Index: linux-2.6.9-full/include/linux/ext3_fs.h
===================================================================
--- linux-2.6.9-full.orig/include/linux/ext3_fs.h 2007-06-08 23:44:08.000000000 +0400
-+++ linux-2.6.9-full/include/linux/ext3_fs.h 2007-07-19 09:39:07.000000000 +0400
++++ linux-2.6.9-full/include/linux/ext3_fs.h 2007-10-17 22:25:01.000000000 +0400
@@ -57,6 +57,30 @@ struct statfs;
#define ext3_debug(f, a...) do {} while (0)
#endif
Index: linux-2.6.9-full/include/linux/ext3_fs_sb.h
===================================================================
--- linux-2.6.9-full.orig/include/linux/ext3_fs_sb.h 2007-06-08 23:44:07.000000000 +0400
-+++ linux-2.6.9-full/include/linux/ext3_fs_sb.h 2007-07-19 09:39:07.000000000 +0400
++++ linux-2.6.9-full/include/linux/ext3_fs_sb.h 2007-10-17 22:25:01.000000000 +0400
@@ -81,6 +81,61 @@ struct ext3_sb_info {
char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */
int s_jquota_fmt; /* Format of quota to use */
Index: linux-2.6.9-full/fs/ext3/super.c
===================================================================
--- linux-2.6.9-full.orig/fs/ext3/super.c 2007-06-08 23:44:08.000000000 +0400
-+++ linux-2.6.9-full/fs/ext3/super.c 2007-07-19 09:39:07.000000000 +0400
++++ linux-2.6.9-full/fs/ext3/super.c 2007-10-17 22:26:27.000000000 +0400
@@ -394,6 +394,7 @@ void ext3_put_super (struct super_block
struct ext3_super_block *es = sbi->s_es;
int i;
journal_destroy(sbi->s_journal);
@@ -463,6 +464,8 @@ static struct inode *ext3_alloc_inode(st
ei->vfs_inode.i_version = 1;
-
+
memset(&ei->i_cached_extent, 0, sizeof(ei->i_cached_extent));
+ INIT_LIST_HEAD(&ei->i_prealloc_list);
+ spin_lock_init(&ei->i_prealloc_lock);
int ext3_prep_san_write(struct inode *inode, long *blocks,
Index: linux-2.6.9-full/fs/ext3/mballoc.c
===================================================================
---- linux-2.6.9-full.orig/fs/ext3/mballoc.c 2007-07-14 04:24:39.138985848 +0400
-+++ linux-2.6.9-full/fs/ext3/mballoc.c 2007-07-20 11:31:03.000000000 +0400
-@@ -0,0 +1,4392 @@
+--- linux-2.6.9-full.orig/fs/ext3/mballoc.c 2007-10-17 21:59:51.072534980 +0400
++++ linux-2.6.9-full/fs/ext3/mballoc.c 2007-10-17 23:09:22.000000000 +0400
+@@ -0,0 +1,4398 @@
+/*
+ * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
+ * Written by Alex Tomas <alex@clusterfs.com>
+ ext2_set_bit(bit, addr);
+}
+
-+static inline void mb_set_bit_atomic(int bit, void *addr)
++static inline void mb_set_bit_atomic(spinlock_t *lock, int bit, void *addr)
+{
+ mb_correct_addr_and_bit(bit,addr);
-+ ext2_set_bit_atomic(NULL, bit, addr);
++ ext2_set_bit_atomic(lock, bit, addr);
+}
+
+static inline void mb_clear_bit(int bit, void *addr)
