+#ifdef HAVE_EXT_PBLOCK /* Name changed to ext4_ext_pblock for kernel 2.6.35 */
+#define ldiskfs_ext_pblock(ex) ext_pblock((ex))
+#endif
+
+struct bpointers {
+ unsigned long *blocks;
+ unsigned long start;
+ int num;
+ int init_num;
+ int create;
+};
+
+static long ldiskfs_ext_find_goal(struct inode *inode,
+ struct ldiskfs_ext_path *path,
+ unsigned long block, int *aflags)
+{
+ struct ldiskfs_inode_info *ei = LDISKFS_I(inode);
+ unsigned long bg_start;
+ unsigned long colour;
+ int depth;
+
+ if (path) {
+ struct ldiskfs_extent *ex;
+ depth = path->p_depth;
+
+ /* try to predict block placement */
+ if ((ex = path[depth].p_ext))
+ return ldiskfs_ext_pblock(ex) +
+ (block - le32_to_cpu(ex->ee_block));
+
+ /* it looks index is empty
+ * try to find starting from index itself */
+ if (path[depth].p_bh)
+ return path[depth].p_bh->b_blocknr;
+ }
+
+ /* OK. use inode's group */
+ bg_start = (ei->i_block_group * LDISKFS_BLOCKS_PER_GROUP(inode->i_sb)) +
+ le32_to_cpu(LDISKFS_SB(inode->i_sb)->s_es->s_first_data_block);
+ colour = (current->pid % 16) *
+ (LDISKFS_BLOCKS_PER_GROUP(inode->i_sb) / 16);
+ return bg_start + colour + block;
+}
+
+static unsigned long new_blocks(handle_t *handle, struct inode *inode,
+ struct ldiskfs_ext_path *path,
+ unsigned long block, unsigned long *count,
+ int *err)
+{
+ struct ldiskfs_allocation_request ar;
+ unsigned long pblock;
+ int aflags;
+
+ /* find neighbour allocated blocks */
+ ar.lleft = block;
+ *err = ldiskfs_ext_search_left(inode, path, &ar.lleft, &ar.pleft);
+ if (*err)
+ return 0;
+ ar.lright = block;
+ *err = ldiskfs_ext_search_right(inode, path, &ar.lright, &ar.pright);
+ if (*err)
+ return 0;
+
+ /* allocate new block */
+ ar.goal = ldiskfs_ext_find_goal(inode, path, block, &aflags);
+ ar.inode = inode;
+ ar.logical = block;
+ ar.len = *count;
+ ar.flags = LDISKFS_MB_HINT_DATA;
+ pblock = ldiskfs_mb_new_blocks(handle, &ar, err);
+ *count = ar.len;
+ return pblock;
+}
+
+static int ldiskfs_ext_new_extent_cb(struct inode *inode,
+ struct ldiskfs_ext_path *path,
+ struct ldiskfs_ext_cache *cex,
+#ifdef HAVE_EXT_PREPARE_CB_EXTENT
+ struct ldiskfs_extent *ex,
+#endif
+ void *cbdata)
+{
+ struct bpointers *bp = cbdata;
+ struct ldiskfs_extent nex;
+ unsigned long pblock;
+ unsigned long tgen;
+ int err, i;
+ unsigned long count;
+ handle_t *handle;
+
+#ifdef LDISKFS_EXT_CACHE_EXTENT /* until kernel 2.6.37 */
+ if (cex->ec_type == LDISKFS_EXT_CACHE_EXTENT) {
+#else
+ if ((cex->ec_len != 0) && (cex->ec_start != 0)) {
+#endif
+ err = EXT_CONTINUE;
+ goto map;
+ }
+
+ if (bp->create == 0) {
+ i = 0;
+ if (cex->ec_block < bp->start)
+ i = bp->start - cex->ec_block;
+ if (i >= cex->ec_len)
+ CERROR("nothing to do?! i = %d, e_num = %u\n",
+ i, cex->ec_len);
+ for (; i < cex->ec_len && bp->num; i++) {
+ *(bp->blocks) = 0;
+ bp->blocks++;
+ bp->num--;
+ bp->start++;
+ }
+
+ return EXT_CONTINUE;
+ }
+
+ tgen = LDISKFS_I(inode)->i_ext_generation;
