X-Git-Url: http://pilppa.org/gitweb/?a=blobdiff_plain;f=fs%2Focfs2%2Falloc.c;h=a164e09491f89984534e1c38fa5873808042efad;hb=31d33073ca38603dea705dae45e094a64ca062d6;hp=377acb24f67b68422862f201b241776da6d2653b;hpb=231b87d10920e024efaf0f9e86e1bab7bced1620;p=linux-2.6-omap-h63xx.git diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c index 377acb24f67..a164e09491f 100644 --- a/fs/ocfs2/alloc.c +++ b/fs/ocfs2/alloc.c @@ -49,6 +49,340 @@ #include "buffer_head_io.h" + +/* + * Operations for a specific extent tree type. + * + * To implement an on-disk btree (extent tree) type in ocfs2, add + * an ocfs2_extent_tree_operations structure and the matching + * ocfs2_init__extent_tree() function. That's pretty much it + * for the allocation portion of the extent tree. + */ +struct ocfs2_extent_tree_operations { + /* + * last_eb_blk is the block number of the right most leaf extent + * block. Most on-disk structures containing an extent tree store + * this value for fast access. The ->eo_set_last_eb_blk() and + * ->eo_get_last_eb_blk() operations access this value. They are + * both required. + */ + void (*eo_set_last_eb_blk)(struct ocfs2_extent_tree *et, + u64 blkno); + u64 (*eo_get_last_eb_blk)(struct ocfs2_extent_tree *et); + + /* + * The on-disk structure usually keeps track of how many total + * clusters are stored in this extent tree. This function updates + * that value. new_clusters is the delta, and must be + * added to the total. Required. + */ + void (*eo_update_clusters)(struct inode *inode, + struct ocfs2_extent_tree *et, + u32 new_clusters); + + /* + * If ->eo_insert_check() exists, it is called before rec is + * inserted into the extent tree. It is optional. + */ + int (*eo_insert_check)(struct inode *inode, + struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec); + int (*eo_sanity_check)(struct inode *inode, struct ocfs2_extent_tree *et); + + /* + * -------------------------------------------------------------- + * The remaining are internal to ocfs2_extent_tree and don't have + * accessor functions + */ + + /* + * ->eo_fill_root_el() takes et->et_object and sets et->et_root_el. + * It is required. + */ + void (*eo_fill_root_el)(struct ocfs2_extent_tree *et); + + /* + * ->eo_fill_max_leaf_clusters sets et->et_max_leaf_clusters if + * it exists. If it does not, et->et_max_leaf_clusters is set + * to 0 (unlimited). Optional. + */ + void (*eo_fill_max_leaf_clusters)(struct inode *inode, + struct ocfs2_extent_tree *et); +}; + + +/* + * Pre-declare ocfs2_dinode_et_ops so we can use it as a sanity check + * in the methods. + */ +static u64 ocfs2_dinode_get_last_eb_blk(struct ocfs2_extent_tree *et); +static void ocfs2_dinode_set_last_eb_blk(struct ocfs2_extent_tree *et, + u64 blkno); +static void ocfs2_dinode_update_clusters(struct inode *inode, + struct ocfs2_extent_tree *et, + u32 clusters); +static int ocfs2_dinode_insert_check(struct inode *inode, + struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec); +static int ocfs2_dinode_sanity_check(struct inode *inode, + struct ocfs2_extent_tree *et); +static void ocfs2_dinode_fill_root_el(struct ocfs2_extent_tree *et); +static struct ocfs2_extent_tree_operations ocfs2_dinode_et_ops = { + .eo_set_last_eb_blk = ocfs2_dinode_set_last_eb_blk, + .eo_get_last_eb_blk = ocfs2_dinode_get_last_eb_blk, + .eo_update_clusters = ocfs2_dinode_update_clusters, + .eo_insert_check = ocfs2_dinode_insert_check, + .eo_sanity_check = ocfs2_dinode_sanity_check, + .eo_fill_root_el = ocfs2_dinode_fill_root_el, +}; + +static void ocfs2_dinode_set_last_eb_blk(struct ocfs2_extent_tree *et, + u64 blkno) +{ + struct ocfs2_dinode *di = et->et_object; + + BUG_ON(et->et_ops != &ocfs2_dinode_et_ops); + di->i_last_eb_blk = cpu_to_le64(blkno); +} + +static u64 ocfs2_dinode_get_last_eb_blk(struct ocfs2_extent_tree *et) +{ + struct ocfs2_dinode *di = et->et_object; + + BUG_ON(et->et_ops != &ocfs2_dinode_et_ops); + return le64_to_cpu(di->i_last_eb_blk); +} + +static void ocfs2_dinode_update_clusters(struct inode *inode, + struct ocfs2_extent_tree *et, + u32 clusters) +{ + struct ocfs2_dinode *di = et->et_object; + + le32_add_cpu(&di->i_clusters, clusters); + spin_lock(&OCFS2_I(inode)->ip_lock); + OCFS2_I(inode)->ip_clusters = le32_to_cpu(di->i_clusters); + spin_unlock(&OCFS2_I(inode)->ip_lock); +} + +static int ocfs2_dinode_insert_check(struct inode *inode, + struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec) +{ + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + BUG_ON(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL); + mlog_bug_on_msg(!ocfs2_sparse_alloc(osb) && + (OCFS2_I(inode)->ip_clusters != rec->e_cpos), + "Device %s, asking for sparse allocation: inode %llu, " + "cpos %u, clusters %u\n", + osb->dev_str, + (unsigned long long)OCFS2_I(inode)->ip_blkno, + rec->e_cpos, + OCFS2_I(inode)->ip_clusters); + + return 0; +} + +static int ocfs2_dinode_sanity_check(struct inode *inode, + struct ocfs2_extent_tree *et) +{ + int ret = 0; + struct ocfs2_dinode *di; + + BUG_ON(et->et_ops != &ocfs2_dinode_et_ops); + + di = et->et_object; + if (!OCFS2_IS_VALID_DINODE(di)) { + ret = -EIO; + ocfs2_error(inode->i_sb, + "Inode %llu has invalid path root", + (unsigned long long)OCFS2_I(inode)->ip_blkno); + } + + return ret; +} + +static void ocfs2_dinode_fill_root_el(struct ocfs2_extent_tree *et) +{ + struct ocfs2_dinode *di = et->et_object; + + et->et_root_el = &di->id2.i_list; +} + + +static void ocfs2_xattr_value_fill_root_el(struct ocfs2_extent_tree *et) +{ + struct ocfs2_xattr_value_root *xv = et->et_object; + + et->et_root_el = &xv->xr_list; +} + +static void ocfs2_xattr_value_set_last_eb_blk(struct ocfs2_extent_tree *et, + u64 blkno) +{ + struct ocfs2_xattr_value_root *xv = + (struct ocfs2_xattr_value_root *)et->et_object; + + xv->xr_last_eb_blk = cpu_to_le64(blkno); +} + +static u64 ocfs2_xattr_value_get_last_eb_blk(struct ocfs2_extent_tree *et) +{ + struct ocfs2_xattr_value_root *xv = + (struct ocfs2_xattr_value_root *) et->et_object; + + return le64_to_cpu(xv->xr_last_eb_blk); +} + +static void ocfs2_xattr_value_update_clusters(struct inode *inode, + struct ocfs2_extent_tree *et, + u32 clusters) +{ + struct ocfs2_xattr_value_root *xv = + (struct ocfs2_xattr_value_root *)et->et_object; + + le32_add_cpu(&xv->xr_clusters, clusters); +} + +static struct ocfs2_extent_tree_operations ocfs2_xattr_value_et_ops = { + .eo_set_last_eb_blk = ocfs2_xattr_value_set_last_eb_blk, + .eo_get_last_eb_blk = ocfs2_xattr_value_get_last_eb_blk, + .eo_update_clusters = ocfs2_xattr_value_update_clusters, + .eo_fill_root_el = ocfs2_xattr_value_fill_root_el, +}; + +static void ocfs2_xattr_tree_fill_root_el(struct ocfs2_extent_tree *et) +{ + struct ocfs2_xattr_block *xb = et->et_object; + + et->et_root_el = &xb->xb_attrs.xb_root.xt_list; +} + +static void ocfs2_xattr_tree_fill_max_leaf_clusters(struct inode *inode, + struct ocfs2_extent_tree *et) +{ + et->et_max_leaf_clusters = + ocfs2_clusters_for_bytes(inode->i_sb, + OCFS2_MAX_XATTR_TREE_LEAF_SIZE); +} + +static void ocfs2_xattr_tree_set_last_eb_blk(struct ocfs2_extent_tree *et, + u64 blkno) +{ + struct ocfs2_xattr_block *xb = et->et_object; + struct ocfs2_xattr_tree_root *xt = &xb->xb_attrs.xb_root; + + xt->xt_last_eb_blk = cpu_to_le64(blkno); +} + +static u64 ocfs2_xattr_tree_get_last_eb_blk(struct ocfs2_extent_tree *et) +{ + struct ocfs2_xattr_block *xb = et->et_object; + struct ocfs2_xattr_tree_root *xt = &xb->xb_attrs.xb_root; + + return le64_to_cpu(xt->xt_last_eb_blk); +} + +static void ocfs2_xattr_tree_update_clusters(struct inode *inode, + struct ocfs2_extent_tree *et, + u32 clusters) +{ + struct ocfs2_xattr_block *xb = et->et_object; + + le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, clusters); +} + +static struct ocfs2_extent_tree_operations ocfs2_xattr_tree_et_ops = { + .eo_set_last_eb_blk = ocfs2_xattr_tree_set_last_eb_blk, + .eo_get_last_eb_blk = ocfs2_xattr_tree_get_last_eb_blk, + .eo_update_clusters = ocfs2_xattr_tree_update_clusters, + .eo_fill_root_el = ocfs2_xattr_tree_fill_root_el, + .eo_fill_max_leaf_clusters = ocfs2_xattr_tree_fill_max_leaf_clusters, +}; + +static void __ocfs2_init_extent_tree(struct ocfs2_extent_tree *et, + struct inode *inode, + struct buffer_head *bh, + void *obj, + struct ocfs2_extent_tree_operations *ops) +{ + et->et_ops = ops; + et->et_root_bh = bh; + if (!obj) + obj = (void *)bh->b_data; + et->et_object = obj; + + et->et_ops->eo_fill_root_el(et); + if (!et->et_ops->eo_fill_max_leaf_clusters) + et->et_max_leaf_clusters = 0; + else + et->et_ops->eo_fill_max_leaf_clusters(inode, et); +} + +void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et, + struct inode *inode, + struct buffer_head *bh) +{ + __ocfs2_init_extent_tree(et, inode, bh, NULL, &ocfs2_dinode_et_ops); +} + +void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et, + struct inode *inode, + struct buffer_head *bh) +{ + __ocfs2_init_extent_tree(et, inode, bh, NULL, + &ocfs2_xattr_tree_et_ops); +} + +void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et, + struct inode *inode, + struct buffer_head *bh, + struct ocfs2_xattr_value_root *xv) +{ + __ocfs2_init_extent_tree(et, inode, bh, xv, + &ocfs2_xattr_value_et_ops); +} + +static inline void ocfs2_et_set_last_eb_blk(struct ocfs2_extent_tree *et, + u64 new_last_eb_blk) +{ + et->et_ops->eo_set_last_eb_blk(et, new_last_eb_blk); +} + +static inline u64 ocfs2_et_get_last_eb_blk(struct ocfs2_extent_tree *et) +{ + return et->et_ops->eo_get_last_eb_blk(et); +} + +static inline void ocfs2_et_update_clusters(struct inode *inode, + struct ocfs2_extent_tree *et, + u32 clusters) +{ + et->et_ops->eo_update_clusters(inode, et, clusters); +} + +static inline int ocfs2_et_insert_check(struct inode *inode, + struct ocfs2_extent_tree *et, + struct ocfs2_extent_rec *rec) +{ + int ret = 0; + + if (et->et_ops->eo_insert_check) + ret = et->et_ops->eo_insert_check(inode, et, rec); + return ret; +} + +static inline int ocfs2_et_sanity_check(struct inode *inode, + struct ocfs2_extent_tree *et) +{ + int ret = 0; + + if (et->et_ops->eo_sanity_check) + ret = et->et_ops->eo_sanity_check(inode, et); + return ret; +} + static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc); static int ocfs2_cache_extent_block_free(struct ocfs2_cached_dealloc_ctxt *ctxt, struct ocfs2_extent_block *eb); @@ -204,17 +538,6 @@ static struct ocfs2_path *ocfs2_new_path(struct buffer_head *root_bh, return path; } -/* - * Allocate and initialize a new path based on a disk inode tree. - */ -static struct ocfs2_path *ocfs2_new_inode_path(struct buffer_head *di_bh) -{ - struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; - struct ocfs2_extent_list *el = &di->id2.i_list; - - return ocfs2_new_path(di_bh, el); -} - /* * Convenience function to journal all components in a path. */ @@ -368,40 +691,35 @@ struct ocfs2_merge_ctxt { */ int ocfs2_num_free_extents(struct ocfs2_super *osb, struct inode *inode, - struct buffer_head *bh) + struct ocfs2_extent_tree *et) { int retval; - struct ocfs2_extent_list *el; + struct ocfs2_extent_list *el = NULL; struct ocfs2_extent_block *eb; struct buffer_head *eb_bh = NULL; - struct ocfs2_dinode *fe = (struct ocfs2_dinode *)bh->b_data; + u64 last_eb_blk = 0; mlog_entry_void(); - if (!OCFS2_IS_VALID_DINODE(fe)) { - OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe); - retval = -EIO; - goto bail; - } + el = et->et_root_el; + last_eb_blk = ocfs2_et_get_last_eb_blk(et); - if (fe->i_last_eb_blk) { - retval = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk), - &eb_bh, OCFS2_BH_CACHED, inode); + if (last_eb_blk) { + retval = ocfs2_read_block(inode, last_eb_blk, + &eb_bh, OCFS2_BH_CACHED); if (retval < 0) { mlog_errno(retval); goto bail; } eb = (struct ocfs2_extent_block *) eb_bh->b_data; el = &eb->h_list; - } else - el = &fe->id2.i_list; + } BUG_ON(el->l_tree_depth != 0); retval = le16_to_cpu(el->l_count) - le16_to_cpu(el->l_next_free_rec); bail: - if (eb_bh) - brelse(eb_bh); + brelse(eb_bh); mlog_exit(retval); return retval; @@ -487,8 +805,7 @@ static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb, bail: if (status < 0) { for(i = 0; i < wanted; i++) { - if (bhs[i]) - brelse(bhs[i]); + brelse(bhs[i]); bhs[i] = NULL; } } @@ -532,7 +849,7 @@ static inline u32 ocfs2_sum_rightmost_rec(struct ocfs2_extent_list *el) static int ocfs2_add_branch(struct ocfs2_super *osb, handle_t *handle, struct inode *inode, - struct buffer_head *fe_bh, + struct ocfs2_extent_tree *et, struct buffer_head *eb_bh, struct buffer_head **last_eb_bh, struct ocfs2_alloc_context *meta_ac) @@ -541,7 +858,6 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, u64 next_blkno, new_last_eb_blk; struct buffer_head *bh; struct buffer_head **new_eb_bhs = NULL; - struct ocfs2_dinode *fe; struct ocfs2_extent_block *eb; struct ocfs2_extent_list *eb_el; struct ocfs2_extent_list *el; @@ -551,13 +867,11 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, BUG_ON(!last_eb_bh || !