diff options
Diffstat (limited to 'fs/btrfs/send.c')
| -rw-r--r-- | fs/btrfs/send.c | 915 |
1 files changed, 501 insertions, 414 deletions
diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c index fb5ffe9..c7beb54 100644 --- a/fs/btrfs/send.c +++ b/fs/btrfs/send.c @@ -107,7 +107,6 @@ struct send_ctx { int cur_inode_new; int cur_inode_new_gen; int cur_inode_deleted; - int cur_inode_first_ref_orphan; u64 cur_inode_size; u64 cur_inode_mode; @@ -126,7 +125,15 @@ struct send_ctx { struct name_cache_entry { struct list_head list; - struct list_head use_list; + /* + * radix_tree has only 32bit entries but we need to handle 64bit inums. + * We use the lower 32bit of the 64bit inum to store it in the tree. If + * more then one inum would fall into the same entry, we use radix_list + * to store the additional entries. radix_list is also used to store + * entries where two entries have the same inum but different + * generations. + */ + struct list_head radix_list; u64 ino; u64 gen; u64 parent_ino; @@ -328,6 +335,7 @@ out: return ret; } +#if 0 static void fs_path_remove(struct fs_path *p) { BUG_ON(p->reversed); @@ -335,6 +343,7 @@ static void fs_path_remove(struct fs_path *p) p->end--; *p->end = 0; } +#endif static int fs_path_copy(struct fs_path *p, struct fs_path *from) { @@ -377,7 +386,7 @@ static struct btrfs_path *alloc_path_for_send(void) return path; } -static int write_buf(struct send_ctx *sctx, const void *buf, u32 len) +int write_buf(struct file *filp, const void *buf, u32 len, loff_t *off) { int ret; mm_segment_t old_fs; @@ -387,8 +396,7 @@ static int write_buf(struct send_ctx *sctx, const void *buf, u32 len) set_fs(KERNEL_DS); while (pos < len) { - ret = vfs_write(sctx->send_filp, (char *)buf + pos, len - pos, - &sctx->send_off); + ret = vfs_write(filp, (char *)buf + pos, len - pos, off); /* TODO handle that correctly */ /*if (ret == -ERESTARTSYS) { continue; @@ -544,7 +552,8 @@ static int send_header(struct send_ctx *sctx) strcpy(hdr.magic, BTRFS_SEND_STREAM_MAGIC); hdr.version = cpu_to_le32(BTRFS_SEND_STREAM_VERSION); - return write_buf(sctx, &hdr, sizeof(hdr)); + return write_buf(sctx->send_filp, &hdr, sizeof(hdr), + &sctx->send_off); } /* @@ -581,7 +590,8 @@ static int send_cmd(struct send_ctx *sctx) crc = crc32c(0, (unsigned char *)sctx->send_buf, sctx->send_size); hdr->crc = cpu_to_le32(crc); - ret = write_buf(sctx, sctx->send_buf, sctx->send_size); + ret = write_buf(sctx->send_filp, sctx->send_buf, sctx->send_size, + &sctx->send_off); sctx->total_send_size += sctx->send_size; sctx->cmd_send_size[le16_to_cpu(hdr->cmd)] += sctx->send_size; @@ -687,7 +697,8 @@ out: */ static int get_inode_info(struct btrfs_root *root, u64 ino, u64 *size, u64 *gen, - u64 *mode, u64 *uid, u64 *gid) + u64 *mode, u64 *uid, u64 *gid, + u64 *rdev) { int ret; struct btrfs_inode_item *ii; @@ -721,6 +732,8 @@ static int get_inode_info(struct btrfs_root *root, *uid = btrfs_inode_uid(path->nodes[0], ii); if (gid) *gid = btrfs_inode_gid(path->nodes[0], ii); + if (rdev) + *rdev = btrfs_inode_rdev(path->nodes[0], ii); out: btrfs_free_path(path); @@ -852,7 +865,6 @@ static int iterate_dir_item(struct send_ctx *sctx, struct extent_buffer *eb; struct btrfs_item *item; struct btrfs_dir_item *di; - struct btrfs_path *tmp_path = NULL; struct btrfs_key di_key; char *buf = NULL; char *buf2 = NULL; @@ -874,12 +886,6 @@ static int iterate_dir_item(struct send_ctx *sctx, goto out; } - tmp_path = alloc_path_for_send(); - if (!tmp_path) { - ret = -ENOMEM; - goto out; - } - eb = path->nodes[0]; slot = path->slots[0]; item = btrfs_item_nr(eb, slot); @@ -941,7 +947,6 @@ static int iterate_dir_item(struct send_ctx *sctx, } out: - btrfs_free_path(tmp_path); if (buf_virtual) vfree(buf); else @@ -1026,12 +1031,12 @@ struct backref_ctx { u64 extent_len; /* Just to check for bugs in backref resolving */ - int found_in_send_root; + int found_itself; }; static int __clone_root_cmp_bsearch(const void *key, const void *elt) { - u64 root = (u64)key; + u64 root = (u64)(uintptr_t)key; struct clone_root *cr = (struct clone_root *)elt; if (root < cr->root->objectid) @@ -1055,6 +1060,7 @@ static int __clone_root_cmp_sort(const void *e1, const void *e2) /* * Called for every backref that is found for the current extent. + * Results are collected in sctx->clone_roots->ino/offset/found_refs */ static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_) { @@ -1064,7 +1070,7 @@ static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_) u64 i_size; /* First check if the root is in the list of accepted clone sources */ - found = bsearch((void *)root, bctx->sctx->clone_roots, + found = bsearch((void *)(uintptr_t)root, bctx->sctx->clone_roots, bctx->sctx->clone_roots_cnt, sizeof(struct clone_root), __clone_root_cmp_bsearch); @@ -1074,14 +1080,15 @@ static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_) if (found->root == bctx->sctx->send_root && ino == bctx->cur_objectid && offset == bctx->cur_offset) { - bctx->found_in_send_root = 1; + bctx->found_itself = 1; } /* - * There are inodes that have extents that lie behind it's i_size. Don't + * There are inodes that have extents that lie behind its i_size. Don't * accept clones from these extents. */ - ret = get_inode_info(found->root, ino, &i_size, NULL, NULL, NULL, NULL); + ret = get_inode_info(found->root, ino, &i_size, NULL, NULL, NULL, NULL, + NULL); if (ret < 0) return ret; @@ -1101,16 +1108,12 @@ static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_) */ if (ino >= bctx->cur_objectid) return 0; - /*if (ino > ctx->cur_objectid) +#if 0 + if (ino > bctx->cur_objectid) return 0; - if (offset + ctx->extent_len > ctx->cur_offset) - return 0;*/ - - bctx->found++; - found->found_refs++; - found->ino = ino; - found->offset = offset; - return 0; + if (offset + bctx->extent_len > bctx->cur_offset) + return 0; +#endif } bctx->found++; @@ -1130,6 +1133,12 @@ static int __iterate_backrefs(u64 ino, u64 offset, u64 root, void *ctx_) } /* + * Given an inode, offset and extent item, it finds a good clone for a clone + * instruction. Returns -ENOENT when none could be found. The function makes + * sure that the returned clone is usable at the point where sending is at the + * moment. This means, that no clones are accepted which lie behind the current + * inode+offset. + * * path must point to the extent item when called. */ static int find_extent_clone(struct send_ctx *sctx, @@ -1141,20 +1150,29 @@ static int find_extent_clone(struct send_ctx *sctx, int ret; int extent_type; u64 logical; + u64 disk_byte; u64 num_bytes; u64 extent_item_pos; + u64 flags = 0; struct btrfs_file_extent_item *fi; struct extent_buffer *eb = path->nodes[0]; - struct backref_ctx backref_ctx; + struct backref_ctx *backref_ctx = NULL; struct clone_root *cur_clone_root; struct btrfs_key found_key; struct btrfs_path *tmp_path; + int compressed; u32 i; tmp_path = alloc_path_for_send(); if (!tmp_path) return -ENOMEM; + backref_ctx = kmalloc(sizeof(*backref_ctx), GFP_NOFS); + if (!backref_ctx) { + ret = -ENOMEM; + goto