/*
 * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
 * Written by Alex Tomas <alex@clusterfs.com>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public Licens
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-
 */


/*
 * mballoc.c contains the multiblocks allocation routines
 */

#include "ext4_jbd2.h"
#include "mballoc.h"
#include <linux/debugfs.h>
#include <linux/slab.h>
#include <trace/events/ext4.h>

/*
 * MUSTDO:
 *   - test ext4_ext_search_left() and ext4_ext_search_right()
 *   - search for metadata in few groups
 *
 * TODO v4:
 *   - normalization should take into account whether file is still open
 *   - discard preallocations if no free space left (policy?)
 *   - don't normalize tails
 *   - quota
 *   - reservation for superuser
 *
 * TODO v3:
 *   - bitmap read-ahead (proposed by Oleg Drokin aka green)
 *   - track min/max extents in each group for better group selection
 *   - mb_mark_used() may allocate chunk right after splitting buddy
 *   - tree of groups sorted by number of free blocks
 *   - error handling
 */

/*
 * The allocation request involve request for multiple number of blocks
 * near to the goal(block) value specified.
 *
 * During initialization phase of the allocator we decide to use the
 * group preallocation or inode preallocation depending on the size of
 * the file. The size of the file could be the resulting file size we
 * would have after allocation, or the current file size, which ever
 * is larger. If the size is less than sbi->s_mb_stream_request we
 * select to use the group preallocation. The default value of
 * s_mb_stream_request is 16 blocks. This can also be tuned via
 * /sys/fs/ext4/<partition>/mb_stream_req. The value is represented in
 * terms of number of blocks.
 *
 * The main motivation for having small file use group preallocation is to
 * ensure that we have small files closer together on the disk.
 *
 * First stage the allocator looks at the inode prealloc list,
 * ext4_inode_info->i_prealloc_list, which contains list of prealloc
 * spaces for this particular inode. The inode prealloc space is
 * represented as:
 *
 * pa_lstart -> the logical start block for this prealloc space
 * pa_pstart -> the physical start block for this prealloc space
 * pa_len    -> length for this prealloc space (in clusters)
 * pa_free   ->  free space available in this prealloc space (in clusters)
 *
 * The inode preallocation space is used looking at the _logical_ start
 * block. If only the logical file block falls within the range of prealloc
 * space we will consume the particular prealloc space. This makes sure that
 * we have contiguous physical blocks representing the file blocks
 *
 * The important thing to be noted in case of inode prealloc space is that
 * we don't modify the values associated to inode prealloc space except
 * pa_free.
 *
 * If we are not able to find blocks in the inode prealloc space and if we
 * have the group allocation flag set then we look at the locality group
 * prealloc space. These are per CPU prealloc list represented as
 *
 * ext4_sb_info.s_locality_groups[smp_processor_id()]
 *
 * The reason for having a per cpu locality group is to reduce the contention
 * between CPUs. It is possible to get scheduled at this point.
 *
 * The locality group prealloc space is used looking at whether we have
 * enough free space (pa_free) within the prealloc space.
 *
 * If we can't allocate blocks via inode prealloc or/and locality group
 * prealloc then we look at the buddy cache. The buddy cache is represented
 * by ext4_sb_info.s_buddy_cache (struct inode) whose file offset gets
 * mapped to the buddy and bitmap information regarding different
 * groups. The buddy information is attached to buddy cache inode so that
 * we can access them through the page cache. The information regarding
 * each group is loaded via ext4_mb_load_buddy.  The information involve
 * block bitmap and buddy information. The information are stored in the
 * inode as:
 *
 *  {                        page                        }
 *  [ group 0 bitmap][ group 0 buddy] [group 1][ group 1]...
 *
 *
 * one block each for bitmap and buddy information.  So for each group we
 * take up 2 blocks. A page can contain blocks_per_page (PAGE_CACHE_SIZE /
 * blocksize) blocks.  So it can have information regarding groups_per_page
 * which is blocks_per_page/2
 *
 * The buddy cache inode is not stored on disk. The inode is thrown
 * away when the filesystem is unmounted.
 *
 * We look for count number of blocks in the buddy cache. If we were able
 * to locate that many free blocks we return with additional information
 * regarding rest of the contiguous physical block available
 *
 * Before allocating blocks via buddy cache we normalize the request
 * blocks. This ensure we ask for more blocks that we needed. The extra
 * blocks that we get after allocation is added to the respective prealloc
 * list. In case of inode preallocation we follow a list of heuristics
 * based on file size. This can be found in ext4_mb_normalize_request. If
 * we are doing a group prealloc we try to normalize the request to
 * sbi->s_mb_group_prealloc.  The default value of s_mb_group_prealloc is
 * dependent on the cluster size; for non-bigalloc file systems, it is
 * 512 blocks. This can be tuned via
 * /sys/fs/ext4/<partition>/mb_group_prealloc. The value is represented in
 * terms of number of blocks. If we have mounted the file system with -O
 * stripe=<value> option the group prealloc request is normalized to the
 * the smallest multiple of the stripe value (sbi->s_stripe) which is
 * greater than the default mb_group_prealloc.
 *
 * The regular allocator (using the buddy cache) supports a few tunables.
 *
 * /sys/fs/ext4/<partition>/mb_min_to_scan
 * /sys/fs/ext4/<partition>/mb_max_to_scan
 * /sys/fs/ext4/<partition>/mb_order2_req
 *
 * The regular allocator uses buddy scan only if the request len is power of
 * 2 blocks and the order of allocation is >= sbi->s_mb_order2_reqs. The
 * value of s_mb_order2_reqs can be tuned via
 * /sys/fs/ext4/<partition>/mb_order2_req.  If the request len is equal to
 * stripe size (sbi->s_stripe), we try to search for contiguous block in
 * stripe size. This should result in better allocation on RAID setups. If
 * not, we search in the specific group using bitmap for best extents. The
 * tunable min_to_scan and max_to_scan control the behaviour here.
 * min_to_scan indicate how long the mballoc __must__ look for a best
 * extent and max_to_scan indicates how long the mballoc __can__ look for a
 * best extent in the found extents. Searching for the blocks starts with
 * the group specified as the goal value in allocation context via
 * ac_g_ex. Each group is first checked based on the criteria whether it
 * can be used for allocation. ext4_mb_good_group explains how the groups are
 * checked.
 *
 * Both the prealloc space are getting populated as above. So for the first
 * request we will hit the buddy cache which will result in this prealloc
 * space getting filled. The prealloc space is then later used for the
 * subsequent request.
 */

/*
 * mballoc operates on the following data:
 *  - on-disk bitmap
 *  - in-core buddy (actually includes buddy and bitmap)
 *  - preallocation descriptors (PAs)
 *
 * there are two types of preallocations:
 *  - inode
 *    assiged to specific inode and can be used for this inode only.
 *    it describes part of inode's space preallocated to specific
 *    physical blocks. any block from that preallocated can be used
 *    independent. the descriptor just tracks number of blocks left
 *    unused. so, before taking some block from descriptor, one must
 *    make sure corresponded logical block isn't allocated yet. this
 *    also means that freeing any block within descriptor's range
 *    must discard all preallocated blocks.
 *  - locality group
 *    assigned to specific locality group which does not translate to
 *    permanent set of inodes: inode can join and leave group. space
 *    from this type of preallocation can be used for any inode. thus
 *    it's consumed from the beginning to the end.
 *
 * relation between them can be expressed as:
 *    in-core buddy = on-disk bitmap + preallocation descriptors
 *
 * this mean blocks mballoc considers used are:
 *  - allocated blocks (persistent)
 *  - preallocated blocks (non-persistent)
 *
 * consistency in mballoc world means that at any time a block is either
 * free or used in ALL structures. notice: "any time" should not be read
 * literally -- time is discrete and delimited by locks.
 *
 *  to keep it simple, we don't use block numbers, instead we count number of
 *  blocks: how many blocks marked used/free in on-disk bitmap, buddy and PA.
 *
 * all operations can be expressed as:
 *  - init buddy:                       buddy = on-disk + PAs
 *  - new PA:                           buddy += N; PA = N
 *  - use inode PA:                     on-disk += N; PA -= N
 *  - discard inode PA                  buddy -= on-disk - PA; PA = 0
 *  - use locality group PA             on-disk += N; PA -= N
 *  - discard locality group PA         buddy -= PA; PA = 0
 *  note: 'buddy -= on-disk - PA' is used to show that on-disk bitmap
 *        is used in real operation because we can't know actual used
 *        bits from PA, only from on-disk bitmap
 *
 * if we follow this strict logic, then all operations above should be atomic.
 * given some of them can block, we'd have to use something like semaphores
 * killing performance on high-end SMP hardware. let's try to relax it using
 * the following knowledge:
 *  1) if buddy is referenced, it's already initialized
 *  2) while block is used in buddy and the buddy is referenced,
 *     nobody can re-allocate that block
 *  3) we work on bitmaps and '+' actually means 'set bits'. if on-disk has
 *     bit set and PA claims same block, it's OK. IOW, one can set bit in
 *     on-disk bitmap if buddy has same bit set or/and PA covers corresponded
 *     block
 *
 * so, now we're building a concurrency table:
 *  - init buddy vs.
 *    - new PA
 *      blocks for PA are allocated in the buddy, buddy must be referenced
 *      until PA is linked to allocation group to avoid concurrent buddy init
 *    - use inode PA
 *      we need to make sure that either on-disk bitmap or PA has uptodate data
 *      given (3) we care that PA-=N operation doesn't interfere with init
 *    - discard inode PA
 *      the simplest way would be to have buddy initialized by the discard
 *    - use locality group PA
 *      again PA-=N must be serialized with init
 *    - discard locality group PA
 *      the simplest way would be to have buddy initialized by the discard
 *  - new PA vs.
 *    - use inode PA
 *      i_data_sem serializes them
 *    - discard inode PA
 *      discard process must wait until PA isn't used by another process
 *    - use locality group PA
 *      some mutex should serialize them
 *    - discard locality group PA
 *      discard process must wait until PA isn't used by another process
 *  - use inode PA
 *    - use inode PA
 *      i_data_sem or another mutex should serializes them
 *    - discard inode PA
 *      discard process must wait until PA isn't used by another process
 *    - use locality group PA
 *      nothing wrong here -- they're different PAs covering different blocks
 *    - discard locality group PA
 *      discard process must wait until PA isn't used by another process
 *
 * now we're ready to make few consequences:
 *  - PA is referenced and while it is no discard is possible
 *  - PA is referenced until block isn't marked in on-disk bitmap
 *  - PA changes only after on-disk bitmap
 *  - discard must not compete with init. either init is done before
 *    any discard or they're serialized somehow
 *  - buddy init as sum of on-disk bitmap and PAs is done atomically
 *
 * a special case when we've used PA to emptiness. no need to modify buddy
 * in this case, but we should care about concurrent init
 *
 */

 /*
 * Logic in few words:
 *
 *  - allocation:
 *    load group
 *    find blocks
 *    mark bits in on-disk bitmap
 *    release group
 *
 *  - use preallocation:
 *    find proper PA (per-inode or group)
 *    load group
 *    mark bits in on-disk bitmap
 *    release group
 *    release PA
 *
 *  - free:
 *    load group
 *    mark bits in on-disk bitmap
 *    release group
 *
 *  - discard preallocations in group:
 *    mark PAs deleted
 *    move them onto local list
 *    load on-disk bitmap
 *    load group
 *    remove PA from object (inode or locality group)
 *    mark free blocks in-core
 *
 *  - discard inode's preallocations:
 */

/*
 * Locking rules
 *
 * Locks:
 *  - bitlock on a group        (group)
 *  - object (inode/locality)   (object)
 *  - per-pa lock               (pa)
 *
 * Paths:
 *  - new pa
 *    object
 *    group
 *
 *  - find and use pa:
 *    pa
 *
 *  - release consumed pa:
 *    pa
 *    group
 *    object
 *
 *  - generate in-core bitmap:
 *    group
 *        pa
 *
 *  - discard all for given object (inode, locality group):
 *    object
 *        pa
 *    group
 *
 *  - discard all for given group:
 *    group
 *        pa
 *    group
 *        object
 *
 */
static struct kmem_cache *ext4_pspace_cachep;
static struct kmem_cache *ext4_ac_cachep;
static struct kmem_cache *ext4_free_data_cachep;

