From: johann Date: Wed, 18 Jul 2007 20:56:05 +0000 (+0000) Subject: Branch b1_6 X-Git-Tag: v1_7_100~7 X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=commitdiff_plain;h=038b92d58c79afbf348dfdef908fff049793eb91 Branch b1_6 b=11039 i=adilger i=scjody Remove 2.4 and 2.6.12 kernel patches. --- diff --git a/ldiskfs/kernel_patches/patches/ext3-extents-2.6.12.patch b/ldiskfs/kernel_patches/patches/ext3-extents-2.6.12.patch deleted file mode 100644 index f421f88..0000000 --- a/ldiskfs/kernel_patches/patches/ext3-extents-2.6.12.patch +++ /dev/null @@ -1,2940 +0,0 @@ -Index: linux-2.6.12-rc6/fs/ext3/extents.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/extents.c 2005-06-14 16:31:25.756503133 +0200 -+++ linux-2.6.12-rc6/fs/ext3/extents.c 2005-06-14 16:31:25.836581257 +0200 -@@ -0,0 +1,2359 @@ -+/* -+ * Copyright(c) 2003, 2004, 2005, Cluster File Systems, Inc, info@clusterfs.com -+ * Written by Alex Tomas -+ * -+ * 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- -+ */ -+ -+/* -+ * Extents support for EXT3 -+ * -+ * TODO: -+ * - ext3_ext_walk_space() sould not use ext3_ext_find_extent() -+ * - ext3_ext_calc_credits() could take 'mergable' into account -+ * - ext3*_error() should be used in some situations -+ * - find_goal() [to be tested and improved] -+ * - smart tree reduction -+ * - arch-independence -+ * common on-disk format for big/little-endian arch -+ */ -+ -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+ -+ -+static inline int ext3_ext_check_header(struct ext3_extent_header *eh) -+{ -+ if (eh->eh_magic != EXT3_EXT_MAGIC) { -+ printk(KERN_ERR "EXT3-fs: invalid magic = 0x%x\n", -+ (unsigned)eh->eh_magic); -+ return -EIO; -+ } -+ if (eh->eh_max == 0) { -+ printk(KERN_ERR "EXT3-fs: invalid eh_max = %u\n", -+ (unsigned)eh->eh_max); -+ return -EIO; -+ } -+ if (eh->eh_entries > eh->eh_max) { -+ printk(KERN_ERR "EXT3-fs: invalid eh_entries = %u\n", -+ (unsigned)eh->eh_entries); -+ return -EIO; -+ } -+ return 0; -+} -+ -+static handle_t *ext3_ext_journal_restart(handle_t *handle, int needed) -+{ -+ int err; -+ -+ if (handle->h_buffer_credits > needed) -+ return handle; -+ if (!ext3_journal_extend(handle, needed)) -+ return handle; -+ err = ext3_journal_restart(handle, needed); -+ -+ return handle; -+} -+ -+static int inline -+ext3_ext_get_access_for_root(handle_t *h, struct ext3_extents_tree *tree) -+{ -+ if (tree->ops->get_write_access) -+ return tree->ops->get_write_access(h,tree->buffer); -+ else -+ return 0; -+} -+ -+static int inline -+ext3_ext_mark_root_dirty(handle_t *h, struct ext3_extents_tree *tree) -+{ -+ if (tree->ops->mark_buffer_dirty) -+ return tree->ops->mark_buffer_dirty(h,tree->buffer); -+ else -+ return 0; -+} -+ -+/* -+ * could return: -+ * - EROFS -+ * - ENOMEM -+ */ -+static int ext3_ext_get_access(handle_t *handle, -+ struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path) -+{ -+ int err; -+ -+ if (path->p_bh) { -+ /* path points to block */ -+ err = ext3_journal_get_write_access(handle, path->p_bh); -+ } else { -+ /* path points to leaf/index in inode body */ -+ err = ext3_ext_get_access_for_root(handle, tree); -+ } -+ return err; -+} -+ -+/* -+ * could return: -+ * - EROFS -+ * - ENOMEM -+ * - EIO -+ */ -+static int ext3_ext_dirty(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path) -+{ -+ int err; -+ if (path->p_bh) { -+ /* path points to block */ -+ err =ext3_journal_dirty_metadata(handle, path->p_bh); -+ } else { -+ /* path points to leaf/index in inode body */ -+ err = ext3_ext_mark_root_dirty(handle, tree); -+ } -+ return err; -+} -+ -+static int inline -+ext3_ext_new_block(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, struct ext3_extent *ex, -+ int *err) -+{ -+ int goal, depth, newblock; -+ struct inode *inode; -+ -+ EXT_ASSERT(tree); -+ if (tree->ops->new_block) -+ return tree->ops->new_block(handle, tree, path, ex, err); -+ -+ inode = tree->inode; -+ depth = EXT_DEPTH(tree); -+ if (path && depth > 0) { -+ goal = path[depth-1].p_block; -+ } else { -+ struct ext3_inode_info *ei = EXT3_I(inode); -+ unsigned long bg_start; -+ unsigned long colour; -+ -+ bg_start = (ei->i_block_group * -+ EXT3_BLOCKS_PER_GROUP(inode->i_sb)) + -+ le32_to_cpu(EXT3_SB(inode->i_sb)->s_es->s_first_data_block); -+ colour = (current->pid % 16) * -+ (EXT3_BLOCKS_PER_GROUP(inode->i_sb) / 16); -+ goal = bg_start + colour; -+ } -+ -+ newblock = ext3_new_block(handle, inode, goal, err); -+ return newblock; -+} -+ -+static inline void ext3_ext_tree_changed(struct ext3_extents_tree *tree) -+{ -+ struct ext3_extent_header *neh = EXT_ROOT_HDR(tree); -+ neh->eh_generation = ((EXT_FLAGS(neh) & ~EXT_FLAGS_CLR_UNKNOWN) << 24) | -+ (EXT_HDR_GEN(neh) + 1); -+} -+ -+static inline int ext3_ext_space_block(struct ext3_extents_tree *tree) -+{ -+ int size; -+ -+ size = (tree->inode->i_sb->s_blocksize - -+ sizeof(struct ext3_extent_header)) / -+ sizeof(struct ext3_extent); -+#ifdef AGRESSIVE_TEST -+ size = 6; -+#endif -+ return size; -+} -+ -+static inline int ext3_ext_space_block_idx(struct ext3_extents_tree *tree) -+{ -+ int size; -+ -+ size = (tree->inode->i_sb->s_blocksize - -+ sizeof(struct ext3_extent_header)) / -+ sizeof(struct ext3_extent_idx); -+#ifdef AGRESSIVE_TEST -+ size = 5; -+#endif -+ return size; -+} -+ -+static inline int ext3_ext_space_root(struct ext3_extents_tree *tree) -+{ -+ int size; -+ -+ size = (tree->buffer_len - sizeof(struct ext3_extent_header)) / -+ sizeof(struct ext3_extent); -+#ifdef AGRESSIVE_TEST -+ size = 3; -+#endif -+ return size; -+} -+ -+static inline int ext3_ext_space_root_idx(struct ext3_extents_tree *tree) -+{ -+ int size; -+ -+ size = (tree->buffer_len - sizeof(struct ext3_extent_header)) / -+ sizeof(struct ext3_extent_idx); -+#ifdef AGRESSIVE_TEST -+ size = 4; -+#endif -+ return size; -+} -+ -+static void ext3_ext_show_path(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path) -+{ -+#ifdef EXT_DEBUG -+ int k, l = path->p_depth; -+ -+ ext_debug(tree, "path:"); -+ for (k = 0; k <= l; k++, path++) { -+ if (path->p_idx) { -+ ext_debug(tree, " %d->%d", path->p_idx->ei_block, -+ path->p_idx->ei_leaf); -+ } else if (path->p_ext) { -+ ext_debug(tree, " %d:%d:%d", -+ path->p_ext->ee_block, -+ path->p_ext->ee_len, -+ path->p_ext->ee_start); -+ } else -+ ext_debug(tree, " []"); -+ } -+ ext_debug(tree, "\n"); -+#endif -+} -+ -+static void ext3_ext_show_leaf(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path) -+{ -+#ifdef EXT_DEBUG -+ int depth = EXT_DEPTH(tree); -+ struct ext3_extent_header *eh; -+ struct ext3_extent *ex; -+ int i; -+ -+ if (!path) -+ return; -+ -+ eh = path[depth].p_hdr; -+ ex = EXT_FIRST_EXTENT(eh); -+ -+ for (i = 0; i < eh->eh_entries; i++, ex++) { -+ ext_debug(tree, "%d:%d:%d ", -+ ex->ee_block, ex->ee_len, ex->ee_start); -+ } -+ ext_debug(tree, "\n"); -+#endif -+} -+ -+static void ext3_ext_drop_refs(struct ext3_ext_path *path) -+{ -+ int depth = path->p_depth; -+ int i; -+ -+ for (i = 0; i <= depth; i++, path++) { -+ if (path->p_bh) { -+ brelse(path->p_bh); -+ path->p_bh = NULL; -+ } -+ } -+} -+ -+/* -+ * binary search for closest index by given block -+ */ -+static inline void -+ext3_ext_binsearch_idx(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, int block) -+{ -+ struct ext3_extent_header *eh = path->p_hdr; -+ struct ext3_extent_idx *ix; -+ int l = 0, k, r; -+ -+ EXT_ASSERT(eh->eh_magic == EXT3_EXT_MAGIC); -+ EXT_ASSERT(eh->eh_entries <= eh->eh_max); -+ EXT_ASSERT(eh->eh_entries > 0); -+ -+ ext_debug(tree, "binsearch for %d(idx): ", block); -+ -+ path->p_idx = ix = EXT_FIRST_INDEX(eh); -+ -+ r = k = eh->eh_entries; -+ while (k > 1) { -+ k = (r - l) / 2; -+ if (block < ix[l + k].ei_block) -+ r -= k; -+ else -+ l += k; -+ ext_debug(tree, "%d:%d:%d ", k, l, r); -+ } -+ -+ ix += l; -+ path->p_idx = ix; -+ ext_debug(tree," -> %d->%d ",path->p_idx->ei_block,path->p_idx->ei_leaf); -+ -+ while (l++ < r) { -+ if (block < ix->ei_block) -+ break; -+ path->p_idx = ix++; -+ } -+ ext_debug(tree, " -> %d->%d\n", path->p_idx->ei_block, -+ path->p_idx->ei_leaf); -+ -+#ifdef CHECK_BINSEARCH -+ { -+ struct ext3_extent_idx *chix; -+ -+ chix = ix = EXT_FIRST_INDEX(eh); -+ for (k = 0; k < eh->eh_entries; k++, ix++) { -+ if (k != 0 && ix->ei_block <= ix[-1].ei_block) { -+ printk("k=%d, ix=0x%p, first=0x%p\n", k, -+ ix, EXT_FIRST_INDEX(eh)); -+ printk("%u <= %u\n", -+ ix->ei_block,ix[-1].ei_block); -+ } -+ EXT_ASSERT(k == 0 || ix->ei_block > ix[-1].ei_block); -+ if (block < ix->ei_block) -+ break; -+ chix = ix; -+ } -+ EXT_ASSERT(chix == path->p_idx); -+ } -+#endif -+} -+ -+/* -+ * binary search for closest extent by given block -+ */ -+static inline void -+ext3_ext_binsearch(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, int block) -+{ -+ struct ext3_extent_header *eh = path->p_hdr; -+ struct ext3_extent *ex; -+ int l = 0, k, r; -+ -+ EXT_ASSERT(eh->eh_magic == EXT3_EXT_MAGIC); -+ EXT_ASSERT(eh->eh_entries <= eh->eh_max); -+ -+ if (eh->eh_entries == 0) { -+ /* -+ * this leaf is empty yet: -+ * we get such a leaf in split/add case -+ */ -+ return; -+ } -+ -+ ext_debug(tree, "binsearch for %d: ", block); -+ -+ path->p_ext = ex = EXT_FIRST_EXTENT(eh); -+ -+ r = k = eh->eh_entries; -+ while (k > 1) { -+ k = (r - l) / 2; -+ if (block < ex[l + k].ee_block) -+ r -= k; -+ else -+ l += k; -+ ext_debug(tree, "%d:%d:%d ", k, l, r); -+ } -+ -+ ex += l; -+ path->p_ext = ex; -+ ext_debug(tree, " -> %d:%d:%d ", path->p_ext->ee_block, -+ path->p_ext->ee_start, path->p_ext->ee_len); -+ -+ while (l++ < r) { -+ if (block < ex->ee_block) -+ break; -+ path->p_ext = ex++; -+ } -+ ext_debug(tree, " -> %d:%d:%d\n", path->p_ext->ee_block, -+ path->p_ext->ee_start, path->p_ext->ee_len); -+ -+#ifdef CHECK_BINSEARCH -+ { -+ struct ext3_extent *chex; -+ -+ chex = ex = EXT_FIRST_EXTENT(eh); -+ for (k = 0; k < eh->eh_entries; k++, ex++) { -+ EXT_ASSERT(k == 0 || ex->ee_block > ex[-1].ee_block); -+ if (block < ex->ee_block) -+ break; -+ chex = ex; -+ } -+ EXT_ASSERT(chex == path->p_ext); -+ } -+#endif -+} -+ -+int ext3_extent_tree_init(handle_t *handle, struct ext3_extents_tree *tree) -+{ -+ struct ext3_extent_header *eh; -+ -+ BUG_ON(tree->buffer_len == 0); -+ ext3_ext_get_access_for_root(handle, tree); -+ eh = EXT_ROOT_HDR(tree); -+ eh->eh_depth = 0; -+ eh->eh_entries = 0; -+ eh->eh_magic = EXT3_EXT_MAGIC; -+ eh->eh_max = ext3_ext_space_root(tree); -+ ext3_ext_mark_root_dirty(handle, tree); -+ ext3_ext_invalidate_cache(tree); -+ return 0; -+} -+ -+struct ext3_ext_path * -+ext3_ext_find_extent(struct ext3_extents_tree *tree, int block, -+ struct ext3_ext_path *path) -+{ -+ struct ext3_extent_header *eh; -+ struct buffer_head *bh; -+ int depth, i, ppos = 0; -+ -+ EXT_ASSERT(tree); -+ EXT_ASSERT(tree->inode); -+ EXT_ASSERT(tree->root); -+ -+ eh = EXT_ROOT_HDR(tree); -+ EXT_ASSERT(eh); -+ if (ext3_ext_check_header(eh)) { -+ /* don't free previously allocated path -+ * -- caller should take care */ -+ path = NULL; -+ goto err; -+ } -+ -+ i = depth = EXT_DEPTH(tree); -+ EXT_ASSERT(eh->eh_max); -+ EXT_ASSERT(eh->eh_magic == EXT3_EXT_MAGIC); -+ -+ /* account possible depth increase */ -+ if (!path) { -+ path = kmalloc(sizeof(struct ext3_ext_path) * (depth + 2), -+ GFP_NOFS); -+ if (!path) -+ return ERR_PTR(-ENOMEM); -+ } -+ memset(path, 0, sizeof(struct ext3_ext_path) * (depth + 1)); -+ path[0].p_hdr = eh; -+ -+ /* walk through the tree */ -+ while (i) { -+ ext_debug(tree, "depth %d: num %d, max %d\n", -+ ppos, eh->eh_entries, eh->eh_max); -+ ext3_ext_binsearch_idx(tree, path + ppos, block); -+ path[ppos].p_block = path[ppos].p_idx->ei_leaf; -+ path[ppos].p_depth = i; -+ path[ppos].p_ext = NULL; -+ -+ bh = sb_bread(tree->inode->i_sb, path[ppos].p_block); -+ if (!bh) -+ goto err; -+ -+ eh = EXT_BLOCK_HDR(bh); -+ ppos++; -+ EXT_ASSERT(ppos <= depth); -+ path[ppos].p_bh = bh; -+ path[ppos].p_hdr = eh; -+ i--; -+ -+ if (ext3_ext_check_header(eh)) -+ goto err; -+ } -+ -+ path[ppos].p_depth = i; -+ path[ppos].p_hdr = eh; -+ path[ppos].p_ext = NULL; -+ path[ppos].p_idx = NULL; -+ -+ if (ext3_ext_check_header(eh)) -+ goto err; -+ -+ /* find extent */ -+ ext3_ext_binsearch(tree, path + ppos, block); -+ -+ ext3_ext_show_path(tree, path); -+ -+ return path; -+ -+err: -+ printk(KERN_ERR "EXT3-fs: header is corrupted!\n"); -+ if (path) { -+ ext3_ext_drop_refs(path); -+ kfree(path); -+ } -+ return ERR_PTR(-EIO); -+} -+ -+/* -+ * insert new index [logical;ptr] into the block at cupr -+ * it check where to insert: before curp or after curp -+ */ -+static int ext3_ext_insert_index(handle_t *handle, -+ struct ext3_extents_tree *tree, -+ struct ext3_ext_path *curp, -+ int logical, int ptr) -+{ -+ struct ext3_extent_idx *ix; -+ int len, err; -+ -+ if ((err = ext3_ext_get_access(handle, tree, curp))) -+ return err; -+ -+ EXT_ASSERT(logical != curp->p_idx->ei_block); -+ len = EXT_MAX_INDEX(curp->p_hdr) - curp->p_idx; -+ if (logical > curp->p_idx->ei_block) { -+ /* insert after */ -+ if (curp->p_idx != EXT_LAST_INDEX(curp->p_hdr)) { -+ len = (len - 1) * sizeof(struct ext3_extent_idx); -+ len = len < 0 ? 0 : len; -+ ext_debug(tree, "insert new index %d after: %d. " -+ "move %d from 0x%p to 0x%p\n", -+ logical, ptr, len, -+ (curp->p_idx + 1), (curp->p_idx + 2)); -+ memmove(curp->p_idx + 2, curp->p_idx + 1, len); -+ } -+ ix = curp->p_idx + 1; -+ } else { -+ /* insert before */ -+ len = len * sizeof(struct ext3_extent_idx); -+ len = len < 0 ? 0 : len; -+ ext_debug(tree, "insert new index %d before: %d. " -+ "move %d from 0x%p to 0x%p\n", -+ logical, ptr, len, -+ curp->p_idx, (curp->p_idx + 1)); -+ memmove(curp->p_idx + 1, curp->p_idx, len); -+ ix = curp->p_idx; -+ } -+ -+ ix->ei_block = logical; -+ ix->ei_leaf = ptr; -+ ix->ei_leaf_hi = ix->ei_unused = 0; -+ curp->p_hdr->eh_entries++; -+ -+ EXT_ASSERT(curp->p_hdr->eh_entries <= curp->p_hdr->eh_max); -+ EXT_ASSERT(ix <= EXT_LAST_INDEX(curp->p_hdr)); -+ -+ err = ext3_ext_dirty(handle, tree, curp); -+ ext3_std_error(tree->inode->i_sb, err); -+ -+ return err; -+} -+ -+/* -+ * routine inserts new subtree into the path, using free index entry -+ * at depth 'at: -+ * - allocates all needed blocks (new leaf and all intermediate index blocks) -+ * - makes decision where to split -+ * - moves remaining extens and index entries (right to the split point) -+ * into the newly allocated blocks -+ * - initialize subtree -+ */ -+static int ext3_ext_split(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, -+ struct ext3_extent *newext, int at) -+{ -+ struct buffer_head *bh = NULL; -+ int depth = EXT_DEPTH(tree); -+ struct ext3_extent_header *neh; -+ struct ext3_extent_idx *fidx; -+ struct ext3_extent *ex; -+ int i = at, k, m, a; -+ unsigned long newblock, oldblock, border; -+ int *ablocks = NULL; /* array of allocated blocks */ -+ int err = 0; -+ -+ /* make decision: where to split? */ -+ /* FIXME: now desicion is simplest: at current extent */ -+ -+ /* if current leaf will be splitted, then we should use -+ * border from split point */ -+ EXT_ASSERT(path[depth].p_ext <= EXT_MAX_EXTENT(path[depth].p_hdr)); -+ if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) { -+ border = path[depth].p_ext[1].ee_block; -+ ext_debug(tree, "leaf will be splitted." -+ " next leaf starts at %d\n", -+ (int)border); -+ } else { -+ border = newext->ee_block; -+ ext_debug(tree, "leaf will be added." -+ " next leaf starts at %d\n", -+ (int)border); -+ } -+ -+ /* -+ * if error occurs, then we break processing -+ * and turn filesystem read-only. so, index won't -+ * be inserted and tree will be in consistent -+ * state. next mount will repair buffers too -+ */ -+ -+ /* -+ * get array to track all allocated blocks -+ * we need this to handle errors and free blocks -+ * upon them -+ */ -+ ablocks = kmalloc(sizeof(unsigned long) * depth, GFP_NOFS); -+ if (!ablocks) -+ return -ENOMEM; -+ memset(ablocks, 0, sizeof(unsigned long) * depth); -+ -+ /* allocate all needed blocks */ -+ ext_debug(tree, "allocate %d blocks for indexes/leaf\n", depth - at); -+ for (a = 0; a < depth - at; a++) { -+ newblock = ext3_ext_new_block(handle, tree, path, newext, &err); -+ if (newblock == 0) -+ goto cleanup; -+ ablocks[a] = newblock; -+ } -+ -+ /* initialize new leaf */ -+ newblock = ablocks[--a]; -+ EXT_ASSERT(newblock); -+ bh = sb_getblk(tree->inode->i_sb, newblock); -+ if (!bh) { -+ err = -EIO; -+ goto cleanup; -+ } -+ lock_buffer(bh); -+ -+ if ((err = ext3_journal_get_create_access(handle, bh))) -+ goto cleanup; -+ -+ neh = EXT_BLOCK_HDR(bh); -+ neh->eh_entries = 0; -+ neh->eh_max = ext3_ext_space_block(tree); -+ neh->eh_magic = EXT3_EXT_MAGIC; -+ neh->eh_depth = 0; -+ ex = EXT_FIRST_EXTENT(neh); -+ -+ /* move remain of path[depth] to the new leaf */ -+ EXT_ASSERT(path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max); -+ /* start copy from next extent */ -+ /* TODO: we could do it by single memmove */ -+ m = 0; -+ path[depth].p_ext++; -+ while (path[depth].p_ext <= -+ EXT_MAX_EXTENT(path[depth].p_hdr)) { -+ ext_debug(tree, "move %d:%d:%d in new leaf %lu\n", -+ path[depth].p_ext->ee_block, -+ path[depth].p_ext->ee_start, -+ path[depth].p_ext->ee_len, -+ newblock); -+ memmove(ex++, path[depth].p_ext++, sizeof(struct ext3_extent)); -+ neh->eh_entries++; -+ m++; -+ } -+ set_buffer_uptodate(bh); -+ unlock_buffer(bh); -+ -+ if ((err = ext3_journal_dirty_metadata(handle, bh))) -+ goto cleanup; -+ brelse(bh); -+ bh = NULL; -+ -+ /* correct old leaf */ -+ if (m) { -+ if ((err = ext3_ext_get_access(handle, tree, path + depth))) -+ goto cleanup; -+ path[depth].p_hdr->eh_entries -= m; -+ if ((err = ext3_ext_dirty(handle, tree, path + depth))) -+ goto cleanup; -+ -+ } -+ -+ /* create intermediate indexes */ -+ k = depth - at - 1; -+ EXT_ASSERT(k >= 0); -+ if (k) -+ ext_debug(tree, "create %d intermediate indices\n", k); -+ /* insert new index into current index block */ -+ /* current depth stored in i var */ -+ i = depth - 1; -+ while (k--) { -+ oldblock = newblock; -+ newblock = ablocks[--a]; -+ bh = sb_getblk(tree->inode->i_sb, newblock); -+ if (!bh) { -+ err = -EIO; -+ goto cleanup; -+ } -+ lock_buffer(bh); -+ -+ if ((err = ext3_journal_get_create_access(handle, bh))) -+ goto cleanup; -+ -+ neh = EXT_BLOCK_HDR(bh); -+ neh->eh_entries = 1; -+ neh->eh_magic = EXT3_EXT_MAGIC; -+ neh->eh_max = ext3_ext_space_block_idx(tree); -+ neh->eh_depth = depth - i; -+ fidx = EXT_FIRST_INDEX(neh); -+ fidx->ei_block = border; -+ fidx->ei_leaf = oldblock; -+ fidx->ei_leaf_hi = fidx->ei_unused = 0; -+ -+ ext_debug(tree, "int.index at %d (block %lu): %lu -> %lu\n", -+ i, newblock, border, oldblock); -+ /* copy indexes */ -+ m = 0; -+ path[i].p_idx++; -+ -+ ext_debug(tree, "cur 0x%p, last 0x%p\n", path[i].p_idx, -+ EXT_MAX_INDEX(path[i].p_hdr)); -+ EXT_ASSERT(EXT_MAX_INDEX(path[i].p_hdr) == -+ EXT_LAST_INDEX(path[i].p_hdr)); -+ while (path[i].p_idx <= EXT_MAX_INDEX(path[i].p_hdr)) { -+ ext_debug(tree, "%d: move %d:%d in new index %lu\n", -+ i, path[i].p_idx->ei_block, -+ path[i].p_idx->ei_leaf, newblock); -+ memmove(++fidx, path[i].p_idx++, -+ sizeof(struct ext3_extent_idx)); -+ neh->eh_entries++; -+ EXT_ASSERT(neh->eh_entries <= neh->eh_max); -+ m++; -+ } -+ set_buffer_uptodate(bh); -+ unlock_buffer(bh); -+ -+ if ((err = ext3_journal_dirty_metadata(handle, bh))) -+ goto cleanup; -+ brelse(bh); -+ bh = NULL; -+ -+ /* correct old index */ -+ if (m) { -+ err = ext3_ext_get_access(handle, tree, path + i); -+ if (err) -+ goto cleanup; -+ path[i].p_hdr->eh_entries -= m; -+ err = ext3_ext_dirty(handle, tree, path + i); -+ if (err) -+ goto cleanup; -+ } -+ -+ i--; -+ } -+ -+ /* insert new index */ -+ if (!err) -+ err = ext3_ext_insert_index(handle, tree, path + at, -+ border, newblock); -+ -+cleanup: -+ if (bh) { -+ if (buffer_locked(bh)) -+ unlock_buffer(bh); -+ brelse(bh); -+ } -+ -+ if (err) { -+ /* free all allocated blocks in error case */ -+ for (i = 0; i < depth; i++) { -+ if (!ablocks[i]) -+ continue; -+ ext3_free_blocks(handle, tree->inode, ablocks[i], 1); -+ } -+ } -+ kfree(ablocks); -+ -+ return err; -+} -+ -+/* -+ * routine implements tree growing procedure: -+ * - allocates new block -+ * - moves top-level data (index block or leaf) into the new block -+ * - initialize new top-level, creating index that points to the -+ * just created block -+ */ -+static int ext3_ext_grow_indepth(handle_t *handle, -+ struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, -+ struct ext3_extent *newext) -+{ -+ struct ext3_ext_path *curp = path; -+ struct ext3_extent_header *neh; -+ struct ext3_extent_idx *fidx; -+ struct buffer_head *bh; -+ unsigned long newblock; -+ int err = 0; -+ -+ newblock = ext3_ext_new_block(handle, tree, path, newext, &err); -+ if (newblock == 0) -+ return err; -+ -+ bh = sb_getblk(tree->inode->i_sb, newblock); -+ if (!bh) { -+ err = -EIO; -+ ext3_std_error(tree->inode->i_sb, err); -+ return err; -+ } -+ lock_buffer(bh); -+ -+ if ((err = ext3_journal_get_create_access(handle, bh))) { -+ unlock_buffer(bh); -+ goto out; -+ } -+ -+ /* move top-level index/leaf into new block */ -+ memmove(bh->b_data, curp->p_hdr, tree->buffer_len); -+ -+ /* set size of new block */ -+ neh = EXT_BLOCK_HDR(bh); -+ /* old root could have indexes or leaves -+ * so calculate eh_max right way */ -+ if (EXT_DEPTH(tree)) -+ neh->eh_max = ext3_ext_space_block_idx(tree); -+ else -+ neh->eh_max = ext3_ext_space_block(tree); -+ neh->eh_magic = EXT3_EXT_MAGIC; -+ set_buffer_uptodate(bh); -+ unlock_buffer(bh); -+ -+ if ((err = ext3_journal_dirty_metadata(handle, bh))) -+ goto out; -+ -+ /* create index in new top-level index: num,max,pointer */ -+ if ((err = ext3_ext_get_access(handle, tree, curp))) -+ goto out; -+ -+ curp->p_hdr->eh_magic = EXT3_EXT_MAGIC; -+ curp->p_hdr->eh_max = ext3_ext_space_root_idx(tree); -+ curp->p_hdr->eh_entries = 1; -+ curp->p_idx = EXT_FIRST_INDEX(curp->p_hdr); -+ /* FIXME: it works, but actually path[0] can be index */ -+ curp->p_idx->ei_block = EXT_FIRST_EXTENT(path[0].p_hdr)->ee_block; -+ curp->p_idx->ei_leaf = newblock; -+ curp->p_idx->ei_leaf_hi = curp->p_idx->ei_unused = 0; -+ -+ neh = EXT_ROOT_HDR(tree); -+ fidx = EXT_FIRST_INDEX(neh); -+ ext_debug(tree, "new root: num %d(%d), lblock %d, ptr %d\n", -+ neh->eh_entries, neh->eh_max, fidx->ei_block, fidx->ei_leaf); -+ -+ neh->eh_depth = path->p_depth + 1; -+ err = ext3_ext_dirty(handle, tree, curp); -+out: -+ brelse(bh); -+ -+ return err; -+} -+ -+/* -+ * routine finds empty index and adds new leaf. if no free index found -+ * then it requests in-depth growing -+ */ -+static int ext3_ext_create_new_leaf(handle_t *handle, -+ struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, -+ struct ext3_extent *newext) -+{ -+ struct ext3_ext_path *curp; -+ int depth, i, err = 0; -+ -+repeat: -+ i = depth = EXT_DEPTH(tree); -+ -+ /* walk up to the tree and look for free index entry */ -+ curp = path + depth; -+ while (i > 0 && !EXT_HAS_FREE_INDEX(curp)) { -+ i--; -+ curp--; -+ } -+ -+ /* we use already allocated block for index block -+ * so, subsequent data blocks should be contigoues */ -+ if (EXT_HAS_FREE_INDEX(curp)) { -+ /* if we found index with free entry, then use that -+ * entry: create all needed subtree and add new leaf */ -+ err = ext3_ext_split(handle, tree, path, newext, i); -+ -+ /* refill path */ -+ ext3_ext_drop_refs(path); -+ path = ext3_ext_find_extent(tree, newext->ee_block, path); -+ if (IS_ERR(path)) -+ err = PTR_ERR(path); -+ } else { -+ /* tree is full, time to grow in depth */ -+ err = ext3_ext_grow_indepth(handle, tree, path, newext); -+ -+ /* refill path */ -+ ext3_ext_drop_refs(path); -+ path = ext3_ext_find_extent(tree, newext->ee_block, path); -+ if (IS_ERR(path)) -+ err = PTR_ERR(path); -+ -+ /* -+ * only first (depth 0 -> 1) produces free space -+ * in all other cases we have to split growed tree -+ */ -+ depth = EXT_DEPTH(tree); -+ if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) { -+ /* now we need split */ -+ goto repeat; -+ } -+ } -+ -+ if (err) -+ return err; -+ -+ return 0; -+} -+ -+/* -+ * returns allocated block in subsequent extent or EXT_MAX_BLOCK -+ * NOTE: it consider block number from index entry as -+ * allocated block. thus, index entries have to be consistent -+ * with leafs -+ */ -+static unsigned long -+ext3_ext_next_allocated_block(struct ext3_ext_path *path) -+{ -+ int depth; -+ -+ EXT_ASSERT(path != NULL); -+ depth = path->p_depth; -+ -+ if (depth == 0 && path->p_ext == NULL) -+ return EXT_MAX_BLOCK; -+ -+ /* FIXME: what if index isn't full ?! */ -+ while (depth >= 0) { -+ if (depth == path->p_depth) { -+ /* leaf */ -+ if (path[depth].p_ext != -+ EXT_LAST_EXTENT(path[depth].p_hdr)) -+ return path[depth].p_ext[1].ee_block; -+ } else { -+ /* index */ -+ if (path[depth].p_idx != -+ EXT_LAST_INDEX(path[depth].p_hdr)) -+ return path[depth].p_idx[1].ei_block; -+ } -+ depth--; -+ } -+ -+ return EXT_MAX_BLOCK; -+} -+ -+/* -+ * returns first allocated block from next leaf or EXT_MAX_BLOCK -+ */ -+static unsigned ext3_ext_next_leaf_block(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path) -+{ -+ int depth; -+ -+ EXT_ASSERT(path != NULL); -+ depth = path->p_depth; -+ -+ /* zero-tree has no leaf blocks at all */ -+ if (depth == 0) -+ return EXT_MAX_BLOCK; -+ -+ /* go to index block */ -+ depth--; -+ -+ while (depth >= 0) { -+ if (path[depth].p_idx != -+ EXT_LAST_INDEX(path[depth].p_hdr)) -+ return path[depth].p_idx[1].ei_block; -+ depth--; -+ } -+ -+ return EXT_MAX_BLOCK; -+} -+ -+/* -+ * if leaf gets modified and modified extent is first in the leaf -+ * then we have to correct all indexes above -+ * TODO: do we need to correct tree in all cases? -+ */ -+int ext3_ext_correct_indexes(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path) -+{ -+ struct ext3_extent_header *eh; -+ int depth = EXT_DEPTH(tree); -+ struct ext3_extent *ex; -+ unsigned long border; -+ int k, err = 0; -+ -+ eh = path[depth].p_hdr; -+ ex = path[depth].p_ext; -+ EXT_ASSERT(ex); -+ EXT_ASSERT(eh); -+ -+ if (depth == 0) { -+ /* there is no tree at all */ -+ return 0; -+ } -+ -+ if (ex != EXT_FIRST_EXTENT(eh)) { -+ /* we correct tree if first leaf got modified only */ -+ return 0; -+ } -+ -+ /* -+ * TODO: we need correction if border is smaller then current one -+ */ -+ k = depth - 1; -+ border = path[depth].p_ext->ee_block; -+ if ((err = ext3_ext_get_access(handle, tree, path + k))) -+ return err; -+ path[k].p_idx->ei_block = border; -+ if ((err = ext3_ext_dirty(handle, tree, path + k))) -+ return err; -+ -+ while (k--) { -+ /* change all left-side indexes */ -+ if (path[k+1].p_idx != EXT_FIRST_INDEX(path[k+1].p_hdr)) -+ break; -+ if ((err = ext3_ext_get_access(handle, tree, path + k))) -+ break; -+ path[k].p_idx->ei_block = border; -+ if ((err = ext3_ext_dirty(handle, tree, path + k))) -+ break; -+ } -+ -+ return err; -+} -+ -+static int inline -+ext3_can_extents_be_merged(struct ext3_extents_tree *tree, -+ struct ext3_extent *ex1, -+ struct ext3_extent *ex2) -+{ -+ if (ex1->ee_block + ex1->ee_len != ex2->ee_block) -+ return 0; -+ -+#ifdef AGRESSIVE_TEST -+ if (ex1->ee_len >= 4) -+ return 0; -+#endif -+ -+ if (!tree->ops->mergable) -+ return 1; -+ -+ return tree->ops->mergable(ex1, ex2); -+} -+ -+/* -+ * this routine tries to merge requsted extent into the existing -+ * extent or inserts requested extent as new one into the tree, -+ * creating new leaf in no-space case -+ */ -+int ext3_ext_insert_extent(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, -+ struct ext3_extent *newext) -+{ -+ struct ext3_extent_header * eh; -+ struct ext3_extent *ex, *fex; -+ struct ext3_extent *nearex; /* nearest extent */ -+ struct ext3_ext_path *npath = NULL; -+ int depth, len, err, next; -+ -+ EXT_ASSERT(newext->ee_len > 0); -+ depth = EXT_DEPTH(tree); -+ ex = path[depth].p_ext; -+ EXT_ASSERT(path[depth].p_hdr); -+ -+ /* try to insert block into found extent and return */ -+ if (ex && ext3_can_extents_be_merged(tree, ex, newext)) { -+ ext_debug(tree, "append %d block to %d:%d (from %d)\n", -+ newext->ee_len, ex->ee_block, ex->ee_len, -+ ex->ee_start); -+ if ((err = ext3_ext_get_access(handle, tree, path + depth))) -+ return err; -+ ex->ee_len += newext->ee_len; -+ eh = path[depth].p_hdr; -+ nearex = ex; -+ goto merge; -+ } -+ -+repeat: -+ depth = EXT_DEPTH(tree); -+ eh = path[depth].p_hdr; -+ if (eh->eh_entries < eh->eh_max) -+ goto has_space; -+ -+ /* probably next leaf has space for us? */ -+ fex = EXT_LAST_EXTENT(eh); -+ next = ext3_ext_next_leaf_block(tree, path); -+ if (newext->ee_block > fex->ee_block && next != EXT_MAX_BLOCK) { -+ ext_debug(tree, "next leaf block - %d\n", next); -+ EXT_ASSERT(!npath); -+ npath = ext3_ext_find_extent(tree, next, NULL); -+ if (IS_ERR(npath)) -+ return PTR_ERR(npath); -+ EXT_ASSERT(npath->p_depth == path->p_depth); -+ eh = npath[depth].p_hdr; -+ if (eh->eh_entries < eh->eh_max) { -+ ext_debug(tree, "next leaf isnt full(%d)\n", -+ eh->eh_entries); -+ path = npath; -+ goto repeat; -+ } -+ ext_debug(tree, "next leaf hasno free space(%d,%d)\n", -+ eh->eh_entries, eh->eh_max); -+ } -+ -+ /* -+ * there is no free space in found leaf -+ * we're gonna add new leaf in the tree -+ */ -+ err = ext3_ext_create_new_leaf(handle, tree, path, newext); -+ if (err) -+ goto cleanup; -+ depth = EXT_DEPTH(tree); -+ eh = path[depth].p_hdr; -+ -+has_space: -+ nearex = path[depth].p_ext; -+ -+ if ((err = ext3_ext_get_access(handle, tree, path + depth))) -+ goto cleanup; -+ -+ if (!nearex) { -+ /* there is no extent in this leaf, create first one */ -+ ext_debug(tree, "first extent in the leaf: %d:%d:%d\n", -+ newext->ee_block, newext->ee_start, -+ newext->ee_len); -+ path[depth].p_ext = EXT_FIRST_EXTENT(eh); -+ } else if (newext->ee_block > nearex->ee_block) { -+ EXT_ASSERT(newext->ee_block != nearex->ee_block); -+ if (nearex != EXT_LAST_EXTENT(eh)) { -+ len = EXT_MAX_EXTENT(eh) - nearex; -+ len = (len - 1) * sizeof(struct ext3_extent); -+ len = len < 0 ? 0 : len; -+ ext_debug(tree, "insert %d:%d:%d after: nearest 0x%p, " -+ "move %d from 0x%p to 0x%p\n", -+ newext->ee_block, newext->ee_start, -+ newext->ee_len, -+ nearex, len, nearex + 1, nearex + 2); -+ memmove(nearex + 2, nearex + 1, len); -+ } -+ path[depth].p_ext = nearex + 1; -+ } else { -+ EXT_ASSERT(newext->ee_block != nearex->ee_block); -+ len = (EXT_MAX_EXTENT(eh) - nearex) * sizeof(struct ext3_extent); -+ len = len < 0 ? 