*last_eb_bh); - fe = (struct ocfs2_dinode *) fe_bh->b_data; - if (eb_bh) { eb = (struct ocfs2_extent_block *) eb_bh->b_data; el = &eb->h_list; } else - el = &fe->id2.i_list; + el = et->et_root_el; /* we never add a branch to a leaf. */ BUG_ON(!el->l_tree_depth); @@ -647,7 +961,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, mlog_errno(status); goto bail; } - status = ocfs2_journal_access(handle, inode, fe_bh, + status = ocfs2_journal_access(handle, inode, et->et_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -663,7 +977,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, } /* Link the new branch into the rest of the tree (el will - * either be on the fe, or the extent block passed in. */ + * either be on the root_bh, or the extent block passed in. */ i = le16_to_cpu(el->l_next_free_rec); el->l_recs[i].e_blkno = cpu_to_le64(next_blkno); el->l_recs[i].e_cpos = cpu_to_le32(new_cpos); @@ -672,7 +986,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, /* fe needs a new last extent block pointer, as does the * next_leaf on the previously last-extent-block. */ - fe->i_last_eb_blk = cpu_to_le64(new_last_eb_blk); + ocfs2_et_set_last_eb_blk(et, new_last_eb_blk); eb = (struct ocfs2_extent_block *) (*last_eb_bh)->b_data; eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk); @@ -680,7 +994,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, status = ocfs2_journal_dirty(handle, *last_eb_bh); if (status < 0) mlog_errno(status); - status = ocfs2_journal_dirty(handle, fe_bh); + status = ocfs2_journal_dirty(handle, et->et_root_bh); if (status < 0) mlog_errno(status); if (eb_bh) { @@ -701,8 +1015,7 @@ static int ocfs2_add_branch(struct ocfs2_super *osb, bail: if (new_eb_bhs) { for (i = 0; i < new_blocks; i++) - if (new_eb_bhs[i]) - brelse(new_eb_bhs[i]); + brelse(new_eb_bhs[i]); kfree(new_eb_bhs); } @@ -718,16 +1031,15 @@ bail: static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, handle_t *handle, struct inode *inode, - struct buffer_head *fe_bh, + struct ocfs2_extent_tree *et, struct ocfs2_alloc_context *meta_ac, struct buffer_head **ret_new_eb_bh) { int status, i; u32 new_clusters; struct buffer_head *new_eb_bh = NULL; - struct ocfs2_dinode *fe; struct ocfs2_extent_block *eb; - struct ocfs2_extent_list *fe_el; + struct ocfs2_extent_list *root_el; struct ocfs2_extent_list *eb_el; mlog_entry_void(); @@ -747,8 +1059,7 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, } eb_el = &eb->h_list; - fe = (struct ocfs2_dinode *) fe_bh->b_data; - fe_el = &fe->id2.i_list; + root_el = et->et_root_el; status = ocfs2_journal_access(handle, inode, new_eb_bh, OCFS2_JOURNAL_ACCESS_CREATE); @@ -757,11 +1068,11 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, goto bail; } - /* copy the fe data into the new extent block */ - eb_el->l_tree_depth = fe_el->l_tree_depth; - eb_el->l_next_free_rec = fe_el->l_next_free_rec; - for(i = 0; i < le16_to_cpu(fe_el->l_next_free_rec); i++) - eb_el->l_recs[i] = fe_el->l_recs[i]; + /* copy the root extent list data into the new extent block */ + eb_el->l_tree_depth = root_el->l_tree_depth; + eb_el->l_next_free_rec = root_el->l_next_free_rec; + for (i = 0; i < le16_to_cpu(root_el->l_next_free_rec); i++) + eb_el->l_recs[i] = root_el->l_recs[i]; status = ocfs2_journal_dirty(handle, new_eb_bh); if (status < 0) { @@ -769,7 +1080,7 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, goto bail; } - status = ocfs2_journal_access(handle, inode, fe_bh, + status = ocfs2_journal_access(handle, inode, et->et_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (status < 0) { mlog_errno(status); @@ -778,21 +1089,21 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, new_clusters = ocfs2_sum_rightmost_rec(eb_el); - /* update fe now */ - le16_add_cpu(&fe_el->l_tree_depth, 1); - fe_el->l_recs[0].e_cpos = 0; - fe_el->l_recs[0].e_blkno = eb->h_blkno; - fe_el->l_recs[0].e_int_clusters = cpu_to_le32(new_clusters); - for(i = 1; i < le16_to_cpu(fe_el->l_next_free_rec); i++) - memset(&fe_el->l_recs[i], 0, sizeof(struct ocfs2_extent_rec)); - fe_el->l_next_free_rec = cpu_to_le16(1); + /* update root_bh now */ + le16_add_cpu(&root_el->l_tree_depth, 1); + root_el->l_recs[0].e_cpos = 0; + root_el->l_recs[0].e_blkno = eb->h_blkno; + root_el->l_recs[0].e_int_clusters = cpu_to_le32(new_clusters); + for (i = 1; i < le16_to_cpu(root_el->l_next_free_rec); i++) + memset(&root_el->l_recs[i], 0, sizeof(struct ocfs2_extent_rec)); + root_el->l_next_free_rec = cpu_to_le16(1); /* If this is our 1st tree depth shift, then last_eb_blk * becomes the allocated extent block */ - if (fe_el->l_tree_depth == cpu_to_le16(1)) - fe->i_last_eb_blk = eb->h_blkno; + if (root_el->l_tree_depth == cpu_to_le16(1)) + ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno)); - status = ocfs2_journal_dirty(handle, fe_bh); + status = ocfs2_journal_dirty(handle, et->et_root_bh); if (status < 0) { mlog_errno(status); goto bail; @@ -802,8 +1113,7 @@ static int ocfs2_shift_tree_depth(struct ocfs2_super *osb, new_eb_bh = NULL; status = 0; bail: - if (new_eb_bh) - brelse(new_eb_bh); + brelse(new_eb_bh); mlog_exit(status); return status; @@ -818,22 +1128,21 @@ bail: * 1) a lowest extent block is found, then we pass it back in * *lowest_eb_bh and return '0' * - * 2) the search fails to find anything, but the dinode has room. We + * 2) the search fails to find anything, but the root_el has room. We * pass NULL back in *lowest_eb_bh, but still return '0' * - * 3) the search fails to find anything AND the dinode is full, in + * 3) the search fails to find anything AND the root_el is full, in * which case we return > 0 * * return status < 0 indicates an error. */ static int ocfs2_find_branch_target(struct ocfs2_super *osb, struct inode *inode, - struct buffer_head *fe_bh, + struct ocfs2_extent_tree *et, struct buffer_head **target_bh) { int status = 0, i; u64 blkno; - struct ocfs2_dinode *fe; struct ocfs2_extent_block *eb; struct ocfs2_extent_list *el; struct buffer_head *bh = NULL; @@ -843,8 +1152,7 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb, *target_bh = NULL; - fe = (struct ocfs2_dinode *) fe_bh->b_data; - el = &fe->id2.i_list; + el = et->et_root_el; while(le16_to_cpu(el->l_tree_depth) > 1) { if (le16_to_cpu(el->l_next_free_rec) == 0) { @@ -865,13 +1173,10 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb, goto bail; } - if (bh) { - brelse(bh); - bh = NULL; - } + brelse(bh); + bh = NULL; - status = ocfs2_read_block(osb, blkno, &bh, OCFS2_BH_CACHED, - inode); + status = ocfs2_read_block(inode, blkno, &bh, OCFS2_BH_CACHED); if (status < 0) { mlog_errno(status); goto bail; @@ -887,8 +1192,7 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb, if (le16_to_cpu(el->l_next_free_rec) < le16_to_cpu(el->l_count)) { - if (lowest_bh) - brelse(lowest_bh); + brelse(lowest_bh); lowest_bh = bh; get_bh(lowest_bh); } @@ -896,14 +1200,13 @@ static int ocfs2_find_branch_target(struct ocfs2_super *osb, /* If we didn't find one and the fe doesn't have any room, * then return '1' */ - if (!lowest_bh - && (fe->id2.i_list.l_next_free_rec == fe->id2.i_list.l_count)) + el = et->et_root_el; + if (!lowest_bh && (el->l_next_free_rec == el->l_count)) status = 1; *target_bh = lowest_bh; bail: - if (bh) - brelse(bh); + brelse(bh); mlog_exit(status); return status; @@ -920,19 +1223,19 @@ bail: * *last_eb_bh will be updated by ocfs2_add_branch(). */ static int ocfs2_grow_tree(struct inode *inode, handle_t *handle, - struct buffer_head *di_bh, int *final_depth, + struct ocfs2_extent_tree *et, int *final_depth, struct buffer_head **last_eb_bh, struct ocfs2_alloc_context *meta_ac) { int ret, shift; - struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; - int depth = le16_to_cpu(di->id2.i_list.l_tree_depth); + struct ocfs2_extent_list *el = et->et_root_el; + int