/* We create slab caches for groupinfo data structures based on the
 * superblock block size.  There will be one per mounted filesystem for
 * each unique s_blocksize_bits */
#define NR_GRPINFO_CACHES 8
static struct kmem_cache *ext4_groupinfo_caches[NR_GRPINFO_CACHES];

static const char *ext4_groupinfo_slab_names[NR_GRPINFO_CACHES] = {
        "ext4_groupinfo_1k", "ext4_groupinfo_2k", "ext4_groupinfo_4k",
        "ext4_groupinfo_8k", "ext4_groupinfo_16k", "ext4_groupinfo_32k",
        "ext4_groupinfo_64k", "ext4_groupinfo_128k"
};

static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
                                        ext4_group_t group);
static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap,
                                                ext4_group_t group);
static void ext4_free_data_callback(struct super_block *sb,
                                struct ext4_journal_cb_entry *jce, int rc);

static inline void *mb_correct_addr_and_bit(int *bit, void *addr)
{
#if BITS_PER_LONG == 64
        *bit += ((unsigned long) addr & 7UL) << 3;
        addr = (void *) ((unsigned long) addr & ~7UL);
#elif BITS_PER_LONG == 32
        *bit += ((unsigned long) addr & 3UL) << 3;
        addr = (void *) ((unsigned long) addr & ~3UL);
#else
#error "how many bits you are?!"
#endif
        return addr;
}

static inline int mb_test_bit(int bit, void *addr)
{
        /*
         * ext4_test_bit on architecture like powerpc
         * needs unsigned long aligned address
         */
        addr = mb_correct_addr_and_bit(&bit, addr);
        return ext4_test_bit(bit, addr);
}

static inline void mb_set_bit(int bit, void *addr)
{
        addr = mb_correct_addr_and_bit(&bit, addr);
        ext4_set_bit(bit, addr);
}

static inline void mb_clear_bit(int bit, void *addr)
{
        addr = mb_correct_addr_and_bit(&bit, addr);
        ext4_clear_bit(bit, addr);
}

static inline int mb_find_next_zero_bit(void *addr, int max, int start)
{
        int fix = 0, ret, tmpmax;
        addr = mb_correct_addr_and_bit(&fix, addr);
        tmpmax = max + fix;
        start += fix;

        ret = ext4_find_next_zero_bit(addr, tmpmax, start) - fix;
        if (ret > max)
                return max;
        return ret;
}

static inline int mb_find_next_bit(void *addr, int max, int start)
{
        int fix = 0, ret, tmpmax;
        addr = mb_correct_addr_and_bit(&fix, addr);
        tmpmax = max + fix;
        start += fix;

        ret = ext4_find_next_bit(addr, tmpmax, start) - fix;
        if (ret > max)
                return max;
        return ret;
}

static void *mb_find_buddy(struct ext4_buddy *e4b, int order, int *max)
{
        char *bb;

        BUG_ON(e4b->bd_bitmap == e4b->bd_buddy);
        BUG_ON(max == NULL);

        if (order > e4b->bd_blkbits + 1) {
                *max = 0;
                return NULL;
        }

        /* at order 0 we see each particular block */
        if (order == 0) {
                *max = 1 << (e4b->bd_blkbits + 3);
                return e4b->bd_bitmap;
        }

        bb = e4b->bd_buddy + EXT4_SB(e4b->bd_sb)->s_mb_offsets[order];
        *max = EXT4_SB(e4b->bd_sb)->s_mb_maxs[order];

        return bb;
}

#ifdef DOUBLE_CHECK
static void mb_free_blocks_double(struct inode *inode, struct ext4_buddy *e4b,
                           int first, int count)
{
        int i;
        struct super_block *sb = e4b->bd_sb;

        if (unlikely(e4b->bd_info->bb_bitmap == NULL))
                return;
        assert_spin_locked(ext4_group_lock_ptr(sb, e4b->bd_group));
        for (i = 0; i < count; i++) {
                if (!mb_test_bit(first + i, e4b->bd_info->bb_bitmap)) {
                        ext4_fsblk_t blocknr;

                        blocknr = ext4_group_first_block_no(sb, e4b->bd_group);
                        blocknr += EXT4_C2B(EXT4_SB(sb), first + i);
                        ext4_grp_locked_error(sb, e4b->bd_group,
                                              inode ? inode->i_ino : 0,
                                              blocknr,
                                              "freeing block already freed "
                                              "(bit %u)",
                                              first + i);
                }
                mb_clear_bit(first + i, e4b->bd_info->bb_bitmap);
        }
}

static void mb_mark_used_double(struct ext4_buddy *e4b, int first, int count)
{
        int i;

        if (unlikely(e4b->bd_info->bb_bitmap == NULL))
                return;
        assert_spin_locked(ext4_group_lock_ptr(e4b->bd_sb, e4b->bd_group));
        for (i = 0; i < count; i++) {
                BUG_ON(mb_test_bit(first + i, e4b->bd_info->bb_bitmap));
                mb_set_bit(first + i, e4b->bd_info->bb_bitmap);
        }
}

static void mb_cmp_bitmaps(struct ext4_buddy *e4b, void *bitmap)
{
        if (memcmp(e4b->bd_info->bb_bitmap, bitmap, e4b->bd_sb->s_blocksize)) {
                unsigned char *b1, *b2;
                int i;
                b1 = (unsigned char *) e4b->bd_info->bb_bitmap;
                b2 = (unsigned char *) bitmap;
                for (i = 0; i < e4b->bd_sb->s_blocksize; i++) {
                        if (b1[i] != b2[i]) {
                                ext4_msg(e4b->bd_sb, KERN_ERR,
                                         "corruption in group %u "
                                         "at byte %u(%u): %x in copy != %x "
                                         "on disk/prealloc",
                                         e4b->bd_group, i, i * 8, b1[i], b2[i]);
                                BUG();
                        }
                }
        }
}

#else
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

#define MB_CHECK_ASSERT(assert)                                         \
do {                                                                    \
        if (!(assert)) {                                                \
                printk(KERN_EMERG                                       \
                        "Assertion failure in %s() at %s:%d: \"%s\"\n", \
                        function, file, line, # assert);                \
                BUG();                                                  \
        }                                                               \
} while (0)

static int __mb_check_buddy(struct ext4_buddy *e4b, char *file,
                                const char *function, int line)
{
        struct super_block *sb = e4b->bd_sb;
        int order = e4b->bd_blkbits + 1;
        int max;
        int max2;
        int i;
        int j;
        int k;
        int count;
        struct ext4_group_info *grp;
        int fragments = 0;
        int fstart;
        struct list_head *cur;
        void *buddy;
        void *buddy2;

        {
                static int mb_check_counter;
                if (mb_check_counter++ % 100 != 0)
                        return 0;
        }

        while (order > 1) {
                buddy = mb_find_buddy(e4b, order, &max);
                MB_CHECK_ASSERT(buddy);
                buddy2 = mb_find_buddy(e4b, order - 1, &max2);
                MB_CHECK_ASSERT(buddy2);
                MB_CHECK_ASSERT(buddy != buddy2);
                MB_CHECK_ASSERT(max * 2 == max2);

                count = 0;
                for (i = 0; i < max; i++) {

                        if (mb_test_bit(i, buddy)) {
                                /* only single bit in buddy2 may be 1 */
                                if (!mb_test_bit(i << 1, buddy2)) {
                                        MB_CHECK_ASSERT(
                                                mb_test_bit((i<<1)+1, buddy2));
                                } else if (!mb_test_bit((i << 1) + 1, buddy2)) {
                                        MB_CHECK_ASSERT(
                                                mb_test_bit(i << 1, buddy2));
                                }
                                continue;
                        }

                        /* both bits in buddy2 must be 1 */
                        MB_CHECK_ASSERT(mb_test_bit(i << 1, buddy2));
                        MB_CHECK_ASSERT(mb_test_bit((i << 1) + 1, buddy2));

                        for (j = 0; j < (1 << order); j++) {
                                k = (i * (1 << order)) + j;
                                MB_CHECK_ASSERT(
                                        !mb_test_bit(k, e4b->bd_bitmap));
                        }
                        count++;
                }
                MB_CHECK_ASSERT(e4b->bd_info->bb_counters[order] == count);
                order--;
        }

        fstart = -1;
        buddy = mb_find_buddy(e4b, 0, &max);
        for (i = 0; i < max; i++) {
                if (!mb_test_bit(i, buddy)) {
                        MB_CHECK_ASSERT(i >= e4b->bd_info->bb_first_free);
                        if (fstart == -1) {
                                fragments++;
                                fstart = i;
                        }
                        continue;
                }
                fstart = -1;
                /* check used bits only */
                for (j = 0; j < e4b->bd_blkbits + 1; j++) {
                        buddy2 = mb_find_buddy(e4b, j, &max2);
                        k = i >> j;
                        MB_CHECK_ASSERT(k < max2);
                        MB_CHECK_ASSERT(mb_test_bit(k, buddy2));
                }
        }
        MB_CHECK_ASSERT(!EXT4_MB_GRP_NEED_INIT(e4b->bd_info));
        MB_CHECK_ASSERT(e4b->bd_info->bb_fragments == fragments);

        grp = ext4_get_group_info(sb, e4b->bd_group);
        list_for_each(cur, &grp->bb_prealloc_list) {
                ext4_group_t groupnr;
                struct ext4_prealloc_space *pa;
                pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list);
                ext4_get_group_no_and_offset(sb, pa->pa_pstart, &groupnr, &k);
                MB_CHECK_ASSERT(groupnr == e4b->bd_group);
                for (i = 0; i < pa->pa_len; i++)
                        MB_CHECK_ASSERT(mb_test_bit(k + i, buddy));
        }
        return 0;
}
#undef MB_CHECK_ASSERT
#define mb_check_buddy(e4b) __mb_check_buddy(e4b,       \
                                        __FILE__, __func__, __LINE__)
#else
#define mb_check_buddy(e4b)
#endif

/*
 * Divide blocks started from @first with length @len into
 * smaller chunks with power of 2 blocks.
 * Clear the bits in bitmap which the blocks of the chunk(s) covered,
 * then increase bb_counters[] for corresponded chunk size.
 */
static void ext4_mb_mark_free_simple(struct super_block *sb,
                                void *buddy, ext4_grpblk_t first, ext4_grpblk_t len,
                                        struct ext4_group_info *grp)
{
        struct ext4_sb_info *sbi = EXT4_SB(sb);
        ext4_grpblk_t min;
        ext4_grpblk_t max;
        ext4_grpblk_t chunk;
        unsigned short border;

        BUG_ON(len > EXT4_CLUSTERS_PER_GROUP(sb));

        border = 2 << sb->s_blocksize_bits;

        while (len > 0) {
                /* find how many blocks can be covered since this position */
                max = ffs(first | border) - 1;

                /* find how many blocks of power 2 we need to mark */
                min = fls(len) - 1;

                if (max < min)
                        min = max;
                chunk = 1 << min;

                /* mark multiblock chunks only */
                grp->bb_counters[min]++;
                if (min > 0)
                        mb_clear_bit(first >> min,
                                     buddy + sbi->s_mb_offsets[min]);

                len -= chunk;
                first += chunk;
        }
}

/*
 * Cache the order of the largest free extent we have available in this block
 * group.
 */
static void
mb_set_largest_free_order(struct super_block *sb, struct ext4_group_info *grp)
{
        int i;
        int bits;

        grp->bb_largest_free_order = -1; /* uninit */

        bits = sb->s_blocksize_bits + 1;
        for (i = bits; i >= 0; i--) {
                if (grp->bb_counters[i] > 0) {
                        grp->bb_largest_free_order = i;
                        break;
                }
        }
}

static noinline_for_stack
void ext4_mb_generate_buddy(struct super_block *sb,
                                void *buddy, void *bitmap, ext4_group_t group)
{
        struct ext4_group_info *grp = ext4_get_group_info(sb, group);
        ext4_grpblk_t max = EXT4_CLUSTERS_PER_GROUP(sb);
        ext4_grpblk_t i = 0;
        ext4_grpblk_t first;
        ext4_grpblk_t len;
        unsigned free = 0;
        unsigned fragments = 0;
        unsigned long long period = get_cycles();

        /* initialize buddy from bitmap which is aggregation
         * of on-disk bitmap and preallocations */
        i = mb_find_next_zero_bit(bitmap, max, 0);
        grp->bb_first_free = i;
        while (i < max) {
                fragments++;
                first = i;
                i = mb_find_next_bit(bitmap, max, i);
                len = i - first;
                free += len;
                if (len > 1)
                        ext4_mb_mark_free_simple(sb, buddy, first, len, grp);
                else
                        grp->bb_counters[0]++;
                if (i < max)
                        i = mb_find_next_zero_bit(bitmap, max, i);
        }
        grp->bb_fragments = fragments;

        if (free != grp->bb_free) {
                ext4_grp_locked_error(sb, group, 0, 0,
                                      "%u clusters in bitmap, %u in gd",
                                      free, grp->bb_free);
                /*
                 * If we intent to continue, we consider group descritor
                 * corrupt and update bb_free using bitmap value
                 */
                grp->bb_free = free;
        }
        mb_set_largest_free_order(sb, grp);

        clear_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state));

        period = get_cycles() - period;
        spin_lock(&EXT4_SB(sb)->s_bal_lock);
        EXT4_SB(sb)->s_mb_buddies_generated++;
        EXT4_SB(sb)->s_mb_generation_time += period;
        spin_unlock(&EXT4_SB(sb)->s_bal_lock);
}

