diff options
Diffstat (limited to 'fs/ocfs2')
36 files changed, 2673 insertions, 1219 deletions
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c index f5e11f4fa95..4ba7f0bdc24 100644 --- a/fs/ocfs2/alloc.c +++ b/fs/ocfs2/alloc.c @@ -354,7 +354,6 @@ struct ocfs2_insert_type { enum ocfs2_append_type ins_appending; enum ocfs2_contig_type ins_contig; int ins_contig_index; - int ins_free_records; int ins_tree_depth; }; @@ -362,7 +361,6 @@ struct ocfs2_merge_ctxt { enum ocfs2_contig_type c_contig_type; int c_has_empty_extent; int c_split_covers_rec; - int c_used_tail_recs; }; /* @@ -2808,36 +2806,28 @@ static int ocfs2_try_to_merge_extent(struct inode *inode, struct ocfs2_merge_ctxt *ctxt) { - int ret = 0, delete_tail_recs = 0; + int ret = 0; struct ocfs2_extent_list *el = path_leaf_el(left_path); struct ocfs2_extent_rec *rec = &el->l_recs[split_index]; BUG_ON(ctxt->c_contig_type == CONTIG_NONE); - if (ctxt->c_split_covers_rec) { - delete_tail_recs++; - - if (ctxt->c_contig_type == CONTIG_LEFTRIGHT || - ctxt->c_has_empty_extent) - delete_tail_recs++; - - if (ctxt->c_has_empty_extent) { - /* - * The merge code will need to create an empty - * extent to take the place of the newly - * emptied slot. Remove any pre-existing empty - * extents - having more than one in a leaf is - * illegal. - */ - ret = ocfs2_rotate_tree_left(inode, handle, left_path, - dealloc); - if (ret) { - mlog_errno(ret); - goto out; - } - split_index--; - rec = &el->l_recs[split_index]; + if (ctxt->c_split_covers_rec && ctxt->c_has_empty_extent) { + /* + * The merge code will need to create an empty + * extent to take the place of the newly + * emptied slot. Remove any pre-existing empty + * extents - having more than one in a leaf is + * illegal. + */ + ret = ocfs2_rotate_tree_left(inode, handle, left_path, + dealloc); + if (ret) { + mlog_errno(ret); + goto out; } + split_index--; + rec = &el->l_recs[split_index]; } if (ctxt->c_contig_type == CONTIG_LEFTRIGHT) { @@ -3593,6 +3583,7 @@ static int ocfs2_figure_insert_type(struct inode *inode, struct buffer_head *di_bh, struct buffer_head **last_eb_bh, struct ocfs2_extent_rec *insert_rec, + int *free_records, struct ocfs2_insert_type *insert) { int ret; @@ -3633,7 +3624,7 @@ static int ocfs2_figure_insert_type(struct inode *inode, * XXX: This test is simplistic, we can search for empty * extent records too. */ - insert->ins_free_records = le16_to_cpu(el->l_count) - + *free_records = le16_to_cpu(el->l_count) - le16_to_cpu(el->l_next_free_rec); if (!insert->ins_tree_depth) { @@ -3730,11 +3721,13 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, struct ocfs2_alloc_context *meta_ac) { int status; + int uninitialized_var(free_records); struct buffer_head *last_eb_bh = NULL; - struct buffer_head *bh = NULL; 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); @@ -3753,7 +3746,7 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, rec.e_flags = flags; status = ocfs2_figure_insert_type(inode, fe_bh, &last_eb_bh, &rec, - &insert); + &free_records, &insert); if (status < 0) { mlog_errno(status); goto bail; @@ -3763,9 +3756,9 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, "Insert.contig_index: %d, Insert.free_records: %d, " "Insert.tree_depth: %d\n", insert.ins_appending, insert.ins_contig, insert.ins_contig_index, - insert.ins_free_records, insert.ins_tree_depth); + free_records, insert.ins_tree_depth); - if (insert.ins_contig == CONTIG_NONE && insert.ins_free_records == 0) { + if (insert.ins_contig == CONTIG_NONE && free_records == 0) { status = ocfs2_grow_tree(inode, handle, fe_bh, &insert.ins_tree_depth, &last_eb_bh, meta_ac); @@ -3783,9 +3776,6 @@ int ocfs2_insert_extent(struct ocfs2_super *osb, ocfs2_extent_map_insert_rec(inode, &rec); bail: - if (bh) - brelse(bh); - if (last_eb_bh) brelse(last_eb_bh); @@ -3851,26 +3841,17 @@ leftright: if (le16_to_cpu(rightmost_el->l_next_free_rec) == le16_to_cpu(rightmost_el->l_count)) { - int old_depth = depth; - ret = ocfs2_grow_tree(inode, handle, di_bh, &depth, last_eb_bh, meta_ac); if (ret) { mlog_errno(ret); goto out; } - - if (old_depth != depth) { - eb = (struct ocfs2_extent_block *)(*last_eb_bh)->b_data; - rightmost_el = &eb->h_list; - } } memset(&insert, 0, sizeof(struct ocfs2_insert_type)); insert.ins_appending = APPEND_NONE; insert.ins_contig = CONTIG_NONE; - insert.ins_free_records = le16_to_cpu(rightmost_el->l_count) - - le16_to_cpu(rightmost_el->l_next_free_rec); insert.ins_tree_depth = depth; insert_range = le32_to_cpu(split_rec.e_cpos) + @@ -4019,11 +4000,6 @@ static int __ocfs2_mark_extent_written(struct inode *inode, } else rightmost_el = path_root_el(path); - ctxt.c_used_tail_recs = le16_to_cpu(rightmost_el->l_next_free_rec); - if (ctxt.c_used_tail_recs > 0 && - ocfs2_is_empty_extent(&rightmost_el->l_recs[0])) - ctxt.c_used_tail_recs--; - if (rec->e_cpos == split_rec->e_cpos && rec->e_leaf_clusters == split_rec->e_leaf_clusters) ctxt.c_split_covers_rec = 1; @@ -4032,10 +4008,9 @@ static int __ocfs2_mark_extent_written(struct inode *inode, ctxt.c_has_empty_extent = ocfs2_is_empty_extent(&el->l_recs[0]); - mlog(0, "index: %d, contig: %u, used_tail_recs: %u, " - "has_empty: %u, split_covers: %u\n", split_index, - ctxt.c_contig_type, ctxt.c_used_tail_recs, - ctxt.c_has_empty_extent, ctxt.c_split_covers_rec); + mlog(0, "index: %d, contig: %u, has_empty: %u, split_covers: %u\n", + split_index, ctxt.c_contig_type, ctxt.c_has_empty_extent, + ctxt.c_split_covers_rec); if (ctxt.c_contig_type == CONTIG_NONE) { if (ctxt.c_split_covers_rec) @@ -4184,27 +4159,18 @@ 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)) { - int old_depth = depth; - ret = ocfs2_grow_tree(inode, handle, di_bh, &depth, &last_eb_bh, meta_ac); if (ret) { mlog_errno(ret); goto out; } - - if (old_depth != depth) { - eb = (struct ocfs2_extent_block *)last_eb_bh->b_data; - rightmost_el = &eb->h_list; - } } memset(&insert, 0, sizeof(struct ocfs2_insert_type)); insert.ins_appending = APPEND_NONE; insert.ins_contig = CONTIG_NONE; insert.ins_split = SPLIT_RIGHT; - insert.ins_free_records = le16_to_cpu(rightmost_el->l_count) - - le16_to_cpu(rightmost_el->l_next_free_rec); insert.ins_tree_depth = depth; ret = ocfs2_do_insert_extent(inode, handle, di_bh, &split_rec, &insert); @@ -5606,6 +5572,7 @@ static int ocfs2_do_truncate(struct ocfs2_super *osb, clusters_to_del; spin_unlock(&OCFS2_I(inode)->ip_lock); le32_add_cpu(&fe->i_clusters, -clusters_to_del); + inode->i_blocks = ocfs2_inode_sector_count(inode); status = ocfs2_trim_tree(inode, path, handle, tc, clusters_to_del, &delete_blk); @@ -5668,12 +5635,50 @@ static int ocfs2_ordered_zero_func(handle_t *handle, struct buffer_head *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) +{ + int ret, partial = 0; + + ret = ocfs2_map_page_blocks(page, phys, inode, from, to, 0); + if (ret) + mlog_errno(ret); + + if (zero) + zero_user_page(page, from, to - from, KM_USER0); + + /* + * Need to set the buffers we zero'd into uptodate + * 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_writeback_zero_func); + if (ret < 0) + mlog_errno(ret); + } + + if (!partial) + SetPageUptodate(page); + + flush_dcache_page(page); +} + static void ocfs2_zero_cluster_pages(struct inode *inode, loff_t start, loff_t end, struct page **pages, int numpages, u64 phys, handle_t *handle) { - int i, ret, partial = 0; - void *kaddr; + int i; struct page *page; unsigned int from, to = PAGE_CACHE_SIZE; struct super_block *sb = inode->i_sb; @@ -5694,87 +5699,31 @@ static void ocfs2_zero_cluster_pages(struct inode *inode, loff_t start, BUG_ON(from > PAGE_CACHE_SIZE); BUG_ON(to > PAGE_CACHE_SIZE); - ret = ocfs2_map_page_blocks(page, &phys, inode, from, to, 0); - if (ret) - mlog_errno(ret); - - kaddr = kmap_atomic(page, KM_USER0); - memset(kaddr + from, 0, to - from); - kunmap_atomic(kaddr, KM_USER0); - - /* - * Need to set the buffers we zero'd into uptodate - * 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_writeback_zero_func); - if (ret < 0) - mlog_errno(ret); - } - - if (!partial) - SetPageUptodate(page); - - flush_dcache_page(page); + ocfs2_map_and_dirty_page(inode, handle, from, to, page, 1, + &phys); start = (page->index + 1) << PAGE_CACHE_SHIFT; } out: - if (pages) { - for (i = 0; i < numpages; i++) { - page = pages[i]; - unlock_page(page); - mark_page_accessed(page); - page_cache_release(page); - } - } + if (pages) + ocfs2_unlock_and_free_pages(pages, numpages); } static int ocfs2_grab_eof_pages(struct inode *inode, loff_t start, loff_t end, - struct page **pages, int *num, u64 *phys) + struct page **pages, int *num) { - int i, numpages = 0, ret = 0; - unsigned int ext_flags; + int numpages, ret = 0; struct super_block *sb = inode->i_sb; struct address_space *mapping = inode->i_mapping; unsigned long index; loff_t last_page_bytes; - BUG_ON(!ocfs2_sparse_alloc(OCFS2_SB(sb))); BUG_ON(start > end); - if (start == end) - goto out; - BUG_ON(start >> OCFS2_SB(sb)->s_clustersize_bits != (end - 1) >> OCFS2_SB(sb)->s_clustersize_bits); - ret = ocfs2_extent_map_get_blocks(inode, start >> sb->s_blocksize_bits, - phys, NULL, &ext_flags); - if (ret) { - mlog_errno(ret); - goto out; - } - - /* Tail is a hole. */ - if (*phys == 0) - goto out; - - /* Tail is marked as unwritten, we can count on write to zero - * in that case. */ - if (ext_flags & OCFS2_EXT_UNWRITTEN) - goto out; - + numpages = 0; last_page_bytes = PAGE_ALIGN(end); index = start >> PAGE_CACHE_SHIFT; do { @@ -5791,14 +5740,8 @@ static int ocfs2_grab_eof_pages(struct inode *inode, loff_t start, loff_t end, out: if (ret != 0) { - if (pages) { - for (i = 0; i < numpages; i++) { - if (pages[i]) { - unlock_page(pages[i]); - page_cache_release(pages[i]); - } - } - } + if (pages) + ocfs2_unlock_and_free_pages(pages, numpages); numpages = 0; } @@ -5819,18 +5762,20 @@ out: int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, u64 range_start, u64 range_end) { - int ret, numpages; + int ret = 0, numpages; struct page **pages = NULL; u64 phys; + unsigned int ext_flags; + struct super_block *sb = inode->i_sb; /* * File systems which don't support sparse files zero on every * extend. */ - if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) + if (!ocfs2_sparse_alloc(OCFS2_SB(sb))) return 0; - pages = kcalloc(ocfs2_pages_per_cluster(inode->i_sb), + pages = kcalloc(ocfs2_pages_per_cluster(sb), sizeof(struct page *), GFP_NOFS); if (pages == NULL) { ret = -ENOMEM; @@ -5838,16 +5783,31 @@ int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, goto out; } - ret = ocfs2_grab_eof_pages(inode, range_start, range_end, pages, - &numpages, &phys); + if (range_start == range_end) + goto out; + + ret = ocfs2_extent_map_get_blocks(inode, + range_start >> sb->s_blocksize_bits, + &phys, NULL, &ext_flags); if (ret) { mlog_errno(ret); goto out; } - if (numpages == 0) + /* + * Tail is a hole, or is marked unwritten. In either case, we + * can count on read and write to return/push zero's. + */ + if (phys == 0 || ext_flags & OCFS2_EXT_UNWRITTEN) goto out; + ret = ocfs2_grab_eof_pages(inode, range_start, range_end, pages, + &numpages); + if (ret) { + mlog_errno(ret); + goto out; + } + ocfs2_zero_cluster_pages(inode, range_start, range_end, pages, numpages, phys, handle); @@ -5868,6 +5828,178 @@ out: return ret; } +static void ocfs2_zero_dinode_id2(struct inode *inode, struct ocfs2_dinode *di) +{ + unsigned int blocksize = 1 << inode->i_sb->s_blocksize_bits; + + 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); + 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)); +} + +void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di) +{ + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_inline_data *idata = &di->id2.i_data; + + spin_lock(&oi->ip_lock); + oi->ip_dyn_features |= OCFS2_INLINE_DATA_FL; + di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); + spin_unlock(&oi->ip_lock); + + /* + * We clear the entire i_data structure here so that all + * fields can be properly initialized. + */ + ocfs2_zero_dinode_id2(inode, di); + + idata->id_count = cpu_to_le16(ocfs2_max_inline_data(inode->i_sb)); +} + +int ocfs2_convert_inline_data_to_extents(struct inode *inode, + struct buffer_head *di_bh) +{ + int ret, i, has_data, num_pages = 0; + handle_t *handle; + u64 uninitialized_var(block); + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_alloc_context *data_ac = NULL; + struct page **pages = NULL; + loff_t end = osb->s_clustersize; + + has_data = i_size_read(inode) ? 1 : 0; + + if (has_data) { + pages = kcalloc(ocfs2_pages_per_cluster(osb->sb), + sizeof(struct page *), GFP_NOFS); + if (pages == NULL) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + + ret = ocfs2_reserve_clusters(osb, 1, &data_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + handle = ocfs2_start_trans(osb, OCFS2_INLINE_TO_EXTENTS_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out_unlock; + } + + ret = ocfs2_journal_access(handle, inode, di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + if (has_data) { + u32 bit_off, num; + unsigned int page_end; + u64 phys; + + ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, + &num); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* + * Save two copies, one for insert, and one that can + * be changed by ocfs2_map_and_dirty_page() below. + */ + block = phys = ocfs2_clusters_to_blocks(inode->i_sb, bit_off); + + /* + * Non sparse file systems zero on extend, so no need + * to do that now. + */ + if (!ocfs2_sparse_alloc(osb) && + PAGE_CACHE_SIZE < osb->s_clustersize) + end = PAGE_CACHE_SIZE; + + ret = ocfs2_grab_eof_pages(inode, 0, end, pages, &num_pages); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* + * This should populate the 1st page for us and mark + * it up to date. + */ + ret = ocfs2_read_inline_data(inode, pages[0], di_bh); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + page_end = PAGE_CACHE_SIZE; + if (PAGE_CACHE_SIZE > osb->s_clustersize) + page_end = osb->s_clustersize; + + for (i = 0; i < num_pages; i++) + ocfs2_map_and_dirty_page(inode, handle, 0, page_end, + pages[i], i > 0, &phys); + } + + spin_lock(&oi->ip_lock); + oi->ip_dyn_features &= ~OCFS2_INLINE_DATA_FL; + di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); + spin_unlock(&oi->ip_lock); + + ocfs2_dinode_new_extent_list(inode, di); + + ocfs2_journal_dirty(handle, di_bh); + + if (has_data) { + /* + * An error at this point should be extremely rare. If + * 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, + 0, block, 1, 0, NULL); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + inode->i_blocks = ocfs2_inode_sector_count(inode); + } + +out_commit: + ocfs2_commit_trans(osb, handle); + +out_unlock: + if (data_ac) + ocfs2_free_alloc_context(data_ac); + +out: + if (pages) { + ocfs2_unlock_and_free_pages(pages, num_pages); + kfree(pages); + } + + return ret; +} + /* * It is expected, that by the time you call this function, * inode->i_size and fe->i_size have been adjusted. @@ -6093,6 +6225,81 @@ bail: return status; } +/* + * 'start' is inclusive, 'end' is not. + */ +int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, + unsigned int start, unsigned int end, int trunc) +{ + int ret; + unsigned int numbytes; + handle_t *handle; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_inline_data *idata = &di->id2.i_data; + + if (end > i_size_read(inode)) + end = i_size_read(inode); + + BUG_ON(start >= end); + + if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) || + !(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL) || + !ocfs2_supports_inline_data(osb)) { + ocfs2_error(inode->i_sb, + "Inline data flags for inode %llu don't agree! " + "Disk: 0x%x, Memory: 0x%x, Superblock: 0x%x\n", + (unsigned long long)OCFS2_I(inode)->ip_blkno, + le16_to_cpu(di->i_dyn_features), + OCFS2_I(inode)->ip_dyn_features, + osb->s_feature_incompat); + ret = -EROFS; + goto out; + } + + handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access(handle, inode, di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + numbytes = end - start; + memset(idata->id_data + start, 0, numbytes); + + /* + * No need to worry about the data page here - it's been + * truncated already and inline data doesn't need it for + * pushing zero's to disk, so we'll let readpage pick it up + * later. + */ + if (trunc) { + i_size_write(inode, start); + di->i_size = cpu_to_le64(start); + } + + inode->i_blocks = ocfs2_inode_sector_count(inode); + inode->i_ctime = inode->i_mtime = CURRENT_TIME; + + di->i_ctime = di->i_mtime = cpu_to_le64(inode->i_ctime.tv_sec); + di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); + + ocfs2_journal_dirty(handle, di_bh); + +out_commit: + ocfs2_commit_trans(osb, handle); + +out: + return ret; +} + static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc) { /* diff --git a/fs/ocfs2/alloc.h b/fs/ocfs2/alloc.h index 990df48ae8d..42ff94bd801 100644 --- a/fs/ocfs2/alloc.h +++ b/fs/ocfs2/alloc.h @@ -62,6 +62,11 @@ static inline int ocfs2_extend_meta_needed(struct ocfs2_dinode *fe) return le16_to_cpu(fe->id2.i_list.l_tree_depth) + 2; } +void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di); +void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di); +int ocfs2_convert_inline_data_to_extents(struct inode *inode, + struct buffer_head *di_bh); + int ocfs2_truncate_log_init(struct ocfs2_super *osb); void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb); void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb, @@ -115,6 +120,8 @@ int ocfs2_commit_truncate(struct ocfs2_super *osb, struct inode *inode, struct buffer_head *fe_bh, struct ocfs2_truncate_context *tc); +int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, + unsigned int start, unsigned int end, int trunc); int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el, u32 cpos, struct buffer_head **leaf_bh); diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 84bf6e79de2..34d10452c56 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -206,9 +206,70 @@ bail: return err; } +int ocfs2_read_inline_data(struct inode *inode, struct page *page, + struct buffer_head *di_bh) +{ + void *kaddr; + unsigned int size; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + + if (!