+ count = ldiskfs_ext_calc_credits_for_insert(inode, path);
+
+ handle = ldiskfs_journal_start(inode, count + LDISKFS_ALLOC_NEEDED + 1);
+ if (IS_ERR(handle)) {
+ return PTR_ERR(handle);
+ }
+
+ if (tgen != LDISKFS_I(inode)->i_ext_generation) {
+ /* the tree has changed. so path can be invalid at moment */
+ ldiskfs_journal_stop(handle);
+ return EXT_REPEAT;
+ }
+
+ /* In 2.6.32 kernel, ldiskfs_ext_walk_space()'s callback func is not
+ * protected by i_data_sem as whole. so we patch it to store
+ * generation to path and now verify the tree hasn't changed */
+ down_write((&LDISKFS_I(inode)->i_data_sem));
+
+ /* validate extent, make sure the extent tree does not changed */
+ if (LDISKFS_I(inode)->i_ext_generation != path[0].p_generation) {
+ /* cex is invalid, try again */
+ up_write(&LDISKFS_I(inode)->i_data_sem);
+ ldiskfs_journal_stop(handle);
+ return EXT_REPEAT;
+ }
+
+ count = cex->ec_len;
+ pblock = new_blocks(handle, inode, path, cex->ec_block, &count, &err);
+ if (!pblock)
+ goto out;
+ BUG_ON(count > cex->ec_len);
+
+ /* insert new extent */
+ nex.ee_block = cpu_to_le32(cex->ec_block);
+ ldiskfs_ext_store_pblock(&nex, pblock);
+ nex.ee_len = cpu_to_le16(count);
+ err = ldiskfs_ext_insert_extent(handle, inode, path, &nex, 0);
+ if (err) {
+ /* free data blocks we just allocated */
+ /* not a good idea to call discard here directly,
+ * but otherwise we'd need to call it every free() */
+ ldiskfs_discard_preallocations(inode);
+#ifdef HAVE_EXT_FREE_BLOCK_WITH_BUFFER_HEAD /* Introduced in 2.6.32-rc7 */
+ ldiskfs_free_blocks(handle, inode, NULL,
+ ldiskfs_ext_pblock(&nex),
+ le16_to_cpu(nex.ee_len), 0);
+#else
+ ldiskfs_free_blocks(handle, inode, ldiskfs_ext_pblock(&nex),
+ le16_to_cpu(nex.ee_len), 0);
+#endif
+ goto out;
+ }
+
+ /*
+ * Putting len of the actual extent we just inserted,
+ * we are asking ldiskfs_ext_walk_space() to continue
+ * scaning after that block
+ */
+ cex->ec_len = le16_to_cpu(nex.ee_len);
+ cex->ec_start = ldiskfs_ext_pblock(&nex);
+ BUG_ON(le16_to_cpu(nex.ee_len) == 0);
+ BUG_ON(le32_to_cpu(nex.ee_block) != cex->ec_block);
+
+out:
+ up_write((&LDISKFS_I(inode)->i_data_sem));
+ ldiskfs_journal_stop(handle);
+map:
+ if (err >= 0) {
+ /* map blocks */
+ if (bp->num == 0) {
+ CERROR("hmm. why do we find this extent?\n");
+ CERROR("initial space: %lu:%u\n",
+ bp->start, bp->init_num);
+#ifdef LDISKFS_EXT_CACHE_EXTENT /* until kernel 2.6.37 */
+ CERROR("current extent: %u/%u/%llu %d\n",
+ cex->ec_block, cex->ec_len,
+ (unsigned long long)cex->ec_start,
+ cex->ec_type);
+#else
+ CERROR("current extent: %u/%u/%llu\n",
+ cex->ec_block, cex->ec_len,
+ (unsigned long long)cex->ec_start);
+#endif
+ }
+ i = 0;
+ if (cex->ec_block < bp->start)
+ i = bp->start - cex->ec_block;
+ if (i >= cex->ec_len)
+ CERROR("nothing to do?! i = %d, e_num = %u\n",
+ i, cex->ec_len);
+ for (; i < cex->ec_len && bp->num; i++) {
+ *(bp->blocks) = cex->ec_start + i;