/* The buddy information is attached the buddy cache inode
 * for convenience. The information regarding each group
 * is loaded via ext4_mb_load_buddy. The information involve
 * block bitmap and buddy information. The information are
 * stored in the inode as
 *
 * {                        page                        }
 * [ group 0 bitmap][ group 0 buddy] [group 1][ group 1]...
 *
 *
 * one block each for bitmap and buddy information.
 * So for each group we take up 2 blocks. A page can
 * contain blocks_per_page (PAGE_CACHE_SIZE / blocksize)  blocks.
 * So it can have information regarding groups_per_page which
 * is blocks_per_page/2
 *
 * Locking note:  This routine takes the block group lock of all groups
 * for this page; do not hold this lock when calling this routine!
 */

static int ext4_mb_init_cache(struct page *page, char *incore)
{
        ext4_group_t ngroups;
        int blocksize;
        int blocks_per_page;
        int groups_per_page;
        int err = 0;
        int i;
        ext4_group_t first_group, group;
        int first_block;
        struct super_block *sb;
        struct buffer_head *bhs;
        struct buffer_head **bh = NULL;
        struct inode *inode;
        char *data;
        char *bitmap;
        struct ext4_group_info *grinfo;

        mb_debug(1, "init page %lu\n", page->index);

        inode = page->mapping->host;
        sb = inode->i_sb;
        ngroups = ext4_get_groups_count(sb);
        blocksize = 1 << inode->i_blkbits;
        blocks_per_page = PAGE_CACHE_SIZE / blocksize;

        groups_per_page = blocks_per_page >> 1;
        if (groups_per_page == 0)
                groups_per_page = 1;

        /* allocate buffer_heads to read bitmaps */
        if (groups_per_page > 1) {
                i = sizeof(struct buffer_head *) * groups_per_page;
                bh = kzalloc(i, GFP_NOFS);
                if (bh == NULL) {
                        err = -ENOMEM;
                        goto out;
                }
        } else
                bh = &bhs;

        first_group = page->index * blocks_per_page / 2;

        /* read all groups the page covers into the cache */
        for (i = 0, group = first_group; i < groups_per_page; i++, group++) {
                if (group >= ngroups)
                        break;

                grinfo = ext4_get_group_info(sb, group);
                /*
                 * If page is uptodate then we came here after online resize
                 * which added some new uninitialized group info structs, so
                 * we must skip all initialized uptodate buddies on the page,
                 * which may be currently in use by an allocating task.
                 */
                if (PageUptodate(page) && !EXT4_MB_GRP_NEED_INIT(grinfo)) {
                        bh[i] = NULL;
                        continue;
                }
                if (!(bh[i] = ext4_read_block_bitmap_nowait(sb, group))) {
                        err = -ENOMEM;
                        goto out;
                }
                mb_debug(1, "read bitmap for group %u\n", group);
        }

        /* wait for I/O completion */
        for (i = 0, group = first_group; i < groups_per_page; i++, group++) {
                if (bh[i] && ext4_wait_block_bitmap(sb, group, bh[i])) {
                        err = -EIO;
                        goto out;
                }
        }

        first_block = page->index * blocks_per_page;
        for (i = 0; i < blocks_per_page; i++) {
                int group;

                group = (first_block + i) >> 1;
                if (group >= ngroups)
                        break;

                if (!bh[group - first_group])
                        /* skip initialized uptodate buddy */
                        continue;

                /*
                 * data carry information regarding this
                 * particular group in the format specified
                 * above
                 *
                 */
                data = page_address(page) + (i * blocksize);
                bitmap = bh[group - first_group]->b_data;

                /*
                 * We place the buddy block and bitmap block
                 * close together
                 */
                if ((first_block + i) & 1) {
                        /* this is block of buddy */
                        BUG_ON(incore == NULL);
                        mb_debug(1, "put buddy for group %u in page %lu/%x\n",
                                group, page->index, i * blocksize);
                        trace_ext4_mb_buddy_bitmap_load(sb, group);
                        grinfo = ext4_get_group_info(sb, group);
                        grinfo->bb_fragments = 0;
                        memset(grinfo->bb_counters, 0,
                               sizeof(*grinfo->bb_counters) *
                                (sb->s_blocksize_bits+2));
                        /*
                         * incore got set to the group block bitmap below
                         */
                        ext4_lock_group(sb, group);
                        /* init the buddy */
                        memset(data, 0xff, blocksize);
                        ext4_mb_generate_buddy(sb, data, incore, group);
                        ext4_unlock_group(sb, group);
                        incore = NULL;
                } else {
                        /* this is block of bitmap */
                        BUG_ON(incore != NULL);
                        mb_debug(1, "put bitmap for group %u in page %lu/%x\n",
                                group, page->index, i * blocksize);
                        trace_ext4_mb_bitmap_load(sb, group);

                        /* see comments in ext4_mb_put_pa() */
                        ext4_lock_group(sb, group);
                        memcpy(data, bitmap, blocksize);

                        /* mark all preallocated blks used in in-core bitmap */
                        ext4_mb_generate_from_pa(sb, data, group);
                        ext4_mb_generate_from_freelist(sb, data, group);
                        ext4_unlock_group(sb, group);

                        /* set incore so that the buddy information can be
                         * generated using this
                         */
                        incore = data;
                }
        }
        SetPageUptodate(page);

out:
        if (bh) {
                for (i = 0; i < groups_per_page; i++)
                        brelse(bh[i]);
                if (bh != &bhs)
                        kfree(bh);
        }
        return err;
}

/*
 * Lock the buddy and bitmap pages. This make sure other parallel init_group
 * on the same buddy page doesn't happen whild holding the buddy page lock.
 * Return locked buddy and bitmap pages on e4b struct. If buddy and bitmap
 * are on the same page e4b->bd_buddy_page is NULL and return value is 0.
 */
static int ext4_mb_get_buddy_page_lock(struct super_block *sb,
                ext4_group_t group, struct ext4_buddy *e4b)
{
        struct inode *inode = EXT4_SB(sb)->s_buddy_cache;
        int block, pnum, poff;
        int blocks_per_page;
        struct page *page;

        e4b->bd_buddy_page = NULL;
        e4b->bd_bitmap_page = NULL;

        blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize;
        /*
         * the buddy cache inode stores the block bitmap
         * and buddy information in consecutive blocks.
         * So for each group we need two blocks.
         */
        block = group * 2;
        pnum = block / blocks_per_page;
        poff = block % blocks_per_page;
        page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
        if (!page)
                return -EIO;
        BUG_ON(page->mapping != inode->i_mapping);
        e4b->bd_bitmap_page = page;
        e4b->bd_bitmap = page_address(page) + (poff * sb->s_blocksize);

        if (blocks_per_page >= 2) {
                /* buddy and bitmap are on the same page */
                return 0;
        }

        block++;
        pnum = block / blocks_per_page;
        page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
        if (!page)
                return -EIO;
        BUG_ON(page->mapping != inode->i_mapping);
        e4b->bd_buddy_page = page;
        return 0;
}

static void ext4_mb_put_buddy_page_lock(struct ext4_buddy *e4b)
{
        if (e4b->bd_bitmap_page) {
                unlock_page(e4b->bd_bitmap_page);
                page_cache_release(e4b->bd_bitmap_page);
        }
        if (e4b->bd_buddy_page) {
                unlock_page(e4b->bd_buddy_page);
                page_cache_release(e4b->bd_buddy_page);
        }
}

/*
 * Locking note:  This routine calls ext4_mb_init_cache(), which takes the
 * block group lock of all groups for this page; do not hold the BG lock when
 * calling this routine!
 */
static noinline_for_stack
int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
{

        struct ext4_group_info *this_grp;
        struct ext4_buddy e4b;
        struct page *page;
        int ret = 0;

        mb_debug(1, "init group %u\n", group);
        this_grp = ext4_get_group_info(sb, group);
        /*
         * This ensures that we don't reinit the buddy cache
         * page which map to the group from which we are already
         * allocating. If we are looking at the buddy cache we would
         * have taken a reference using ext4_mb_load_buddy and that
         * would have pinned buddy page to page cache.
         */
        ret = ext4_mb_get_buddy_page_lock(sb, group, &e4b);
        if (ret || !EXT4_MB_GRP_NEED_INIT(this_grp)) {
                /*
                 * somebody initialized the group
                 * return without doing anything
                 */
                goto err;
        }

        page = e4b.bd_bitmap_page;
        ret = ext4_mb_init_cache(page, NULL);
        if (ret)
                goto err;
        if (!PageUptodate(page)) {
                ret = -EIO;
                goto err;
        }
        mark_page_accessed(page);

        if (e4b.bd_buddy_page == NULL) {
                /*
                 * If both the bitmap and buddy are in
                 * the same page we don't need to force
                 * init the buddy
                 */
                ret = 0;
                goto err;
        }
        /* init buddy cache */
        page = e4b.bd_buddy_page;
        ret = ext4_mb_init_cache(page, e4b.bd_bitmap);
        if (ret)
                goto err;
        if (!PageUptodate(page)) {
                ret = -EIO;
                goto err;
        }
        mark_page_accessed(page);
err:
        ext4_mb_put_buddy_page_lock(&e4b);
        return ret;
}

/*
 * Locking note:  This routine calls ext4_mb_init_cache(), which takes the
 * block group lock of all groups for this page; do not hold the BG lock when
 * calling this routine!
 */
static noinline_for_stack int
ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
                                        struct ext4_buddy *e4b)
{
        int blocks_per_page;
        int block;
        int pnum;
        int poff;
        struct page *page;
        int ret;
        struct ext4_group_info *grp;
        struct ext4_sb_info *sbi = EXT4_SB(sb);
        struct inode *inode = sbi->s_buddy_cache;

        mb_debug(1, "load group %u\n", group);

        blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize;
        grp = ext4_get_group_info(sb, group);

        e4b->bd_blkbits = sb->s_blocksize_bits;
        e4b->bd_info = grp;
        e4b->bd_sb = sb;
        e4b->bd_group = group;
        e4b->bd_buddy_page = NULL;
        e4b->bd_bitmap_page = NULL;

        if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) {
                /*
                 * we need full data about the group
                 * to make a good selection
                 */
                ret = ext4_mb_init_group(sb, group);
                if (ret)
                        return ret;
        }

        /*
         * the buddy cache inode stores the block bitmap
         * and buddy information in consecutive blocks.
         * So for each group we need two blocks.
         */
        block = group * 2;
        pnum = block / blocks_per_page;
        poff = block % blocks_per_page;

        /* we could use find_or_create_page(), but it locks page
         * what we'd like to avoid in fast path ... */
        page = find_get_page(inode->i_mapping, pnum);
        if (page == NULL || !PageUptodate(page)) {
                if (page)
                        /*
                         * drop the page reference and try
                         * to get the page with lock. If we
                         * are not uptodate that implies
                         * somebody just created the page but
                         * is yet to initialize the same. So
                         * wait for it to initialize.
                         */
                        page_cache_release(page);
                page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
                if (page) {
                        BUG_ON(page->mapping != inode->i_mapping);
                        if (!PageUptodate(page)) {
                                ret = ext4_mb_init_cache(page, NULL);
                                if (ret) {
                                        unlock_page(page);
                                        goto err;
                                }
                                mb_cmp_bitmaps(e4b, page_address(page) +
                                               (poff * sb->s_blocksize));
                        }
                        unlock_page(page);
                }
        }
        if (page == NULL || !PageUptodate(page)) {
                ret = -EIO;
                goto err;
        }
        e4b->bd_bitmap_page = page;
        e4b->bd_bitmap = page_address(page) + (poff * sb->s_blocksize);
        mark_page_accessed(page);

        block++;
        pnum = block / blocks_per_page;
        poff = block % blocks_per_page;

        page = find_get_page(inode->i_mapping, pnum);
        if (page == NULL || !PageUptodate(page)) {
                if (page)
                        page_cache_release(page);
                page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
                if (page) {
                        BUG_ON(page->mapping != inode->i_mapping);
                        if (!PageUptodate(page)) {
                                ret = ext4_mb_init_cache(page, e4b->bd_bitmap);
                                if (ret) {
                                        unlock_page(page);
                                        goto err;
                                }
                        }
                        unlock_page(page);
                }
        }
        if (page == NULL || !PageUptodate(page)) {
                ret = -EIO;
                goto err;
        }
        e4b->bd_buddy_page = page;
        e4b->bd_buddy = page_address(page) + (poff * sb->s_blocksize);
        mark_page_accessed(page);