(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL)) { + ocfs2_error(inode->i_sb, "Inode %llu lost inline data flag", + (unsigned long long)OCFS2_I(inode)->ip_blkno); + return -EROFS; + } + + size = i_size_read(inode); + + if (size > PAGE_CACHE_SIZE || + size > ocfs2_max_inline_data(inode->i_sb)) { + ocfs2_error(inode->i_sb, + "Inode %llu has with inline data has bad size: %u", + (unsigned long long)OCFS2_I(inode)->ip_blkno, size); + return -EROFS; + } + + kaddr = kmap_atomic(page, KM_USER0); + if (size) + memcpy(kaddr, di->id2.i_data.id_data, size); + /* Clear the remaining part of the page */ + memset(kaddr + size, 0, PAGE_CACHE_SIZE - size); + flush_dcache_page(page); + kunmap_atomic(kaddr, KM_USER0); + + SetPageUptodate(page); + + return 0; +} + +static int ocfs2_readpage_inline(struct inode *inode, struct page *page) +{ + int ret; + struct buffer_head *di_bh = NULL; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + + BUG_ON(!PageLocked(page)); + BUG_ON(!OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL); + + ret = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, &di_bh, + OCFS2_BH_CACHED, inode); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_read_inline_data(inode, page, di_bh); +out: + unlock_page(page); + + brelse(di_bh); + return ret; +} + static int ocfs2_readpage(struct file *file, struct page *page) { struct inode *inode = page->mapping->host; + struct ocfs2_inode_info *oi = OCFS2_I(inode); loff_t start = (loff_t)page->index << PAGE_CACHE_SHIFT; int ret, unlock = 1; @@ -222,7 +283,7 @@ static int ocfs2_readpage(struct file *file, struct page *page) goto out; } - if (down_read_trylock(&OCFS2_I(inode)->ip_alloc_sem) == 0) { + if (down_read_trylock(&oi->ip_alloc_sem) == 0) { ret = AOP_TRUNCATED_PAGE; goto out_meta_unlock; } @@ -232,7 +293,7 @@ static int ocfs2_readpage(struct file *file, struct page *page) * might now be discovering a truncate that hit on another node. * block_read_full_page->get_block freaks out if it is asked to read * beyond the end of a file, so we check here. Callers - * (generic_file_read, fault->nopage) are clever enough to check i_size + * (generic_file_read, vm_ops->fault) are clever enough to check i_size * and notice that the page they just read isn't needed. * * XXX sys_readahead() seems to get that wrong? @@ -252,7 +313,10 @@ static int ocfs2_readpage(struct file *file, struct page *page) goto out_alloc; } - ret = block_read_full_page(page, ocfs2_get_block); + if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) + ret = ocfs2_readpage_inline(inode, page); + else + ret = block_read_full_page(page, ocfs2_get_block); unlock = 0; ocfs2_data_unlock(inode, 0); @@ -301,12 +365,8 @@ int ocfs2_prepare_write_nolock(struct inode *inode, struct page *page, { int ret; - down_read(&OCFS2_I(inode)->ip_alloc_sem); - ret = block_prepare_write(page, from, to, ocfs2_get_block); - up_read(&OCFS2_I(inode)->ip_alloc_sem); - return ret; } @@ -401,7 +461,9 @@ static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block) down_read(&OCFS2_I(inode)->ip_alloc_sem); } - err = ocfs2_extent_map_get_blocks(inode, block, &p_blkno, NULL, NULL); + if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)) + err = ocfs2_extent_map_get_blocks(inode, block, &p_blkno, NULL, + NULL); if (!INODE_JOURNAL(inode)) { up_read(&OCFS2_I(inode)->ip_alloc_sem); @@ -415,7 +477,6 @@ static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block) goto bail; } - bail: status = err ? 0 : p_blkno; @@ -570,6 +631,13 @@ static ssize_t ocfs2_direct_IO(int rw, mlog_entry_void(); + /* + * Fallback to buffered I/O if we see an inode without + * extents. + */ + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) + return 0; + if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) { /* * We get PR data locks even for O_DIRECT. This @@ -834,18 +902,22 @@ struct ocfs2_write_ctxt { struct ocfs2_cached_dealloc_ctxt w_dealloc; }; -static void ocfs2_free_write_ctxt(struct ocfs2_write_ctxt *wc) +void ocfs2_unlock_and_free_pages(struct page **pages, int num_pages) { int i; - for(i = 0; i < wc->w_num_pages; i++) { - if (wc->w_pages[i] == NULL) - continue; - - unlock_page(wc->w_pages[i]); - mark_page_accessed(wc->w_pages[i]); - page_cache_release(wc->w_pages[i]); + for(i = 0; i < num_pages; i++) { + if (pages[i]) { + unlock_page(pages[i]); + mark_page_accessed(pages[i]); + page_cache_release(pages[i]); + } } +} + +static void ocfs2_free_write_ctxt(struct ocfs2_write_ctxt *wc) +{ + ocfs2_unlock_and_free_pages(wc->w_pages, wc->w_num_pages); brelse(wc->w_di_bh); kfree(wc); @@ -855,6 +927,7 @@ static int ocfs2_alloc_write_ctxt(struct ocfs2_write_ctxt **wcp, struct ocfs2_super *osb, loff_t pos, unsigned len, struct buffer_head *di_bh) { + u32 cend; struct ocfs2_write_ctxt *wc; wc = kzalloc(sizeof(struct ocfs2_write_ctxt), GFP_NOFS); @@ -862,7 +935,8 @@ static int ocfs2_alloc_write_ctxt(struct ocfs2_write_ctxt **wcp, return -ENOMEM; wc->w_cpos = pos >> osb->s_clustersize_bits; - wc->w_clen = ocfs2_clusters_for_bytes(osb->sb, len); + cend = (pos + len - 1) >> osb->s_clustersize_bits; + wc->w_clen = cend - wc->w_cpos + 1; get_bh(di_bh); wc->w_di_bh = di_bh; @@ -928,18 +1002,11 @@ static void ocfs2_write_failure(struct inode *inode, loff_t user_pos, unsigned user_len) { int i; - unsigned from, to; + unsigned from = user_pos & (PAGE_CACHE_SIZE - 1), + to = user_pos + user_len; struct page *tmppage; - ocfs2_zero_new_buffers(wc->w_target_page, user_pos, user_len); - - if (wc->w_large_pages) { - from = wc->w_target_from; - to = wc->w_target_to; - } else { - from = 0; - to = PAGE_CACHE_SIZE; - } + ocfs2_zero_new_buffers(wc->w_target_page, from, to); for(i = 0; i < wc->w_num_pages; i++) { tmppage = wc->w_pages[i]; @@ -989,9 +1056,6 @@ static int ocfs2_prepare_page_for_write(struct inode *inode, u64 *p_blkno, map_from = cluster_start; map_to = cluster_end; } - - wc->w_target_from = map_from; - wc->w_target_to = map_to; } else { /* * If we haven't allocated the new page yet, we @@ -1209,18 +1273,33 @@ static int ocfs2_write_cluster_by_desc(struct address_space *mapping, loff_t pos, unsigned len) { int ret, i; + loff_t cluster_off; + unsigned int local_len = len; struct ocfs2_write_cluster_desc *desc; + struct ocfs2_super *osb = OCFS2_SB(mapping->host->i_sb); for (i = 0; i < wc->w_clen; i++) { desc = &wc->w_desc[i]; + /* + * We have to make sure that the total write passed in + * doesn't extend past a single cluster. + */ + local_len = len; + cluster_off = pos & (osb->s_clustersize - 1); + if ((cluster_off + local_len) > osb->s_clustersize) + local_len = osb->s_clustersize - cluster_off; + ret = ocfs2_write_cluster(mapping, desc->c_phys, desc->c_unwritten, data_ac, meta_ac, - wc, desc->c_cpos, pos, len); + wc, desc->c_cpos, pos, local_len); if (ret) { mlog_errno(ret); goto out; } + + len -= local_len; + pos += local_len; } ret = 0; @@ -1353,6 +1432,160 @@ out: return ret; } +static int ocfs2_write_begin_inline(struct address_space *mapping, + struct inode *inode, + struct ocfs2_write_ctxt *wc) +{ + int ret; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct page *page; + handle_t *handle; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)wc->w_di_bh->b_data; + + page = find_or_create_page(mapping, 0, GFP_NOFS); + if (!page) { + ret = -ENOMEM; + mlog_errno(ret); + goto out; + } + /* + * If we don't set w_num_pages then this page won't get unlocked + * and freed on cleanup of the write context. + */ + wc->w_pages[0] = wc->w_target_page = page; + wc->w_num_pages = 1; + + handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_access(handle, inode, wc->w_di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + ocfs2_commit_trans(osb, handle); + + mlog_errno(ret); + goto out; + } + + if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)) + ocfs2_set_inode_data_inline(inode, di); + + if (!PageUptodate(page)) { + ret = ocfs2_read_inline_data(inode, page, wc->w_di_bh); + if (ret) { + ocfs2_commit_trans(osb, handle); + + goto out; + } + } + + wc->w_handle = handle; +out: + return ret; +} + +int ocfs2_size_fits_inline_data(struct buffer_head *di_bh, u64 new_size) +{ + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + + if (new_size < le16_to_cpu(di->id2.i_data.id_count)) + return 1; + return 0; +} + +static int ocfs2_try_to_write_inline_data(struct address_space *mapping, + struct inode *inode, loff_t pos, + unsigned len, struct page *mmap_page, + struct ocfs2_write_ctxt *wc) +{ + int ret, written = 0; + loff_t end = pos + len; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + + mlog(0, "Inode %llu, write of %u bytes at off %llu. features: 0x%x\n", + (unsigned long long)oi->ip_blkno, len, (unsigned long long)pos, + oi->ip_dyn_features); + + /* + * Handle inodes which already have inline data 1st. + */ + if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + if (mmap_page == NULL && + ocfs2_size_fits_inline_data(wc->w_di_bh, end)) + goto do_inline_write; + + /* + * The write won't fit - we have to give this inode an + * inline extent list now. + */ + ret = ocfs2_convert_inline_data_to_extents(inode, wc->w_di_bh); + if (ret) + mlog_errno(ret); + goto out; + } + + /* + * Check whether the inode can accept inline data. + */ + if (oi->ip_clusters != 0 || i_size_read(inode) != 0) + return 0; + + /* + * Check whether the write can fit. + */ + if (mmap_page || end > ocfs2_max_inline_data(inode->i_sb)) + return 0; + +do_inline_write: + ret = ocfs2_write_begin_inline(mapping, inode, wc); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * This signals to the caller that the data can be written + * inline. + */ + written = 1; +out: + return written ? written : ret; +} + +/* + * This function only does anything for file systems which can't + * handle sparse files. + * + * What we want to do here is fill in any hole between the current end + * of allocation and the end of our write. That way the rest of the + * write path can treat it as an non-allocating write, which has no + * special case code for sparse/nonsparse files. + */ +static int ocfs2_expand_nonsparse_inode(struct inode *inode, loff_t pos, + unsigned len, + struct ocfs2_write_ctxt *wc) +{ + int ret; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + loff_t newsize = pos + len; + + if (ocfs2_sparse_alloc(osb)) + return 0; + + if (newsize <= i_size_read(inode)) + return 0; + + ret = ocfs2_extend_no_holes(inode, newsize, newsize - len); + if (ret) + mlog_errno(ret); + + return ret; +} + int ocfs2_write_begin_nolock(struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata, @@ -1374,6 +1607,25 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, return ret; } + if (ocfs2_supports_inline_data(osb)) { + ret = ocfs2_try_to_write_inline_data(mapping, inode, pos, len, + mmap_page, wc); + if (ret == 1) { + ret = 0; + goto success; + } + if (ret < 0) { + mlog_errno(ret); + goto out; + } + } + + ret = ocfs2_expand_nonsparse_inode(inode, pos, len, wc); + if (ret) { + mlog_errno(ret); + goto out; + } + ret = ocfs2_populate_write_desc(inode, wc, &clusters_to_alloc, &extents_to_split); if (ret) { @@ -1455,6 +1707,7 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, if (meta_ac) ocfs2_free_alloc_context(meta_ac); +success: *pagep = wc->w_target_page; *fsdata = wc; return 0; @@ -1522,6 +1775,31 @@ out_fail: return ret; } +static void ocfs2_write_end_inline(struct inode *inode, loff_t pos, + unsigned len, unsigned *copied, + struct ocfs2_dinode *di, + struct ocfs2_write_ctxt *wc) +{ + void *kaddr; + + if (unlikely(*copied < len)) { + if (!PageUptodate(wc->w_target_page)) { + *copied = 0; + return; + } + } + + kaddr = kmap_atomic(wc->w_target_page, KM_USER0); + memcpy(di->id2.i_data.id_data + pos, kaddr + pos, *copied); + kunmap_atomic(kaddr, KM_USER0); + + mlog(0, "Data written to inode at offset %llu. " + "id_count = %u, copied = %u, i_dyn_features = 0x%x\n", + (unsigned long long)pos, *copied, + le16_to_cpu(di->id2.i_data.id_count), + le16_to_cpu(di->i_dyn_features)); +} + int ocfs2_write_end_nolock(struct address_space *mapping, loff_t pos, unsigned len, unsigned copied, struct page *page, void *fsdata) @@ -1535,6 +1813,11 @@ int ocfs2_write_end_nolock(struct address_space *mapping, handle_t *handle = wc->w_handle; struct page *tmppage; + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + ocfs2_write_end_inline(inode, pos, len, &copied, di, wc); + goto out_write_size; + } + if (unlikely(copied < len)) { if (!PageUptodate(wc->w_target_page)) copied = 0; @@ -1572,6 +1855,7 @@ int ocfs2_write_end_nolock(struct address_space *mapping, block_commit_write(tmppage, from, to); } +out_write_size: pos += copied; if (pos > inode->i_size) { i_size_write(inode, pos); diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h index 389579bd64e..113560877db 100644 --- a/fs/ocfs2/aops.h +++ b/fs/ocfs2/aops.h @@ -34,6 +34,8 @@ int ocfs2_map_page_blocks(struct page *page, u64 *p_blkno, struct inode *inode, unsigned int from, unsigned int to, int new); +void ocfs2_unlock_and_free_pages(struct page **pages, int num_pages); + int walk_page_buffers( handle_t *handle, struct buffer_head *head, unsigned from, @@ -59,6 +61,10 @@ int ocfs2_write_begin_nolock(struct address_space *mapping, struct page **pagep, void **fsdata, struct buffer_head *di_bh, struct page *mmap_page); +int ocfs2_read_inline_data(struct inode *inode, struct page *page, + struct buffer_head *di_bh); +int ocfs2_size_fits_inline_data(struct buffer_head *di_bh, u64 new_size); + /* all ocfs2_dio_end_io()'s fault */ #define ocfs2_iocb_is_rw_locked(iocb) \ test_bit(0, (unsigned long *)&iocb->private) diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c index 2bd7f788cf3..f14b541fab9 100644 --- a/fs/ocfs2/cluster/heartbeat.c +++ b/fs/ocfs2/cluster/heartbeat.c @@ -216,8 +216,7 @@ static void o2hb_wait_on_io(struct o2hb_region *reg, wait_for_completion(&wc->wc_io_complete); } -static int o2hb_bio_end_io(struct bio *bio, - unsigned int bytes_done, +static void o2hb_bio_end_io(struct bio *bio, int error) { struct o2hb_bio_wait_ctxt *wc = bio->bi_private; @@ -227,12 +226,8 @@ static int o2hb_bio_end_io(struct bio *bio, wc->wc_error = error; } - if (bio->bi_size) - return 1; - o2hb_bio_wait_dec(wc, 1); bio_put(bio); - return 0; } /* Setup a Bio to cover I/O against num_slots slots starting at diff --git a/fs/ocfs2/cluster/masklog.c b/fs/ocfs2/cluster/masklog.c index e9e042b93db..a4882c8df94 100644 --- a/fs/ocfs2/cluster/masklog.c +++ b/fs/ocfs2/cluster/masklog.c @@ -143,7 +143,7 @@ static struct kobj_type mlog_ktype = { }; static struct kset mlog_kset = { - .kobj = {.name = "logmask", .ktype = &mlog_ktype}, + .kobj = {.ktype = &mlog_ktype}, }; int mlog_sys_init(struct kset *o2cb_subsys) @@ -156,6 +156,7 @@ int mlog_sys_init(struct kset *o2cb_subsys) } mlog_attr_ptrs[i] = NULL; + kobject_set_name(&mlog_kset.kobj, "logmask"); kobj_set_kset_s(&mlog_kset, *o2cb_subsys); return kset_register(&mlog_kset); } diff --git a/fs/ocfs2/cluster/tcp.c b/fs/ocfs2/cluster/tcp.c index f0bdfd944c4..685c18065c8 100644 --- a/fs/ocfs2/cluster/tcp.c +++ b/fs/ocfs2/cluster/tcp.c @@ -854,17 +854,25 @@ static void o2net_sendpage(struct o2net_sock_container *sc, struct o2net_node *nn = o2net_nn_from_num(sc->sc_node->nd_num); ssize_t ret; - - mutex_lock(&sc->sc_send_lock); - ret = sc->sc_sock->ops->sendpage(sc->sc_sock, - virt_to_page(kmalloced_virt), - (long)kmalloced_virt & ~PAGE_MASK, - size, MSG_DONTWAIT); - mutex_unlock(&sc->sc_send_lock); - if (ret != size) { + while (1) { + mutex_lock(&sc->sc_send_lock); + ret = sc->sc_sock->ops->sendpage(sc->sc_sock, + virt_to_page(kmalloced_virt), + (long)kmalloced_virt & ~PAGE_MASK, + size, MSG_DONTWAIT); + mutex_unlock(&sc->sc_send_lock); + if (ret == size) + break; + if (ret == (ssize_t)-EAGAIN) { + mlog(0, "sendpage of size %zu to " SC_NODEF_FMT + " returned EAGAIN\n", size, SC_NODEF_ARGS(sc)); + cond_resched(); + continue; + } mlog(ML_ERROR, "sendpage of size %zu to " SC_NODEF_FMT " failed with %zd\n", size, SC_NODEF_ARGS(sc), ret); o2net_ensure_shutdown(nn, sc, 0); + break; } } diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c index 0d5fdde959c..7453b70c1a1 100644 --- a/fs/ocfs2/dir.c +++ b/fs/ocfs2/dir.c @@ -55,10 +55,16 @@ #include "journal.h" #include "namei.h" #include "suballoc.h" +#include "super.h" #include "uptodate.h" #include "buffer_head_io.h" +#define NAMEI_RA_CHUNKS 2 +#define NAMEI_RA_BLOCKS 4 +#define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) +#define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) + static unsigned char ocfs2_filetype_table[] = { DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK }; @@ -66,12 +72,614 @@ static unsigned char ocfs2_filetype_table[] = { static int ocfs2_extend_dir(struct ocfs2_super *osb, struct inode *dir, struct buffer_head *parent_fe_bh, + unsigned int blocks_wanted, struct buffer_head **new_de_bh); +static int ocfs2_do_extend_dir(struct super_block *sb, + handle_t *handle, + struct inode *dir, + struct buffer_head *parent_fe_bh, + struct ocfs2_alloc_context *data_ac, + struct ocfs2_alloc_context *meta_ac, + struct buffer_head **new_bh); + /* - * ocfs2_readdir() + * bh passed here can be an inode block or a dir data block, depending + * on the inode inline data flag. + */ +static int ocfs2_check_dir_entry(struct inode * dir, + struct ocfs2_dir_entry * de, + struct buffer_head * bh, + unsigned long offset) +{ + const char *error_msg = NULL; + const int rlen = le16_to_cpu(de->rec_len); + + if (rlen < OCFS2_DIR_REC_LEN(1)) + error_msg = "rec_len is smaller than minimal"; + else if (rlen % 4 != 0) + error_msg = "rec_len % 4 != 0"; + else if (rlen < OCFS2_DIR_REC_LEN(de->name_len)) + error_msg = "rec_len is too small for name_len"; + else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize) + error_msg = "directory entry across blocks"; + + if (error_msg != NULL) + mlog(ML_ERROR, "bad entry in directory #%llu: %s - " + "offset=%lu, inode=%llu, rec_len=%d, name_len=%d\n", + (unsigned long long)OCFS2_I(dir)->ip_blkno, error_msg, + offset, (unsigned long long)le64_to_cpu(de->inode), rlen, + de->name_len); + return error_msg == NULL ? 