0 : len; -+ ext_debug(tree, "insert %d:%d:%d before: nearest 0x%p, " -+ "move %d from 0x%p to 0x%p\n", -+ newext->ee_block, newext->ee_start, newext->ee_len, -+ nearex, len, nearex + 1, nearex + 2); -+ memmove(nearex + 1, nearex, len); -+ path[depth].p_ext = nearex; -+ } -+ -+ eh->eh_entries++; -+ nearex = path[depth].p_ext; -+ nearex->ee_block = newext->ee_block; -+ nearex->ee_start = newext->ee_start; -+ nearex->ee_len = newext->ee_len; -+ /* FIXME: support for large fs */ -+ nearex->ee_start_hi = 0; -+ -+merge: -+ /* try to merge extents to the right */ -+ while (nearex < EXT_LAST_EXTENT(eh)) { -+ if (!ext3_can_extents_be_merged(tree, nearex, nearex + 1)) -+ break; -+ /* merge with next extent! */ -+ nearex->ee_len += nearex[1].ee_len; -+ if (nearex + 1 < EXT_LAST_EXTENT(eh)) { -+ len = (EXT_LAST_EXTENT(eh) - nearex - 1) * -+ sizeof(struct ext3_extent); -+ memmove(nearex + 1, nearex + 2, len); -+ } -+ eh->eh_entries--; -+ EXT_ASSERT(eh->eh_entries > 0); -+ } -+ -+ /* try to merge extents to the left */ -+ -+ /* time to correct all indexes above */ -+ err = ext3_ext_correct_indexes(handle, tree, path); -+ if (err) -+ goto cleanup; -+ -+ err = ext3_ext_dirty(handle, tree, path + depth); -+ -+cleanup: -+ if (npath) { -+ ext3_ext_drop_refs(npath); -+ kfree(npath); -+ } -+ ext3_ext_tree_changed(tree); -+ ext3_ext_invalidate_cache(tree); -+ return err; -+} -+ -+int ext3_ext_walk_space(struct ext3_extents_tree *tree, unsigned long block, -+ unsigned long num, ext_prepare_callback func) -+{ -+ struct ext3_ext_path *path = NULL; -+ struct ext3_ext_cache cbex; -+ struct ext3_extent *ex; -+ unsigned long next, start = 0, end = 0; -+ unsigned long last = block + num; -+ int depth, exists, err = 0; -+ -+ EXT_ASSERT(tree); -+ EXT_ASSERT(func); -+ EXT_ASSERT(tree->inode); -+ EXT_ASSERT(tree->root); -+ -+ while (block < last && block != EXT_MAX_BLOCK) { -+ num = last - block; -+ /* find extent for this block */ -+ path = ext3_ext_find_extent(tree, block, path); -+ if (IS_ERR(path)) { -+ err = PTR_ERR(path); -+ path = NULL; -+ break; -+ } -+ -+ depth = EXT_DEPTH(tree); -+ EXT_ASSERT(path[depth].p_hdr); -+ ex = path[depth].p_ext; -+ next = ext3_ext_next_allocated_block(path); -+ -+ exists = 0; -+ if (!ex) { -+ /* there is no extent yet, so try to allocate -+ * all requested space */ -+ start = block; -+ end = block + num; -+ } else if (ex->ee_block > block) { -+ /* need to allocate space before found extent */ -+ start = block; -+ end = ex->ee_block; -+ if (block + num < end) -+ end = block + num; -+ } else if (block >= ex->ee_block + ex->ee_len) { -+ /* need to allocate space after found extent */ -+ start = block; -+ end = block + num; -+ if (end >= next) -+ end = next; -+ } else if (block >= ex->ee_block) { -+ /* -+ * some part of requested space is covered -+ * by found extent -+ */ -+ start = block; -+ end = ex->ee_block + ex->ee_len; -+ if (block + num < end) -+ end = block + num; -+ exists = 1; -+ } else { -+ BUG(); -+ } -+ EXT_ASSERT(end > start); -+ -+ if (!exists) { -+ cbex.ec_block = start; -+ cbex.ec_len = end - start; -+ cbex.ec_start = 0; -+ cbex.ec_type = EXT3_EXT_CACHE_GAP; -+ } else { -+ cbex.ec_block = ex->ee_block; -+ cbex.ec_len = ex->ee_len; -+ cbex.ec_start = ex->ee_start; -+ cbex.ec_type = EXT3_EXT_CACHE_EXTENT; -+ } -+ -+ EXT_ASSERT(cbex.ec_len > 0); -+ EXT_ASSERT(path[depth].p_hdr); -+ err = func(tree, path, &cbex); -+ ext3_ext_drop_refs(path); -+ -+ if (err < 0) -+ break; -+ if (err == EXT_REPEAT) -+ continue; -+ else if (err == EXT_BREAK) { -+ err = 0; -+ break; -+ } -+ -+ if (EXT_DEPTH(tree) != depth) { -+ /* depth was changed. we have to realloc path */ -+ kfree(path); -+ path = NULL; -+ } -+ -+ block = cbex.ec_block + cbex.ec_len; -+ } -+ -+ if (path) { -+ ext3_ext_drop_refs(path); -+ kfree(path); -+ } -+ -+ return err; -+} -+ -+static inline void -+ext3_ext_put_in_cache(struct ext3_extents_tree *tree, __u32 block, -+ __u32 len, __u32 start, int type) -+{ -+ EXT_ASSERT(len > 0); -+ if (tree->cex) { -+ tree->cex->ec_type = type; -+ tree->cex->ec_block = block; -+ tree->cex->ec_len = len; -+ tree->cex->ec_start = start; -+ } -+} -+ -+/* -+ * this routine calculate boundaries of the gap requested block fits into -+ * and cache this gap -+ */ -+static inline void -+ext3_ext_put_gap_in_cache(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, -+ unsigned long block) -+{ -+ int depth = EXT_DEPTH(tree); -+ unsigned long lblock, len; -+ struct ext3_extent *ex; -+ -+ if (!tree->cex) -+ return; -+ -+ ex = path[depth].p_ext; -+ if (ex == NULL) { -+ /* there is no extent yet, so gap is [0;-] */ -+ lblock = 0; -+ len = EXT_MAX_BLOCK; -+ ext_debug(tree, "cache gap(whole file):"); -+ } else if (block < ex->ee_block) { -+ lblock = block; -+ len = ex->ee_block - block; -+ ext_debug(tree, "cache gap(before): %lu [%lu:%lu]", -+ (unsigned long) block, -+ (unsigned long) ex->ee_block, -+ (unsigned long) ex->ee_len); -+ } else if (block >= ex->ee_block + ex->ee_len) { -+ lblock = ex->ee_block + ex->ee_len; -+ len = ext3_ext_next_allocated_block(path); -+ ext_debug(tree, "cache gap(after): [%lu:%lu] %lu", -+ (unsigned long) ex->ee_block, -+ (unsigned long) ex->ee_len, -+ (unsigned long) block); -+ EXT_ASSERT(len > lblock); -+ len = len - lblock; -+ } else { -+ lblock = len = 0; -+ BUG(); -+ } -+ -+ ext_debug(tree, " -> %lu:%lu\n", (unsigned long) lblock, len); -+ ext3_ext_put_in_cache(tree, lblock, len, 0, EXT3_EXT_CACHE_GAP); -+} -+ -+static inline int -+ext3_ext_in_cache(struct ext3_extents_tree *tree, unsigned long block, -+ struct ext3_extent *ex) -+{ -+ struct ext3_ext_cache *cex = tree->cex; -+ -+ /* is there cache storage at all? */ -+ if (!cex) -+ return EXT3_EXT_CACHE_NO; -+ -+ /* has cache valid data? */ -+ if (cex->ec_type == EXT3_EXT_CACHE_NO) -+ return EXT3_EXT_CACHE_NO; -+ -+ EXT_ASSERT(cex->ec_type == EXT3_EXT_CACHE_GAP || -+ cex->ec_type == EXT3_EXT_CACHE_EXTENT); -+ if (block >= cex->ec_block && block < cex->ec_block + cex->ec_len) { -+ ex->ee_block = cex->ec_block; -+ ex->ee_start = cex->ec_start; -+ ex->ee_start_hi = 0; -+ ex->ee_len = cex->ec_len; -+ ext_debug(tree, "%lu cached by %lu:%lu:%lu\n", -+ (unsigned long) block, -+ (unsigned long) ex->ee_block, -+ (unsigned long) ex->ee_len, -+ (unsigned long) ex->ee_start); -+ return cex->ec_type; -+ } -+ -+ /* not in cache */ -+ return EXT3_EXT_CACHE_NO; -+} -+ -+/* -+ * routine removes index from the index block -+ * it's used in truncate case only. thus all requests are for -+ * last index in the block only -+ */ -+int ext3_ext_rm_idx(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path) -+{ -+ struct buffer_head *bh; -+ int err; -+ -+ /* free index block */ -+ path--; -+ EXT_ASSERT(path->p_hdr->eh_entries); -+ if ((err = ext3_ext_get_access(handle, tree, path))) -+ return err; -+ path->p_hdr->eh_entries--; -+ if ((err = ext3_ext_dirty(handle, tree, path))) -+ return err; -+ ext_debug(tree, "index is empty, remove it, free block %d\n", -+ path->p_idx->ei_leaf); -+ bh = sb_find_get_block(tree->inode->i_sb, path->p_idx->ei_leaf); -+ ext3_forget(handle, 1, tree->inode, bh, path->p_idx->ei_leaf); -+ ext3_free_blocks(handle, tree->inode, path->p_idx->ei_leaf, 1); -+ return err; -+} -+ -+int ext3_ext_calc_credits_for_insert(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path) -+{ -+ int depth = EXT_DEPTH(tree); -+ int needed; -+ -+ if (path) { -+ /* probably there is space in leaf? */ -+ if (path[depth].p_hdr->eh_entries < path[depth].p_hdr->eh_max) -+ return 1; -+ } -+ -+ /* -+ * the worste case we're expecting is creation of the -+ * new root (growing in depth) with index splitting -+ * for splitting we have to consider depth + 1 because -+ * previous growing could increase it -+ */ -+ depth = depth + 1; -+ -+ /* -+ * growing in depth: -+ * block allocation + new root + old root -+ */ -+ needed = EXT3_ALLOC_NEEDED + 2; -+ -+ /* index split. we may need: -+ * allocate intermediate indexes and new leaf -+ * change two blocks at each level, but root -+ * modify root block (inode) -+ */ -+ needed += (depth * EXT3_ALLOC_NEEDED) + (2 * depth) + 1; -+ -+ return needed; -+} -+ -+static int -+ext3_ext_split_for_rm(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, unsigned long start, -+ unsigned long end) -+{ -+ struct ext3_extent *ex, tex; -+ struct ext3_ext_path *npath; -+ int depth, creds, err; -+ -+ depth = EXT_DEPTH(tree); -+ ex = path[depth].p_ext; -+ EXT_ASSERT(ex); -+ EXT_ASSERT(end < ex->ee_block + ex->ee_len - 1); -+ EXT_ASSERT(ex->ee_block < start); -+ -+ /* calculate tail extent */ -+ tex.ee_block = end + 1; -+ EXT_ASSERT(tex.ee_block < ex->ee_block + ex->ee_len); -+ tex.ee_len = ex->ee_block + ex->ee_len - tex.ee_block; -+ -+ creds = ext3_ext_calc_credits_for_insert(tree, path); -+ handle = ext3_ext_journal_restart(handle, creds); -+ if (IS_ERR(handle)) -+ return PTR_ERR(handle); -+ -+ /* calculate head extent. use primary extent */ -+ err = ext3_ext_get_access(handle, tree, path + depth); -+ if (err) -+ return err; -+ ex->ee_len = start - ex->ee_block; -+ err = ext3_ext_dirty(handle, tree, path + depth); -+ if (err) -+ return err; -+ -+ /* FIXME: some callback to free underlying resource -+ * and correct ee_start? */ -+ ext_debug(tree, "split extent: head %u:%u, tail %u:%u\n", -+ ex->ee_block, ex->ee_len, tex.ee_block, tex.ee_len); -+ -+ npath = ext3_ext_find_extent(tree, ex->ee_block, NULL); -+ if (IS_ERR(npath)) -+ return PTR_ERR(npath); -+ depth = EXT_DEPTH(tree); -+ EXT_ASSERT(npath[depth].p_ext->ee_block == ex->ee_block); -+ EXT_ASSERT(npath[depth].p_ext->ee_len == ex->ee_len); -+ -+ err = ext3_ext_insert_extent(handle, tree, npath, &tex); -+ ext3_ext_drop_refs(npath); -+ kfree(npath); -+ -+ return err; -+} -+ -+static int -+ext3_ext_rm_leaf(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, unsigned long start, -+ unsigned long end) -+{ -+ struct ext3_extent *ex, *fu = NULL, *lu, *le; -+ int err = 0, correct_index = 0; -+ int depth = EXT_DEPTH(tree), credits; -+ struct ext3_extent_header *eh; -+ unsigned a, b, block, num; -+ -+ ext_debug(tree, "remove [%lu:%lu] in leaf\n", start, end); -+ if (!path[depth].p_hdr) -+ path[depth].p_hdr = EXT_BLOCK_HDR(path[depth].p_bh); -+ eh = path[depth].p_hdr; -+ EXT_ASSERT(eh); -+ EXT_ASSERT(eh->eh_entries <= eh->eh_max); -+ EXT_ASSERT(eh->eh_magic == EXT3_EXT_MAGIC); -+ -+ /* find where to start removing */ -+ le = ex = EXT_LAST_EXTENT(eh); -+ while (ex != EXT_FIRST_EXTENT(eh)) { -+ if (ex->ee_block <= end) -+ break; -+ ex--; -+ } -+ -+ if (start > ex->ee_block && end < ex->ee_block + ex->ee_len - 1) { -+ /* removal of internal part of the extent requested -+ * tail and head must be placed in different extent -+ * so, we have to insert one more extent */ -+ path[depth].p_ext = ex; -+ return ext3_ext_split_for_rm(handle, tree, path, start, end); -+ } -+ -+ lu = ex; -+ while (ex >= EXT_FIRST_EXTENT(eh) && ex->ee_block + ex->ee_len > start) { -+ ext_debug(tree, "remove ext %u:%u\n", ex->ee_block, ex->ee_len); -+ path[depth].p_ext = ex; -+ -+ a = ex->ee_block > start ? ex->ee_block : start; -+ b = ex->ee_block + ex->ee_len - 1 < end ? -+ ex->ee_block + ex->ee_len - 1 : end; -+ -+ ext_debug(tree, " border %u:%u\n", a, b); -+ -+ if (a != ex->ee_block && b != ex->ee_block + ex->ee_len - 1) { -+ block = 0; -+ num = 0; -+ BUG(); -+ } else if (a != ex->ee_block) { -+ /* remove tail of the extent */ -+ block = ex->ee_block; -+ num = a - block; -+ } else if (b != ex->ee_block + ex->ee_len - 1) { -+ /* remove head of the extent */ -+ block = a; -+ num = b - a; -+ } else { -+ /* remove whole extent: excelent! */ -+ block = ex->ee_block; -+ num = 0; -+ EXT_ASSERT(a == ex->ee_block && -+ b == ex->ee_block + ex->ee_len - 1); -+ } -+ -+ if (ex == EXT_FIRST_EXTENT(eh)) -+ correct_index = 1; -+ -+ credits = 1; -+ if (correct_index) -+ credits += (EXT_DEPTH(tree) * EXT3_ALLOC_NEEDED) + 1; -+ if (tree->ops->remove_extent_credits) -+ credits+=tree->ops->remove_extent_credits(tree,ex,a,b); -+ -+ handle = ext3_ext_journal_restart(handle, credits); -+ if (IS_ERR(handle)) { -+ err = PTR_ERR(handle); -+ goto out; -+ } -+ -+ err = ext3_ext_get_access(handle, tree, path + depth); -+ if (err) -+ goto out; -+ -+ if (tree->ops->remove_extent) -+ err = tree->ops->remove_extent(tree, ex, a, b); -+ if (err) -+ goto out; -+ -+ if (num == 0) { -+ /* this extent is removed entirely mark slot unused */ -+ ex->ee_start = ex->ee_start_hi = 0; -+ eh->eh_entries--; -+ fu = ex; -+ } -+ -+ ex->ee_block = block; -+ ex->ee_len = num; -+ -+ err = ext3_ext_dirty(handle, tree, path + depth); -+ if (err) -+ goto out; -+ -+ ext_debug(tree, "new extent: %u:%u:%u\n", -+ ex->ee_block, ex->ee_len, ex->ee_start); -+ ex--; -+ } -+ -+ if (fu) { -+ /* reuse unused slots */ -+ while (lu < le) { -+ if (lu->ee_start) { -+ *fu = *lu; -+ lu->ee_start = lu->ee_start_hi = 0; -+ fu++; -+ } -+ lu++; -+ } -+ } -+ -+ if (correct_index && eh->eh_entries) -+ err = ext3_ext_correct_indexes(handle, tree, path); -+ -+ /* if this leaf is free, then we should -+ * remove it from index block above */ -+ if (err == 0 && eh->eh_entries == 0 && path[depth].p_bh != NULL) -+ err = ext3_ext_rm_idx(handle, tree, path + depth); -+ -+out: -+ return err; -+} -+ -+ -+static struct ext3_extent_idx * -+ext3_ext_last_covered(struct ext3_extent_header *hdr, unsigned long block) -+{ -+ struct ext3_extent_idx *ix; -+ -+ ix = EXT_LAST_INDEX(hdr); -+ while (ix != EXT_FIRST_INDEX(hdr)) { -+ if (ix->ei_block <= block) -+ break; -+ ix--; -+ } -+ return ix; -+} -+ -+/* -+ * returns 1 if current index have to be freed (even partial) -+ */ -+static int inline -+ext3_ext_more_to_rm(struct ext3_ext_path *path) -+{ -+ EXT_ASSERT(path->p_idx); -+ -+ if (path->p_idx < EXT_FIRST_INDEX(path->p_hdr)) -+ return 0; -+ -+ /* -+ * if truncate on deeper level happened it it wasn't partial -+ * so we have to consider current index for truncation -+ */ -+ if (path->p_hdr->eh_entries == path->p_block) -+ return 0; -+ return 1; -+} -+ -+int ext3_ext_remove_space(struct ext3_extents_tree *tree, -+ unsigned long start, unsigned long end) -+{ -+ struct inode *inode = tree->inode; -+ struct super_block *sb = inode->i_sb; -+ int depth = EXT_DEPTH(tree); -+ struct ext3_ext_path *path; -+ handle_t *handle; -+ int i = 0, err = 0; -+ -+ ext_debug(tree, "space to be removed: %lu:%lu\n", start, end); -+ -+ /* probably first extent we're gonna free will be last in block */ -+ handle = ext3_journal_start(inode, depth + 1); -+ if (IS_ERR(handle)) -+ return PTR_ERR(handle); -+ -+ ext3_ext_invalidate_cache(tree); -+ -+ /* -+ * we start scanning from right side freeing all the blocks -+ * after i_size and walking into the deep -+ */ -+ path = kmalloc(sizeof(struct ext3_ext_path) * (depth + 1), GFP_KERNEL); -+ if (IS_ERR(path)) { -+ ext3_error(sb, __FUNCTION__, "Can't allocate path array"); -+ ext3_journal_stop(handle); -+ return -ENOMEM; -+ } -+ memset(path, 0, sizeof(struct ext3_ext_path) * (depth + 1)); -+ path[i].p_hdr = EXT_ROOT_HDR(tree); -+ -+ while (i >= 0 && err == 0) { -+ if (i == depth) { -+ /* this is leaf block */ -+ err = ext3_ext_rm_leaf(handle, tree, path, start, end); -+ /* root level have p_bh == NULL, brelse() eats this */ -+ brelse(path[i].p_bh); -+ i--; -+ continue; -+ } -+ -+ /* this is index block */ -+ if (!path[i].p_hdr) { -+ ext_debug(tree, "initialize header\n"); -+ path[i].p_hdr = EXT_BLOCK_HDR(path[i].p_bh); -+ } -+ -+ EXT_ASSERT(path[i].p_hdr->eh_entries <= path[i].p_hdr->eh_max); -+ EXT_ASSERT(path[i].p_hdr->eh_magic == EXT3_EXT_MAGIC); -+ -+ if (!path[i].p_idx) { -+ /* this level hasn't touched yet */ -+ path[i].p_idx = -+ ext3_ext_last_covered(path[i].p_hdr, end); -+ path[i].p_block = path[i].p_hdr->eh_entries + 1; -+ ext_debug(tree, "init index ptr: hdr 0x%p, num %d\n", -+ path[i].p_hdr, path[i].p_hdr->eh_entries); -+ } else { -+ /* we've already was here, see at next index */ -+ path[i].p_idx--; -+ } -+ -+ ext_debug(tree, "level %d - index, first 0x%p, cur 0x%p\n", -+ i, EXT_FIRST_INDEX(path[i].p_hdr), -+ path[i].p_idx); -+ if (ext3_ext_more_to_rm(path + i)) { -+ /* go to the next level */ -+ ext_debug(tree, "move to level %d (block %d)\n", -+ i + 1, path[i].p_idx->ei_leaf); -+ memset(path + i + 1, 0, sizeof(*path)); -+ path[i+1].p_bh = sb_bread(sb, path[i].p_idx->ei_leaf); -+ if (!path[i+1].p_bh) { -+ /* should we reset i_size? */ -+ err = -EIO; -+ break; -+ } -+ /* put actual number of indexes to know is this -+ * number got changed at the next iteration */ -+ path[i].p_block = path[i].p_hdr->eh_entries; -+ i++; -+ } else { -+ /* we finish processing this index, go up */ -+ if (path[i].p_hdr->eh_entries == 0 && i > 0) { -+ /* index is empty, remove it -+ * handle must be already prepared by the -+ * truncatei_leaf() */ -+ err = ext3_ext_rm_idx(handle, tree, path + i); -+ } -+ /* root level have p_bh == NULL, brelse() eats this */ -+ brelse(path[i].p_bh); -+ i--; -+ ext_debug(tree, "return to level %d\n", i); -+ } -+ } -+ -+ /* TODO: flexible tree reduction should be here */ -+ if (path->p_hdr->eh_entries == 0) { -+ /* -+ * truncate to zero freed all the tree -+ * so, we need to correct eh_depth -+ */ -+ err = ext3_ext_get_access(handle, tree, path); -+ if (err == 0) { -+ EXT_ROOT_HDR(tree)->eh_depth = 0; -+ EXT_ROOT_HDR(tree)->eh_max = ext3_ext_space_root(tree); -+ err = ext3_ext_dirty(handle, tree, path); -+ } -+ } -+ ext3_ext_tree_changed(tree); -+ -+ kfree(path); -+ ext3_journal_stop(handle); -+ -+ return err; -+} -+ -+int ext3_ext_calc_metadata_amount(struct ext3_extents_tree *tree, int blocks) -+{ -+ int lcap, icap, rcap, leafs, idxs, num; -+ -+ rcap = ext3_ext_space_root(tree); -+ if (blocks <= rcap) { -+ /* all extents fit to the root */ -+ return 0; -+ } -+ -+ rcap = ext3_ext_space_root_idx(tree); -+ lcap = ext3_ext_space_block(tree); -+ icap = ext3_ext_space_block_idx(tree); -+ -+ num = leafs = (blocks + lcap - 1) / lcap; -+ if (leafs <= rcap) { -+ /* all pointers to leafs fit to the root */ -+ return leafs; -+ } -+ -+ /* ok. we need separate index block(s) to link all leaf blocks */ -+ idxs = (leafs + icap - 1) / icap; -+ do { -+ num += idxs; -+ idxs = (idxs + icap - 1) / icap; -+ } while (idxs > rcap); -+ -+ return num; -+} -+ -+/* -+ * called at mount time -+ */ -+void ext3_ext_init(struct super_block *sb) -+{ -+ /* -+ * possible initialization would be here -+ */ -+ -+ if (test_opt(sb, EXTENTS)) { -+ printk("EXT3-fs: file extents enabled"); -+#ifdef AGRESSIVE_TEST -+ printk(", agressive tests"); -+#endif -+#ifdef CHECK_BINSEARCH -+ printk(", check binsearch"); -+#endif -+ printk("\n"); -+ } -+} -+ -+/* -+ * called at umount time -+ */ -+void ext3_ext_release(struct super_block *sb) -+{ -+} -+ -+/************************************************************************ -+ * VFS related routines -+ ************************************************************************/ -+ -+static int ext3_get_inode_write_access(handle_t *handle, void *buffer) -+{ -+ /* we use in-core data, not bh */ -+ return 0; -+} -+ -+static int ext3_mark_buffer_dirty(handle_t *handle, void *buffer) -+{ -+ struct inode *inode = buffer; -+ return ext3_mark_inode_dirty(handle, inode); -+} -+ -+static int ext3_ext_mergable(struct ext3_extent *ex1, -+ struct ext3_extent *ex2) -+{ -+ /* FIXME: support for large fs */ -+ if (ex1->ee_start + ex1->ee_len == ex2->ee_start) -+ return 1; -+ return 0; -+} -+ -+static int -+ext3_remove_blocks_credits(struct ext3_extents_tree *tree, -+ struct ext3_extent *ex, -+ unsigned long from, unsigned long to) -+{ -+ int needed; -+ -+ /* at present, extent can't cross block group */; -+ needed = 4; /* bitmap + group desc + sb + inode */ -+ -+#ifdef CONFIG_QUOTA -+ needed += 2 * EXT3_SINGLEDATA_TRANS_BLOCKS; -+#endif -+ return needed; -+} -+ -+static int -+ext3_remove_blocks(struct ext3_extents_tree *tree, -+ struct ext3_extent *ex, -+ unsigned long from, unsigned long to) -+{ -+ int needed = ext3_remove_blocks_credits(tree, ex, from, to); -+ handle_t *handle = ext3_journal_start(tree->inode, needed); -+ struct buffer_head *bh; -+ int i; -+ -+ if (IS_ERR(handle)) -+ return PTR_ERR(handle); -+ if (from >= ex->ee_block && to == ex->ee_block + ex->ee_len - 1) { -+ /* tail removal */ -+ unsigned long num, start; -+ num = ex->ee_block + ex->ee_len - from; -+ start = ex->ee_start + ex->ee_len - num; -+ ext_debug(tree, "free last %lu blocks starting %lu\n", -+ num, start); -+ for (i = 0; i < num; i++) { -+ bh = sb_find_get_block(tree->inode->i_sb, start + i); -+ ext3_forget(handle, 0, tree->inode, bh, start + i); -+ } -+ ext3_free_blocks(handle, tree->inode, start, num); -+ } else if (from == ex->ee_block && to <= ex->ee_block + ex->ee_len - 1) { -+ printk("strange request: removal %lu-%lu from %u:%u\n", -+ from, to, ex->ee_block, ex->ee_len); -+ } else { -+ printk("strange request: removal(2) %lu-%lu from %u:%u\n", -+ from, to, ex->ee_block, ex->ee_len); -+ } -+ ext3_journal_stop(handle); -+ return 0; -+} -+ -+static int ext3_ext_find_goal(struct inode *inode, -+ struct ext3_ext_path *path, unsigned long block) -+{ -+ struct ext3_inode_info *ei = EXT3_I(inode); -+ unsigned long bg_start; -+ unsigned long colour; -+ int depth; -+ -+ if (path) { -+ struct ext3_extent *ex; -+ depth = path->p_depth; -+ -+ /* try to predict block placement */ -+ if ((ex = path[depth].p_ext)) -+ return ex->ee_start + (block - ex->ee_block); -+ -+ /* it looks index is empty -+ * try to find starting from index itself */ -+ if (path[depth].p_bh) -+ return path[depth].p_bh->b_blocknr; -+ } -+ -+ /* OK. use inode's group */ -+ bg_start = (ei->i_block_group * EXT3_BLOCKS_PER_GROUP(inode->i_sb)) + -+ le32_to_cpu(EXT3_SB(inode->i_sb)->s_es->s_first_data_block); -+ colour = (current->pid % 16) * -+ (EXT3_BLOCKS_PER_GROUP(inode->i_sb) / 16); -+ return bg_start + colour + block; -+} -+ -+static int ext3_new_block_cb(handle_t *handle, struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, -+ struct ext3_extent *ex, int *err) -+{ -+ struct inode *inode = tree->inode; -+ int newblock, goal; -+ -+ EXT_ASSERT(path); -+ EXT_ASSERT(ex); -+ EXT_ASSERT(ex->ee_start); -+ EXT_ASSERT(ex->ee_len); -+ -+ /* reuse block from the extent to order data/metadata */ -+ newblock = ex->ee_start++; -+ ex->ee_len--; -+ if (ex->ee_len == 0) { -+ ex->ee_len = 1; -+ /* allocate new block for the extent */ -+ goal = ext3_ext_find_goal(inode, path, ex->ee_block); -+ ex->ee_start = ext3_new_block(handle, inode, goal, err); -+ ex->ee_start_hi = 0; -+ if (ex->ee_start == 0) { -+ /* error occured: restore old extent */ -+ ex->ee_start = newblock; -+ return 0; -+ } -+ } -+ return newblock; -+} -+ -+static struct ext3_extents_helpers ext3_blockmap_helpers = { -+ .get_write_access = ext3_get_inode_write_access, -+ .mark_buffer_dirty = ext3_mark_buffer_dirty, -+ .mergable = ext3_ext_mergable, -+ .new_block = ext3_new_block_cb, -+ .remove_extent = ext3_remove_blocks, -+ .remove_extent_credits = ext3_remove_blocks_credits, -+}; -+ -+void ext3_init_tree_desc(struct ext3_extents_tree *tree, -+ struct inode *inode) -+{ -+ tree->inode = inode; -+ tree->root = (void *) EXT3_I(inode)->i_data; -+ tree->buffer = (void *) inode; -+ tree->buffer_len = sizeof(EXT3_I(inode)->i_data); -+ tree->cex = (struct ext3_ext_cache *) &EXT3_I(inode)->i_cached_extent; -+ tree->ops = &ext3_blockmap_helpers; -+} -+ -+int ext3_ext_get_block(handle_t *handle, struct inode *inode, -+ long iblock, struct buffer_head *bh_result, -+ int create, int extend_disksize) -+{ -+ struct ext3_ext_path *path = NULL; -+ struct ext3_extent newex; -+ struct ext3_extent *ex; -+ int goal, newblock, err = 0, depth; -+ struct ext3_extents_tree tree; -+ -+ clear_buffer_new(bh_result); -+ ext3_init_tree_desc(&tree, inode); -+ ext_debug(&tree, "block %d requested for inode %u\n", -+ (int) iblock, (unsigned) inode->i_ino); -+ down(&EXT3_I(inode)->truncate_sem); -+ -+ /* check in cache */ -+ if ((goal = ext3_ext_in_cache(&tree, iblock, &newex))) { -+ if (goal == EXT3_EXT_CACHE_GAP) { -+ if (!create) { -+ /* block isn't allocated yet and -+ * user don't want to allocate it */ -+ goto out2; -+ } -+ /* we should allocate requested block */ -+ } else if (goal == EXT3_EXT_CACHE_EXTENT) { -+ /* block is already allocated */ -+ newblock = iblock - newex.ee_block + newex.ee_start; -+ goto out; -+ } else { -+ EXT_ASSERT(0); -+ } -+ } -+ -+ /* find extent for this block */ -+ path = ext3_ext_find_extent(&tree, iblock, NULL); -+ if (IS_ERR(path)) { -+ err = PTR_ERR(path); -+ path = NULL; -+ goto out2; -+ } -+ -+ depth = EXT_DEPTH(&tree); -+ -+ /* -+ * consistent leaf must not be empty -+ * this situations is possible, though, _during_ tree modification -+ * this is why assert can't be put in ext3_ext_find_extent() -+ */ -+ EXT_ASSERT(path[depth].p_ext != NULL || depth == 0); -+ -+ if ((ex = path[depth].p_ext)) { -+ /* if found exent covers block, simple return it */ -+ if (iblock >= ex->ee_block && iblock < ex->ee_block + ex->ee_len) { -+ newblock = iblock - ex->ee_block + ex->ee_start; -+ ext_debug(&tree, "%d fit into %d:%d -> %d\n", -+ (int) iblock, ex->ee_block, ex->ee_len, -+ newblock); -+ ext3_ext_put_in_cache(&tree, ex->ee_block, -+ ex->ee_len, ex->ee_start, -+ EXT3_EXT_CACHE_EXTENT); -+ goto out; -+ } -+ } -+ -+ /* -+ * requested block isn't allocated yet -+ * we couldn't try to create block if create flag is zero -+ */ -+ if (!create) { -+ /* put just found gap into cache to speedup subsequest reqs */ -+ ext3_ext_put_gap_in_cache(&tree, path, iblock); -+ goto out2; -+ } -+ -+ /* allocate new block */ -+ goal = ext3_ext_find_goal(inode, path, iblock); -+ newblock = ext3_new_block(handle, inode, goal, &err); -+ if (!newblock) -+ goto out2; -+ ext_debug(&tree, "allocate new block: goal %d, found %d\n", -+ goal, newblock); -+ -+ /* try to insert new extent into found leaf and return */ -+ newex.ee_block = iblock; -+ newex.ee_start = newblock; -+ newex.ee_start_hi = 0; -+ newex.ee_len = 1; -+ err = ext3_ext_insert_extent(handle, &tree, path, &newex); -+ if (err) -+ goto out2; -+ -+ if (extend_disksize && inode->i_size > EXT3_I(inode)->i_disksize) -+ EXT3_I(inode)->i_disksize = inode->i_size; -+ -+ /* previous routine could use block we allocated */ -+ newblock = newex.ee_start; -+ set_buffer_new(bh_result); -+ -+ ext3_ext_put_in_cache(&tree, newex.ee_block, newex.ee_len, -+ newex.ee_start, EXT3_EXT_CACHE_EXTENT); -+out: -+ ext3_ext_show_leaf(&tree, path); -+ map_bh(bh_result, inode->i_sb, newblock); -+out2: -+ if (path) { -+ ext3_ext_drop_refs(path); -+ kfree(path); -+ } -+ up(&EXT3_I(inode)->truncate_sem); -+ -+ return err; -+} -+ -+void ext3_ext_truncate(struct inode * inode, struct page *page) -+{ -+ struct address_space *mapping = inode->i_mapping; -+ struct super_block *sb = inode->i_sb; -+ struct ext3_extents_tree tree; -+ unsigned long last_block; -+ handle_t *handle; -+ int err = 0; -+ -+ ext3_init_tree_desc(&tree, inode); -+ -+ /* -+ * probably first extent we're gonna free will be last in block -+ */ -+ err = ext3_writepage_trans_blocks(inode) + 3; -+ handle = ext3_journal_start(inode, err); -+ if (IS_ERR(handle)) { -+ if (page) { -+ clear_highpage(page); -+ flush_dcache_page(page); -+ unlock_page(page); -+ page_cache_release(page); -+ } -+ return; -+ } -+ -+ if (page) -+ ext3_block_truncate_page(handle, page, mapping, inode->i_size); -+ -+ down(&EXT3_I(inode)->truncate_sem); -+ ext3_ext_invalidate_cache(&tree); -+ -+ /* -+ * TODO: optimization is possible here -+ * probably we need not scaning at all, -+ * because page truncation is enough -+ */ -+ if (ext3_orphan_add(handle, inode)) -+ goto out_stop; -+ -+ /* we have to know where to truncate from in crash case */ -+ EXT3_I(inode)->i_disksize = inode->i_size; -+ ext3_mark_inode_dirty(handle, inode); -+ -+ last_block = (inode->i_size + sb->s_blocksize - 1) >> -+ EXT3_BLOCK_SIZE_BITS(sb); -+ err = ext3_ext_remove_space(&tree, last_block, EXT_MAX_BLOCK); -+ -+ /* In a multi-transaction truncate, we only make the final -+ * transaction synchronous */ -+ if (IS_SYNC(inode)) -+ handle->h_sync = 1; -+ -+out_stop: -+ /* -+ * If this was a simple ftruncate(), and the file will remain alive -+ * then we need to clear up the orphan record which we created above. -+ * However, if this was a real unlink then we were called by -+ * ext3_delete_inode(), and we allow that function to clean up the -+ * orphan info for us. -+ */ -+ if (inode->i_nlink) -+ ext3_orphan_del(handle, inode); -+ -+ up(&EXT3_I(inode)->truncate_sem); -+ ext3_journal_stop(handle); -+} -+ -+/* -+ * this routine calculate max number of blocks we could modify -+ * in order to allocate new block for an inode -+ */ -+int ext3_ext_writepage_trans_blocks(struct inode *inode, int num) -+{ -+ struct ext3_extents_tree tree; -+ int needed; -+ -+ ext3_init_tree_desc(&tree, inode); -+ -+ needed = ext3_ext_calc_credits_for_insert(&tree, NULL); -+ -+ /* caller want to allocate num blocks */ -+ needed *= num; -+ -+#ifdef CONFIG_QUOTA -+ /* -+ * FIXME: real calculation should be here -+ * it depends on blockmap format of qouta file -+ */ -+ needed += 2 * EXT3_SINGLEDATA_TRANS_BLOCKS; -+#endif -+ -+ return needed; -+} -+ -+void ext3_extents_initialize_blockmap(handle_t *handle, struct inode *inode) -+{ -+ struct ext3_extents_tree tree; -+ -+ ext3_init_tree_desc(&tree, inode); -+ ext3_extent_tree_init(handle, &tree); -+} -+ -+int ext3_ext_calc_blockmap_metadata(struct inode *inode, int blocks) -+{ -+ struct ext3_extents_tree tree; -+ -+ ext3_init_tree_desc(&tree, inode); -+ return ext3_ext_calc_metadata_amount(&tree, blocks); -+} -+ -+static int -+ext3_ext_store_extent_cb(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, -+ struct ext3_ext_cache *newex) -+{ -+ struct ext3_extent_buf *buf = (struct ext3_extent_buf *) tree->private; -+ -+ if (newex->ec_type != EXT3_EXT_CACHE_EXTENT) -+ return EXT_CONTINUE; -+ -+ if (buf->err < 0) -+ return EXT_BREAK; -+ if (buf->cur - buf->buffer + sizeof(*newex) > buf->buflen) -+ return EXT_BREAK; -+ -+ if (!copy_to_user(buf->cur, newex, sizeof(*newex))) { -+ buf->err++; -+ buf->cur += sizeof(*newex); -+ } else { -+ buf->err = -EFAULT; -+ return EXT_BREAK; -+ } -+ return EXT_CONTINUE; -+} -+ -+static int -+ext3_ext_collect_stats_cb(struct ext3_extents_tree *tree, -+ struct ext3_ext_path *path, -+ struct ext3_ext_cache *ex) -+{ -+ struct ext3_extent_tree_stats *buf = -+ (struct ext3_extent_tree_stats *) tree->private; -+ int depth; -+ -+ if (ex->ec_type != EXT3_EXT_CACHE_EXTENT) -+ return EXT_CONTINUE; -+ -+ depth = EXT_DEPTH(tree); -+ buf->extents_num++; -+ if (path[depth].p_ext == EXT_FIRST_EXTENT(path[depth].p_hdr)) -+ buf->leaf_num++; -+ return EXT_CONTINUE; -+} -+ -+int ext3_ext_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, -+ unsigned long arg) -+{ -+ int err = 0; -+ -+ if (!(EXT3_I(inode)->i_flags & EXT3_EXTENTS_FL)) -+ return -EINVAL; -+ -+ if (cmd == EXT3_IOC_GET_EXTENTS) { -+ struct ext3_extent_buf buf; -+ struct ext3_extents_tree tree; -+ -+ if (copy_from_user(&buf, (void *) arg, sizeof(buf))) -+ return -EFAULT; -+ -+ ext3_init_tree_desc(&tree, inode); -+ buf.cur = buf.buffer; -+ buf.err = 0; -+ tree.private = &buf; -+ down(&EXT3_I(inode)->truncate_sem); -+ err = ext3_ext_walk_space(&tree, buf.start, EXT_MAX_BLOCK, -+ ext3_ext_store_extent_cb); -+ up(&EXT3_I(inode)->truncate_sem); -+ if (err == 0) -+ err = buf.err; -+ } else if (cmd == EXT3_IOC_GET_TREE_STATS) { -+ struct ext3_extent_tree_stats buf; -+ struct ext3_extents_tree tree; -+ -+ ext3_init_tree_desc(&tree, inode); -+ down(&EXT3_I(inode)->truncate_sem); -+ buf.depth = EXT_DEPTH(&tree); -+ buf.extents_num = 0; -+ buf.leaf_num = 0; -+ tree.private = &buf; -+ err = ext3_ext_walk_space(&tree, 0, EXT_MAX_BLOCK, -+ ext3_ext_collect_stats_cb); -+ up(&EXT3_I(inode)->truncate_sem); -+ if (!err) -+ err = copy_to_user((void *) arg, &buf, sizeof(buf)); -+ } else if (cmd == EXT3_IOC_GET_TREE_DEPTH) { -+ struct ext3_extents_tree tree; -+ ext3_init_tree_desc(&tree, inode); -+ down(&EXT3_I(inode)->truncate_sem); -+ err = EXT_DEPTH(&tree); -+ up(&EXT3_I(inode)->truncate_sem); -+ } -+ -+ return err; -+} -+ -+EXPORT_SYMBOL(ext3_init_tree_desc); -+EXPORT_SYMBOL(ext3_mark_inode_dirty); -+EXPORT_SYMBOL(ext3_ext_invalidate_cache); -+EXPORT_SYMBOL(ext3_ext_insert_extent); -+EXPORT_SYMBOL(ext3_ext_walk_space); -+EXPORT_SYMBOL(ext3_ext_find_goal); -+EXPORT_SYMBOL(ext3_ext_calc_credits_for_insert); -Index: linux-2.6.12-rc6/fs/ext3/ialloc.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/ialloc.c 2005-06-14 16:31:08.634433030 +0200 -+++ linux-2.6.12-rc6/fs/ext3/ialloc.c 2005-06-14 16:31:25.846346882 +0200 -@@ -598,7 +598,7 @@ - ei->i_dir_start_lookup = 0; - ei->i_disksize = 0; - -- ei->i_flags = EXT3_I(dir)->i_flags & ~EXT3_INDEX_FL; -+ ei->i_flags = EXT3_I(dir)->i_flags & ~(EXT3_INDEX_FL|EXT3_EXTENTS_FL); - if (S_ISLNK(mode)) - ei->i_flags &= ~(EXT3_IMMUTABLE_FL|EXT3_APPEND_FL); - /* dirsync only applies to directories */ -@@ -639,6 +639,18 @@ - DQUOT_FREE_INODE(inode); - goto fail2; - } -+ if (test_opt(sb, EXTENTS) && S_ISREG(inode->i_mode)) { -+ EXT3_I(inode)->i_flags |= EXT3_EXTENTS_FL; -+ ext3_extents_initialize_blockmap(handle, inode); -+ if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_EXTENTS)) { -+ err = ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); -+ if (err) goto fail; -+ EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_EXTENTS); -+ BUFFER_TRACE(EXT3_SB(sb)->s_sbh, "call ext3_journal_dirty_metadata"); -+ err = ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); -+ } -+ } -+ - err = ext3_mark_inode_dirty(handle, inode); - if (err) { - ext3_std_error(sb, err); -Index: linux-2.6.12-rc6/fs/ext3/inode.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/inode.c 2005-06-14 16:31:09.701815830 +0200 -+++ linux-2.6.12-rc6/fs/ext3/inode.c 2005-06-14 16:31:25.861971882 +0200 -@@ -40,7 +40,7 @@ - #include "iopen.h" - #include "acl.h" - --static int ext3_writepage_trans_blocks(struct inode *inode); -+int ext3_writepage_trans_blocks(struct inode *inode); - - /* - * Test whether an inode is a fast symlink. -@@ -784,6 +784,17 @@ - return err; - } - -+static inline int -+ext3_get_block_wrap(handle_t *handle, struct inode *inode, long block, -+ struct buffer_head *bh, int create, int extend_disksize) -+{ -+ if (EXT3_I(inode)->i_flags & EXT3_EXTENTS_FL) -+ return ext3_ext_get_block(handle, inode, block, bh, create, -+ extend_disksize); -+ return ext3_get_block_handle(handle, inode, block, bh, create, -+ extend_disksize); -+} -+ - static int ext3_get_block(struct inode *inode, sector_t iblock, - struct buffer_head *bh_result, int create) - { -@@ -794,8 +805,8 @@ - handle = ext3_journal_current_handle(); - J_ASSERT(handle != 0); - } -- ret = ext3_get_block_handle(handle, inode, iblock, -- bh_result, create, 1); -+ ret = ext3_get_block_wrap(handle, inode, iblock, -+ bh_result, create, 1); - return ret; - } - -@@ -839,7 +850,7 @@ - - get_block: - if (ret == 0) -- ret = ext3_get_block_handle(handle, inode, iblock, -+ ret = ext3_get_block_wrap(handle, inode, iblock, - bh_result, create, 0); - bh_result->b_size = (1 << inode->i_blkbits); - return ret; -@@ -859,7 +870,7 @@ - dummy.b_state = 0; - dummy.b_blocknr = -1000; - buffer_trace_init(&dummy.b_history); -- *errp = ext3_get_block_handle(handle, inode, block, &dummy, create, 1); -+ *errp = ext3_get_block_wrap(handle, inode, block, &dummy, create, 1); - if (!*errp && buffer_mapped(&dummy)) { - struct buffer_head *bh; - bh = sb_getblk(inode->i_sb, dummy.b_blocknr); -@@ -1593,7 +1604,7 @@ - * This required during truncate. We need to physically zero the tail end - * of that block so it doesn't yield old data if the file is later grown. - */ --static int ext3_block_truncate_page(handle_t *handle, struct page *page, -+int ext3_block_truncate_page(handle_t *handle, struct page *page, - struct address_space *mapping, loff_t from) - { - unsigned long index = from >> PAGE_CACHE_SHIFT; -@@ -2104,6 +2115,9 @@ - return; - } - -+ if (EXT3_I(inode)->i_flags & EXT3_EXTENTS_FL) -+ return ext3_ext_truncate(inode, page); -+ - handle = start_transaction(inode); - if (IS_ERR(handle)) { - if (page) { -@@ -2850,12 +2864,15 @@ - * block and work out the exact number of indirects which are touched. Pah. - */ - --static int ext3_writepage_trans_blocks(struct inode *inode) -+int ext3_writepage_trans_blocks(struct inode *inode) - { - int bpp = ext3_journal_blocks_per_page(inode); - int indirects = (EXT3_NDIR_BLOCKS % bpp) ? 