        BUG_ON(e4b->bd_bitmap_page == NULL);
        BUG_ON(e4b->bd_buddy_page == NULL);

        return 0;

err:
        if (page)
                page_cache_release(page);
        if (e4b->bd_bitmap_page)
                page_cache_release(e4b->bd_bitmap_page);
        if (e4b->bd_buddy_page)
                page_cache_release(e4b->bd_buddy_page);
        e4b->bd_buddy = NULL;
        e4b->bd_bitmap = NULL;
        return ret;
}

static void ext4_mb_unload_buddy(struct ext4_buddy *e4b)
{
        if (e4b->bd_bitmap_page)
                page_cache_release(e4b->bd_bitmap_page);
        if (e4b->bd_buddy_page)
                page_cache_release(e4b->bd_buddy_page);
}


static int mb_find_order_for_block(struct ext4_buddy *e4b, int block)
{
        int order = 1;
        void *bb;

        BUG_ON(e4b->bd_bitmap == e4b->bd_buddy);
        BUG_ON(block >= (1 << (e4b->bd_blkbits + 3)));

        bb = e4b->bd_buddy;
        while (order <= e4b->bd_blkbits + 1) {
                block = block >> 1;
                if (!mb_test_bit(block, bb)) {
                        /* this block is part of buddy of order 'order' */
                        return order;
                }
                bb += 1 << (e4b->bd_blkbits - order);
                order++;
        }
        return 0;
}

static void mb_clear_bits(void *bm, int cur, int len)
{
        __u32 *addr;

        len = cur + len;
        while (cur < len) {
                if ((cur & 31) == 0 && (len - cur) >= 32) {
                        /* fast path: clear whole word at once */
                        addr = bm + (cur >> 3);
                        *addr = 0;
                        cur += 32;
                        continue;
                }
                mb_clear_bit(cur, bm);
                cur++;
        }
}

void ext4_set_bits(void *bm, int cur, int len)
{
        __u32 *addr;

        len = cur + len;
        while (cur < len) {
                if ((cur & 31) == 0 && (len - cur) >= 32) {
                        /* fast path: set whole word at once */
                        addr = bm + (cur >> 3);
                        *addr = 0xffffffff;
                        cur += 32;
                        continue;
                }
                mb_set_bit(cur, bm);
                cur++;
        }
}

static void mb_free_blocks(struct inode *inode, struct ext4_buddy *e4b,
                          int first, int count)
{
        int block = 0;
        int max = 0;
        int order;
        void *buddy;
        void *buddy2;
        struct super_block *sb = e4b->bd_sb;

        BUG_ON(first + count > (sb->s_blocksize << 3));
        assert_spin_locked(ext4_group_lock_ptr(sb, e4b->bd_group));
        mb_check_buddy(e4b);
        mb_free_blocks_double(inode, e4b, first, count);

        e4b->bd_info->bb_free += count;
        if (first < e4b->bd_info->bb_first_free)
                e4b->bd_info->bb_first_free = first;

        /* let's maintain fragments counter */
        if (first != 0)
                block = !mb_test_bit(first - 1, e4b->bd_bitmap);
        if (first + count < EXT4_SB(sb)->s_mb_maxs[0])
                max = !mb_test_bit(first + count, e4b->bd_bitmap);
        if (block && max)
                e4b->bd_info->bb_fragments--;
        else if (!block && !max)
                e4b->bd_info->bb_fragments++;

        /* let's maintain buddy itself */
        while (count-- > 0) {
                block = first++;
                order = 0;

                if (!mb_test_bit(block, e4b->bd_bitmap)) {
                        ext4_fsblk_t blocknr;

                        blocknr = ext4_group_first_block_no(sb, e4b->bd_group);
                        blocknr += EXT4_C2B(EXT4_SB(sb), block);
                        ext4_grp_locked_error(sb, e4b->bd_group,
                                              inode ? inode->i_ino : 0,
                                              blocknr,
                                              "freeing already freed block "
                                              "(bit %u)", block);
                }
                mb_clear_bit(block, e4b->bd_bitmap);
                e4b->bd_info->bb_counters[order]++;

                /* start of the buddy */
                buddy = mb_find_buddy(e4b, order, &max);

                do {
                        block &= ~1UL;
                        if (mb_test_bit(block, buddy) ||
                                        mb_test_bit(block + 1, buddy))
                                break;

                        /* both the buddies are free, try to coalesce them */
                        buddy2 = mb_find_buddy(e4b, order + 1, &max);

                        if (!buddy2)
                                break;

                        if (order > 0) {
                                /* for special purposes, we don't set
                                 * free bits in bitmap */
                                mb_set_bit(block, buddy);
                                mb_set_bit(block + 1, buddy);
                        }
                        e4b->bd_info->bb_counters[order]--;
                        e4b->bd_info->bb_counters[order]--;

                        block = block >> 1;
                        order++;
                        e4b->bd_info->bb_counters[order]++;

                        mb_clear_bit(block, buddy2);
                        buddy = buddy2;
                } while (1);
        }
        mb_set_largest_free_order(sb, e4b->bd_info);
        mb_check_buddy(e4b);
}

static int mb_find_extent(struct ext4_buddy *e4b, int order, int block,
                                int needed, struct ext4_free_extent *ex)
{
        int next = block;
        int max;
        void *buddy;

        assert_spin_locked(ext4_group_lock_ptr(e4b->bd_sb, e4b->bd_group));
        BUG_ON(ex == NULL);

        buddy = mb_find_buddy(e4b, order, &max);
        BUG_ON(buddy == NULL);
        BUG_ON(block >= max);
        if (mb_test_bit(block, buddy)) {
                ex->fe_len = 0;
                ex->fe_start = 0;
                ex->fe_group = 0;
                return 0;
        }

        /* FIXME dorp order completely ? */
        if (likely(order == 0)) {
                /* find actual order */
                order = mb_find_order_for_block(e4b, block);
                block = block >> order;
        }

        ex->fe_len = 1 << order;
        ex->fe_start = block << order;
        ex->fe_group = e4b->bd_group;

        /* calc difference from given start */
        next = next - ex->fe_start;
        ex->fe_len -= next;
        ex->fe_start += next;

        while (needed > ex->fe_len &&
               (buddy = mb_find_buddy(e4b, order, &max))) {

                if (block + 1 >= max)
                        break;

                next = (block + 1) * (1 << order);
                if (mb_test_bit(next, e4b->bd_bitmap))
                        break;

                order = mb_find_order_for_block(e4b, next);

                block = next >> order;
                ex->fe_len += 1 << order;
        }

        BUG_ON(ex->fe_start + ex->fe_len > (1 << (e4b->bd_blkbits + 3)));
        return ex->fe_len;
}

static int mb_mark_used(struct ext4_buddy *e4b, struct ext4_free_extent *ex)
{
        int ord;
        int mlen = 0;
        int max = 0;
        int cur;
        int start = ex->fe_start;
        int len = ex->fe_len;
        unsigned ret = 0;
        int len0 = len;
        void *buddy;

        BUG_ON(start + len > (e4b->bd_sb->s_blocksize << 3));
        BUG_ON(e4b->bd_group != ex->fe_group);
        assert_spin_locked(ext4_group_lock_ptr(e4b->bd_sb, e4b->bd_group));
        mb_check_buddy(e4b);
        mb_mark_used_double(e4b, start, len);

        e4b->bd_info->bb_free -= len;
        if (e4b->bd_info->bb_first_free == start)
                e4b->bd_info->bb_first_free += len;

        /* let's maintain fragments counter */
        if (start != 0)
                mlen = !mb_test_bit(start - 1, e4b->bd_bitmap);
        if (start + len < EXT4_SB(e4b->bd_sb)->s_mb_maxs[0])
                max = !mb_test_bit(start + len, e4b->bd_bitmap);
        if (mlen && max)
                e4b->bd_info->bb_fragments++;
        else if (!mlen && !max)
                e4b->bd_info->bb_fragments--;

        /* let's maintain buddy itself */
        while (len) {
                ord = mb_find_order_for_block(e4b, start);

                if (((start >> ord) << ord) == start && len >= (1 << ord)) {
                        /* the whole chunk may be allocated at once! */
                        mlen = 1 << ord;
                        buddy = mb_find_buddy(e4b, ord, &max);
                        BUG_ON((start >> ord) >= max);
                        mb_set_bit(start >> ord, buddy);
                        e4b->bd_info->bb_counters[ord]--;
                        start += mlen;
                        len -= mlen;
                        BUG_ON(len < 0);
                        continue;
                }

                /* store for history */
                if (ret == 0)
                        ret = len | (ord << 16);

                /* we have to split large buddy */
                BUG_ON(ord <= 0);
                buddy = mb_find_buddy(e4b, ord, &max);
                mb_set_bit(start >> ord, buddy);
                e4b->bd_info->bb_counters[ord]--;

                ord--;
                cur = (start >> ord) & ~1U;
                buddy = mb_find_buddy(e4b, ord, &max);
                mb_clear_bit(cur, buddy);
                mb_clear_bit(cur + 1, buddy);
                e4b->bd_info->bb_counters[ord]++;
                e4b->bd_info->bb_counters[ord]++;
        }
        mb_set_largest_free_order(e4b->bd_sb, e4b->bd_info);

        ext4_set_bits(e4b->bd_bitmap, ex->fe_start, len0);
        mb_check_buddy(e4b);

        return ret;
}

/*
 * Must be called under group lock!
 */
static void ext4_mb_use_best_found(struct ext4_allocation_context *ac,
                                        struct ext4_buddy *e4b)
{
        struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb);
        int ret;

        BUG_ON(ac->ac_b_ex.fe_group != e4b->bd_group);
        BUG_ON(ac->ac_status == AC_STATUS_FOUND);

        ac->ac_b_ex.fe_len = min(ac->ac_b_ex.fe_len, ac->ac_g_ex.fe_len);
        ac->ac_b_ex.fe_logical = ac->ac_g_ex.fe_logical;
        ret = mb_mark_used(e4b, &ac->ac_b_ex);

        /* preallocation can change ac_b_ex, thus we store actually
         * allocated blocks for history */
        ac->ac_f_ex = ac->ac_b_ex;

        ac->ac_status = AC_STATUS_FOUND;
        ac->ac_tail = ret & 0xffff;
        ac->ac_buddy = ret >> 16;

        /*
         * take the page reference. We want the page to be pinned
         * so that we don't get a ext4_mb_init_cache_call for this
         * group until we update the bitmap. That would mean we
         * double allocate blocks. The reference is dropped
         * in ext4_mb_release_context
         */
        ac->ac_bitmap_page = e4b->bd_bitmap_page;
        get_page(ac->ac_bitmap_page);
        ac->ac_buddy_page = e4b->bd_buddy_page;
        get_page(ac->ac_buddy_page);
        /* store last allocated for subsequent stream allocation */
        if (ac->ac_flags & EXT4_MB_STREAM_ALLOC) {
                spin_lock(&sbi->s_md_lock);
                sbi->s_mb_last_group = ac->ac_f_ex.fe_group;
                sbi->s_mb_last_start = ac->ac_f_ex.fe_start;
                spin_unlock(&sbi->s_md_lock);
        }
}

/*
 * regular allocator, for general purposes allocation
 */

static void ext4_mb_check_limits(struct ext4_allocation_context *ac,
                                        struct ext4_buddy *e4b,
                                        int finish_group)
{
        struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb);
        struct ext4_free_extent *bex = &ac->ac_b_ex;
        struct ext4_free_extent *gex = &ac->ac_g_ex;
        struct ext4_free_extent ex;
        int max;

        if (ac->ac_status == AC_STATUS_FOUND)
                return;
        /*
         * We don't want to scan for a whole year
         */
        if (ac->ac_found > sbi->s_mb_max_to_scan &&
                        !(ac->ac_flags & EXT4_MB_HINT_FIRST)) {
                ac->ac_status = AC_STATUS_BREAK;
                return;
        }

        /*
         * Haven't found good chunk so far, let's continue
         */
        if (bex->fe_len < gex->fe_len)
                return;

        if ((finish_group || ac->ac_found > sbi->s_mb_min_to_scan)
                        && bex->fe_group == e4b->bd_group) {
                /* recheck chunk's availability - we don't know
                 * when it was found (within this lock-unlock
                 * period or not) */
                max = mb_find_extent(e4b, 0, bex->fe_start, gex->fe_len, &ex);
                if (max >= gex->fe_len) {
                        ext4_mb_use_best_found(ac, e4b);
                        return;
                }
        }
}

/*
 * The routine checks whether found extent is good enough. If it is,
 * then the extent gets marked used and flag is set to the context
 * to stop scanning. Otherwise, the extent is compared with the
 * previous found extent and if new one is better, then it's stored
 * in the context. Later, the best found extent will be used, if
 * mballoc can't find good enough extent.
 *
 * FIXME: real allocation policy is to be designed yet!
 */
static void ext4_mb_measure_extent(struct ext4_allocation_context *ac,
                                        struct ext4_free_extent *ex,
                                        struct ext4_buddy *e4b)
{
        struct ext4_free_extent *bex = &ac->ac_b_ex;
        struct ext4_free_extent *gex = &ac->ac_g_ex;