1 : 0; +} + +static inline int ocfs2_match(int len, + const char * const name, + struct ocfs2_dir_entry *de) +{ + if (len != de->name_len) + return 0; + if (!de->inode) + return 0; + return !memcmp(name, de->name, len); +} + +/* + * Returns 0 if not found, -1 on failure, and 1 on success + */ +static int inline ocfs2_search_dirblock(struct buffer_head *bh, + struct inode *dir, + const char *name, int namelen, + unsigned long offset, + char *first_de, + unsigned int bytes, + struct ocfs2_dir_entry **res_dir) +{ + struct ocfs2_dir_entry *de; + char *dlimit, *de_buf; + int de_len; + int ret = 0; + + mlog_entry_void(); + + de_buf = first_de; + dlimit = de_buf + bytes; + + while (de_buf < dlimit) { + /* this code is executed quadratically often */ + /* do minimal checking `by hand' */ + + de = (struct ocfs2_dir_entry *) de_buf; + + if (de_buf + namelen <= dlimit && + ocfs2_match(namelen, name, de)) { + /* found a match - just to be sure, do a full check */ + if (!ocfs2_check_dir_entry(dir, de, bh, offset)) { + ret = -1; + goto bail; + } + *res_dir = de; + ret = 1; + goto bail; + } + + /* prevent looping on a bad block */ + de_len = le16_to_cpu(de->rec_len); + if (de_len <= 0) { + ret = -1; + goto bail; + } + + de_buf += de_len; + offset += de_len; + } + +bail: + mlog_exit(ret); + return ret; +} + +static struct buffer_head *ocfs2_find_entry_id(const char *name, + int namelen, + struct inode *dir, + struct ocfs2_dir_entry **res_dir) +{ + int ret, found; + struct buffer_head *di_bh = NULL; + struct ocfs2_dinode *di; + struct ocfs2_inline_data *data; + + ret = ocfs2_read_block(OCFS2_SB(dir->i_sb), OCFS2_I(dir)->ip_blkno, + &di_bh, OCFS2_BH_CACHED, dir); + if (ret) { + mlog_errno(ret); + goto out; + } + + di = (struct ocfs2_dinode *)di_bh->b_data; + data = &di->id2.i_data; + + found = ocfs2_search_dirblock(di_bh, dir, name, namelen, 0, + data->id_data, i_size_read(dir), res_dir); + if (found == 1) + return di_bh; + + brelse(di_bh); +out: + return NULL; +} + +struct buffer_head *ocfs2_find_entry_el(const char *name, int namelen, + struct inode *dir, + struct ocfs2_dir_entry **res_dir) +{ + struct super_block *sb; + struct buffer_head *bh_use[NAMEI_RA_SIZE]; + struct buffer_head *bh, *ret = NULL; + unsigned long start, block, b; + int ra_max = 0; /* Number of bh's in the readahead + buffer, bh_use[] */ + int ra_ptr = 0; /* Current index into readahead + buffer */ + int num = 0; + int nblocks, i, err; + + mlog_entry_void(); + + sb = dir->i_sb; + + nblocks = i_size_read(dir) >> sb->s_blocksize_bits; + start = OCFS2_I(dir)->ip_dir_start_lookup; + if (start >= nblocks) + start = 0; + block = start; + +restart: + do { + /* + * We deal with the read-ahead logic here. + */ + if (ra_ptr >= ra_max) { + /* Refill the readahead buffer */ + ra_ptr = 0; + b = block; + for (ra_max = 0; ra_max < NAMEI_RA_SIZE; ra_max++) { + /* + * Terminate if we reach the end of the + * directory and must wrap, or if our + * search has finished at this block. + */ + if (b >= nblocks || (num && block == start)) { + bh_use[ra_max] = NULL; + break; + } + num++; + + bh = ocfs2_bread(dir, b++, &err, 1); + bh_use[ra_max] = bh; + } + } + if ((bh = bh_use[ra_ptr++]) == NULL) + goto next; + wait_on_buffer(bh); + if (!buffer_uptodate(bh)) { + /* read error, skip block & hope for the best */ + ocfs2_error(dir->i_sb, "reading directory %llu, " + "offset %lu\n", + (unsigned long long)OCFS2_I(dir)->ip_blkno, + block); + brelse(bh); + goto next; + } + i = ocfs2_search_dirblock(bh, dir, name, namelen, + block << sb->s_blocksize_bits, + bh->b_data, sb->s_blocksize, + res_dir); + if (i == 1) { + OCFS2_I(dir)->ip_dir_start_lookup = block; + ret = bh; + goto cleanup_and_exit; + } else { + brelse(bh); + if (i < 0) + goto cleanup_and_exit; + } + next: + if (++block >= nblocks) + block = 0; + } while (block != start); + + /* + * If the directory has grown while we were searching, then + * search the last part of the directory before giving up. + */ + block = nblocks; + nblocks = i_size_read(dir) >> sb->s_blocksize_bits; + if (block < nblocks) { + start = 0; + goto restart; + } + +cleanup_and_exit: + /* Clean up the read-ahead blocks */ + for (; ra_ptr < ra_max; ra_ptr++) + brelse(bh_use[ra_ptr]); + + mlog_exit_ptr(ret); + return ret; +} + +/* + * Try to find an entry of the provided name within 'dir'. * + * If nothing was found, NULL is returned. Otherwise, a buffer_head + * and pointer to the dir entry are passed back. + * + * Caller can NOT assume anything about the contents of the + * buffer_head - it is passed back only so that it can be passed into + * any one of the manipulation functions (add entry, delete entry, + * etc). As an example, bh in the extent directory case is a data + * block, in the inline-data case it actually points to an inode. */ -int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir) +struct buffer_head *ocfs2_find_entry(const char *name, int namelen, + struct inode *dir, + struct ocfs2_dir_entry **res_dir) +{ + *res_dir = NULL; + + if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) + return ocfs2_find_entry_id(name, namelen, dir, res_dir); + + return ocfs2_find_entry_el(name, namelen, dir, res_dir); +} + +/* + * Update inode number and type of a previously found directory entry. + */ +int ocfs2_update_entry(struct inode *dir, handle_t *handle, + struct buffer_head *de_bh, struct ocfs2_dir_entry *de, + struct inode *new_entry_inode) +{ + int ret; + + /* + * The same code works fine for both inline-data and extent + * based directories, so no need to split this up. + */ + + ret = ocfs2_journal_access(handle, dir, de_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; + } + + de->inode = cpu_to_le64(OCFS2_I(new_entry_inode)->ip_blkno); + ocfs2_set_de_type(de, new_entry_inode->i_mode); + + ocfs2_journal_dirty(handle, de_bh); + +out: + return ret; +} + +static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir, + struct ocfs2_dir_entry *de_del, + struct buffer_head *bh, char *first_de, + unsigned int bytes) +{ + struct ocfs2_dir_entry *de, *pde; + int i, status = -ENOENT; + + mlog_entry("(0x%p, 0x%p, 0x%p, 0x%p)\n", handle, dir, de_del, bh); + + i = 0; + pde = NULL; + de = (struct ocfs2_dir_entry *) first_de; + while (i < bytes) { + if (!ocfs2_check_dir_entry(dir, de, bh, i)) { + status = -EIO; + mlog_errno(status); + goto bail; + } + if (de == de_del) { + status = ocfs2_journal_access(handle, dir, bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (status < 0) { + status = -EIO; + mlog_errno(status); + goto bail; + } + if (pde) + pde->rec_len = + cpu_to_le16(le16_to_cpu(pde->rec_len) + + le16_to_cpu(de->rec_len)); + else + de->inode = 0; + dir->i_version++; + status = ocfs2_journal_dirty(handle, bh); + goto bail; + } + i += le16_to_cpu(de->rec_len); + pde = de; + de = (struct ocfs2_dir_entry *)((char *)de + le16_to_cpu(de->rec_len)); + } +bail: + mlog_exit(status); + return status; +} + +static inline int ocfs2_delete_entry_id(handle_t *handle, + struct inode *dir, + struct ocfs2_dir_entry *de_del, + struct buffer_head *bh) +{ + int ret; + struct buffer_head *di_bh = NULL; + struct ocfs2_dinode *di; + struct ocfs2_inline_data *data; + + ret = ocfs2_read_block(OCFS2_SB(dir->i_sb), OCFS2_I(dir)->ip_blkno, + &di_bh, OCFS2_BH_CACHED, dir); + if (ret) { + mlog_errno(ret); + goto out; + } + + di = (struct ocfs2_dinode *)di_bh->b_data; + data = &di->id2.i_data; + + ret = __ocfs2_delete_entry(handle, dir, de_del, bh, data->id_data, + i_size_read(dir)); + + brelse(di_bh); +out: + return ret; +} + +static inline int ocfs2_delete_entry_el(handle_t *handle, + struct inode *dir, + struct ocfs2_dir_entry *de_del, + struct buffer_head *bh) +{ + return __ocfs2_delete_entry(handle, dir, de_del, bh, bh->b_data, + bh->b_size); +} + +/* + * ocfs2_delete_entry deletes a directory entry by merging it with the + * previous entry + */ +int ocfs2_delete_entry(handle_t *handle, + struct inode *dir, + struct ocfs2_dir_entry *de_del, + struct buffer_head *bh) +{ + if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) + return ocfs2_delete_entry_id(handle, dir, de_del, bh); + + return ocfs2_delete_entry_el(handle, dir, de_del, bh); +} + +/* + * Check whether 'de' has enough room to hold an entry of + * 'new_rec_len' bytes. + */ +static inline int ocfs2_dirent_would_fit(struct ocfs2_dir_entry *de, + unsigned int new_rec_len) +{ + unsigned int de_really_used; + + /* Check whether this is an empty record with enough space */ + if (le64_to_cpu(de->inode) == 0 && + le16_to_cpu(de->rec_len) >= new_rec_len) + return 1; + + /* + * Record might have free space at the end which we can + * use. + */ + de_really_used = OCFS2_DIR_REC_LEN(de->name_len); + if (le16_to_cpu(de->rec_len) >= (de_really_used + new_rec_len)) + return 1; + + return 0; +} + +/* we don't always have a dentry for what we want to add, so people + * like orphan dir can call this instead. + * + * If you pass me insert_bh, I'll skip the search of the other dir + * blocks and put the record in there. + */ +int __ocfs2_add_entry(handle_t *handle, + struct inode *dir, + const char *name, int namelen, + struct inode *inode, u64 blkno, + struct buffer_head *parent_fe_bh, + struct buffer_head *insert_bh) +{ + unsigned long offset; + unsigned short rec_len; + struct ocfs2_dir_entry *de, *de1; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)parent_fe_bh->b_data; + struct super_block *sb = dir->i_sb; + int retval, status; + unsigned int size = sb->s_blocksize; + char *data_start = insert_bh->b_data; + + mlog_entry_void(); + + if (!namelen) + return -EINVAL; + + if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + data_start = di->id2.i_data.id_data; + size = i_size_read(dir); + + BUG_ON(insert_bh != parent_fe_bh); + } + + rec_len = OCFS2_DIR_REC_LEN(namelen); + offset = 0; + de = (struct ocfs2_dir_entry *) data_start; + while (1) { + BUG_ON((char *)de >= (size + data_start)); + + /* These checks should've already been passed by the + * prepare function, but I guess we can leave them + * here anyway. */ + if (!ocfs2_check_dir_entry(dir, de, insert_bh, offset)) { + retval = -ENOENT; + goto bail; + } + if (ocfs2_match(namelen, name, de)) { + retval = -EEXIST; + goto bail; + } + + if (ocfs2_dirent_would_fit(de, rec_len)) { + dir->i_mtime = dir->i_ctime = CURRENT_TIME; + retval = ocfs2_mark_inode_dirty(handle, dir, parent_fe_bh); + if (retval < 0) { + mlog_errno(retval); + goto bail; + } + + status = ocfs2_journal_access(handle, dir, insert_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + /* By now the buffer is marked for journaling */ + offset += le16_to_cpu(de->rec_len); + if (le64_to_cpu(de->inode)) { + de1 = (struct ocfs2_dir_entry *)((char *) de + + OCFS2_DIR_REC_LEN(de->name_len)); + de1->rec_len = + cpu_to_le16(le16_to_cpu(de->rec_len) - + OCFS2_DIR_REC_LEN(de->name_len)); + de->rec_len = cpu_to_le16(OCFS2_DIR_REC_LEN(de->name_len)); + de = de1; + } + de->file_type = OCFS2_FT_UNKNOWN; + if (blkno) { + de->inode = cpu_to_le64(blkno); + ocfs2_set_de_type(de, inode->i_mode); + } else + de->inode = 0; + de->name_len = namelen; + memcpy(de->name, name, namelen); + + dir->i_version++; + status = ocfs2_journal_dirty(handle, insert_bh); + retval = 0; + goto bail; + } + offset += le16_to_cpu(de->rec_len); + de = (struct ocfs2_dir_entry *) ((char *) de + le16_to_cpu(de->rec_len)); + } + + /* when you think about it, the assert above should prevent us + * from ever getting here. */ + retval = -ENOSPC; +bail: + + mlog_exit(retval); + return retval; +} + +static int ocfs2_dir_foreach_blk_id(struct inode *inode, + unsigned long *f_version, + loff_t *f_pos, void *priv, + filldir_t filldir, int *filldir_err) +{ + int ret, i, filldir_ret; + unsigned long offset = *f_pos; + struct buffer_head *di_bh = NULL; + struct ocfs2_dinode *di; + struct ocfs2_inline_data *data; + struct ocfs2_dir_entry *de; + + ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), OCFS2_I(inode)->ip_blkno, + &di_bh, OCFS2_BH_CACHED, inode); + if (ret) { + mlog(ML_ERROR, "Unable to read inode block for dir %llu\n", + (unsigned long long)OCFS2_I(inode)->ip_blkno); + goto out; + } + + di = (struct ocfs2_dinode *)di_bh->b_data; + data = &di->id2.i_data; + + while (*f_pos < i_size_read(inode)) { +revalidate: + /* If the dir block has changed since the last call to + * readdir(2), then we might be pointing to an invalid + * dirent right now. Scan from the start of the block + * to make sure. */ + if (*f_version != inode->i_version) { + for (i = 0; i < i_size_read(inode) && i < offset; ) { + de = (struct ocfs2_dir_entry *) + (data->id_data + i); + /* It's too expensive to do a full + * dirent test each time round this + * loop, but we do have to test at + * least that it is non-zero. A + * failure will be detected in the + * dirent test below. */ + if (le16_to_cpu(de->rec_len) < + OCFS2_DIR_REC_LEN(1)) + break; + i += le16_to_cpu(de->rec_len); + } + *f_pos = offset = i; + *f_version = inode->i_version; + } + + de = (struct ocfs2_dir_entry *) (data->id_data + *f_pos); + if (!ocfs2_check_dir_entry(inode, de, di_bh, *f_pos)) { + /* On error, skip the f_pos to the end. */ + *f_pos = i_size_read(inode); + goto out; + } + offset += le16_to_cpu(de->rec_len); + if (le64_to_cpu(de->inode)) { + /* We might block in the next section + * if the data destination is + * currently swapped out. So, use a + * version stamp to detect whether or + * not the directory has been modified + * during the copy operation. + */ + unsigned long version = *f_version; + unsigned char d_type = DT_UNKNOWN; + + if (de->file_type < OCFS2_FT_MAX) + d_type = ocfs2_filetype_table[de->file_type]; + + filldir_ret = filldir(priv, de->name, + de->name_len, + *f_pos, + le64_to_cpu(de->inode), + d_type); + if (filldir_ret) { + if (filldir_err) + *filldir_err = filldir_ret; + break; + } + if (version != *f_version) + goto revalidate; + } + *f_pos += le16_to_cpu(de->rec_len); + } + +out: + brelse(di_bh); + + return 0; +} + +static int ocfs2_dir_foreach_blk_el(struct inode *inode, + unsigned long *f_version, + loff_t *f_pos, void *priv, + filldir_t filldir, int *filldir_err) { int error = 0; unsigned long offset, blk, last_ra_blk = 0; @@ -79,45 +687,23 @@ int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir) struct buffer_head * bh, * tmp; struct ocfs2_dir_entry * de; int err; - struct inode *inode = filp->f_path.dentry->d_inode; struct super_block * sb = inode->i_sb; unsigned int ra_sectors = 16; - int lock_level = 0; - - mlog_entry("dirino=%llu\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno); stored = 0; bh = NULL; - error = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level); - if (lock_level && error >= 0) { - /* We release EX lock which used to update atime - * and get PR lock again to reduce contention - * on commonly accessed directories. */ - ocfs2_meta_unlock(inode, 1); - lock_level = 0; - error = ocfs2_meta_lock(inode, NULL, 0); - } - if (error < 0) { - if (error != -ENOENT) - mlog_errno(error); - /* we haven't got any yet, so propagate the error. */ - stored = error; - goto bail_nolock; - } + offset = (*f_pos) & (sb->s_blocksize - 1); - offset = filp->f_pos & (sb->s_blocksize - 1); - - while (!error && !stored && filp->f_pos < i_size_read(inode)) { - blk = (filp->f_pos) >> sb->s_blocksize_bits; + while (!error && !stored && *f_pos < i_size_read(inode)) { + blk = (*f_pos) >> sb->s_blocksize_bits; bh = ocfs2_bread(inode, blk, &err, 0); if (!bh) { mlog(ML_ERROR, "directory #%llu contains a hole at offset %lld\n", (unsigned long long)OCFS2_I(inode)->ip_blkno, - filp->f_pos); - filp->f_pos += sb->s_blocksize - offset; + *f_pos); + *f_pos += sb->s_blocksize - offset; continue; } @@ -143,7 +729,7 @@ revalidate: * readdir(2), then we might be pointing to an invalid * dirent right now. Scan from the start of the block * to make sure. */ - if (filp->f_version != inode->i_version) { + if (*f_version != inode->i_version) { for (i = 0; i < sb->s_blocksize && i < offset; ) { de = (struct ocfs2_dir_entry *) (bh->b_data + i); /* It's too expensive to do a full @@ -158,21 +744,20 @@ revalidate: i += le16_to_cpu(de->rec_len); } offset = i; - filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1)) + *f_pos = ((*f_pos) & ~(sb->s_blocksize - 1)) | offset; - filp->f_version = inode->i_version; + *f_version = inode->i_version; } - while (!error && filp->f_pos < i_size_read(inode) + while (!error && *f_pos < i_size_read(inode) && offset < sb->s_blocksize) { de = (struct ocfs2_dir_entry *) (bh->b_data + offset); if (!ocfs2_check_dir_entry(inode, de, bh, offset)) { /* On error, skip the f_pos to the next block. */ - filp->f_pos = (filp->f_pos | - (sb->s_blocksize - 1)) + 1; + *f_pos = ((*f_pos) | (sb->s_blocksize - 1)) + 1; brelse(bh); - goto bail; + goto out; } offset += le16_to_cpu(de->rec_len); if (le64_to_cpu(de->inode)) { @@ -183,36 +768,109 @@ revalidate: * not the directory has been modified * during the copy operation. */ - unsigned long version = filp->f_version; + unsigned long version = *f_version; unsigned char d_type = DT_UNKNOWN; if (de->file_type < OCFS2_FT_MAX) d_type = ocfs2_filetype_table[de->file_type]; - error = filldir(dirent, de->name, + error = filldir(priv, de->name, de->name_len, - filp->f_pos, - ino_from_blkno(sb, le64_to_cpu(de->inode)), + *f_pos, + le64_to_cpu(de->inode), d_type); - if (error) + if (error) { + if (filldir_err) + *filldir_err = error; break; - if (version != filp->f_version) + } + if (version != *f_version) goto revalidate; stored ++; } - filp->f_pos += le16_to_cpu(de->rec_len); + *f_pos += le16_to_cpu(de->rec_len); } offset = 0; brelse(bh); } stored = 0; -bail: +out: + return stored; +} + +static int ocfs2_dir_foreach_blk(struct inode *inode, unsigned long *f_version, + loff_t *f_pos, void *priv, filldir_t filldir, + int *filldir_err) +{ + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) + return ocfs2_dir_foreach_blk_id(inode, f_version, f_pos, priv, + filldir, filldir_err); + + return ocfs2_dir_foreach_blk_el(inode, f_version, f_pos, priv, filldir, + filldir_err); +} + +/* + * This is intended to be called from inside other kernel functions, + * so we fake some arguments. + */ +int ocfs2_dir_foreach(struct inode *inode, loff_t *f_pos, void *priv, + filldir_t filldir) +{ + int ret = 0, filldir_err = 0; + unsigned long version = inode->i_version; + + while (*f_pos < i_size_read(inode)) { + ret = ocfs2_dir_foreach_blk(inode, &version, f_pos, priv, + filldir, &filldir_err); + if (ret || filldir_err) + break; + } + + if (ret > 0) + ret = -EIO; + + return 0; +} + +/* + * ocfs2_readdir() + * + */ +int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir) +{ + int error = 0; + struct inode *inode = filp->f_path.dentry->d_inode; + int lock_level = 0; + + mlog_entry("dirino=%llu\n", + (unsigned long long)OCFS2_I(inode)->ip_blkno); + + error = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level); + if (lock_level && error >= 0) { + /* We release EX lock which used to update atime + * and get PR lock again to reduce contention + * on commonly accessed directories. */ + ocfs2_meta_unlock(inode, 1); + lock_level = 0; + error = ocfs2_meta_lock(inode, NULL, 0); + } + if (error < 0) { + if (error != -ENOENT) + mlog_errno(error); + /* we haven't got any yet, so propagate the error. */ + goto bail_nolock; + } + + error = ocfs2_dir_foreach_blk(inode, &filp->f_version, &filp->f_pos, + dirent, filldir, NULL); + ocfs2_meta_unlock(inode, lock_level); bail_nolock: - mlog_exit(stored); + mlog_exit(error); - return stored; + return error; } /* @@ -252,6 +910,23 @@ leave: return status; } +/* + * Convenience function for callers which just want the block number + * mapped to a name and don't require the full dirent info, etc. + */ +int ocfs2_lookup_ino_from_name(struct inode *dir, const char *name, + int namelen, u64 *blkno) +{ + int ret; + struct buffer_head *bh = NULL; + struct ocfs2_dir_entry *dirent = NULL; + + ret = ocfs2_find_files_on_disk(name, namelen, blkno, dir, &bh, &dirent); + brelse(bh); + + return ret; +} + /* Check for a name within a directory. * * Return 0 if the name does not exist @@ -284,77 +959,414 @@ bail: return ret; } +struct ocfs2_empty_dir_priv { + unsigned seen_dot; + unsigned seen_dot_dot; + unsigned seen_other; +}; +static int ocfs2_empty_dir_filldir(void *priv, const char *name, int name_len, + loff_t pos, u64 ino, unsigned type) +{ + struct ocfs2_empty_dir_priv *p = priv; + + /* + * Check the positions of "." and ".." records to be sure + * they're in the correct place. + */ + if (name_len == 1 && !strncmp(".", name, 1) && pos == 0) { + p->seen_dot = 1; + return 0; + } + + if (name_len == 2 && !strncmp("..", name, 2) && + pos == OCFS2_DIR_REC_LEN(1)) { + p->seen_dot_dot = 1; + return 0; + } + + p->seen_other = 1; + return 1; +} /* * routine to check that the specified directory is empty (for rmdir) + * + * Returns 1 if dir is empty, zero otherwise. */ int ocfs2_empty_dir(struct inode *inode) { - unsigned long offset; - struct buffer_head * bh; - struct ocfs2_dir_entry * de, * de1; - struct super_block * sb; - int err; + int ret; + loff_t start = 0; + struct ocfs2_empty_dir_priv priv; + + memset(&priv, 0, sizeof(priv)); + + ret = ocfs2_dir_foreach(inode, &start, &priv, ocfs2_empty_dir_filldir); + if (ret) + mlog_errno(ret); - sb = inode->i_sb; - if ((i_size_read(inode) < - (OCFS2_DIR_REC_LEN(1) + OCFS2_DIR_REC_LEN(2))) || - !