+#ifdef LDISKFS_EXT_CACHE_EXTENT /* until kernel 2.6.37 */
+ if (cex->ec_type != LDISKFS_EXT_CACHE_EXTENT) {
+#else
+ if ((cex->ec_len == 0) || (cex->ec_start == 0)) {
+#endif
+ /* unmap any possible underlying metadata from
+ * the block device mapping. bug 6998. */
+ unmap_underlying_metadata(inode->i_sb->s_bdev,
+ *(bp->blocks));
+ }
+ bp->blocks++;
+ bp->num--;
+ bp->start++;
+ }
+ }
+ return err;
+}
+
+int osd_ldiskfs_map_nblocks(struct inode *inode, unsigned long block,
+ unsigned long num, unsigned long *blocks,
+ int create)
+{
+ struct bpointers bp;
+ int err;
+
+ CDEBUG(D_OTHER, "blocks %lu-%lu requested for inode %u\n",
+ block, block + num - 1, (unsigned) inode->i_ino);
+
+ bp.blocks = blocks;
+ bp.start = block;
+ bp.init_num = bp.num = num;
+ bp.create = create;
+
+ err = ldiskfs_ext_walk_space(inode, block, num,
+ ldiskfs_ext_new_extent_cb, &bp);
+ ldiskfs_ext_invalidate_cache(inode);
+
+ return err;
+}
+
+int osd_ldiskfs_map_ext_inode_pages(struct inode *inode, struct page **page,
+ int pages, unsigned long *blocks,
+ int create)
+{
+ int blocks_per_page = PAGE_CACHE_SIZE >> inode->i_blkbits;
+ int rc = 0, i = 0;
+ struct page *fp = NULL;
+ int clen = 0;
+
+ CDEBUG(D_OTHER, "inode %lu: map %d pages from %lu\n",
+ inode->i_ino, pages, (*page)->index);
+
+ /* pages are sorted already. so, we just have to find
+ * contig. space and process them properly */
+ while (i < pages) {
+ if (fp == NULL) {
+ /* start new extent */
+ fp = *page++;
+ clen = 1;
+ i++;
+ continue;
+ } else if (fp->index + clen == (*page)->index) {
+ /* continue the extent */
+ page++;
+ clen++;
+ i++;
+ continue;
+ }
+
+ /* process found extent */
+ rc = osd_ldiskfs_map_nblocks(inode, fp->index * blocks_per_page,
+ clen * blocks_per_page, blocks,
+ create);
+ if (rc)
+ GOTO(cleanup, rc);
+
+ /* look for next extent */
+ fp = NULL;
+ blocks += blocks_per_page * clen;
+ }
+
+ if (fp)
+ rc = osd_ldiskfs_map_nblocks(inode, fp->index * blocks_per_page,
+ clen * blocks_per_page, blocks,
+ create);
+cleanup:
+ return rc;
+}
+
+int osd_ldiskfs_map_bm_inode_pages(struct inode *inode, struct page **page,
+ int pages, unsigned long *blocks,
+ int create)
+{
+ int blocks_per_page = PAGE_CACHE_SIZE >> inode->i_blkbits;
+ unsigned long *b;
+ int rc = 0, i;
+
+ for (i = 0, b = blocks; i < pages; i++, page++) {
+ rc = ldiskfs_map_inode_page(inode, *page, b, create);
+ if (rc) {
+ CERROR("ino %lu, blk %lu create %d: rc %d\n",
+ inode->i_ino, *b, create, rc);
+ break;
+ }
+
+ b += blocks_per_page;
+ }
+ return rc;
+}
+
+static int osd_ldiskfs_map_inode_pages(struct inode *inode, struct page **page,
+ int pages, unsigned long *blocks,
+ int create)
+{
+ int rc;
+
+ if (LDISKFS_I(inode)->i_flags & LDISKFS_EXTENTS_FL) {
+ rc = osd_ldiskfs_map_ext_inode_pages(inode, page, pages,
+ blocks, create);
+ return rc;
+ }
+ rc = osd_ldiskfs_map_bm_inode_pages(inode, page, pages, blocks, create);
+
+ return rc;
+}
+