5 : 3; - int ret; - -+ if (EXT3_I(inode)->i_flags & EXT3_EXTENTS_FL) -+ return ext3_ext_writepage_trans_blocks(inode, bpp); -+ - if (ext3_should_journal_data(inode)) - ret = 3 * (bpp + indirects) + 2; - else -Index: linux-2.6.12-rc6/fs/ext3/Makefile -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/Makefile 2005-06-14 16:31:09.179354899 +0200 -+++ linux-2.6.12-rc6/fs/ext3/Makefile 2005-06-14 16:31:25.872714069 +0200 -@@ -5,7 +5,8 @@ - obj-$(CONFIG_EXT3_FS) += ext3.o - - ext3-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o iopen.o \ -- ioctl.o namei.o super.o symlink.o hash.o resize.o -+ ioctl.o namei.o super.o symlink.o hash.o resize.o \ -+ extents.o - - ext3-$(CONFIG_EXT3_FS_XATTR) += xattr.o xattr_user.o xattr_trusted.o - ext3-$(CONFIG_EXT3_FS_POSIX_ACL) += acl.o -Index: linux-2.6.12-rc6/fs/ext3/super.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/super.c 2005-06-14 16:31:09.950839264 +0200 -+++ linux-2.6.12-rc6/fs/ext3/super.c 2005-06-14 16:31:25.886385944 +0200 -@@ -387,6 +387,7 @@ - struct ext3_super_block *es = sbi->s_es; - int i; - -+ ext3_ext_release(sb); - ext3_xattr_put_super(sb); - journal_destroy(sbi->s_journal); - if (!(sb->s_flags & MS_RDONLY)) { -@@ -451,6 +452,8 @@ - #endif - ei->i_block_alloc_info = NULL; - ei->vfs_inode.i_version = 1; -+ -+ memset(&ei->i_cached_extent, 0, sizeof(ei->i_cached_extent)); - return &ei->vfs_inode; - } - -@@ -593,6 +596,7 @@ - Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, - Opt_ignore, Opt_barrier, Opt_err, Opt_resize, - Opt_iopen, Opt_noiopen, Opt_iopen_nopriv, -+ Opt_extents, Opt_noextents, Opt_extdebug, - }; - - static match_table_t tokens = { -@@ -644,6 +647,9 @@ - {Opt_iopen, "iopen"}, - {Opt_noiopen, "noiopen"}, - {Opt_iopen_nopriv, "iopen_nopriv"}, -+ {Opt_extents, "extents"}, -+ {Opt_noextents, "noextents"}, -+ {Opt_extdebug, "extdebug"}, - {Opt_barrier, "barrier=%u"}, - {Opt_err, NULL}, - {Opt_resize, "resize"}, -@@ -953,6 +958,15 @@ - case Opt_nobh: - set_opt(sbi->s_mount_opt, NOBH); - break; -+ case Opt_extents: -+ set_opt (sbi->s_mount_opt, EXTENTS); -+ break; -+ case Opt_noextents: -+ clear_opt (sbi->s_mount_opt, EXTENTS); -+ break; -+ case Opt_extdebug: -+ set_opt (sbi->s_mount_opt, EXTDEBUG); -+ break; - default: - printk (KERN_ERR - "EXT3-fs: Unrecognized mount option \"%s\" " -@@ -1668,6 +1681,7 @@ - percpu_counter_mod(&sbi->s_dirs_counter, - ext3_count_dirs(sb)); - -+ ext3_ext_init(sb); - lock_kernel(); - return 0; - -Index: linux-2.6.12-rc6/fs/ext3/ioctl.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/ioctl.c 2005-06-14 16:31:08.646151780 +0200 -+++ linux-2.6.12-rc6/fs/ext3/ioctl.c 2005-06-14 16:31:25.897128131 +0200 -@@ -124,6 +124,10 @@ - err = ext3_change_inode_journal_flag(inode, jflag); - return err; - } -+ case EXT3_IOC_GET_EXTENTS: -+ case EXT3_IOC_GET_TREE_STATS: -+ case EXT3_IOC_GET_TREE_DEPTH: -+ return ext3_ext_ioctl(inode, filp, cmd, arg); - case EXT3_IOC_GETVERSION: - case EXT3_IOC_GETVERSION_OLD: - return put_user(inode->i_generation, (int __user *) arg); -Index: linux-2.6.12-rc6/include/linux/ext3_fs.h -=================================================================== ---- linux-2.6.12-rc6.orig/include/linux/ext3_fs.h 2005-06-14 16:31:10.185214261 +0200 -+++ linux-2.6.12-rc6/include/linux/ext3_fs.h 2005-06-14 16:31:52.859041864 +0200 -@@ -186,8 +186,9 @@ - #define EXT3_NOTAIL_FL 0x00008000 /* don't merge file tail */ - #define EXT3_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */ - #define EXT3_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/ -+#define EXT3_EXTENTS_FL 0x00080000 /* Inode uses extents */ - #define EXT3_RESERVED_FL 0x80000000 /* reserved for ext3 lib */ - --#define EXT3_FL_USER_VISIBLE 0x0003DFFF /* User visible flags */ -+#define EXT3_FL_USER_VISIBLE 0x000BDFFF /* User visible flags */ - #define EXT3_FL_USER_MODIFIABLE 0x000380FF /* User modifiable flags */ - -@@ -237,6 +238,9 @@ - #endif - #define EXT3_IOC_GETRSVSZ _IOR('f', 5, long) - #define EXT3_IOC_SETRSVSZ _IOW('f', 6, long) -+#define EXT3_IOC_GET_EXTENTS _IOR('f', 7, long) -+#define EXT3_IOC_GET_TREE_DEPTH _IOR('f', 8, long) -+#define EXT3_IOC_GET_TREE_STATS _IOR('f', 9, long) - - /* - * Structure of an inode on the disk -@@ -360,6 +364,8 @@ - #define EXT3_MOUNT_NOBH 0x40000 /* No bufferheads */ - #define EXT3_MOUNT_IOPEN 0x80000 /* Allow access via iopen */ - #define EXT3_MOUNT_IOPEN_NOPRIV 0x100000/* Make iopen world-readable */ -+#define EXT3_MOUNT_EXTENTS 0x200000/* Extents support */ -+#define EXT3_MOUNT_EXTDEBUG 0x400000/* Extents debug */ - - /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ - #ifndef clear_opt -@@ -548,11 +554,13 @@ - #define EXT3_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */ - #define EXT3_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */ - #define EXT3_FEATURE_INCOMPAT_META_BG 0x0010 -+#define EXT3_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */ - - #define EXT3_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR - #define EXT3_FEATURE_INCOMPAT_SUPP (EXT3_FEATURE_INCOMPAT_FILETYPE| \ - EXT3_FEATURE_INCOMPAT_RECOVER| \ -- EXT3_FEATURE_INCOMPAT_META_BG) -+ EXT3_FEATURE_INCOMPAT_META_BG| \ -+ EXT3_FEATURE_INCOMPAT_EXTENTS) - #define EXT3_FEATURE_RO_COMPAT_SUPP (EXT3_FEATURE_RO_COMPAT_SPARSE_SUPER| \ - EXT3_FEATURE_RO_COMPAT_LARGE_FILE| \ - EXT3_FEATURE_RO_COMPAT_BTREE_DIR) -@@ -759,6 +767,9 @@ - - - /* inode.c */ -+extern int ext3_block_truncate_page(handle_t *, struct page *, -+ struct address_space *, loff_t); -+extern int ext3_writepage_trans_blocks(struct inode *inode); - extern int ext3_forget(handle_t *, int, struct inode *, struct buffer_head *, int); - extern struct buffer_head * ext3_getblk (handle_t *, struct inode *, long, int, int *); - extern struct buffer_head * ext3_bread (handle_t *, struct inode *, int, int, int *); -@@ -828,6 +837,16 @@ - extern struct inode_operations ext3_symlink_inode_operations; - extern struct inode_operations ext3_fast_symlink_inode_operations; - -+/* extents.c */ -+extern int ext3_ext_writepage_trans_blocks(struct inode *, int); -+extern int ext3_ext_get_block(handle_t *, struct inode *, long, -+ struct buffer_head *, int, int); -+extern void ext3_ext_truncate(struct inode *, struct page *); -+extern void ext3_ext_init(struct super_block *); -+extern void ext3_ext_release(struct super_block *); -+extern void ext3_extents_initialize_blockmap(handle_t *, struct inode *); -+extern int ext3_ext_ioctl(struct inode *inode, struct file *filp, -+ unsigned int cmd, unsigned long arg); - - #endif /* __KERNEL__ */ - -Index: linux-2.6.12-rc6/include/linux/ext3_extents.h -=================================================================== ---- linux-2.6.12-rc6.orig/include/linux/ext3_extents.h 2005-06-14 16:31:25.780917195 +0200 -+++ linux-2.6.12-rc6/include/linux/ext3_extents.h 2005-06-14 16:31:25.932284381 +0200 -@@ -0,0 +1,262 @@ -+/* -+ * Copyright (c) 2003, Cluster File Systems, Inc, info@clusterfs.com -+ * Written by Alex Tomas -+ * -+ * 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- -+ */ -+ -+#ifndef _LINUX_EXT3_EXTENTS -+#define _LINUX_EXT3_EXTENTS -+ -+/* -+ * with AGRESSIVE_TEST defined capacity of index/leaf blocks -+ * become very little, so index split, in-depth growing and -+ * other hard changes happens much more often -+ * this is for debug purposes only -+ */ -+#define AGRESSIVE_TEST_ -+ -+/* -+ * if CHECK_BINSEARCH defined, then results of binary search -+ * will be checked by linear search -+ */ -+#define CHECK_BINSEARCH_ -+ -+/* -+ * if EXT_DEBUG is defined you can use 'extdebug' mount option -+ * to get lots of info what's going on -+ */ -+#define EXT_DEBUG_ -+#ifdef EXT_DEBUG -+#define ext_debug(tree,fmt,a...) \ -+do { \ -+ if (test_opt((tree)->inode->i_sb, EXTDEBUG)) \ -+ printk(fmt, ##a); \ -+} while (0); -+#else -+#define ext_debug(tree,fmt,a...) -+#endif -+ -+/* -+ * if EXT_STATS is defined then stats numbers are collected -+ * these number will be displayed at umount time -+ */ -+#define EXT_STATS_ -+ -+ -+#define EXT3_ALLOC_NEEDED 3 /* block bitmap + group desc. + sb */ -+ -+/* -+ * ext3_inode has i_block array (total 60 bytes) -+ * first 4 bytes are used to store: -+ * - tree depth (0 mean there is no tree yet. all extents in the inode) -+ * - number of alive extents in the inode -+ */ -+ -+/* -+ * this is extent on-disk structure -+ * it's used at the bottom of the tree -+ */ -+struct ext3_extent { -+ __u32 ee_block; /* first logical block extent covers */ -+ __u16 ee_len; /* number of blocks covered by extent */ -+ __u16 ee_start_hi; /* high 16 bits of physical block */ -+ __u32 ee_start; /* low 32 bigs of physical block */ -+}; -+ -+/* -+ * this is index on-disk structure -+ * it's used at all the levels, but the bottom -+ */ -+struct ext3_extent_idx { -+ __u32 ei_block; /* index covers logical blocks from 'block' */ -+ __u32 ei_leaf; /* pointer to the physical block of the next * -+ * level. leaf or next index could bet here */ -+ __u16 ei_leaf_hi; /* high 16 bits of physical block */ -+ __u16 ei_unused; -+}; -+ -+/* -+ * each block (leaves and indexes), even inode-stored has header -+ */ -+struct ext3_extent_header { -+ __u16 eh_magic; /* probably will support different formats */ -+ __u16 eh_entries; /* number of valid entries */ -+ __u16 eh_max; /* capacity of store in entries */ -+ __u16 eh_depth; /* has tree real underlaying blocks? */ -+ __u32 eh_generation; /* flags(8 bits) | generation of the tree */ -+}; -+ -+#define EXT3_EXT_MAGIC 0xf30a -+ -+/* -+ * array of ext3_ext_path contains path to some extent -+ * creation/lookup routines use it for traversal/splitting/etc -+ * truncate uses it to simulate recursive walking -+ */ -+struct ext3_ext_path { -+ __u32 p_block; -+ __u16 p_depth; -+ struct ext3_extent *p_ext; -+ struct ext3_extent_idx *p_idx; -+ struct ext3_extent_header *p_hdr; -+ struct buffer_head *p_bh; -+}; -+ -+/* -+ * structure for external API -+ */ -+ -+/* -+ * storage for cached extent -+ */ -+struct ext3_ext_cache { -+ __u32 ec_start; -+ __u32 ec_block; -+ __u32 ec_len; -+ __u32 ec_type; -+}; -+ -+#define EXT3_EXT_CACHE_NO 0 -+#define EXT3_EXT_CACHE_GAP 1 -+#define EXT3_EXT_CACHE_EXTENT 2 -+ -+/* -+ * ext3_extents_tree is used to pass initial information -+ * to top-level extents API -+ */ -+struct ext3_extents_helpers; -+struct ext3_extents_tree { -+ struct inode *inode; /* inode which tree belongs to */ -+ void *root; /* ptr to data top of tree resides at */ -+ void *buffer; /* will be passed as arg to ^^ routines */ -+ int buffer_len; -+ void *private; -+ struct ext3_ext_cache *cex;/* last found extent */ -+ struct ext3_extents_helpers *ops; -+}; -+ -+struct ext3_extents_helpers { -+ int (*get_write_access)(handle_t *h, void *buffer); -+ int (*mark_buffer_dirty)(handle_t *h, void *buffer); -+ int (*mergable)(struct ext3_extent *ex1, struct ext3_extent *ex2); -+ int (*remove_extent_credits)(struct ext3_extents_tree *, -+ struct ext3_extent *, unsigned long, -+ unsigned long); -+ int (*remove_extent)(struct ext3_extents_tree *, -+ struct ext3_extent *, unsigned long, -+ unsigned long); -+ int (*new_block)(handle_t *, struct ext3_extents_tree *, -+ struct ext3_ext_path *, struct ext3_extent *, -+ int *); -+}; -+ -+/* -+ * to be called by ext3_ext_walk_space() -+ * negative retcode - error -+ * positive retcode - signal for ext3_ext_walk_space(), see below -+ * callback must return valid extent (passed or newly created) -+ */ -+typedef int (*ext_prepare_callback)(struct ext3_extents_tree *, -+ struct ext3_ext_path *, -+ struct ext3_ext_cache *); -+ -+#define EXT_CONTINUE 0 -+#define EXT_BREAK 1 -+#define EXT_REPEAT 2 -+ -+ -+#define EXT_MAX_BLOCK 0xffffffff -+ -+ -+#define EXT_FIRST_EXTENT(__hdr__) \ -+ ((struct ext3_extent *) (((char *) (__hdr__)) + \ -+ sizeof(struct ext3_extent_header))) -+#define EXT_FIRST_INDEX(__hdr__) \ -+ ((struct ext3_extent_idx *) (((char *) (__hdr__)) + \ -+ sizeof(struct ext3_extent_header))) -+#define EXT_HAS_FREE_INDEX(__path__) \ -+ ((__path__)->p_hdr->eh_entries < (__path__)->p_hdr->eh_max) -+#define EXT_LAST_EXTENT(__hdr__) \ -+ (EXT_FIRST_EXTENT((__hdr__)) + (__hdr__)->eh_entries - 1) -+#define EXT_LAST_INDEX(__hdr__) \ -+ (EXT_FIRST_INDEX((__hdr__)) + (__hdr__)->eh_entries - 1) -+#define EXT_MAX_EXTENT(__hdr__) \ -+ (EXT_FIRST_EXTENT((__hdr__)) + (__hdr__)->eh_max - 1) -+#define EXT_MAX_INDEX(__hdr__) \ -+ (EXT_FIRST_INDEX((__hdr__)) + (__hdr__)->eh_max - 1) -+#define EXT_HDR_GEN(__hdr__) ((__hdr__)->eh_generation & 0x00ffffff) -+#define EXT_FLAGS(__hdr__) ((__hdr__)->eh_generation >> 24) -+#define EXT_FLAGS_CLR_UNKNOWN 0x7 /* Flags cleared on modification */ -+ -+#define EXT_BLOCK_HDR(__bh__) ((struct ext3_extent_header *)(__bh__)->b_data) -+#define EXT_ROOT_HDR(__tree__) ((struct ext3_extent_header *)(__tree__)->root) -+#define EXT_DEPTH(__tree__) (EXT_ROOT_HDR(__tree__)->eh_depth) -+#define EXT_GENERATION(__tree__) EXT_HDR_GEN(EXT_ROOT_HDR(__tree__)) -+ -+#define EXT_ASSERT(__x__) if (!(__x__)) BUG(); -+ -+#define EXT_CHECK_PATH(tree,path) \ -+{ \ -+ int depth = EXT_DEPTH(tree); \ -+ BUG_ON((unsigned long) (path) < __PAGE_OFFSET); \ -+ BUG_ON((unsigned long) (path)[depth].p_idx < \ -+ __PAGE_OFFSET && (path)[depth].p_idx != NULL); \ -+ BUG_ON((unsigned long) (path)[depth].p_ext < \ -+ __PAGE_OFFSET && (path)[depth].p_ext != NULL); \ -+ BUG_ON((unsigned long) (path)[depth].p_hdr < __PAGE_OFFSET); \ -+ BUG_ON((unsigned long) (path)[depth].p_bh < __PAGE_OFFSET \ -+ && depth != 0); \ -+ BUG_ON((path)[0].p_depth != depth); \ -+} -+ -+ -+/* -+ * this structure is used to gather extents from the tree via ioctl -+ */ -+struct ext3_extent_buf { -+ unsigned long start; -+ int buflen; -+ void *buffer; -+ void *cur; -+ int err; -+}; -+ -+/* -+ * this structure is used to collect stats info about the tree -+ */ -+struct ext3_extent_tree_stats { -+ int depth; -+ int extents_num; -+ int leaf_num; -+}; -+ -+extern void ext3_init_tree_desc(struct ext3_extents_tree *, struct inode *); -+extern int ext3_extent_tree_init(handle_t *, struct ext3_extents_tree *); -+extern int ext3_ext_calc_credits_for_insert(struct ext3_extents_tree *, struct ext3_ext_path *); -+extern int ext3_ext_insert_extent(handle_t *, struct ext3_extents_tree *, struct ext3_ext_path *, struct ext3_extent *); -+extern int ext3_ext_walk_space(struct ext3_extents_tree *, unsigned long, unsigned long, ext_prepare_callback); -+extern int ext3_ext_remove_space(struct ext3_extents_tree *, unsigned long, unsigned long); -+extern struct ext3_ext_path * ext3_ext_find_extent(struct ext3_extents_tree *, int, struct ext3_ext_path *); -+extern int ext3_ext_calc_blockmap_metadata(struct inode *, int); -+ -+static inline void -+ext3_ext_invalidate_cache(struct ext3_extents_tree *tree) -+{ -+ if (tree->cex) -+ tree->cex->ec_type = EXT3_EXT_CACHE_NO; -+} -+ -+ -+#endif /* _LINUX_EXT3_EXTENTS */ -Index: linux-2.6.12-rc6/include/linux/ext3_fs_i.h -=================================================================== ---- linux-2.6.12-rc6.orig/include/linux/ext3_fs_i.h 2005-06-06 17:22:29.000000000 +0200 -+++ linux-2.6.12-rc6/include/linux/ext3_fs_i.h 2005-06-14 16:31:25.941073443 +0200 -@@ -133,6 +133,8 @@ - */ - struct semaphore truncate_sem; - struct inode vfs_inode; -+ -+ __u32 i_cached_extent[4]; - }; - - #endif /* _LINUX_EXT3_FS_I */ diff --git a/ldiskfs/kernel_patches/patches/ext3-external-journal-2.6.12.patch b/ldiskfs/kernel_patches/patches/ext3-external-journal-2.6.12.patch deleted file mode 100644 index bcfdae2..0000000 --- a/ldiskfs/kernel_patches/patches/ext3-external-journal-2.6.12.patch +++ /dev/null @@ -1,148 +0,0 @@ -Signed-off-by: Johann Lombardi - ---- linux-2.6.12.orig/fs/ext3/super.c 2005-06-17 21:48:29.000000000 +0200 -+++ linux-2.6.12/fs/ext3/super.c 2005-11-07 13:37:30.000000000 +0100 -@@ -39,7 +39,8 @@ - #include "xattr.h" - #include "acl.h" - --static int ext3_load_journal(struct super_block *, struct ext3_super_block *); -+static int ext3_load_journal(struct super_block *, struct ext3_super_block *, -+ unsigned long journal_devnum); - static int ext3_create_journal(struct super_block *, struct ext3_super_block *, - int); - static void ext3_commit_super (struct super_block * sb, -@@ -586,7 +587,7 @@ enum { - Opt_nouid32, Opt_check, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov, - Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, - Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, -- Opt_commit, Opt_journal_update, Opt_journal_inum, -+ Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev, - Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, - Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, - Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, -@@ -624,6 +625,7 @@ static match_table_t tokens = { - {Opt_commit, "commit=%u"}, - {Opt_journal_update, "journal=update"}, - {Opt_journal_inum, "journal=%u"}, -+ {Opt_journal_dev, "journal_dev=%u"}, - {Opt_abort, "abort"}, - {Opt_data_journal, "data=journal"}, - {Opt_data_ordered, "data=ordered"}, -@@ -663,8 +665,9 @@ static unsigned long get_sb_block(void * - return sb_block; - } - --static int parse_options (char * options, struct super_block *sb, -- unsigned long * inum, unsigned long *n_blocks_count, int is_remount) -+static int parse_options (char *options, struct super_block *sb, -+ unsigned long *inum, unsigned long *journal_devnum, -+ unsigned long *n_blocks_count, int is_remount) - { - struct ext3_sb_info *sbi = EXT3_SB(sb); - char * p; -@@ -805,6 +808,16 @@ static int parse_options (char * options - return 0; - *inum = option; - break; -+ case Opt_journal_dev: -+ if (is_remount) { -+ printk(KERN_ERR "EXT3-fs: cannot specify " -+ "journal on remount\n"); -+ return 0; -+ } -+ if (match_int(&args[0], &option)) -+ return 0; -+ *journal_devnum = option; -+ break; - case Opt_noload: - set_opt (sbi->s_mount_opt, NOLOAD); - break; -@@ -1250,6 +1263,7 @@ static int ext3_fill_super (struct super - unsigned long logic_sb_block; - unsigned long offset = 0; - unsigned long journal_inum = 0; -+ unsigned long journal_devnum = 0; - unsigned long def_mount_opts; - struct inode *root; - int blocksize; -@@ -1330,7 +1344,8 @@ static int ext3_fill_super (struct super - - set_opt(sbi->s_mount_opt, RESERVATION); - -- if (!parse_options ((char *) data, sb, &journal_inum, NULL, 0)) -+ if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum, -+ NULL, 0)) - goto failed_mount; - - sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | -@@ -1541,7 +1556,7 @@ static int ext3_fill_super (struct super - */ - if (!test_opt(sb, NOLOAD) && - EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) { -- if (ext3_load_journal(sb, es)) -+ if (ext3_load_journal(sb, es, journal_devnum)) - goto failed_mount2; - } else if (journal_inum) { - if (ext3_create_journal(sb, es, journal_inum)) -@@ -1821,15 +1836,24 @@ out_bdev: - return NULL; - } - --static int ext3_load_journal(struct super_block * sb, -- struct ext3_super_block * es) -+static int ext3_load_journal(struct super_block *sb, -+ struct ext3_super_block *es, -+ unsigned long journal_devnum) - { - journal_t *journal; - int journal_inum = le32_to_cpu(es->s_journal_inum); -- dev_t journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); -+ dev_t journal_dev; - int err = 0; - int really_read_only; - -+ if (journal_devnum && -+ journal_devnum != le32_to_cpu(es->s_journal_dev)) { -+ printk(KERN_INFO "EXT3-fs: external journal device major/minor " -+ "numbers have changed\n"); -+ journal_dev = new_decode_dev(journal_devnum); -+ } else -+ journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); -+ - really_read_only = bdev_read_only(sb->s_bdev); - - /* -@@ -1888,6 +1912,16 @@ static int ext3_load_journal(struct supe - - EXT3_SB(sb)->s_journal = journal; - ext3_clear_journal_err(sb, es); -+ -+ if (journal_devnum && -+ journal_devnum != le32_to_cpu(es->s_journal_dev)) { -+ es->s_journal_dev = cpu_to_le32(journal_devnum); -+ sb->s_dirt = 1; -+ -+ /* Make sure we flush the recovery flag to disk. */ -+ ext3_commit_super(sb, es, 1); -+ } -+ - return 0; - } - -@@ -2093,13 +2127,13 @@ static int ext3_remount (struct super_bl - { - struct ext3_super_block * es; - struct ext3_sb_info *sbi = EXT3_SB(sb); -- unsigned long tmp; -+ unsigned long tmp1, tmp2; - unsigned long n_blocks_count = 0; - - /* - * Allow the "check" option to be passed as a remount option. - */ -- if (!parse_options(data, sb, &tmp, &n_blocks_count, 1)) -+ if (!parse_options(data, sb, &tmp1, &tmp2, &n_blocks_count, 1)) - return -EINVAL; - - if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) diff --git a/ldiskfs/kernel_patches/patches/ext3-mballoc2-2.6.12.patch b/ldiskfs/kernel_patches/patches/ext3-mballoc2-2.6.12.patch deleted file mode 100644 index ad07714..0000000 --- a/ldiskfs/kernel_patches/patches/ext3-mballoc2-2.6.12.patch +++ /dev/null @@ -1,3102 +0,0 @@ -Index: linux-2.6.12.6-bull/include/linux/ext3_fs.h -=================================================================== ---- linux-2.6.12.6-bull.orig/include/linux/ext3_fs.h 2006-04-29 20:39:09.000000000 +0400 -+++ linux-2.6.12.6-bull/include/linux/ext3_fs.h 2006-04-29 20:39:10.000000000 +0400 -@@ -57,6 +57,14 @@ struct statfs; - #define ext3_debug(f, a...) do {} while (0) - #endif - -+#define EXT3_MULTIBLOCK_ALLOCATOR 1 -+ -+#define EXT3_MB_HINT_MERGE 1 -+#define EXT3_MB_HINT_RESERVED 2 -+#define EXT3_MB_HINT_METADATA 4 -+#define EXT3_MB_HINT_FIRST 8 -+#define EXT3_MB_HINT_BEST 16 -+ - /* - * Special inodes numbers - */ -@@ -366,6 +374,7 @@ struct ext3_inode { - #define EXT3_MOUNT_IOPEN_NOPRIV 0x100000/* Make iopen world-readable */ - #define EXT3_MOUNT_EXTENTS 0x200000/* Extents support */ - #define EXT3_MOUNT_EXTDEBUG 0x400000/* Extents debug */ -+#define EXT3_MOUNT_MBALLOC 0x800000/* Buddy allocation support */ - - /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ - #ifndef clear_opt -@@ -387,6 +396,14 @@ struct ext3_inode { - #define ext3_find_first_zero_bit ext2_find_first_zero_bit - #define ext3_find_next_zero_bit ext2_find_next_zero_bit - -+#ifndef ext2_find_next_le_bit -+#ifdef __LITTLE_ENDIAN -+#define ext2_find_next_le_bit(addr, size, off) find_next_bit((addr), (size), (off)) -+#else -+#error "mballoc needs a patch for big-endian systems - CFS bug 10634" -+#endif /* __LITTLE_ENDIAN */ -+#endif /* !ext2_find_next_le_bit */ -+ - /* - * Maximal mount counts between two filesystem checks - */ -@@ -727,7 +736,8 @@ extern int ext3_bg_has_super(struct supe - extern unsigned long ext3_bg_num_gdb(struct super_block *sb, int group); - extern int ext3_new_block (handle_t *, struct inode *, unsigned long, int *); - extern void ext3_free_blocks (handle_t *, struct inode *, unsigned long, -- unsigned long); -+ unsigned long, int); -+extern int ext3_new_block_old(handle_t *, struct inode *, unsigned long, int *); - extern void ext3_free_blocks_sb (handle_t *, struct super_block *, - unsigned long, unsigned long, int *); - extern unsigned long ext3_count_free_blocks (struct super_block *); -@@ -848,6 +857,17 @@ extern void ext3_extents_initialize_bloc - extern int ext3_ext_ioctl(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); - -+/* mballoc.c */ -+extern long ext3_mb_stats; -+extern long ext3_mb_max_to_scan; -+extern int ext3_mb_init(struct super_block *, int); -+extern int ext3_mb_release(struct super_block *); -+extern int ext3_mb_new_blocks(handle_t *, struct inode *, unsigned long, int *, int, int *); -+extern int ext3_mb_reserve_blocks(struct super_block *, int); -+extern void ext3_mb_release_blocks(struct super_block *, int); -+int __init init_ext3_proc(void); -+void exit_ext3_proc(void); -+ - #endif /* __KERNEL__ */ - - /* EXT3_IOC_CREATE_INUM at bottom of file (visible to kernel and user). */ -Index: linux-2.6.12.6-bull/include/linux/ext3_fs_sb.h -=================================================================== ---- linux-2.6.12.6-bull.orig/include/linux/ext3_fs_sb.h 2005-08-29 20:55:27.000000000 +0400 -+++ linux-2.6.12.6-bull/include/linux/ext3_fs_sb.h 2006-04-29 20:39:10.000000000 +0400 -@@ -21,8 +21,14 @@ - #include - #include - #include -+#include - #endif - #include -+#include -+ -+struct ext3_buddy_group_blocks; -+struct ext3_mb_history; -+#define EXT3_BB_MAX_BLOCKS - - /* - * third extended-fs super-block data in memory -@@ -78,6 +84,43 @@ struct ext3_sb_info { - char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */ - int s_jquota_fmt; /* Format of quota to use */ - #endif -+ -+ /* for buddy allocator */ -+ struct ext3_group_info ***s_group_info; -+ struct inode *s_buddy_cache; -+ long s_blocks_reserved; -+ spinlock_t s_reserve_lock; -+ struct list_head s_active_transaction; -+ struct list_head s_closed_transaction; -+ struct list_head s_committed_transaction; -+ spinlock_t s_md_lock; -+ tid_t s_last_transaction; -+ int s_mb_factor; -+ unsigned short *s_mb_offsets, *s_mb_maxs; -+ unsigned long s_stripe; -+ -+ /* history to debug policy */ -+ struct ext3_mb_history *s_mb_history; -+ int s_mb_history_cur; -+ int s_mb_history_max; -+ struct proc_dir_entry *s_mb_proc; -+ spinlock_t s_mb_history_lock; -+ -+ /* stats for buddy allocator */ -+ atomic_t s_bal_reqs; /* number of reqs with len > 1 */ -+ atomic_t s_bal_success; /* we found long enough chunks */ -+ atomic_t s_bal_allocated; /* in blocks */ -+ atomic_t s_bal_ex_scanned; /* total extents scanned */ -+ atomic_t s_bal_goals; /* goal hits */ -+ atomic_t s_bal_breaks; /* too long searches */ -+ atomic_t s_bal_2orders; /* 2^order hits */ -+ spinlock_t s_bal_lock; -+ unsigned long s_mb_buddies_generated; -+ unsigned long long s_mb_generation_time; - }; -+ -+#define EXT3_GROUP_INFO(sb, group) \ -+ EXT3_SB(sb)->s_group_info[(group) >> EXT3_DESC_PER_BLOCK_BITS(sb)] \ -+ [(group) & (EXT3_DESC_PER_BLOCK(sb) - 1)] - - #endif /* _LINUX_EXT3_FS_SB */ -Index: linux-2.6.12.6-bull/fs/ext3/super.c -=================================================================== ---- linux-2.6.12.6-bull.orig/fs/ext3/super.c 2006-04-29 20:39:09.000000000 +0400 -+++ linux-2.6.12.6-bull/fs/ext3/super.c 2006-04-29 20:39:10.000000000 +0400 -@@ -387,6 +387,7 @@ static void ext3_put_super (struct super - struct ext3_super_block *es = sbi->s_es; - int i; - -+ ext3_mb_release(sb); - ext3_ext_release(sb); - ext3_xattr_put_super(sb); - journal_destroy(sbi->s_journal); -@@ -597,6 +598,7 @@ enum { - Opt_ignore, Opt_barrier, Opt_err, Opt_resize, - Opt_iopen, Opt_noiopen, Opt_iopen_nopriv, - Opt_extents, Opt_noextents, Opt_extdebug, -+ Opt_mballoc, Opt_nomballoc, Opt_stripe, - }; - - static match_table_t tokens = { -@@ -650,6 +651,9 @@ static match_table_t tokens = { - {Opt_extents, "extents"}, - {Opt_noextents, "noextents"}, - {Opt_extdebug, "extdebug"}, -+ {Opt_mballoc, "mballoc"}, -+ {Opt_nomballoc, "nomballoc"}, -+ {Opt_stripe, "stripe=%u"}, - {Opt_barrier, "barrier=%u"}, - {Opt_err, NULL}, - {Opt_resize, "resize"}, -@@ -965,6 +967,19 @@ clear_qf_name: - case Opt_extdebug: - set_opt (sbi->s_mount_opt, EXTDEBUG); - break; -+ case Opt_mballoc: -+ set_opt(sbi->s_mount_opt, MBALLOC); -+ break; -+ case Opt_nomballoc: -+ clear_opt(sbi->s_mount_opt, MBALLOC); -+ break; -+ case Opt_stripe: -+ if (match_int(&args[0], &option)) -+ return 0; -+ if (option < 0) -+ return 0; -+ sbi->s_stripe = option; -+ break; - default: - printk (KERN_ERR - "EXT3-fs: Unrecognized mount option \"%s\" " -@@ -1670,6 +1675,7 @@ static int ext3_fill_super (struct super - ext3_count_dirs(sb)); - - ext3_ext_init(sb); -+ ext3_mb_init(sb, needs_recovery); - lock_kernel(); - return 0; - -@@ -2549,7 +2555,13 @@ static struct file_system_type ext3_fs_t - - static int __init init_ext3_fs(void) - { -- int err = init_ext3_xattr(); -+ int err; -+ -+ err = init_ext3_proc(); -+ if (err) -+ return err; -+ -+ err = init_ext3_xattr(); - if (err) - return err; - err = init_inodecache(); -@@ -2571,6 +2583,7 @@ static void __exit exit_ext3_fs(void) - unregister_filesystem(&ext3_fs_type); - destroy_inodecache(); - exit_ext3_xattr(); -+ exit_ext3_proc(); - } - - int ext3_prep_san_write(struct inode *inode, long *blocks, -Index: linux-2.6.12.6-bull/fs/ext3/extents.c -=================================================================== ---- linux-2.6.12.6-bull.orig/fs/ext3/extents.c 2006-04-29 20:39:09.000000000 +0400 -+++ linux-2.6.12.6-bull/fs/ext3/extents.c 2006-04-29 20:39:10.000000000 +0400 -@@ -777,7 +777,7 @@ cleanup: - for (i = 0; i < depth; i++) { - if (!ablocks[i]) - continue; -- ext3_free_blocks(handle, tree->inode, ablocks[i], 1); -+ ext3_free_blocks(handle, tree->inode, ablocks[i], 1, 1); - } - } - kfree(ablocks); -@@ -1434,7 +1434,7 @@ int ext3_ext_rm_idx(handle_t *handle, st - path->p_idx->ei_leaf); - bh = sb_find_get_block(tree->inode->i_sb, path->p_idx->ei_leaf); - ext3_forget(handle, 1, tree->inode, bh, path->p_idx->ei_leaf); -- ext3_free_blocks(handle, tree->inode, path->p_idx->ei_leaf, 1); -+ ext3_free_blocks(handle, tree->inode, path->p_idx->ei_leaf, 1, 1); - return err; - } - -@@ -1919,10 +1919,12 @@ ext3_remove_blocks(struct ext3_extents_t - int needed = ext3_remove_blocks_credits(tree, ex, from, to); - handle_t *handle = ext3_journal_start(tree->inode, needed); - struct buffer_head *bh; -- int i; -+ int i, metadata = 0; - - if (IS_ERR(handle)) - return PTR_ERR(handle); -+ if (S_ISDIR(tree->inode->i_mode) || S_ISLNK(tree->inode->i_mode)) -+ metadata = 1; - if (from >= ex->ee_block && to == ex->ee_block + ex->ee_len - 1) { - /* tail removal */ - unsigned long num, start; -@@ -1934,7 +1936,7 @@ ext3_remove_blocks(struct ext3_extents_t - bh = sb_find_get_block(tree->inode->i_sb, start + i); - ext3_forget(handle, 0, tree->inode, bh, start + i); - } -- ext3_free_blocks(handle, tree->inode, start, num); -+ ext3_free_blocks(handle, tree->inode, start, num, metadata); - } else if (from == ex->ee_block && to <= ex->ee_block + ex->ee_len - 1) { - printk("strange request: removal %lu-%lu from %u:%u\n", - from, to, ex->ee_block, ex->ee_len); -Index: linux-2.6.12.6-bull/fs/ext3/inode.c -=================================================================== ---- linux-2.6.12.6-bull.orig/fs/ext3/inode.c 2006-04-29 20:39:09.000000000 +0400 -+++ linux-2.6.12.6-bull/fs/ext3/inode.c 2006-04-29 20:39:10.000000000 +0400 -@@ -564,7 +564,7 @@ static int ext3_alloc_branch(handle_t *h - ext3_journal_forget(handle, branch[i].bh); - } - for (i = 0; i < keys; i++) -- ext3_free_blocks(handle, inode, le32_to_cpu(branch[i].key), 1); -+ ext3_free_blocks(handle, inode, le32_to_cpu(branch[i].key), 1, 1); - return err; - } - -@@ -1850,7 +1850,7 @@ ext3_clear_blocks(handle_t *handle, stru - } - } - -- ext3_free_blocks(handle, inode, block_to_free, count); -+ ext3_free_blocks(handle, inode, block_to_free, count, 1); - } - - /** -@@ -2023,7 +2023,7 @@ static void ext3_free_branches(handle_t - ext3_journal_test_restart(handle, inode); - } - -- ext3_free_blocks(handle, inode, nr, 1); -+ ext3_free_blocks(handle, inode, nr, 1, 1); - - if (parent_bh) { - /* -Index: linux-2.6.12.6-bull/fs/ext3/balloc.c -=================================================================== ---- linux-2.6.12.6-bull.orig/fs/ext3/balloc.c 2005-08-29 20:55:27.000000000 +0400 -+++ linux-2.6.12.6-bull/fs/ext3/balloc.c 2006-04-29 20:39:10.000000000 +0400 -@@ -79,7 +79,7 @@ struct ext3_group_desc * ext3_get_group_ - * - * Return buffer_head on success or NULL in case of failure. - */ --static struct buffer_head * -+struct buffer_head * - read_block_bitmap(struct super_block *sb, unsigned int block_group) - { - struct ext3_group_desc * desc; -@@ -490,24 +490,6 @@ error_return: - return; - } - --/* Free given blocks, update quota and i_blocks field */ --void ext3_free_blocks(handle_t *handle, struct inode *inode, -- unsigned long block, unsigned long count) --{ -- struct super_block * sb; -- int dquot_freed_blocks; -- -- sb = inode->i_sb; -- if (!sb) { -- printk ("ext3_free_blocks: nonexistent device"); -- return; -- } -- ext3_free_blocks_sb(handle, sb, block, count, &dquot_freed_blocks); -- if (dquot_freed_blocks) -- DQUOT_FREE_BLOCK(inode, dquot_freed_blocks); -- return; --} -- - /* - * For ext3 allocations, we must not reuse any blocks which are - * allocated in the bitmap buffer's "last committed data" copy. This -@@ -1162,7 +1144,7 @@ int ext3_should_retry_alloc(struct super - * bitmap, and then for any free bit if that fails. - * This function also updates quota and i_blocks field. - */ --int ext3_new_block(handle_t *handle, struct inode *inode, -+int ext3_new_block_old(handle_t *handle, struct inode *inode, - unsigned long goal, int *errp) - { - struct buffer_head *bitmap_bh = NULL; -Index: linux-2.6.12.6-bull/fs/ext3/xattr.c -=================================================================== ---- linux-2.6.12.6-bull.orig/fs/ext3/xattr.c 