        BUG_ON(ex->fe_len <= 0);
        BUG_ON(ex->fe_len > EXT4_CLUSTERS_PER_GROUP(ac->ac_sb));
        BUG_ON(ex->fe_start >= EXT4_CLUSTERS_PER_GROUP(ac->ac_sb));
        BUG_ON(ac->ac_status != AC_STATUS_CONTINUE);

        ac->ac_found++;

        /*
         * The special case - take what you catch first
         */
        if (unlikely(ac->ac_flags & EXT4_MB_HINT_FIRST)) {
                *bex = *ex;
                ext4_mb_use_best_found(ac, e4b);
                return;
        }

        /*
         * Let's check whether the chuck is good enough
         */
        if (ex->fe_len == gex->fe_len) {
                *bex = *ex;
                ext4_mb_use_best_found(ac, e4b);
                return;
        }

        /*
         * If this is first found extent, just store it in the context
         */
        if (bex->fe_len == 0) {
                *bex = *ex;
                return;
        }

        /*
         * If new found extent is better, store it in the context
         */
        if (bex->fe_len < gex->fe_len) {
                /* if the request isn't satisfied, any found extent
                 * larger than previous best one is better */
                if (ex->fe_len > bex->fe_len)
                        *bex = *ex;
        } else if (ex->fe_len > gex->fe_len) {
                /* if the request is satisfied, then we try to find
                 * an extent that still satisfy the request, but is
                 * smaller than previous one */
                if (ex->fe_len < bex->fe_len)
                        *bex = *ex;
        }

        ext4_mb_check_limits(ac, e4b, 0);
}

static noinline_for_stack
int ext4_mb_try_best_found(struct ext4_allocation_context *ac,
                                        struct ext4_buddy *e4b)
{
        struct ext4_free_extent ex = ac->ac_b_ex;
        ext4_group_t group = ex.fe_group;
        int max;
        int err;

        BUG_ON(ex.fe_len <= 0);
        err = ext4_mb_load_buddy(ac->ac_sb, group, e4b);
        if (err)
                return err;

        ext4_lock_group(ac->ac_sb, group);
        max = mb_find_extent(e4b, 0, ex.fe_start, ex.fe_len, &ex);

        if (max > 0) {
                ac->ac_b_ex = ex;
                ext4_mb_use_best_found(ac, e4b);
        }

        ext4_unlock_group(ac->ac_sb, group);
        ext4_mb_unload_buddy(e4b);

        return 0;
}

static noinline_for_stack
int ext4_mb_find_by_goal(struct ext4_allocation_context *ac,
                                struct ext4_buddy *e4b)
{
        ext4_group_t group = ac->ac_g_ex.fe_group;
        int max;
        int err;
        struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb);
        struct ext4_free_extent ex;

        if (!(ac->ac_flags & EXT4_MB_HINT_TRY_GOAL))
                return 0;

        err = ext4_mb_load_buddy(ac->ac_sb, group, e4b);
        if (err)
                return err;

        ext4_lock_group(ac->ac_sb, group);
        max = mb_find_extent(e4b, 0, ac->ac_g_ex.fe_start,
                             ac->ac_g_ex.fe_len, &ex);

        if (max >= ac->ac_g_ex.fe_len && ac->ac_g_ex.fe_len == sbi->s_stripe) {
                ext4_fsblk_t start;

                start = ext4_group_first_block_no(ac->ac_sb, e4b->bd_group) +
                        ex.fe_start;
                /* use do_div to get remainder (would be 64-bit modulo) */
                if (do_div(start, sbi->s_stripe) == 0) {
                        ac->ac_found++;
                        ac->ac_b_ex = ex;
                        ext4_mb_use_best_found(ac, e4b);
                }
        } else if (max >= ac->ac_g_ex.fe_len) {
                BUG_ON(ex.fe_len <= 0);
                BUG_ON(ex.fe_group != ac->ac_g_ex.fe_group);
                BUG_ON(ex.fe_start != ac->ac_g_ex.fe_start);
                ac->ac_found++;
                ac->ac_b_ex = ex;
                ext4_mb_use_best_found(ac, e4b);
        } else if (max > 0 && (ac->ac_flags & EXT4_MB_HINT_MERGE)) {
                /* Sometimes, caller may want to merge even small
                 * number of blocks to an existing extent */
                BUG_ON(ex.fe_len <= 0);
                BUG_ON(ex.fe_group != ac->ac_g_ex.fe_group);
                BUG_ON(ex.fe_start != ac->ac_g_ex.fe_start);
                ac->ac_found++;
                ac->ac_b_ex = ex;
                ext4_mb_use_best_found(ac, e4b);
        }
        ext4_unlock_group(ac->ac_sb, group);
        ext4_mb_unload_buddy(e4b);

        return 0;
}

/*
 * The routine scans buddy structures (not bitmap!) from given order
 * to max order and tries to find big enough chunk to satisfy the req
 */
static noinline_for_stack
void ext4_mb_simple_scan_group(struct ext4_allocation_context *ac,
                                        struct ext4_buddy *e4b)
{
        struct super_block *sb = ac->ac_sb;
        struct ext4_group_info *grp = e4b->bd_info;
        void *buddy;
        int i;
        int k;
        int max;

        BUG_ON(ac->ac_2order <= 0);
        for (i = ac->ac_2order; i <= sb->s_blocksize_bits + 1; i++) {
                if (grp->bb_counters[i] == 0)
                        continue;

                buddy = mb_find_buddy(e4b, i, &max);
                BUG_ON(buddy == NULL);

                k = mb_find_next_zero_bit(buddy, max, 0);
                BUG_ON(k >= max);

                ac->ac_found++;

                ac->ac_b_ex.fe_len = 1 << i;
                ac->ac_b_ex.fe_start = k << i;
                ac->ac_b_ex.fe_group = e4b->bd_group;

                ext4_mb_use_best_found(ac, e4b);

                BUG_ON(ac->ac_b_ex.fe_len != ac->ac_g_ex.fe_len);

                if (EXT4_SB(sb)->s_mb_stats)
                        atomic_inc(&EXT4_SB(sb)->s_bal_2orders);

                break;
        }
}

/*
 * The routine scans the group and measures all found extents.
 * In order to optimize scanning, caller must pass number of
 * free blocks in the group, so the routine can know upper limit.
 */
static noinline_for_stack
void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac,
                                        struct ext4_buddy *e4b)
{
        struct super_block *sb = ac->ac_sb;
        void *bitmap = e4b->bd_bitmap;
        struct ext4_free_extent ex;
        int i;
        int free;

        free = e4b->bd_info->bb_free;
        BUG_ON(free <= 0);

        i = e4b->bd_info->bb_first_free;

        while (free && ac->ac_status == AC_STATUS_CONTINUE) {
                i = mb_find_next_zero_bit(bitmap,
                                                EXT4_CLUSTERS_PER_GROUP(sb), i);
                if (i >= EXT4_CLUSTERS_PER_GROUP(sb)) {
                        /*
                         * IF we have corrupt bitmap, we won't find any
                         * free blocks even though group info says we
                         * we have free blocks
                         */
                        ext4_grp_locked_error(sb, e4b->bd_group, 0, 0,
                                        "%d free clusters as per "
                                        "group info. But bitmap says 0",
                                        free);
                        break;
                }

                mb_find_extent(e4b, 0, i, ac->ac_g_ex.fe_len, &ex);
                BUG_ON(ex.fe_len <= 0);
                if (free < ex.fe_len) {
                        ext4_grp_locked_error(sb, e4b->bd_group, 0, 0,
                                        "%d free clusters as per "
                                        "group info. But got %d blocks",
                                        free, ex.fe_len);
                        /*
                         * The number of free blocks differs. This mostly
                         * indicate that the bitmap is corrupt. So exit
                         * without claiming the space.
                         */
                        break;
                }

                ext4_mb_measure_extent(ac, &ex, e4b);

                i += ex.fe_len;
                free -= ex.fe_len;
        }

        ext4_mb_check_limits(ac, e4b, 1);
}

/*
 * This is a special case for storages like raid5
 * we try to find stripe-aligned chunks for stripe-size-multiple requests
 */
static noinline_for_stack
void ext4_mb_scan_aligned(struct ext4_allocation_context *ac,
                                 struct ext4_buddy *e4b)
{
        struct super_block *sb = ac->ac_sb;
        struct ext4_sb_info *sbi = EXT4_SB(sb);
        void *bitmap = e4b->bd_bitmap;
        struct ext4_free_extent ex;
        ext4_fsblk_t first_group_block;
        ext4_fsblk_t a;
        ext4_grpblk_t i;
        int max;

        BUG_ON(sbi->s_stripe == 0);

        /* find first stripe-aligned block in group */
        first_group_block = ext4_group_first_block_no(sb, e4b->bd_group);

        a = first_group_block + sbi->s_stripe - 1;
        do_div(a, sbi->s_stripe);
        i = (a * sbi->s_stripe) - first_group_block;

        while (i < EXT4_CLUSTERS_PER_GROUP(sb)) {
                if (!mb_test_bit(i, bitmap)) {
                        max = mb_find_extent(e4b, 0, i, sbi->s_stripe, &ex);
                        if (max >= sbi->s_stripe) {
                                ac->ac_found++;
                                ac->ac_b_ex = ex;
                                ext4_mb_use_best_found(ac, e4b);
                                break;
                        }
                }
                i += sbi->s_stripe;
        }
}

/* This is now called BEFORE we load the buddy bitmap. */
static int ext4_mb_good_group(struct ext4_allocation_context *ac,
                                ext4_group_t group, int cr)
{
        unsigned free, fragments;
        int flex_size = ext4_flex_bg_size(EXT4_SB(ac->ac_sb));
        struct ext4_group_info *grp = ext4_get_group_info(ac->ac_sb, group);

        BUG_ON(cr < 0 || cr >= 4);

        /* We only do this if the grp has never been initialized */
        if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) {
                int ret = ext4_mb_init_group(ac->ac_sb, group);
                if (ret)
                        return 0;
        }

        free = grp->bb_free;
        fragments = grp->bb_fragments;
        if (free == 0)
                return 0;
        if (fragments == 0)
                return 0;

        switch (cr) {
        case 0:
                BUG_ON(ac->ac_2order == 0);

                if (grp->bb_largest_free_order < ac->ac_2order)
                        return 0;

                /* Avoid using the first bg of a flexgroup for data files */
                if ((ac->ac_flags & EXT4_MB_HINT_DATA) &&
                    (flex_size >= EXT4_FLEX_SIZE_DIR_ALLOC_SCHEME) &&
                    ((group % flex_size) == 0))
                        return 0;

                return 1;
        case 1:
                if ((free / fragments) >= ac->ac_g_ex.fe_len)
                        return 1;
                break;
        case 2:
                if (free >= ac->ac_g_ex.fe_len)
                        return 1;
                break;
        case 3:
                return 1;
        default:
                BUG();
        }

        return 0;
}

static noinline_for_stack int
ext4_mb_regular_allocator(struct ext4_allocation_context *ac)
{
        ext4_group_t ngroups, group, i;
        int cr;
        int err = 0;
        struct ext4_sb_info *sbi;
        struct super_block *sb;
        struct ext4_buddy e4b;

        sb = ac->ac_sb;
        sbi = EXT4_SB(sb);
        ngroups = ext4_get_groups_count(sb);
        /* non-extent files are limited to low blocks/groups */
        if (!(ext4_test_inode_flag(ac->ac_inode, EXT4_INODE_EXTENTS)))
                ngroups = sbi->s_blockfile_groups;

        BUG_ON(ac->ac_status == AC_STATUS_FOUND);

        /* first, try the goal */
        err = ext4_mb_find_by_goal(ac, &e4b);
        if (err || ac->ac_status == AC_STATUS_FOUND)
                goto out;

        if (unlikely(ac->ac_flags & EXT4_MB_HINT_GOAL_ONLY))
                goto out;

        /*
         * ac->ac2_order is set only if the fe_len is a power of 2
         * if ac2_order is set we also set criteria to 0 so that we
         * try exact allocation using buddy.
         */
        i = fls(ac->ac_g_ex.fe_len);
        ac->ac_2order = 0;
        /*
         * We search using buddy data only if the order of the request
         * is greater than equal to the sbi_s_mb_order2_reqs
         * You can tune it via /sys/fs/ext4/<partition>/mb_order2_req
         */
        if (i >= sbi->s_mb_order2_reqs) {
                /*
                 * This should tell if fe_len is exactly power of 2
                 */
                if ((ac->ac_g_ex.fe_len & (~(1 << (i - 1)))) == 0)
                        ac->ac_2order = i - 1;
        }