+ ext2_clear_bit(bit, addr);
+}
+
-+static inline void mb_clear_bit_atomic(int bit, void *addr)
++static inline void mb_clear_bit_atomic(spinlock_t *lock, int bit, void *addr)
+{
+ mb_correct_addr_and_bit(bit,addr);
-+ ext2_clear_bit_atomic(NULL, bit, addr);
++ ext2_clear_bit_atomic(lock, bit, addr);
+}
+
+static inline int mb_find_next_zero_bit(void *addr, int max, int start)
+ return 0;
+}
+
-+static inline void mb_clear_bits(void *bm, int cur, int len)
++static inline void mb_clear_bits(spinlock_t *lock, void *bm, int cur, int len)
+{
+ __u32 *addr;
+
+ cur += 32;
+ continue;
+ }
-+ mb_clear_bit_atomic(cur, bm);
++ mb_clear_bit_atomic(lock, cur, bm);
+ cur++;
+ }
+}
+
-+static inline void mb_set_bits(void *bm, int cur, int len)
++static inline void mb_set_bits(spinlock_t *lock, void *bm, int cur, int len)
+{
+ __u32 *addr;
+
+ cur += 32;
+ continue;
+ }
-+ mb_set_bit_atomic(cur, bm);
++ mb_set_bit_atomic(lock, cur, bm);
+ cur++;
+ }
+}
+ e3b->bd_info->bb_counters[ord]++;
+ }
+
-+ mb_set_bits(EXT3_MB_BITMAP(e3b), ex->fe_start, len0);
++ mb_set_bits(sb_bgl_lock(EXT3_SB(e3b->bd_sb), ex->fe_group),
++ EXT3_MB_BITMAP(e3b), ex->fe_start, len0);
+ mb_check_buddy(e3b);
+
+ return ret;
+ struct ext3_free_extent *gex = &ac->ac_g_ex;
+
+ BUG_ON(ex->fe_len <= 0);
-+ BUG_ON(ex->fe_len >= (1 << ac->ac_sb->s_blocksize_bits) * 8);
-+ BUG_ON(ex->fe_start >= (1 << ac->ac_sb->s_blocksize_bits) * 8);
++ BUG_ON(ex->fe_len >= EXT3_BLOCKS_PER_GROUP(ac->ac_sb));
++ BUG_ON(ex->fe_start >= EXT3_BLOCKS_PER_GROUP(ac->ac_sb));
+ BUG_ON(ac->ac_status != AC_STATUS_CONTINUE);
+
+ ac->ac_found++;
+ i = e3b->bd_info->bb_first_free;
+
+ while (free && ac->ac_status == AC_STATUS_CONTINUE) {
-+ i = mb_find_next_zero_bit(bitmap, sb->s_blocksize * 8, i);
-+ if (i >= sb->s_blocksize * 8) {
++ i = mb_find_next_zero_bit(bitmap, EXT3_BLOCKS_PER_GROUP(sb), i);
++ if (i >= EXT3_BLOCKS_PER_GROUP(sb)) {
+ BUG_ON(free != 0);
+ break;
+ }
+ i = (i - le32_to_cpu(sbi->s_es->s_first_data_block))
+ % EXT3_BLOCKS_PER_GROUP(sb);
+
-+ while (i < sb->s_blocksize * 8) {
++ while (i < EXT3_BLOCKS_PER_GROUP(sb)) {
+ if (!mb_test_bit(i, bitmap)) {
+ max = mb_find_extent(e3b, 0, i, sbi->s_stripe, &ex);
+ if (max >= sbi->s_stripe) {
+ spin_lock_init(&sbi->s_mb_history_lock);
+ i = sbi->s_mb_history_max * sizeof(struct ext3_mb_history);
+ sbi->s_mb_history = kmalloc(i, GFP_KERNEL);
-+ if (likely(sbi->s_mb_history != NULL))
-+ memset(sbi->s_mb_history, 0, i);
++ if (likely(sbi->s_mb_history != NULL))
++ memset(sbi->s_mb_history, 0, i);
+ /* if we can't allocate history, then we simple won't use it */
+}
+
+ sbi->s_mb_order2_reqs = MB_DEFAULT_ORDER2_REQS;
