode->i_sb->s_root->d_inode,
> S_IFREG);
> -
> if (IS_ERR(tmp_inode)) {
> retval = -ENOMEM;
> ext4_journal_stop(handle);
> tmp_inode = NULL;
> goto err_out;
> }
> -
> i_size_write(tmp_inode, i_size_read(inode));
> /*
> * We don't want the inode to be reclaimed
> @@ -523,7 +482,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
> */
> handle = ext4_journal_start(inode, 1);
> for (i = 0; i < EXT4_NDIR_BLOCKS; i++, blk_count++) {
> -
> if (i_data[i]) {
> retval = update_extent_range(handle, tmp_inode,
> le32_to_cpu(i_data[i]),
> @@ -532,7 +490,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
> goto err_out;
> }
> }
> -
> if (i_data[EXT4_IND_BLOCK]) {
> retval = update_ind_extent_range(handle, tmp_inode,
> le32_to_cpu(i_data[EXT4_IND_BLOCK]),
> @@ -542,7 +499,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
> } else {
> blk_count += max_entries;
> }
> -
> if (i_data[EXT4_DIND_BLOCK]) {
> retval = update_dind_extent_range(handle, tmp_inode,
> le32_to_cpu(i_data[EXT4_DIND_BLOCK]),
> @@ -552,8 +508,6 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
> } else {
> blk_count += max_entries * max_entries;
> }
> -
> -
> if (i_data[EXT4_TIND_BLOCK]) {
> retval = update_tind_extent_range(handle, tmp_inode,
> le32_to_cpu(i_data[EXT4_TIND_BLOCK]),
> @@ -561,12 +515,10 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
> if (retval)
> goto err_out;
> }
> -
> /*
> * Build the last extent
> */
> retval = finish_range(handle, tmp_inode, &lb);
> -
> err_out:
> /*
> * We are either freeing extent information or indirect
> @@ -577,14 +529,12 @@ err_out:
> *
> * FIXME!! we may be touching bitmaps in different block groups.
> */
> -
> if (ext4_journal_extend(handle,
> 4 + 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb)) != 0) {
>
> ext4_journal_restart(handle,
> 4 + 2*EXT4_QUOTA_TRANS_BLOCKS(inode->i_sb));
> }
> -
> if (retval) {
> /*
> * Failure case delete the extent information with the
> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
> index cf2f612..416d919 100644
> --- a/fs/ext4/super.c
> +++ b/fs/ext4/super.c
> @@ -1841,13 +1841,14 @@ static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
> unsigned long stripe_width =
> le32_to_cpu(sbi->s_es->s_raid_stripe_width);
>
> - if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) {
> + if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
> return sbi->s_stripe;
> - } else if (stripe_width <= sbi->s_blocks_per_group) {
> +
> + if (stripe_width <= sbi->s_blocks_per_group)
> return stripe_width;
> - } else if (stride <= sbi->s_blocks_per_group) {
> +
> + if (stride <= sbi->s_blocks_per_group)
> return stride;
> - }
>
> return 0;
> }
> diff --git a/include/linux/ext4_fs_extents.h b/include/linux/ext4_fs_extents.h
> index be4ada4..8bece0e 100644
> --- a/include/linux/ext4_fs_extents.h
> +++ b/include/linux/ext4_fs_extents.h
> @@ -225,6 +225,7 @@ static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
> (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
> }
>
> +extern void ext4_ext_store_pblock(struct ext4_extent *, ext4_fsblk_t);
> extern int ext4_extent_tree_init(handle_t *, struct inode *);
> extern int ext4_ext_calc_credits_for_insert(struct inode *, struct ext4_ext_path *);
> extern int ext4_ext_try_to_merge(struct inode *inode,
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html