        /* if stream allocation is enabled, use global goal */
        if (ac->ac_flags & EXT4_MB_STREAM_ALLOC) {
                /* TBD: may be hot point */
                spin_lock(&sbi->s_md_lock);
                ac->ac_g_ex.fe_group = sbi->s_mb_last_group;
                ac->ac_g_ex.fe_start = sbi->s_mb_last_start;
                spin_unlock(&sbi->s_md_lock);
        }

        /* Let's just scan groups to find more-less suitable blocks */
        cr = ac->ac_2order ? 0 : 1;
        /*
         * cr == 0 try to get exact allocation,
         * cr == 3  try to get anything
         */
repeat:
        for (; cr < 4 && ac->ac_status == AC_STATUS_CONTINUE; cr++) {
                ac->ac_criteria = cr;
                /*
                 * searching for the right group start
                 * from the goal value specified
                 */
                group = ac->ac_g_ex.fe_group;

                for (i = 0; i < ngroups; group++, i++) {
                        if (group == ngroups)
                                group = 0;

                        /* This now checks without needing the buddy page */
                        if (!ext4_mb_good_group(ac, group, cr))
                                continue;

                        err = ext4_mb_load_buddy(sb, group, &e4b);
                        if (err)
                                goto out;

                        ext4_lock_group(sb, group);

                        /*
                         * We need to check again after locking the
                         * block group
                         */
                        if (!ext4_mb_good_group(ac, group, cr)) {
                                ext4_unlock_group(sb, group);

                                ext4_mb_unload_buddy(&e4b);
                                continue;
                        }

                        ac->ac_groups_scanned++;
                        if (cr == 0)
                                ext4_mb_simple_scan_group(ac, &e4b);
                        else if (cr == 1 && sbi->s_stripe &&
                                        !(ac->ac_g_ex.fe_len % sbi->s_stripe))
                                ext4_mb_scan_aligned(ac, &e4b);
                        else
                                ext4_mb_complex_scan_group(ac, &e4b);

                        ext4_unlock_group(sb, group);
                        ext4_mb_unload_buddy(&e4b);

                        if (ac->ac_status != AC_STATUS_CONTINUE)
                                break;
                }
        }

        if (ac->ac_b_ex.fe_len > 0 && ac->ac_status != AC_STATUS_FOUND &&
            !(ac->ac_flags & EXT4_MB_HINT_FIRST)) {
                /*
                 * We've been searching too long. Let's try to allocate
                 * the best chunk we've found so far
                 */

                ext4_mb_try_best_found(ac, &e4b);
                if (ac->ac_status != AC_STATUS_FOUND) {
                        /*
                         * Someone more lucky has already allocated it.
                         * The only thing we can do is just take first
                         * found block(s)
                        printk(KERN_DEBUG "EXT4-fs: someone won our chunk\n");
                         */
                        ac->ac_b_ex.fe_group = 0;
                        ac->ac_b_ex.fe_start = 0;
                        ac->ac_b_ex.fe_len = 0;
                        ac->ac_status = AC_STATUS_CONTINUE;
                        ac->ac_flags |= EXT4_MB_HINT_FIRST;
                        cr = 3;
                        atomic_inc(&sbi->s_mb_lost_chunks);
                        goto repeat;
                }
        }
out:
        return err;
}

static void *ext4_mb_seq_groups_start(struct seq_file *seq, loff_t *pos)
{
        struct super_block *sb = seq->private;
        ext4_group_t group;

        if (*pos < 0 || *pos >= ext4_get_groups_count(sb))
                return NULL;
        group = *pos + 1;
        return (void *) ((unsigned long) group);
}

static void *ext4_mb_seq_groups_next(struct seq_file *seq, void *v, loff_t *pos)
{
        struct super_block *sb = seq->private;
        ext4_group_t group;

        ++*pos;
        if (*pos < 0 || *pos >= ext4_get_groups_count(sb))
                return NULL;
        group = *pos + 1;
        return (void *) ((unsigned long) group);
}

static int ext4_mb_seq_groups_show(struct seq_file *seq, void *v)
{
        struct super_block *sb = seq->private;
        ext4_group_t group = (ext4_group_t) ((unsigned long) v);
        int i;
        int err, buddy_loaded = 0;
        struct ext4_buddy e4b;
        struct ext4_group_info *grinfo;
        struct sg {
                struct ext4_group_info info;
                ext4_grpblk_t counters[16];
        } sg;

        group--;
        if (group == 0)
                seq_printf(seq, "#%-5s: %-5s %-5s %-5s "
                                "[ %-5s %-5s %-5s %-5s %-5s %-5s %-5s "
                                  "%-5s %-5s %-5s %-5s %-5s %-5s %-5s ]\n",
                           "group", "free", "frags", "first",
                           "2^0", "2^1", "2^2", "2^3", "2^4", "2^5", "2^6",
                           "2^7", "2^8", "2^9", "2^10", "2^11", "2^12", "2^13");

        i = (sb->s_blocksize_bits + 2) * sizeof(sg.info.bb_counters[0]) +
                sizeof(struct ext4_group_info);
        grinfo = ext4_get_group_info(sb, group);
        /* Load the group info in memory only if not already loaded. */
        if (unlikely(EXT4_MB_GRP_NEED_INIT(grinfo))) {
                err = ext4_mb_load_buddy(sb, group, &e4b);
                if (err) {
                        seq_printf(seq, "#%-5u: I/O error\n", group);
                        return 0;
                }
                buddy_loaded = 1;
        }

        memcpy(&sg, ext4_get_group_info(sb, group), i);

        if (buddy_loaded)
                ext4_mb_unload_buddy(&e4b);

        seq_printf(seq, "#%-5u: %-5u %-5u %-5u [", group, sg.info.bb_free,
                        sg.info.bb_fragments, sg.info.bb_first_free);
        for (i = 0; i <= 13; i++)
                seq_printf(seq, " %-5u", i <= sb->s_blocksize_bits + 1 ?
                                sg.info.bb_counters[i] : 0);
        seq_printf(seq, " ]\n");

        return 0;
}

static void ext4_mb_seq_groups_stop(struct seq_file *seq, void *v)
{
}

static const struct seq_operations ext4_mb_seq_groups_ops = {
        .start  = ext4_mb_seq_groups_start,
        .next   = ext4_mb_seq_groups_next,
        .stop   = ext4_mb_seq_groups_stop,
        .show   = ext4_mb_seq_groups_show,
};

static int ext4_mb_seq_groups_open(struct inode *inode, struct file *file)
{
        struct super_block *sb = PDE(inode)->data;
        int rc;

        rc = seq_open(file, &ext4_mb_seq_groups_ops);
        if (rc == 0) {
                struct seq_file *m = file->private_data;
                m->private = sb;
        }
        return rc;

}

static const struct file_operations ext4_mb_seq_groups_fops = {
        .owner          = THIS_MODULE,
        .open           = ext4_mb_seq_groups_open,
        .read           = seq_read,
        .llseek         = seq_lseek,
        .release        = seq_release,
};

static struct kmem_cache *get_groupinfo_cache(int blocksize_bits)
{
        int cache_index = blocksize_bits - EXT4_MIN_BLOCK_LOG_SIZE;
        struct kmem_cache *cachep = ext4_groupinfo_caches[cache_index];

        BUG_ON(!cachep);
        return cachep;
}

/* Create and initialize ext4_group_info data for the given group. */
int ext4_mb_add_groupinfo(struct super_block *sb, ext4_group_t group,
                          struct ext4_group_desc *desc)
{
        int i;
        int metalen = 0;
        struct ext4_sb_info *sbi = EXT4_SB(sb);
        struct ext4_group_info **meta_group_info;
        struct kmem_cache *cachep = get_groupinfo_cache(sb->s_blocksize_bits);

        /*
         * First check if this group is the first of a reserved block.
         * If it's true, we have to allocate a new table of pointers
         * to ext4_group_info structures
         */
        if (group % EXT4_DESC_PER_BLOCK(sb) == 0) {
                metalen = sizeof(*meta_group_info) <<
                        EXT4_DESC_PER_BLOCK_BITS(sb);
                meta_group_info = kmalloc(metalen, GFP_KERNEL);
                if (meta_group_info == NULL) {
                        ext4_msg(sb, KERN_ERR, "can't allocate mem "
                                 "for a buddy group");
                        goto exit_meta_group_info;
                }
                sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)] =
                        meta_group_info;
        }

        meta_group_info =
                sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)];
        i = group & (EXT4_DESC_PER_BLOCK(sb) - 1);

        meta_group_info[i] = kmem_cache_alloc(cachep, GFP_KERNEL);
        if (meta_group_info[i] == NULL) {
                ext4_msg(sb, KERN_ERR, "can't allocate buddy mem");
                goto exit_group_info;
        }
        memset(meta_group_info[i], 0, kmem_cache_size(cachep));
        set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT,
                &(meta_group_info[i]->bb_state));

        /*
         * initialize bb_free to be able to skip
         * empty groups without initialization
         */
        if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
                meta_group_info[i]->bb_free =
                        ext4_free_clusters_after_init(sb, group, desc);
        } else {
                meta_group_info[i]->bb_free =
                        ext4_free_group_clusters(sb, desc);
        }

        INIT_LIST_HEAD(&meta_group_info[i]->bb_prealloc_list);
        init_rwsem(&meta_group_info[i]->alloc_sem);
        meta_group_info[i]->bb_free_root = RB_ROOT;
        meta_group_info[i]->bb_largest_free_order = -1;  /* uninit */

#ifdef DOUBLE_CHECK
        {
                struct buffer_head *bh;
                meta_group_info[i]->bb_bitmap =
                        kmalloc(sb->s_blocksize, GFP_KERNEL);
                BUG_ON(meta_group_info[i]->bb_bitmap == NULL);
                bh = ext4_read_block_bitmap(sb, group);
                BUG_ON(bh == NULL);
                memcpy(meta_group_info[i]->bb_bitmap, bh->b_data,
                        sb->s_blocksize);
                put_bh(bh);
        }
#endif

        return 0;

exit_group_info:
        /* If a meta_group_info table has been allocated, release it now */
        if (group % EXT4_DESC_PER_BLOCK(sb) == 0) {
                kfree(sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)]);
                sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)] = NULL;
        }
exit_meta_group_info:
        return -ENOMEM;
} /* ext4_mb_add_groupinfo */

static int ext4_mb_init_backend(struct super_block *sb)
{
        ext4_group_t ngroups = ext4_get_groups_count(sb);
        ext4_group_t i;
        struct ext4_sb_info *sbi = EXT4_SB(sb);
        struct ext4_super_block *es = sbi->s_es;
        int num_meta_group_infos;
        int num_meta_group_infos_max;
        int array_size;
        struct ext4_group_desc *desc;
        struct kmem_cache *cachep;

        /* This is the number of blocks used by GDT */
        num_meta_group_infos = (ngroups + EXT4_DESC_PER_BLOCK(sb) -
                                1) >> EXT4_DESC_PER_BLOCK_BITS(sb);

        /*
         * This is the total number of blocks used by GDT including
         * the number of reserved blocks for GDT.
         * The s_group_info array is allocated with this value
         * to allow a clean online resize without a complex
         * manipulation of pointer.
         * The drawback is the unused memory when no resize
         * occurs but it's very low in terms of pages
         * (see comments below)
         * Need to handle this properly when META_BG resizing is allowed
         */
        num_meta_group_infos_max = num_meta_group_infos +
                                le16_to_cpu(es->s_reserved_gdt_blocks);