depth = le16_to_cpu(el->l_tree_depth); struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct buffer_head *bh = NULL; BUG_ON(meta_ac == NULL); - shift = ocfs2_find_branch_target(osb, inode, di_bh, &bh); + shift = ocfs2_find_branch_target(osb, inode, et, &bh); if (shift < 0) { ret = shift; mlog_errno(ret); @@ -949,7 +1252,7 @@ static int ocfs2_grow_tree(struct inode *inode, handle_t *handle, /* ocfs2_shift_tree_depth will return us a buffer with * the new extent block (so we can pass that to * ocfs2_add_branch). */ - ret = ocfs2_shift_tree_depth(osb, handle, inode, di_bh, + ret = ocfs2_shift_tree_depth(osb, handle, inode, et, meta_ac, &bh); if (ret < 0) { mlog_errno(ret); @@ -976,7 +1279,7 @@ static int ocfs2_grow_tree(struct inode *inode, handle_t *handle, /* call ocfs2_add_branch to add the final part of the tree with * the new data. */ mlog(0, "add branch. bh = %p\n", bh); - ret = ocfs2_add_branch(osb, handle, inode, di_bh, bh, last_eb_bh, + ret = ocfs2_add_branch(osb, handle, inode, et, bh, last_eb_bh, meta_ac); if (ret < 0) { mlog_errno(ret); @@ -1237,8 +1540,7 @@ static int __ocfs2_find_path(struct inode *inode, brelse(bh); bh = NULL; - ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), blkno, - &bh, OCFS2_BH_CACHED, inode); + ret = ocfs2_read_block(inode, blkno, &bh, OCFS2_BH_CACHED); if (ret) { mlog_errno(ret); goto out; @@ -2059,11 +2361,11 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, struct ocfs2_path *right_path, int subtree_index, struct ocfs2_cached_dealloc_ctxt *dealloc, - int *deleted) + int *deleted, + struct ocfs2_extent_tree *et) { int ret, i, del_right_subtree = 0, right_has_empty = 0; - struct buffer_head *root_bh, *di_bh = path_root_bh(right_path); - struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct buffer_head *root_bh, *et_root_bh = path_root_bh(right_path); struct ocfs2_extent_list *right_leaf_el, *left_leaf_el; struct ocfs2_extent_block *eb; @@ -2115,7 +2417,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, * We have to update i_last_eb_blk during the meta * data delete. */ - ret = ocfs2_journal_access(handle, inode, di_bh, + ret = ocfs2_journal_access(handle, inode, et_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -2190,7 +2492,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, ocfs2_update_edge_lengths(inode, handle, left_path); eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data; - di->i_last_eb_blk = eb->h_blkno; + ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno)); /* * Removal of the extent in the left leaf was skipped @@ -2200,7 +2502,7 @@ static int ocfs2_rotate_subtree_left(struct inode *inode, handle_t *handle, if (right_has_empty) ocfs2_remove_empty_extent(left_leaf_el); - ret = ocfs2_journal_dirty(handle, di_bh); + ret = ocfs2_journal_dirty(handle, et_root_bh); if (ret) mlog_errno(ret); @@ -2323,7 +2625,8 @@ static int __ocfs2_rotate_tree_left(struct inode *inode, handle_t *handle, int orig_credits, struct ocfs2_path *path, struct ocfs2_cached_dealloc_ctxt *dealloc, - struct ocfs2_path **empty_extent_path) + struct ocfs2_path **empty_extent_path, + struct ocfs2_extent_tree *et) { int ret, subtree_root, deleted; u32 right_cpos; @@ -2396,7 +2699,7 @@ static int __ocfs2_rotate_tree_left(struct inode *inode, ret = ocfs2_rotate_subtree_left(inode, handle, left_path, right_path, subtree_root, - dealloc, &deleted); + dealloc, &deleted, et); if (ret == -EAGAIN) { /* * The rotation has to temporarily stop due to @@ -2439,29 +2742,20 @@ out: } static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle, - struct ocfs2_path *path, - struct ocfs2_cached_dealloc_ctxt *dealloc) + struct ocfs2_path *path, + struct ocfs2_cached_dealloc_ctxt *dealloc, + struct ocfs2_extent_tree *et) { int ret, subtree_index; u32 cpos; struct ocfs2_path *left_path = NULL; - struct ocfs2_dinode *di; struct ocfs2_extent_block *eb; struct ocfs2_extent_list *el; - /* - * XXX: This code assumes that the root is an inode, which is - * true for now but may change as tree code gets generic. - */ - di = (struct ocfs2_dinode *)path_root_bh(path)->b_data; - if (!OCFS2_IS_VALID_DINODE(di)) { - ret = -EIO; - ocfs2_error(inode->i_sb, - "Inode %llu has invalid path root", - (unsigned long long)OCFS2_I(inode)->ip_blkno); - goto out; - } + ret = ocfs2_et_sanity_check(inode, et); + if (ret) + goto out; /* * There's two ways we handle this depending on * whether path is the only existing one. @@ -2518,7 +2812,7 @@ static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle, ocfs2_update_edge_lengths(inode, handle, left_path); eb = (struct ocfs2_extent_block *)path_leaf_bh(left_path)->b_data; - di->i_last_eb_blk = eb->h_blkno; + ocfs2_et_set_last_eb_blk(et, le64_to_cpu(eb->h_blkno)); } else { /* * 'path' is also the leftmost path which @@ -2529,12 +2823,12 @@ static int ocfs2_remove_rightmost_path(struct inode *inode, handle_t *handle, */ ocfs2_unlink_path(inode, handle, dealloc, path, 1); - el = &di->id2.i_list; + el = et->et_root_el; el->l_tree_depth = 0; el->l_next_free_rec = 0; memset(&el->l_recs[0], 0, sizeof(struct ocfs2_extent_rec)); - di->i_last_eb_blk = 0; + ocfs2_et_set_last_eb_blk(et, 0); } ocfs2_journal_dirty(handle, path_root_bh(path)); @@ -2562,7 +2856,8 @@ out: */ static int ocfs2_rotate_tree_left(struct inode *inode, handle_t *handle, struct ocfs2_path *path, - struct ocfs2_cached_dealloc_ctxt *dealloc) + struct ocfs2_cached_dealloc_ctxt *dealloc, + struct ocfs2_extent_tree *et) { int ret, orig_credits = handle->h_buffer_credits; struct ocfs2_path *tmp_path = NULL, *restart_path = NULL; @@ -2576,7 +2871,7 @@ static int ocfs2_rotate_tree_left(struct inode *inode, handle_t *handle, if (path->p_tree_depth == 0) { rightmost_no_delete: /* - * In-inode extents. This is trivially handled, so do + * Inline extents. This is trivially handled, so do * it up front. */ ret = ocfs2_rotate_rightmost_leaf_left(inode, handle, @@ -2630,7 +2925,7 @@ rightmost_no_delete: */ ret = ocfs2_remove_rightmost_path(inode, handle, path, - dealloc); + dealloc, et); if (ret) mlog_errno(ret); goto out; @@ -2642,7 +2937,7 @@ rightmost_no_delete: */ try_rotate: ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits, path, - dealloc, &restart_path); + dealloc, &restart_path, et); if (ret && ret != -EAGAIN) { mlog_errno(ret); goto out; @@ -2654,7 +2949,7 @@ try_rotate: ret = __ocfs2_rotate_tree_left(inode, handle, orig_credits, tmp_path, dealloc, - &restart_path); + &restart_path, et); if (ret && ret != -EAGAIN) { mlog_errno(ret); goto out; @@ -2940,6 +3235,7 @@ static int ocfs2_merge_rec_left(struct inode *inode, handle_t *handle, struct ocfs2_extent_rec *split_rec, struct ocfs2_cached_dealloc_ctxt *dealloc, + struct ocfs2_extent_tree *et, int index) { int ret, i, subtree_index = 0, has_empty_extent = 0; @@ -3060,7 +3356,8 @@ static int ocfs2_merge_rec_left(struct inode *inode, le16_to_cpu(el->l_next_free_rec) == 1) { ret = ocfs2_remove_rightmost_path(inode, handle, - right_path, dealloc); + right_path, + dealloc, et); if (ret) { mlog_errno(ret); goto out; @@ -3087,7 +3384,8 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, int split_index, struct ocfs2_extent_rec *split_rec, struct ocfs2_cached_dealloc_ctxt *dealloc, - struct ocfs2_merge_ctxt *ctxt) + struct ocfs2_merge_ctxt *ctxt, + struct ocfs2_extent_tree *et) { int ret = 0; @@ -3105,7 +3403,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, * illegal. */ ret = ocfs2_rotate_tree_left(inode, handle, path, - dealloc); + dealloc, et); if (ret) { mlog_errno(ret); goto out; @@ -3148,7 +3446,8 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, BUG_ON(!ocfs2_is_empty_extent(&el->l_recs[0])); /* The merge left us with an empty extent, remove it. */ - ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc); + ret = ocfs2_rotate_tree_left(inode, handle, path, + dealloc, et); if (ret) { mlog_errno(ret); goto out; @@ -3162,7 +3461,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, */ ret = ocfs2_merge_rec_left(inode, path, handle, rec, - dealloc, + dealloc, et, split_index); if (ret) { @@ -3171,7 +3470,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, } ret = ocfs2_rotate_tree_left(inode, handle, path, - dealloc); + dealloc, et); /* * Error from this last rotate is not critical, so * print but don't bubble it up. @@ -3191,7 +3490,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, ret = ocfs2_merge_rec_left(inode, path, handle, split_rec, - dealloc, + dealloc, et, split_index); if (ret) { mlog_errno(ret); @@ -3214,7 +3513,7 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, * our leaf. Try to rotate it away. */ ret = ocfs2_rotate_tree_left(inode, handle, path, - dealloc); + dealloc, et); if (ret) mlog_errno(ret); ret = 0; @@ -3348,16 +3647,6 @@ rotate: ocfs2_rotate_leaf(el, insert_rec); } -static inline void ocfs2_update_dinode_clusters(struct inode *inode, - struct ocfs2_dinode *di, - u32 clusters) -{ - le32_add_cpu(&di->i_clusters, clusters); - spin_lock(&OCFS2_I(inode)->ip_lock); - OCFS2_I(inode)->ip_clusters = le32_to_cpu(di->i_clusters); - spin_unlock(&OCFS2_I(inode)->ip_lock); -} - static void ocfs2_adjust_rightmost_records(struct inode *inode, handle_t *handle, struct ocfs2_path *path, @@ -3559,8 +3848,8 @@ static void ocfs2_split_record(struct inode *inode, } /* - * This function only does inserts on an allocation b-tree. For dinode - * lists, ocfs2_insert_at_leaf() is called directly. + * This function only does inserts on an allocation b-tree. For tree + * depth = 0, ocfs2_insert_at_leaf() is called directly. * * right_path is the path we want to do the actual insert * in. left_path should only be passed in if we need to update that @@ -3657,7 +3946,7 @@ out: static int ocfs2_do_insert_extent(struct inode *inode, handle_t *handle, - struct buffer_head *di_bh, + struct ocfs2_extent_tree *et, struct ocfs2_extent_rec *insert_rec, struct ocfs2_insert_type *type) { @@ -3665,13 +3954,11 @@ static int ocfs2_do_insert_extent(struct inode *inode, u32 cpos; struct ocfs2_path *right_path = NULL; struct ocfs2_path *left_path = NULL; - struct ocfs2_dinode *di; struct ocfs2_extent_list *el; - di = (struct ocfs2_dinode *) di_bh->b_data; - el = &di->id2.i_list; + el = et->et_root_el; - ret = ocfs2_journal_access(handle, inode, di_bh, + ret = ocfs2_journal_access(handle, inode, et->et_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -3683,7 +3970,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, goto out_update_clusters; } - right_path = ocfs2_new_inode_path(di_bh); + right_path = ocfs2_new_path(et->et_root_bh, et->et_root_el); if (!right_path) { ret = -ENOMEM; mlog_errno(ret); @@ -3733,7 +4020,7 @@ static int ocfs2_do_insert_extent(struct inode *inode, * ocfs2_rotate_tree_right() might have extended the * transaction without re-journaling our tree root. */ - ret = ocfs2_journal_access(handle, inode, di_bh, + ret = ocfs2_journal_access(handle, inode, et->et_root_bh, OCFS2_JOURNAL_ACCESS_WRITE); if (ret) { mlog_errno(ret); @@ -3758,10 +4045,10 @@ static int ocfs2_do_insert_extent(struct inode *inode, out_update_clusters: if (type->ins_split == SPLIT_NONE) - ocfs2_update_dinode_clusters(inode, di, - le16_to_cpu(insert_rec->e_leaf_clusters)); + ocfs2_et_update_clusters(inode, et, + le16_to_cpu(insert_rec->e_leaf_clusters)); - ret = ocfs2_journal_dirty(handle, di_bh); + ret = ocfs2_journal_dirty(handle, et->et_root_bh); if (ret) mlog_errno(ret); @@ -3891,7 +4178,8 @@ out: static void ocfs2_figure_contig_type(struct inode *inode, struct ocfs2_insert_type *insert, struct ocfs2_extent_list *el, - struct ocfs2_extent_rec *insert_rec) + struct ocfs2_extent_rec *insert_rec, + struct ocfs2_extent_tree *et) { int i; enum ocfs2_contig_type contig_type = CONTIG_NONE; @@ -3907,6 +4195,21 @@ static void ocfs2_figure_contig_type(struct inode *inode, } } insert->ins_contig = contig_type; + + if (insert->ins_contig != CONTIG_NONE) { + struct ocfs2_extent_rec *rec = + &el->l_recs[insert->ins_contig_index]; + unsigned int len = le16_to_cpu(rec->e_leaf_clusters) + + le16_to_cpu(insert_rec->e_leaf_clusters); + + /* + * Caller might want us to limit the size of extents, don't + * calculate contiguousness if we might exceed that limit. + */ + if (et->et_max_leaf_clusters && + (len > et->et_max_leaf_clusters)) + insert->ins_contig = CONTIG_NONE; + } } /* @@ -3915,8 +4218,8 @@ static void ocfs2_figure_contig_type(struct inode *inode, * ocfs2_figure_appending_type() will figure out whether we'll have to * insert at the tail of the rightmost leaf. * - * This should also work against the dinode list for tree's with 0 - * depth. If we consider the dinode list to be the rightmost leaf node + * This should also work against the root extent list for tree's with 0 + * depth. If we consider the root extent list to be the rightmost leaf node * then the logic here makes sense. */ static void ocfs2_figure_appending_type(struct ocfs2_insert_type *insert, @@ -3967,14 +4270,13 @@ set_tail_append: * structure. */ static int ocfs2_figure_insert_type(struct inode *inode, - struct buffer_head *di_bh, + struct ocfs2_extent_tree *et, struct buffer_head **last_eb_bh, struct ocfs2_extent_rec *insert_rec, int *free_records, struct ocfs2_insert_type *insert) { int ret; - struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; struct ocfs2_extent_block *eb; struct ocfs2_extent_list *el; struct ocfs2_path *path = NULL; @@ -3982,7 +4284,7 @@ static int ocfs2_figure_insert_type(struct inode *inode, insert->ins_split = SPLIT_NONE; - el = &di->id2.i_list; + el = et->et_root_el; insert->ins_tree_depth = le16_to_cpu(el->l_tree_depth); if (el->l_tree_depth) { @@ -3992,9 +4294,9 @@ static int ocfs2_figure_insert_type(struct inode *inode, * ocfs2_figure_insert_type() and ocfs2_add_branch() * may want it later. */ - ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), - le64_to_cpu(di->i_last_eb_blk), &bh, - OCFS2_BH_CACHED, inode); + ret = ocfs2_read_block(inode, + ocfs2_et_get_last_eb_blk(et), &bh, + OCFS2_BH_CACHED); if (ret) { mlog_exit(ret); goto out; @@ -4015,12 +4317,12 @@ static int ocfs2_figure_insert_type(struct inode *inode, le16_to_cpu(el->l_next_free_rec); if (!insert->ins_tree_depth) { - ocfs2_figure_contig_type(inode, insert, el, insert_rec); + ocfs2_figure_contig_type(inode, insert, el, insert_rec, et); ocfs2_figure_appending_type(insert, el, insert_rec); return 0; } - path = ocfs2_new_inode_path(di_bh); + path = ocfs2_new_path(et->et_root_bh, et->et_root_el); if (!path) { ret = -ENOMEM; mlog_errno(ret); @@ -4049,7 +4351,7 @@ static int ocfs2_figure_insert_type(struct inode *inode, * into two types of appends: simple record append, or a * rotate inside the tail leaf. */ - ocfs2_figure_contig_type(inode, insert, el, insert_rec); + ocfs2_figure_contig_type(inode, insert, el, insert_rec, et); /* * The insert code isn't quite ready to deal with all cases of @@ -4070,7 +4372,8 @@ static int ocfs2_figure_insert_type(struct inode *inode, * the case that we're doing a tail append, so maybe we can * take advantage of that information somehow. */ - if (le64_to_cpu(di->i_last_eb_blk) == path_leaf_bh(path)->b_blocknr) { + if (ocfs2_et_get_last_eb_blk(et) == + path_leaf_bh(path)->b_blocknr) { /* * Ok, ocfs2_find_path() returned us the rightmost * tree path. This might be an appending insert. There are @@ -4100,7 +4403,7 @@ out: int ocfs2_insert_extent(struct ocfs2_super *osb, handle_t *handle, struct inode *inode, - struct buffer_head *fe_bh, + struct ocfs2_extent_tree *et, u32 cpos, u64 start_blk, u32 new_clusters, @@ -4113,26 +4416,21 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, struct ocfs2_insert_type insert = {0, }; struct ocfs2_extent_rec rec; - BUG_ON(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL); - mlog(0, "add %u clusters at position %u to inode %llu\n", new_clusters, cpos, (unsigned long long)OCFS2_I(inode)->ip_blkno); - mlog_bug_on_msg(!ocfs2_sparse_alloc(osb) && - (OCFS2_I(inode)->ip_clusters != cpos), - "Device %s, asking for sparse allocation: inode %llu, " - "cpos %u, clusters %u\n", - osb->dev_str, - (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos, - OCFS2_I(inode)->ip_clusters); - memset(&rec, 0, sizeof(rec)); rec.e_cpos = cpu_to_le32(cpos); rec.e_blkno = cpu_to_le64(start_blk); rec.e_leaf_clusters = cpu_to_le16(new_clusters); rec.e_flags = flags; + status = ocfs2_et_insert_check(inode, et, &rec); + if (status) { + mlog_errno(status); + goto bail; + } - status = ocfs2_figure_insert_type(inode, fe_bh, &last_eb_bh, &rec, + status = ocfs2_figure_insert_type(inode, et, &last_eb_bh, &rec, &free_records, &insert); if (status < 0) { mlog_errno(status); @@ -4146,7 +4444,7 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, free_records, insert.ins_tree_depth); if (insert.ins_contig == CONTIG_NONE && free_records == 0) { - status = ocfs2_grow_tree(inode, handle, fe_bh, + status = ocfs2_grow_tree(inode, handle, et, &insert.ins_tree_depth, &last_eb_bh, meta_ac); if (status) { @@ -4156,17 +4454,124 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, } /* Finally, we can add clusters. This might rotate the tree for us. */ - status = ocfs2_do_insert_extent(inode, handle, fe_bh, &rec, &insert); + status = ocfs2_do_insert_extent(inode, handle, et, &rec, &insert); if (status < 0) mlog_errno(status); - else + else if (et->et_ops == &ocfs2_dinode_et_ops) ocfs2_extent_map_insert_rec(inode, &rec); bail: - if (last_eb_bh) - brelse(last_eb_bh); + brelse(last_eb_bh); + + mlog_exit(status); + return status; +} + +/* + * Allcate and add clusters into the extent b-tree. + * The new clusters(clusters_to_add) will be inserted at logical_offset. + * The extent b-tree's root is specified by et, and + * it is not limited to the file storage. Any extent tree can use this + * function if it implements the proper ocfs2_extent_tree. + */ +int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb, + struct inode *inode, + u32 *logical_offset, + u32 clusters_to_add, + int mark_unwritten, + struct ocfs2_extent_tree *et, + handle_t *handle, + struct ocfs2_alloc_context *data_ac, + struct ocfs2_alloc_context *meta_ac, + enum ocfs2_alloc_restarted *reason_ret) +{ + int status = 0; + int free_extents; + enum ocfs2_alloc_restarted reason = RESTART_NONE; + u32 bit_off, num_bits; + u64 block; + u8 flags = 0; + + BUG_ON(!clusters_to_add); + + if (mark_unwritten) + flags = OCFS2_EXT_UNWRITTEN; + + free_extents = ocfs2_num_free_extents(osb, inode, et); + if (free_extents < 0) { + status = free_extents; + mlog_errno(status); + goto leave; + } + + /* there are two cases which could cause us to EAGAIN in the + * we-need-more-metadata case: + * 1) we haven't reserved *any* + * 2) we are so fragmented, we've needed to add metadata too + * many times. */ + if (!free_extents && !meta_ac) { + mlog(0, "we haven't reserved any metadata!\n"); + status = -EAGAIN; + reason = RESTART_META; + goto leave; + } else if ((!free_extents) + && (ocfs2_alloc_context_bits_left(meta_ac) + < ocfs2_extend_meta_needed(et->et_root_el))) { + mlog(0, "filesystem is really fragmented...\n"); + status = -EAGAIN; + reason = RESTART_META; + goto leave; + } + + status = __ocfs2_claim_clusters(osb, handle, data_ac, 1, + clusters_to_add, &bit_off, &num_bits); + if (status < 0) { + if (status != -ENOSPC) + mlog_errno(status); + goto leave; + } + + BUG_ON(num_bits > clusters_to_add); + /* reserve our write early -- insert_extent may update the inode */ + status = ocfs2_journal_access(handle, inode, et->et_root_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (status < 0) { + mlog_errno(status); + goto leave; + } + + block = ocfs2_clusters_to_blocks(osb->sb, bit_off); + mlog(0, "Allocating %u clusters at block %u for inode %llu\n", + num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno); + status = ocfs2_insert_extent(osb, handle, inode, et, + *logical_offset, block, + num_bits, flags, meta_ac); + if (status < 0) { + mlog_errno(status); + goto leave; + } + + status = ocfs2_journal_dirty(handle, et->et_root_bh); + if (status < 0) { + mlog_errno(status); + goto leave; + } + + clusters_to_add -= num_bits; + *logical_offset += num_bits; + + if (clusters_to_add) { + mlog(0, "need to alloc once more, wanted = %u\n", + clusters_to_add); + status = -EAGAIN; + reason = RESTART_TRANS; + } + +leave: mlog_exit(status); + if (reason_ret) + *reason_ret = reason; return status; } @@ -4193,7 +4598,7 @@ static void ocfs2_make_right_split_rec(struct super_block *sb, static int ocfs2_split_and_insert(struct inode *inode, handle_t *handle, struct ocfs2_path *path, - struct buffer_head *di_bh, + struct ocfs2_extent_tree *et, struct buffer_head **last_eb_bh, int split_index, struct ocfs2_extent_rec *orig_split_rec, @@ -4207,7 +4612,6 @@ static int ocfs2_split_and_insert(struct