(bh = ocfs2_bread(inode, 0, &err, 0))) { - mlog(ML_ERROR, "bad directory (dir #%llu) - no data block\n", + if (!priv.seen_dot || !priv.seen_dot_dot) { + mlog(ML_ERROR, "bad directory (dir #%llu) - no `.' or `..'\n", (unsigned long long)OCFS2_I(inode)->ip_blkno); + /* + * XXX: Is it really safe to allow an unlink to continue? + */ return 1; } - de = (struct ocfs2_dir_entry *) bh->b_data; - de1 = (struct ocfs2_dir_entry *) - ((char *)de + le16_to_cpu(de->rec_len)); - if ((le64_to_cpu(de->inode) != OCFS2_I(inode)->ip_blkno) || - !le64_to_cpu(de1->inode) || - strcmp(".", de->name) || - strcmp("..", de1->name)) { - mlog(ML_ERROR, "bad directory (dir #%llu) - no `.' or `..'\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno); - brelse(bh); - return 1; + return !priv.seen_other; +} + +static void ocfs2_fill_initial_dirents(struct inode *inode, + struct inode *parent, + char *start, unsigned int size) +{ + struct ocfs2_dir_entry *de = (struct ocfs2_dir_entry *)start; + + de->inode = cpu_to_le64(OCFS2_I(inode)->ip_blkno); + de->name_len = 1; + de->rec_len = + cpu_to_le16(OCFS2_DIR_REC_LEN(de->name_len)); + strcpy(de->name, "."); + ocfs2_set_de_type(de, S_IFDIR); + + de = (struct ocfs2_dir_entry *) ((char *)de + le16_to_cpu(de->rec_len)); + de->inode = cpu_to_le64(OCFS2_I(parent)->ip_blkno); + de->rec_len = cpu_to_le16(size - OCFS2_DIR_REC_LEN(1)); + de->name_len = 2; + strcpy(de->name, ".."); + ocfs2_set_de_type(de, S_IFDIR); +} + +/* + * This works together with code in ocfs2_mknod_locked() which sets + * the inline-data flag and initializes the inline-data section. + */ +static int ocfs2_fill_new_dir_id(struct ocfs2_super *osb, + handle_t *handle, + struct inode *parent, + struct inode *inode, + struct buffer_head *di_bh) +{ + int ret; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_inline_data *data = &di->id2.i_data; + unsigned int size = le16_to_cpu(data->id_count); + + ret = ocfs2_journal_access(handle, inode, di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out; } - offset = le16_to_cpu(de->rec_len) + le16_to_cpu(de1->rec_len); - de = (struct ocfs2_dir_entry *)((char *)de1 + le16_to_cpu(de1->rec_len)); - while (offset < i_size_read(inode) ) { - if (!bh || (void *)de >= (void *)(bh->b_data + sb->s_blocksize)) { - brelse(bh); - bh = ocfs2_bread(inode, - offset >> sb->s_blocksize_bits, &err, 0); - if (!bh) { - mlog(ML_ERROR, "dir %llu has a hole at %lu\n", - (unsigned long long)OCFS2_I(inode)->ip_blkno, offset); - offset += sb->s_blocksize; - continue; - } - de = (struct ocfs2_dir_entry *) bh->b_data; - } - if (!ocfs2_check_dir_entry(inode, de, bh, offset)) { - brelse(bh); - return 1; + + ocfs2_fill_initial_dirents(inode, parent, data->id_data, size); + + ocfs2_journal_dirty(handle, di_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + + i_size_write(inode, size); + inode->i_nlink = 2; + inode->i_blocks = ocfs2_inode_sector_count(inode); + + ret = ocfs2_mark_inode_dirty(handle, inode, di_bh); + if (ret < 0) + mlog_errno(ret); + +out: + return ret; +} + +static int ocfs2_fill_new_dir_el(struct ocfs2_super *osb, + handle_t *handle, + struct inode *parent, + struct inode *inode, + struct buffer_head *fe_bh, + struct ocfs2_alloc_context *data_ac) +{ + int status; + struct buffer_head *new_bh = NULL; + + mlog_entry_void(); + + status = ocfs2_do_extend_dir(osb->sb, handle, inode, fe_bh, + data_ac, NULL, &new_bh); + if (status < 0) { + mlog_errno(status); + goto bail; + } + + ocfs2_set_new_buffer_uptodate(inode, new_bh); + + status = ocfs2_journal_access(handle, inode, new_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (status < 0) { + mlog_errno(status); + goto bail; + } + memset(new_bh->b_data, 0, osb->sb->s_blocksize); + + ocfs2_fill_initial_dirents(inode, parent, new_bh->b_data, + osb->sb->s_blocksize); + + status = ocfs2_journal_dirty(handle, new_bh); + if (status < 0) { + mlog_errno(status); + goto bail; + } + + i_size_write(inode, inode->i_sb->s_blocksize); + inode->i_nlink = 2; + inode->i_blocks = ocfs2_inode_sector_count(inode); + status = ocfs2_mark_inode_dirty(handle, inode, fe_bh); + if (status < 0) { + mlog_errno(status); + goto bail; + } + + status = 0; +bail: + if (new_bh) + brelse(new_bh); + + mlog_exit(status); + return status; +} + +int ocfs2_fill_new_dir(struct ocfs2_super *osb, + handle_t *handle, + struct inode *parent, + struct inode *inode, + struct buffer_head *fe_bh, + struct ocfs2_alloc_context *data_ac) +{ + BUG_ON(!ocfs2_supports_inline_data(osb) && data_ac == NULL); + + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) + return ocfs2_fill_new_dir_id(osb, handle, parent, inode, fe_bh); + + return ocfs2_fill_new_dir_el(osb, handle, parent, inode, fe_bh, + data_ac); +} + +static void ocfs2_expand_last_dirent(char *start, unsigned int old_size, + unsigned int new_size) +{ + struct ocfs2_dir_entry *de; + struct ocfs2_dir_entry *prev_de; + char *de_buf, *limit; + unsigned int bytes = new_size - old_size; + + limit = start + old_size; + de_buf = start; + de = (struct ocfs2_dir_entry *)de_buf; + do { + prev_de = de; + de_buf += le16_to_cpu(de->rec_len); + de = (struct ocfs2_dir_entry *)de_buf; + } while (de_buf < limit); + + le16_add_cpu(&prev_de->rec_len, bytes); +} + +/* + * We allocate enough clusters to fulfill "blocks_wanted", but set + * i_size to exactly one block. Ocfs2_extend_dir() will handle the + * rest automatically for us. + * + * *first_block_bh is a pointer to the 1st data block allocated to the + * directory. + */ +static int ocfs2_expand_inline_dir(struct inode *dir, struct buffer_head *di_bh, + unsigned int blocks_wanted, + struct buffer_head **first_block_bh) +{ + int ret, credits = OCFS2_INLINE_TO_EXTENTS_CREDITS; + u32 alloc, bit_off, len; + struct super_block *sb = dir->i_sb; + u64 blkno, bytes = blocks_wanted << sb->s_blocksize_bits; + struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); + struct ocfs2_inode_info *oi = OCFS2_I(dir); + struct ocfs2_alloc_context *data_ac; + struct buffer_head *dirdata_bh = NULL; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + handle_t *handle; + + alloc = ocfs2_clusters_for_bytes(sb, bytes); + + /* + * We should never need more than 2 clusters for this - + * maximum dirent size is far less than one block. In fact, + * the only time we'd need more than one cluster is if + * blocksize == clustersize and the dirent won't fit in the + * extra space that the expansion to a single block gives. As + * of today, that only happens on 4k/4k file systems. + */ + BUG_ON(alloc > 2); + + ret = ocfs2_reserve_clusters(osb, alloc, &data_ac); + if (ret) { + mlog_errno(ret); + goto out; + } + + down_write(&oi->ip_alloc_sem); + + /* + * Prepare for worst case allocation scenario of two seperate + * extents. + */ + if (alloc == 2) + credits += OCFS2_SUBALLOC_ALLOC; + + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto out_sem; + } + + /* + * Try to claim as many clusters as the bitmap can give though + * if we only get one now, that's enough to continue. The rest + * will be claimed after the conversion to extents. + */ + ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, &len); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* + * Operations are carefully ordered so that we set up the new + * data block first. The conversion from inline data to + * extents follows. + */ + blkno = ocfs2_clusters_to_blocks(dir->i_sb, bit_off); + dirdata_bh = sb_getblk(sb, blkno); + if (!dirdata_bh) { + ret = -EIO; + mlog_errno(ret); + goto out_commit; + } + + ocfs2_set_new_buffer_uptodate(dir, dirdata_bh); + + ret = ocfs2_journal_access(handle, dir, dirdata_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + memcpy(dirdata_bh->b_data, di->id2.i_data.id_data, i_size_read(dir)); + memset(dirdata_bh->b_data + i_size_read(dir), 0, + sb->s_blocksize - i_size_read(dir)); + ocfs2_expand_last_dirent(dirdata_bh->b_data, i_size_read(dir), + sb->s_blocksize); + + ret = ocfs2_journal_dirty(handle, dirdata_bh); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* + * Set extent, i_size, etc on the directory. After this, the + * inode should contain the same exact dirents as before and + * be fully accessible from system calls. + * + * We let the later dirent insert modify c/mtime - to the user + * the data hasn't changed. + */ + ret = ocfs2_journal_access(handle, dir, di_bh, + OCFS2_JOURNAL_ACCESS_CREATE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + spin_lock(&oi->ip_lock); + oi->ip_dyn_features &= ~OCFS2_INLINE_DATA_FL; + di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); + spin_unlock(&oi->ip_lock); + + ocfs2_dinode_new_extent_list(dir, di); + + i_size_write(dir, sb->s_blocksize); + dir->i_mtime = dir->i_ctime = CURRENT_TIME; + + di->i_size = cpu_to_le64(sb->s_blocksize); + di->i_ctime = di->i_mtime = cpu_to_le64(dir->i_ctime.tv_sec); + di->i_ctime_nsec = di->i_mtime_nsec = cpu_to_le32(dir->i_ctime.tv_nsec); + dir->i_blocks = ocfs2_inode_sector_count(dir); + + /* + * This should never fail as our extent list is empty and all + * related blocks have been journaled already. + */ + ret = ocfs2_insert_extent(osb, handle, dir, di_bh, 0, blkno, len, 0, + NULL); + if (ret) { + mlog_errno(ret); + goto out; + } + + ret = ocfs2_journal_dirty(handle, di_bh); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* + * We asked for two clusters, but only got one in the 1st + * pass. Claim the 2nd cluster as a separate extent. + */ + if (alloc > len) { + ret = ocfs2_claim_clusters(osb, handle, data_ac, 1, &bit_off, + &len); + if (ret) { + mlog_errno(ret); + goto out_commit; } - if (le64_to_cpu(de->inode)) { - brelse(bh); - return 0; + blkno = ocfs2_clusters_to_blocks(dir->i_sb, bit_off); + + ret = ocfs2_insert_extent(osb, handle, dir, di_bh, 1, blkno, + len, 0, NULL); + if (ret) { + mlog_errno(ret); + goto out; } - offset += le16_to_cpu(de->rec_len); - de = (struct ocfs2_dir_entry *) - ((char *)de + le16_to_cpu(de->rec_len)); } - brelse(bh); - return 1; + + *first_block_bh = dirdata_bh; + dirdata_bh = NULL; + +out_commit: + ocfs2_commit_trans(osb, handle); + +out_sem: + up_write(&oi->ip_alloc_sem); + +out: + if (data_ac) + ocfs2_free_alloc_context(data_ac); + + brelse(dirdata_bh); + + return ret; } /* returns a bh of the 1st new block in the allocation. */ -int ocfs2_do_extend_dir(struct super_block *sb, - handle_t *handle, - struct inode *dir, - struct buffer_head *parent_fe_bh, - struct ocfs2_alloc_context *data_ac, - struct ocfs2_alloc_context *meta_ac, - struct buffer_head **new_bh) +static int ocfs2_do_extend_dir(struct super_block *sb, + handle_t *handle, + struct inode *dir, + struct buffer_head *parent_fe_bh, + struct ocfs2_alloc_context *data_ac, + struct ocfs2_alloc_context *meta_ac, + struct buffer_head **new_bh) { int status; int extend; @@ -396,10 +1408,18 @@ bail: return status; } -/* assumes you already have a cluster lock on the directory. */ +/* + * Assumes you already have a cluster lock on the directory. + * + * 'blocks_wanted' is only used if we have an inline directory which + * is to be turned into an extent based one. The size of the dirent to + * insert might be larger than the space gained by growing to just one + * block, so we may have to grow the inode by two blocks in that case. + */ static int ocfs2_extend_dir(struct ocfs2_super *osb, struct inode *dir, struct buffer_head *parent_fe_bh, + unsigned int blocks_wanted, struct buffer_head **new_de_bh) { int status = 0; @@ -415,6 +1435,38 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb, mlog_entry_void(); + if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + status = ocfs2_expand_inline_dir(dir, parent_fe_bh, + blocks_wanted, &new_bh); + if (status) { + mlog_errno(status); + goto bail; + } + + if (blocks_wanted == 1) { + /* + * If the new dirent will fit inside the space + * created by pushing out to one block, then + * we can complete the operation + * here. Otherwise we have to expand i_size + * and format the 2nd block below. + */ + BUG_ON(new_bh == NULL); + goto bail_bh; + } + + /* + * Get rid of 'new_bh' - we want to format the 2nd + * data block and return that instead. + */ + brelse(new_bh); + new_bh = NULL; + + dir_i_size = i_size_read(dir); + credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS; + goto do_extend; + } + dir_i_size = i_size_read(dir); mlog(0, "extending dir %llu (i_size = %lld)\n", (unsigned long long)OCFS2_I(dir)->ip_blkno, dir_i_size); @@ -452,6 +1504,7 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb, credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS; } +do_extend: down_write(&OCFS2_I(dir)->ip_alloc_sem); drop_alloc_sem = 1; @@ -497,6 +1550,7 @@ static int ocfs2_extend_dir(struct ocfs2_super *osb, goto bail; } +bail_bh: *new_de_bh = new_bh; get_bh(*new_de_bh); bail: @@ -517,41 +1571,71 @@ bail: return status; } -/* - * Search the dir for a good spot, extending it if necessary. The - * block containing an appropriate record is returned in ret_de_bh. - */ -int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb, - struct inode *dir, - struct buffer_head *parent_fe_bh, - const char *name, - int namelen, - struct buffer_head **ret_de_bh) +static int ocfs2_find_dir_space_id(struct inode *dir, struct buffer_head *di_bh, + const char *name, int namelen, + struct buffer_head **ret_de_bh, + unsigned int *blocks_wanted) { - unsigned long offset; - struct buffer_head * bh = NULL; - unsigned short rec_len; - struct ocfs2_dinode *fe; - struct ocfs2_dir_entry *de; - struct super_block *sb; - int status; + int ret; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; + struct ocfs2_dir_entry *de, *last_de = NULL; + char *de_buf, *limit; + unsigned long offset = 0; + unsigned int rec_len, new_rec_len; + + de_buf = di->id2.i_data.id_data; + limit = de_buf + i_size_read(dir); + rec_len = OCFS2_DIR_REC_LEN(namelen); - mlog_entry_void(); + while (de_buf < limit) { + de = (struct ocfs2_dir_entry *)de_buf; - mlog(0, "getting ready to insert namelen %d into dir %llu\n", - namelen, (unsigned long long)OCFS2_I(dir)->ip_blkno); + if (!ocfs2_check_dir_entry(dir, de, di_bh, offset)) { + ret = -ENOENT; + goto out; + } + if (ocfs2_match(namelen, name, de)) { + ret = -EEXIST; + goto out; + } + if (ocfs2_dirent_would_fit(de, rec_len)) { + /* Ok, we found a spot. Return this bh and let + * the caller actually fill it in. */ + *ret_de_bh = di_bh; + get_bh(*ret_de_bh); + ret = 0; + goto out; + } - BUG_ON(!S_ISDIR(dir->i_mode)); - fe = (struct ocfs2_dinode *) parent_fe_bh->b_data; - BUG_ON(le64_to_cpu(fe->i_size) != i_size_read(dir)); + last_de = de; + de_buf += le16_to_cpu(de->rec_len); + offset += le16_to_cpu(de->rec_len); + } - sb = dir->i_sb; + /* + * We're going to require expansion of the directory - figure + * out how many blocks we'll need so that a place for the + * dirent can be found. + */ + *blocks_wanted = 1; + new_rec_len = le16_to_cpu(last_de->rec_len) + (dir->i_sb->s_blocksize - i_size_read(dir)); + if (new_rec_len < (rec_len + OCFS2_DIR_REC_LEN(last_de->name_len))) + *blocks_wanted = 2; + + ret = -ENOSPC; +out: + return ret; +} - if (!namelen) { - status = -EINVAL; - mlog_errno(status); - goto bail; - } +static int ocfs2_find_dir_space_el(struct inode *dir, const char *name, + int namelen, struct buffer_head **ret_de_bh) +{ + unsigned long offset; + struct buffer_head *bh = NULL; + unsigned short rec_len; + struct ocfs2_dir_entry *de; + struct super_block *sb = dir->i_sb; + int status; bh = ocfs2_bread(dir, 0, &status, 0); if (!bh) { @@ -568,17 +1652,11 @@ int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb, bh = NULL; if (i_size_read(dir) <= offset) { - status = ocfs2_extend_dir(osb, - dir, - parent_fe_bh, - &bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - BUG_ON(!bh); - *ret_de_bh = bh; - get_bh(*ret_de_bh); + /* + * Caller will have to expand this + * directory. + */ + status = -ENOSPC; goto bail; } bh = ocfs2_bread(dir, @@ -600,10 +1678,7 @@ int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb, status = -EEXIST; goto bail; } - if (((le64_to_cpu(de->inode) == 0) && - (le16_to_cpu(de->rec_len) >= rec_len)) || - (le16_to_cpu(de->rec_len) >= - (OCFS2_DIR_REC_LEN(de->name_len) + rec_len))) { + if (ocfs2_dirent_would_fit(de, rec_len)) { /* Ok, we found a spot. Return this bh and let * the caller actually fill it in. */ *ret_de_bh = bh; @@ -623,3 +1698,61 @@ bail: mlog_exit(status); return status; } + +int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb, + struct inode *dir, + struct buffer_head *parent_fe_bh, + const char *name, + int namelen, + struct buffer_head **ret_de_bh) +{ + int ret; + unsigned int blocks_wanted = 1; + struct buffer_head *bh = NULL; + + mlog(0, "getting ready to insert namelen %d into dir %llu\n", + namelen, (unsigned long long)OCFS2_I(dir)->ip_blkno); + + *ret_de_bh = NULL; + + if (!namelen) { + ret = -EINVAL; + mlog_errno(ret); + goto out; + } + + if (OCFS2_I(dir)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + ret = ocfs2_find_dir_space_id(dir, parent_fe_bh, name, + namelen, &bh, &blocks_wanted); + } else + ret = ocfs2_find_dir_space_el(dir, name, namelen, &bh); + + if (ret && ret != -ENOSPC) { + mlog_errno(ret); + goto out; + } + + if (ret == -ENOSPC) { + /* + * We have to expand the directory to add this name. + */ + BUG_ON(bh); + + ret = ocfs2_extend_dir(osb, dir, parent_fe_bh, blocks_wanted, + &bh); + if (ret) { + if (ret != -ENOSPC) + mlog_errno(ret); + goto out; + } + + BUG_ON(!bh); + } + + *ret_de_bh = bh; + bh = NULL; +out: + if (bh) + brelse(bh); + return ret; +} diff --git a/fs/ocfs2/dir.h b/fs/ocfs2/dir.h index 3f67e146864..ce48b9080d8 100644 --- a/fs/ocfs2/dir.h +++ b/fs/ocfs2/dir.h @@ -26,17 +26,49 @@ #ifndef OCFS2_DIR_H #define OCFS2_DIR_H +struct buffer_head *ocfs2_find_entry(const char *name, + int namelen, + struct inode *dir, + struct ocfs2_dir_entry **res_dir); +int ocfs2_delete_entry(handle_t *handle, + struct inode *dir, + struct ocfs2_dir_entry *de_del, + struct buffer_head *bh); +int __ocfs2_add_entry(handle_t *handle, + struct inode *dir, + const char *name, int namelen, + struct inode *inode, u64 blkno, + struct buffer_head *parent_fe_bh, + struct buffer_head *insert_bh); +static inline int ocfs2_add_entry(handle_t *handle, + struct dentry *dentry, + struct inode *inode, u64 blkno, + struct buffer_head *parent_fe_bh, + struct buffer_head *insert_bh) +{ + return __ocfs2_add_entry(handle, dentry->d_parent->d_inode, + dentry->d_name.name, dentry->d_name.len, + inode, blkno, parent_fe_bh, insert_bh); +} +int ocfs2_update_entry(struct inode *dir, handle_t *handle, + struct buffer_head *de_bh, struct ocfs2_dir_entry *de, + struct inode *new_entry_inode); + int ocfs2_check_dir_for_entry(struct inode *dir, const char *name, int namelen); -int ocfs2_empty_dir(struct inode *inode); /* FIXME: to namei.c */ +int ocfs2_empty_dir(struct inode *inode); int ocfs2_find_files_on_disk(const char *name, int namelen, u64 *blkno, struct inode *inode, struct buffer_head **dirent_bh, struct ocfs2_dir_entry **dirent); +int ocfs2_lookup_ino_from_name(struct inode *dir, const char *name, + int namelen, u64 *blkno); int ocfs2_readdir(struct file *filp, void *dirent, filldir_t filldir); +int ocfs2_dir_foreach(struct inode *inode, loff_t *f_pos, void *priv, + filldir_t filldir); int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb, struct inode *dir, struct buffer_head *parent_fe_bh, @@ -44,11 +76,11 @@ int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb, int namelen, struct buffer_head **ret_de_bh); struct ocfs2_alloc_context; -int ocfs2_do_extend_dir(struct super_block *sb, - handle_t *handle, - struct inode *dir, - struct buffer_head *parent_fe_bh, - struct ocfs2_alloc_context *data_ac, - struct ocfs2_alloc_context *meta_ac, - struct buffer_head **new_bh); +int ocfs2_fill_new_dir(struct ocfs2_super *osb, + handle_t *handle, + struct inode *parent, + struct inode *inode, + struct buffer_head *fe_bh, + struct ocfs2_alloc_context *data_ac); + #endif /* OCFS2_DIR_H */ diff --git a/fs/ocfs2/dlm/dlmfs.c b/fs/ocfs2/dlm/dlmfs.c index fd8cb1badc9..7418dc83de1 100644 --- a/fs/ocfs2/dlm/dlmfs.c +++ b/fs/ocfs2/dlm/dlmfs.c @@ -592,7 +592,7 @@ static int __init init_dlmfs_fs(void) sizeof(struct dlmfs_inode_private), 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| SLAB_MEM_SPREAD), - dlmfs_init_once, NULL); + dlmfs_init_once); if (!dlmfs_inode_cache) return -ENOMEM; cleanup_inode = 1; diff --git a/fs/ocfs2/dlm/dlmmaster.c b/fs/ocfs2/dlm/dlmmaster.c index 65b2b9b9268..62e4a7daa28 100644 --- a/fs/ocfs2/dlm/dlmmaster.c +++ b/fs/ocfs2/dlm/dlmmaster.c @@ -510,7 +510,7 @@ int dlm_init_mle_cache(void) dlm_mle_cache = kmem_cache_create("dlm_mle_cache", sizeof(struct dlm_master_list_entry), 0, SLAB_HWCACHE_ALIGN, - NULL, NULL); + NULL); if (dlm_mle_cache == NULL) return -ENOMEM; return 0; diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index f71250ed166..41c76ff2fcf 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -1482,6 +1482,7 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) lvb->lvb_imtime_packed = cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime)); lvb->lvb_iattr = cpu_to_be32(oi->ip_attr); + lvb->lvb_idynfeatures = cpu_to_be16(oi->ip_dyn_features); lvb->lvb_igeneration = cpu_to_be32(inode->i_generation); out: @@ -1515,6 +1516,7 @@ static void ocfs2_refresh_inode_from_lvb(struct inode *inode) i_size_write(inode, be64_to_cpu(lvb->lvb_isize)); oi->ip_attr = be32_to_cpu(lvb->lvb_iattr); + oi->ip_dyn_features = be16_to_cpu(lvb->lvb_idynfeatures); ocfs2_set_inode_flags(inode); /* fast-symlinks are a special case */ diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h index 492bad32a8c..87a785e4120 100644 --- a/fs/ocfs2/dlmglue.h +++ b/fs/ocfs2/dlmglue.h @@ -29,12 +29,12 @@ #include "dcache.h" -#define OCFS2_LVB_VERSION 4 +#define OCFS2_LVB_VERSION 5 struct ocfs2_meta_lvb { __u8 lvb_version; __u8 lvb_reserved0; - __be16 lvb_reserved1; + __be16 lvb_idynfeatures; __be32 lvb_iclusters; __be32 lvb_iuid; __be32 lvb_igid; diff --git a/fs/ocfs2/export.c b/fs/ocfs2/export.c index bc48177bd18..c3bbc198f9c 100644 --- a/fs/ocfs2/export.c +++ b/fs/ocfs2/export.c @@ -88,8 +88,6 @@ static struct dentry *ocfs2_get_parent(struct dentry *child) struct dentry *parent; struct inode *inode; struct inode *dir = child->d_inode; - struct buffer_head *dirent_bh = NULL; - struct ocfs2_dir_entry *dirent; mlog_entry("(0x%p, '%.*s')\n", child, child->d_name.len, child->d_name.name); @@ -105,8 +103,7 @@ static struct dentry *ocfs2_get_parent(struct dentry *child) goto bail; } - status = ocfs2_find_files_on_disk("..", 2, &blkno, dir, &dirent_bh, - &dirent); + status = ocfs2_lookup_ino_from_name(dir, "..", 2, &blkno); if (status < 0) { parent = ERR_PTR(-ENOENT); goto bail_unlock; @@ -131,9 +128,6 @@ static struct dentry *ocfs2_get_parent(struct dentry *child) bail_unlock: ocfs2_meta_unlock(dir, 0); - if (dirent_bh) - brelse(dirent_bh); - bail: mlog_exit_ptr(parent); diff --git a/fs/ocfs2/extent_map.c b/fs/ocfs2/extent_map.c index 03c1d365c78..c58668a326f 100644 --- a/fs/ocfs2/extent_map.c +++ b/fs/ocfs2/extent_map.c @@ -387,6 +387,12 @@ int ocfs2_get_clusters(struct inode *inode, u32 v_cluster, struct ocfs2_extent_rec *rec; u32 coff; + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + ret = -ERANGE; + mlog_errno(ret); + goto out; + } + ret = ocfs2_extent_map_lookup(inode, v_cluster, p_cluster, num_clusters, extent_flags); if (ret == 0) diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index 004c2abbc73..a62b14eb406 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -34,6 +34,7 @@ #include <linux/splice.h> #include <linux/mount.h> #include <linux/writeback.h> +#include <linux/falloc.h> #define MLOG_MASK_PREFIX ML_INODE #include <cluster/masklog.h> @@ -186,6 +187,7 @@ int ocfs2_update_inode_atime(struct inode *inode, int ret; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); handle_t *handle; + struct ocfs2_dinode *di = (struct ocfs2_dinode *) bh->b_data; mlog_entry_void(); @@ -196,11 +198,27 @@ int ocfs2_update_inode_atime(struct inode *inode, goto out; } + ret = ocfs2_journal_access(handle, inode, bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto out_commit; + } + + /* + * Don't use ocfs2_mark_inode_dirty() here as we don't always + * have i_mutex to guard against concurrent changes to other + * inode fields. + */ inode->i_atime = CURRENT_TIME; - ret = ocfs2_mark_inode_dirty(handle, inode, bh); + di->i_atime = cpu_to_le64(inode->i_atime.tv_sec); + di->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec); + + ret = ocfs2_journal_dirty(handle, bh); if (ret < 0) mlog_errno(ret); +out_commit: ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); out: mlog_exit(ret); @@ -296,7 +314,6 @@ static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb, } i_size_write(inode, new_i_size); - inode->i_blocks = ocfs2_align_bytes_to_sectors(new_i_size); inode->i_ctime = inode->i_mtime = CURRENT_TIME; di = (struct ocfs2_dinode *) fe_bh->b_data; @@ -380,6 +397,15 @@ static int ocfs2_truncate_file(struct inode *inode, unmap_mapping_range(inode->i_mapping, new_i_size + PAGE_SIZE - 1, 0, 1); truncate_inode_pages(inode->i_mapping, new_i_size); + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + status = ocfs2_truncate_inline(inode, di_bh, new_i_size, + i_size_read(inode), 0); + if (status) + mlog_errno(status); + + goto bail_unlock_data; + } + /* alright, we're going to need to do a full blown alloc size * change. Orphan the inode so that recovery can complete the * truncate if necessary. This does the task of marking @@ -474,8 +500,8 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb, goto leave; } - status = ocfs2_claim_clusters(osb, handle, data_ac, 1, - &bit_off, &num_bits); + 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); @@ -762,25 +788,6 @@ leave: return status; } -static int ocfs2_extend_allocation(struct inode *inode, u32 logical_start, - u32 clusters_to_add, int mark_unwritten) -{ - int ret; - - /* - * The alloc sem blocks peope in read/write from reading our - * allocation until we're done changing it. We depend on - * i_mutex to block other extend/truncate calls while we're - * here. - */ - down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = __ocfs2_extend_allocation(inode, logical_start, clusters_to_add, - mark_unwritten); - up_write(&OCFS2_I(inode)->ip_alloc_sem); - - return ret; -} - /* Some parts of this taken from generic_cont_expand, which turned out * to be too fragile to do exactly what we need without us having to * worry about recursive locking in ->prepare_write() and @@ -872,25 +879,48 @@ out: return ret; } -/* - * A tail_to_skip value > 0 indicates that we're being called from - * ocfs2_file_aio_write(). This has the following implications: - * - * - we don't want to update i_size - * - di_bh will be NULL, which is fine because it's only used in the - * case where we want to update i_size. - * - ocfs2_zero_extend() will then only be filling the hole created - * between i_size and the start of the write. - */ +int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size, u64 zero_to) +{ + int ret; + u32 clusters_to_add; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + + clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size); + if (clusters_to_add < oi->ip_clusters) + clusters_to_add = 0; + else + clusters_to_add -= oi->ip_clusters; + + if (clusters_to_add) { + ret = __ocfs2_extend_allocation(inode, oi->ip_clusters, + clusters_to_add, 0); + if (ret) { + mlog_errno(ret); + goto out; + } + } + + /* + * Call this even if we don't add any clusters to the tree. We + * still need to zero the area between the old i_size and the + * new i_size. + */ + ret = ocfs2_zero_extend(inode, zero_to); + if (ret < 0) + mlog_errno(ret); + +out: + return ret; +} + static int ocfs2_extend_file(struct inode *inode, struct buffer_head *di_bh, - u64 new_i_size, - size_t tail_to_skip) + u64 new_i_size) { - int ret = 0; - u32 clusters_to_add = 0; + int ret = 0, data_locked = 0; + struct ocfs2_inode_info *oi = OCFS2_I(inode); - BUG_ON(!tail_to_skip && !di_bh); + BUG_ON(!di_bh); /* setattr sometimes calls us like this. */ if (new_i_size == 0) @@ -900,13 +930,18 @@ static int ocfs2_extend_file(struct inode *inode, goto out; BUG_ON(new_i_size < i_size_read(inode)); - if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) { - BUG_ON(tail_to_skip != 0); + /* + * Fall through for converting inline data, even if the fs + * supports sparse files. + * + * The check for inline data here is legal - nobody can add + * the feature since we have i_mutex. We must check it again + * after acquiring ip_alloc_sem though, as paths like mmap + * might have raced us to converting the inode to extents. + */ + if (!(oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) + && ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) goto out_update_size; - } - - clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size) - - OCFS2_I(inode)->ip_clusters; /* * protect the pages that ocfs2_zero_extend is going to be @@ -920,39 +955,52 @@ static int ocfs2_extend_file(struct inode *inode, mlog_errno(ret); goto out; } + data_locked = 1; + + /* + * The alloc sem blocks people in read/write from reading our + * allocation until we're done changing it. We depend on + * i_mutex to block other extend/truncate calls while we're + * here. + */ + down_write(&oi->ip_alloc_sem); + + if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + /* + * We can optimize small extends by keeping the inodes + * inline data. + */ + if (ocfs2_size_fits_inline_data(di_bh, new_i_size)) { + up_write(&oi->ip_alloc_sem); + goto out_update_size; + } + + ret = ocfs2_convert_inline_data_to_extents(inode, di_bh); + if (ret) { + up_write(&oi->ip_alloc_sem); - if (clusters_to_add) { - ret = ocfs2_extend_allocation(inode, - OCFS2_I(inode)->ip_clusters, - clusters_to_add, 0); - if (ret < 0) { mlog_errno(ret); goto out_unlock; } } - /* - * Call this even if we don't add any clusters to the tree. We - * still need to zero the area between the old i_size and the - * new i_size. - */ - ret = ocfs2_zero_extend(inode, (u64)new_i_size - tail_to_skip); + if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) + ret = ocfs2_extend_no_holes(inode, new_i_size, new_i_size); + + up_write(&oi->ip_alloc_sem); + if (ret < 0) { mlog_errno(ret); goto out_unlock; } out_update_size: - if (!tail_to_skip) { - /* We're being called from ocfs2_setattr() which wants - * us to update i_size */ - ret = ocfs2_simple_size_update(inode, di_bh, new_i_size); - if (ret < 0) - mlog_errno(ret); - } + ret = ocfs2_simple_size_update(inode, di_bh, new_i_size); + if (ret < 0) + mlog_errno(ret); out_unlock: - if (!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) + if (data_locked) ocfs2_data_unlock(inode, 1); out: @@ -1010,10 +1058,15 @@ int ocfs2_setattr(struct dentry *dentry, struct iattr *attr) } if (size_change && attr->ia_size != i_size_read(inode)) { + if (attr->ia_size > sb->s_maxbytes) { + status = -EFBIG; + goto bail_unlock; + } + if (i_size_read(inode) > attr->ia_size) status = ocfs2_truncate_file(inode, bh, attr->ia_size); else - status = ocfs2_extend_file(inode, bh, attr->ia_size, 0); + status = ocfs2_extend_file(inode, bh, attr->ia_size); if (status < 0) { if (status != -ENOSPC) mlog_errno(status); @@ -1221,6 +1274,31 @@ static int ocfs2_allocate_unwritten_extents(struct inode *inode, { int ret; u32 cpos, phys_cpos, clusters, alloc_size; + u64 end = start + len; + struct buffer_head *di_bh = NULL; + + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), + OCFS2_I(inode)->ip_blkno, &di_bh, + OCFS2_BH_CACHED, inode); + if (ret) { + mlog_errno(ret); + goto out; + } + + /* + * Nothing to do if the requested reservation range + * fits within the inode. + */ + if (ocfs2_size_fits_inline_data(di_bh, end)) + goto out; + + ret = ocfs2_convert_inline_data_to_extents(inode, di_bh); + if (ret) { + mlog_errno(ret); + goto out; + } + } /* * We consider both start and len to be inclusive. @@ -1266,6 +1344,8 @@ next: ret = 0; out: + + brelse(di_bh); return ret; } @@ -1447,6 +1527,14 @@ static int ocfs2_remove_inode_range(struct inode *inode, if (byte_len == 0) return 0; + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + ret = ocfs2_truncate_inline(inode, di_bh, byte_start, + byte_start + byte_len, 1); + if (ret) + mlog_errno(ret); + return ret; + } + trunc_start = ocfs2_clusters_for_bytes(osb->sb, byte_start); trunc_len = (byte_start + byte_len) >> osb->s_clustersize_bits; if (trunc_len >= trunc_start) @@ -1504,29 +1592,18 @@ out: /* * Parts of this function taken from xfs_change_file_space() */ -int ocfs2_change_file_space(struct file *file, unsigned int cmd, - struct ocfs2_space_resv *sr) +static int __ocfs2_change_file_space(struct file *file, struct inode *inode, + loff_t f_pos, unsigned int cmd, + struct ocfs2_space_resv *sr, + int change_size) { int ret; s64 llen; - struct inode *inode = file->f_path.dentry->d_inode; + loff_t size; struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); struct buffer_head *di_bh = NULL; handle_t *handle; - unsigned long long max_off = ocfs2_max_file_offset(inode->i_sb->s_blocksize_bits); - - if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) && - !ocfs2_writes_unwritten_extents(osb)) - return -ENOTTY; - else if ((cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) && - !ocfs2_sparse_alloc(osb)) - return -ENOTTY; - - if (!S_ISREG(inode->i_mode)) - return -EINVAL; - - if (!(file->f_mode & FMODE_WRITE)) - return -EBADF; + unsigned long long max_off = inode->i_sb->s_maxbytes; if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) return -EROFS; @@ -1557,7 +1634,7 @@ int ocfs2_change_file_space(struct file *file, unsigned int cmd, case 0: /*SEEK_SET*/ break; case 1: /*SEEK_CUR*/ - sr->l_start += file->f_pos; + sr->l_start += f_pos; break; case 2: /*SEEK_END*/ sr->l_start += i_size_read(inode); @@ -1577,6 +1654,7 @@ int ocfs2_change_file_space(struct file *file, unsigned int cmd, ret = -EINVAL; goto out_meta_unlock; } + size = sr->l_start + sr->l_len; if (cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) { if (sr->l_len <= 0) { @@ -1585,7 +1663,7 @@ int ocfs2_change_file_space(struct file *file, unsigned int cmd, } } - if (should_remove_suid(file->f_path.dentry)) { + if (file && should_remove_suid(file->f_path.dentry)) { ret = __ocfs2_write_remove_suid(inode, di_bh); if (ret) { mlog_errno(ret); @@ -1628,6 +1706,9 @@ int ocfs2_change_file_space(struct file *file, unsigned int cmd, goto out_meta_unlock; } + if (change_size && i_size_read(inode) < size) + i_size_write(inode, size); + inode->i_ctime = inode->i_mtime = CURRENT_TIME; ret = ocfs2_mark_inode_dirty(handle, inode, di_bh); if (ret < 0) @@ -1646,21 +1727,65 @@ out: return ret; } +int ocfs2_change_file_space(struct file *file, unsigned int cmd, + struct ocfs2_space_resv *sr) +{ + struct inode *inode = file->f_path.dentry->d_inode; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);; + + if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) && + !ocfs2_writes_unwritten_extents(osb)) + return -ENOTTY; + else if ((cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) && + !ocfs2_sparse_alloc(osb)) + return -ENOTTY; + + if (!S_ISREG(inode->i_mode)) + return -EINVAL; + + if (!(file->f_mode & FMODE_WRITE)) + return -EBADF; + + return __ocfs2_change_file_space(file, inode, file->f_pos, cmd, sr, 0); +} + +static long ocfs2_fallocate(struct inode *inode, int mode, loff_t offset, + loff_t len) +{ + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_space_resv sr; + int change_size = 1; + + if (!ocfs2_writes_unwritten_extents(osb)) + return -EOPNOTSUPP; + + if (S_ISDIR(inode->i_mode)) + return -ENODEV; + + if (mode & FALLOC_FL_KEEP_SIZE) + change_size = 0; + + sr.l_whence = 0; + sr.l_start = (s64)offset; + sr.l_len = (s64)len; + + return __ocfs2_change_file_space(NULL, inode, offset, + OCFS2_IOC_RESVSP64, &sr, change_size); +} + static int ocfs2_prepare_inode_for_write(struct dentry *dentry, loff_t *ppos, size_t count, int appending, int *direct_io) { - int ret = 0, meta_level = appending; + int ret = 0, meta_level = 0; struct inode *inode = dentry->d_inode; - u32 clusters; - loff_t newsize, saved_pos; + loff_t saved_pos, end; /* - * We sample i_size under a read level meta lock to see if our write - * is extending the file, if it is we back off and get a write level - * meta lock. + * We start with a read level meta lock and only jump to an ex + * if we need to make modifications here. */ for(;;) { ret = ocfs2_meta_lock(inode, NULL, meta_level); @@ -1702,87 +1827,47 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, saved_pos = *ppos; } - if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) { - loff_t end = saved_pos + count; - - /* - * Skip the O_DIRECT checks if we don't need - * them. - */ - if (!direct_io || !(*direct_io)) - break; - - /* - * Allowing concurrent direct writes means - * i_size changes wouldn't be synchronized, so - * one node could wind up truncating another - * nodes writes. - */ - if (end > i_size_read(inode)) { - *direct_io = 0; - break; - } + end = saved_pos + count; - /* - * We don't fill holes during direct io, so - * check for them here. If any are found, the - * caller will have to retake some cluster - * locks and initiate the io as buffered. - */ - ret = ocfs2_check_range_for_holes(inode, saved_pos, - count); - if (ret == 1) { - *direct_io = 0; - ret = 0; - } else if (ret < 0) - mlog_errno(ret); + /* + * Skip the O_DIRECT checks if we don't need + * them. + */ + if (!direct_io || !