2005-08-29 20:55:27.000000000 +0400 -+++ linux-2.6.12.6-bull/fs/ext3/xattr.c 2006-04-29 20:39:10.000000000 +0400 -@@ -484,7 +484,7 @@ ext3_xattr_release_block(handle_t *handl - ea_bdebug(bh, "refcount now=0; freeing"); - if (ce) - mb_cache_entry_free(ce); -- ext3_free_blocks(handle, inode, bh->b_blocknr, 1); -+ ext3_free_blocks(handle, inode, bh->b_blocknr, 1, 1); - get_bh(bh); - ext3_forget(handle, 1, inode, bh, bh->b_blocknr); - } else { -@@ -804,7 +804,7 @@ inserted: - new_bh = sb_getblk(sb, block); - if (!new_bh) { - getblk_failed: -- ext3_free_blocks(handle, inode, block, 1); -+ ext3_free_blocks(handle, inode, block, 1, 1); - error = -EIO; - goto cleanup; - } -Index: linux-2.6.12.6-bull/fs/ext3/mballoc.c -=================================================================== ---- linux-2.6.12.6-bull.orig/fs/ext3/mballoc.c 2006-04-22 17:31:47.543334750 +0400 -+++ linux-2.6.12.6-bull/fs/ext3/mballoc.c 2006-04-30 01:24:11.000000000 +0400 -@@ -0,0 +1,2725 @@ -+/* -+ * Copyright (c) 2003-2005, Cluster File Systems, Inc, info@clusterfs.com -+ * Written by Alex Tomas -+ * -+ * 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 -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+ -+/* -+ * TODO: -+ * - 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 -+ * - special flag to advice allocator to look for requested + N blocks -+ * this may improve interaction between extents and mballoc -+ * - tree of groups sorted by number of free blocks -+ * - percpu reservation code (hotpath) -+ * - error handling -+ */ -+ -+/* -+ * with AGRESSIVE_CHECK allocator runs consistency checks over -+ * structures. these checks slow things down a lot -+ */ -+#define AGGRESSIVE_CHECK__ -+ -+/* -+ */ -+#define MB_DEBUG__ -+#ifdef MB_DEBUG -+#define mb_debug(fmt,a...) printk(fmt, ##a) -+#else -+#define mb_debug(fmt,a...) -+#endif -+ -+/* -+ * with EXT3_MB_HISTORY mballoc stores last N allocations in memory -+ * and you can monitor it in /proc/fs/ext3//mb_history -+ */ -+#define EXT3_MB_HISTORY -+ -+/* -+ * How long mballoc can look for a best extent (in found extents) -+ */ -+long ext3_mb_max_to_scan = 500; -+ -+/* -+ * How long mballoc must look for a best extent -+ */ -+long ext3_mb_min_to_scan = 30; -+ -+/* -+ * with 'ext3_mb_stats' allocator will collect stats that will be -+ * shown at umount. The collecting costs though! -+ */ -+ -+long ext3_mb_stats = 1; -+ -+/* -+ * for which requests use 2^N search using buddies -+ */ -+long ext3_mb_order2_reqs = 8; -+ -+#ifdef EXT3_BB_MAX_BLOCKS -+#undef EXT3_BB_MAX_BLOCKS -+#endif -+#define EXT3_BB_MAX_BLOCKS 30 -+ -+struct ext3_free_metadata { -+ unsigned short group; -+ unsigned short num; -+ unsigned short blocks[EXT3_BB_MAX_BLOCKS]; -+ struct list_head list; -+}; -+ -+struct ext3_group_info { -+ unsigned long bb_state; -+ unsigned long bb_tid; -+ struct ext3_free_metadata *bb_md_cur; -+ unsigned short bb_first_free; -+ unsigned short bb_free; -+ unsigned short bb_fragments; -+ unsigned short bb_counters[]; -+}; -+ -+ -+#define EXT3_GROUP_INFO_NEED_INIT_BIT 0 -+#define EXT3_GROUP_INFO_LOCKED_BIT 1 -+ -+#define EXT3_MB_GRP_NEED_INIT(grp) \ -+ (test_bit(EXT3_GROUP_INFO_NEED_INIT_BIT, &(grp)->bb_state)) -+ -+struct ext3_free_extent { -+ __u16 fe_start; -+ __u16 fe_len; -+ __u16 fe_group; -+}; -+ -+struct ext3_allocation_context { -+ struct super_block *ac_sb; -+ -+ /* search goals */ -+ struct ext3_free_extent ac_g_ex; -+ -+ /* the best found extent */ -+ struct ext3_free_extent ac_b_ex; -+ -+ /* number of iterations done. we have to track to limit searching */ -+ unsigned long ac_ex_scanned; -+ __u16 ac_groups_scanned; -+ __u16 ac_found; -+ __u16 ac_tail; -+ __u16 ac_buddy; -+ __u8 ac_status; -+ __u8 ac_flags; /* allocation hints */ -+ __u8 ac_criteria; -+ __u8 ac_repeats; -+ __u8 ac_2order; /* if request is to allocate 2^N blocks and -+ * N > 0, the field stores N, otherwise 0 */ -+ -+ struct page *ac_buddy_page; -+ struct page *ac_bitmap_page; -+}; -+ -+#define AC_STATUS_CONTINUE 1 -+#define AC_STATUS_FOUND 2 -+#define AC_STATUS_BREAK 3 -+ -+struct ext3_mb_history { -+ struct ext3_free_extent goal; /* goal allocation */ -+ struct ext3_free_extent result; /* result allocation */ -+ unsigned pid; -+ unsigned ino; -+ __u16 found; /* how many extents have been found */ -+ __u16 groups; /* how many groups have been scanned */ -+ __u16 tail; /* what tail broke some buddy */ -+ __u16 buddy; /* buddy the tail ^^^ broke */ -+ __u8 cr; /* which phase the result extent was found at */ -+ __u8 merged; -+}; -+ -+struct ext3_buddy { -+ struct page *bd_buddy_page; -+ void *bd_buddy; -+ struct page *bd_bitmap_page; -+ void *bd_bitmap; -+ struct ext3_group_info *bd_info; -+ struct super_block *bd_sb; -+ __u16 bd_blkbits; -+ __u16 bd_group; -+}; -+#define EXT3_MB_BITMAP(e3b) ((e3b)->bd_bitmap) -+#define EXT3_MB_BUDDY(e3b) ((e3b)->bd_buddy) -+ -+#ifndef EXT3_MB_HISTORY -+#define ext3_mb_store_history(sb,ino,ac) -+#else -+static void ext3_mb_store_history(struct super_block *, unsigned ino, -+ struct ext3_allocation_context *ac); -+#endif -+ -+#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1) -+ -+static struct proc_dir_entry *proc_root_ext3; -+ -+struct buffer_head * read_block_bitmap(struct super_block *, unsigned int); -+void ext3_mb_poll_new_transaction(struct super_block *, handle_t *); -+void ext3_mb_free_committed_blocks(struct super_block *); -+ -+#if BITS_PER_LONG == 64 -+#define mb_correct_addr_and_bit(bit,addr) \ -+{ \ -+ bit += ((unsigned long) addr & 7UL) << 3; \ -+ addr = (void *) ((unsigned long) addr & ~7UL); \ -+} -+#elif BITS_PER_LONG == 32 -+#define mb_correct_addr_and_bit(bit,addr) \ -+{ \ -+ bit += ((unsigned long) addr & 3UL) << 3; \ -+ addr = (void *) ((unsigned long) addr & ~3UL); \ -+} -+#else -+#error "how many bits you are?!" -+#endif -+ -+static inline int mb_test_bit(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ return ext2_test_bit(bit, addr); -+} -+ -+static inline void mb_set_bit(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ ext2_set_bit(bit, addr); -+} -+ -+static inline void mb_set_bit_atomic(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ ext2_set_bit_atomic(NULL, bit, addr); -+} -+ -+static inline void mb_clear_bit(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ ext2_clear_bit(bit, addr); -+} -+ -+static inline void mb_clear_bit_atomic(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ ext2_clear_bit_atomic(NULL, bit, addr); -+} -+ -+static inline int mb_find_next_zero_bit(void *addr, int max, int start) -+{ -+ int fix; -+#if BITS_PER_LONG == 64 -+ fix = ((unsigned long) addr & 7UL) << 3; -+ addr = (void *) ((unsigned long) addr & ~7UL); -+#elif BITS_PER_LONG == 32 -+ fix = ((unsigned long) addr & 3UL) << 3; -+ addr = (void *) ((unsigned long) addr & ~3UL); -+#else -+#error "how many bits you are?!" -+#endif -+ max += fix; -+ start += fix; -+ return ext2_find_next_zero_bit(addr, max, start) - fix; -+} -+ -+static inline void *mb_find_buddy(struct ext3_buddy *e3b, int order, int *max) -+{ -+ char *bb; -+ -+ J_ASSERT(EXT3_MB_BITMAP(e3b) != EXT3_MB_BUDDY(e3b)); -+ J_ASSERT(max != NULL); -+ -+ if (order > e3b->bd_blkbits + 1) { -+ *max = 0; -+ return NULL; -+ } -+ -+ /* at order 0 we see each particular block */ -+ *max = 1 << (e3b->bd_blkbits + 3); -+ if (order == 0) -+ return EXT3_MB_BITMAP(e3b); -+ -+ bb = EXT3_MB_BUDDY(e3b) + EXT3_SB(e3b->bd_sb)->s_mb_offsets[order]; -+ *max = EXT3_SB(e3b->bd_sb)->s_mb_maxs[order]; -+ -+ return bb; -+} -+ -+#ifdef AGGRESSIVE_CHECK -+ -+static void mb_check_buddy(struct ext3_buddy *e3b) -+{ -+ int order = e3b->bd_blkbits + 1; -+ int max, max2, i, j, k, count; -+ int fragments = 0, fstart; -+ void *buddy, *buddy2; -+ -+ if (!test_opt(e3b->bd_sb, MBALLOC)) -+ return; -+ -+ { -+ static int mb_check_counter = 0; -+ if (mb_check_counter++ % 300 != 0) -+ return; -+ } -+ -+ while (order > 1) { -+ buddy = mb_find_buddy(e3b, order, &max); -+ J_ASSERT(buddy); -+ buddy2 = mb_find_buddy(e3b, order - 1, &max2); -+ J_ASSERT(buddy2); -+ J_ASSERT(buddy != buddy2); -+ J_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)) -+ J_ASSERT(mb_test_bit((i<<1)+1, buddy2)); -+ else if (!mb_test_bit((i << 1) + 1, buddy2)) -+ J_ASSERT(mb_test_bit(i << 1, buddy2)); -+ continue; -+ } -+ -+ /* both bits in buddy2 must be 0 */ -+ J_ASSERT(mb_test_bit(i << 1, buddy2)); -+ J_ASSERT(mb_test_bit((i << 1) + 1, buddy2)); -+ -+ for (j = 0; j < (1 << order); j++) { -+ k = (i * (1 << order)) + j; -+ J_ASSERT(!mb_test_bit(k, EXT3_MB_BITMAP(e3b))); -+ } -+ count++; -+ } -+ J_ASSERT(e3b->bd_info->bb_counters[order] == count); -+ order--; -+ } -+ -+ fstart = -1; -+ buddy = mb_find_buddy(e3b, 0, &max); -+ for (i = 0; i < max; i++) { -+ if (!mb_test_bit(i, buddy)) { -+ J_ASSERT(i >= e3b->bd_info->bb_first_free); -+ if (fstart == -1) { -+ fragments++; -+ fstart = i; -+ } -+ continue; -+ } -+ fstart = -1; -+ /* check used bits only */ -+ for (j = 0; j < e3b->bd_blkbits + 1; j++) { -+ buddy2 = mb_find_buddy(e3b, j, &max2); -+ k = i >> j; -+ J_ASSERT(k < max2); -+ J_ASSERT(mb_test_bit(k, buddy2)); -+ } -+ } -+ J_ASSERT(!EXT3_MB_GRP_NEED_INIT(e3b->bd_info)); -+ J_ASSERT(e3b->bd_info->bb_fragments == fragments); -+} -+ -+#else -+#define mb_check_buddy(e3b) -+#endif -+ -+/* find most significant bit */ -+static int inline fmsb(unsigned short word) -+{ -+ int order; -+ -+ if (word > 255) { -+ order = 7; -+ word >>= 8; -+ } else { -+ order = -1; -+ } -+ -+ do { -+ order++; -+ word >>= 1; -+ } while (word != 0); -+ -+ return order; -+} -+ -+static void inline -+ext3_mb_mark_free_simple(struct super_block *sb, void *buddy, unsigned first, -+ int len, struct ext3_group_info *grp) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ unsigned short min, max, chunk, border; -+ -+ mb_debug("mark %u/%u free\n", first, len); -+ J_ASSERT(len < EXT3_BLOCKS_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 = fmsb(len); -+ -+ mb_debug(" %u/%u -> max %u, min %u\n", -+ first & ((2 << sb->s_blocksize_bits) - 1), -+ len, max, min); -+ -+ if (max < min) -+ min = max; -+ chunk = 1 << min; -+ -+ /* mark multiblock chunks only */ -+ grp->bb_counters[min]++; -+ if (min > 0) { -+ mb_debug(" set %u at %u \n", first >> min, -+ sbi->s_mb_offsets[min]); -+ mb_clear_bit(first >> min, buddy + sbi->s_mb_offsets[min]); -+ } -+ -+ len -= chunk; -+ first += chunk; -+ } -+} -+ -+static void -+ext3_mb_generate_buddy(struct super_block *sb, void *buddy, void *bitmap, -+ int group) -+{ -+ struct ext3_group_info *grp = EXT3_GROUP_INFO(sb, group); -+ unsigned short max = EXT3_BLOCKS_PER_GROUP(sb); -+ unsigned short i = 0, first, len; -+ unsigned free = 0, fragments = 0; -+ unsigned long long period = get_cycles(); -+ -+ i = mb_find_next_zero_bit(bitmap, max, 0); -+ grp->bb_first_free = i; -+ while (i < max) { -+ fragments++; -+ first = i; -+ i = ext2_find_next_le_bit(bitmap, max, i); -+ len = i - first; -+ free += len; -+ if (len > 1) -+ ext3_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; -+ -+ /* bb_state shouldn't being modified because all -+ * others waits for init completion on page lock */ -+ clear_bit(EXT3_GROUP_INFO_NEED_INIT_BIT, &grp->bb_state); -+ if (free != grp->bb_free) { -+ printk("EXT3-fs: group %u: %u blocks in bitmap, %u in gd\n", -+ group, free, grp->bb_free); -+ grp->bb_free = free; -+ } -+ -+ period = get_cycles() - period; -+ spin_lock(&EXT3_SB(sb)->s_bal_lock); -+ EXT3_SB(sb)->s_mb_buddies_generated++; -+ EXT3_SB(sb)->s_mb_generation_time += period; -+ spin_unlock(&EXT3_SB(sb)->s_bal_lock); -+} -+ -+static int ext3_mb_init_cache(struct page *page) -+{ -+ int blocksize, blocks_per_page, groups_per_page; -+ int err = 0, i, first_group, first_block; -+ struct super_block *sb; -+ struct buffer_head *bhs; -+ struct buffer_head **bh; -+ struct inode *inode; -+ char *data, *bitmap; -+ -+ mb_debug("init page %lu\n", page->index); -+ -+ inode = page->mapping->host; -+ sb = inode->i_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) { -+ err = -ENOMEM; -+ i = sizeof(struct buffer_head *) * groups_per_page; -+ bh = kmalloc(i, GFP_NOFS); -+ if (bh == NULL) -+ goto out; -+ memset(bh, 0, i); -+ } else -+ bh = &bhs; -+ -+ first_group = page->index * blocks_per_page / 2; -+ -+ /* read all groups the page covers into the cache */ -+ for (i = 0; i < groups_per_page; i++) { -+ struct ext3_group_desc * desc; -+ -+ if (first_group + i >= EXT3_SB(sb)->s_groups_count) -+ break; -+ -+ err = -EIO; -+ desc = ext3_get_group_desc(sb, first_group + i, NULL); -+ if (desc == NULL) -+ goto out; -+ -+ err = -ENOMEM; -+ bh[i] = sb_getblk(sb, le32_to_cpu(desc->bg_block_bitmap)); -+ if (bh[i] == NULL) -+ goto out; -+ -+ if (buffer_uptodate(bh[i])) -+ continue; -+ -+ lock_buffer(bh[i]); -+ if (buffer_uptodate(bh[i])) { -+ unlock_buffer(bh[i]); -+ continue; -+ } -+ -+ get_bh(bh[i]); -+ bh[i]->b_end_io = end_buffer_read_sync; -+ submit_bh(READ, bh[i]); -+ mb_debug("read bitmap for group %u\n", first_group + i); -+ } -+ -+ /* wait for I/O completion */ -+ for (i = 0; i < groups_per_page && bh[i]; i++) -+ wait_on_buffer(bh[i]); -+ -+ err = -EIO; -+ for (i = 0; i < groups_per_page && bh[i]; i++) -+ if (!buffer_uptodate(bh[i])) -+ 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 >= EXT3_SB(sb)->s_groups_count) -+ break; -+ -+ data = page_address(page) + (i * blocksize); -+ bitmap = bh[group - first_group]->b_data; -+ -+ if ((first_block + i) & 1) { -+ /* this is block of buddy */ -+ mb_debug("put buddy for group %u in page %lu/%x\n", -+ group, page->index, i * blocksize); -+ memset(data, 0xff, blocksize); -+ EXT3_GROUP_INFO(sb, group)->bb_fragments = 0; -+ memset(EXT3_GROUP_INFO(sb, group)->bb_counters, 0, -+ sizeof(unsigned short)*(sb->s_blocksize_bits+2)); -+ ext3_mb_generate_buddy(sb, data, bitmap, group); -+ } else { -+ /* this is block of bitmap */ -+ mb_debug("put bitmap for group %u in page %lu/%x\n", -+ group, page->index, i * blocksize); -+ memcpy(data, bitmap, blocksize); -+ } -+ } -+ SetPageUptodate(page); -+ -+out: -+ if (bh) { -+ for (i = 0; i < groups_per_page && bh[i]; i++) -+ brelse(bh[i]); -+ if (bh != &bhs) -+ kfree(bh); -+ } -+ return err; -+} -+ -+static int ext3_mb_load_buddy(struct super_block *sb, int group, -+ struct ext3_buddy *e3b) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct inode *inode = sbi->s_buddy_cache; -+ int blocks_per_page, block, pnum, poff; -+ struct page *page; -+ -+ mb_debug("load group %u\n", group); -+ -+ blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; -+ -+ e3b->bd_blkbits = sb->s_blocksize_bits; -+ e3b->bd_info = EXT3_GROUP_INFO(sb, group); -+ e3b->bd_sb = sb; -+ e3b->bd_group = group; -+ e3b->bd_buddy_page = NULL; -+ e3b->bd_bitmap_page = NULL; -+ -+ 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) -+ 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)) -+ ext3_mb_init_cache(page); -+ unlock_page(page); -+ } -+ } -+ if (page == NULL || !PageUptodate(page)) -+ goto err; -+ e3b->bd_bitmap_page = page; -+ e3b->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)) -+ ext3_mb_init_cache(page); -+ unlock_page(page); -+ } -+ } -+ if (page == NULL || !PageUptodate(page)) -+ goto err; -+ e3b->bd_buddy_page = page; -+ e3b->bd_buddy = page_address(page) + (poff * sb->s_blocksize); -+ mark_page_accessed(page); -+ -+ J_ASSERT(e3b->bd_bitmap_page != NULL); -+ J_ASSERT(e3b->bd_buddy_page != NULL); -+ -+ return 0; -+ -+err: -+ if (e3b->bd_bitmap_page) -+ page_cache_release(e3b->bd_bitmap_page); -+ if (e3b->bd_buddy_page) -+ page_cache_release(e3b->bd_buddy_page); -+ e3b->bd_buddy = NULL; -+ e3b->bd_bitmap = NULL; -+ return -EIO; -+} -+ -+static void ext3_mb_release_desc(struct ext3_buddy *e3b) -+{ -+ if (e3b->bd_bitmap_page) -+ page_cache_release(e3b->bd_bitmap_page); -+ if (e3b->bd_buddy_page) -+ page_cache_release(e3b->bd_buddy_page); -+} -+ -+ -+static inline void -+ext3_lock_group(struct super_block *sb, int group) -+{ -+ bit_spin_lock(EXT3_GROUP_INFO_LOCKED_BIT, -+ &EXT3_GROUP_INFO(sb, group)->bb_state); -+} -+ -+static inline void -+ext3_unlock_group(struct super_block *sb, int group) -+{ -+ bit_spin_unlock(EXT3_GROUP_INFO_LOCKED_BIT, -+ &EXT3_GROUP_INFO(sb, group)->bb_state); -+} -+ -+static int mb_find_order_for_block(struct ext3_buddy *e3b, int block) -+{ -+ int order = 1; -+ void *bb; -+ -+ J_ASSERT(EXT3_MB_BITMAP(e3b) != EXT3_MB_BUDDY(e3b)); -+ J_ASSERT(block < (1 << (e3b->bd_blkbits + 3))); -+ -+ bb = EXT3_MB_BUDDY(e3b); -+ while (order <= e3b->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 << (e3b->bd_blkbits - order); -+ order++; -+ } -+ return 0; -+} -+ -+static inline 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_atomic(cur, bm); -+ cur++; -+ } -+} -+ -+static inline void mb_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: clear whole word at once */ -+ addr = bm + (cur >> 3); -+ *addr = 0xffffffff; -+ cur += 32; -+ continue; -+ } -+ mb_set_bit_atomic(cur, bm); -+ cur++; -+ } -+} -+ -+static int mb_free_blocks(struct ext3_buddy *e3b, int first, int count) -+{ -+ int block = 0, max = 0, order; -+ void *buddy, *buddy2; -+ -+ mb_check_buddy(e3b); -+ -+ e3b->bd_info->bb_free += count; -+ if (first < e3b->bd_info->bb_first_free) -+ e3b->bd_info->bb_first_free = first; -+ -+ /* let's maintain fragments counter */ -+ if (first != 0) -+ block = !mb_test_bit(first - 1, EXT3_MB_BITMAP(e3b)); -+ if (first + count < EXT3_SB(e3b->bd_sb)->s_mb_maxs[0]) -+ max = !mb_test_bit(first + count, EXT3_MB_BITMAP(e3b)); -+ if (block && max) -+ e3b->bd_info->bb_fragments--; -+ else if (!block && !max) -+ e3b->bd_info->bb_fragments++; -+ -+ /* let's maintain buddy itself */ -+ while (count-- > 0) { -+ block = first++; -+ order = 0; -+ -+ J_ASSERT(mb_test_bit(block, EXT3_MB_BITMAP(e3b))); -+ mb_clear_bit(block, EXT3_MB_BITMAP(e3b)); -+ e3b->bd_info->bb_counters[order]++; -+ -+ /* start of the buddy */ -+ buddy = mb_find_buddy(e3b, 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(e3b, 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); -+ } -+ e3b->bd_info->bb_counters[order]--; -+ e3b->bd_info->bb_counters[order]--; -+ -+ block = block >> 1; -+ order++; -+ e3b->bd_info->bb_counters[order]++; -+ -+ mb_clear_bit(block, buddy2); -+ buddy = buddy2; -+ } while (1); -+ } -+ mb_check_buddy(e3b); -+ -+ return 0; -+} -+ -+static int mb_find_extent(struct ext3_buddy *e3b, int order, int block, -+ int needed, struct ext3_free_extent *ex) -+{ -+ int next = block, max, ord; -+ void *buddy; -+ -+ J_ASSERT(ex != NULL); -+ -+ buddy = mb_find_buddy(e3b, order, &max); -+ J_ASSERT(buddy); -+ J_ASSERT(block < max); -+ if (mb_test_bit(block, buddy)) { -+ ex->fe_len = 0; -+ ex->fe_start = 0; -+ ex->fe_group = 0; -+ return 0; -+ } -+ -+ if (likely(order == 0)) { -+ /* find actual order */ -+ order = mb_find_order_for_block(e3b, block); -+ block = block >> order; -+ } -+ -+ ex->fe_len = 1 << order; -+ ex->fe_start = block << order; -+ ex->fe_group = e3b->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(e3b, order, &max))) { -+ -+ if (block + 1 >= max) -+ break; -+ -+ next = (block + 1) * (1 << order); -+ if (mb_test_bit(next, EXT3_MB_BITMAP(e3b))) -+ break; -+ -+ ord = mb_find_order_for_block(e3b, next); -+ -+ order = ord; -+ block = next >> order; -+ ex->fe_len += 1 << order; -+ } -+ -+ J_ASSERT(ex->fe_start + ex->fe_len <= (1 << (e3b->bd_blkbits + 3))); -+ return ex->fe_len; -+} -+ -+static int mb_mark_used(struct ext3_buddy *e3b, struct ext3_free_extent *ex) -+{ -+ int ord, mlen = 0, max = 0, cur; -+ int start = ex->fe_start; -+ int len = ex->fe_len; -+ unsigned ret = 0; -+ int len0 = len; -+ void *buddy; -+ -+ mb_check_buddy(e3b); -+ -+ e3b->bd_info->bb_free -= len; -+ if (e3b->bd_info->bb_first_free == start) -+ e3b->bd_info->bb_first_free += len; -+ -+ /* let's maintain fragments counter */ -+ if (start != 0) -+ mlen = !mb_test_bit(start - 1, EXT3_MB_BITMAP(e3b)); -+ if (start + len < EXT3_SB(e3b->bd_sb)->s_mb_maxs[0]) -+ max = !mb_test_bit(start + len, EXT3_MB_BITMAP(e3b)); -+ if (mlen && max) -+ e3b->bd_info->bb_fragments++; -+ else if (!mlen && !max) -+ e3b->bd_info->bb_fragments--; -+ -+ /* let's maintain buddy itself */ -+ while (len) { -+ ord = mb_find_order_for_block(e3b, start); -+ -+ if (((start >> ord) << ord) == start && len >= (1 << ord)) { -+ /* the whole chunk may be allocated at once! */ -+ mlen = 1 << ord; -+ buddy = mb_find_buddy(e3b, ord, &max); -+ J_ASSERT((start >> ord) < max); -+ mb_set_bit(start >> ord, buddy); -+ e3b->bd_info->bb_counters[ord]--; -+ start += mlen; -+ len -= mlen; -+ J_ASSERT(len >= 0); -+ continue; -+ } -+ -+ /* store for history */ -+ if (ret == 0) -+ ret = len | (ord << 16); -+ -+ /* we have to split large buddy */ -+ J_ASSERT(ord > 0); -+ buddy = mb_find_buddy(e3b, ord, &max); -+ mb_set_bit(start >> ord, buddy); -+ e3b->bd_info->bb_counters[ord]--; -+ -+ ord--; -+ cur = (start >> ord) & ~1U; -+ buddy = mb_find_buddy(e3b, ord, &max); -+ mb_clear_bit(cur, buddy); -+ mb_clear_bit(cur + 1, buddy); -+ e3b->bd_info->bb_counters[ord]++; -+ e3b->bd_info->bb_counters[ord]++; -+ } -+ -+ /* now drop all the bits in bitmap */ -+ mb_set_bits(EXT3_MB_BITMAP(e3b), ex->fe_start, len0); -+ -+ mb_check_buddy(e3b); -+ -+ return ret; -+} -+ -+/* -+ * Must be called under group lock! -+ */ -+static void ext3_mb_use_best_found(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ unsigned long ret; -+ -+ ac->ac_b_ex.fe_len = min(ac->ac_b_ex.fe_len, ac->ac_g_ex.fe_len); -+ ret = mb_mark_used(e3b, &ac->ac_b_ex); -+ -+ ac->ac_status = AC_STATUS_FOUND; -+ ac->ac_tail = ret & 0xffff; -+ ac->ac_buddy = ret >> 16; -+ -+ /* hold in-core structures until allocated -+ * blocks are marked non-free in on-disk bitmap */ -+ ac->ac_buddy_page = e3b->bd_buddy_page; -+ page_cache_get(e3b->bd_buddy_page); -+ ac->ac_bitmap_page = e3b->bd_bitmap_page; -+ page_cache_get(e3b->bd_bitmap_page); -+} -+ -+/* -+ * 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 ext3_mb_measure_extent(struct ext3_allocation_context *ac, -+ struct ext3_free_extent *ex, -+ struct ext3_buddy *e3b) -+{ -+ struct ext3_free_extent *bex = &ac->ac_b_ex; -+ struct ext3_free_extent *gex = &ac->ac_g_ex; -+ -+ J_ASSERT(ex->fe_len > 0); -+ J_ASSERT(ex->fe_len < (1 << ac->ac_sb->s_blocksize_bits) * 8); -+ J_ASSERT(ex->fe_start < (1 << ac->ac_sb->s_blocksize_bits) * 8); -+ -+ ac->ac_found++; -+ -+ /* -+ * The special case - take what you catch first -+ */ -+ if (unlikely(ac->ac_flags & EXT3_MB_HINT_FIRST)) { -+ *bex = *ex; -+ ext3_mb_use_best_found(ac, e3b); -+ return; -+ } -+ -+ /* -+ * Let's check whether the chunk is good enough -+ */ -+ if (ex->fe_len == gex->fe_len) { -+ *bex = *ex; -+ ext3_mb_use_best_found(ac, e3b); -+ 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 */ -+ *bex = *ex; -+ } -+ -+ /* -+ * Let's scan at least few extents and don't pick up a first one -+ */ -+ if (bex->fe_len > gex->fe_len && ac->ac_found > ext3_mb_min_to_scan) -+ ac->ac_status = AC_STATUS_BREAK; -+ -+ /* -+ * We don't want to scan for a whole year -+ */ -+ if (ac->ac_found > ext3_mb_max_to_scan) -+ ac->ac_status = AC_STATUS_BREAK; -+} -+ -+static int ext3_mb_try_best_found(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ struct ext3_free_extent ex = ac->ac_b_ex; -+ int group = ex.fe_group, max, err; -+ -+ J_ASSERT(ex.fe_len > 0); -+ err = ext3_mb_load_buddy(ac->ac_sb, group, e3b); -+ if (err) -+ return err; -+ -+ ext3_lock_group(ac->ac_sb, group); -+ max = mb_find_extent(e3b, 0, ex.fe_start, ex.fe_len, &ex); -+ -+ if (max > 0) { -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ } -+ -+ ext3_unlock_group(ac->ac_sb, group); -+ -+ ext3_mb_release_desc(e3b); -+ -+ return 0; -+} -+ -+static int ext3_mb_find_by_goal(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ int group = ac->ac_g_ex.fe_group, max, err; -+ struct ext3_sb_info *sbi = EXT3_SB(ac->ac_sb); -+ struct ext3_super_block *es = sbi->s_es; -+ struct ext3_free_extent ex; -+ -+ err = ext3_mb_load_buddy(ac->ac_sb, group, e3b); -+ if (err) -+ return err; -+ -+ ext3_lock_group(ac->ac_sb, group); -+ max = mb_find_extent(e3b, 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) { -+ unsigned long start; -+ start = (e3b->bd_group * EXT3_BLOCKS_PER_GROUP(ac->ac_sb) + -+ ex.fe_start + le32_to_cpu(es->s_first_data_block)); -+ if (start % sbi->s_stripe == 0) { -+ ac->ac_found++; -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ } -+ } else if (max >= ac->ac_g_ex.fe_len) { -+ J_ASSERT(ex.fe_len > 0); -+ J_ASSERT(ex.fe_group == ac->ac_g_ex.fe_group); -+ J_ASSERT(ex.fe_start == ac->ac_g_ex.fe_start); -+ ac->ac_found++; -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ } else if (max > 0 && (ac->ac_flags & EXT3_MB_HINT_MERGE)) { -+ /* Sometimes, caller may want to merge even small -+ * number of blocks to an existing extent */ -+ J_ASSERT(ex.fe_len > 0); -+ J_ASSERT(ex.fe_group == ac->ac_g_ex.fe_group); -+ J_ASSERT(ex.fe_start == ac->ac_g_ex.fe_start); -+ ac->ac_found++; -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ } -+ ext3_unlock_group(ac->ac_sb, group); -+ -+ ext3_mb_release_desc(e3b); -+ -+ 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 void ext3_mb_simple_scan_group(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ struct super_block *sb = ac->ac_sb; -+ struct ext3_group_info *grp = e3b->bd_info; -+ void *buddy; -+ int i, k, max; -+ -+ J_ASSERT(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(e3b, i, &max); -+ if (buddy == NULL) { -+ printk(KERN_ALERT "looking for wrong order?\n"); -+ break; -+ } -+ -+ k = mb_find_next_zero_bit(buddy, max, 0); -+ J_ASSERT(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 = e3b->bd_group; -+ -+ ext3_mb_use_best_found(ac, e3b); -+ J_ASSERT(ac->ac_b_ex.fe_len == ac->ac_g_ex.fe_len); -+ -+ if (unlikely(ext3_mb_stats)) -+ atomic_inc(&EXT3_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 void ext3_mb_complex_scan_group(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ struct super_block *sb = ac->ac_sb; -+ void *bitmap = EXT3_MB_BITMAP(e3b); -+ struct ext3_free_extent ex; -+ int i, free; -+ -+ free = e3b->bd_info->bb_free; -+ J_ASSERT(free > 0); -+ -+ i = e3b->bd_info->bb_first_free; -+ -+ while (free && ac->ac_status == AC_STATUS_CONTINUE) { -+ i = mb_find_next_zero_bit(bitmap, sb->s_blocksize * 8, i); -+ if (i >= sb->s_blocksize * 8) { -+ J_ASSERT(free == 0); -+ break; -+ } -+ -+ mb_find_extent(e3b, 0, i, ac->ac_g_ex.fe_len, &ex); -+ J_ASSERT(ex.fe_len > 0); -+ J_ASSERT(free >= ex.fe_len); -+ -+ ext3_mb_measure_extent(ac, &ex, e3b); -+ -+ i += ex.fe_len; -+ free -= ex.fe_len; -+ } -+} -+ -+/* -+ * This is a special case for storages like raid5 -+ * we try to find stripe-aligned chunks for stripe-size requests -+ */ -+static void ext3_mb_scan_aligned(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ struct super_block *sb = ac->ac_sb; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ void *bitmap = EXT3_MB_BITMAP(e3b); -+ struct ext3_free_extent ex; -+ unsigned long i, max; -+ -+ J_ASSERT(sbi->s_stripe != 0); -+ -+ /* find first stripe-aligned block */ -+ i = e3b->bd_group * EXT3_BLOCKS_PER_GROUP(sb) + -+ le32_to_cpu(sbi->s_es->s_first_data_block); -+ i = ((i + sbi->s_stripe - 1) / sbi->s_stripe) * sbi->s_stripe; -+ i = (i - le32_to_cpu(sbi->s_es->s_first_data_block)) % -+ EXT3_BLOCKS_PER_GROUP(sb); -+ -+ while (i < sb->s_blocksize * 8) { -+ if (!mb_test_bit(i, bitmap)) { -+ max = mb_find_extent(e3b, 0, i, sbi->s_stripe, &ex); -+ if (max >= sbi->s_stripe) { -+ ac->ac_found++; -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ break; -+ } -+ } -+ i += sbi->s_stripe; -+ } -+} -+ -+static int ext3_mb_good_group(struct ext3_allocation_context *ac, -+ int group, int cr) -+{ -+ struct ext3_group_info *grp = EXT3_GROUP_INFO(ac->ac_sb, group); -+ unsigned free, fragments, i, bits; -+ -+ J_ASSERT(cr >= 0 && cr < 4); -+ J_ASSERT(!EXT3_MB_GRP_NEED_INIT(grp)); -+ -+ free = grp->bb_free; -+ fragments = grp->bb_fragments; -+ if (free == 0) -+ return 0; -+ if (fragments == 0) -+ return 0; -+ -+ switch (cr) { -+ case 0: -+ J_ASSERT(ac->ac_2order != 0); -+ bits = ac->ac_sb->s_blocksize_bits + 1; -+ for (i = ac->ac_2order; i <= bits; i++) -+ if (grp->bb_counters[i] > 0) -+ return 1; -+ break; -+ 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; -+} -+ -+int ext3_mb_new_blocks(handle_t *handle, struct inode *inode, -+ unsigned long goal, int *len, int flags, int *errp) -+{ -+ struct buffer_head *bitmap_bh = NULL; -+ struct ext3_allocation_context ac; -+ int i, group, block, cr, err = 0; -+ struct ext3_group_desc *gdp; -+ struct ext3_super_block *es; -+ struct buffer_head *gdp_bh; -+ struct ext3_sb_info *sbi; -+ struct super_block *sb; -+ struct ext3_buddy e3b; -+ -+ J_ASSERT(len != NULL); -+ J_ASSERT(*len > 0); -+ -+ sb = inode->i_sb; -+ if (!sb) { -+ printk("ext3_mb_new_nblocks: nonexistent device"); -+ return 0; -+ } -+ -+ if (!test_opt(sb, MBALLOC)) { -+ static int ext3_mballoc_warning = 0; -+ if (ext3_mballoc_warning == 0) { -+ printk(KERN_ERR "EXT3-fs: multiblock request with " -+ "mballoc disabled!\n"); -+ ext3_mballoc_warning++; -+ } -+ *len = 1; -+ err = ext3_new_block_old(handle, inode, goal, errp); -+ return err; -+ } -+ -+ ext3_mb_poll_new_transaction(sb, handle); -+ -+ sbi = EXT3_SB(sb); -+ es = EXT3_SB(sb)->s_es; -+ -+ /* -+ * We can't allocate > group size -+ */ -+ if (*len >= EXT3_BLOCKS_PER_GROUP(sb) - 10) -+ *len = EXT3_BLOCKS_PER_GROUP(sb) - 10; -+ -+ if (!(flags & EXT3_MB_HINT_RESERVED)) { -+ /* someone asks for non-reserved blocks */ -+ BUG_ON(*len > 1); -+ err = ext3_mb_reserve_blocks(sb, 1); -+ if (err) { -+ *errp = err; -+ return 0; -+ } -+ } -+ -+ ac.ac_buddy_page = NULL; -+ ac.ac_bitmap_page = NULL; -+ -+ /* -+ * Check quota for allocation of this blocks. -+ */ -+ while (*len && DQUOT_ALLOC_BLOCK(inode, *len)) -+ *len -= 1; -+ if (*len == 0) { -+ *errp = -EDQUOT; -+ block = 0; -+ goto out; -+ } -+ -+ /* start searching from the goal */ -+ if (goal < le32_to_cpu(es->s_first_data_block) || -+ goal >= le32_to_cpu(es->s_blocks_count)) -+ goal = le32_to_cpu(es->s_first_data_block); -+ group = (goal - le32_to_cpu(es->s_first_data_block)) / -+ EXT3_BLOCKS_PER_GROUP(sb); -+ block = ((goal - le32_to_cpu(es->s_first_data_block)) % -+ EXT3_BLOCKS_PER_GROUP(sb)); -+ -+ /* set up allocation goals */ -+ 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_groups_scanned = 0; -+ ac.ac_ex_scanned = 0; -+ ac.ac_found = 0; -+ ac.ac_sb = inode->i_sb; -+ ac.ac_g_ex.fe_group = group; -+ ac.ac_g_ex.fe_start = block; -+ ac.ac_g_ex.fe_len = *len; -+ ac.ac_flags = flags; -+ ac.ac_2order = 0; -+ ac.ac_criteria = 0; -+ -+ if (*len == 1 && sbi->s_stripe) { -+ /* looks like a metadata, let's use a dirty hack for raid5 -+ * move all metadata in first groups in hope to hit cached -+ * sectors and thus avoid read-modify cycles in raid5 */ -+ ac.ac_g_ex.fe_group = group = 0; -+ } -+ -+ /* probably, the request is for 2^8+ blocks (1/2/3/... MB) */ -+ i = ffs(*len); -+ if (i >= ext3_mb_order2_reqs) { -+ i--; -+ if ((*len & (~(1 << i))) == 0) -+ ac.ac_2order = i; -+ } -+ -+ /* first, try the goal */ -+ err = ext3_mb_find_by_goal(&ac, &e3b); -+ if (err) -+ goto out_err; -+ if (ac.ac_status == AC_STATUS_FOUND) -+ goto found; -+ -+ /* Let's just scan groups to find more-less suitable blocks */ -+ cr = ac.ac_2order ? 