        /*
         * array_size is the size of s_group_info array. We round it
         * to the next power of two because this approximation is done
         * internally by kmalloc so we can have some more memory
         * for free here (e.g. may be used for META_BG resize).
         */
        array_size = 1;
        while (array_size < sizeof(*sbi->s_group_info) *
               num_meta_group_infos_max)
                array_size = array_size << 1;
        /* An 8TB filesystem with 64-bit pointers requires a 4096 byte
         * kmalloc. A 128kb malloc should suffice for a 256TB filesystem.
         * So a two level scheme suffices for now. */
        sbi->s_group_info = ext4_kvzalloc(array_size, GFP_KERNEL);
        if (sbi->s_group_info == NULL) {
                ext4_msg(sb, KERN_ERR, "can't allocate buddy meta group");
                return -ENOMEM;
        }
        sbi->s_buddy_cache = new_inode(sb);
        if (sbi->s_buddy_cache == NULL) {
                ext4_msg(sb, KERN_ERR, "can't get new inode");
                goto err_freesgi;
        }
        /* To avoid potentially colliding with an valid on-disk inode number,
         * use EXT4_BAD_INO for the buddy cache inode number.  This inode is
         * not in the inode hash, so it should never be found by iget(), but
         * this will avoid confusion if it ever shows up during debugging. */
        sbi->s_buddy_cache->i_ino = EXT4_BAD_INO;
        EXT4_I(sbi->s_buddy_cache)->i_disksize = 0;
        for (i = 0; i < ngroups; i++) {
                desc = ext4_get_group_desc(sb, i, NULL);
                if (desc == NULL) {
                        ext4_msg(sb, KERN_ERR, "can't read descriptor %u", i);
                        goto err_freebuddy;
                }
                if (ext4_mb_add_groupinfo(sb, i, desc) != 0)
                        goto err_freebuddy;
        }

        return 0;

err_freebuddy:
        cachep = get_groupinfo_cache(sb->s_blocksize_bits);
        while (i-- > 0)
                kmem_cache_free(cachep, ext4_get_group_info(sb, i));
        i = num_meta_group_infos;
        while (i-- > 0)
                kfree(sbi->s_group_info[i]);
        iput(sbi->s_buddy_cache);
err_freesgi:
        ext4_kvfree(sbi->s_group_info);
        return -ENOMEM;
}

static void ext4_groupinfo_destroy_slabs(void)
{
        int i;

        for (i = 0; i < NR_GRPINFO_CACHES; i++) {
                if (ext4_groupinfo_caches[i])
                        kmem_cache_destroy(ext4_groupinfo_caches[i]);
                ext4_groupinfo_caches[i] = NULL;
        }
}

static int ext4_groupinfo_create_slab(size_t size)
{
        static DEFINE_MUTEX(ext4_grpinfo_slab_create_mutex);
        int slab_size;
        int blocksize_bits = order_base_2(size);
        int cache_index = blocksize_bits - EXT4_MIN_BLOCK_LOG_SIZE;
        struct kmem_cache *cachep;

        if (cache_index >= NR_GRPINFO_CACHES)
                return -EINVAL;

        if (unlikely(cache_index < 0))
                cache_index = 0;

        mutex_lock(&ext4_grpinfo_slab_create_mutex);
        if (ext4_groupinfo_caches[cache_index]) {
                mutex_unlock(&ext4_grpinfo_slab_create_mutex);
                return 0;       /* Already created */
        }

        slab_size = offsetof(struct ext4_group_info,
                                bb_counters[blocksize_bits + 2]);

        cachep = kmem_cache_create(ext4_groupinfo_slab_names[cache_index],
                                        slab_size, 0, SLAB_RECLAIM_ACCOUNT,
                                        NULL);

        ext4_groupinfo_caches[cache_index] = cachep;

        mutex_unlock(&ext4_grpinfo_slab_create_mutex);
        if (!cachep) {
                printk(KERN_EMERG
                       "EXT4-fs: no memory for groupinfo slab cache\n");
                return -ENOMEM;
        }

        return 0;
}

int ext4_mb_init(struct super_block *sb)
{
        struct ext4_sb_info *sbi = EXT4_SB(sb);
        unsigned i, j;
        unsigned offset;
        unsigned max;
        int ret;

        i = (sb->s_blocksize_bits + 2) * sizeof(*sbi->s_mb_offsets);

        sbi->s_mb_offsets = kmalloc(i, GFP_KERNEL);
        if (sbi->s_mb_offsets == NULL) {
                ret = -ENOMEM;
                goto out;
        }

        i = (sb->s_blocksize_bits + 2) * sizeof(*sbi->s_mb_maxs);
        sbi->s_mb_maxs = kmalloc(i, GFP_KERNEL);
        if (sbi->s_mb_maxs == NULL) {
                ret = -ENOMEM;
                goto out;
        }

        ret = ext4_groupinfo_create_slab(sb->s_blocksize);
        if (ret < 0)
                goto out;

        /* order 0 is regular bitmap */
        sbi->s_mb_maxs[0] = sb->s_blocksize << 3;
        sbi->s_mb_offsets[0] = 0;

        i = 1;
        offset = 0;
        max = sb->s_blocksize << 2;
        do {
                sbi->s_mb_offsets[i] = offset;
                sbi->s_mb_maxs[i] = max;
                offset += 1 << (sb->s_blocksize_bits - i);
                max = max >> 1;
                i++;
        } while (i <= sb->s_blocksize_bits + 1);

        spin_lock_init(&sbi->s_md_lock);
        spin_lock_init(&sbi->s_bal_lock);

        sbi->s_mb_max_to_scan = MB_DEFAULT_MAX_TO_SCAN;
        sbi->s_mb_min_to_scan = MB_DEFAULT_MIN_TO_SCAN;
        sbi->s_mb_stats = MB_DEFAULT_STATS;
        sbi->s_mb_stream_request = MB_DEFAULT_STREAM_THRESHOLD;
        sbi->s_mb_order2_reqs = MB_DEFAULT_ORDER2_REQS;
        /*
         * The default group preallocation is 512, which for 4k block
         * sizes translates to 2 megabytes.  However for bigalloc file
         * systems, this is probably too big (i.e, if the cluster size
         * is 1 megabyte, then group preallocation size becomes half a
         * gigabyte!).  As a default, we will keep a two megabyte
         * group pralloc size for cluster sizes up to 64k, and after
         * that, we will force a minimum group preallocation size of
         * 32 clusters.  This translates to 8 megs when the cluster
         * size is 256k, and 32 megs when the cluster size is 1 meg,
         * which seems reasonable as a default.
         */
        sbi->s_mb_group_prealloc = max(MB_DEFAULT_GROUP_PREALLOC >>
                                       sbi->s_cluster_bits, 32);
        /*
         * If there is a s_stripe > 1, then we set the s_mb_group_prealloc
         * to the lowest multiple of s_stripe which is bigger than
         * the s_mb_group_prealloc as determined above. We want
         * the preallocation size to be an exact multiple of the
         * RAID stripe size so that preallocations don't fragment
         * the stripes.
         */
        if (sbi->s_stripe > 1) {
                sbi->s_mb_group_prealloc = roundup(
                        sbi->s_mb_group_prealloc, sbi->s_stripe);
        }

        sbi->s_locality_groups = alloc_percpu(struct ext4_locality_group);
        if (sbi->s_locality_groups == NULL) {
                ret = -ENOMEM;
                goto out_free_groupinfo_slab;
        }
        for_each_possible_cpu(i) {
                struct ext4_locality_group *lg;
                lg = per_cpu_ptr(sbi->s_locality_groups, i);
                mutex_init(&lg->lg_mutex);
                for (j = 0; j < PREALLOC_TB_SIZE; j++)
                        INIT_LIST_HEAD(&lg->lg_prealloc_list[j]);
                spin_lock_init(&lg->lg_prealloc_lock);
        }

        /* init file for buddy data */
        ret = ext4_mb_init_backend(sb);
        if (ret != 0)
                goto out_free_locality_groups;

        if (sbi->s_proc)
                proc_create_data("mb_groups", S_IRUGO, sbi->s_proc,
                                 &ext4_mb_seq_groups_fops, sb);

        return 0;

out_free_locality_groups:
        free_percpu(sbi->s_locality_groups);
        sbi->s_locality_groups = NULL;
out_free_groupinfo_slab:
        ext4_groupinfo_destroy_slabs();
out:
        kfree(sbi->s_mb_offsets);
        sbi->s_mb_offsets = NULL;
        kfree(sbi->s_mb_maxs);
        sbi->s_mb_maxs = NULL;
        return ret;
}

/* need to called with the ext4 group lock held */
static void ext4_mb_cleanup_pa(struct ext4_group_info *grp)
{
        struct ext4_prealloc_space *pa;
        struct list_head *cur, *tmp;
        int count = 0;

        list_for_each_safe(cur, tmp, &grp->bb_prealloc_list) {
                pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list);
                list_del(&pa->pa_group_list);
                count++;
                kmem_cache_free(ext4_pspace_cachep, pa);
        }
        if (count)
                mb_debug(1, "mballoc: %u PAs left\n", count);

}

int ext4_mb_release(struct super_block *sb)
{
        ext4_group_t ngroups = ext4_get_groups_count(sb);
        ext4_group_t i;
        int num_meta_group_infos;
        struct ext4_group_info *grinfo;
        struct ext4_sb_info *sbi = EXT4_SB(sb);
        struct kmem_cache *cachep = get_groupinfo_cache(sb->s_blocksize_bits);

        if (sbi->s_proc)
                remove_proc_entry("mb_groups", sbi->s_proc);

        if (sbi->s_group_info) {
                for (i = 0; i < ngroups; i++) {
                        grinfo = ext4_get_group_info(sb, i);
#ifdef DOUBLE_CHECK
                        kfree(grinfo->bb_bitmap);
#endif
                        ext4_lock_group(sb, i);
                        ext4_mb_cleanup_pa(grinfo);
                        ext4_unlock_group(sb, i);
                        kmem_cache_free(cachep, grinfo);
                }
                num_meta_group_infos = (ngroups +
                                EXT4_DESC_PER_BLOCK(sb) - 1) >>
                        EXT4_DESC_PER_BLOCK_BITS(sb);
                for (i = 0; i < num_meta_group_infos; i++)
                        kfree(sbi->s_group_info[i]);
                ext4_kvfree(sbi->s_group_info);
        }
        kfree(sbi->s_mb_offsets);
        kfree(sbi->s_mb_maxs);
        if (sbi->s_buddy_cache)
                iput(sbi->s_buddy_cache);
        if (sbi->s_mb_stats) {
                ext4_msg(sb, KERN_INFO,
                       "mballoc: %u blocks %u reqs (%u success)",
                                atomic_read(&sbi->s_bal_allocated),
                                atomic_read(&sbi->s_bal_reqs),
                                atomic_read(&sbi->s_bal_success));
                ext4_msg(sb, KERN_INFO,
                      "mballoc: %u extents scanned, %u goal hits, "
                                "%u 2^N hits, %u breaks, %u lost",
                                atomic_read(&sbi->s_bal_ex_scanned),
                                atomic_read(&sbi->s_bal_goals),
                                atomic_read(&sbi->s_bal_2orders),
                                atomic_read(&sbi->s_bal_breaks),
                                atomic_read(&sbi->s_mb_lost_chunks));
                ext4_msg(sb, KERN_INFO,
                       "mballoc: %lu generated and it took %Lu",
                                sbi->s_mb_buddies_generated,
                                sbi->s_mb_generation_time);
                ext4_msg(sb, KERN_INFO,
                       "mballoc: %u preallocated, %u discarded",
                                atomic_read(&sbi->s_mb_preallocated),
                                atomic_read(&sbi->s_mb_discarded));
        }

        free_percpu(sbi->s_locality_groups);

        return 0;
}

static inline int ext4_issue_discard(struct super_block *sb,
                ext4_group_t block_group, ext4_grpblk_t cluster, int count)
{
        ext4_fsblk_t discard_block;

        discard_block = (EXT4_C2B(EXT4_SB(sb), cluster) +
                         ext4_group_first_block_no(sb, block_group));
        count = EXT4_C2B(EXT4_SB(sb), count);
        trace_ext4_discard_blocks(sb,
                        (unsigned long long) discard_block, count);
        return sb_issue_discard(sb, discard_block, count, GFP_NOFS, 0);
}

/*
 * This function is called by the jbd2 layer once the commit has finished,
 * so we know we can free the blocks that were released with that commit.
 */
static void ext4_free_data_callback(struct super_block *sb,
                                    struct ext4_journal_cb_entry *jce,
                                    int rc)
{
        struct ext4_free_data *entry = (struct ext4_free_data *)jce;
        struct ext4_buddy e4b;
        struct ext4_group_info *db;
        int err, count = 0, count2 = 0;

        mb_debug(1, "gonna free %u blocks in group %u (0x%p):",
                 entry->efd_count, entry->efd_group, entry);

        if (test_opt(sb, DISCARD))
                ext4_issue_discard(sb, entry->efd_group,
                                   entry->efd_start_cluster, entry->efd_count);

        err = ext4_mb_load_buddy(sb, entry->efd_group, &e4b);
        /* we expect to find existing buddy because it's pinned */
        BUG_ON(err != 0);


        db = e4b.bd_info;
        /* there are blocks to put in buddy to make them really free */
        count += entry->efd_count;
        count2++;
        ext4_lock_group(sb, entry->efd_group);
        /* Take it out of per group rb tree */
        rb_erase(&entry->efd_node, &(db->bb_free_root));
        mb_free_blocks(NULL, &e4b, entry->efd_start_cluster, entry->efd_count);