inode *inode, struct ocfs2_extent_rec split_rec = *orig_split_rec; struct ocfs2_insert_type insert; struct ocfs2_extent_block *eb; - struct ocfs2_dinode *di; leftright: /* @@ -4216,8 +4620,7 @@ leftright: */ rec = path_leaf_el(path)->l_recs[split_index]; - di = (struct ocfs2_dinode *)di_bh->b_data; - rightmost_el = &di->id2.i_list; + rightmost_el = et->et_root_el; depth = le16_to_cpu(rightmost_el->l_tree_depth); if (depth) { @@ -4228,8 +4631,8 @@ leftright: if (le16_to_cpu(rightmost_el->l_next_free_rec) == le16_to_cpu(rightmost_el->l_count)) { - ret = ocfs2_grow_tree(inode, handle, di_bh, &depth, last_eb_bh, - meta_ac); + ret = ocfs2_grow_tree(inode, handle, et, + &depth, last_eb_bh, meta_ac); if (ret) { mlog_errno(ret); goto out; @@ -4266,8 +4669,7 @@ leftright: do_leftright = 1; } - ret = ocfs2_do_insert_extent(inode, handle, di_bh, &split_rec, - &insert); + ret = ocfs2_do_insert_extent(inode, handle, et, &split_rec, &insert); if (ret) { mlog_errno(ret); goto out; @@ -4309,8 +4711,9 @@ out: * of the tree is required. All other cases will degrade into a less * optimal tree layout. * - * last_eb_bh should be the rightmost leaf block for any inode with a - * btree. Since a split may grow the tree or a merge might shrink it, the caller cannot trust the contents of that buffer after this call. + * last_eb_bh should be the rightmost leaf block for any extent + * btree. Since a split may grow the tree or a merge might shrink it, + * the caller cannot trust the contents of that buffer after this call. * * This code is optimized for readability - several passes might be * made over certain portions of the tree. All of those blocks will @@ -4318,7 +4721,7 @@ out: * extra overhead is not expressed in terms of disk reads. */ static int __ocfs2_mark_extent_written(struct inode *inode, - struct buffer_head *di_bh, + struct ocfs2_extent_tree *et, handle_t *handle, struct ocfs2_path *path, int split_index, @@ -4358,11 +4761,10 @@ static int __ocfs2_mark_extent_written(struct inode *inode, */ if (path->p_tree_depth) { struct ocfs2_extent_block *eb; - struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; - ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), - le64_to_cpu(di->i_last_eb_blk), - &last_eb_bh, OCFS2_BH_CACHED, inode); + ret = ocfs2_read_block(inode, + ocfs2_et_get_last_eb_blk(et), + &last_eb_bh, OCFS2_BH_CACHED); if (ret) { mlog_exit(ret); goto out; @@ -4395,7 +4797,7 @@ static int __ocfs2_mark_extent_written(struct inode *inode, if (ctxt.c_split_covers_rec) el->l_recs[split_index] = *split_rec; else - ret = ocfs2_split_and_insert(inode, handle, path, di_bh, + ret = ocfs2_split_and_insert(inode, handle, path, et, &last_eb_bh, split_index, split_rec, meta_ac); if (ret) @@ -4403,7 +4805,7 @@ static int __ocfs2_mark_extent_written(struct inode *inode, } else { ret = ocfs2_try_to_merge_extent(inode, handle, path, split_index, split_rec, - dealloc, &ctxt); + dealloc, &ctxt, et); if (ret) mlog_errno(ret); } @@ -4421,7 +4823,8 @@ out: * * The caller is responsible for passing down meta_ac if we'll need it. */ -int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh, +int ocfs2_mark_extent_written(struct inode *inode, + struct ocfs2_extent_tree *et, handle_t *handle, u32 cpos, u32 len, u32 phys, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) @@ -4447,10 +4850,14 @@ int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh, /* * XXX: This should be fixed up so that we just re-insert the * next extent records. + * + * XXX: This is a hack on the extent tree, maybe it should be + * an op? */ - ocfs2_extent_map_trunc(inode, 0); + if (et->et_ops == &ocfs2_dinode_et_ops) + ocfs2_extent_map_trunc(inode, 0); - left_path = ocfs2_new_inode_path(di_bh); + left_path = ocfs2_new_path(et->et_root_bh, et->et_root_el); if (!left_path) { ret = -ENOMEM; mlog_errno(ret); @@ -4481,8 +4888,9 @@ int ocfs2_mark_extent_written(struct inode *inode, struct buffer_head *di_bh, split_rec.e_flags = path_leaf_el(left_path)->l_recs[index].e_flags; split_rec.e_flags &= ~OCFS2_EXT_UNWRITTEN; - ret = __ocfs2_mark_extent_written(inode, di_bh, handle, left_path, - index, &split_rec, meta_ac, dealloc); + ret = __ocfs2_mark_extent_written(inode, et, handle, left_path, + index, &split_rec, meta_ac, + dealloc); if (ret) mlog_errno(ret); @@ -4491,13 +4899,12 @@ out: return ret; } -static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh, +static int ocfs2_split_tree(struct inode *inode, struct ocfs2_extent_tree *et, handle_t *handle, struct ocfs2_path *path, int index, u32 new_range, struct ocfs2_alloc_context *meta_ac) { int ret, depth, credits = handle->h_buffer_credits; - struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; struct buffer_head *last_eb_bh = NULL; struct ocfs2_extent_block *eb; struct ocfs2_extent_list *rightmost_el, *el; @@ -4514,9 +4921,9 @@ static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh, depth = path->p_tree_depth; if (depth > 0) { - ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), - le64_to_cpu(di->i_last_eb_blk), - &last_eb_bh, OCFS2_BH_CACHED, inode); + ret = ocfs2_read_block(inode, + ocfs2_et_get_last_eb_blk(et), + &last_eb_bh, OCFS2_BH_CACHED); if (ret < 0) { mlog_errno(ret); goto out; @@ -4527,7 +4934,8 @@ static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh, } else rightmost_el = path_leaf_el(path); - credits += path->p_tree_depth + ocfs2_extend_meta_needed(di); + credits += path->p_tree_depth + + ocfs2_extend_meta_needed(et->et_root_el); ret = ocfs2_extend_trans(handle, credits); if (ret) { mlog_errno(ret); @@ -4536,7 +4944,7 @@ static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh, if (le16_to_cpu(rightmost_el->l_next_free_rec) == le16_to_cpu(rightmost_el->l_count)) { - ret = ocfs2_grow_tree(inode, handle, di_bh, &depth, &last_eb_bh, + ret = ocfs2_grow_tree(inode, handle, et, &depth, &last_eb_bh, meta_ac); if (ret) { mlog_errno(ret); @@ -4550,7 +4958,7 @@ static int ocfs2_split_tree(struct inode *inode, struct buffer_head *di_bh, insert.ins_split = SPLIT_RIGHT; insert.ins_tree_depth = depth; - ret = ocfs2_do_insert_extent(inode, handle, di_bh, &split_rec, &insert); + ret = ocfs2_do_insert_extent(inode, handle, et, &split_rec, &insert); if (ret) mlog_errno(ret); @@ -4562,7 +4970,8 @@ out: static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, struct ocfs2_path *path, int index, struct ocfs2_cached_dealloc_ctxt *dealloc, - u32 cpos, u32 len) + u32 cpos, u32 len, + struct ocfs2_extent_tree *et) { int ret; u32 left_cpos, rec_range, trunc_range; @@ -4574,7 +4983,7 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, struct ocfs2_extent_block *eb; if (ocfs2_is_empty_extent(&el->l_recs[0]) && index > 0) { - ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc); + ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc, et); if (ret) { mlog_errno(ret); goto out; @@ -4705,7 +5114,7 @@ static int ocfs2_truncate_rec(struct