+ sbi->s_mb_history_filter = EXT3_MB_HISTORY_DEFAULT;
+
-+ i = sizeof(struct ext3_locality_group) * NR_CPUS;
++ i = sizeof(struct ext3_locality_group) * num_possible_cpus();
+ sbi->s_locality_groups = kmalloc(i, GFP_NOFS);
+ if (sbi->s_locality_groups == NULL) {
+ clear_opt(sbi->s_mount_opt, MBALLOC);
+ kfree(sbi->s_mb_maxs);
+ return -ENOMEM;
+ }
-+ for (i = 0; i < NR_CPUS; i++) {
++ for (i = 0; i < num_possible_cpus(); i++) {
+ struct ext3_locality_group *lg;
+ lg = &sbi->s_locality_groups[i];
+ sema_init(&lg->lg_sem, 1);
+ }
+ }
+#endif
-+ mb_set_bits(bitmap_bh->b_data, ac->ac_b_ex.fe_start, ac->ac_b_ex.fe_len);
++ mb_set_bits(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group), bitmap_bh->b_data,
++ ac->ac_b_ex.fe_start, ac->ac_b_ex.fe_len);
+
+ spin_lock(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group));
+ gdp->bg_free_blocks_count =
+ struct ext3_inode_info *ei = EXT3_I(ac->ac_inode);
+ loff_t start, end, size, orig_size, orig_start;
+ struct list_head *cur;
-+ int bsbits;
++ int bsbits, max;
+
+ /* do normalize only data requests, metadata requests
+ do not need preallocation */
+ if (size < i_size_read(ac->ac_inode))
+ size = i_size_read(ac->ac_inode);
+
++ /* max available blocks in a free group */
++ max = EXT3_BLOCKS_PER_GROUP(ac->ac_sb) - 1 - 1
++ - EXT3_SB(ac->ac_sb)->s_itb_per_group;
++
++#define NRL_CHECK_SIZE(req,size,max,bits) \
++ (req <= (size) || max <= ((size) >> bits))
++
+ /* first, try to predict filesize */
+ /* XXX: should this table be tunable? */
+ start = 0;
+ size = 512 * 1024;
+ } else if (size <= 1024 * 1024) {
+ size = 1024 * 1024;
-+ } else if (size < 4 * 1024 * 1024) {
++ } else if (NRL_CHECK_SIZE(size, 4 * 1024 * 1024, max, bsbits)) {
+ start = ac->ac_o_ex.fe_logical << bsbits;
+ start = (start / (1024 * 1024)) * (1024 * 1024);
+ size = 1024 * 1024;
-+ } else if (size < 8 * 1024 * 1024) {
++ } else if (NRL_CHECK_SIZE(size, 8 * 1024 * 1024, max, bsbits)) {
+ start = ac->ac_o_ex.fe_logical << bsbits;
+ start = (start / (4 * (1024 * 1024))) * 4 * (1024 * 1024);
+ size = 4 * 1024 * 1024;
-+ } else if (ac->ac_o_ex.fe_len < ((8 << 20) >> bsbits)) {
++ } else if(NRL_CHECK_SIZE(ac->ac_o_ex.fe_len,(8<<20)>>bsbits,max,bsbits)){
+ start = ac->ac_o_ex.fe_logical;
+ start = start << bsbits;
+ start = (start / (8 * (1024 * 1024))) * 8 * (1024 * 1024);
+ ext3_get_group_no_and_offset(sb, pa->pa_pstart, &groupnr, &start);
+ len = pa->pa_len;
+ spin_unlock(&pa->pa_lock);
-+ BUG_ON(groupnr != group);
-+ mb_set_bits(bitmap, start, len);
++ if (unlikely(len == 0))
++ continue;
++ BUG_ON(groupnr != group && len != 0);
++ mb_set_bits(sb_bgl_lock(EXT3_SB(sb), group), bitmap, start,len);