        /*
         * Clear the trimmed flag for the group so that the next
         * ext4_trim_fs can trim it.
         * If the volume is mounted with -o discard, online discard
         * is supported and the free blocks will be trimmed online.
         */
        if (!test_opt(sb, DISCARD))
                EXT4_MB_GRP_CLEAR_TRIMMED(db);

        if (!db->bb_free_root.rb_node) {
                /* No more items in the per group rb tree
                 * balance refcounts from ext4_mb_free_metadata()
                 */
                page_cache_release(e4b.bd_buddy_page);
                page_cache_release(e4b.bd_bitmap_page);
        }
        ext4_unlock_group(sb, entry->efd_group);
        kmem_cache_free(ext4_free_data_cachep, entry);
        ext4_mb_unload_buddy(&e4b);

        mb_debug(1, "freed %u blocks in %u structures\n", count, count2);
}

#ifdef CONFIG_EXT4_DEBUG
u8 mb_enable_debug __read_mostly;

static struct dentry *debugfs_dir;
static struct dentry *debugfs_debug;

static void __init ext4_create_debugfs_entry(void)
{
        debugfs_dir = debugfs_create_dir("ext4", NULL);
        if (debugfs_dir)
                debugfs_debug = debugfs_create_u8("mballoc-debug",
                                                  S_IRUGO | S_IWUSR,
                                                  debugfs_dir,
                                                  &mb_enable_debug);
}

static void ext4_remove_debugfs_entry(void)
{
        debugfs_remove(debugfs_debug);
        debugfs_remove(debugfs_dir);
}

#else

static void __init ext4_create_debugfs_entry(void)
{
}

static void ext4_remove_debugfs_entry(void)
{
}

#endif

int __init ext4_init_mballoc(void)
{
        ext4_pspace_cachep = KMEM_CACHE(ext4_prealloc_space,
                                        SLAB_RECLAIM_ACCOUNT);
        if (ext4_pspace_cachep == NULL)
                return -ENOMEM;

        ext4_ac_cachep = KMEM_CACHE(ext4_allocation_context,
                                    SLAB_RECLAIM_ACCOUNT);
        if (ext4_ac_cachep == NULL) {
                kmem_cache_destroy(ext4_pspace_cachep);
                return -ENOMEM;
        }

        ext4_free_data_cachep = KMEM_CACHE(ext4_free_data,
                                           SLAB_RECLAIM_ACCOUNT);
        if (ext4_free_data_cachep == NULL) {
                kmem_cache_destroy(ext4_pspace_cachep);
                kmem_cache_destroy(ext4_ac_cachep);
                return -ENOMEM;
        }
        ext4_create_debugfs_entry();
        return 0;
}

void ext4_exit_mballoc(void)
{
        /*
         * Wait for completion of call_rcu()'s on ext4_pspace_cachep
         * before destroying the slab cache.
         */
        rcu_barrier();
        kmem_cache_destroy(ext4_pspace_cachep);
        kmem_cache_destroy(ext4_ac_cachep);
        kmem_cache_destroy(ext4_free_data_cachep);
        ext4_groupinfo_destroy_slabs();
        ext4_remove_debugfs_entry();
}


/*
 * Check quota and mark chosen space (ac->ac_b_ex) non-free in bitmaps
 * Returns 0 if success or error code
 */
static noinline_for_stack int
ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
                                handle_t *handle, unsigned int reserv_clstrs)
{
        struct buffer_head *bitmap_bh = NULL;
        struct ext4_group_desc *gdp;
        struct buffer_head *gdp_bh;
        struct ext4_sb_info *sbi;
        struct super_block *sb;
        ext4_fsblk_t block;
        int err, len;

        BUG_ON(ac->ac_status != AC_STATUS_FOUND);
        BUG_ON(ac->ac_b_ex.fe_len <= 0);

        sb = ac->ac_sb;
        sbi = EXT4_SB(sb);

        err = -EIO;
        bitmap_bh = ext4_read_block_bitmap(sb, ac->ac_b_ex.fe_group);
        if (!bitmap_bh)
                goto out_err;

        err = ext4_journal_get_write_access(handle, bitmap_bh);
        if (err)
                goto out_err;

        err = -EIO;
        gdp = ext4_get_group_desc(sb, ac->ac_b_ex.fe_group, &gdp_bh);
        if (!gdp)
                goto out_err;

        ext4_debug("using block group %u(%d)\n", ac->ac_b_ex.fe_group,
                        ext4_free_group_clusters(sb, gdp));

        err = ext4_journal_get_write_access(handle, gdp_bh);
        if (err)
                goto out_err;

        block = ext4_grp_offs_to_block(sb, &ac->ac_b_ex);

        len = EXT4_C2B(sbi, ac->ac_b_ex.fe_len);
        if (!ext4_data_block_valid(sbi, block, len)) {
                ext4_error(sb, "Allocating blocks %llu-%llu which overlap "
                           "fs metadata", block, block+len);
                /* File system mounted not to panic on error
                 * Fix the bitmap and repeat the block allocation
                 * We leak some of the blocks here.
                 */
                ext4_lock_group(sb, ac->ac_b_ex.fe_group);
                ext4_set_bits(bitmap_bh->b_data, ac->ac_b_ex.fe_start,
                              ac->ac_b_ex.fe_len);
                ext4_unlock_group(sb, ac->ac_b_ex.fe_group);
                err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
                if (!err)
                        err = -EAGAIN;
                goto out_err;
        }

        ext4_lock_group(sb, ac->ac_b_ex.fe_group);
#ifdef AGGRESSIVE_CHECK
        {
                int i;
                for (i = 0; i < ac->ac_b_ex.fe_len; i++) {
                        BUG_ON(mb_test_bit(ac->ac_b_ex.fe_start + i,
                                                bitmap_bh->b_data));
                }
        }
#endif
        ext4_set_bits(bitmap_bh->b_data, ac->ac_b_ex.fe_start,
                      ac->ac_b_ex.fe_len);
        if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
                gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT);
                ext4_free_group_clusters_set(sb, gdp,
                                             ext4_free_clusters_after_init(sb,
                                                ac->ac_b_ex.fe_group, gdp));
        }
        len = ext4_free_group_clusters(sb, gdp) - ac->ac_b_ex.fe_len;
        ext4_free_group_clusters_set(sb, gdp, len);
        ext4_block_bitmap_csum_set(sb, ac->ac_b_ex.fe_group, gdp, bitmap_bh);
        ext4_group_desc_csum_set(sb, ac->ac_b_ex.fe_group, gdp);

        ext4_unlock_group(sb, ac->ac_b_ex.fe_group);
        percpu_counter_sub(&sbi->s_freeclusters_counter, ac->ac_b_ex.fe_len);
        /*
         * Now reduce the dirty block count also. Should not go negative
         */
        if (!(ac->ac_flags & EXT4_MB_DELALLOC_RESERVED))
                /* release all the reserved blocks if non delalloc */
                percpu_counter_sub(&sbi->s_dirtyclusters_counter,
                                   reserv_clstrs);

        if (sbi->s_log_groups_per_flex) {
                ext4_group_t flex_group = ext4_flex_group(sbi,
                                                          ac->ac_b_ex.fe_group);
                atomic_sub(ac->ac_b_ex.fe_len,
                           &sbi->s_flex_groups[flex_group].free_clusters);
        }

        err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
        if (err)
                goto out_err;
        err = ext4_handle_dirty_metadata(handle, NULL, gdp_bh);

out_err:
        brelse(bitmap_bh);
        return err;
}

/*
 * here we normalize request for locality group
 * Group request are normalized to s_mb_group_prealloc, which goes to
 * s_strip if we set the same via mount option.
 * s_mb_group_prealloc can be configured via
 * /sys/fs/ext4/<partition>/mb_group_prealloc
 *
 * XXX: should we try to preallocate more than the group has now?
 */
static void ext4_mb_normalize_group_request(struct ext4_allocation_context *ac)
{
        struct super_block *sb = ac->ac_sb;
        struct ext4_locality_group *lg = ac->ac_lg;

        BUG_ON(lg == NULL);
        ac->ac_g_ex.fe_len = EXT4_SB(sb)->s_mb_group_prealloc;
        mb_debug(1, "#%u: goal %u blocks for locality group\n",
                current->pid, ac->ac_g_ex.fe_len);
}

/*
 * Normalization means making request better in terms of
 * size and alignment
 */
static noinline_for_stack void
ext4_mb_normalize_request(struct ext4_allocation_context *ac,
                                struct ext4_allocation_request *ar)
{
        struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb);
        int bsbits, max;
        ext4_lblk_t end;
        loff_t size, start_off;
        loff_t orig_size __maybe_unused;
        ext4_lblk_t start;
        struct ext4_inode_info *ei = EXT4_I(ac->ac_inode);
        struct ext4_prealloc_space *pa;

        /* do normalize only data requests, metadata requests
           do not need preallocation */
        if (!(ac->ac_flags & EXT4_MB_HINT_DATA))
                return;

        /* sometime caller may want exact blocks */
        if (unlikely(ac->ac_flags & EXT4_MB_HINT_GOAL_ONLY))
                return;

        /* caller may indicate that preallocation isn't
         * required (it's a tail, for example) */
        if (ac->ac_flags & EXT4_MB_HINT_NOPREALLOC)
                return;

        if (ac->ac_flags & EXT4_MB_HINT_GROUP_ALLOC) {
                ext4_mb_normalize_group_request(ac);
                return ;
        }

        bsbits = ac->ac_sb->s_blocksize_bits;

        /* first, let's learn actual file size
         * given current request is allocated */
        size = ac->ac_o_ex.fe_logical + EXT4_C2B(sbi, ac->ac_o_ex.fe_len);
        size = size << bsbits;
        if (size < i_size_read(ac->ac_inode))
                size = i_size_read(ac->ac_inode);
        orig_size = size;

        /* max size of free chunks */
        max = 2 << bsbits;

#define NRL_CHECK_SIZE(req, size, max, chunk_size)      \
                (req <= (size) || max <= (chunk_size))

        /* first, try to predict filesize */
        /* XXX: should this table be tunable? */
        start_off = 0;
        if (size <= 16 * 1024) {
                size = 16 * 1024;
        } else if (size <= 32 * 1024) {
                size = 32 * 1024;
        } else if (size <= 64 * 1024) {
                size = 64 * 1024;
        } else if (size <= 128 * 1024) {
                size = 128 * 1024;
        } else if (size <= 256 * 1024) {
                size = 256 * 1024;
        } else if (size <= 512 * 1024) {
                size = 512 * 1024;
        } else if (size <= 1024 * 1024) {
                size = 1024 * 1024;
        } else if (NRL_CHECK_SIZE(size, 4 * 1024 * 1024, max, 2 * 1024)) {
                start_off = ((loff_t)ac->ac_o_ex.fe_logical >>
                                                (21 - bsbits)) << 21;
                size = 2 * 1024 * 1024;
        } else if (NRL_CHECK_SIZE(size, 8 * 1024 * 1024, max, 4 * 1024)) {
                start_off = ((loff_t)ac->ac_o_ex.fe_logical >>
                                                        (22 - bsbits)) << 22;
                size = 4 * 1024 * 1024;
        } else if (NRL_CHECK_SIZE(ac->ac_o_ex.fe_len,
                                        (8<<20)>>bsbits, max, 8 * 1024)) {
                start_off = ((loff_t)ac->ac_o_ex.fe_logical >>
                                                        (23 - bsbits)) << 23;
                size = 8 * 1024 * 1024;
        } else {
                start_off = (loff_t)ac->ac_o_ex.fe_logical << bsbits;
                size      = ac->ac_o_ex.fe_len << bsbits;
        }
        size = size >> bsbits;
        start = start_off >> bsbits;

        /* don't cover already allocated blocks in selected range */
        if (ar->pleft && start <= ar->lleft) {
                size -= ar->lleft + 1 - start;
                start = ar->lleft + 1;
        }
        if (ar->pright && start + size - 1 >= ar->lright)
                size -= start + size - ar->lright;

        end = start + size;

        /* check we don't cross already preallocated blocks */
        rcu_read_lock();
        list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) {
                ext4_lblk_t pa_end;

                if (pa->pa_deleted)
                        continue;
                spin_lock(&pa->pa_lock);
                if (pa->pa_deleted) {
                        spin_unlock(&pa->pa_lock);
                        continue;
                }

                pa_end = pa->pa_lstart + EXT4_C2B(EXT4_SB(ac->ac_sb),
                                                  pa->pa_len);

                /* PA must not overlap original request */
                BUG_ON(!(ac->ac_o_ex.fe_logical >= pa_end ||
                        ac->ac_o_ex.fe_logical < pa->pa_lstart));

                /* skip PAs this normalized request doesn't overlap with */
                if (pa->pa_lstart >= end || pa_end <= start) {
                        spin_unlock(&pa->pa_lock);
                        continue;
                }
                BUG_ON(pa->pa_lstart <= start && pa_end >= end);

                /* adjust start or end to be adjacent to this pa */
                if (pa_end <= ac->ac_o_ex.fe_logical) {
                        BUG_ON(pa_end < start);
                        start = pa_end;
                } else if (pa->pa_lstart > ac->ac_o_ex.fe_logical) {
                        BUG_ON(pa->pa_lstart > end);
                        end = pa->pa_lstart;
                }
                spin_unlock(&pa->pa_lock);
        }
        rcu_read_unlock();
        size = end - start;

        /* XXX: extra loop to check we really don't overlap preallocations */
        rcu_read_lock();
        list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) {
                ext4_lblk_t pa_end;