(*direct_io)) break; - } /* - * The rest of this loop is concerned with legacy file - * systems which don't support sparse files. + * There's no sane way to do direct writes to an inode + * with inline data. */ - - newsize = count + saved_pos; - - mlog(0, "pos=%lld newsize=%lld cursize=%lld\n", - (long long) saved_pos, (long long) newsize, - (long long) i_size_read(inode)); - - /* No need for a higher level metadata lock if we're - * never going past i_size. */ - if (newsize <= i_size_read(inode)) + if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) { + *direct_io = 0; break; - - if (meta_level == 0) { - ocfs2_meta_unlock(inode, meta_level); - meta_level = 1; - continue; } - spin_lock(&OCFS2_I(inode)->ip_lock); - clusters = ocfs2_clusters_for_bytes(inode->i_sb, newsize) - - OCFS2_I(inode)->ip_clusters; - spin_unlock(&OCFS2_I(inode)->ip_lock); - - mlog(0, "Writing at EOF, may need more allocation: " - "i_size = %lld, newsize = %lld, need %u clusters\n", - (long long) i_size_read(inode), (long long) newsize, - clusters); - - /* We only want to continue the rest of this loop if - * our extend will actually require more - * allocation. */ - if (!clusters) + /* + * Allowing concurrent direct writes means + * i_size changes wouldn't be synchronized, so + * one node could wind up truncating another + * nodes writes. + */ + if (end > i_size_read(inode)) { + *direct_io = 0; break; - - ret = ocfs2_extend_file(inode, NULL, newsize, count); - if (ret < 0) { - if (ret != -ENOSPC) - mlog_errno(ret); - goto out_unlock; } + + /* + * We don't fill holes during direct io, so + * check for them here. If any are found, the + * caller will have to retake some cluster + * locks and initiate the io as buffered. + */ + ret = ocfs2_check_range_for_holes(inode, saved_pos, count); + if (ret == 1) { + *direct_io = 0; + ret = 0; + } else if (ret < 0) + mlog_errno(ret); break; } @@ -1902,7 +1987,7 @@ static ssize_t ocfs2_file_buffered_write(struct file *file, loff_t *ppos, } dst = kmap_atomic(page, KM_USER0); - memcpy(dst + (pos & (PAGE_CACHE_SIZE - 1)), buf, bytes); + memcpy(dst + (pos & (loff_t)(PAGE_CACHE_SIZE - 1)), buf, bytes); kunmap_atomic(dst, KM_USER0); flush_dcache_page(page); ocfs2_put_write_source(user_page); @@ -2113,7 +2198,7 @@ static int ocfs2_splice_write_actor(struct pipe_inode_info *pipe, src = buf->ops->map(pipe, buf, 1); dst = kmap_atomic(page, KM_USER1); memcpy(dst + offset, src + buf->offset, count); - kunmap_atomic(page, KM_USER1); + kunmap_atomic(dst, KM_USER1); buf->ops->unmap(pipe, buf, src); copied = ocfs2_write_end(file, file->f_mapping, sd->pos, count, count, @@ -2312,6 +2397,7 @@ const struct inode_operations ocfs2_file_iops = { .setattr = ocfs2_setattr, .getattr = ocfs2_getattr, .permission = ocfs2_permission, + .fallocate = ocfs2_fallocate, }; const struct inode_operations ocfs2_special_file_iops = { diff --git a/fs/ocfs2/file.h b/fs/ocfs2/file.h index 36fe27f268e..066f14add3a 100644 --- a/fs/ocfs2/file.h +++ b/fs/ocfs2/file.h @@ -47,6 +47,8 @@ int ocfs2_do_extend_allocation(struct ocfs2_super *osb, struct ocfs2_alloc_context *data_ac, struct ocfs2_alloc_context *meta_ac, enum ocfs2_alloc_restarted *reason_ret); +int ocfs2_extend_no_holes(struct inode *inode, u64 new_i_size, + u64 zero_to); int ocfs2_lock_allocators(struct inode *inode, struct ocfs2_dinode *di, u32 clusters_to_add, u32 extents_to_split, struct ocfs2_alloc_context **data_ac, diff --git a/fs/ocfs2/heartbeat.c b/fs/ocfs2/heartbeat.c index 352eb4a13f9..c4c36171240 100644 --- a/fs/ocfs2/heartbeat.c +++ b/fs/ocfs2/heartbeat.c @@ -209,7 +209,7 @@ void ocfs2_stop_heartbeat(struct ocfs2_super *osb) envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin"; envp[2] = NULL; - ret = call_usermodehelper(argv[0], argv, envp, 1); + ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC); if (ret < 0) mlog_errno(ret); } diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c index c53a6763bbb..1d5e0cb0fda 100644 --- a/fs/ocfs2/inode.c +++ b/fs/ocfs2/inode.c @@ -241,6 +241,7 @@ int ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters); OCFS2_I(inode)->ip_attr = le32_to_cpu(fe->i_attr); + OCFS2_I(inode)->ip_dyn_features = le16_to_cpu(fe->i_dyn_features); inode->i_version = 1; inode->i_generation = le32_to_cpu(fe->i_generation); @@ -513,6 +514,10 @@ static int ocfs2_truncate_for_delete(struct ocfs2_super *osb, fe = (struct ocfs2_dinode *) fe_bh->b_data; + /* + * This check will also skip truncate of inodes with inline + * data and fast symlinks. + */ if (fe->i_clusters) { handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); if (IS_ERR(handle)) { @@ -1220,6 +1225,7 @@ int ocfs2_mark_inode_dirty(handle_t *handle, fe->i_clusters = cpu_to_le32(OCFS2_I(inode)->ip_clusters); ocfs2_get_inode_flags(OCFS2_I(inode)); fe->i_attr = cpu_to_le32(OCFS2_I(inode)->ip_attr); + fe->i_dyn_features = cpu_to_le16(OCFS2_I(inode)->ip_dyn_features); spin_unlock(&OCFS2_I(inode)->ip_lock); fe->i_size = cpu_to_le64(i_size_read(inode)); @@ -1257,6 +1263,7 @@ void ocfs2_refresh_inode(struct inode *inode, OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters); OCFS2_I(inode)->ip_attr = le32_to_cpu(fe->i_attr); + OCFS2_I(inode)->ip_dyn_features = le16_to_cpu(fe->i_dyn_features); ocfs2_set_inode_flags(inode); i_size_write(inode, le64_to_cpu(fe->i_size)); inode->i_nlink = le16_to_cpu(fe->i_links_count); diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h index a41d0817121..70e881c5553 100644 --- a/fs/ocfs2/inode.h +++ b/fs/ocfs2/inode.h @@ -51,6 +51,7 @@ struct ocfs2_inode_info u32 ip_flags; /* see below */ u32 ip_attr; /* inode attributes */ + u16 ip_dyn_features; /* protected by recovery_lock. */ struct inode *ip_next_orphan; diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c index dbfb20bb27e..f9d01e25298 100644 --- a/fs/ocfs2/journal.c +++ b/fs/ocfs2/journal.c @@ -35,13 +35,13 @@ #include "ocfs2.h" #include "alloc.h" +#include "dir.h" #include "dlmglue.h" #include "extent_map.h" #include "heartbeat.h" #include "inode.h" #include "journal.h" #include "localalloc.h" -#include "namei.h" #include "slot_map.h" #include "super.h" #include "vote.h" @@ -1213,17 +1213,49 @@ bail: return status; } +struct ocfs2_orphan_filldir_priv { + struct inode *head; + struct ocfs2_super *osb; +}; + +static int ocfs2_orphan_filldir(void *priv, const char *name, int name_len, + loff_t pos, u64 ino, unsigned type) +{ + struct ocfs2_orphan_filldir_priv *p = priv; + struct inode *iter; + + if (name_len == 1 && !strncmp(".", name, 1)) + return 0; + if (name_len == 2 && !strncmp("..", name, 2)) + return 0; + + /* Skip bad inodes so that recovery can continue */ + iter = ocfs2_iget(p->osb, ino, + OCFS2_FI_FLAG_ORPHAN_RECOVERY); + if (IS_ERR(iter)) + return 0; + + mlog(0, "queue orphan %llu\n", + (unsigned long long)OCFS2_I(iter)->ip_blkno); + /* No locking is required for the next_orphan queue as there + * is only ever a single process doing orphan recovery. */ + OCFS2_I(iter)->ip_next_orphan = p->head; + p->head = iter; + + return 0; +} + static int ocfs2_queue_orphans(struct ocfs2_super *osb, int slot, struct inode **head) { int status; struct inode *orphan_dir_inode = NULL; - struct inode *iter; - unsigned long offset, blk, local; - struct buffer_head *bh = NULL; - struct ocfs2_dir_entry *de; - struct super_block *sb = osb->sb; + struct ocfs2_orphan_filldir_priv priv; + loff_t pos = 0; + + priv.osb = osb; + priv.head = *head; orphan_dir_inode = ocfs2_get_system_file_inode(osb, ORPHAN_DIR_SYSTEM_INODE, @@ -1241,77 +1273,15 @@ static int ocfs2_queue_orphans(struct ocfs2_super *osb, goto out; } - offset = 0; - iter = NULL; - while(offset < i_size_read(orphan_dir_inode)) { - blk = offset >> sb->s_blocksize_bits; - - bh = ocfs2_bread(orphan_dir_inode, blk, &status, 0); - if (!bh) - status = -EINVAL; - if (status < 0) { - if (bh) - brelse(bh); - mlog_errno(status); - goto out_unlock; - } - - local = 0; - while(offset < i_size_read(orphan_dir_inode) - && local < sb->s_blocksize) { - de = (struct ocfs2_dir_entry *) (bh->b_data + local); - - if (!ocfs2_check_dir_entry(orphan_dir_inode, - de, bh, local)) { - status = -EINVAL; - mlog_errno(status); - brelse(bh); - goto out_unlock; - } - - local += le16_to_cpu(de->rec_len); - offset += le16_to_cpu(de->rec_len); - - /* I guess we silently fail on no inode? */ - if (!le64_to_cpu(de->inode)) - continue; - if (de->file_type > OCFS2_FT_MAX) { - mlog(ML_ERROR, - "block %llu contains invalid de: " - "inode = %llu, rec_len = %u, " - "name_len = %u, file_type = %u, " - "name='%.*s'\n", - (unsigned long long)bh->b_blocknr, - (unsigned long long)le64_to_cpu(de->inode), - le16_to_cpu(de->rec_len), - de->name_len, - de->file_type, - de->name_len, - de->name); - continue; - } - if (de->name_len == 1 && !strncmp(".", de->name, 1)) - continue; - if (de->name_len == 2 && !strncmp("..", de->name, 2)) - continue; - - iter = ocfs2_iget(osb, le64_to_cpu(de->inode), - OCFS2_FI_FLAG_ORPHAN_RECOVERY); - if (IS_ERR(iter)) - continue; - - mlog(0, "queue orphan %llu\n", - (unsigned long long)OCFS2_I(iter)->ip_blkno); - /* No locking is required for the next_orphan - * queue as there is only ever a single - * process doing orphan recovery. */ - OCFS2_I(iter)->ip_next_orphan = *head; - *head = iter; - } - brelse(bh); + status = ocfs2_dir_foreach(orphan_dir_inode, &pos, &priv, + ocfs2_orphan_filldir); + if (status) { + mlog_errno(status); + goto out; } -out_unlock: + *head = priv.head; + ocfs2_meta_unlock(orphan_dir_inode, 0); out: mutex_unlock(&orphan_dir_inode->i_mutex); diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h index ce60aab013a..4b32e096156 100644 --- a/fs/ocfs2/journal.h +++ b/fs/ocfs2/journal.h @@ -282,6 +282,9 @@ int ocfs2_journal_dirty_data(handle_t *handle, * prev. group desc. if we relink. */ #define OCFS2_SUBALLOC_ALLOC (3) +#define OCFS2_INLINE_TO_EXTENTS_CREDITS (OCFS2_SUBALLOC_ALLOC \ + + OCFS2_INODE_UPDATE_CREDITS) + /* dinode + group descriptor update. We don't relink on free yet. */ #define OCFS2_SUBALLOC_FREE (2) diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c index 545f7892cdf..d272847d5a0 100644 --- a/fs/ocfs2/localalloc.c +++ b/fs/ocfs2/localalloc.c @@ -514,8 +514,10 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb, ac->ac_bh = osb->local_alloc_bh; status = 0; bail: - if (status < 0 && local_alloc_inode) + if (status < 0 && local_alloc_inode) { + mutex_unlock(&local_alloc_inode->i_mutex); iput(local_alloc_inode); + } mlog_exit(status); return status; @@ -524,13 +526,12 @@ bail: int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, handle_t *handle, struct ocfs2_alloc_context *ac, - u32 min_bits, + u32 bits_wanted, u32 *bit_off, u32 *num_bits) { int status, start; struct inode *local_alloc_inode; - u32 bits_wanted; void *bitmap; struct ocfs2_dinode *alloc; struct ocfs2_local_alloc *la; @@ -538,7 +539,6 @@ int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, mlog_entry_void(); BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL); - bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given; local_alloc_inode = ac->ac_inode; alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data; la = OCFS2_LOCAL_ALLOC(alloc); diff --git a/fs/ocfs2/localalloc.h b/fs/ocfs2/localalloc.h index 385a10152f9..3f76631e110 100644 --- a/fs/ocfs2/localalloc.h +++ b/fs/ocfs2/localalloc.h @@ -48,7 +48,7 @@ int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb, int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb, handle_t *handle, struct ocfs2_alloc_context *ac, - u32 min_bits, + u32 bits_wanted, u32 *bit_off, u32 *num_bits); diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c index d79aa12137d..98756156d29 100644 --- a/fs/ocfs2/mmap.c +++ b/fs/ocfs2/mmap.c @@ -60,31 +60,28 @@ static inline int ocfs2_vm_op_unblock_sigs(sigset_t *oldset) return sigprocmask(SIG_SETMASK, oldset, NULL); } -static struct page *ocfs2_nopage(struct vm_area_struct * area, - unsigned long address, - int *type) +static int ocfs2_fault(struct vm_area_struct *area, struct vm_fault *vmf) { - struct page *page = NOPAGE_SIGBUS; sigset_t blocked, oldset; - int ret; + int error, ret; - mlog_entry("(area=%p, address=%lu, type=%p)\n", area, address, - type); + mlog_entry("(area=%p, page offset=%lu)\n", area, vmf->pgoff); - ret = ocfs2_vm_op_block_sigs(&blocked, &oldset); - if (ret < 0) { - mlog_errno(ret); + error = ocfs2_vm_op_block_sigs(&blocked, &oldset); + if (error < 0) { + mlog_errno(error); + ret = VM_FAULT_SIGBUS; goto out; } - page = filemap_nopage(area, address, type); + ret = filemap_fault(area, vmf); - ret = ocfs2_vm_op_unblock_sigs(&oldset); - if (ret < 0) - mlog_errno(ret); + error = ocfs2_vm_op_unblock_sigs(&oldset); + if (error < 0) + mlog_errno(error); out: - mlog_exit_ptr(page); - return page; + mlog_exit_ptr(vmf->page); + return ret; } static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh, @@ -92,7 +89,7 @@ static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh, { int ret; struct address_space *mapping = inode->i_mapping; - loff_t pos = page->index << PAGE_CACHE_SHIFT; + loff_t pos = page_offset(page); unsigned int len = PAGE_CACHE_SIZE; pgoff_t last_index; struct page *locked_page = NULL; @@ -209,7 +206,7 @@ out: } static struct vm_operations_struct ocfs2_file_vm_ops = { - .nopage = ocfs2_nopage, + .fault = ocfs2_fault, .page_mkwrite = ocfs2_page_mkwrite, }; @@ -226,6 +223,7 @@ int ocfs2_mmap(struct file *file, struct vm_area_struct *vma) ocfs2_meta_unlock(file->f_dentry->d_inode, lock_level); out: vma->vm_ops = &ocfs2_file_vm_ops; + vma->vm_flags |= VM_CAN_NONLINEAR; return 0; } diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c index d430fdab16e..729259016c1 100644 --- a/fs/ocfs2/namei.c +++ b/fs/ocfs2/namei.c @@ -64,29 +64,6 @@ #include "buffer_head_io.h" -#define NAMEI_RA_CHUNKS 2 -#define NAMEI_RA_BLOCKS 4 -#define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) -#define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) - -static int inline ocfs2_search_dirblock(struct buffer_head *bh, - struct inode *dir, - const char *name, int namelen, - unsigned long offset, - struct ocfs2_dir_entry **res_dir); - -static int ocfs2_delete_entry(handle_t *handle, - struct inode *dir, - struct ocfs2_dir_entry *de_del, - struct buffer_head *bh); - -static int __ocfs2_add_entry(handle_t *handle, - struct inode *dir, - const char *name, int namelen, - struct inode *inode, u64 blkno, - struct buffer_head *parent_fe_bh, - struct buffer_head *insert_bh); - static int ocfs2_mknod_locked(struct ocfs2_super *osb, struct inode *dir, struct dentry *dentry, int mode, @@ -97,13 +74,6 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, struct inode **ret_inode, struct ocfs2_alloc_context *inode_ac); -static int ocfs2_fill_new_dir(struct ocfs2_super *osb, - handle_t *handle, - struct inode *parent, - struct inode *inode, - struct buffer_head *fe_bh, - struct ocfs2_alloc_context *data_ac); - static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb, struct inode **ret_orphan_dir, struct inode *inode, @@ -123,17 +93,6 @@ static int ocfs2_create_symlink_data(struct ocfs2_super *osb, struct inode *inode, const char *symname); -static inline int ocfs2_add_entry(handle_t *handle, - struct dentry *dentry, - struct inode *inode, u64 blkno, - struct buffer_head *parent_fe_bh, - struct buffer_head *insert_bh) -{ - return __ocfs2_add_entry(handle, dentry->d_parent->d_inode, - dentry->d_name.name, dentry->d_name.len, - inode, blkno, parent_fe_bh, insert_bh); -} - /* An orphan dir name is an 8 byte value, printed as a hex string */ #define OCFS2_ORPHAN_NAMELEN ((int)(2 * sizeof(u64))) @@ -142,10 +101,8 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, { int status; u64 blkno; - struct buffer_head *dirent_bh = NULL; struct inode *inode = NULL; struct dentry *ret; - struct ocfs2_dir_entry *dirent; struct ocfs2_inode_info *oi; mlog_entry("(0x%p, 0x%p, '%.*s')\n", dir, dentry, @@ -167,9 +124,8 @@ static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry, goto bail; } - status = ocfs2_find_files_on_disk(dentry->d_name.name, - dentry->d_name.len, &blkno, - dir, &dirent_bh, &dirent); + status = ocfs2_lookup_ino_from_name(dir, dentry->d_name.name, + dentry->d_name.len, &blkno); if (status < 0) goto bail_add; @@ -224,83 +180,12 @@ bail_unlock: ocfs2_meta_unlock(dir, 0); bail: - if (dirent_bh) - brelse(dirent_bh); mlog_exit_ptr(ret); return ret; } -static int ocfs2_fill_new_dir(struct ocfs2_super *osb, - handle_t *handle, - struct inode *parent, - struct inode *inode, - struct buffer_head *fe_bh, - struct ocfs2_alloc_context *data_ac) -{ - int status; - struct buffer_head *new_bh = NULL; - struct ocfs2_dir_entry *de = NULL; - - mlog_entry_void(); - - status = ocfs2_do_extend_dir(osb->sb, handle, inode, fe_bh, - data_ac, NULL, &new_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - - ocfs2_set_new_buffer_uptodate(inode, new_bh); - - status = ocfs2_journal_access(handle, inode, new_bh, - OCFS2_JOURNAL_ACCESS_CREATE); - if (status < 0) { - mlog_errno(status); - goto bail; - } - memset(new_bh->b_data, 0, osb->sb->s_blocksize); - - de = (struct ocfs2_dir_entry *) new_bh->b_data; - de->inode = cpu_to_le64(OCFS2_I(inode)->ip_blkno); - de->name_len = 1; - de->rec_len = - cpu_to_le16(OCFS2_DIR_REC_LEN(de->name_len)); - strcpy(de->name, "."); - ocfs2_set_de_type(de, S_IFDIR); - de = (struct ocfs2_dir_entry *) ((char *)de + le16_to_cpu(de->rec_len)); - de->inode = cpu_to_le64(OCFS2_I(parent)->ip_blkno); - de->rec_len = cpu_to_le16(inode->i_sb->s_blocksize - - OCFS2_DIR_REC_LEN(1)); - de->name_len = 2; - strcpy(de->name, ".."); - ocfs2_set_de_type(de, S_IFDIR); - - status = ocfs2_journal_dirty(handle, new_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - - i_size_write(inode, inode->i_sb->s_blocksize); - inode->i_nlink = 2; - inode->i_blocks = ocfs2_inode_sector_count(inode); - status = ocfs2_mark_inode_dirty(handle, inode, fe_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } - - status = 0; -bail: - if (new_bh) - brelse(new_bh); - - mlog_exit(status); - return status; -} - static int ocfs2_mknod(struct inode *dir, struct dentry *dentry, int mode, @@ -365,9 +250,8 @@ static int ocfs2_mknod(struct inode *dir, goto leave; } - /* are we making a directory? If so, reserve a cluster for his - * 1st extent. */ - if (S_ISDIR(mode)) { + /* Reserve a cluster if creating an extent based directory. */ + if (S_ISDIR(mode) && !ocfs2_supports_inline_data(osb)) { status = ocfs2_reserve_clusters(osb, 1, &data_ac); if (status < 0) { if (status != -ENOSPC) @@ -564,10 +448,21 @@ static int ocfs2_mknod_locked(struct ocfs2_super *osb, cpu_to_le32(CURRENT_TIME.tv_nsec); fe->i_dtime = 0; - fel = &fe->id2.i_list; - fel->l_tree_depth = 0; - fel->l_next_free_rec = 0; - fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb)); + /* + * If supported, directories start with inline data. + */ + if (S_ISDIR(mode) && ocfs2_supports_inline_data(osb)) { + u16 feat = le16_to_cpu(fe->i_dyn_features); + + fe->i_dyn_features = cpu_to_le16(feat | OCFS2_INLINE_DATA_FL); + + fe->id2.i_data.id_count = cpu_to_le16(ocfs2_max_inline_data(osb->sb)); + } else { + fel = &fe->id2.i_list; + fel->l_tree_depth = 0; + fel->l_next_free_rec = 0; + fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb)); + } status = ocfs2_journal_dirty(handle, *new_fe_bh); if (status < 0) { @@ -1048,11 +943,6 @@ static void ocfs2_double_unlock(struct inode *inode1, struct inode *inode2) ocfs2_meta_unlock(inode2, 1); } -#define PARENT_INO(buffer) \ - ((struct ocfs2_dir_entry *) \ - ((char *)buffer + \ - le16_to_cpu(((struct ocfs2_dir_entry *)buffer)->rec_len)))->inode - static int ocfs2_rename(struct inode *old_dir, struct dentry *old_dentry, struct inode *new_dir, @@ -1070,16 +960,17 @@ static int ocfs2_rename(struct inode *old_dir, struct buffer_head *old_inode_bh = NULL; struct buffer_head *insert_entry_bh = NULL; struct ocfs2_super *osb = NULL; - u64 newfe_blkno; + u64 newfe_blkno, old_de_ino; handle_t *handle = NULL; struct buffer_head *old_dir_bh = NULL; struct buffer_head *new_dir_bh = NULL; - struct ocfs2_dir_entry *old_de = NULL, *new_de = NULL; // dirent for old_dentry - // and new_dentry + struct ocfs2_dir_entry *old_inode_dot_dot_de = NULL, *old_de = NULL, + *new_de = NULL; struct buffer_head *new_de_bh = NULL, *old_de_bh = NULL; // bhs for above struct buffer_head *old_inode_de_bh = NULL; // if old_dentry is a dir, // this is the 1st dirent bh nlink_t old_dir_nlink = old_dir->i_nlink; + struct ocfs2_dinode *old_di; /* At some point it might be nice to break this function up a * bit. */ @@ -1158,27 +1049,35 @@ static int ocfs2_rename(struct inode *old_dir, } if (S_ISDIR(old_inode->i_mode)) { - status = -EIO; - old_inode_de_bh = ocfs2_bread(old_inode, 0, &status, 0); - if (!old_inode_de_bh) + u64 old_inode_parent; + + status = ocfs2_find_files_on_disk("..", 2, &old_inode_parent, + old_inode, &old_inode_de_bh, + &old_inode_dot_dot_de); + if (status) { + status = -EIO; goto bail; + } - status = -EIO; - if (le64_to_cpu(PARENT_INO(old_inode_de_bh->b_data)) != - OCFS2_I(old_dir)->ip_blkno) + if (old_inode_parent != OCFS2_I(old_dir)->ip_blkno) { + status = -EIO; goto bail; - status = -EMLINK; - if (!new_inode && new_dir!