0 : 1; -+repeat: -+ for (; cr < 4 && ac.ac_status == AC_STATUS_CONTINUE; cr++) { -+ ac.ac_criteria = cr; -+ for (i = 0; i < EXT3_SB(sb)->s_groups_count; group++, i++) { -+ if (group == EXT3_SB(sb)->s_groups_count) -+ group = 0; -+ -+ if (EXT3_MB_GRP_NEED_INIT(EXT3_GROUP_INFO(sb, group))) { -+ /* we need full data about the group -+ * to make a good selection */ -+ err = ext3_mb_load_buddy(ac.ac_sb, group, &e3b); -+ if (err) -+ goto out_err; -+ ext3_mb_release_desc(&e3b); -+ } -+ -+ /* check is group good for our criteries */ -+ if (!ext3_mb_good_group(&ac, group, cr)) -+ continue; -+ -+ err = ext3_mb_load_buddy(ac.ac_sb, group, &e3b); -+ if (err) -+ goto out_err; -+ -+ ext3_lock_group(sb, group); -+ if (!ext3_mb_good_group(&ac, group, cr)) { -+ /* someone did allocation from this group */ -+ ext3_unlock_group(sb, group); -+ ext3_mb_release_desc(&e3b); -+ continue; -+ } -+ -+ ac.ac_groups_scanned++; -+ if (cr == 0) -+ ext3_mb_simple_scan_group(&ac, &e3b); -+ else if (cr == 1 && *len == sbi->s_stripe) -+ ext3_mb_scan_aligned(&ac, &e3b); -+ else -+ ext3_mb_complex_scan_group(&ac, &e3b); -+ -+ ext3_unlock_group(sb, group); -+ -+ ext3_mb_release_desc(&e3b); -+ -+ 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 & EXT3_MB_HINT_FIRST)) { -+ /* -+ * We've been searching too long. Let's try to allocate -+ * the best chunk we've found so far -+ */ -+ -+ /*if (ac.ac_found > ext3_mb_max_to_scan) -+ printk(KERN_DEBUG "EXT3-fs: too long searching at " -+ "%u (%d/%d)\n", cr, ac.ac_b_ex.fe_len, -+ ac.ac_g_ex.fe_len);*/ -+ ext3_mb_try_best_found(&ac, &e3b); -+ 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 "EXT3-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 |= EXT3_MB_HINT_FIRST; -+ cr = 3; -+ goto repeat; -+ } -+ } -+ -+ if (ac.ac_status != AC_STATUS_FOUND) { -+ /* -+ * We aren't lucky definitely -+ */ -+ DQUOT_FREE_BLOCK(inode, *len); -+ *errp = -ENOSPC; -+ block = 0; -+#if 1 -+ printk(KERN_ERR "EXT3-fs: can't allocate: status %d flags %d\n", -+ ac.ac_status, ac.ac_flags); -+ printk(KERN_ERR "EXT3-fs: goal %d, best found %d/%d/%d cr %d\n", -+ ac.ac_g_ex.fe_len, ac.ac_b_ex.fe_group, -+ ac.ac_b_ex.fe_start, ac.ac_b_ex.fe_len, cr); -+ printk(KERN_ERR "EXT3-fs: %lu block reserved, %d found\n", -+ sbi->s_blocks_reserved, ac.ac_found); -+ printk("EXT3-fs: groups: "); -+ for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) -+ printk("%d: %d ", i, EXT3_GROUP_INFO(sb, i)->bb_free); -+ printk("\n"); -+#endif -+ goto out; -+ } -+ -+found: -+ J_ASSERT(ac.ac_b_ex.fe_len > 0); -+ -+ /* good news - free block(s) have been found. now it's time -+ * to mark block(s) in good old journaled bitmap */ -+ block = ac.ac_b_ex.fe_group * EXT3_BLOCKS_PER_GROUP(sb) -+ + ac.ac_b_ex.fe_start -+ + le32_to_cpu(es->s_first_data_block); -+ -+ /* we made a desicion, now mark found blocks in good old -+ * bitmap to be journaled */ -+ -+ ext3_debug("using block group %d(%d)\n", -+ ac.ac_b_group.group, gdp->bg_free_blocks_count); -+ -+ bitmap_bh = read_block_bitmap(sb, ac.ac_b_ex.fe_group); -+ if (!bitmap_bh) { -+ *errp = -EIO; -+ goto out_err; -+ } -+ -+ err = ext3_journal_get_write_access(handle, bitmap_bh); -+ if (err) { -+ *errp = err; -+ goto out_err; -+ } -+ -+ gdp = ext3_get_group_desc(sb, ac.ac_b_ex.fe_group, &gdp_bh); -+ if (!gdp) { -+ *errp = -EIO; -+ goto out_err; -+ } -+ -+ err = ext3_journal_get_write_access(handle, gdp_bh); -+ if (err) -+ goto out_err; -+ -+ block = ac.ac_b_ex.fe_group * EXT3_BLOCKS_PER_GROUP(sb) -+ + ac.ac_b_ex.fe_start -+ + le32_to_cpu(es->s_first_data_block); -+ -+ if (block == le32_to_cpu(gdp->bg_block_bitmap) || -+ block == le32_to_cpu(gdp->bg_inode_bitmap) || -+ in_range(block, le32_to_cpu(gdp->bg_inode_table), -+ EXT3_SB(sb)->s_itb_per_group)) -+ ext3_error(sb, "ext3_new_block", -+ "Allocating block in system zone - " -+ "block = %u", block); -+#ifdef AGGRESSIVE_CHECK -+ for (i = 0; i < ac.ac_b_ex.fe_len; i++) -+ J_ASSERT(!mb_test_bit(ac.ac_b_ex.fe_start + i, bitmap_bh->b_data)); -+#endif -+ mb_set_bits(bitmap_bh->b_data, ac.ac_b_ex.fe_start, ac.ac_b_ex.fe_len); -+ -+ spin_lock(sb_bgl_lock(sbi, ac.ac_b_ex.fe_group)); -+ gdp->bg_free_blocks_count = -+ cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count) -+ - ac.ac_b_ex.fe_len); -+ spin_unlock(sb_bgl_lock(sbi, ac.ac_b_ex.fe_group)); -+ percpu_counter_mod(&sbi->s_freeblocks_counter, - ac.ac_b_ex.fe_len); -+ -+ err = ext3_journal_dirty_metadata(handle, bitmap_bh); -+ if (err) -+ goto out_err; -+ err = ext3_journal_dirty_metadata(handle, gdp_bh); -+ if (err) -+ goto out_err; -+ -+ sb->s_dirt = 1; -+ *errp = 0; -+ brelse(bitmap_bh); -+ -+ /* drop non-allocated, but dquote'd blocks */ -+ J_ASSERT(*len >= ac.ac_b_ex.fe_len); -+ DQUOT_FREE_BLOCK(inode, *len - ac.ac_b_ex.fe_len); -+ -+ *len = ac.ac_b_ex.fe_len; -+ J_ASSERT(*len > 0); -+ J_ASSERT(block != 0); -+ goto out; -+ -+out_err: -+ /* if we've already allocated something, roll it back */ -+ if (ac.ac_status == AC_STATUS_FOUND) { -+ /* FIXME: free blocks here */ -+ } -+ -+ DQUOT_FREE_BLOCK(inode, *len); -+ brelse(bitmap_bh); -+ *errp = err; -+ block = 0; -+out: -+ if (ac.ac_buddy_page) -+ page_cache_release(ac.ac_buddy_page); -+ if (ac.ac_bitmap_page) -+ page_cache_release(ac.ac_bitmap_page); -+ -+ if (!(flags & EXT3_MB_HINT_RESERVED)) { -+ /* block wasn't reserved before and we reserved it -+ * at the beginning of allocation. it doesn't matter -+ * whether we allocated anything or we failed: time -+ * to release reservation. NOTE: because I expect -+ * any multiblock request from delayed allocation -+ * path only, here is single block always */ -+ ext3_mb_release_blocks(sb, 1); -+ } -+ -+ if (unlikely(ext3_mb_stats) && ac.ac_g_ex.fe_len > 1) { -+ atomic_inc(&sbi->s_bal_reqs); -+ atomic_add(*len, &sbi->s_bal_allocated); -+ if (*len >= ac.ac_g_ex.fe_len) -+ atomic_inc(&sbi->s_bal_success); -+ atomic_add(ac.ac_found, &sbi->s_bal_ex_scanned); -+ if (ac.ac_g_ex.fe_start == ac.ac_b_ex.fe_start && -+ ac.ac_g_ex.fe_group == ac.ac_b_ex.fe_group) -+ atomic_inc(&sbi->s_bal_goals); -+ if (ac.ac_found > ext3_mb_max_to_scan) -+ atomic_inc(&sbi->s_bal_breaks); -+ } -+ -+ ext3_mb_store_history(sb, inode->i_ino, &ac); -+ -+ return block; -+} -+EXPORT_SYMBOL(ext3_mb_new_blocks); -+ -+#ifdef EXT3_MB_HISTORY -+struct ext3_mb_proc_session { -+ struct ext3_mb_history *history; -+ struct super_block *sb; -+ int start; -+ int max; -+}; -+ -+static void *ext3_mb_history_skip_empty(struct ext3_mb_proc_session *s, -+ struct ext3_mb_history *hs, -+ int first) -+{ -+ if (hs == s->history + s->max) -+ hs = s->history; -+ if (!first && hs == s->history + s->start) -+ return NULL; -+ while (hs->goal.fe_len == 0) { -+ hs++; -+ if (hs == s->history + s->max) -+ hs = s->history; -+ if (hs == s->history + s->start) -+ return NULL; -+ } -+ return hs; -+} -+ -+static void *ext3_mb_seq_history_start(struct seq_file *seq, loff_t *pos) -+{ -+ struct ext3_mb_proc_session *s = seq->private; -+ struct ext3_mb_history *hs; -+ int l = *pos; -+ -+ if (l == 0) -+ return SEQ_START_TOKEN; -+ hs = ext3_mb_history_skip_empty(s, s->history + s->start, 1); -+ if (!hs) -+ return NULL; -+ while (--l && (hs = ext3_mb_history_skip_empty(s, ++hs, 0)) != NULL); -+ return hs; -+} -+ -+static void *ext3_mb_seq_history_next(struct seq_file *seq, void *v, loff_t *pos) -+{ -+ struct ext3_mb_proc_session *s = seq->private; -+ struct ext3_mb_history *hs = v; -+ -+ ++*pos; -+ if (v == SEQ_START_TOKEN) -+ return ext3_mb_history_skip_empty(s, s->history + s->start, 1); -+ else -+ return ext3_mb_history_skip_empty(s, ++hs, 0); -+} -+ -+static int ext3_mb_seq_history_show(struct seq_file *seq, void *v) -+{ -+ struct ext3_mb_history *hs = v; -+ char buf[20], buf2[20]; -+ -+ if (v == SEQ_START_TOKEN) { -+ seq_printf(seq, "%-5s %-8s %-17s %-17s %-5s %-5s %-2s %-5s %-5s %-6s\n", -+ "pid", "inode", "goal", "result", "found", "grps", "cr", -+ "merge", "tail", "broken"); -+ return 0; -+ } -+ -+ sprintf(buf, "%u/%u/%u", hs->goal.fe_group, -+ hs->goal.fe_start, hs->goal.fe_len); -+ sprintf(buf2, "%u/%u/%u", hs->result.fe_group, -+ hs->result.fe_start, hs->result.fe_len); -+ seq_printf(seq, "%-5u %-8u %-17s %-17s %-5u %-5u %-2u %-5s %-5u %-6u\n", -+ hs->pid, hs->ino, buf, buf2, hs->found, hs->groups, -+ hs->cr, hs->merged ? "M" : "", hs->tail, -+ hs->buddy ? 1 << hs->buddy : 0); -+ return 0; -+} -+ -+static void ext3_mb_seq_history_stop(struct seq_file *seq, void *v) -+{ -+} -+ -+static struct seq_operations ext3_mb_seq_history_ops = { -+ .start = ext3_mb_seq_history_start, -+ .next = ext3_mb_seq_history_next, -+ .stop = ext3_mb_seq_history_stop, -+ .show = ext3_mb_seq_history_show, -+}; -+ -+static int ext3_mb_seq_history_open(struct inode *inode, struct file *file) -+{ -+ struct super_block *sb = PDE(inode)->data; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct ext3_mb_proc_session *s; -+ int rc, size; -+ -+ s = kmalloc(sizeof(*s), GFP_KERNEL); -+ if (s == NULL) -+ return -EIO; -+ size = sizeof(struct ext3_mb_history) * sbi->s_mb_history_max; -+ s->history = kmalloc(size, GFP_KERNEL); -+ if (s->history == NULL) { -+ kfree(s); -+ return -EIO; -+ } -+ -+ spin_lock(&sbi->s_mb_history_lock); -+ memcpy(s->history, sbi->s_mb_history, size); -+ s->max = sbi->s_mb_history_max; -+ s->start = sbi->s_mb_history_cur % s->max; -+ spin_unlock(&sbi->s_mb_history_lock); -+ -+ rc = seq_open(file, &ext3_mb_seq_history_ops); -+ if (rc == 0) { -+ struct seq_file *m = (struct seq_file *)file->private_data; -+ m->private = s; -+ } else { -+ kfree(s->history); -+ kfree(s); -+ } -+ return rc; -+ -+} -+ -+static int ext3_mb_seq_history_release(struct inode *inode, struct file *file) -+{ -+ struct seq_file *seq = (struct seq_file *)file->private_data; -+ struct ext3_mb_proc_session *s = seq->private; -+ kfree(s->history); -+ kfree(s); -+ return seq_release(inode, file); -+} -+ -+static struct file_operations ext3_mb_seq_history_fops = { -+ .owner = THIS_MODULE, -+ .open = ext3_mb_seq_history_open, -+ .read = seq_read, -+ .llseek = seq_lseek, -+ .release = ext3_mb_seq_history_release, -+}; -+ -+static void *ext3_mb_seq_groups_start(struct seq_file *seq, loff_t *pos) -+{ -+ struct super_block *sb = seq->private; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ long group; -+ -+ if (*pos < 0 || *pos >= sbi->s_groups_count) -+ return NULL; -+ -+ group = *pos + 1; -+ return (void *) group; -+} -+ -+static void *ext3_mb_seq_groups_next(struct seq_file *seq, void *v, loff_t *pos) -+{ -+ struct super_block *sb = seq->private; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ long group; -+ -+ ++*pos; -+ if (*pos < 0 || *pos >= sbi->s_groups_count) -+ return NULL; -+ group = *pos + 1; -+ return (void *) group;; -+} -+ -+static int ext3_mb_seq_groups_show(struct seq_file *seq, void *v) -+{ -+ struct super_block *sb = seq->private; -+ long group = (long) v, i; -+ struct sg { -+ struct ext3_group_info info; -+ unsigned short 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 ext3_group_info); -+ ext3_lock_group(sb, group); -+ memcpy(&sg, EXT3_GROUP_INFO(sb, group), i); -+ ext3_unlock_group(sb, group); -+ -+ if (EXT3_MB_GRP_NEED_INIT(&sg.info)) -+ return 0; -+ -+ seq_printf(seq, "#%-5lu: %-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 ext3_mb_seq_groups_stop(struct seq_file *seq, void *v) -+{ -+} -+ -+static struct seq_operations ext3_mb_seq_groups_ops = { -+ .start = ext3_mb_seq_groups_start, -+ .next = ext3_mb_seq_groups_next, -+ .stop = ext3_mb_seq_groups_stop, -+ .show = ext3_mb_seq_groups_show, -+}; -+ -+static int ext3_mb_seq_groups_open(struct inode *inode, struct file *file) -+{ -+ struct super_block *sb = PDE(inode)->data; -+ int rc; -+ -+ rc = seq_open(file, &ext3_mb_seq_groups_ops); -+ if (rc == 0) { -+ struct seq_file *m = (struct seq_file *)file->private_data; -+ m->private = sb; -+ } -+ return rc; -+ -+} -+ -+static struct file_operations ext3_mb_seq_groups_fops = { -+ .owner = THIS_MODULE, -+ .open = ext3_mb_seq_groups_open, -+ .read = seq_read, -+ .llseek = seq_lseek, -+ .release = seq_release, -+}; -+ -+static void ext3_mb_history_release(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ char name[64]; -+ -+ snprintf(name, sizeof(name) - 1, "%s", bdevname(sb->s_bdev, name)); -+ remove_proc_entry("mb_groups", sbi->s_mb_proc); -+ remove_proc_entry("mb_history", sbi->s_mb_proc); -+ remove_proc_entry(name, proc_root_ext3); -+ -+ if (sbi->s_mb_history) -+ kfree(sbi->s_mb_history); -+} -+ -+static void ext3_mb_history_init(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ char name[64]; -+ int i; -+ -+ snprintf(name, sizeof(name) - 1, "%s", bdevname(sb->s_bdev, name)); -+ sbi->s_mb_proc = proc_mkdir(name, proc_root_ext3); -+ if (sbi->s_mb_proc != NULL) { -+ struct proc_dir_entry *p; -+ p = create_proc_entry("mb_history", S_IRUGO, sbi->s_mb_proc); -+ if (p) { -+ p->proc_fops = &ext3_mb_seq_history_fops; -+ p->data = sb; -+ } -+ p = create_proc_entry("mb_groups", S_IRUGO, sbi->s_mb_proc); -+ if (p) { -+ p->proc_fops = &ext3_mb_seq_groups_fops; -+ p->data = sb; -+ } -+ } -+ -+ sbi->s_mb_history_max = 1000; -+ sbi->s_mb_history_cur = 0; -+ spin_lock_init(&sbi->s_mb_history_lock); -+ i = sbi->s_mb_history_max * sizeof(struct ext3_mb_history); -+ sbi->s_mb_history = kmalloc(i, GFP_KERNEL); -+ memset(sbi->s_mb_history, 0, i); -+ /* if we can't allocate history, then we simple won't use it */ -+} -+ -+static void -+ext3_mb_store_history(struct super_block *sb, unsigned ino, -+ struct ext3_allocation_context *ac) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct ext3_mb_history h; -+ -+ if (likely(sbi->s_mb_history == NULL)) -+ return; -+ -+ h.pid = current->pid; -+ h.ino = ino; -+ h.goal = ac->ac_g_ex; -+ h.result = ac->ac_b_ex; -+ h.found = ac->ac_found; -+ h.cr = ac->ac_criteria; -+ h.groups = ac->ac_groups_scanned; -+ h.tail = ac->ac_tail; -+ h.buddy = ac->ac_buddy; -+ h.merged = 0; -+ if (ac->ac_g_ex.fe_start == ac->ac_b_ex.fe_start && -+ ac->ac_g_ex.fe_group == ac->ac_b_ex.fe_group) -+ h.merged = 1; -+ -+ spin_lock(&sbi->s_mb_history_lock); -+ memcpy(sbi->s_mb_history + sbi->s_mb_history_cur, &h, sizeof(h)); -+ if (++sbi->s_mb_history_cur >= sbi->s_mb_history_max) -+ sbi->s_mb_history_cur = 0; -+ spin_unlock(&sbi->s_mb_history_lock); -+} -+ -+#else -+#define ext3_mb_history_release(sb) -+#define ext3_mb_history_init(sb) -+#endif -+ -+int ext3_mb_init_backend(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ int i, j, len, metalen; -+ int num_meta_group_infos = -+ (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) >> -+ EXT3_DESC_PER_BLOCK_BITS(sb); -+ struct ext3_group_info **meta_group_info; -+ -+ /* 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 = kmalloc(sizeof(*sbi->s_group_info) * -+ num_meta_group_infos, GFP_KERNEL); -+ if (sbi->s_group_info == NULL) { -+ printk(KERN_ERR "EXT3-fs: can't allocate buddy meta group\n"); -+ return -ENOMEM; -+ } -+ sbi->s_buddy_cache = new_inode(sb); -+ if (sbi->s_buddy_cache == NULL) { -+ printk(KERN_ERR "EXT3-fs: can't get new inode\n"); -+ goto err_freesgi; -+ } -+ -+ metalen = sizeof(*meta_group_info) << EXT3_DESC_PER_BLOCK_BITS(sb); -+ for (i = 0; i < num_meta_group_infos; i++) { -+ if ((i + 1) == num_meta_group_infos) -+ metalen = sizeof(*meta_group_info) * -+ (sbi->s_groups_count - -+ (i << EXT3_DESC_PER_BLOCK_BITS(sb))); -+ meta_group_info = kmalloc(metalen, GFP_KERNEL); -+ if (meta_group_info == NULL) { -+ printk(KERN_ERR "EXT3-fs: can't allocate mem for a " -+ "buddy group\n"); -+ goto err_freemeta; -+ } -+ sbi->s_group_info[i] = meta_group_info; -+ } -+ -+ /* -+ * calculate needed size. if change bb_counters size, -+ * don't forget about ext3_mb_generate_buddy() -+ */ -+ len = sizeof(struct ext3_group_info); -+ len += sizeof(unsigned short) * (sb->s_blocksize_bits + 2); -+ for (i = 0; i < sbi->s_groups_count; i++) { -+ struct ext3_group_desc * desc; -+ -+ meta_group_info = -+ sbi->s_group_info[i >> EXT3_DESC_PER_BLOCK_BITS(sb)]; -+ j = i & (EXT3_DESC_PER_BLOCK(sb) - 1); -+ -+ meta_group_info[j] = kmalloc(len, GFP_KERNEL); -+ if (meta_group_info[j] == NULL) { -+ printk(KERN_ERR "EXT3-fs: can't allocate buddy mem\n"); -+ i--; -+ goto err_freebuddy; -+ } -+ desc = ext3_get_group_desc(sb, i, NULL); -+ if (desc == NULL) { -+ printk(KERN_ERR"EXT3-fs: can't read descriptor %u\n",i); -+ goto err_freebuddy; -+ } -+ memset(meta_group_info[j], 0, len); -+ set_bit(EXT3_GROUP_INFO_NEED_INIT_BIT, -+ &meta_group_info[j]->bb_state); -+ meta_group_info[j]->bb_free = -+ le16_to_cpu(desc->bg_free_blocks_count); -+ } -+ -+ return 0; -+ -+err_freebuddy: -+ while (i >= 0) { -+ kfree(EXT3_GROUP_INFO(sb, i)); -+ i--; -+ } -+ i = num_meta_group_infos; -+err_freemeta: -+ while (--i >= 0) -+ kfree(sbi->s_group_info[i]); -+ iput(sbi->s_buddy_cache); -+err_freesgi: -+ kfree(sbi->s_group_info); -+ return -ENOMEM; -+} -+ -+int ext3_mb_init(struct super_block *sb, int needs_recovery) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct inode *root = sb->s_root->d_inode; -+ unsigned i, offset, max; -+ struct dentry *dentry; -+ -+ if (!test_opt(sb, MBALLOC)) -+ return 0; -+ -+ i = (sb->s_blocksize_bits + 2) * sizeof(unsigned short); -+ -+ sbi->s_mb_offsets = kmalloc(i, GFP_KERNEL); -+ if (sbi->s_mb_offsets == NULL) { -+ clear_opt(sbi->s_mount_opt, MBALLOC); -+ return -ENOMEM; -+ } -+ sbi->s_mb_maxs = kmalloc(i, GFP_KERNEL); -+ if (sbi->s_mb_maxs == NULL) { -+ clear_opt(sbi->s_mount_opt, MBALLOC); -+ kfree(sbi->s_mb_maxs); -+ return -ENOMEM; -+ } -+ -+ /* 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); -+ -+ /* init file for buddy data */ -+ if ((i = ext3_mb_init_backend(sb))) { -+ clear_opt(sbi->s_mount_opt, MBALLOC); -+ kfree(sbi->s_mb_offsets); -+ kfree(sbi->s_mb_maxs); -+ return i; -+ } -+ -+ spin_lock_init(&sbi->s_reserve_lock); -+ spin_lock_init(&sbi->s_md_lock); -+ INIT_LIST_HEAD(&sbi->s_active_transaction); -+ INIT_LIST_HEAD(&sbi->s_closed_transaction); -+ INIT_LIST_HEAD(&sbi->s_committed_transaction); -+ spin_lock_init(&sbi->s_bal_lock); -+ -+ /* remove old on-disk buddy file */ -+ down(&root->i_sem); -+ dentry = lookup_one_len(".buddy", sb->s_root, strlen(".buddy")); -+ if (dentry->d_inode != NULL) { -+ i = vfs_unlink(root, dentry); -+ if (i != 0) -+ printk("EXT3-fs: can't remove .buddy file: %d\n", i); -+ } -+ dput(dentry); -+ up(&root->i_sem); -+ -+ ext3_mb_history_init(sb); -+ -+ printk("EXT3-fs: mballoc enabled\n"); -+ return 0; -+} -+ -+int ext3_mb_release(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ int i, num_meta_group_infos; -+ -+ if (!test_opt(sb, MBALLOC)) -+ return 0; -+ -+ /* release freed, non-committed blocks */ -+ spin_lock(&sbi->s_md_lock); -+ list_splice_init(&sbi->s_closed_transaction, -+ &sbi->s_committed_transaction); -+ list_splice_init(&sbi->s_active_transaction, -+ &sbi->s_committed_transaction); -+ spin_unlock(&sbi->s_md_lock); -+ ext3_mb_free_committed_blocks(sb); -+ -+ if (sbi->s_group_info) { -+ for (i = 0; i < sbi->s_groups_count; i++) -+ kfree(EXT3_GROUP_INFO(sb, i)); -+ num_meta_group_infos = (sbi->s_groups_count + -+ EXT3_DESC_PER_BLOCK(sb) - 1) >> -+ EXT3_DESC_PER_BLOCK_BITS(sb); -+ for (i = 0; i < num_meta_group_infos; i++) -+ kfree(sbi->s_group_info[i]); -+ kfree(sbi->s_group_info); -+ } -+ if (sbi->s_mb_offsets) -+ kfree(sbi->s_mb_offsets); -+ if (sbi->s_mb_maxs) -+ kfree(sbi->s_mb_maxs); -+ if (sbi->s_buddy_cache) -+ iput(sbi->s_buddy_cache); -+ if (sbi->s_blocks_reserved) -+ printk("ext3-fs: %ld blocks being reserved at umount!\n", -+ sbi->s_blocks_reserved); -+ if (ext3_mb_stats) { -+ printk("EXT3-fs: mballoc: %u blocks %u reqs (%u success)\n", -+ atomic_read(&sbi->s_bal_allocated), -+ atomic_read(&sbi->s_bal_reqs), -+ atomic_read(&sbi->s_bal_success)); -+ printk("EXT3-fs: mballoc: %u extents scanned, %u goal hits, " -+ "%u 2^N hits, %u breaks\n", -+ 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)); -+ printk("EXT3-fs: mballoc: %lu generated and it took %Lu\n", -+ sbi->s_mb_buddies_generated++, -+ sbi->s_mb_generation_time); -+ } -+ -+ ext3_mb_history_release(sb); -+ -+ return 0; -+} -+ -+void ext3_mb_free_committed_blocks(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ int err, i, count = 0, count2 = 0; -+ struct ext3_free_metadata *md; -+ struct ext3_buddy e3b; -+ -+ if (list_empty(&sbi->s_committed_transaction)) -+ return; -+ -+ /* there is committed blocks to be freed yet */ -+ do { -+ /* get next array of blocks */ -+ md = NULL; -+ spin_lock(&sbi->s_md_lock); -+ if (!list_empty(&sbi->s_committed_transaction)) { -+ md = list_entry(sbi->s_committed_transaction.next, -+ struct ext3_free_metadata, list); -+ list_del(&md->list); -+ } -+ spin_unlock(&sbi->s_md_lock); -+ -+ if (md == NULL) -+ break; -+ -+ mb_debug("gonna free %u blocks in group %u (0x%p):", -+ md->num, md->group, md); -+ -+ err = ext3_mb_load_buddy(sb, md->group, &e3b); -+ /* we expect to find existing buddy because it's pinned */ -+ BUG_ON(err != 0); -+ -+ /* there are blocks to put in buddy to make them really free */ -+ count += md->num; -+ count2++; -+ ext3_lock_group(sb, md->group); -+ for (i = 0; i < md->num; i++) { -+ mb_debug(" %u", md->blocks[i]); -+ mb_free_blocks(&e3b, md->blocks[i], 1); -+ } -+ mb_debug("\n"); -+ ext3_unlock_group(sb, md->group); -+ -+ /* balance refcounts from ext3_mb_free_metadata() */ -+ page_cache_release(e3b.bd_buddy_page); -+ page_cache_release(e3b.bd_bitmap_page); -+ -+ kfree(md); -+ ext3_mb_release_desc(&e3b); -+ -+ } while (md); -+ mb_debug("freed %u blocks in %u structures\n", count, count2); -+} -+ -+void ext3_mb_poll_new_transaction(struct super_block *sb, handle_t *handle) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ -+ if (sbi->s_last_transaction == handle->h_transaction->t_tid) -+ return; -+ -+ /* new transaction! time to close last one and free blocks for -+ * committed transaction. we know that only transaction can be -+ * active, so previos transaction can be being logged and we -+ * know that transaction before previous is known to be already -+ * logged. this means that now we may free blocks freed in all -+ * transactions before previous one. hope I'm clear enough ... */ -+ -+ spin_lock(&sbi->s_md_lock); -+ if (sbi->s_last_transaction != handle->h_transaction->t_tid) { -+ mb_debug("new transaction %lu, old %lu\n", -+ (unsigned long) handle->h_transaction->t_tid, -+ (unsigned long) sbi->s_last_transaction); -+ list_splice_init(&sbi->s_closed_transaction, -+ &sbi->s_committed_transaction); -+ list_splice_init(&sbi->s_active_transaction, -+ &sbi->s_closed_transaction); -+ sbi->s_last_transaction = handle->h_transaction->t_tid; -+ } -+ spin_unlock(&sbi->s_md_lock); -+ -+ ext3_mb_free_committed_blocks(sb); -+} -+ -+int ext3_mb_free_metadata(handle_t *handle, struct ext3_buddy *e3b, -+ int group, int block, int count) -+{ -+ struct ext3_group_info *db = e3b->bd_info; -+ struct super_block *sb = e3b->bd_sb; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct ext3_free_metadata *md; -+ int i; -+ -+ J_ASSERT(e3b->bd_bitmap_page != NULL); -+ J_ASSERT(e3b->bd_buddy_page != NULL); -+ -+ ext3_lock_group(sb, group); -+ for (i = 0; i < count; i++) { -+ md = db->bb_md_cur; -+ if (md && db->bb_tid != handle->h_transaction->t_tid) { -+ db->bb_md_cur = NULL; -+ md = NULL; -+ } -+ -+ if (md == NULL) { -+ ext3_unlock_group(sb, group); -+ md = kmalloc(sizeof(*md), GFP_KERNEL); -+ if (md == NULL) -+ return -ENOMEM; -+ md->num = 0; -+ md->group = group; -+ -+ ext3_lock_group(sb, group); -+ if (db->bb_md_cur == NULL) { -+ spin_lock(&sbi->s_md_lock); -+ list_add(&md->list, &sbi->s_active_transaction); -+ spin_unlock(&sbi->s_md_lock); -+ /* protect buddy cache from being freed, -+ * otherwise we'll refresh it from -+ * on-disk bitmap and lose not-yet-available -+ * blocks */ -+ page_cache_get(e3b->bd_buddy_page); -+ page_cache_get(e3b->bd_bitmap_page); -+ db->bb_md_cur = md; -+ db->bb_tid = handle->h_transaction->t_tid; -+ mb_debug("new md 0x%p for group %u\n", -+ md, md->group); -+ } else { -+ kfree(md); -+ md = db->bb_md_cur; -+ } -+ } -+ -+ BUG_ON(md->num >= EXT3_BB_MAX_BLOCKS); -+ md->blocks[md->num] = block + i; -+ md->num++; -+ if (md->num == EXT3_BB_MAX_BLOCKS) { -+ /* no more space, put full container on a sb's list */ -+ db->bb_md_cur = NULL; -+ } -+ } -+ ext3_unlock_group(sb, group); -+ return 0; -+} -+ -+void ext3_mb_free_blocks(handle_t *handle, struct inode *inode, -+ unsigned long block, unsigned long count, -+ int metadata, int *freed) -+{ -+ struct buffer_head *bitmap_bh = NULL; -+ struct ext3_group_desc *gdp; -+ struct ext3_super_block *es; -+ unsigned long bit, overflow; -+ struct buffer_head *gd_bh; -+ unsigned long block_group; -+ struct ext3_sb_info *sbi; -+ struct super_block *sb; -+ struct ext3_buddy e3b; -+ int err = 0, ret; -+ -+ *freed = 0; -+ sb = inode->i_sb; -+ if (!sb) { -+ printk ("ext3_free_blocks: nonexistent device"); -+ return; -+ } -+ -+ ext3_mb_poll_new_transaction(sb, handle); -+ -+ sbi = EXT3_SB(sb); -+ es = EXT3_SB(sb)->s_es; -+ if (block < le32_to_cpu(es->s_first_data_block) || -+ block + count < block || -+ block + count > le32_to_cpu(es->s_blocks_count)) { -+ ext3_error (sb, "ext3_free_blocks", -+ "Freeing blocks not in datazone - " -+ "block = %lu, count = %lu", block, count); -+ goto error_return; -+ } -+ -+ ext3_debug("freeing block %lu\n", block); -+ -+do_more: -+ overflow = 0; -+ block_group = (block - le32_to_cpu(es->s_first_data_block)) / -+ EXT3_BLOCKS_PER_GROUP(sb); -+ bit = (block - le32_to_cpu(es->s_first_data_block)) % -+ EXT3_BLOCKS_PER_GROUP(sb); -+ /* -+ * Check to see if we are freeing blocks across a group -+ * boundary. -+ */ -+ if (bit + count > EXT3_BLOCKS_PER_GROUP(sb)) { -+ overflow = bit + count - EXT3_BLOCKS_PER_GROUP(sb); -+ count -= overflow; -+ } -+ brelse(bitmap_bh); -+ bitmap_bh = read_block_bitmap(sb, block_group); -+ if (!bitmap_bh) -+ goto error_return; -+ gdp = ext3_get_group_desc (sb, block_group, &gd_bh); -+ if (!gdp) -+ goto error_return; -+ -+ if (in_range (le32_to_cpu(gdp->bg_block_bitmap), block, count) || -+ in_range (le32_to_cpu(gdp->bg_inode_bitmap), block, count) || -+ in_range (block, le32_to_cpu(gdp->bg_inode_table), -+ EXT3_SB(sb)->s_itb_per_group) || -+ in_range (block + count - 1, le32_to_cpu(gdp->bg_inode_table), -+ EXT3_SB(sb)->s_itb_per_group)) -+ ext3_error (sb, "ext3_free_blocks", -+ "Freeing blocks in system zones - " -+ "Block = %lu, count = %lu", -+ block, count); -+ -+ BUFFER_TRACE(bitmap_bh, "getting write access"); -+ err = ext3_journal_get_write_access(handle, bitmap_bh); -+ if (err) -+ goto error_return; -+ -+ /* -+ * We are about to modify some metadata. Call the journal APIs -+ * to unshare ->b_data if a currently-committing transaction is -+ * using it -+ */ -+ BUFFER_TRACE(gd_bh, "get_write_access"); -+ err = ext3_journal_get_write_access(handle, gd_bh); -+ if (err) -+ goto error_return; -+ -+ err = ext3_mb_load_buddy(sb, block_group, &e3b); -+ if (err) -+ goto error_return; -+ -+#ifdef AGGRESSIVE_CHECK -+ { -+ int i; -+ for (i = 0; i < count; i++) -+ J_ASSERT(mb_test_bit(bit + i, bitmap_bh->b_data)); -+ } -+#endif -+ mb_clear_bits(bitmap_bh->b_data, bit, count); -+ -+ /* We dirtied the bitmap block */ -+ BUFFER_TRACE(bitmap_bh, "dirtied bitmap block"); -+ err = ext3_journal_dirty_metadata(handle, bitmap_bh); -+ -+ if (metadata) { -+ /* blocks being freed are metadata. these blocks shouldn't -+ * be used until this transaction is committed */ -+ ext3_mb_free_metadata(handle, &e3b, block_group, bit, count); -+ } else { -+ ext3_lock_group(sb, block_group); -+ mb_free_blocks(&e3b, bit, count); -+ ext3_unlock_group(sb, block_group); -+ } -+ -+ spin_lock(sb_bgl_lock(sbi, block_group)); -+ gdp->bg_free_blocks_count = -+ cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count) + count); -+ spin_unlock(sb_bgl_lock(sbi, block_group)); -+ percpu_counter_mod(&sbi->s_freeblocks_counter, count); -+ -+ ext3_mb_release_desc(&e3b); -+ -+ *freed = count; -+ -+ /* And the group descriptor block */ -+ BUFFER_TRACE(gd_bh, "dirtied group descriptor block"); -+ ret = ext3_journal_dirty_metadata(handle, gd_bh); -+ if (!err) err = ret; -+ -+ if (overflow && !err) { -+ block += count; -+ count = overflow; -+ goto do_more; -+ } -+ sb->s_dirt = 1; -+error_return: -+ brelse(bitmap_bh); -+ ext3_std_error(sb, err); -+ return; -+} -+ -+int ext3_mb_reserve_blocks(struct super_block *sb, int blocks) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ int free, ret = -ENOSPC; -+ -+ BUG_ON(blocks < 0); -+ spin_lock(&sbi->s_reserve_lock); -+ free = percpu_counter_read_positive(&sbi->s_freeblocks_counter); -+ if (blocks <= free - sbi->s_blocks_reserved) { -+ sbi->s_blocks_reserved += blocks; -+ ret = 0; -+ } -+ spin_unlock(&sbi->s_reserve_lock); -+ return ret; -+} -+ -+void ext3_mb_release_blocks(struct super_block *sb, int blocks) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ -+ BUG_ON(blocks < 0); -+ spin_lock(&sbi->s_reserve_lock); -+ sbi->s_blocks_reserved -= blocks; -+ WARN_ON(sbi->s_blocks_reserved < 0); -+ if (sbi->s_blocks_reserved < 0) -+ sbi->s_blocks_reserved = 0; -+ spin_unlock(&sbi->s_reserve_lock); -+} -+ -+int ext3_new_block(handle_t *handle, struct inode *inode, -+ unsigned long goal, int *errp) -+{ -+ int ret, len; -+ -+ if (!test_opt(inode->i_sb, MBALLOC)) { -+ ret = ext3_new_block_old(handle, inode, goal, errp); -+ goto out; -+ } -+ len = 1; -+ ret = ext3_mb_new_blocks(handle, inode, goal, &len, 0, errp); -+out: -+ return ret; -+} -+ -+ -+void ext3_free_blocks(handle_t *handle, struct inode * inode, -+ unsigned long block, unsigned long count, int metadata) -+{ -+ struct super_block *sb; -+ int freed; -+ -+ sb = inode->i_sb; -+ if (!test_opt(sb, MBALLOC) || !EXT3_SB(sb)->s_group_info) -+ ext3_free_blocks_sb(handle, sb, block, count, &freed); -+ else -+ ext3_mb_free_blocks(handle, inode, block, count, metadata, &freed); -+ if (freed) -+ DQUOT_FREE_BLOCK(inode, freed); -+ return; -+} -+ -+#define EXT3_ROOT "ext3" -+#define EXT3_MB_STATS_NAME "mb_stats" -+#define EXT3_MB_MAX_TO_SCAN_NAME "mb_max_to_scan" -+#define EXT3_MB_MIN_TO_SCAN_NAME "mb_min_to_scan" -+#define EXT3_MB_ORDER2_REQ "mb_order2_req" -+ -+static int ext3_mb_stats_read(char *page, char **start, off_t off, -+ int count, int *eof, void *data) -+{ -+ int len; -+ -+ *eof = 1; -+ if (off != 0) -+ return 0; -+ -+ len = sprintf(page, "%ld\n", ext3_mb_stats); -+ *start = page; -+ return len; -+} -+ -+static int ext3_mb_stats_write(struct file *file, const char *buffer, -+ unsigned long count, void *data) -+{ -+ char str[32]; -+ -+ if (count >= sizeof(str)) { -+ printk(KERN_ERR "EXT3-fs: %s string too long, max %u bytes\n", -+ EXT3_MB_STATS_NAME, (int)sizeof(str)); -+ return -EOVERFLOW; -+ } -+ -+ if (copy_from_user(str, buffer, count)) -+ return -EFAULT; -+ -+ /* Only set to 0 or 1 respectively; zero->0; non-zero->1 */ -+ ext3_mb_stats = (simple_strtol(str, NULL, 0) != 0); -+ return count; -+} -+ -+static int ext3_mb_max_to_scan_read(char *page, char **start, off_t off, -+ int count, int *eof, void *data) -+{ -+ int len; -+ -+ *eof = 1; -+ if (off != 0) -+ return 0; -+ -+ len = sprintf(page, "%ld\n", ext3_mb_max_to_scan); -+ *start = page; -+ return len; -+} -+ -+static int ext3_mb_max_to_scan_write(struct file *file, const char *buffer, -+ unsigned long count, void *data) -+{ -+ char str[32]; -+ long value; -+ -+ if (count >= sizeof(str)) { -+ printk(KERN_ERR "EXT3-fs: %s string too long, max %u bytes\n", -+ EXT3_MB_MAX_TO_SCAN_NAME, (int)sizeof(str)); -+ return -EOVERFLOW; -+ } -+ -+ if (copy_from_user(str, buffer, count)) -+ return -EFAULT; -+ -+ /* Only set to 0 or 1 respectively; zero->0; non-zero->1 */ -+ value = simple_strtol(str, NULL, 0); -+ if (value <= 0) -+ return -ERANGE; -+ -+ ext3_mb_max_to_scan = value; -+ -+ return count; -+} -+ -+static int ext3_mb_min_to_scan_read(char *page, char **start, off_t off, -+ int count, int *eof, void *data) -+{ -+ int len; -+ -+ *eof = 1; -+ if (off != 0) -+ return 0; -+ -+ len = sprintf(page, "%ld\n", ext3_mb_min_to_scan); -+ *start = page; -+ return len; -+} -+ -+static int ext3_mb_min_to_scan_write(struct file *file, const char *buffer, -+ unsigned long count, void *data) -+{ -+ char str[32]; -+ long value; -+ -+ if (count >= sizeof(str)) { -+ printk(KERN_ERR "EXT3-fs: %s string too long, max %u bytes\n", -+ EXT3_MB_MIN_TO_SCAN_NAME, (int)sizeof(str)); -+ return -EOVERFLOW; -+ } -+ -+ if (copy_from_user(str, buffer, count)) -+ return -EFAULT; -+ -+ /* Only set to 0 or 1 respectively; zero->0; non-zero->1 */ -+ value = simple_strtol(str, NULL, 0); -+ if (value <= 0) -+ return -ERANGE; -+ -+ ext3_mb_min_to_scan = value; -+ -+ return count; -+} -+ -+static int ext3_mb_order2_req_read(char *page, char **start, off_t off, -+ int count, int *eof, void *data) -+{ -+ int len; -+ -+ *eof = 1; -+ if (off != 0) -+ return 0; -+ -+ len = sprintf(page, "%ld\n", ext3_mb_order2_reqs); -+ *start = page; -+ return len; -+} -+ -+static int ext3_mb_order2_req_write(struct file *file, const char *buffer, -+ unsigned long count, void *data) -+{ -+ char str[32]; -+ long value; -+ -+ if (count >= sizeof(str)) { -+ printk(KERN_ERR "EXT3-fs: %s string too long, max %u bytes\n", -+ EXT3_MB_MIN_TO_SCAN_NAME, (int)sizeof(str)); -+ return -EOVERFLOW; -+ } -+ -+ if (copy_from_user(str, buffer, count)) -+ return -EFAULT; -+ -+ /* Only set to 0 or 1 respectively; zero->0; non-zero->1 */ -+ value = simple_strtol(str, NULL, 0); -+ if (value <= 0) -+ return -ERANGE; -+ -+ ext3_mb_order2_reqs = value; -+ -+ return count; -+} -+ -+int __init init_ext3_proc(void) -+{ -+ struct proc_dir_entry *proc_ext3_mb_stats; -+ struct proc_dir_entry *proc_ext3_mb_max_to_scan; -+ struct proc_dir_entry *proc_ext3_mb_min_to_scan; -+ struct proc_dir_entry *proc_ext3_mb_order2_req; -+ -+ proc_root_ext3 = proc_mkdir(EXT3_ROOT, proc_root_fs); -+ if (proc_root_ext3 == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", EXT3_ROOT); -+ return -EIO; -+ } -+ -+ /* Initialize EXT3_MB_STATS_NAME */ -+ proc_ext3_mb_stats = create_proc_entry(EXT3_MB_STATS_NAME, -+ S_IFREG | S_IRUGO | S_IWUSR, proc_root_ext3); -+ if (proc_ext3_mb_stats == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", -+ EXT3_MB_STATS_NAME); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+ return -EIO; -+ } -+ -+ proc_ext3_mb_stats->data = NULL; -+ proc_ext3_mb_stats->read_proc = ext3_mb_stats_read; -+ proc_ext3_mb_stats->write_proc = ext3_mb_stats_write; -+ -+ /* Initialize EXT3_MAX_TO_SCAN_NAME */ -+ proc_ext3_mb_max_to_scan = create_proc_entry( -+ EXT3_MB_MAX_TO_SCAN_NAME, -+ S_IFREG | S_IRUGO | S_IWUSR, proc_root_ext3); -+ if (proc_ext3_mb_max_to_scan == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", -+ EXT3_MB_MAX_TO_SCAN_NAME); -+ remove_proc_entry(EXT3_MB_STATS_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+ return -EIO; -+ } -+ -+ proc_ext3_mb_max_to_scan->data = NULL; -+ proc_ext3_mb_max_to_scan->read_proc = ext3_mb_max_to_scan_read; -+ proc_ext3_mb_max_to_scan->write_proc = ext3_mb_max_to_scan_write; -+ -+ /* Initialize EXT3_MIN_TO_SCAN_NAME */ -+ proc_ext3_mb_min_to_scan = create_proc_entry( -+ EXT3_MB_MIN_TO_SCAN_NAME, -+ S_IFREG | S_IRUGO | S_IWUSR, proc_root_ext3); -+ if (proc_ext3_mb_min_to_scan == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", -+ EXT3_MB_MIN_TO_SCAN_NAME); -+ remove_proc_entry(EXT3_MB_MAX_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_STATS_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+ return -EIO; -+ } -+ -+ proc_ext3_mb_min_to_scan->data = NULL; -+ proc_ext3_mb_min_to_scan->read_proc = ext3_mb_min_to_scan_read; -+ proc_ext3_mb_min_to_scan->write_proc = ext3_mb_min_to_scan_write; -+ -+ /* Initialize EXT3_ORDER2_REQ */ -+ proc_ext3_mb_order2_req = create_proc_entry( -+ EXT3_MB_ORDER2_REQ, -+ S_IFREG | S_IRUGO | S_IWUSR, proc_root_ext3); -+ if (proc_ext3_mb_order2_req == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", -+ EXT3_MB_ORDER2_REQ); -+ remove_proc_entry(EXT3_MB_MIN_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_MAX_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_STATS_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+ return -EIO; -+ } -+ -+ proc_ext3_mb_order2_req->data = NULL; -+ proc_ext3_mb_order2_req->read_proc = ext3_mb_order2_req_read; -+ proc_ext3_mb_order2_req->write_proc = ext3_mb_order2_req_write; -+ -+ return 0; -+} -+ -+void exit_ext3_proc(void) -+{ -+ remove_proc_entry(EXT3_MB_STATS_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_MAX_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_MIN_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_ORDER2_REQ, proc_root_ext3); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+} -Index: linux-2.6.12.6-bull/fs/ext3/Makefile -=================================================================== ---- linux-2.6.12.6-bull.orig/fs/ext3/Makefile 2006-04-29 20:39:09.000000000 +0400 -+++ linux-2.6.12.6-bull/fs/ext3/Makefile 2006-04-29 20:39:10.000000000 +0400 -@@ -6,7 +6,7 @@ obj-$(CONFIG_EXT3_FS) += ext3.o - - ext3-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o iopen.o \ - ioctl.o namei.o super.o symlink.o hash.o resize.o \ -- extents.o -+ extents.o mballoc.o - - ext3-$(CONFIG_EXT3_FS_XATTR) += xattr.o xattr_user.o xattr_trusted.o - ext3-$(CONFIG_EXT3_FS_POSIX_ACL) += acl.o diff --git a/ldiskfs/kernel_patches/patches/ext3-mballoc2-2.6.18-vanilla.patch b/ldiskfs/kernel_patches/patches/ext3-mballoc2-2.6.18-vanilla.patch deleted file mode 100644 index 702dfcc..0000000 --- a/ldiskfs/kernel_patches/patches/ext3-mballoc2-2.6.18-vanilla.patch +++ /dev/null @@ -1,3140 +0,0 @@ -Index: linux-stage/include/linux/ext3_fs.h -=================================================================== ---- linux-stage.orig/include/linux/ext3_fs.h 2006-07-16 02:29:43.000000000 +0800 -+++ linux-stage/include/linux/ext3_fs.h 2006-07-16 02:29:49.000000000 +0800 -@@ -53,6 +53,14 @@ - #define ext3_debug(f, a...) do {} while (0) - #endif - -+#define EXT3_MULTIBLOCK_ALLOCATOR 1 -+ -+#define EXT3_MB_HINT_MERGE 1 -+#define EXT3_MB_HINT_RESERVED 2 -+#define EXT3_MB_HINT_METADATA 4 -+#define EXT3_MB_HINT_FIRST 8 -+#define EXT3_MB_HINT_BEST 16 -+ - /* - * Special inodes numbers - */ -@@ -379,6 +387,7 @@ struct ext3_inode { - #define EXT3_MOUNT_IOPEN_NOPRIV 0x800000/* Make iopen world-readable */ - #define EXT3_MOUNT_EXTENTS 0x1000000/* Extents support */ - #define EXT3_MOUNT_EXTDEBUG 0x2000000/* Extents debug */ -+#define EXT3_MOUNT_MBALLOC 0x4000000/* Buddy allocation support */ - - /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ - #ifndef clear_opt -@@ -405,6 +413,14 @@ - #define ext3_find_first_zero_bit ext2_find_first_zero_bit - #define ext3_find_next_zero_bit ext2_find_next_zero_bit - -+#ifndef ext2_find_next_le_bit -+#ifdef __LITTLE_ENDIAN -+#define ext2_find_next_le_bit(addr, size, off) find_next_bit((addr), (size), (off)) -+#else -+#error "mballoc needs a patch for big-endian systems - CFS bug 10634" -+#endif /* __LITTLE_ENDIAN */ -+#endif /* !ext2_find_next_le_bit */ -+ - /* - * Maximal mount counts between two filesystem checks - */ -@@ -749,12 +758,12 @@ ext3_group_first_block_no(struct super_b - /* balloc.c */ - extern int ext3_bg_has_super(struct super_block *sb, int group); - extern unsigned long ext3_bg_num_gdb(struct super_block *sb, int group); --extern ext3_fsblk_t ext3_new_block (handle_t *handle, struct inode *inode, -+extern ext3_fsblk_t ext3_new_block_old(handle_t *handle, struct inode *inode, - ext3_fsblk_t goal, int *errp); - extern ext3_fsblk_t ext3_new_blocks (handle_t *handle, struct inode *inode, - ext3_fsblk_t goal, unsigned long *count, int *errp); - extern void ext3_free_blocks (handle_t *handle, struct inode *inode, -- ext3_fsblk_t block, unsigned long count); -+ ext3_fsblk_t block, unsigned long count, int metadata); - extern void ext3_free_blocks_sb (handle_t *handle, struct super_block *sb, - ext3_fsblk_t block, unsigned long count, - unsigned long *pdquot_freed_blocks); -@@ -881,6 +890,21 @@ extern void ext3_extents_initialize_bloc - extern int ext3_ext_ioctl(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); - -+/* mballoc.c */ -+extern long ext3_mb_stats; -+extern long ext3_mb_max_to_scan; -+extern int ext3_mb_init(struct super_block *sb, int needs_recovery); -+extern int ext3_mb_release(struct super_block *sb); -+extern ext3_fsblk_t ext3_new_block(handle_t *handle, struct inode *inode, -+ ext3_fsblk_t goal, int *errp); -+extern ext3_fsblk_t ext3_mb_new_blocks(handle_t *handle, struct inode *inode, -+ ext3_fsblk_t goal, int *len, int flags, -+ int *errp); -+extern int ext3_mb_reserve_blocks(struct super_block *sb, int); -+extern void ext3_mb_release_blocks(struct super_block *sb, int); -+int __init init_ext3_proc(void); -+void exit_ext3_proc(void); -+ - #endif /* __KERNEL__ */ - - /* EXT3_IOC_CREATE_INUM at bottom of file (visible to kernel and user). */ -Index: linux-stage/include/linux/ext3_fs_sb.h -=================================================================== ---- linux-stage.orig/include/linux/ext3_fs_sb.h 2006-07-16 02:29:43.000000000 +0800 -+++ linux-stage/include/linux/ext3_fs_sb.h 2006-07-16 02:29:49.000000000 +0800 -@@ -21,8 +21,14 @@ - #include - #include - #include -+#include - #endif - #include -+#include -+ -+struct ext3_buddy_group_blocks; -+struct ext3_mb_history; -+#define EXT3_BB_MAX_BLOCKS - - /* - * third extended-fs super-block data in memory -@@ -78,6 +84,43 @@ struct ext3_sb_info { - char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */ - int s_jquota_fmt; /* Format of quota to use */ - #endif -+ -+ /* for buddy allocator */ -+ struct ext3_group_info ***s_group_info; -+ struct inode *s_buddy_cache; -+ long s_blocks_reserved; -+ spinlock_t s_reserve_lock; -+ struct list_head s_active_transaction; -+ struct list_head s_closed_transaction; -+ struct list_head s_committed_transaction; -+ spinlock_t s_md_lock; -+ tid_t s_last_transaction; -+ int s_mb_factor; -+ unsigned short *s_mb_offsets, *s_mb_maxs; -+ unsigned long s_stripe; -+ -+ /* history to debug policy */ -+ struct ext3_mb_history *s_mb_history; -+ int s_mb_history_cur; -+ int s_mb_history_max; -+ struct proc_dir_entry *s_mb_proc; -+ spinlock_t s_mb_history_lock; -+ -+ /* stats for buddy allocator */ -+ atomic_t s_bal_reqs; /* number of reqs with len > 1 */ -+ atomic_t s_bal_success; /* we found long enough chunks */ -+ atomic_t s_bal_allocated; /* in blocks */ -+ atomic_t s_bal_ex_scanned; /* total extents scanned */ -+ atomic_t s_bal_goals; /* goal hits */ -+ atomic_t s_bal_breaks; /* too long searches */ -+ atomic_t s_bal_2orders; /* 2^order hits */ -+ spinlock_t s_bal_lock; -+ unsigned long s_mb_buddies_generated; -+ unsigned long long s_mb_generation_time; - }; -+ -+#define EXT3_GROUP_INFO(sb, group) \ -+ EXT3_SB(sb)->s_group_info[(group) >> EXT3_DESC_PER_BLOCK_BITS(sb)] \ -+ [(group) & (EXT3_DESC_PER_BLOCK(sb) - 1)] - - #endif /* _LINUX_EXT3_FS_SB */ -Index: linux-stage/fs/ext3/super.c -=================================================================== ---- linux-stage.orig/fs/ext3/super.c 2006-07-16 02:29:43.000000000 +0800 -+++ linux-stage/fs/ext3/super.c 2006-07-16 02:29:49.000000000 +0800 -@@ -391,6 +391,7 @@ static void ext3_put_super (struct super - struct ext3_super_block *es = sbi->s_es; - int i; - -+ ext3_mb_release(sb); - ext3_ext_release(sb); - ext3_xattr_put_super(sb); - journal_destroy(sbi->s_journal); -@@ -642,6 +643,7 @@ enum { - Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota, - Opt_iopen, Opt_noiopen, Opt_iopen_nopriv, - Opt_extents, Opt_noextents, Opt_extdebug, -+ Opt_mballoc, Opt_nomballoc, Opt_stripe, - Opt_grpquota - }; - -@@ -696,6 +697,9 @@ static match_table_t tokens = { - {Opt_extents, "extents"}, - {Opt_noextents, "noextents"}, - {Opt_extdebug, "extdebug"}, -+ {Opt_mballoc, "mballoc"}, -+ {Opt_nomballoc, "nomballoc"}, -+ {Opt_stripe, "stripe=%u"}, - {Opt_barrier, "barrier=%u"}, - {Opt_err, NULL}, - {Opt_resize, "resize"}, -@@ -1047,6 +1049,19 @@ clear_qf_name: - case Opt_extdebug: - set_opt (sbi->s_mount_opt, EXTDEBUG); - break; -+ case Opt_mballoc: -+ set_opt(sbi->s_mount_opt, MBALLOC); -+ break; -+ case Opt_nomballoc: -+ clear_opt(sbi->s_mount_opt, MBALLOC); -+ break; -+ case Opt_stripe: -+ if (match_int(&args[0], &option)) -+ return 0; -+ if (option < 0) -+ return 0; -+ sbi->s_stripe = option; -+ break; - default: - printk (KERN_ERR - "EXT3-fs: Unrecognized mount option \"%s\" " -@@ -1773,6 +1778,7 @@ static int ext3_fill_super (struct super - "writeback"); - - ext3_ext_init(sb); -+ ext3_mb_init(sb, needs_recovery); - lock_kernel(); - return 0; - -@@ -2712,7 +2718,13 @@ static struct file_system_type ext3_fs_t - - static int __init init_ext3_fs(void) - { -- int err = init_ext3_xattr(); -+ int err; -+ -+ err = init_ext3_proc(); -+ if (err) -+ return err; -+ -+ err = init_ext3_xattr(); - if (err) - return err; - err = init_inodecache(); -@@ -2734,6 +2746,7 @@ static void __exit exit_ext3_fs(void) - unregister_filesystem(&ext3_fs_type); - destroy_inodecache(); - exit_ext3_xattr(); -+ exit_ext3_proc(); - } - - int ext3_prep_san_write(struct inode *inode, long *blocks, -Index: linux-stage/fs/ext3/extents.c -=================================================================== ---- linux-stage.orig/fs/ext3/extents.c 2006-07-16 02:29:43.000000000 +0800 -+++ linux-stage/fs/ext3/extents.c 2006-07-16 02:29:49.000000000 +0800 -@@ -771,7 +771,7 @@ cleanup: - for (i = 0; i < depth; i++) { - if (!ablocks[i]) - continue; -- ext3_free_blocks(handle, tree->inode, ablocks[i], 1); -+ ext3_free_blocks(handle, tree->inode, ablocks[i], 1, 1); - } - } - kfree(ablocks); -@@ -1428,7 +1428,7 @@ int ext3_ext_rm_idx(handle_t *handle, st - path->p_idx->ei_leaf); - bh = sb_find_get_block(tree->inode->i_sb, path->p_idx->ei_leaf); - ext3_forget(handle, 1, tree->inode, bh, path->p_idx->ei_leaf); -- ext3_free_blocks(handle, tree->inode, path->p_idx->ei_leaf, 1); -+ ext3_free_blocks(handle, tree->inode, path->p_idx->ei_leaf, 1, 1); - return err; - } - -@@ -1913,10 +1913,12 @@ ext3_remove_blocks(struct ext3_extents_t - int needed = ext3_remove_blocks_credits(tree, ex, from, to); - handle_t *handle = ext3_journal_start(tree->inode, needed); - struct buffer_head *bh; -- int i; -+ int i, metadata = 0; - - if (IS_ERR(handle)) - return PTR_ERR(handle); -+ if (S_ISDIR(tree->inode->i_mode) || S_ISLNK(tree->inode->i_mode)) -+ metadata = 1; - if (from >= ex->ee_block && to == ex->ee_block + ex->ee_len - 1) { - /* tail removal */ - unsigned long num, start; -@@ -1928,7 +1930,7 @@ ext3_remove_blocks(struct ext3_extents_t - bh = sb_find_get_block(tree->inode->i_sb, start + i); - ext3_forget(handle, 0, tree->inode, bh, start + i); - } -- ext3_free_blocks(handle, tree->inode, start, num); -+ ext3_free_blocks(handle, tree->inode, start, num, metadata); - } else if (from == ex->ee_block && to <= ex->ee_block + ex->ee_len - 1) { - printk("strange request: removal %lu-%lu from %u:%u\n", - from, to, ex->ee_block, ex->ee_len); -Index: linux-stage/fs/ext3/inode.c -=================================================================== ---- linux-stage.orig/fs/ext3/inode.c 2006-07-16 02:29:43.000000000 +0800 -+++ linux-stage/fs/ext3/inode.c 2006-07-16 02:29:49.000000000 +0800 -@@ -562,7 +562,7 @@ static int ext3_alloc_blocks(handle_t *h - return ret; - failed_out: - for (i = 0; i i_sb; -- if (!sb) { -- printk ("ext3_free_blocks: nonexistent device"); -- return; -- } -- ext3_free_blocks_sb(handle, sb, block, count, &dquot_freed_blocks); -- if (dquot_freed_blocks) -- DQUOT_FREE_BLOCK(inode, dquot_freed_blocks); -- return; --} -- - /* - * For ext3 allocations, we must not reuse any blocks which are - * allocated in the bitmap buffer's "last committed data" copy. This -@@ -1463,7 +1445,7 @@ out: - return 0; - } - --ext3_fsblk_t ext3_new_block(handle_t *handle, struct inode *inode, -+ext3_fsblk_t ext3_new_block_old(handle_t *handle, struct inode *inode, - ext3_fsblk_t goal, int *errp) - { - unsigned long count = 1; -Index: linux-stage/fs/ext3/xattr.c -=================================================================== ---- linux-stage.orig/fs/ext3/xattr.c 2006-07-16 02:29:43.000000000 +0800 -+++ linux-stage/fs/ext3/xattr.c 2006-07-16 02:29:49.000000000 +0800 -@@ -484,7 +484,7 @@ ext3_xattr_release_block(handle_t *handl - ea_bdebug(bh, "refcount now=0; freeing"); - if (ce) - mb_cache_entry_free(ce); -- ext3_free_blocks(handle, inode, bh->b_blocknr, 1); -+ ext3_free_blocks(handle, inode, bh->b_blocknr, 1, 1); - get_bh(bh); - ext3_forget(handle, 1, inode, bh, bh->b_blocknr); - } else { -@@ -805,7 +805,7 @@ inserted: - new_bh = sb_getblk(sb, block); - if (!new_bh) { - getblk_failed: -- ext3_free_blocks(handle, inode, block, 1); -+ ext3_free_blocks(handle, inode, block, 1, 1); - error = -EIO; - goto cleanup; - } -Index: linux-stage/fs/ext3/mballoc.c -=================================================================== ---- /dev/null 1970-01-01 00:00:00.000000000 +0000 -+++ linux-stage/fs/ext3/mballoc.c 2006-07-16 02:29:49.000000000 +0800 -@@ -0,0 +1,2730 @@ -+/* -+ * Copyright (c) 2003-2005, Cluster File Systems, Inc, info@clusterfs.com -+ * Written by Alex Tomas -+ * -+ * 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 -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+ -+/* -+ * TODO: -+ * - 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 -+ * - special flag to advice allocator to look for requested + N blocks -+ * this may improve interaction between extents and mballoc -+ * - tree of groups sorted by number of free blocks -+ * - percpu reservation code (hotpath) -+ * - error handling -+ */ -+ -+/* -+ * with AGRESSIVE_CHECK allocator runs consistency checks over -+ * structures. these checks slow things down a lot -+ */ -+#define AGGRESSIVE_CHECK__ -+ -+/* -+ */ -+#define MB_DEBUG__ -+#ifdef MB_DEBUG -+#define mb_debug(fmt,a...) printk(fmt, ##a) -+#else -+#define mb_debug(fmt,a...) -+#endif -+ -+/* -+ * with EXT3_MB_HISTORY mballoc stores last N allocations in memory -+ * and you can monitor it in /proc/fs/ext3//mb_history -+ */ -+#define EXT3_MB_HISTORY -+ -+/* -+ * How long mballoc can look for a best extent (in found extents) -+ */ -+long ext3_mb_max_to_scan = 500; -+ -+/* -+ * How long mballoc must look for a best extent -+ */ -+long ext3_mb_min_to_scan = 30; -+ -+/* -+ * with 'ext3_mb_stats' allocator will collect stats that will be -+ * shown at umount. The collecting costs though! -+ */ -+ -+long ext3_mb_stats = 1; -+ -+/* -+ * for which requests use 2^N search using buddies -+ */ -+long ext3_mb_order2_reqs = 8; -+ -+#ifdef EXT3_BB_MAX_BLOCKS -+#undef EXT3_BB_MAX_BLOCKS -+#endif -+#define EXT3_BB_MAX_BLOCKS 30 -+ -+struct ext3_free_metadata { -+ unsigned short group; -+ unsigned short num; -+ unsigned short blocks[EXT3_BB_MAX_BLOCKS]; -+ struct list_head list; -+}; -+ -+struct ext3_group_info { -+ unsigned long bb_state; -+ unsigned long bb_tid; -+ struct ext3_free_metadata *bb_md_cur; -+ unsigned short bb_first_free; -+ unsigned short bb_free; -+ unsigned short bb_fragments; -+ unsigned short bb_counters[]; -+}; -+ -+ -+#define EXT3_GROUP_INFO_NEED_INIT_BIT 0 -+#define EXT3_GROUP_INFO_LOCKED_BIT 1 -+ -+#define EXT3_MB_GRP_NEED_INIT(grp) \ -+ (test_bit(EXT3_GROUP_INFO_NEED_INIT_BIT, &(grp)->bb_state)) -+ -+struct ext3_free_extent { -+ __u16 fe_start; -+ __u16 fe_len; -+ __u16 fe_group; -+}; -+ -+struct ext3_allocation_context { -+ struct super_block *ac_sb; -+ -+ /* search goals */ -+ struct ext3_free_extent ac_g_ex; -+ -+ /* the best found extent */ -+ struct ext3_free_extent ac_b_ex; -+ -+ /* number of iterations done. we have to track to limit searching */ -+ unsigned long ac_ex_scanned; -+ __u16 ac_groups_scanned; -+ __u16 ac_found; -+ __u16 ac_tail; -+ __u16 ac_buddy; -+ __u8 ac_status; -+ __u8 ac_flags; /* allocation hints */ -+ __u8 ac_criteria; -+ __u8 ac_repeats; -+ __u8 ac_2order; /* if request is to allocate 2^N blocks and -+ * N > 0, the field stores N, otherwise 0 */ -+ -+ struct page *ac_buddy_page; -+ struct page *ac_bitmap_page; -+}; -+ -+#define AC_STATUS_CONTINUE 1 -+#define AC_STATUS_FOUND 2 -+#define AC_STATUS_BREAK 3 -+ -+struct ext3_mb_history { -+ struct ext3_free_extent goal; /* goal allocation */ -+ struct ext3_free_extent result; /* result allocation */ -+ unsigned pid; -+ unsigned ino; -+ __u16 found; /* how many extents have been found */ -+ __u16 groups; /* how many groups have been scanned */ -+ __u16 tail; /* what tail broke some buddy */ -+ __u16 buddy; /* buddy the tail ^^^ broke */ -+ __u8 cr; /* which phase the result extent was found at */ -+ __u8 merged; -+}; -+ -+struct ext3_buddy { -+ struct page *bd_buddy_page; -+ void *bd_buddy; -+ struct page *bd_bitmap_page; -+ void *bd_bitmap; -+ struct ext3_group_info *bd_info; -+ struct super_block *bd_sb; -+ __u16 bd_blkbits; -+ __u16 bd_group; -+}; -+#define EXT3_MB_BITMAP(e3b) ((e3b)->bd_bitmap) -+#define EXT3_MB_BUDDY(e3b) ((e3b)->bd_buddy) -+ -+#ifndef EXT3_MB_HISTORY -+#define ext3_mb_store_history(sb,ino,ac) -+#else -+static void ext3_mb_store_history(struct super_block *, unsigned ino, -+ struct ext3_allocation_context *ac); -+#endif -+ -+#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1) -+ -+static struct proc_dir_entry *proc_root_ext3; -+ -+int ext3_create (struct inode *, struct dentry *, int, struct nameidata *); -+struct buffer_head * read_block_bitmap(struct super_block *, unsigned int); -+int ext3_mb_reserve_blocks(struct super_block *, int); -+void ext3_mb_release_blocks(struct super_block *, int); -+void ext3_mb_poll_new_transaction(struct super_block *, handle_t *); -+void ext3_mb_free_committed_blocks(struct super_block *); -+ -+#if BITS_PER_LONG == 64 -+#define mb_correct_addr_and_bit(bit,addr) \ -+{ \ -+ bit += ((unsigned long) addr & 7UL) << 3; \ -+ addr = (void *) ((unsigned long) addr & ~7UL); \ -+} -+#elif BITS_PER_LONG == 32 -+#define mb_correct_addr_and_bit(bit,addr) \ -+{ \ -+ bit += ((unsigned long) addr & 3UL) << 3; \ -+ addr = (void *) ((unsigned long) addr & ~3UL); \ -+} -+#else -+#error "how many bits you are?!" -+#endif -+ -+static inline int mb_test_bit(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ return ext2_test_bit(bit, addr); -+} -+ -+static inline void mb_set_bit(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ ext2_set_bit(bit, addr); -+} -+ -+static inline void mb_set_bit_atomic(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ ext2_set_bit_atomic(NULL, bit, addr); -+} -+ -+static inline void mb_clear_bit(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ ext2_clear_bit(bit, addr); -+} -+ -+static inline void mb_clear_bit_atomic(int bit, void *addr) -+{ -+ mb_correct_addr_and_bit(bit,addr); -+ ext2_clear_bit_atomic(NULL, bit, addr); -+} -+ -+static inline int mb_find_next_zero_bit(void *addr, int max, int start) -+{ -+ int fix; -+#if BITS_PER_LONG == 64 -+ fix = ((unsigned long) addr & 7UL) << 3; -+ addr = (void *) ((unsigned long) addr & ~7UL); -+#elif BITS_PER_LONG == 32 -+ fix = ((unsigned long) addr & 3UL) << 3; -+ addr = (void *) ((unsigned long) addr & ~3UL); -+#else -+#error "how many bits you are?!" -+#endif -+ max += fix; -+ start += fix; -+ return ext2_find_next_zero_bit(addr, max, start) - fix; -+} -+ -+static inline void *mb_find_buddy(struct ext3_buddy *e3b, int order, int *max) -+{ -+ char *bb; -+ -+ J_ASSERT(EXT3_MB_BITMAP(e3b) != EXT3_MB_BUDDY(e3b)); -+ J_ASSERT(max != NULL); -+ -+ if (order > e3b->bd_blkbits + 1) { -+ *max = 0; -+ return NULL; -+ } -+ -+ /* at order 0 we see each particular block */ -+ *max = 1 << (e3b->bd_blkbits + 3); -+ if (order == 0) -+ return EXT3_MB_BITMAP(e3b); -+ -+ bb = EXT3_MB_BUDDY(e3b) + EXT3_SB(e3b->bd_sb)->s_mb_offsets[order]; -+ *max = EXT3_SB(e3b->bd_sb)->s_mb_maxs[order]; -+ -+ return bb; -+} -+ -+#ifdef AGGRESSIVE_CHECK -+ -+static void mb_check_buddy(struct ext3_buddy *e3b) -+{ -+ int order = e3b->bd_blkbits + 1; -+ int max, max2, i, j, k, count; -+ int fragments = 0, fstart; -+ void *buddy, *buddy2; -+ -+ if (!test_opt(e3b->bd_sb, MBALLOC)) -+ return; -+ -+ { -+ static int mb_check_counter = 0; -+ if (mb_check_counter++ % 300 != 0) -+ return; -+ } -+ -+ while (order > 1) { -+ buddy = mb_find_buddy(e3b, order, &max); -+ J_ASSERT(buddy); -+ buddy2 = mb_find_buddy(e3b, order - 1, &max2); -+ J_ASSERT(buddy2); -+ J_ASSERT(buddy != buddy2); -+ J_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)) -+ J_ASSERT(mb_test_bit((i<<1)+1, buddy2)); -+ else if (!mb_test_bit((i << 1) + 1, buddy2)) -+ J_ASSERT(mb_test_bit(i << 1, buddy2)); -+ continue; -+ } -+ -+ /* both bits in buddy2 must be 0 */ -+ J_ASSERT(mb_test_bit(i << 1, buddy2)); -+ J_ASSERT(mb_test_bit((i << 1) + 1, buddy2)); -+ -+ for (j = 0; j < (1 << order); j++) { -+ k = (i * (1 << order)) + j; -+ J_ASSERT(!mb_test_bit(k, EXT3_MB_BITMAP(e3b))); -+ } -+ count++; -+ } -+ J_ASSERT(e3b->bd_info->bb_counters[order] == count); -+ order--; -+ } -+ -+ fstart = -1; -+ buddy = mb_find_buddy(e3b, 0, &max); -+ for (i = 0; i < max; i++) { -+ if (!mb_test_bit(i, buddy)) { -+ J_ASSERT(i >= e3b->bd_info->bb_first_free); -+ if (fstart == -1) { -+ fragments++; -+ fstart = i; -+ } -+ continue; -+ } -+ fstart = -1; -+ /* check used bits only */ -+ for (j = 0; j < e3b->bd_blkbits + 1; j++) { -+ buddy2 = mb_find_buddy(e3b, j, &max2); -+ k = i >> j; -+ J_ASSERT(k < max2); -+ J_ASSERT(mb_test_bit(k, buddy2)); -+ } -+ } -+ J_ASSERT(!EXT3_MB_GRP_NEED_INIT(e3b->bd_info)); -+ J_ASSERT(e3b->bd_info->bb_fragments == fragments); -+} -+ -+#else -+#define mb_check_buddy(e3b) -+#endif -+ -+/* find most significant bit */ -+static int inline fmsb(unsigned short word) -+{ -+ int order; -+ -+ if (word > 255) { -+ order = 7; -+ word >>= 8; -+ } else { -+ order = -1; -+ } -+ -+ do { -+ order++; -+ word >>= 1; -+ } while (word != 0); -+ -+ return order; -+} -+ -+static void inline -+ext3_mb_mark_free_simple(struct super_block *sb, void *buddy, unsigned first, -+ int len, struct ext3_group_info *grp) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ unsigned short min, max, chunk, border; -+ -+ mb_debug("mark %u/%u free\n", first, len); -+ J_ASSERT(len < EXT3_BLOCKS_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 = fmsb(len); -+ -+ mb_debug(" %u/%u -> max %u, min %u\n", -+ first & ((2 << sb->s_blocksize_bits) - 1), -+ len, max, min); -+ -+ if (max < min) -+ min = max; -+ chunk = 1 << min; -+ -+ /* mark multiblock chunks only */ -+ grp->bb_counters[min]++; -+ if (min > 0) { -+ mb_debug(" set %u at %u \n", first >> min, -+ sbi->s_mb_offsets[min]); -+ mb_clear_bit(first >> min, buddy + sbi->s_mb_offsets[min]); -+ } -+ -+ len -= chunk; -+ first += chunk; -+ } -+} -+ -+static void -+ext3_mb_generate_buddy(struct super_block *sb, void *buddy, void *bitmap, -+ int group) -+{ -+ struct ext3_group_info *grp = EXT3_GROUP_INFO(sb, group); -+ unsigned short max = EXT3_BLOCKS_PER_GROUP(sb); -+ unsigned short i = 0, first, len; -+ unsigned free = 0, fragments = 0; -+ unsigned long long period = get_cycles(); -+ -+ i = mb_find_next_zero_bit(bitmap, max, 0); -+ grp->bb_first_free = i; -+ while (i < max) { -+ fragments++; -+ first = i; -+ i = ext2_find_next_le_bit(bitmap, max, i); -+ len = i - first; -+ free += len; -+ if (len > 1) -+ ext3_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; -+ -+ /* bb_state shouldn't being modified because all -+ * others waits for init completion on page lock */ -+ clear_bit(EXT3_GROUP_INFO_NEED_INIT_BIT, &grp->bb_state); -+ if (free != grp->bb_free) { -+ printk("EXT3-fs: group %u: %u blocks in bitmap, %u in gd\n", -+ group, free, grp->bb_free); -+ grp->bb_free = free; -+ } -+ -+ period = get_cycles() - period; -+ spin_lock(&EXT3_SB(sb)->s_bal_lock); -+ EXT3_SB(sb)->s_mb_buddies_generated++; -+ EXT3_SB(sb)->s_mb_generation_time += period; -+ spin_unlock(&EXT3_SB(sb)->s_bal_lock); -+} -+ -+static int ext3_mb_init_cache(struct page *page) -+{ -+ int blocksize, blocks_per_page, groups_per_page; -+ int err = 0, i, first_group, first_block; -+ struct super_block *sb; -+ struct buffer_head *bhs; -+ struct buffer_head **bh; -+ struct inode *inode; -+ char *data, *bitmap; -+ -+ mb_debug("init page %lu\n", page->index); -+ -+ inode = page->mapping->host; -+ sb = inode->i_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) { -+ err = -ENOMEM; -+ i = sizeof(struct buffer_head *) * groups_per_page; -+ bh = kmalloc(i, GFP_NOFS); -+ if (bh == NULL) -+ goto out; -+ memset(bh, 0, i); -+ } else -+ bh = &bhs; -+ -+ first_group = page->index * blocks_per_page / 2; -+ -+ /* read all groups the page covers into the cache */ -+ for (i = 0; i < groups_per_page; i++) { -+ struct ext3_group_desc * desc; -+ -+ if (first_group + i >= EXT3_SB(sb)->s_groups_count) -+ break; -+ -+ err = -EIO; -+ desc = ext3_get_group_desc(sb, first_group + i, NULL); -+ if (desc == NULL) -+ goto out; -+ -+ err = -ENOMEM; -+ bh[i] = sb_getblk(sb, le32_to_cpu(desc->bg_block_bitmap)); -+ if (bh[i] == NULL) -+ goto out; -+ -+ if (buffer_uptodate(bh[i])) -+ continue; -+ -+ lock_buffer(bh[i]); -+ if (buffer_uptodate(bh[i])) { -+ unlock_buffer(bh[i]); -+ continue; -+ } -+ -+ get_bh(bh[i]); -+ bh[i]->b_end_io = end_buffer_read_sync; -+ submit_bh(READ, bh[i]); -+ mb_debug("read bitmap for group %u\n", first_group + i); -+ } -+ -+ /* wait for I/O completion */ -+ for (i = 0; i < groups_per_page && bh[i]; i++) -+ wait_on_buffer(bh[i]); -+ -+ err = -EIO; -+ for (i = 0; i < groups_per_page && bh[i]; i++) -+ if (!buffer_uptodate(bh[i])) -+ 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 >= EXT3_SB(sb)->s_groups_count) -+ break; -+ -+ data = page_address(page) + (i * blocksize); -+ bitmap = bh[group - first_group]->b_data; -+ -+ if ((first_block + i) & 1) { -+ /* this is block of buddy */ -+ mb_debug("put buddy for group %u in page %lu/%x\n", -+ group, page->index, i * blocksize); -+ memset(data, 0xff, blocksize); -+ EXT3_GROUP_INFO(sb, group)->bb_fragments = 0; -+ memset(EXT3_GROUP_INFO(sb, group)->bb_counters, 0, -+ sizeof(unsigned short)*(sb->s_blocksize_bits+2)); -+ ext3_mb_generate_buddy(sb, data, bitmap, group); -+ } else { -+ /* this is block of bitmap */ -+ mb_debug("put bitmap for group %u in page %lu/%x\n", -+ group, page->index, i * blocksize); -+ memcpy(data, bitmap, blocksize); -+ } -+ } -+ SetPageUptodate(page); -+ -+out: -+ if (bh) { -+ for (i = 0; i < groups_per_page && bh[i]; i++) -+ brelse(bh[i]); -+ if (bh != &bhs) -+ kfree(bh); -+ } -+ return err; -+} -+ -+static int ext3_mb_load_buddy(struct super_block *sb, int group, -+ struct ext3_buddy *e3b) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct inode *inode = sbi->s_buddy_cache; -+ int blocks_per_page, block, pnum, poff; -+ struct page *page; -+ -+ mb_debug("load group %u\n", group); -+ -+ blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize; -+ -+ e3b->bd_blkbits = sb->s_blocksize_bits; -+ e3b->bd_info = EXT3_GROUP_INFO(sb, group); -+ e3b->bd_sb = sb; -+ e3b->bd_group = group; -+ e3b->bd_buddy_page = NULL; -+ e3b->bd_bitmap_page = NULL; -+ -+ 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) -+ 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)) -+ ext3_mb_init_cache(page); -+ unlock_page(page); -+ } -+ } -+ if (page == NULL || !PageUptodate(page)) -+ goto err; -+ e3b->bd_bitmap_page = page; -+ e3b->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)) -+ ext3_mb_init_cache(page); -+ unlock_page(page); -+ } -+ } -+ if (page == NULL || !PageUptodate(page)) -+ goto err; -+ e3b->bd_buddy_page = page; -+ e3b->bd_buddy = page_address(page) + (poff * sb->s_blocksize); -+ mark_page_accessed(page); -+ -+ J_ASSERT(e3b->bd_bitmap_page != NULL); -+ J_ASSERT(e3b->bd_buddy_page != NULL); -+ -+ return 0; -+ -+err: -+ if (e3b->bd_bitmap_page) -+ page_cache_release(e3b->bd_bitmap_page); -+ if (e3b->bd_buddy_page) -+ page_cache_release(e3b->bd_buddy_page); -+ e3b->bd_buddy = NULL; -+ e3b->bd_bitmap = NULL; -+ return -EIO; -+} -+ -+static void ext3_mb_release_desc(struct ext3_buddy *e3b) -+{ -+ if (e3b->bd_bitmap_page) -+ page_cache_release(e3b->bd_bitmap_page); -+ if (e3b->bd_buddy_page) -+ page_cache_release(e3b->bd_buddy_page); -+} -+ -+ -+static inline void -+ext3_lock_group(struct super_block *sb, int group) -+{ -+ bit_spin_lock(EXT3_GROUP_INFO_LOCKED_BIT, -+ &EXT3_GROUP_INFO(sb, group)->bb_state); -+} -+ -+static inline void -+ext3_unlock_group(struct super_block *sb, int group) -+{ -+ bit_spin_unlock(EXT3_GROUP_INFO_LOCKED_BIT, -+ &EXT3_GROUP_INFO(sb, group)->bb_state); -+} -+ -+static int mb_find_order_for_block(struct ext3_buddy *e3b, int block) -+{ -+ int order = 1; -+ void *bb; -+ -+ J_ASSERT(EXT3_MB_BITMAP(e3b) != EXT3_MB_BUDDY(e3b)); -+ J_ASSERT(block < (1 << (e3b->bd_blkbits + 3))); -+ -+ bb = EXT3_MB_BUDDY(e3b); -+ while (order <= e3b->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 << (e3b->bd_blkbits - order); -+ order++; -+ } -+ return 0; -+} -+ -+static inline 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_atomic(cur, bm); -+ cur++; -+ } -+} -+ -+static inline void mb_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: clear whole word at once */ -+ addr = bm + (cur >> 3); -+ *addr = 0xffffffff; -+ cur += 32; -+ continue; -+ } -+ mb_set_bit_atomic(cur, bm); -+ cur++; -+ } -+} -+ -+static int mb_free_blocks(struct ext3_buddy *e3b, int first, int count) -+{ -+ int block = 0, max = 0, order; -+ void *buddy, *buddy2; -+ -+ mb_check_buddy(e3b); -+ -+ e3b->bd_info->bb_free += count; -+ if (first < e3b->bd_info->bb_first_free) -+ e3b->bd_info->bb_first_free = first; -+ -+ /* let's maintain fragments counter */ -+ if (first != 0) -+ block = !mb_test_bit(first - 1, EXT3_MB_BITMAP(e3b)); -+ if (first + count < EXT3_SB(e3b->bd_sb)->s_mb_maxs[0]) -+ max = !mb_test_bit(first + count, EXT3_MB_BITMAP(e3b)); -+ if (block && max) -+ e3b->bd_info->bb_fragments--; -+ else if (!block && !max) -+ e3b->bd_info->bb_fragments++; -+ -+ /* let's maintain buddy itself */ -+ while (count-- > 0) { -+ block = first++; -+ order = 0; -+ -+ J_ASSERT(mb_test_bit(block, EXT3_MB_BITMAP(e3b))); -+ mb_clear_bit(block, EXT3_MB_BITMAP(e3b)); -+ e3b->bd_info->bb_counters[order]++; -+ -+ /* start of the buddy */ -+ buddy = mb_find_buddy(e3b, 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(e3b, 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); -+ } -+ e3b->bd_info->bb_counters[order]--; -+ e3b->bd_info->bb_counters[order]--; -+ -+ block = block >> 1; -+ order++; -+ e3b->bd_info->bb_counters[order]++; -+ -+ mb_clear_bit(block, buddy2); -+ buddy = buddy2; -+ } while (1); -+ } -+ mb_check_buddy(e3b); -+ -+ return 0; -+} -+ -+static int mb_find_extent(struct ext3_buddy *e3b, int order, int block, -+ int needed, struct ext3_free_extent *ex) -+{ -+ int next = block, max, ord; -+ void *buddy; -+ -+ J_ASSERT(ex != NULL); -+ -+ buddy = mb_find_buddy(e3b, order, &max); -+ J_ASSERT(buddy); -+ J_ASSERT(block < max); -+ if (mb_test_bit(block, buddy)) { -+ ex->fe_len = 0; -+ ex->fe_start = 0; -+ ex->fe_group = 0; -+ return 0; -+ } -+ -+ if (likely(order == 0)) { -+ /* find actual order */ -+ order = mb_find_order_for_block(e3b, block); -+ block = block >> order; -+ } -+ -+ ex->fe_len = 1 << order; -+ ex->fe_start = block << order; -+ ex->fe_group = e3b->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(e3b, order, &max))) { -+ -+ if (block + 1 >= max) -+ break; -+ -+ next = (block + 1) * (1 << order); -+ if (mb_test_bit(next, EXT3_MB_BITMAP(e3b))) -+ break; -+ -+ ord = mb_find_order_for_block(e3b, next); -+ -+ order = ord; -+ block = next >> order; -+ ex->fe_len += 1 << order; -+ } -+ -+ J_ASSERT(ex->fe_start + ex->fe_len <= (1 << (e3b->bd_blkbits + 3))); -+ return ex->fe_len; -+} -+ -+static int mb_mark_used(struct ext3_buddy *e3b, struct ext3_free_extent *ex) -+{ -+ int ord, mlen = 0, max = 0, cur; -+ int start = ex->fe_start; -+ int len = ex->fe_len; -+ unsigned ret = 0; -+ int len0 = len; -+ void *buddy; -+ -+ mb_check_buddy(e3b); -+ -+ e3b->bd_info->bb_free -= len; -+ if (e3b->bd_info->bb_first_free == start) -+ e3b->bd_info->bb_first_free += len; -+ -+ /* let's maintain fragments counter */ -+ if (start != 0) -+ mlen = !mb_test_bit(start - 1, EXT3_MB_BITMAP(e3b)); -+ if (start + len < EXT3_SB(e3b->bd_sb)->s_mb_maxs[0]) -+ max = !mb_test_bit(start + len, EXT3_MB_BITMAP(e3b)); -+ if (mlen && max) -+ e3b->bd_info->bb_fragments++; -+ else if (!mlen && !max) -+ e3b->bd_info->bb_fragments--; -+ -+ /* let's maintain buddy itself */ -+ while (len) { -+ ord = mb_find_order_for_block(e3b, start); -+ -+ if (((start >> ord) << ord) == start && len >= (1 << ord)) { -+ /* the whole chunk may be allocated at once! */ -+ mlen = 1 << ord; -+ buddy = mb_find_buddy(e3b, ord, &max); -+ J_ASSERT((start >> ord) < max); -+ mb_set_bit(start >> ord, buddy); -+ e3b->bd_info->bb_counters[ord]--; -+ start += mlen; -+ len -= mlen; -+ J_ASSERT(len >= 0); -+ continue; -+ } -+ -+ /* store for history */ -+ if (ret == 0) -+ ret = len | (ord << 16); -+ -+ /* we have to split large buddy */ -+ J_ASSERT(ord > 0); -+ buddy = mb_find_buddy(e3b, ord, &max); -+ mb_set_bit(start >> ord, buddy); -+ e3b->bd_info->bb_counters[ord]--; -+ -+ ord--; -+ cur = (start >> ord) & ~1U; -+ buddy = mb_find_buddy(e3b, ord, &max); -+ mb_clear_bit(cur, buddy); -+ mb_clear_bit(cur + 1, buddy); -+ e3b->bd_info->bb_counters[ord]++; -+ e3b->bd_info->bb_counters[ord]++; -+ } -+ -+ /* now drop all the bits in bitmap */ -+ mb_set_bits(EXT3_MB_BITMAP(e3b), ex->fe_start, len0); -+ -+ mb_check_buddy(e3b); -+ -+ return ret; -+} -+ -+/* -+ * Must be called under group lock! -+ */ -+static void ext3_mb_use_best_found(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ unsigned long ret; -+ -+ ac->ac_b_ex.fe_len = min(ac->ac_b_ex.fe_len, ac->ac_g_ex.fe_len); -+ ret = mb_mark_used(e3b, &ac->ac_b_ex); -+ -+ ac->ac_status = AC_STATUS_FOUND; -+ ac->ac_tail = ret & 0xffff; -+ ac->ac_buddy = ret >> 16; -+ -+ /* hold in-core structures until allocated -+ * blocks are marked non-free in on-disk bitmap */ -+ ac->ac_buddy_page = e3b->bd_buddy_page; -+ page_cache_get(e3b->bd_buddy_page); -+ ac->ac_bitmap_page = e3b->bd_bitmap_page; -+ page_cache_get(e3b->bd_bitmap_page); -+} -+ -+/* -+ * 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 ext3_mb_measure_extent(struct ext3_allocation_context *ac, -+ struct ext3_free_extent *ex, -+ struct ext3_buddy *e3b) -+{ -+ struct ext3_free_extent *bex = &ac->ac_b_ex; -+ struct ext3_free_extent *gex = &ac->ac_g_ex; -+ -+ J_ASSERT(ex->fe_len > 0); -+ J_ASSERT(ex->fe_len < (1 << ac->ac_sb->s_blocksize_bits) * 8); -+ J_ASSERT(ex->fe_start < (1 << ac->ac_sb->s_blocksize_bits) * 8); -+ -+ ac->ac_found++; -+ -+ /* -+ * The special case - take what you catch first -+ */ -+ if (unlikely(ac->ac_flags & EXT3_MB_HINT_FIRST)) { -+ *bex = *ex; -+ ext3_mb_use_best_found(ac, e3b); -+ return; -+ } -+ -+ /* -+ * Let's check whether the chunk is good enough -+ */ -+ if (ex->fe_len == gex->fe_len) { -+ *bex = *ex; -+ ext3_mb_use_best_found(ac, e3b); -+ 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 */ -+ *bex = *ex; -+ } -+ -+ /* -+ * Let's scan at least few extents and don't pick up a first one -+ */ -+ if (bex->fe_len > gex->fe_len && ac->ac_found > ext3_mb_min_to_scan) -+ ac->ac_status = AC_STATUS_BREAK; -+ -+ /* -+ * We don't want to scan for a whole year -+ */ -+ if (ac->ac_found > ext3_mb_max_to_scan) -+ ac->ac_status = AC_STATUS_BREAK; -+} -+ -+static int ext3_mb_try_best_found(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ struct ext3_free_extent ex = ac->ac_b_ex; -+ int group = ex.fe_group, max, err; -+ -+ J_ASSERT(ex.fe_len > 0); -+ err = ext3_mb_load_buddy(ac->ac_sb, group, e3b); -+ if (err) -+ return err; -+ -+ ext3_lock_group(ac->ac_sb, group); -+ max = mb_find_extent(e3b, 0, ex.fe_start, ex.fe_len, &ex); -+ -+ if (max > 0) { -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ } -+ -+ ext3_unlock_group(ac->ac_sb, group); -+ -+ ext3_mb_release_desc(e3b); -+ -+ return 0; -+} -+ -+static int ext3_mb_find_by_goal(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ int group = ac->ac_g_ex.fe_group, max, err; -+ struct ext3_sb_info *sbi = EXT3_SB(ac->ac_sb); -+ struct ext3_super_block *es = sbi->s_es; -+ struct ext3_free_extent ex; -+ -+ err = ext3_mb_load_buddy(ac->ac_sb, group, e3b); -+ if (err) -+ return err; -+ -+ ext3_lock_group(ac->ac_sb, group); -+ max = mb_find_extent(e3b, 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) { -+ ext3_fsblk_t start; -+ start = (e3b->bd_group * EXT3_BLOCKS_PER_GROUP(ac->ac_sb) + -+ ex.fe_start + le32_to_cpu(es->s_first_data_block)); -+ if (start % sbi->s_stripe == 0) { -+ ac->ac_found++; -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ } -+ } else if (max >= ac->ac_g_ex.fe_len) { -+ J_ASSERT(ex.fe_len > 0); -+ J_ASSERT(ex.fe_group == ac->ac_g_ex.fe_group); -+ J_ASSERT(ex.fe_start == ac->ac_g_ex.fe_start); -+ ac->ac_found++; -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ } else if (max > 0 && (ac->ac_flags & EXT3_MB_HINT_MERGE)) { -+ /* Sometimes, caller may want to merge even small -+ * number of blocks to an existing extent */ -+ J_ASSERT(ex.fe_len > 0); -+ J_ASSERT(ex.fe_group == ac->ac_g_ex.fe_group); -+ J_ASSERT(ex.fe_start == ac->ac_g_ex.fe_start); -+ ac->ac_found++; -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ } -+ ext3_unlock_group(ac->ac_sb, group); -+ -+ ext3_mb_release_desc(e3b); -+ -+ 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 void ext3_mb_simple_scan_group(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ struct super_block *sb = ac->ac_sb; -+ struct ext3_group_info *grp = e3b->bd_info; -+ void *buddy; -+ int i, k, max; -+ -+ J_ASSERT(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(e3b, i, &max); -+ if (buddy == NULL) { -+ printk(KERN_ALERT "looking for wrong order?\n"); -+ break; -+ } -+ -+ k = mb_find_next_zero_bit(buddy, max, 0); -+ J_ASSERT(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 = e3b->bd_group; -+ -+ ext3_mb_use_best_found(ac, e3b); -+ J_ASSERT(ac->ac_b_ex.fe_len == ac->ac_g_ex.fe_len); -+ -+ if (unlikely(ext3_mb_stats)) -+ atomic_inc(&EXT3_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 void ext3_mb_complex_scan_group(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ struct super_block *sb = ac->ac_sb; -+ void *bitmap = EXT3_MB_BITMAP(e3b); -+ struct ext3_free_extent ex; -+ int i, free; -+ -+ free = e3b->bd_info->bb_free; -+ J_ASSERT(free > 0); -+ -+ i = e3b->bd_info->bb_first_free; -+ -+ while (free && ac->ac_status == AC_STATUS_CONTINUE) { -+ i = mb_find_next_zero_bit(bitmap, sb->s_blocksize * 8, i); -+ if (i >= sb->s_blocksize * 8) { -+ J_ASSERT(free == 0); -+ break; -+ } -+ -+ mb_find_extent(e3b, 0, i, ac->ac_g_ex.fe_len, &ex); -+ J_ASSERT(ex.fe_len > 0); -+ J_ASSERT(free >= ex.fe_len); -+ -+ ext3_mb_measure_extent(ac, &ex, e3b); -+ -+ i += ex.fe_len; -+ free -= ex.fe_len; -+ } -+} -+ -+/* -+ * This is a special case for storages like raid5 -+ * we try to find stripe-aligned chunks for stripe-size requests -+ */ -+static void ext3_mb_scan_aligned(struct ext3_allocation_context *ac, -+ struct ext3_buddy *e3b) -+{ -+ struct super_block *sb = ac->ac_sb; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ void *bitmap = EXT3_MB_BITMAP(e3b); -+ struct ext3_free_extent ex; -+ ext3_fsblk_t i, max; -+ -+ J_ASSERT(sbi->s_stripe != 0); -+ -+ /* find first stripe-aligned block */ -+ i = e3b->bd_group * EXT3_BLOCKS_PER_GROUP(sb) + -+ le32_to_cpu(sbi->s_es->s_first_data_block); -+ i = ((i + sbi->s_stripe - 1) / sbi->s_stripe) * sbi->s_stripe; -+ i = (i - le32_to_cpu(sbi->s_es->s_first_data_block)) % -+ EXT3_BLOCKS_PER_GROUP(sb); -+ -+ while (i < sb->s_blocksize * 8) { -+ if (!mb_test_bit(i, bitmap)) { -+ max = mb_find_extent(e3b, 0, i, sbi->s_stripe, &ex); -+ if (max >= sbi->s_stripe) { -+ ac->ac_found++; -+ ac->ac_b_ex = ex; -+ ext3_mb_use_best_found(ac, e3b); -+ break; -+ } -+ } -+ i += sbi->s_stripe; -+ } -+} -+ -+static int ext3_mb_good_group(struct ext3_allocation_context *ac, -+ int group, int cr) -+{ -+ struct ext3_group_info *grp = EXT3_GROUP_INFO(ac->ac_sb, group); -+ unsigned free, fragments, i, bits; -+ -+ J_ASSERT(cr >= 0 && cr < 4); -+ J_ASSERT(!EXT3_MB_GRP_NEED_INIT(grp)); -+ -+ free = grp->bb_free; -+ fragments = grp->bb_fragments; -+ if (free == 0) -+ return 0; -+ if (fragments == 0) -+ return 0; -+ -+ switch (cr) { -+ case 0: -+ J_ASSERT(ac->ac_2order != 0); -+ bits = ac->ac_sb->s_blocksize_bits + 1; -+ for (i = ac->ac_2order; i <= bits; i++) -+ if (grp->bb_counters[i] > 0) -+ return 1; -+ break; -+ 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; -+} -+ -+ext3_fsblk_t ext3_mb_new_blocks(handle_t *handle, struct inode *inode, -+ ext3_fsblk_t goal, int *len,int flags,int *errp) -+{ -+ struct buffer_head *bitmap_bh = NULL; -+ struct ext3_allocation_context ac; -+ int i, group, cr, err = 0; -+ struct ext3_group_desc *gdp; -+ struct ext3_super_block *es; -+ struct buffer_head *gdp_bh; -+ struct ext3_sb_info *sbi; -+ struct super_block *sb; -+ struct ext3_buddy e3b; -+ ext3_fsblk_t block; -+ -+ J_ASSERT(len != NULL); -+ J_ASSERT(*len > 0); -+ -+ sb = inode->i_sb; -+ if (!sb) { -+ printk("ext3_mb_new_nblocks: nonexistent device"); -+ return 0; -+ } -+ -+ if (!test_opt(sb, MBALLOC)) { -+ static int ext3_mballoc_warning = 0; -+ if (ext3_mballoc_warning == 0) { -+ printk(KERN_ERR "EXT3-fs: multiblock request with " -+ "mballoc disabled!\n"); -+ ext3_mballoc_warning++; -+ } -+ *len = 1; -+ err = ext3_new_block_old(handle, inode, goal, errp); -+ return err; -+ } -+ -+ ext3_mb_poll_new_transaction(sb, handle); -+ -+ sbi = EXT3_SB(sb); -+ es = EXT3_SB(sb)->s_es; -+ -+ /* -+ * We can't allocate > group size -+ */ -+ if (*len >= EXT3_BLOCKS_PER_GROUP(sb) - 10) -+ *len = EXT3_BLOCKS_PER_GROUP(sb) - 10; -+ -+ if (!(flags & EXT3_MB_HINT_RESERVED)) { -+ /* someone asks for non-reserved blocks */ -+ BUG_ON(*len > 1); -+ err = ext3_mb_reserve_blocks(sb, 1); -+ if (err) { -+ *errp = err; -+ return 0; -+ } -+ } -+ -+ ac.ac_buddy_page = NULL; -+ ac.ac_bitmap_page = NULL; -+ -+ /* -+ * Check quota for allocation of this blocks. -+ */ -+ while (*len && DQUOT_ALLOC_BLOCK(inode, *len)) -+ *len -= 1; -+ if (*len == 0) { -+ *errp = -EDQUOT; -+ block = 0; -+ goto out; -+ } -+ -+ /* start searching from the goal */ -+ if (goal < le32_to_cpu(es->s_first_data_block) || -+ goal >= le32_to_cpu(es->s_blocks_count)) -+ goal = le32_to_cpu(es->s_first_data_block); -+ group = (goal - le32_to_cpu(es->s_first_data_block)) / -+ EXT3_BLOCKS_PER_GROUP(sb); -+ block = ((goal - le32_to_cpu(es->s_first_data_block)) % -+ EXT3_BLOCKS_PER_GROUP(sb)); -+ -+ /* set up allocation goals */ -+ 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_groups_scanned = 0; -+ ac.ac_ex_scanned = 0; -+ ac.ac_found = 0; -+ ac.ac_sb = inode->i_sb; -+ ac.ac_g_ex.fe_group = group; -+ ac.ac_g_ex.fe_start = block; -+ ac.ac_g_ex.fe_len = *len; -+ ac.ac_flags = flags; -+ ac.ac_2order = 0; -+ ac.ac_criteria = 0; -+ -+ if (*len == 1 && sbi->s_stripe) { -+ /* looks like a metadata, let's use a dirty hack for raid5 -+ * move all metadata in first groups in hope to hit cached -+ * sectors and thus avoid read-modify cycles in raid5 */ -+ ac.ac_g_ex.fe_group = group = 0; -+ } -+ -+ /* probably, the request is for 2^8+ blocks (1/2/3/... MB) */ -+ i = ffs(*len); -+ if (i >= ext3_mb_order2_reqs) { -+ i--; -+ if ((*len & (~(1 << i))) == 0) -+ ac.ac_2order = i; -+ } -+ -+ /* first, try the goal */ -+ err = ext3_mb_find_by_goal(&ac, &e3b); -+ if (err) -+ goto out_err; -+ if (ac.ac_status == AC_STATUS_FOUND) -+ goto found; -+ -+ /* Let's just scan groups to find more-less suitable blocks */ -+ cr = ac.ac_2order ? 0 : 1; -+repeat: -+ for (; cr < 4 && ac.ac_status == AC_STATUS_CONTINUE; cr++) { -+ ac.ac_criteria = cr; -+ for (i = 0; i < EXT3_SB(sb)->s_groups_count; group++, i++) { -+ if (group == EXT3_SB(sb)->s_groups_count) -+ group = 0; -+ -+ if (EXT3_MB_GRP_NEED_INIT(EXT3_GROUP_INFO(sb, group))) { -+ /* we need full data about the group -+ * to make a good selection */ -+ err = ext3_mb_load_buddy(ac.ac_sb, group, &e3b); -+ if (err) -+ goto out_err; -+ ext3_mb_release_desc(&e3b); -+ } -+ -+ /* check is group good for our criteries */ -+ if (!ext3_mb_good_group(&ac, group, cr)) -+ continue; -+ -+ err = ext3_mb_load_buddy(ac.ac_sb, group, &e3b); -+ if (err) -+ goto out_err; -+ -+ ext3_lock_group(sb, group); -+ if (!ext3_mb_good_group(&ac, group, cr)) { -+ /* someone did allocation from this group */ -+ ext3_unlock_group(sb, group); -+ ext3_mb_release_desc(&e3b); -+ continue; -+ } -+ -+ ac.ac_groups_scanned++; -+ if (cr == 0) -+ ext3_mb_simple_scan_group(&ac, &e3b); -+ else if (cr == 1 && *len == sbi->s_stripe) -+ ext3_mb_scan_aligned(&ac, &e3b); -+ else -+ ext3_mb_complex_scan_group(&ac, &e3b); -+ -+ ext3_unlock_group(sb, group); -+ -+ ext3_mb_release_desc(&e3b); -+ -+ 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 & EXT3_MB_HINT_FIRST)) { -+ /* -+ * We've been searching too long. Let's try to allocate -+ * the best chunk we've found so far -+ */ -+ -+ /*if (ac.ac_found > ext3_mb_max_to_scan) -+ printk(KERN_DEBUG "EXT3-fs: too long searching at " -+ "%u (%d/%d)\n", cr, ac.ac_b_ex.fe_len, -+ ac.ac_g_ex.fe_len);*/ -+ ext3_mb_try_best_found(&ac, &e3b); -+ 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 "EXT3-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 |= EXT3_MB_HINT_FIRST; -+ cr = 3; -+ goto repeat; -+ } -+ } -+ -+ if (ac.ac_status != AC_STATUS_FOUND) { -+ /* -+ * We aren't lucky definitely -+ */ -+ DQUOT_FREE_BLOCK(inode, *len); -+ *errp = -ENOSPC; -+ block = 0; -+#if 1 -+ printk(KERN_ERR "EXT3-fs: can't allocate: status %d flags %d\n", -+ ac.ac_status, ac.ac_flags); -+ printk(KERN_ERR "EXT3-fs: goal %d, best found %d/%d/%d cr %d\n", -+ ac.ac_g_ex.fe_len, ac.ac_b_ex.fe_group, -+ ac.ac_b_ex.fe_start, ac.ac_b_ex.fe_len, cr); -+ printk(KERN_ERR "EXT3-fs: %lu block reserved, %d found\n", -+ sbi->s_blocks_reserved, ac.ac_found); -+ printk("EXT3-fs: groups: "); -+ for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) -+ printk("%d: %d ", i, EXT3_GROUP_INFO(sb, i)->bb_free); -+ printk("\n"); -+#endif -+ goto out; -+ } -+ -+found: -+ J_ASSERT(ac.ac_b_ex.fe_len > 0); -+ -+ /* good news - free block(s) have been found. now it's time -+ * to mark block(s) in good old journaled bitmap */ -+ block = ac.ac_b_ex.fe_group * EXT3_BLOCKS_PER_GROUP(sb) -+ + ac.ac_b_ex.fe_start -+ + le32_to_cpu(es->s_first_data_block); -+ -+ /* we made a desicion, now mark found blocks in good old -+ * bitmap to be journaled */ -+ -+ ext3_debug("using block group %d(%d)\n", -+ ac.ac_b_group.group, gdp->bg_free_blocks_count); -+ -+ bitmap_bh = read_block_bitmap(sb, ac.ac_b_ex.fe_group); -+ if (!bitmap_bh) { -+ *errp = -EIO; -+ goto out_err; -+ } -+ -+ err = ext3_journal_get_write_access(handle, bitmap_bh); -+ if (err) { -+ *errp = err; -+ goto out_err; -+ } -+ -+ gdp = ext3_get_group_desc(sb, ac.ac_b_ex.fe_group, &gdp_bh); -+ if (!gdp) { -+ *errp = -EIO; -+ goto out_err; -+ } -+ -+ err = ext3_journal_get_write_access(handle, gdp_bh); -+ if (err) -+ goto out_err; -+ -+ block = ac.ac_b_ex.fe_group * EXT3_BLOCKS_PER_GROUP(sb) -+ + ac.ac_b_ex.fe_start -+ + le32_to_cpu(es->s_first_data_block); -+ -+ if (block == le32_to_cpu(gdp->bg_block_bitmap) || -+ block == le32_to_cpu(gdp->bg_inode_bitmap) || -+ in_range(block, le32_to_cpu(gdp->bg_inode_table), -+ EXT3_SB(sb)->s_itb_per_group)) -+ ext3_error(sb, "ext3_new_block", -+ "Allocating block in system zone - " -+ "block = "E3FSBLK, block); -+#ifdef AGGRESSIVE_CHECK -+ for (i = 0; i < ac.ac_b_ex.fe_len; i++) -+ J_ASSERT(!mb_test_bit(ac.ac_b_ex.fe_start + i, bitmap_bh->b_data)); -+#endif -+ mb_set_bits(bitmap_bh->b_data, ac.ac_b_ex.fe_start, ac.ac_b_ex.fe_len); -+ -+ spin_lock(sb_bgl_lock(sbi, ac.ac_b_ex.fe_group)); -+ gdp->bg_free_blocks_count = -+ cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count) -+ - ac.ac_b_ex.fe_len); -+ spin_unlock(sb_bgl_lock(sbi, ac.ac_b_ex.fe_group)); -+ percpu_counter_mod(&sbi->s_freeblocks_counter, - ac.ac_b_ex.fe_len); -+ -+ err = ext3_journal_dirty_metadata(handle, bitmap_bh); -+ if (err) -+ goto out_err; -+ err = ext3_journal_dirty_metadata(handle, gdp_bh); -+ if (err) -+ goto out_err; -+ -+ sb->s_dirt = 1; -+ *errp = 0; -+ brelse(bitmap_bh); -+ -+ /* drop non-allocated, but dquote'd blocks */ -+ J_ASSERT(*len >= ac.ac_b_ex.fe_len); -+ DQUOT_FREE_BLOCK(inode, *len - ac.ac_b_ex.fe_len); -+ -+ *len = ac.ac_b_ex.fe_len; -+ J_ASSERT(*len > 0); -+ J_ASSERT(block != 0); -+ goto out; -+ -+out_err: -+ /* if we've already allocated something, roll it back */ -+ if (ac.ac_status == AC_STATUS_FOUND) { -+ /* FIXME: free blocks here */ -+ } -+ -+ DQUOT_FREE_BLOCK(inode, *len); -+ brelse(bitmap_bh); -+ *errp = err; -+ block = 0; -+out: -+ if (ac.ac_buddy_page) -+ page_cache_release(ac.ac_buddy_page); -+ if (ac.ac_bitmap_page) -+ page_cache_release(ac.ac_bitmap_page); -+ -+ if (!(flags & EXT3_MB_HINT_RESERVED)) { -+ /* block wasn't reserved before and we reserved it -+ * at the beginning of allocation. it doesn't matter -+ * whether we allocated anything or we failed: time -+ * to release reservation. NOTE: because I expect -+ * any multiblock request from delayed allocation -+ * path only, here is single block always */ -+ ext3_mb_release_blocks(sb, 1); -+ } -+ -+ if (unlikely(ext3_mb_stats) && ac.ac_g_ex.fe_len > 1) { -+ atomic_inc(&sbi->s_bal_reqs); -+ atomic_add(*len, &sbi->s_bal_allocated); -+ if (*len >= ac.ac_g_ex.fe_len) -+ atomic_inc(&sbi->s_bal_success); -+ atomic_add(ac.ac_found, &sbi->s_bal_ex_scanned); -+ if (ac.ac_g_ex.fe_start == ac.ac_b_ex.fe_start && -+ ac.ac_g_ex.fe_group == ac.ac_b_ex.fe_group) -+ atomic_inc(&sbi->s_bal_goals); -+ if (ac.ac_found > ext3_mb_max_to_scan) -+ atomic_inc(&sbi->s_bal_breaks); -+ } -+ -+ ext3_mb_store_history(sb, inode->i_ino, &ac); -+ -+ return block; -+} -+EXPORT_SYMBOL(ext3_mb_new_blocks); -+ -+#ifdef EXT3_MB_HISTORY -+struct ext3_mb_proc_session { -+ struct ext3_mb_history *history; -+ struct super_block *sb; -+ int start; -+ int max; -+}; -+ -+static void *ext3_mb_history_skip_empty(struct ext3_mb_proc_session *s, -+ struct ext3_mb_history *hs, -+ int first) -+{ -+ if (hs == s->history + s->max) -+ hs = s->history; -+ if (!first && hs == s->history + s->start) -+ return NULL; -+ while (hs->goal.fe_len == 0) { -+ hs++; -+ if (hs == s->history + s->max) -+ hs = s->history; -+ if (hs == s->history + s->start) -+ return NULL; -+ } -+ return hs; -+} -+ -+static void *ext3_mb_seq_history_start(struct seq_file *seq, loff_t *pos) -+{ -+ struct ext3_mb_proc_session *s = seq->private; -+ struct ext3_mb_history *hs; -+ int l = *pos; -+ -+ if (l == 0) -+ return SEQ_START_TOKEN; -+ hs = ext3_mb_history_skip_empty(s, s->history + s->start, 1); -+ if (!hs) -+ return NULL; -+ while (--l && (hs = ext3_mb_history_skip_empty(s, ++hs, 0)) != NULL); -+ return hs; -+} -+ -+static void *ext3_mb_seq_history_next(struct seq_file *seq, void *v, loff_t *pos) -+{ -+ struct ext3_mb_proc_session *s = seq->private; -+ struct ext3_mb_history *hs = v; -+ -+ ++*pos; -+ if (v == SEQ_START_TOKEN) -+ return ext3_mb_history_skip_empty(s, s->history + s->start, 1); -+ else -+ return ext3_mb_history_skip_empty(s, ++hs, 0); -+} -+ -+static int ext3_mb_seq_history_show(struct seq_file *seq, void *v) -+{ -+ struct ext3_mb_history *hs = v; -+ char buf[20], buf2[20]; -+ -+ if (v == SEQ_START_TOKEN) { -+ seq_printf(seq, "%-5s %-8s %-17s %-17s %-5s %-5s %-2s %-5s %-5s %-6s\n", -+ "pid", "inode", "goal", "result", "found", "grps", "cr", -+ "merge", "tail", "broken"); -+ return 0; -+ } -+ -+ sprintf(buf, "%u/%u/%u", hs->goal.fe_group, -+ hs->goal.fe_start, hs->goal.fe_len); -+ sprintf(buf2, "%u/%u/%u", hs->result.fe_group, -+ hs->result.fe_start, hs->result.fe_len); -+ seq_printf(seq, "%-5u %-8u %-17s %-17s %-5u %-5u %-2u %-5s %-5u %-6u\n", -+ hs->pid, hs->ino, buf, buf2, hs->found, hs->groups, -+ hs->cr, hs->merged ? "M" : "", hs->tail, -+ hs->buddy ? 1 << hs->buddy : 0); -+ return 0; -+} -+ -+static void ext3_mb_seq_history_stop(struct seq_file *seq, void *v) -+{ -+} -+ -+static struct seq_operations ext3_mb_seq_history_ops = { -+ .start = ext3_mb_seq_history_start, -+ .next = ext3_mb_seq_history_next, -+ .stop = ext3_mb_seq_history_stop, -+ .show = ext3_mb_seq_history_show, -+}; -+ -+static int ext3_mb_seq_history_open(struct inode *inode, struct file *file) -+{ -+ struct super_block *sb = PDE(inode)->data; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct ext3_mb_proc_session *s; -+ int rc, size; -+ -+ s = kmalloc(sizeof(*s), GFP_KERNEL); -+ if (s == NULL) -+ return -EIO; -+ size = sizeof(struct ext3_mb_history) * sbi->s_mb_history_max; -+ s->history = kmalloc(size, GFP_KERNEL); -+ if (s == NULL) { -+ kfree(s); -+ return -EIO; -+ } -+ -+ spin_lock(&sbi->s_mb_history_lock); -+ memcpy(s->history, sbi->s_mb_history, size); -+ s->max = sbi->s_mb_history_max; -+ s->start = sbi->s_mb_history_cur % s->max; -+ spin_unlock(&sbi->s_mb_history_lock); -+ -+ rc = seq_open(file, &ext3_mb_seq_history_ops); -+ if (rc == 0) { -+ struct seq_file *m = (struct seq_file *)file->private_data; -+ m->private = s; -+ } else { -+ kfree(s->history); -+ kfree(s); -+ } -+ return rc; -+ -+} -+ -+static int ext3_mb_seq_history_release(struct inode *inode, struct file *file) -+{ -+ struct seq_file *seq = (struct seq_file *)file->private_data; -+ struct ext3_mb_proc_session *s = seq->private; -+ kfree(s->history); -+ kfree(s); -+ return seq_release(inode, file); -+} -+ -+static struct file_operations ext3_mb_seq_history_fops = { -+ .owner = THIS_MODULE, -+ .open = ext3_mb_seq_history_open, -+ .read = seq_read, -+ .llseek = seq_lseek, -+ .release = ext3_mb_seq_history_release, -+}; -+ -+static void *ext3_mb_seq_groups_start(struct seq_file *seq, loff_t *pos) -+{ -+ struct super_block *sb = seq->private; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ long group; -+ -+ if (*pos < 0 || *pos >= sbi->s_groups_count) -+ return NULL; -+ -+ group = *pos + 1; -+ return (void *) group; -+} -+ -+static void *ext3_mb_seq_groups_next(struct seq_file *seq, void *v, loff_t *pos) -+{ -+ struct super_block *sb = seq->private; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ long group; -+ -+ ++*pos; -+ if (*pos < 0 || *pos >= sbi->s_groups_count) -+ return NULL; -+ group = *pos + 1; -+ return (void *) group;; -+} -+ -+static int ext3_mb_seq_groups_show(struct seq_file *seq, void *v) -+{ -+ struct super_block *sb = seq->private; -+ long group = (long) v, i; -+ struct sg { -+ struct ext3_group_info info; -+ unsigned short 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 ext3_group_info); -+ ext3_lock_group(sb, group); -+ memcpy(&sg, EXT3_GROUP_INFO(sb, group), i); -+ ext3_unlock_group(sb, group); -+ -+ if (EXT3_MB_GRP_NEED_INIT(&sg.info)) -+ return 0; -+ -+ seq_printf(seq, "#%-5lu: %-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 ext3_mb_seq_groups_stop(struct seq_file *seq, void *v) -+{ -+} -+ -+static struct seq_operations ext3_mb_seq_groups_ops = { -+ .start = ext3_mb_seq_groups_start, -+ .next = ext3_mb_seq_groups_next, -+ .stop = ext3_mb_seq_groups_stop, -+ .show = ext3_mb_seq_groups_show, -+}; -+ -+static int ext3_mb_seq_groups_open(struct inode *inode, struct file *file) -+{ -+ struct super_block *sb = PDE(inode)->data; -+ int rc; -+ -+ rc = seq_open(file, &ext3_mb_seq_groups_ops); -+ if (rc == 0) { -+ struct seq_file *m = (struct seq_file *)file->private_data; -+ m->private = sb; -+ } -+ return rc; -+ -+} -+ -+static struct file_operations ext3_mb_seq_groups_fops = { -+ .owner = THIS_MODULE, -+ .open = ext3_mb_seq_groups_open, -+ .read = seq_read, -+ .llseek = seq_lseek, -+ .release = seq_release, -+}; -+ -+static void ext3_mb_history_release(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ char name[64]; -+ -+ snprintf(name, sizeof(name) - 1, "%s", bdevname(sb->s_bdev, name)); -+ remove_proc_entry("mb_groups", sbi->s_mb_proc); -+ remove_proc_entry("mb_history", sbi->s_mb_proc); -+ remove_proc_entry(name, proc_root_ext3); -+ -+ if (sbi->s_mb_history) -+ kfree(sbi->s_mb_history); -+} -+ -+static void ext3_mb_history_init(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ char name[64]; -+ int i; -+ -+ snprintf(name, sizeof(name) - 1, "%s", bdevname(sb->s_bdev, name)); -+ sbi->s_mb_proc = proc_mkdir(name, proc_root_ext3); -+ if (sbi->s_mb_proc != NULL) { -+ struct proc_dir_entry *p; -+ p = create_proc_entry("mb_history", S_IRUGO, sbi->s_mb_proc); -+ if (p) { -+ p->proc_fops = &ext3_mb_seq_history_fops; -+ p->data = sb; -+ } -+ p = create_proc_entry("mb_groups", S_IRUGO, sbi->s_mb_proc); -+ if (p) { -+ p->proc_fops = &ext3_mb_seq_groups_fops; -+ p->data = sb; -+ } -+ } -+ -+ sbi->s_mb_history_max = 1000; -+ sbi->s_mb_history_cur = 0; -+ spin_lock_init(&sbi->s_mb_history_lock); -+ i = sbi->s_mb_history_max * sizeof(struct ext3_mb_history); -+ sbi->s_mb_history = kmalloc(i, GFP_KERNEL); -+ memset(sbi->s_mb_history, 0, i); -+ /* if we can't allocate history, then we simple won't use it */ -+} -+ -+static void -+ext3_mb_store_history(struct super_block *sb, unsigned ino, -+ struct ext3_allocation_context *ac) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct ext3_mb_history h; -+ -+ if (likely(sbi->s_mb_history == NULL)) -+ return; -+ -+ h.pid = current->pid; -+ h.ino = ino; -+ h.goal = ac->ac_g_ex; -+ h.result = ac->ac_b_ex; -+ h.found = ac->ac_found; -+ h.cr = ac->ac_criteria; -+ h.groups = ac->ac_groups_scanned; -+ h.tail = ac->ac_tail; -+ h.buddy = ac->ac_buddy; -+ h.merged = 0; -+ if (ac->ac_g_ex.fe_start == ac->ac_b_ex.fe_start && -+ ac->ac_g_ex.fe_group == ac->ac_b_ex.fe_group) -+ h.merged = 1; -+ -+ spin_lock(&sbi->s_mb_history_lock); -+ memcpy(sbi->s_mb_history + sbi->s_mb_history_cur, &h, sizeof(h)); -+ if (++sbi->s_mb_history_cur >= sbi->s_mb_history_max) -+ sbi->s_mb_history_cur = 0; -+ spin_unlock(&sbi->s_mb_history_lock); -+} -+ -+#else -+#define ext3_mb_history_release(sb) -+#define ext3_mb_history_init(sb) -+#endif -+ -+int ext3_mb_init_backend(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ int i, j, len, metalen; -+ int num_meta_group_infos = -+ (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) >> -+ EXT3_DESC_PER_BLOCK_BITS(sb); -+ struct ext3_group_info **meta_group_info; -+ -+ /* 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 = kmalloc(sizeof(*sbi->s_group_info) * -+ num_meta_group_infos, GFP_KERNEL); -+ if (sbi->s_group_info == NULL) { -+ printk(KERN_ERR "EXT3-fs: can't allocate buddy meta group\n"); -+ return -ENOMEM; -+ } -+ sbi->s_buddy_cache = new_inode(sb); -+ if (sbi->s_buddy_cache == NULL) { -+ printk(KERN_ERR "EXT3-fs: can't get new inode\n"); -+ goto err_freesgi; -+ } -+ -+ metalen = sizeof(*meta_group_info) << EXT3_DESC_PER_BLOCK_BITS(sb); -+ for (i = 0; i < num_meta_group_infos; i++) { -+ if ((i + 1) == num_meta_group_infos) -+ metalen = sizeof(*meta_group_info) * -+ (sbi->s_groups_count - -+ (i << EXT3_DESC_PER_BLOCK_BITS(sb))); -+ meta_group_info = kmalloc(metalen, GFP_KERNEL); -+ if (meta_group_info == NULL) { -+ printk(KERN_ERR "EXT3-fs: can't allocate mem for a " -+ "buddy group\n"); -+ goto err_freemeta; -+ } -+ sbi->s_group_info[i] = meta_group_info; -+ } -+ -+ /* -+ * calculate needed size. if change bb_counters size, -+ * don't forget about ext3_mb_generate_buddy() -+ */ -+ len = sizeof(struct ext3_group_info); -+ len += sizeof(unsigned short) * (sb->s_blocksize_bits + 2); -+ for (i = 0; i < sbi->s_groups_count; i++) { -+ struct ext3_group_desc * desc; -+ -+ meta_group_info = -+ sbi->s_group_info[i >> EXT3_DESC_PER_BLOCK_BITS(sb)]; -+ j = i & (EXT3_DESC_PER_BLOCK(sb) - 1); -+ -+ meta_group_info[j] = kmalloc(len, GFP_KERNEL); -+ if (meta_group_info[j] == NULL) { -+ printk(KERN_ERR "EXT3-fs: can't allocate buddy mem\n"); -+ i--; -+ goto err_freebuddy; -+ } -+ desc = ext3_get_group_desc(sb, i, NULL); -+ if (desc == NULL) { -+ printk(KERN_ERR"EXT3-fs: can't read descriptor %u\n",i); -+ goto err_freebuddy; -+ } -+ memset(meta_group_info[j], 0, len); -+ set_bit(EXT3_GROUP_INFO_NEED_INIT_BIT, -+ &meta_group_info[j]->bb_state); -+ meta_group_info[j]->bb_free = -+ le16_to_cpu(desc->bg_free_blocks_count); -+ } -+ -+ return 0; -+ -+err_freebuddy: -+ while (i >= 0) { -+ kfree(EXT3_GROUP_INFO(sb, i)); -+ i--; -+ } -+ i = num_meta_group_infos; -+err_freemeta: -+ while (--i >= 0) -+ kfree(sbi->s_group_info[i]); -+ iput(sbi->s_buddy_cache); -+err_freesgi: -+ kfree(sbi->s_group_info); -+ return -ENOMEM; -+} -+ -+int ext3_mb_init(struct super_block *sb, int needs_recovery) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct inode *root = sb->s_root->d_inode; -+ unsigned i, offset, max; -+ struct dentry *dentry; -+ -+ if (!test_opt(sb, MBALLOC)) -+ return 0; -+ -+ i = (sb->s_blocksize_bits + 2) * sizeof(unsigned short); -+ -+ sbi->s_mb_offsets = kmalloc(i, GFP_KERNEL); -+ if (sbi->s_mb_offsets == NULL) { -+ clear_opt(sbi->s_mount_opt, MBALLOC); -+ return -ENOMEM; -+ } -+ sbi->s_mb_maxs = kmalloc(i, GFP_KERNEL); -+ if (sbi->s_mb_maxs == NULL) { -+ clear_opt(sbi->s_mount_opt, MBALLOC); -+ kfree(sbi->s_mb_maxs); -+ return -ENOMEM; -+ } -+ -+ /* 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); -+ -+ /* init file for buddy data */ -+ if ((i = ext3_mb_init_backend(sb))) { -+ clear_opt(sbi->s_mount_opt, MBALLOC); -+ kfree(sbi->s_mb_offsets); -+ kfree(sbi->s_mb_maxs); -+ return i; -+ } -+ -+ spin_lock_init(&sbi->s_reserve_lock); -+ spin_lock_init(&sbi->s_md_lock); -+ INIT_LIST_HEAD(&sbi->s_active_transaction); -+ INIT_LIST_HEAD(&sbi->s_closed_transaction); -+ INIT_LIST_HEAD(&sbi->s_committed_transaction); -+ spin_lock_init(&sbi->s_bal_lock); -+ -+ /* remove old on-disk buddy file */ -+ mutex_lock(&root->i_mutex); -+ dentry = lookup_one_len(".buddy", sb->s_root, strlen(".buddy")); -+ if (dentry->d_inode != NULL) { -+ i = vfs_unlink(root, dentry); -+ if (i != 0) -+ printk("EXT3-fs: can't remove .buddy file: %d\n", i); -+ } -+ dput(dentry); -+ mutex_unlock(&root->i_mutex); -+ -+ ext3_mb_history_init(sb); -+ -+ printk("EXT3-fs: mballoc enabled\n"); -+ return 0; -+} -+ -+int ext3_mb_release(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ int i, num_meta_group_infos; -+ -+ if (!test_opt(sb, MBALLOC)) -+ return 0; -+ -+ /* release freed, non-committed blocks */ -+ spin_lock(&sbi->s_md_lock); -+ list_splice_init(&sbi->s_closed_transaction, -+ &sbi->s_committed_transaction); -+ list_splice_init(&sbi->s_active_transaction, -+ &sbi->s_committed_transaction); -+ spin_unlock(&sbi->s_md_lock); -+ ext3_mb_free_committed_blocks(sb); -+ -+ if (sbi->s_group_info) { -+ for (i = 0; i < sbi->s_groups_count; i++) -+ kfree(EXT3_GROUP_INFO(sb, i)); -+ num_meta_group_infos = (sbi->s_groups_count + -+ EXT3_DESC_PER_BLOCK(sb) - 1) >> -+ EXT3_DESC_PER_BLOCK_BITS(sb); -+ for (i = 0; i < num_meta_group_infos; i++) -+ kfree(sbi->s_group_info[i]); -+ kfree(sbi->s_group_info); -+ } -+ if (sbi->s_mb_offsets) -+ kfree(sbi->s_mb_offsets); -+ if (sbi->s_mb_maxs) -+ kfree(sbi->s_mb_maxs); -+ if (sbi->s_buddy_cache) -+ iput(sbi->s_buddy_cache); -+ if (sbi->s_blocks_reserved) -+ printk("ext3-fs: %ld blocks being reserved at umount!