+ preallocated += len;
+ count++;
+ }
+ BUG_ON(ac->ac_status != AC_STATUS_FOUND);
+ BUG_ON(!S_ISREG(ac->ac_inode->i_mode));
+
-+ pa = kmem_cache_alloc(ext3_pspace_cachep, SLAB_NOFS);
++ pa = kmem_cache_alloc(ext3_pspace_cachep, GFP_NOFS);
+ if (pa == NULL)
+ return -ENOMEM;
+
+ BUG_ON(!S_ISREG(ac->ac_inode->i_mode));
+
+ BUG_ON(ext3_pspace_cachep == NULL);
-+ pa = kmem_cache_alloc(ext3_pspace_cachep, SLAB_NOFS);
++ pa = kmem_cache_alloc(ext3_pspace_cachep, GFP_NOFS);
+ if (pa == NULL)
+ return -ENOMEM;
+
+
+ BUG_ON(pa->pa_deleted == 0);
+ ext3_get_group_no_and_offset(sb, pa->pa_pstart, &group, &bit);
-+ BUG_ON(group != e3b->bd_group);
++ BUG_ON(group != e3b->bd_group && pa->pa_len != 0);
+ end = bit + pa->pa_len;
+
+ ac.ac_sb = sb;
+
+ BUG_ON(pa->pa_deleted == 0);
+ ext3_get_group_no_and_offset(sb, pa->pa_pstart, &group, &bit);
-+ BUG_ON(group != e3b->bd_group);
++ BUG_ON(group != e3b->bd_group && pa->pa_len != 0);
+ mb_free_blocks(pa->pa_inode, e3b, bit, pa->pa_len);
+ atomic_add(pa->pa_len, &EXT3_SB(sb)->s_mb_discarded);
+
+ spin_lock(&pa->pa_lock);
+ if (atomic_read(&pa->pa_count)) {
+ spin_unlock(&pa->pa_lock);
-+ printk("uh! busy PA\n");
-+ dump_stack();
+ busy = 1;
+ continue;
+ }
+ * use preallocation while we're discarding it */
+ spin_unlock(&pa->pa_lock);
+ spin_unlock(&ei->i_prealloc_lock);
-+ printk("uh-oh! used pa while discarding\n");
-+ dump_stack();
+ current->state = TASK_UNINTERRUPTIBLE;
+ schedule_timeout(HZ);
+ goto repeat;
+ * add a flag to force wait only in case
+ * of ->clear_inode(), but not in case of
+ * regular truncate */
-+ printk("uh-oh! some one just deleted it\n");
-+ dump_stack();
+ current->state = TASK_UNINTERRUPTIBLE;
+ schedule_timeout(HZ);
+ goto repeat;
+ BUG_ON(err != 0); /* error handling here */
+
+ bitmap_bh = read_block_bitmap(sb, group);
-+ if (bitmap_bh == NULL) {
-+ /* error handling here */
-+ ext3_mb_release_desc(&e3b);
-+ BUG_ON(bitmap_bh == NULL);
-+ }
+
+ ext3_lock_group(sb, group);
+ list_del_rcu(&pa->pa_group_list);
-+ ext3_mb_release_inode_pa(&e3b, bitmap_bh, pa);
++
++ /* can be NULL due to IO error, at worst
++ * we leave some free blocks unavailable
++ * do not go RO - no need for */
++ if (bitmap_bh != NULL)
++ ext3_mb_release_inode_pa(&e3b, bitmap_bh, pa);
+ ext3_unlock_group(sb, group);
+
+ ext3_mb_release_desc(&e3b);
+ BUG_ON(!mb_test_bit(bit + i, bitmap_bh->b_data));
+ }
+#endif
-+ mb_clear_bits(bitmap_bh->b_data, bit, count);
++ mb_clear_bits(sb_bgl_lock(sbi, block_group), bitmap_bh->b_data, bit,
++ count);
+
+ /* We dirtied the bitmap block */
+ BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
git://git.whamcloud.com / fs / lustre-release.git / blobdiff