=old_dir && - new_dir->i_nlink >= OCFS2_LINK_MAX) + } + + if (!new_inode && new_dir != old_dir && + new_dir->i_nlink >= OCFS2_LINK_MAX) { + status = -EMLINK; goto bail; + } } - status = -ENOENT; - old_de_bh = ocfs2_find_entry(old_dentry->d_name.name, - old_dentry->d_name.len, - old_dir, &old_de); - if (!old_de_bh) + status = ocfs2_lookup_ino_from_name(old_dir, old_dentry->d_name.name, + old_dentry->d_name.len, + &old_de_ino); + if (status) { + status = -ENOENT; goto bail; + } /* * Check for inode number is _not_ due to possible IO errors. @@ -1186,8 +1085,10 @@ static int ocfs2_rename(struct inode *old_dir, * and merrily kill the link to whatever was created under the * same name. Goodbye sticky bit ;-< */ - if (le64_to_cpu(old_de->inode) != OCFS2_I(old_inode)->ip_blkno) + if (old_de_ino != OCFS2_I(old_inode)->ip_blkno) { + status = -ENOENT; goto bail; + } /* check if the target already exists (in which case we need * to delete it */ @@ -1320,20 +1221,13 @@ static int ocfs2_rename(struct inode *old_dir, } /* change the dirent to point to the correct inode */ - status = ocfs2_journal_access(handle, new_dir, new_de_bh, - OCFS2_JOURNAL_ACCESS_WRITE); + status = ocfs2_update_entry(new_dir, handle, new_de_bh, + new_de, old_inode); if (status < 0) { mlog_errno(status); goto bail; } - new_de->inode = cpu_to_le64(OCFS2_I(old_inode)->ip_blkno); - new_de->file_type = old_de->file_type; new_dir->i_version++; - status = ocfs2_journal_dirty(handle, new_de_bh); - if (status < 0) { - mlog_errno(status); - goto bail; - } if (S_ISDIR(new_inode->i_mode)) newfe->i_links_count = 0; @@ -1354,9 +1248,36 @@ static int ocfs2_rename(struct inode *old_dir, old_inode->i_ctime = CURRENT_TIME; mark_inode_dirty(old_inode); - ocfs2_mark_inode_dirty(handle, old_inode, old_inode_bh); - /* now that the name has been added to new_dir, remove the old name */ + status = ocfs2_journal_access(handle, old_inode, old_inode_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (status >= 0) { + old_di = (struct ocfs2_dinode *) old_inode_bh->b_data; + + old_di->i_ctime = cpu_to_le64(old_inode->i_ctime.tv_sec); + old_di->i_ctime_nsec = cpu_to_le32(old_inode->i_ctime.tv_nsec); + + status = ocfs2_journal_dirty(handle, old_inode_bh); + if (status < 0) + mlog_errno(status); + } else + mlog_errno(status); + + /* + * Now that the name has been added to new_dir, remove the old name. + * + * We don't keep any directory entry context around until now + * because the insert might have changed the type of directory + * we're dealing with. + */ + old_de_bh = ocfs2_find_entry(old_dentry->d_name.name, + old_dentry->d_name.len, + old_dir, &old_de); + if (!old_de_bh) { + status = -EIO; + goto bail; + } + status = ocfs2_delete_entry(handle, old_dir, old_de, old_de_bh); if (status < 0) { mlog_errno(status); @@ -1369,12 +1290,8 @@ static int ocfs2_rename(struct inode *old_dir, } old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME; if (old_inode_de_bh) { - status = ocfs2_journal_access(handle, old_inode, - old_inode_de_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - PARENT_INO(old_inode_de_bh->b_data) = - cpu_to_le64(OCFS2_I(new_dir)->ip_blkno); - status = ocfs2_journal_dirty(handle, old_inode_de_bh); + status = ocfs2_update_entry(old_inode, handle, old_inode_de_bh, + old_inode_dot_dot_de, new_dir); old_dir->i_nlink--; if (new_inode) { new_inode->i_nlink--; @@ -1753,329 +1670,6 @@ bail: return status; } -int ocfs2_check_dir_entry(struct inode * dir, - struct ocfs2_dir_entry * de, - struct buffer_head * bh, - unsigned long offset) -{ - const char *error_msg = NULL; - const int rlen = le16_to_cpu(de->rec_len); - - if (rlen < OCFS2_DIR_REC_LEN(1)) - error_msg = "rec_len is smaller than minimal"; - else if (rlen % 4 != 0) - error_msg = "rec_len % 4 != 0"; - else if (rlen < OCFS2_DIR_REC_LEN(de->name_len)) - error_msg = "rec_len is too small for name_len"; - else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize) - error_msg = "directory entry across blocks"; - - if (error_msg != NULL) - mlog(ML_ERROR, "bad entry in directory #%llu: %s - " - "offset=%lu, inode=%llu, rec_len=%d, name_len=%d\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, error_msg, - offset, (unsigned long long)le64_to_cpu(de->inode), rlen, - de->name_len); - return error_msg == NULL ? 1 : 0; -} - -/* we don't always have a dentry for what we want to add, so people - * like orphan dir can call this instead. - * - * If you pass me insert_bh, I'll skip the search of the other dir - * blocks and put the record in there. - */ -static int __ocfs2_add_entry(handle_t *handle, - struct inode *dir, - const char *name, int namelen, - struct inode *inode, u64 blkno, - struct buffer_head *parent_fe_bh, - struct buffer_head *insert_bh) -{ - unsigned long offset; - unsigned short rec_len; - struct ocfs2_dir_entry *de, *de1; - struct super_block *sb; - int retval, status; - - mlog_entry_void(); - - sb = dir->i_sb; - - if (!namelen) - return -EINVAL; - - rec_len = OCFS2_DIR_REC_LEN(namelen); - offset = 0; - de = (struct ocfs2_dir_entry *) insert_bh->b_data; - while (1) { - BUG_ON((char *)de >= sb->s_blocksize + insert_bh->b_data); - /* These checks should've already been passed by the - * prepare function, but I guess we can leave them - * here anyway. */ - if (!ocfs2_check_dir_entry(dir, de, insert_bh, offset)) { - retval = -ENOENT; - goto bail; - } - if (ocfs2_match(namelen, name, de)) { - retval = -EEXIST; - goto bail; - } - if (((le64_to_cpu(de->inode) == 0) && - (le16_to_cpu(de->rec_len) >= rec_len)) || - (le16_to_cpu(de->rec_len) >= - (OCFS2_DIR_REC_LEN(de->name_len) + rec_len))) { - dir->i_mtime = dir->i_ctime = CURRENT_TIME; - retval = ocfs2_mark_inode_dirty(handle, dir, parent_fe_bh); - if (retval < 0) { - mlog_errno(retval); - goto bail; - } - - status = ocfs2_journal_access(handle, dir, insert_bh, - OCFS2_JOURNAL_ACCESS_WRITE); - /* By now the buffer is marked for journaling */ - offset += le16_to_cpu(de->rec_len); - if (le64_to_cpu(de->inode)) { - de1 = (struct ocfs2_dir_entry *)((char *) de + - OCFS2_DIR_REC_LEN(de->name_len)); - de1->rec_len = - cpu_to_le16(le16_to_cpu(de->rec_len) - - OCFS2_DIR_REC_LEN(de->name_len)); - de->rec_len = cpu_to_le16(OCFS2_DIR_REC_LEN(de->name_len)); - de = de1; - } - de->file_type = OCFS2_FT_UNKNOWN; - if (blkno) { - de->inode = cpu_to_le64(blkno); - ocfs2_set_de_type(de, inode->i_mode); - } else - de->inode = 0; - de->name_len = namelen; - memcpy(de->name, name, namelen); - - dir->i_version++; - status = ocfs2_journal_dirty(handle, insert_bh); - retval = 0; - goto bail; - } - offset += le16_to_cpu(de->rec_len); - de = (struct ocfs2_dir_entry *) ((char *) de + le16_to_cpu(de->rec_len)); - } - - /* when you think about it, the assert above should prevent us - * from ever getting here. */ - retval = -ENOSPC; -bail: - - mlog_exit(retval); - return retval; -} - - -/* - * ocfs2_delete_entry deletes a directory entry by merging it with the - * previous entry - */ -static int ocfs2_delete_entry(handle_t *handle, - struct inode *dir, - struct ocfs2_dir_entry *de_del, - struct buffer_head *bh) -{ - struct ocfs2_dir_entry *de, *pde; - int i, status = -ENOENT; - - mlog_entry("(0x%p, 0x%p, 0x%p, 0x%p)\n", handle, dir, de_del, bh); - - i = 0; - pde = NULL; - de = (struct ocfs2_dir_entry *) bh->b_data; - while (i < bh->b_size) { - if (!ocfs2_check_dir_entry(dir, de, bh, i)) { - status = -EIO; - mlog_errno(status); - goto bail; - } - if (de == de_del) { - status = ocfs2_journal_access(handle, dir, bh, - OCFS2_JOURNAL_ACCESS_WRITE); - if (status < 0) { - status = -EIO; - mlog_errno(status); - goto bail; - } - if (pde) - pde->rec_len = - cpu_to_le16(le16_to_cpu(pde->rec_len) + - le16_to_cpu(de->rec_len)); - else - de->inode = 0; - dir->i_version++; - status = ocfs2_journal_dirty(handle, bh); - goto bail; - } - i += le16_to_cpu(de->rec_len); - pde = de; - de = (struct ocfs2_dir_entry *)((char *)de + le16_to_cpu(de->rec_len)); - } -bail: - mlog_exit(status); - return status; -} - -/* - * Returns 0 if not found, -1 on failure, and 1 on success - */ -static int inline ocfs2_search_dirblock(struct buffer_head *bh, - struct inode *dir, - const char *name, int namelen, - unsigned long offset, - struct ocfs2_dir_entry **res_dir) -{ - struct ocfs2_dir_entry *de; - char *dlimit, *de_buf; - int de_len; - int ret = 0; - - mlog_entry_void(); - - de_buf = bh->b_data; - dlimit = de_buf + dir->i_sb->s_blocksize; - - while (de_buf < dlimit) { - /* this code is executed quadratically often */ - /* do minimal checking `by hand' */ - - de = (struct ocfs2_dir_entry *) de_buf; - - if (de_buf + namelen <= dlimit && - ocfs2_match(namelen, name, de)) { - /* found a match - just to be sure, do a full check */ - if (!ocfs2_check_dir_entry(dir, de, bh, offset)) { - ret = -1; - goto bail; - } - *res_dir = de; - ret = 1; - goto bail; - } - - /* prevent looping on a bad block */ - de_len = le16_to_cpu(de->rec_len); - if (de_len <= 0) { - ret = -1; - goto bail; - } - - de_buf += de_len; - offset += de_len; - } - -bail: - mlog_exit(ret); - return ret; -} - -struct buffer_head *ocfs2_find_entry(const char *name, int namelen, - struct inode *dir, - struct ocfs2_dir_entry **res_dir) -{ - struct super_block *sb; - struct buffer_head *bh_use[NAMEI_RA_SIZE]; - struct buffer_head *bh, *ret = NULL; - unsigned long start, block, b; - int ra_max = 0; /* Number of bh's in the readahead - buffer, bh_use[] */ - int ra_ptr = 0; /* Current index into readahead - buffer */ - int num = 0; - int nblocks, i, err; - - mlog_entry_void(); - - *res_dir = NULL; - sb = dir->i_sb; - - nblocks = i_size_read(dir) >> sb->s_blocksize_bits; - start = OCFS2_I(dir)->ip_dir_start_lookup; - if (start >= nblocks) - start = 0; - block = start; - -restart: - do { - /* - * We deal with the read-ahead logic here. - */ - if (ra_ptr >= ra_max) { - /* Refill the readahead buffer */ - ra_ptr = 0; - b = block; - for (ra_max = 0; ra_max < NAMEI_RA_SIZE; ra_max++) { - /* - * Terminate if we reach the end of the - * directory and must wrap, or if our - * search has finished at this block. - */ - if (b >= nblocks || (num && block == start)) { - bh_use[ra_max] = NULL; - break; - } - num++; - - bh = ocfs2_bread(dir, b++, &err, 1); - bh_use[ra_max] = bh; - } - } - if ((bh = bh_use[ra_ptr++]) == NULL) - goto next; - wait_on_buffer(bh); - if (!buffer_uptodate(bh)) { - /* read error, skip block & hope for the best */ - ocfs2_error(dir->i_sb, "reading directory %llu, " - "offset %lu\n", - (unsigned long long)OCFS2_I(dir)->ip_blkno, - block); - brelse(bh); - goto next; - } - i = ocfs2_search_dirblock(bh, dir, name, namelen, - block << sb->s_blocksize_bits, - res_dir); - if (i == 1) { - OCFS2_I(dir)->ip_dir_start_lookup = block; - ret = bh; - goto cleanup_and_exit; - } else { - brelse(bh); - if (i < 0) - goto cleanup_and_exit; - } - next: - if (++block >= nblocks) - block = 0; - } while (block != start); - - /* - * If the directory has grown while we were searching, then - * search the last part of the directory before giving up. - */ - block = nblocks; - nblocks = i_size_read(dir) >> sb->s_blocksize_bits; - if (block < nblocks) { - start = 0; - goto restart; - } - -cleanup_and_exit: - /* Clean up the read-ahead blocks */ - for (; ra_ptr < ra_max; ra_ptr++) - brelse(bh_use[ra_ptr]); - - mlog_exit_ptr(ret); - return ret; -} - static int ocfs2_blkno_stringify(u64 blkno, char *name) { int status, namelen; diff --git a/fs/ocfs2/namei.h b/fs/ocfs2/namei.h index 0975c7b7212..688aef64c87 100644 --- a/fs/ocfs2/namei.h +++ b/fs/ocfs2/namei.h @@ -30,29 +30,10 @@ extern const struct inode_operations ocfs2_dir_iops; struct dentry *ocfs2_get_parent(struct dentry *child); -int ocfs2_check_dir_entry (struct inode *dir, - struct ocfs2_dir_entry *de, - struct buffer_head *bh, - unsigned long offset); -struct buffer_head *ocfs2_find_entry(const char *name, - int namelen, - struct inode *dir, - struct ocfs2_dir_entry **res_dir); int ocfs2_orphan_del(struct ocfs2_super *osb, handle_t *handle, struct inode *orphan_dir_inode, struct inode *inode, struct buffer_head *orphan_dir_bh); -static inline int ocfs2_match(int len, - const char * const name, - struct ocfs2_dir_entry *de) -{ - if (len != de->name_len) - return 0; - if (!de->inode) - return 0; - return !memcmp(name, de->name, len); -} - #endif /* OCFS2_NAMEI_H */ diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h index 5cc90a40b3c..60a23e1906b 100644 --- a/fs/ocfs2/ocfs2.h +++ b/fs/ocfs2/ocfs2.h @@ -319,6 +319,13 @@ static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb) return 0; } +static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb) +{ + if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA) + return 1; + return 0; +} + /* set / clear functions because cluster events can make these happen * in parallel so we want the transitions to be atomic. this also * means that any future flags osb_flags must be protected by spinlock @@ -494,16 +501,16 @@ static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb, /* * Find the 1st page index which covers the given clusters. */ -static inline unsigned long ocfs2_align_clusters_to_page_index(struct super_block *sb, +static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb, u32 clusters) { unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; - unsigned long index = clusters; + pgoff_t index = clusters; if (PAGE_CACHE_SHIFT > cbits) { - index = clusters >> (PAGE_CACHE_SHIFT - cbits); + index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits); } else if (PAGE_CACHE_SHIFT < cbits) { - index = clusters << (cbits - PAGE_CACHE_SHIFT); + index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT); } return index; diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h index 82f8a75b207..6ef876759a7 100644 --- a/fs/ocfs2/ocfs2_fs.h +++ b/fs/ocfs2/ocfs2_fs.h @@ -87,7 +87,8 @@ #define OCFS2_FEATURE_COMPAT_SUPP OCFS2_FEATURE_COMPAT_BACKUP_SB #define OCFS2_FEATURE_INCOMPAT_SUPP (OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT \ - | OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC) + | OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC \ + | OCFS2_FEATURE_INCOMPAT_INLINE_DATA) #define OCFS2_FEATURE_RO_COMPAT_SUPP OCFS2_FEATURE_RO_COMPAT_UNWRITTEN /* @@ -111,6 +112,20 @@ #define OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC 0x0010 /* + * Tunefs sets this incompat flag before starting an operation which + * would require cleanup on abort. This is done to protect users from + * inadvertently mounting the fs after an aborted run without + * fsck-ing. + * + * s_tunefs_flags on the super block describes precisely which + * operations were in progress. + */ +#define OCFS2_FEATURE_INCOMPAT_TUNEFS_INPROG 0x0020 + +/* Support for data packed into inode blocks */ +#define OCFS2_FEATURE_INCOMPAT_INLINE_DATA 0x0040 + +/* * backup superblock flag is used to indicate that this volume * has backup superblocks. */ @@ -130,6 +145,11 @@ #define OCFS2_MAX_BACKUP_SUPERBLOCKS 6 /* + * Flags on ocfs2_super_block.s_tunefs_flags + */ +#define OCFS2_TUNEFS_INPROG_REMOVE_SLOT 0x0001 /* Removing slots */ + +/* * Flags on ocfs2_dinode.i_flags */ #define OCFS2_VALID_FL (0x00000001) /* Inode is valid */ @@ -146,6 +166,17 @@ #define OCFS2_CHAIN_FL (0x00000400) /* Chain allocator */ #define OCFS2_DEALLOC_FL (0x00000800) /* Truncate log */ +/* + * Flags on ocfs2_dinode.i_dyn_features + * + * These can change much more often than i_flags. When adding flags, + * keep in mind that i_dyn_features is only 16 bits wide. + */ +#define OCFS2_INLINE_DATA_FL (0x0001) /* Data stored in inode block */ +#define OCFS2_HAS_XATTR_FL (0x0002) +#define OCFS2_INLINE_XATTR_FL (0x0004) +#define OCFS2_INDEXED_DIR_FL (0x0008) + /* Inode attributes, keep in sync with EXT2 */ #define OCFS2_SECRM_FL (0x00000001) /* Secure deletion */ #define OCFS2_UNRM_FL (0x00000002) /* Undelete */ @@ -447,8 +478,8 @@ struct ocfs2_super_block { __le32 s_clustersize_bits; /* Clustersize for this fs */ /*40*/ __le16 s_max_slots; /* Max number of simultaneous mounts before tunefs required */ - __le16 s_reserved1; - __le32 s_reserved2; + __le16 s_tunefs_flag; + __le32 s_reserved1; __le64 s_first_cluster_group; /* Block offset of 1st cluster * group header */ /*50*/ __u8 s_label[OCFS2_MAX_VOL_LABEL_LEN]; /* Label for mounting, etc. */ @@ -471,6 +502,19 @@ struct ocfs2_local_alloc }; /* + * Data-in-inode header. This is only used if i_dyn_features has + * OCFS2_INLINE_DATA_FL set. + */ +struct ocfs2_inline_data +{ +/*00*/ __le16 id_count; /* Number of bytes that can be used + * for data, starting at id_data */ + __le16 id_reserved0; + __le32 id_reserved1; + __u8 id_data[0]; /* Start of user data */ +}; + +/* * On disk inode for OCFS2 */ struct ocfs2_dinode { @@ -502,7 +546,7 @@ struct ocfs2_dinode { __le32 i_attr; __le16 i_orphaned_slot; /* Only valid when OCFS2_ORPHANED_FL was set in i_flags */ - __le16 i_reserved1; + __le16 i_dyn_features; /*70*/ __le64 i_reserved2[8]; /*B8*/ union { __le64 i_pad1; /* Generic way to refer to this @@ -528,6 +572,7 @@ struct ocfs2_dinode { struct ocfs2_chain_list i_chain; struct ocfs2_extent_list i_list; struct ocfs2_truncate_log i_dealloc; + struct ocfs2_inline_data i_data; __u8 i_symlink[0]; } id2; /* Actual on-disk size is one block */ @@ -577,6 +622,12 @@ static inline int ocfs2_fast_symlink_chars(struct super_block *sb) offsetof(struct ocfs2_dinode, id2.i_symlink); } +static inline int ocfs2_max_inline_data(struct super_block *sb) +{ + return sb->s_blocksize - + offsetof(struct ocfs2_dinode, id2.i_data.id_data); +} + static inline int ocfs2_extent_recs_per_inode(struct super_block *sb) { int size; @@ -656,6 +707,11 @@ static inline int ocfs2_fast_symlink_chars(int blocksize) return blocksize - offsetof(struct ocfs2_dinode, id2.i_symlink); } +static inline int ocfs2_max_inline_data(int blocksize) +{ + return blocksize - offsetof(struct ocfs2_dinode, id2.i_data.id_data); +} + static inline int ocfs2_extent_recs_per_inode(int blocksize) { int size; diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c index d9c5c9fcb30..8f09f5235e3 100644 --- a/fs/ocfs2/suballoc.c +++ b/fs/ocfs2/suballoc.c @@ -1486,21 +1486,21 @@ static inline void ocfs2_block_to_cluster_group(struct inode *inode, * contig. allocation, set to '1' to indicate we can deal with extents * of any size. */ -int ocfs2_claim_clusters(struct ocfs2_super *osb, - handle_t *handle, - struct ocfs2_alloc_context *ac, - u32 min_clusters, - u32 *cluster_start, - u32 *num_clusters) +int __ocfs2_claim_clusters(struct ocfs2_super *osb, + handle_t *handle, + struct ocfs2_alloc_context *ac, + u32 min_clusters, + u32 max_clusters, + u32 *cluster_start, + u32 *num_clusters) { int status; - unsigned int bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given; + unsigned int bits_wanted = max_clusters; u64 bg_blkno = 0; u16 bg_bit_off; mlog_entry_void(); - BUG_ON(!ac); BUG_ON(ac->ac_bits_given >= ac->ac_bits_wanted); BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL @@ -1557,6 +1557,19 @@ bail: return status; } +int ocfs2_claim_clusters(struct ocfs2_super *osb, + handle_t *handle, + struct ocfs2_alloc_context *ac, + u32 min_clusters, + u32 *cluster_start, + u32 *num_clusters) +{ + unsigned int bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given; + + return __ocfs2_claim_clusters(osb, handle, ac, min_clusters, + bits_wanted, cluster_start, num_clusters); +} + static inline int ocfs2_block_group_clear_bits(handle_t *handle, struct inode *alloc_inode, struct ocfs2_group_desc *bg, diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h index f212dc01a84..cafe9370309 100644 --- a/fs/ocfs2/suballoc.h +++ b/fs/ocfs2/suballoc.h @@ -85,6 +85,17 @@ int ocfs2_claim_clusters(struct ocfs2_super *osb, u32 min_clusters, u32 *cluster_start, u32 *num_clusters); +/* + * Use this variant of ocfs2_claim_clusters to specify a maxiumum + * number of clusters smaller than the allocation reserved. + */ +int __ocfs2_claim_clusters(struct ocfs2_super *osb, + handle_t *handle, + struct ocfs2_alloc_context *ac, + u32 min_clusters, + u32 max_clusters, + u32 *cluster_start, + u32 *num_clusters); int ocfs2_free_suballoc_bits(handle_t *handle, struct inode *alloc_inode, diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c index 3a5a1ed09ac..0e2a1b45bf9 100644 --- a/fs/ocfs2/super.c +++ b/fs/ocfs2/super.c @@ -39,6 +39,7 @@ #include <linux/parser.h> #include <linux/crc32.h> #include <linux/debugfs.h> +#include <linux/mount.h> #include <cluster/nodemanager.h> @@ -81,9 +82,17 @@ static struct dentry *ocfs2_debugfs_root = NULL; MODULE_AUTHOR("Oracle"); MODULE_LICENSE("GPL"); +struct mount_options +{ + unsigned long mount_opt; + unsigned int atime_quantum; + signed short slot; +}; + static int ocfs2_parse_options(struct super_block *sb, char *options, - unsigned long *mount_opt, s16 *slot, + struct mount_options *mopt, int is_remount); +static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt); static void ocfs2_put_super(struct super_block *sb); static int ocfs2_mount_volume(struct super_block *sb); static int ocfs2_remount(struct super_block *sb, int *flags, char *data); @@ -98,7 +107,7 @@ static int ocfs2_sync_fs(struct super_block *sb, int wait); static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb); static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb); -static int ocfs2_release_system_inodes(struct ocfs2_super *osb); +static void ocfs2_release_system_inodes(struct ocfs2_super *osb); static int ocfs2_fill_local_node_info(struct ocfs2_super *osb); static int ocfs2_check_volume(struct ocfs2_super *osb); static int ocfs2_verify_volume(struct ocfs2_dinode *di, @@ -126,6 +135,7 @@ static const struct super_operations ocfs2_sops = { .write_super = ocfs2_write_super, .put_super = ocfs2_put_super, .remount_fs = ocfs2_remount, + .show_options = ocfs2_show_options, }; enum { @@ -170,7 +180,7 @@ static void ocfs2_write_super(struct super_block *sb) static int ocfs2_sync_fs(struct super_block *sb, int wait) { - int status = 0; + int status; tid_t target; struct ocfs2_super *osb = OCFS2_SB(sb); @@ -268,9 +278,9 @@ bail: return status; } -static int ocfs2_release_system_inodes(struct ocfs2_super *osb) +static void ocfs2_release_system_inodes(struct ocfs2_super *osb) { - int status = 0, i; + int i; struct inode *inode; mlog_entry_void(); @@ -295,8 +305,7 @@ static int ocfs2_release_system_inodes(struct ocfs2_super *osb) osb->root_inode = NULL; } - mlog_exit(status); - return status; + mlog_exit(0); } /* We're allocating fs objects, use GFP_NOFS */ @@ -316,63 +325,74 @@ static void ocfs2_destroy_inode(struct inode *inode) kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); } -/* From xfs_super.c:xfs_max_file_offset - * Copyright (c) 2000-2004 Silicon Graphics, Inc. - */ -unsigned long long ocfs2_max_file_offset(unsigned int blockshift) +static unsigned long long ocfs2_max_file_offset(unsigned int bbits, + unsigned int cbits) { - unsigned int pagefactor = 1; - unsigned int bitshift = BITS_PER_LONG - 1; - - /* Figure out maximum filesize, on Linux this can depend on - * the filesystem blocksize (on 32 bit platforms). - * __block_prepare_write does this in an [unsigned] long... - * page->index << (PAGE_CACHE_SHIFT - bbits) - * So, for page sized blocks (4K on 32 bit platforms), - * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is - * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) - * but for smaller blocksizes it is less (bbits = log2 bsize). - * Note1: get_block_t takes a long (implicit cast from above) - * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch - * can optionally convert the [unsigned] long from above into - * an [unsigned] long long. + unsigned int bytes = 1 << cbits; + unsigned int trim = bytes; + unsigned int bitshift = 32; + + /* + * i_size and all block offsets in ocfs2 are always 64 bits + * wide. i_clusters is 32 bits, in cluster-sized units. So on + * 64 bit platforms, cluster size will be the limiting factor. */ #if BITS_PER_LONG == 32 # if defined(CONFIG_LBD) BUILD_BUG_ON(sizeof(sector_t) != 8); - pagefactor = PAGE_CACHE_SIZE; - bitshift = BITS_PER_LONG; + /* + * We might be limited by page cache size. + */ + if (bytes > PAGE_CACHE_SIZE) { + bytes = PAGE_CACHE_SIZE; + trim = 1; + /* + * Shift by 31 here so that we don't get larger than + * MAX_LFS_FILESIZE + */ + bitshift = 31; + } # else - pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); + /* + * We are limited by the size of sector_t. Use block size, as + * that's what we expose to the VFS. + */ + bytes = 1 << bbits; + trim = 1; + bitshift = 31; # endif #endif - return (((unsigned long long)pagefactor) << bitshift) - 1; + /* + * Trim by a whole cluster when we can actually approach the + * on-disk limits. Otherwise we can overflow i_clusters when + * an extent start is at the max offset. + */ + return (((unsigned long long)bytes) << bitshift) - trim; } static int ocfs2_remount(struct super_block *sb, int *flags, char *data) { int incompat_features; int ret = 0; - unsigned long parsed_options; - s16 slot; + struct mount_options parsed_options; struct ocfs2_super *osb = OCFS2_SB(sb); - if (!ocfs2_parse_options(sb, data, &parsed_options, &slot, 1)) { + if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) { ret = -EINVAL; goto out; } if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) != - (parsed_options & OCFS2_MOUNT_HB_LOCAL)) { + (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) { ret = -EINVAL; mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); goto out; } if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) != - (parsed_options & OCFS2_MOUNT_DATA_WRITEBACK)) { + (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) { ret = -EINVAL; mlog(ML_ERROR, "Cannot change data mode on remount\n"); goto out; @@ -423,7 +443,9 @@ unlock_osb: /* Only save off the new mount options in case of a successful * remount. */ - osb->s_mount_opt = parsed_options; + osb->s_mount_opt = parsed_options.mount_opt; + osb->s_atime_quantum = parsed_options.atime_quantum; + osb->preferred_slot = parsed_options.slot; } out: return ret; @@ -433,7 +455,7 @@ static int ocfs2_sb_probe(struct super_block *sb, struct buffer_head **bh, int *sector_size) { - int status = 0, tmpstat; + int status, tmpstat; struct ocfs1_vol_disk_hdr *hdr; struct ocfs2_dinode *di; int blksize; @@ -535,8 +557,7 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) { struct dentry *root; int status, sector_size; - unsigned long parsed_opt; - s16 slot; + struct mount_options parsed_options; struct inode *inode = NULL; struct ocfs2_super *osb = NULL; struct buffer_head *bh = NULL; @@ -544,14 +565,14 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) mlog_entry("%p, %p, %i", sb, data, silent); - if (!ocfs2_parse_options(sb, data, &parsed_opt, &slot, 0)) { + if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { status = -EINVAL; goto read_super_error; } /* for now we only have one cluster/node, make sure we see it * in the heartbeat universe */ - if (parsed_opt & OCFS2_MOUNT_HB_LOCAL) { + if (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL) { if (!o2hb_check_local_node_heartbeating()) { status = -EINVAL; goto read_super_error; @@ -573,8 +594,9 @@ static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) } brelse(bh); bh = NULL; - osb->s_mount_opt = parsed_opt; - osb->preferred_slot = slot; + osb->s_mount_opt = parsed_options.mount_opt; + osb->s_atime_quantum = parsed_options.atime_quantum; + osb->preferred_slot = parsed_options.slot; sb->s_magic = OCFS2_SUPER_MAGIC; @@ -716,8 +738,7 @@ static struct file_system_type ocfs2_fs_type = { static int ocfs2_parse_options(struct super_block *sb, char *options, - unsigned long *mount_opt, - s16 *slot, + struct mount_options *mopt, int is_remount) { int status; @@ -726,8 +747,9 @@ static int ocfs2_parse_options(struct super_block *sb, mlog_entry("remount: %d, options: \"%s\"\n", is_remount, options ? options : "(none)"); - *mount_opt = 0; - *slot = OCFS2_INVALID_SLOT; + mopt->mount_opt = 0; + mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; + mopt->slot = OCFS2_INVALID_SLOT; if (!options) { status = 1; @@ -737,7 +759,6 @@ static int ocfs2_parse_options(struct super_block *sb, while ((p = strsep(&options, ",")) != NULL) { int token, option; substring_t args[MAX_OPT_ARGS]; - struct ocfs2_super * osb = OCFS2_SB(sb); if (!*p) continue; @@ -745,10 +766,10 @@ static int ocfs2_parse_options(struct super_block *sb, token = match_token(p, tokens, args); switch (token) { case Opt_hb_local: - *mount_opt |= OCFS2_MOUNT_HB_LOCAL; + mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; break; case Opt_hb_none: - *mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; + mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; break; case Opt_barrier: if (match_int(&args[0], &option)) { @@ -756,27 +777,27 @@ static int ocfs2_parse_options(struct super_block *sb, goto bail; } if (option) - *mount_opt |= OCFS2_MOUNT_BARRIER; + mopt->mount_opt |= OCFS2_MOUNT_BARRIER; else - *mount_opt &= ~OCFS2_MOUNT_BARRIER; + mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; break; case Opt_intr: - *mount_opt &= ~OCFS2_MOUNT_NOINTR; + mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; break; case Opt_nointr: - *mount_opt |= OCFS2_MOUNT_NOINTR; + mopt->mount_opt |= OCFS2_MOUNT_NOINTR; break; case Opt_err_panic: - *mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; + mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; break; case Opt_err_ro: - *mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; + mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; break; case Opt_data_ordered: - *mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; + mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; break; case Opt_data_writeback: - *mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; + mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; break; case Opt_atime_quantum: if (match_int(&args[0], &option)) { @@ -784,9 +805,7 @@ static int ocfs2_parse_options(struct super_block *sb, goto bail; } if (option >= 0) - osb->s_atime_quantum = option; - else - osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; + mopt->atime_quantum = option; break; case Opt_slot: option = 0; @@ -795,7 +814,7 @@ static int ocfs2_parse_options(struct super_block *sb, goto bail; } if (option) - *slot = (s16)option; + mopt->slot = (s16)option; break; default: mlog(ML_ERROR, @@ -813,6 +832,41 @@ bail: return status; } +static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) +{ + struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb); + unsigned long opts = osb->s_mount_opt; + + if (opts & OCFS2_MOUNT_HB_LOCAL) + seq_printf(s, ",_netdev,heartbeat=local"); + else + seq_printf(s, ",heartbeat=none"); + + if (opts & OCFS2_MOUNT_NOINTR) + seq_printf(s, ",nointr"); + + if (opts & OCFS2_MOUNT_DATA_WRITEBACK) + seq_printf(s, ",data=writeback"); + else + seq_printf(s, ",data=ordered"); + + if (opts & OCFS2_MOUNT_BARRIER) + seq_printf(s, ",barrier=1"); + + if (opts & OCFS2_MOUNT_ERRORS_PANIC) + seq_printf(s, ",errors=panic"); + else + seq_printf(s, ",errors=remount-ro"); + + if (osb->preferred_slot != OCFS2_INVALID_SLOT) + seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); + + if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM) + seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); + + return 0; +} + static int __init ocfs2_init(void) { int status; @@ -984,7 +1038,7 @@ static int ocfs2_initialize_mem_caches(void) 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| SLAB_MEM_SPREAD), - ocfs2_inode_init_once, NULL); + ocfs2_inode_init_once); if (!ocfs2_inode_cachep) return -ENOMEM; @@ -1192,12 +1246,13 @@ static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) tmp = ocfs2_request_umount_vote(osb); if (tmp < 0) mlog_errno(tmp); + } - if (osb->slot_num != OCFS2_INVALID_SLOT) - ocfs2_put_slot(osb); + if (osb->slot_num != OCFS2_INVALID_SLOT) + ocfs2_put_slot(osb); + if (osb->dlm) ocfs2_super_unlock(osb, 1); - } ocfs2_release_system_inodes(osb); @@ -1258,9 +1313,9 @@ static int ocfs2_initialize_super(struct super_block *sb, struct buffer_head *bh, int sector_size) { - int status = 0; - int i; - struct ocfs2_dinode *di = NULL; + int status; + int i, cbits, bbits; + struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; struct inode *inode = NULL; struct buffer_head *bitmap_bh = NULL; struct ocfs2_journal *journal; @@ -1279,9 +1334,12 @@ static int ocfs2_initialize_super(struct super_block *sb, sb->s_fs_info = osb; sb->s_op = &ocfs2_sops; sb->s_export_op = &ocfs2_export_ops; + sb->s_time_gran = 1; sb->s_flags |= MS_NOATIME; /* this is needed to support O_LARGEFILE */ - sb->s_maxbytes = ocfs2_max_file_offset(sb->s_blocksize_bits); + cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); + bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); + sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); osb->sb = sb; /* Save off for ocfs2_rw_direct */ @@ -1341,8 +1399,6 @@ static int ocfs2_initialize_super(struct super_block *sb, goto bail; } - di = (struct ocfs2_dinode *)bh->b_data; - osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { mlog(ML_ERROR, "Invalid number of node slots (%u)\n", @@ -1578,7 +1634,7 @@ static int ocfs2_verify_volume(struct ocfs2_dinode *di, static int ocfs2_check_volume(struct ocfs2_super *osb) { - int status = 0; + int status; int dirty; int local; struct ocfs2_dinode *local_alloc = NULL; /* only used if we diff --git a/fs/ocfs2/super.h b/fs/ocfs2/super.h index 3b9cb3d0b00..783f5270f2a 100644 --- a/fs/ocfs2/super.h +++ b/fs/ocfs2/super.h @@ -45,6 +45,4 @@ void __ocfs2_abort(struct super_block *sb, #define ocfs2_abort(sb, fmt, args...) __ocfs2_abort(sb, __PRETTY_FUNCTION__, fmt, ##args) -unsigned long long ocfs2_max_file_offset(unsigned int blockshift); - #endif /* OCFS2_SUPER_H */ diff --git a/fs/ocfs2/sysfile.c b/fs/ocfs2/sysfile.c index 5df6e35d09b..fd2e846e3e6 100644 --- a/fs/ocfs2/sysfile.c +++ b/fs/ocfs2/sysfile.c @@ -100,17 +100,14 @@ static struct inode * _ocfs2_get_system_file_inode(struct ocfs2_super *osb, char namebuf[40]; struct inode *inode = NULL; u64 blkno; - struct buffer_head *dirent_bh = NULL; - struct ocfs2_dir_entry *de = NULL; int status = 0; ocfs2_sprintf_system_inode_name(namebuf, sizeof(namebuf), type, slot); - status = ocfs2_find_files_on_disk(namebuf, strlen(namebuf), - &blkno, osb->sys_root_inode, - &dirent_bh, &de); + status = ocfs2_lookup_ino_from_name(osb->sys_root_inode, namebuf, + strlen(namebuf), &blkno); if (status < 0) { goto bail; } @@ -122,8 +119,7 @@ static struct inode * _ocfs2_get_system_file_inode(struct ocfs2_super *osb, goto bail; } bail: - if (dirent_bh) - brelse(dirent_bh); + return inode; } diff --git a/fs/ocfs2/uptodate.c b/fs/ocfs2/uptodate.c index 39814b900fc..4da8851f2b2 100644 --- a/fs/ocfs2/uptodate.c +++ b/fs/ocfs2/uptodate.c @@ -548,7 +548,7 @@ int __init init_ocfs2_uptodate_cache(void) { ocfs2_uptodate_cachep = kmem_cache_create("ocfs2_uptodate", sizeof(struct ocfs2_meta_cache_item), - 0, SLAB_HWCACHE_ALIGN, NULL, NULL); + 0, SLAB_HWCACHE_ALIGN, NULL); if (!ocfs2_uptodate_cachep) return -ENOMEM; diff --git a/fs/ocfs2/vote.c b/fs/ocfs2/vote.c index 66a13ee63d4..c05358538f2 100644 --- a/fs/ocfs2/vote.c +++ b/fs/ocfs2/vote.c @@ -66,7 +66,7 @@ struct ocfs2_vote_msg { struct ocfs2_msg_hdr v_hdr; __be32 v_reserved1; -}; +} __attribute__ ((packed)); /* Responses are given these values to maintain backwards * compatibility with older ocfs2 versions */ @@ -78,7 +78,7 @@ struct ocfs2_response_msg { struct ocfs2_msg_hdr r_hdr; __be32 r_response; -}; +} __attribute__ ((packed)); struct ocfs2_vote_work { struct list_head w_list; |