inode *inode, handle_t *handle, ocfs2_journal_dirty(handle, path_leaf_bh(path)); - ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc); + ret = ocfs2_rotate_tree_left(inode, handle, path, dealloc, et); if (ret) { mlog_errno(ret); goto out; @@ -4716,7 +5125,8 @@ out: return ret; } -int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh, +int ocfs2_remove_extent(struct inode *inode, + struct ocfs2_extent_tree *et, u32 cpos, u32 len, handle_t *handle, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) @@ -4725,11 +5135,11 @@ int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh, u32 rec_range, trunc_range; struct ocfs2_extent_rec *rec; struct ocfs2_extent_list *el; - struct ocfs2_path *path; + struct ocfs2_path *path = NULL; ocfs2_extent_map_trunc(inode, 0); - path = ocfs2_new_inode_path(di_bh); + path = ocfs2_new_path(et->et_root_bh, et->et_root_el); if (!path) { ret = -ENOMEM; mlog_errno(ret); @@ -4782,13 +5192,13 @@ int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh, if (le32_to_cpu(rec->e_cpos) == cpos || rec_range == trunc_range) { ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc, - cpos, len); + cpos, len, et); if (ret) { mlog_errno(ret); goto out; } } else { - ret = ocfs2_split_tree(inode, di_bh, handle, path, index, + ret = ocfs2_split_tree(inode, et, handle, path, index, trunc_range, meta_ac); if (ret) { mlog_errno(ret); @@ -4837,7 +5247,7 @@ int ocfs2_remove_extent(struct inode *inode, struct buffer_head *di_bh, } ret = ocfs2_truncate_rec(inode, handle, path, index, dealloc, - cpos, len); + cpos, len, et); if (ret) { mlog_errno(ret); goto out; @@ -5180,8 +5590,8 @@ static int ocfs2_get_truncate_log_info(struct ocfs2_super *osb, goto bail; } - status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, &bh, - OCFS2_BH_CACHED, inode); + status = ocfs2_read_block(inode, OCFS2_I(inode)->ip_blkno, &bh, + OCFS2_BH_CACHED); if (status < 0) { iput(inode); mlog_errno(status); @@ -5256,8 +5666,7 @@ int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, bail: if (tl_inode) iput(tl_inode); - if (tl_bh) - brelse(tl_bh); + brelse(tl_bh); if (status < 0 && (*tl_copy)) { kfree(*tl_copy); @@ -6000,20 +6409,13 @@ bail: return status; } -static int ocfs2_writeback_zero_func(handle_t *handle, struct buffer_head *bh) +static int ocfs2_zero_func(handle_t *handle, struct buffer_head *bh) { set_buffer_uptodate(bh); mark_buffer_dirty(bh); return 0; } -static int ocfs2_ordered_zero_func(handle_t *handle, struct buffer_head *bh) -{ - set_buffer_uptodate(bh); - mark_buffer_dirty(bh); - return ocfs2_journal_dirty_data(handle, bh); -} - static void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle, unsigned int from, unsigned int to, struct page *page, int zero, u64 *phys) @@ -6032,17 +6434,18 @@ static void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle, * here if they aren't - ocfs2_map_page_blocks() * might've skipped some */ - if (ocfs2_should_order_data(inode)) { - ret = walk_page_buffers(handle, - page_buffers(page), - from, to, &partial, - ocfs2_ordered_zero_func); - if (ret < 0) - mlog_errno(ret); - } else { + ret = walk_page_buffers(handle, page_buffers(page), + from, to, &partial, + ocfs2_zero_func); + if (ret < 0) + mlog_errno(ret); + else if (ocfs2_should_order_data(inode)) { + ret = ocfs2_jbd2_file_inode(handle, inode); +#ifdef CONFIG_OCFS2_COMPAT_JBD ret = walk_page_buffers(handle, page_buffers(page), from, to, &partial, - ocfs2_writeback_zero_func); + ocfs2_journal_dirty_data); +#endif if (ret < 0) mlog_errno(ret); } @@ -6207,20 +6610,29 @@ out: return ret; } -static void ocfs2_zero_dinode_id2(struct inode *inode, struct ocfs2_dinode *di) +static void ocfs2_zero_dinode_id2_with_xattr(struct inode *inode, + struct ocfs2_dinode *di) { unsigned int blocksize = 1 << inode->i_sb->s_blocksize_bits; + unsigned int xattrsize = le16_to_cpu(di->i_xattr_inline_size); - memset(&di->id2, 0, blocksize - offsetof(struct ocfs2_dinode, id2)); + if (le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_XATTR_FL) + memset(&di->id2, 0, blocksize - + offsetof(struct ocfs2_dinode, id2) - + xattrsize); + else + memset(&di->id2, 0, blocksize - + offsetof(struct ocfs2_dinode, id2)); } void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di) { - ocfs2_zero_dinode_id2(inode, di); + ocfs2_zero_dinode_id2_with_xattr(inode, di); di->id2.i_list.l_tree_depth = 0; di->id2.i_list.l_next_free_rec = 0; - di->id2.i_list.l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(inode->i_sb)); + di->id2.i_list.l_count = cpu_to_le16( + ocfs2_extent_recs_per_inode_with_xattr(inode->i_sb, di)); } void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di) @@ -6237,9 +6649,10 @@ void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di) * We clear the entire i_data structure here so that all * fields can be properly initialized. */ - ocfs2_zero_dinode_id2(inode, di); + ocfs2_zero_dinode_id2_with_xattr(inode, di); - idata->id_count = cpu_to_le16(ocfs2_max_inline_data(inode->i_sb)); + idata->id_count = cpu_to_le16( + ocfs2_max_inline_data_with_xattr(inode->i_sb, di)); } int ocfs2_convert_inline_data_to_extents(struct inode *inode, @@ -6254,6 +6667,7 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, struct ocfs2_alloc_context *data_ac = NULL; struct page **pages = NULL; loff_t end = osb->s_clustersize; + struct ocfs2_extent_tree et; has_data = i_size_read(inode) ? 1 : 0; @@ -6353,7 +6767,8 @@ int ocfs2_convert_inline_data_to_extents(struct inode *inode, * this proves to be false, we could always re-build * the in-inode data from our pages. */ - ret = ocfs2_insert_extent(osb, handle, inode, di_bh, + ocfs2_init_dinode_extent_tree(&et, inode, di_bh); + ret = ocfs2_insert_extent(osb, handle, inode, &et, 0, block, 1, 0, NULL); if (ret) { mlog_errno(ret); @@ -6396,13 +6811,14 @@ int ocfs2_commit_truncate(struct ocfs2_super *osb, handle_t *handle = NULL; struct inode *tl_inode = osb->osb_tl_inode; struct ocfs2_path *path = NULL; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; mlog_entry_void(); new_highest_cpos = ocfs2_clusters_for_bytes(osb->sb, i_size_read(inode)); - path = ocfs2_new_inode_path(fe_bh); + path = ocfs2_new_path(fe_bh, &di->id2.i_list); if (!path) { status = -ENOMEM; mlog_errno(status); @@ -6573,8 +6989,8 @@ int ocfs2_prepare_truncate(struct ocfs2_super *osb, ocfs2_init_dealloc_ctxt(&(*tc)->tc_dealloc); if (fe->id2.i_list.l_tree_depth) { - status = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk), - &last_eb_bh, OCFS2_BH_CACHED, inode); + status = ocfs2_read_block(inode, le64_to_cpu(fe->i_last_eb_blk), + &last_eb_bh, OCFS2_BH_CACHED); if (status < 0) { mlog_errno(status); goto bail; @@ -6687,8 +7103,7 @@ static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc) mlog(ML_NOTICE, "Truncate completion has non-empty dealloc context\n"); - if (tc->tc_last_eb_bh) - brelse(tc->tc_last_eb_bh); + brelse(tc->tc_last_eb_bh); kfree(tc); }