\n", -+ sbi->s_blocks_reserved); -+ if (ext3_mb_stats) { -+ printk("EXT3-fs: mballoc: %u blocks %u reqs (%u success)\n", -+ atomic_read(&sbi->s_bal_allocated), -+ atomic_read(&sbi->s_bal_reqs), -+ atomic_read(&sbi->s_bal_success)); -+ printk("EXT3-fs: mballoc: %u extents scanned, %u goal hits, " -+ "%u 2^N hits, %u breaks\n", -+ 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)); -+ printk("EXT3-fs: mballoc: %lu generated and it took %Lu\n", -+ sbi->s_mb_buddies_generated++, -+ sbi->s_mb_generation_time); -+ } -+ -+ ext3_mb_history_release(sb); -+ -+ return 0; -+} -+ -+void ext3_mb_free_committed_blocks(struct super_block *sb) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ int err, i, count = 0, count2 = 0; -+ struct ext3_free_metadata *md; -+ struct ext3_buddy e3b; -+ -+ if (list_empty(&sbi->s_committed_transaction)) -+ return; -+ -+ /* there is committed blocks to be freed yet */ -+ do { -+ /* get next array of blocks */ -+ md = NULL; -+ spin_lock(&sbi->s_md_lock); -+ if (!list_empty(&sbi->s_committed_transaction)) { -+ md = list_entry(sbi->s_committed_transaction.next, -+ struct ext3_free_metadata, list); -+ list_del(&md->list); -+ } -+ spin_unlock(&sbi->s_md_lock); -+ -+ if (md == NULL) -+ break; -+ -+ mb_debug("gonna free %u blocks in group %u (0x%p):", -+ md->num, md->group, md); -+ -+ err = ext3_mb_load_buddy(sb, md->group, &e3b); -+ /* we expect to find existing buddy because it's pinned */ -+ BUG_ON(err != 0); -+ -+ /* there are blocks to put in buddy to make them really free */ -+ count += md->num; -+ count2++; -+ ext3_lock_group(sb, md->group); -+ for (i = 0; i < md->num; i++) { -+ mb_debug(" %u", md->blocks[i]); -+ mb_free_blocks(&e3b, md->blocks[i], 1); -+ } -+ mb_debug("\n"); -+ ext3_unlock_group(sb, md->group); -+ -+ /* balance refcounts from ext3_mb_free_metadata() */ -+ page_cache_release(e3b.bd_buddy_page); -+ page_cache_release(e3b.bd_bitmap_page); -+ -+ kfree(md); -+ ext3_mb_release_desc(&e3b); -+ -+ } while (md); -+ mb_debug("freed %u blocks in %u structures\n", count, count2); -+} -+ -+void ext3_mb_poll_new_transaction(struct super_block *sb, handle_t *handle) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ -+ if (sbi->s_last_transaction == handle->h_transaction->t_tid) -+ return; -+ -+ /* new transaction! time to close last one and free blocks for -+ * committed transaction. we know that only transaction can be -+ * active, so previos transaction can be being logged and we -+ * know that transaction before previous is known to be already -+ * logged. this means that now we may free blocks freed in all -+ * transactions before previous one. hope I'm clear enough ... */ -+ -+ spin_lock(&sbi->s_md_lock); -+ if (sbi->s_last_transaction != handle->h_transaction->t_tid) { -+ mb_debug("new transaction %lu, old %lu\n", -+ (unsigned long) handle->h_transaction->t_tid, -+ (unsigned long) sbi->s_last_transaction); -+ list_splice_init(&sbi->s_closed_transaction, -+ &sbi->s_committed_transaction); -+ list_splice_init(&sbi->s_active_transaction, -+ &sbi->s_closed_transaction); -+ sbi->s_last_transaction = handle->h_transaction->t_tid; -+ } -+ spin_unlock(&sbi->s_md_lock); -+ -+ ext3_mb_free_committed_blocks(sb); -+} -+ -+int ext3_mb_free_metadata(handle_t *handle, struct ext3_buddy *e3b, -+ int group, int block, int count) -+{ -+ struct ext3_group_info *db = e3b->bd_info; -+ struct super_block *sb = e3b->bd_sb; -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ struct ext3_free_metadata *md; -+ int i; -+ -+ J_ASSERT(e3b->bd_bitmap_page != NULL); -+ J_ASSERT(e3b->bd_buddy_page != NULL); -+ -+ ext3_lock_group(sb, group); -+ for (i = 0; i < count; i++) { -+ md = db->bb_md_cur; -+ if (md && db->bb_tid != handle->h_transaction->t_tid) { -+ db->bb_md_cur = NULL; -+ md = NULL; -+ } -+ -+ if (md == NULL) { -+ ext3_unlock_group(sb, group); -+ md = kmalloc(sizeof(*md), GFP_KERNEL); -+ if (md == NULL) -+ return -ENOMEM; -+ md->num = 0; -+ md->group = group; -+ -+ ext3_lock_group(sb, group); -+ if (db->bb_md_cur == NULL) { -+ spin_lock(&sbi->s_md_lock); -+ list_add(&md->list, &sbi->s_active_transaction); -+ spin_unlock(&sbi->s_md_lock); -+ /* protect buddy cache from being freed, -+ * otherwise we'll refresh it from -+ * on-disk bitmap and lose not-yet-available -+ * blocks */ -+ page_cache_get(e3b->bd_buddy_page); -+ page_cache_get(e3b->bd_bitmap_page); -+ db->bb_md_cur = md; -+ db->bb_tid = handle->h_transaction->t_tid; -+ mb_debug("new md 0x%p for group %u\n", -+ md, md->group); -+ } else { -+ kfree(md); -+ md = db->bb_md_cur; -+ } -+ } -+ -+ BUG_ON(md->num >= EXT3_BB_MAX_BLOCKS); -+ md->blocks[md->num] = block + i; -+ md->num++; -+ if (md->num == EXT3_BB_MAX_BLOCKS) { -+ /* no more space, put full container on a sb's list */ -+ db->bb_md_cur = NULL; -+ } -+ } -+ ext3_unlock_group(sb, group); -+ return 0; -+} -+ -+void ext3_mb_free_blocks(handle_t *handle, struct inode *inode, -+ ext3_fsblk_t block, unsigned long count, -+ int metadata, unsigned long *freed) -+{ -+ struct buffer_head *bitmap_bh = NULL; -+ struct ext3_group_desc *gdp; -+ struct ext3_super_block *es; -+ unsigned long bit, overflow; -+ struct buffer_head *gd_bh; -+ unsigned long block_group; -+ struct ext3_sb_info *sbi; -+ struct super_block *sb; -+ struct ext3_buddy e3b; -+ int err = 0, ret; -+ -+ *freed = 0; -+ sb = inode->i_sb; -+ if (!sb) { -+ printk ("ext3_free_blocks: nonexistent device"); -+ return; -+ } -+ -+ ext3_mb_poll_new_transaction(sb, handle); -+ -+ sbi = EXT3_SB(sb); -+ es = EXT3_SB(sb)->s_es; -+ if (block < le32_to_cpu(es->s_first_data_block) || -+ block + count < block || -+ block + count > le32_to_cpu(es->s_blocks_count)) { -+ ext3_error (sb, "ext3_free_blocks", -+ "Freeing blocks not in datazone - " -+ "block = %lu, count = %lu", block, count); -+ goto error_return; -+ } -+ -+ ext3_debug("freeing block %lu\n", block); -+ -+do_more: -+ overflow = 0; -+ block_group = (block - le32_to_cpu(es->s_first_data_block)) / -+ EXT3_BLOCKS_PER_GROUP(sb); -+ bit = (block - le32_to_cpu(es->s_first_data_block)) % -+ EXT3_BLOCKS_PER_GROUP(sb); -+ /* -+ * Check to see if we are freeing blocks across a group -+ * boundary. -+ */ -+ if (bit + count > EXT3_BLOCKS_PER_GROUP(sb)) { -+ overflow = bit + count - EXT3_BLOCKS_PER_GROUP(sb); -+ count -= overflow; -+ } -+ brelse(bitmap_bh); -+ bitmap_bh = read_block_bitmap(sb, block_group); -+ if (!bitmap_bh) -+ goto error_return; -+ gdp = ext3_get_group_desc (sb, block_group, &gd_bh); -+ if (!gdp) -+ goto error_return; -+ -+ if (in_range (le32_to_cpu(gdp->bg_block_bitmap), block, count) || -+ in_range (le32_to_cpu(gdp->bg_inode_bitmap), block, count) || -+ in_range (block, le32_to_cpu(gdp->bg_inode_table), -+ EXT3_SB(sb)->s_itb_per_group) || -+ in_range (block + count - 1, le32_to_cpu(gdp->bg_inode_table), -+ EXT3_SB(sb)->s_itb_per_group)) -+ ext3_error (sb, "ext3_free_blocks", -+ "Freeing blocks in system zones - " -+ "Block = %lu, count = %lu", -+ block, count); -+ -+ BUFFER_TRACE(bitmap_bh, "getting write access"); -+ err = ext3_journal_get_write_access(handle, bitmap_bh); -+ if (err) -+ goto error_return; -+ -+ /* -+ * We are about to modify some metadata. Call the journal APIs -+ * to unshare ->b_data if a currently-committing transaction is -+ * using it -+ */ -+ BUFFER_TRACE(gd_bh, "get_write_access"); -+ err = ext3_journal_get_write_access(handle, gd_bh); -+ if (err) -+ goto error_return; -+ -+ err = ext3_mb_load_buddy(sb, block_group, &e3b); -+ if (err) -+ goto error_return; -+ -+#ifdef AGGRESSIVE_CHECK -+ { -+ int i; -+ for (i = 0; i < count; i++) -+ J_ASSERT(mb_test_bit(bit + i, bitmap_bh->b_data)); -+ } -+#endif -+ mb_clear_bits(bitmap_bh->b_data, bit, count); -+ -+ /* We dirtied the bitmap block */ -+ BUFFER_TRACE(bitmap_bh, "dirtied bitmap block"); -+ err = ext3_journal_dirty_metadata(handle, bitmap_bh); -+ -+ if (metadata) { -+ /* blocks being freed are metadata. these blocks shouldn't -+ * be used until this transaction is committed */ -+ ext3_mb_free_metadata(handle, &e3b, block_group, bit, count); -+ } else { -+ ext3_lock_group(sb, block_group); -+ mb_free_blocks(&e3b, bit, count); -+ ext3_unlock_group(sb, block_group); -+ } -+ -+ spin_lock(sb_bgl_lock(sbi, block_group)); -+ gdp->bg_free_blocks_count = -+ cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count) + count); -+ spin_unlock(sb_bgl_lock(sbi, block_group)); -+ percpu_counter_mod(&sbi->s_freeblocks_counter, count); -+ -+ ext3_mb_release_desc(&e3b); -+ -+ *freed = count; -+ -+ /* And the group descriptor block */ -+ BUFFER_TRACE(gd_bh, "dirtied group descriptor block"); -+ ret = ext3_journal_dirty_metadata(handle, gd_bh); -+ if (!err) err = ret; -+ -+ if (overflow && !err) { -+ block += count; -+ count = overflow; -+ goto do_more; -+ } -+ sb->s_dirt = 1; -+error_return: -+ brelse(bitmap_bh); -+ ext3_std_error(sb, err); -+ return; -+} -+ -+int ext3_mb_reserve_blocks(struct super_block *sb, int blocks) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ int free, ret = -ENOSPC; -+ -+ BUG_ON(blocks < 0); -+ spin_lock(&sbi->s_reserve_lock); -+ free = percpu_counter_read_positive(&sbi->s_freeblocks_counter); -+ if (blocks <= free - sbi->s_blocks_reserved) { -+ sbi->s_blocks_reserved += blocks; -+ ret = 0; -+ } -+ spin_unlock(&sbi->s_reserve_lock); -+ return ret; -+} -+ -+void ext3_mb_release_blocks(struct super_block *sb, int blocks) -+{ -+ struct ext3_sb_info *sbi = EXT3_SB(sb); -+ -+ BUG_ON(blocks < 0); -+ spin_lock(&sbi->s_reserve_lock); -+ sbi->s_blocks_reserved -= blocks; -+ WARN_ON(sbi->s_blocks_reserved < 0); -+ if (sbi->s_blocks_reserved < 0) -+ sbi->s_blocks_reserved = 0; -+ spin_unlock(&sbi->s_reserve_lock); -+} -+ -+ext3_fsblk_t ext3_new_block(handle_t *handle, struct inode *inode, -+ ext3_fsblk_t goal, int *errp) -+{ -+ ext3_fsblk_t ret; -+ int len; -+ -+ if (!test_opt(inode->i_sb, MBALLOC)) { -+ ret = ext3_new_block_old(handle, inode, goal, errp); -+ goto out; -+ } -+ len = 1; -+ ret = ext3_mb_new_blocks(handle, inode, goal, &len, 0, errp); -+out: -+ return ret; -+} -+ -+void ext3_free_blocks(handle_t *handle, struct inode * inode, -+ ext3_fsblk_t block, unsigned long count, int metadata) -+{ -+ struct super_block *sb; -+ unsigned long freed; -+ -+ sb = inode->i_sb; -+ if (!test_opt(sb, MBALLOC) || !EXT3_SB(sb)->s_group_info) -+ ext3_free_blocks_sb(handle, sb, block, count, &freed); -+ else -+ ext3_mb_free_blocks(handle, inode, block, count, metadata, -+ &freed); -+ if (freed) -+ DQUOT_FREE_BLOCK(inode, freed); -+ return; -+} -+ -+#define EXT3_ROOT "ext3" -+#define EXT3_MB_STATS_NAME "mb_stats" -+#define EXT3_MB_MAX_TO_SCAN_NAME "mb_max_to_scan" -+#define EXT3_MB_MIN_TO_SCAN_NAME "mb_min_to_scan" -+#define EXT3_MB_ORDER2_REQ "mb_order2_req" -+ -+static int ext3_mb_stats_read(char *page, char **start, off_t off, -+ int count, int *eof, void *data) -+{ -+ int len; -+ -+ *eof = 1; -+ if (off != 0) -+ return 0; -+ -+ len = sprintf(page, "%ld\n", ext3_mb_stats); -+ *start = page; -+ return len; -+} -+ -+static int ext3_mb_stats_write(struct file *file, const char *buffer, -+ unsigned long count, void *data) -+{ -+ char str[32]; -+ -+ if (count >= sizeof(str)) { -+ printk(KERN_ERR "EXT3-fs: %s string too long, max %u bytes\n", -+ EXT3_MB_STATS_NAME, (int)sizeof(str)); -+ return -EOVERFLOW; -+ } -+ -+ if (copy_from_user(str, buffer, count)) -+ return -EFAULT; -+ -+ /* Only set to 0 or 1 respectively; zero->0; non-zero->1 */ -+ ext3_mb_stats = (simple_strtol(str, NULL, 0) != 0); -+ return count; -+} -+ -+static int ext3_mb_max_to_scan_read(char *page, char **start, off_t off, -+ int count, int *eof, void *data) -+{ -+ int len; -+ -+ *eof = 1; -+ if (off != 0) -+ return 0; -+ -+ len = sprintf(page, "%ld\n", ext3_mb_max_to_scan); -+ *start = page; -+ return len; -+} -+ -+static int ext3_mb_max_to_scan_write(struct file *file, const char *buffer, -+ unsigned long count, void *data) -+{ -+ char str[32]; -+ long value; -+ -+ if (count >= sizeof(str)) { -+ printk(KERN_ERR "EXT3-fs: %s string too long, max %u bytes\n", -+ EXT3_MB_MAX_TO_SCAN_NAME, (int)sizeof(str)); -+ return -EOVERFLOW; -+ } -+ -+ if (copy_from_user(str, buffer, count)) -+ return -EFAULT; -+ -+ /* Only set to 0 or 1 respectively; zero->0; non-zero->1 */ -+ value = simple_strtol(str, NULL, 0); -+ if (value <= 0) -+ return -ERANGE; -+ -+ ext3_mb_max_to_scan = value; -+ -+ return count; -+} -+ -+static int ext3_mb_min_to_scan_read(char *page, char **start, off_t off, -+ int count, int *eof, void *data) -+{ -+ int len; -+ -+ *eof = 1; -+ if (off != 0) -+ return 0; -+ -+ len = sprintf(page, "%ld\n", ext3_mb_min_to_scan); -+ *start = page; -+ return len; -+} -+ -+static int ext3_mb_min_to_scan_write(struct file *file, const char *buffer, -+ unsigned long count, void *data) -+{ -+ char str[32]; -+ long value; -+ -+ if (count >= sizeof(str)) { -+ printk(KERN_ERR "EXT3-fs: %s string too long, max %u bytes\n", -+ EXT3_MB_MIN_TO_SCAN_NAME, (int)sizeof(str)); -+ return -EOVERFLOW; -+ } -+ -+ if (copy_from_user(str, buffer, count)) -+ return -EFAULT; -+ -+ /* Only set to 0 or 1 respectively; zero->0; non-zero->1 */ -+ value = simple_strtol(str, NULL, 0); -+ if (value <= 0) -+ return -ERANGE; -+ -+ ext3_mb_min_to_scan = value; -+ -+ return count; -+} -+ -+static int ext3_mb_order2_req_read(char *page, char **start, off_t off, -+ int count, int *eof, void *data) -+{ -+ int len; -+ -+ *eof = 1; -+ if (off != 0) -+ return 0; -+ -+ len = sprintf(page, "%ld\n", ext3_mb_order2_reqs); -+ *start = page; -+ return len; -+} -+ -+static int ext3_mb_order2_req_write(struct file *file, const char *buffer, -+ unsigned long count, void *data) -+{ -+ char str[32]; -+ long value; -+ -+ if (count >= sizeof(str)) { -+ printk(KERN_ERR "EXT3-fs: %s string too long, max %u bytes\n", -+ EXT3_MB_MIN_TO_SCAN_NAME, (int)sizeof(str)); -+ return -EOVERFLOW; -+ } -+ -+ if (copy_from_user(str, buffer, count)) -+ return -EFAULT; -+ -+ /* Only set to 0 or 1 respectively; zero->0; non-zero->1 */ -+ value = simple_strtol(str, NULL, 0); -+ if (value <= 0) -+ return -ERANGE; -+ -+ ext3_mb_order2_reqs = value; -+ -+ return count; -+} -+ -+int __init init_ext3_proc(void) -+{ -+ struct proc_dir_entry *proc_ext3_mb_stats; -+ struct proc_dir_entry *proc_ext3_mb_max_to_scan; -+ struct proc_dir_entry *proc_ext3_mb_min_to_scan; -+ struct proc_dir_entry *proc_ext3_mb_order2_req; -+ -+ proc_root_ext3 = proc_mkdir(EXT3_ROOT, proc_root_fs); -+ if (proc_root_ext3 == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", EXT3_ROOT); -+ return -EIO; -+ } -+ -+ /* Initialize EXT3_MB_STATS_NAME */ -+ proc_ext3_mb_stats = create_proc_entry(EXT3_MB_STATS_NAME, -+ S_IFREG | S_IRUGO | S_IWUSR, proc_root_ext3); -+ if (proc_ext3_mb_stats == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", -+ EXT3_MB_STATS_NAME); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+ return -EIO; -+ } -+ -+ proc_ext3_mb_stats->data = NULL; -+ proc_ext3_mb_stats->read_proc = ext3_mb_stats_read; -+ proc_ext3_mb_stats->write_proc = ext3_mb_stats_write; -+ -+ /* Initialize EXT3_MAX_TO_SCAN_NAME */ -+ proc_ext3_mb_max_to_scan = create_proc_entry( -+ EXT3_MB_MAX_TO_SCAN_NAME, -+ S_IFREG | S_IRUGO | S_IWUSR, proc_root_ext3); -+ if (proc_ext3_mb_max_to_scan == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", -+ EXT3_MB_MAX_TO_SCAN_NAME); -+ remove_proc_entry(EXT3_MB_STATS_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+ return -EIO; -+ } -+ -+ proc_ext3_mb_max_to_scan->data = NULL; -+ proc_ext3_mb_max_to_scan->read_proc = ext3_mb_max_to_scan_read; -+ proc_ext3_mb_max_to_scan->write_proc = ext3_mb_max_to_scan_write; -+ -+ /* Initialize EXT3_MIN_TO_SCAN_NAME */ -+ proc_ext3_mb_min_to_scan = create_proc_entry( -+ EXT3_MB_MIN_TO_SCAN_NAME, -+ S_IFREG | S_IRUGO | S_IWUSR, proc_root_ext3); -+ if (proc_ext3_mb_min_to_scan == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", -+ EXT3_MB_MIN_TO_SCAN_NAME); -+ remove_proc_entry(EXT3_MB_MAX_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_STATS_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+ return -EIO; -+ } -+ -+ proc_ext3_mb_min_to_scan->data = NULL; -+ proc_ext3_mb_min_to_scan->read_proc = ext3_mb_min_to_scan_read; -+ proc_ext3_mb_min_to_scan->write_proc = ext3_mb_min_to_scan_write; -+ -+ /* Initialize EXT3_ORDER2_REQ */ -+ proc_ext3_mb_order2_req = create_proc_entry( -+ EXT3_MB_ORDER2_REQ, -+ S_IFREG | S_IRUGO | S_IWUSR, proc_root_ext3); -+ if (proc_ext3_mb_order2_req == NULL) { -+ printk(KERN_ERR "EXT3-fs: Unable to create %s\n", -+ EXT3_MB_ORDER2_REQ); -+ remove_proc_entry(EXT3_MB_MIN_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_MAX_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_STATS_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+ return -EIO; -+ } -+ -+ proc_ext3_mb_order2_req->data = NULL; -+ proc_ext3_mb_order2_req->read_proc = ext3_mb_order2_req_read; -+ proc_ext3_mb_order2_req->write_proc = ext3_mb_order2_req_write; -+ -+ return 0; -+} -+ -+void exit_ext3_proc(void) -+{ -+ remove_proc_entry(EXT3_MB_STATS_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_MAX_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_MIN_TO_SCAN_NAME, proc_root_ext3); -+ remove_proc_entry(EXT3_MB_ORDER2_REQ, proc_root_ext3); -+ remove_proc_entry(EXT3_ROOT, proc_root_fs); -+} -Index: linux-stage/fs/ext3/Makefile -=================================================================== ---- linux-stage.orig/fs/ext3/Makefile 2006-07-16 02:29:43.000000000 +0800 -+++ linux-stage/fs/ext3/Makefile 2006-07-16 02:29:49.000000000 +0800 -@@ -6,7 +6,7 @@ obj-$(CONFIG_EXT3_FS) += ext3.o - - ext3-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o iopen.o \ - ioctl.o namei.o super.o symlink.o hash.o resize.o \ -- extents.o -+ extents.o mballoc.o - - ext3-$(CONFIG_EXT3_FS_XATTR) += xattr.o xattr_user.o xattr_trusted.o - ext3-$(CONFIG_EXT3_FS_POSIX_ACL) += acl.o diff --git a/ldiskfs/kernel_patches/patches/ext3-sector_t-overflow-2.6.12.patch b/ldiskfs/kernel_patches/patches/ext3-sector_t-overflow-2.6.12.patch deleted file mode 100644 index ef0f4a4..0000000 --- a/ldiskfs/kernel_patches/patches/ext3-sector_t-overflow-2.6.12.patch +++ /dev/null @@ -1,64 +0,0 @@ -Subject: Avoid disk sector_t overflow for >2TB ext3 filesystem -From: Mingming Cao - - -If ext3 filesystem is larger than 2TB, and sector_t is a u32 (i.e. -CONFIG_LBD not defined in the kernel), the calculation of the disk sector -will overflow. Add check at ext3_fill_super() and ext3_group_extend() to -prevent mount/remount/resize >2TB ext3 filesystem if sector_t size is 4 -bytes. - -Verified this patch on a 32 bit platform without CONFIG_LBD defined -(sector_t is 32 bits long), mount refuse to mount a 10TB ext3. - -Signed-off-by: Mingming Cao -Acked-by: Andreas Dilger -Signed-off-by: Andrew Morton ---- - - fs/ext3/resize.c | 10 ++++++++++ - fs/ext3/super.c | 10 ++++++++++ - 2 files changed, 20 insertions(+) - -diff -puN fs/ext3/resize.c~avoid-disk-sector_t-overflow-for-2tb-ext3-filesystem fs/ext3/resize.c ---- devel/fs/ext3/resize.c~avoid-disk-sector_t-overflow-for-2tb-ext3-filesystem 2006-05-22 14:09:53.000000000 -0700 -+++ devel-akpm/fs/ext3/resize.c 2006-05-22 14:10:56.000000000 -0700 -@@ -926,6 +926,16 @@ int ext3_group_extend(struct super_block - if (n_blocks_count == 0 || n_blocks_count == o_blocks_count) - return 0; - -+ if (n_blocks_count > (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) { -+ printk(KERN_ERR "EXT3-fs: filesystem on %s: " -+ "too large to resize to %lu blocks safely\n", -+ sb->s_id, n_blocks_count); -+ if (sizeof(sector_t) < 8) -+ ext3_warning(sb, __FUNCTION__, -+ "CONFIG_LBD not enabled\n"); -+ return -EINVAL; -+ } -+ - if (n_blocks_count < o_blocks_count) { - ext3_warning(sb, __FUNCTION__, - "can't shrink FS - resize aborted"); -diff -puN fs/ext3/super.c~avoid-disk-sector_t-overflow-for-2tb-ext3-filesystem fs/ext3/super.c ---- devel/fs/ext3/super.c~avoid-disk-sector_t-overflow-for-2tb-ext3-filesystem 2006-05-22 14:09:53.000000000 -0700 -+++ devel-akpm/fs/ext3/super.c 2006-05-22 14:11:10.000000000 -0700 -@@ -1565,6 +1565,17 @@ static int ext3_fill_super (struct super - goto failed_mount; - } - -+ if (le32_to_cpu(es->s_blocks_count) > -+ (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) { -+ printk(KERN_ERR "EXT3-fs: filesystem on %s: " -+ "too large to mount safely - %u blocks\n", sb->s_id, -+ le32_to_cpu(es->s_blocks_count)); -+ if (sizeof(sector_t) < 8) -+ printk(KERN_WARNING -+ "EXT3-fs: CONFIG_LBD not enabled\n"); -+ goto failed_mount; -+ } -+ - if (EXT3_BLOCKS_PER_GROUP(sb) == 0) - goto cantfind_ext3; - sbi->s_groups_count = (le32_to_cpu(es->s_blocks_count) - -_ diff --git a/ldiskfs/kernel_patches/patches/iopen-2.6.12.patch b/ldiskfs/kernel_patches/patches/iopen-2.6.12.patch deleted file mode 100644 index 8d456ac..0000000 --- a/ldiskfs/kernel_patches/patches/iopen-2.6.12.patch +++ /dev/null @@ -1,471 +0,0 @@ -Index: linux-2.6.12-rc6/fs/ext3/Makefile -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/Makefile 2005-06-14 16:00:45.206720992 +0200 -+++ linux-2.6.12-rc6/fs/ext3/Makefile 2005-06-14 16:14:33.595382720 +0200 -@@ -4,7 +4,7 @@ - - obj-$(CONFIG_EXT3_FS) += ext3.o - --ext3-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ -+ext3-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o iopen.o \ - ioctl.o namei.o super.o symlink.o hash.o resize.o - - ext3-$(CONFIG_EXT3_FS_XATTR) += xattr.o xattr_user.o xattr_trusted.o -Index: linux-2.6.12-rc6/fs/ext3/inode.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/inode.c 2005-06-14 16:01:16.272150299 +0200 -+++ linux-2.6.12-rc6/fs/ext3/inode.c 2005-06-14 16:24:55.686195412 +0200 -@@ -37,6 +37,7 @@ - #include - #include - #include "xattr.h" -+#include "iopen.h" - #include "acl.h" - - static int ext3_writepage_trans_blocks(struct inode *inode); -@@ -2437,6 +2438,8 @@ - ei->i_default_acl = EXT3_ACL_NOT_CACHED; - #endif - ei->i_block_alloc_info = NULL; -+ if (ext3_iopen_get_inode(inode)) -+ return; - - if (__ext3_get_inode_loc(inode, &iloc, 0)) - goto bad_inode; -Index: linux-2.6.12-rc6/fs/ext3/iopen.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/iopen.c 2005-06-14 16:14:33.530929595 +0200 -+++ linux-2.6.12-rc6/fs/ext3/iopen.c 2005-06-14 16:14:33.626632719 +0200 -@@ -0,0 +1,278 @@ -+/* -+ * linux/fs/ext3/iopen.c -+ * -+ * Special support for open by inode number -+ * -+ * Copyright (C) 2001 by Theodore Ts'o (tytso@alum.mit.edu). -+ * -+ * This file may be redistributed under the terms of the GNU General -+ * Public License. -+ * -+ * -+ * Invariants: -+ * - there is only ever a single DCACHE_NFSD_DISCONNECTED dentry alias -+ * for an inode at one time. -+ * - there are never both connected and DCACHE_NFSD_DISCONNECTED dentry -+ * aliases on an inode at the same time. -+ * -+ * If we have any connected dentry aliases for an inode, use one of those -+ * in iopen_lookup(). Otherwise, we instantiate a single NFSD_DISCONNECTED -+ * dentry for this inode, which thereafter will be found by the dcache -+ * when looking up this inode number in __iopen__, so we don't return here -+ * until it is gone. -+ * -+ * If we get an inode via a regular name lookup, then we "rename" the -+ * NFSD_DISCONNECTED dentry to the proper name and parent. This ensures -+ * existing users of the disconnected dentry will continue to use the same -+ * dentry as the connected users, and there will never be both kinds of -+ * dentry aliases at one time. -+ */ -+ -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include -+#include "iopen.h" -+ -+#ifndef assert -+#define assert(test) J_ASSERT(test) -+#endif -+ -+#define IOPEN_NAME_LEN 32 -+ -+/* -+ * This implements looking up an inode by number. -+ */ -+static struct dentry *iopen_lookup(struct inode * dir, struct dentry *dentry, -+ struct nameidata *nd) -+{ -+ struct inode *inode; -+ unsigned long ino; -+ struct list_head *lp; -+ struct dentry *alternate; -+ char buf[IOPEN_NAME_LEN]; -+ -+ if (dentry->d_name.len >= IOPEN_NAME_LEN) -+ return ERR_PTR(-ENAMETOOLONG); -+ -+ memcpy(buf, dentry->d_name.name, dentry->d_name.len); -+ buf[dentry->d_name.len] = 0; -+ -+ if (strcmp(buf, ".") == 0) -+ ino = dir->i_ino; -+ else if (strcmp(buf, "..") == 0) -+ ino = EXT3_ROOT_INO; -+ else -+ ino = simple_strtoul(buf, 0, 0); -+ -+ if ((ino != EXT3_ROOT_INO && -+ //ino != EXT3_ACL_IDX_INO && -+ //ino != EXT3_ACL_DATA_INO && -+ ino < EXT3_FIRST_INO(dir->i_sb)) || -+ ino > le32_to_cpu(EXT3_SB(dir->i_sb)->s_es->s_inodes_count)) -+ return ERR_PTR(-ENOENT); -+ -+ inode = iget(dir->i_sb, ino); -+ if (!inode) -+ return ERR_PTR(-EACCES); -+ if (is_bad_inode(inode)) { -+ iput(inode); -+ return ERR_PTR(-ENOENT); -+ } -+ -+ assert(list_empty(&dentry->d_alias)); /* d_instantiate */ -+ assert(d_unhashed(dentry)); /* d_rehash */ -+ -+ /* preferrably return a connected dentry */ -+ spin_lock(&dcache_lock); -+ list_for_each(lp, &inode->i_dentry) { -+ alternate = list_entry(lp, struct dentry, d_alias); -+ assert(!(alternate->d_flags & DCACHE_DISCONNECTED)); -+ } -+ -+ if (!list_empty(&inode->i_dentry)) { -+ alternate = list_entry(inode->i_dentry.next, -+ struct dentry, d_alias); -+ dget_locked(alternate); -+ spin_lock(&alternate->d_lock); -+ alternate->d_flags |= DCACHE_REFERENCED; -+ spin_unlock(&alternate->d_lock); -+ iput(inode); -+ spin_unlock(&dcache_lock); -+ return alternate; -+ } -+ dentry->d_flags |= DCACHE_DISCONNECTED; -+ -+ /* d_add(), but don't drop dcache_lock before adding dentry to inode */ -+ list_add(&dentry->d_alias, &inode->i_dentry); /* d_instantiate */ -+ dentry->d_inode = inode; -+ -+ d_rehash_cond(dentry, 0); /* d_rehash */ -+ spin_unlock(&dcache_lock); -+ -+ return NULL; -+} -+ -+#define do_switch(x,y) do { \ -+ __typeof__ (x) __tmp = x; \ -+ x = y; y = __tmp; } while (0) -+ -+static inline void switch_names(struct dentry *dentry, struct dentry *target) -+{ -+ const unsigned char *old_name, *new_name; -+ -+ memcpy(dentry->d_iname, target->d_iname, DNAME_INLINE_LEN_MIN); -+ old_name = target->d_name.name; -+ new_name = dentry->d_name.name; -+ if (old_name == target->d_iname) -+ old_name = dentry->d_iname; -+ if (new_name == dentry->d_iname) -+ new_name = target->d_iname; -+ target->d_name.name = new_name; -+ dentry->d_name.name = old_name; -+} -+ -+/* This function is spliced into ext3_lookup and does the move of a -+ * disconnected dentry (if it exists) to a connected dentry. -+ */ -+struct dentry *iopen_connect_dentry(struct dentry *dentry, struct inode *inode, -+ int rehash) -+{ -+ struct dentry *tmp, *goal = NULL; -+ struct list_head *lp; -+ -+ /* verify this dentry is really new */ -+ assert(dentry->d_inode == NULL); -+ assert(list_empty(&dentry->d_alias)); /* d_instantiate */ -+ if (rehash) -+ assert(d_unhashed(dentry)); /* d_rehash */ -+ assert(list_empty(&dentry->d_subdirs)); -+ -+ spin_lock(&dcache_lock); -+ if (!inode) -+ goto do_rehash; -+ -+ if (!test_opt(inode->i_sb, IOPEN)) -+ goto do_instantiate; -+ -+ /* preferrably return a connected dentry */ -+ list_for_each(lp, &inode->i_dentry) { -+ tmp = list_entry(lp, struct dentry, d_alias); -+ if (tmp->d_flags & DCACHE_DISCONNECTED) { -+ assert(tmp->d_alias.next == &inode->i_dentry); -+ assert(tmp->d_alias.prev == &inode->i_dentry); -+ goal = tmp; -+ dget_locked(goal); -+ break; -+ } -+ } -+ -+ if (!goal) -+ goto do_instantiate; -+ -+ /* Move the goal to the de hash queue */ -+ goal->d_flags &= ~DCACHE_DISCONNECTED; -+ security_d_instantiate(goal, inode); -+ __d_drop(dentry); -+ d_rehash_cond(dentry, 0); -+ __d_move(goal, dentry); -+ spin_unlock(&dcache_lock); -+ iput(inode); -+ -+ return goal; -+ -+ /* d_add(), but don't drop dcache_lock before adding dentry to inode */ -+do_instantiate: -+ list_add(&dentry->d_alias, &inode->i_dentry); /* d_instantiate */ -+ dentry->d_inode = inode; -+do_rehash: -+ if (rehash) -+ d_rehash_cond(dentry, 0); /* d_rehash */ -+ spin_unlock(&dcache_lock); -+ -+ return NULL; -+} -+ -+/* -+ * These are the special structures for the iopen pseudo directory. -+ */ -+ -+static struct inode_operations iopen_inode_operations = { -+ lookup: iopen_lookup, /* BKL held */ -+}; -+ -+static struct file_operations iopen_file_operations = { -+ read: generic_read_dir, -+}; -+ -+static int match_dentry(struct dentry *dentry, const char *name) -+{ -+ int len; -+ -+ len = strlen(name); -+ if (dentry->d_name.len != len) -+ return 0; -+ if (strncmp(dentry->d_name.name, name, len)) -+ return 0; -+ return 1; -+} -+ -+/* -+ * This function is spliced into ext3_lookup and returns 1 the file -+ * name is __iopen__ and dentry has been filled in appropriately. -+ */ -+int ext3_check_for_iopen(struct inode *dir, struct dentry *dentry) -+{ -+ struct inode *inode; -+ -+ if (dir->i_ino != EXT3_ROOT_INO || -+ !test_opt(dir->i_sb, IOPEN) || -+ !match_dentry(dentry, "__iopen__")) -+ return 0; -+ -+ inode = iget(dir->i_sb, EXT3_BAD_INO); -+ -+ if (!inode) -+ return 0; -+ d_add(dentry, inode); -+ return 1; -+} -+ -+/* -+ * This function is spliced into read_inode; it returns 1 if inode -+ * number is the one for /__iopen__, in which case the inode is filled -+ * in appropriately. Otherwise, this fuction returns 0. -+ */ -+int ext3_iopen_get_inode(struct inode *inode) -+{ -+ if (inode->i_ino != EXT3_BAD_INO) -+ return 0; -+ -+ inode->i_mode = S_IFDIR | S_IRUSR | S_IXUSR; -+ if (test_opt(inode->i_sb, IOPEN_NOPRIV)) -+ inode->i_mode |= 0777; -+ inode->i_uid = 0; -+ inode->i_gid = 0; -+ inode->i_nlink = 1; -+ inode->i_size = 4096; -+ inode->i_atime = CURRENT_TIME; -+ inode->i_ctime = CURRENT_TIME; -+ inode->i_mtime = CURRENT_TIME; -+ EXT3_I(inode)->i_dtime = 0; -+ inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size -+ * (for stat), not the fs block -+ * size */ -+ inode->i_blocks = 0; -+ inode->i_version = 1; -+ inode->i_generation = 0; -+ -+ inode->i_op = &iopen_inode_operations; -+ inode->i_fop = &iopen_file_operations; -+ inode->i_mapping->a_ops = 0; -+ -+ return 1; -+} -Index: linux-2.6.12-rc6/fs/ext3/iopen.h -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/iopen.h 2005-06-14 16:14:33.534835845 +0200 -+++ linux-2.6.12-rc6/fs/ext3/iopen.h 2005-06-14 16:14:33.633468657 +0200 -@@ -0,0 +1,15 @@ -+/* -+ * iopen.h -+ * -+ * Special support for opening files by inode number. -+ * -+ * Copyright (C) 2001 by Theodore Ts'o (tytso@alum.mit.edu). -+ * -+ * This file may be redistributed under the terms of the GNU General -+ * Public License. -+ */ -+ -+extern int ext3_check_for_iopen(struct inode *dir, struct dentry *dentry); -+extern int ext3_iopen_get_inode(struct inode *inode); -+extern struct dentry *iopen_connect_dentry(struct dentry *dentry, -+ struct inode *inode, int rehash); -Index: linux-2.6.12-rc6/fs/ext3/namei.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/namei.c 2005-06-14 16:01:14.701837819 +0200 -+++ linux-2.6.12-rc6/fs/ext3/namei.c 2005-06-14 16:14:33.644210844 +0200 -@@ -37,6 +37,7 @@ - #include - #include - #include "xattr.h" -+#include "iopen.h" - #include "acl.h" - - /* -@@ -985,6 +986,9 @@ - if (dentry->d_name.len > EXT3_NAME_LEN) - return ERR_PTR(-ENAMETOOLONG); - -+ if (ext3_check_for_iopen(dir, dentry)) -+ return NULL; -+ - bh = ext3_find_entry(dentry, &de); - inode = NULL; - if (bh) { -@@ -995,10 +999,8 @@ - if (!inode) - return ERR_PTR(-EACCES); - } -- if (inode) -- return d_splice_alias(inode, dentry); -- d_add(dentry, inode); -- return NULL; -+ -+ return iopen_connect_dentry(dentry, inode, 1); - } - - -@@ -2042,10 +2044,6 @@ - inode->i_nlink); - inode->i_version++; - inode->i_nlink = 0; -- /* There's no need to set i_disksize: the fact that i_nlink is -- * zero will ensure that the right thing happens during any -- * recovery. */ -- inode->i_size = 0; - ext3_orphan_add(handle, inode); - inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME_SEC; - ext3_mark_inode_dirty(handle, inode); -@@ -2168,6 +2166,23 @@ - return err; - } - -+/* Like ext3_add_nondir() except for call to iopen_connect_dentry */ -+static int ext3_add_link(handle_t *handle, struct dentry *dentry, -+ struct inode *inode) -+{ -+ int err = ext3_add_entry(handle, dentry, inode); -+ if (!err) { -+ err = ext3_mark_inode_dirty(handle, inode); -+ if (err == 0) { -+ dput(iopen_connect_dentry(dentry, inode, 0)); -+ return 0; -+ } -+ } -+ ext3_dec_count(handle, inode); -+ iput(inode); -+ return err; -+} -+ - static int ext3_link (struct dentry * old_dentry, - struct inode * dir, struct dentry *dentry) - { -@@ -2191,7 +2206,8 @@ - ext3_inc_count(handle, inode); - atomic_inc(&inode->i_count); - -- err = ext3_add_nondir(handle, dentry, inode); -+ err = ext3_add_link(handle, dentry, inode); -+ ext3_orphan_del(handle, inode); - ext3_journal_stop(handle); - if (err == -ENOSPC && ext3_should_retry_alloc(dir->i_sb, &retries)) - goto retry; -Index: linux-2.6.12-rc6/fs/ext3/super.c -=================================================================== ---- linux-2.6.12-rc6.orig/fs/ext3/super.c 2005-06-14 16:01:16.287775299 +0200 -+++ linux-2.6.12-rc6/fs/ext3/super.c 2005-06-14 16:14:33.656906156 +0200 -@@ -590,6 +590,7 @@ - Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, - Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, - Opt_ignore, Opt_barrier, Opt_err, Opt_resize, -+ Opt_iopen, Opt_noiopen, Opt_iopen_nopriv, - }; - - static match_table_t tokens = { -@@ -638,6 +639,9 @@ - {Opt_ignore, "noquota"}, - {Opt_ignore, "quota"}, - {Opt_ignore, "usrquota"}, -+ {Opt_iopen, "iopen"}, -+ {Opt_noiopen, "noiopen"}, -+ {Opt_iopen_nopriv, "iopen_nopriv"}, - {Opt_barrier, "barrier=%u"}, - {Opt_err, NULL}, - {Opt_resize, "resize"}, -@@ -921,6 +925,18 @@ - else - clear_opt(sbi->s_mount_opt, BARRIER); - break; -+ case Opt_iopen: -+ set_opt (sbi->s_mount_opt, IOPEN); -+ clear_opt (sbi->s_mount_opt, IOPEN_NOPRIV); -+ break; -+ case Opt_noiopen: -+ clear_opt (sbi->s_mount_opt, IOPEN); -+ clear_opt (sbi->s_mount_opt, IOPEN_NOPRIV); -+ break; -+ case Opt_iopen_nopriv: -+ set_opt (sbi->s_mount_opt, IOPEN); -+ set_opt (sbi->s_mount_opt, IOPEN_NOPRIV); -+ break; - case Opt_ignore: - break; - case Opt_resize: -Index: linux-2.6.12-rc6/include/linux/ext3_fs.h -=================================================================== ---- linux-2.6.12-rc6.orig/include/linux/ext3_fs.h 2005-06-14 16:01:14.709650318 +0200 -+++ linux-2.6.12-rc6/include/linux/ext3_fs.h 2005-06-14 16:28:38.452794245 +0200 -@@ -358,6 +358,8 @@ - #define EXT3_MOUNT_RESERVATION 0x10000 /* Preallocation */ - #define EXT3_MOUNT_BARRIER 0x20000 /* Use block barriers */ - #define EXT3_MOUNT_NOBH 0x40000 /* No bufferheads */ -+#define EXT3_MOUNT_IOPEN 0x80000 /* Allow access via iopen */ -+#define EXT3_MOUNT_IOPEN_NOPRIV 0x100000/* Make iopen world-readable */ - - /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ - #ifndef _LINUX_EXT2_FS_H diff --git a/ldiskfs/kernel_patches/series/ldiskfs-2.6.12-vanilla.series b/ldiskfs/kernel_patches/series/ldiskfs-2.6.12-vanilla.series deleted file mode 100644 index 286a3a2..0000000 --- a/ldiskfs/kernel_patches/series/ldiskfs-2.6.12-vanilla.series +++ /dev/null @@ -1,18 +0,0 @@ -ext3-wantedi-2.6-rhel4.patch -ext3-san-jdike-2.6-suse.patch -iopen-2.6.12.patch -ext3-map_inode_page-2.6-suse.patch -export-ext3-2.6-rhel4.patch -ext3-include-fixes-2.6-rhel4.patch -ext3-extents-2.6.12.patch -ext3-mballoc2-2.6.12.patch -ext3-nlinks-2.6.9.patch -ext3-ialloc-2.6.patch -ext3-remove-cond_resched-calls-2.6.12.patch -ext3-htree-dot-2.6.patch -ext3-external-journal-2.6.12.patch -ext3-check-jbd-errors-2.6.12.patch -ext3-check-jbd-errors-2.6.9.patch -ext3-uninit-2.6.9.patch -ext3-nanosecond-2.6-rhel4.patch -