out; + } + if (data_offset >= ino_size) { /* * There may be extents that lie behind the file's size. @@ -1172,22 +1190,23 @@ static int find_extent_clone(struct send_ctx *sctx, ret = -ENOENT; goto out; } + compressed = btrfs_file_extent_compression(eb, fi); num_bytes = btrfs_file_extent_num_bytes(eb, fi); - logical = btrfs_file_extent_disk_bytenr(eb, fi); - if (logical == 0) { + disk_byte = btrfs_file_extent_disk_bytenr(eb, fi); + if (disk_byte == 0) { ret = -ENOENT; goto out; } - logical += btrfs_file_extent_offset(eb, fi); + logical = disk_byte + btrfs_file_extent_offset(eb, fi); - ret = extent_from_logical(sctx->send_root->fs_info, - logical, tmp_path, &found_key); + ret = extent_from_logical(sctx->send_root->fs_info, disk_byte, tmp_path, + &found_key, &flags); btrfs_release_path(tmp_path); if (ret < 0) goto out; - if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK) { + if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) { ret = -EIO; goto out; } @@ -1202,12 +1221,12 @@ static int find_extent_clone(struct send_ctx *sctx, cur_clone_root->found_refs = 0; } - backref_ctx.sctx = sctx; - backref_ctx.found = 0; - backref_ctx.cur_objectid = ino; - backref_ctx.cur_offset = data_offset; - backref_ctx.found_in_send_root = 0; - backref_ctx.extent_len = num_bytes; + backref_ctx->sctx = sctx; + backref_ctx->found = 0; + backref_ctx->cur_objectid = ino; + backref_ctx->cur_offset = data_offset; + backref_ctx->found_itself = 0; + backref_ctx->extent_len = num_bytes; /* * The last extent of a file may be too large due to page alignment. @@ -1215,25 +1234,31 @@ static int find_extent_clone(struct send_ctx *sctx, * __iterate_backrefs work. */ if (data_offset + num_bytes >= ino_size) - backref_ctx.extent_len = ino_size - data_offset; + backref_ctx->extent_len = ino_size - data_offset; /* * Now collect all backrefs. */ + if (compressed == BTRFS_COMPRESS_NONE) + extent_item_pos = logical - found_key.objectid; + else + extent_item_pos = 0; + extent_item_pos = logical - found_key.objectid; ret = iterate_extent_inodes(sctx->send_root->fs_info, found_key.objectid, extent_item_pos, 1, - __iterate_backrefs, &backref_ctx); + __iterate_backrefs, backref_ctx); + if (ret < 0) goto out; - if (!backref_ctx.found_in_send_root) { + if (!backref_ctx->found_itself) { /* found a bug in backref code? */ ret = -EIO; printk(KERN_ERR "btrfs: ERROR did not find backref in " "send_root. inode=%llu, offset=%llu, " - "logical=%llu\n", - ino, data_offset, logical); + "disk_byte=%llu found extent=%llu\n", + ino, data_offset, disk_byte, found_key.objectid); goto out; } @@ -1242,7 +1267,7 @@ verbose_printk(KERN_DEBUG "btrfs: find_extent_clone: data_offset=%llu, " "num_bytes=%llu, logical=%llu\n", data_offset, ino, num_bytes, logical); - if (!backref_ctx.found) + if (!backref_ctx->found) verbose_printk("btrfs: no clones found\n"); cur_clone_root = NULL; @@ -1253,7 +1278,6 @@ verbose_printk(KERN_DEBUG "btrfs: find_extent_clone: data_offset=%llu, " else if (sctx->clone_roots[i].root == sctx->send_root) /* prefer clones from send_root over others */ cur_clone_root = sctx->clone_roots + i; - break; } } @@ -1267,6 +1291,7 @@ verbose_printk(KERN_DEBUG "btrfs: find_extent_clone: data_offset=%llu, " out: btrfs_free_path(tmp_path); + kfree(backref_ctx); return ret; } @@ -1307,8 +1332,6 @@ static int read_symlink(struct send_ctx *sctx, len = btrfs_file_extent_inline_len(path->nodes[0], ei); ret = fs_path_add_from_extent_buffer(dest, path->nodes[0], off, len); - if (ret < 0) - goto out; out: btrfs_free_path(path); @@ -1404,7 +1427,7 @@ static int get_cur_inode_state(struct send_ctx *sctx, u64 ino, u64 gen) u64 right_gen; ret = get_inode_info(sctx->send_root, ino, NULL, &left_gen, NULL, NULL, - NULL); + NULL, NULL); if (ret < 0 && ret != -ENOENT) goto out; left_ret = ret; @@ -1413,16 +1436,16 @@ static int get_cur_inode_state(struct send_ctx *sctx, u64 ino, u64 gen) right_ret = -ENOENT; } else { ret = get_inode_info(sctx->parent_root, ino, NULL, &right_gen, - NULL, NULL, NULL); + NULL, NULL, NULL, NULL); if (ret < 0 && ret != -ENOENT) goto out; right_ret = ret; } if (!left_ret && !right_ret) { - if (left_gen == gen && right_gen == gen) + if (left_gen == gen && right_gen == gen) { ret = inode_state_no_change; - else if (left_gen == gen) { + } else if (left_gen == gen) { if (ino < sctx->send_progress) ret = inode_state_did_create; else @@ -1516,6 +1539,10 @@ out: return ret; } +/* + * Looks up the first btrfs_inode_ref of a given ino. It returns the parent dir, + * generation of the parent dir and the name of the dir entry. + */ static int get_first_ref(struct send_ctx *sctx, struct btrfs_root *root, u64 ino, u64 *dir, u64 *dir_gen, struct fs_path *name) @@ -1557,7 +1584,7 @@ static int get_first_ref(struct send_ctx *sctx, btrfs_release_path(path); ret = get_inode_info(root, found_key.offset, NULL, dir_gen, NULL, NULL, - NULL); + NULL, NULL); if (ret < 0) goto out; @@ -1586,22 +1613,28 @@ static int is_first_ref(struct send_ctx *sctx, if (ret < 0) goto out; - if (name_len != fs_path_len(tmp_name)) { + if (dir != tmp_dir || name_len != fs_path_len(tmp_name)) { ret = 0; goto out; } - ret = memcmp(tmp_name->start, name, name_len); - if (ret) - ret = 0; - else - ret = 1; + ret = !memcmp(tmp_name->start, name, name_len); out: fs_path_free(sctx, tmp_name); return ret; } +/* + * Used by process_recorded_refs to determine if a new ref would overwrite an + * already existing ref. In case it detects an overwrite, it returns the + * inode/gen in who_ino/who_gen. + * When an overwrite is detected, process_recorded_refs does proper orphanizing + * to make sure later references to the overwritten inode are possible. + * Orphanizing is however only required for the first ref of an inode. + * process_recorded_refs does an additional is_first_ref check to see if + * orphanizing is really required. + */ static int will_overwrite_ref(struct send_ctx *sctx, u64 dir, u64 dir_gen, const char *name, int name_len, u64 *who_ino, u64 *who_gen) @@ -1626,9 +1659,14 @@ static int will_overwrite_ref(struct send_ctx *sctx, u64 dir, u64 dir_gen, goto out; } + /* + * Check if the overwritten ref was already processed. If yes, the ref + * was already unlinked/moved, so we can safely assume that we will not + * overwrite anything at this point in time. + */ if (other_inode > sctx->send_progress) { ret = get_inode_info(sctx->parent_root, other_inode, NULL, - who_gen, NULL, NULL, NULL); + who_gen, NULL, NULL, NULL, NULL); if (ret < 0) goto out; @@ -1642,6 +1680,13 @@ out: return ret; } +/* + * Checks if the ref was overwritten by an already processed inode. This is + * used by __get_cur_name_and_parent to find out if the ref was orphanized and + * thus the orphan name needs be used. + * process_recorded_refs also uses it to avoid unlinking of refs that were + * overwritten. + */ static int did_overwrite_ref(struct send_ctx *sctx, u64 dir, u64 dir_gen, u64 ino, u64 ino_gen, @@ -1671,7 +1716,7 @@ static int did_overwrite_ref(struct send_ctx *sctx, } ret = get_inode_info(sctx->send_root, ow_inode, NULL, &gen, NULL, NULL, - NULL); + NULL, NULL); if (ret < 0) goto out; @@ -1690,6 +1735,11 @@ out: return ret; } +/* + * Same as did_overwrite_ref, but also checks if it is the first ref of an inode + * that got overwritten. This is used by process_recorded_refs to determine + * if it has to use the path as returned by get_cur_path or the orphan name. + */ static int did_overwrite_first_ref(struct send_ctx *sctx, u64 ino, u64 gen) { int ret = 0; @@ -1710,39 +1760,40 @@ static int did_overwrite_first_ref(struct send_ctx *sctx, u64 ino, u64 gen) ret = did_overwrite_ref(sctx, dir, dir_gen, ino, gen, name->start, fs_path_len(name)); - if (ret < 0) - goto out; out: fs_path_free(sctx, name); return ret; } +/* + * Insert a name cache entry. On 32bit kernels the radix tree index is 32bit, + * so we need to do some special handling in case we have clashes. This function + * takes care of this with the help of name_cache_entry::radix_list. + * In case of error, nce is kfreed. + */ static int name_cache_insert(struct send_ctx *sctx, struct name_cache_entry *nce) { int ret = 0; - struct name_cache_entry **ncea; - - ncea = radix_tree_lookup(&sctx->name_cache, nce->ino); - if (ncea) { - if (!ncea[0]) - ncea[0] = nce; - else if (!ncea[1]) - ncea[1] = nce; - else - BUG(); - } else { - ncea = kmalloc(sizeof(void *) * 2, GFP_NOFS); - if (!ncea) + struct list_head *nce_head; + + nce_head = radix_tree_lookup(&sctx->name_cache, + (unsigned long)nce->ino); + if (!nce_head) { + nce_head = kmalloc(sizeof(*nce_head), GFP_NOFS); + if (!nce_head) return -ENOMEM; + INIT_LIST_HEAD(nce_head); - ncea[0] = nce; - ncea[1] = NULL; - ret = radix_tree_insert(&sctx->name_cache, nce->ino, ncea); - if (ret < 0) + ret = radix_tree_insert(&sctx->name_cache, nce->ino, nce_head); + if (ret < 0) { + kfree(nce_head); + kfree(nce); return ret; + } } + list_add_tail(&nce->radix_list, nce_head); list_add_tail(&nce->list, &sctx->name_cache_list); sctx->name_cache_size++; @@ -1752,50 +1803,52 @@ static int name_cache_insert(struct send_ctx *sctx, static void name_cache_delete(struct send_ctx *sctx, struct name_cache_entry *nce) { - struct name_cache_entry **ncea; - - ncea = radix_tree_lookup(&sctx->name_cache, nce->ino); - BUG_ON(!ncea); - - if (ncea[0] == nce) - ncea[0] = NULL; - else if (ncea[1] == nce) - ncea[1] = NULL; - else - BUG(); + struct list_head *nce_head; - if (!ncea[0] && !ncea[1]) { - radix_tree_delete(&sctx->name_cache, nce->ino); - kfree(ncea); - } + nce_head = radix_tree_lookup(&sctx->name_cache, + (unsigned long)nce->ino); + BUG_ON(!nce_head); + list_del(&nce->radix_list); list_del(&nce->list); - sctx->name_cache_size--; + + if (list_empty(nce_head)) { + radix_tree_delete(&sctx->name_cache, (unsigned long)nce->ino); + kfree(nce_head); + } } static struct name_cache_entry *name_cache_search(struct send_ctx *sctx, u64 ino, u64 gen) { - struct name_cache_entry **ncea; + struct list_head *nce_head; + struct name_cache_entry *cur; - ncea = radix_tree_lookup(&sctx->name_cache, ino); - if (!ncea) + nce_head = radix_tree_lookup(&sctx->name_cache, (unsigned long)ino); + if (!nce_head) return NULL; - if (ncea[0] && ncea[0]->gen == gen) - return ncea[0]; - else if (ncea[1] && ncea[1]->gen == gen) - return ncea[1]; + list_for_each_entry(cur, nce_head, radix_list) { + if (cur->ino == ino && cur->gen == gen) + return cur; + } return NULL; } +/* + * Removes the entry from the list and adds it back to the end. This marks the + * entry as recently used so that name_cache_clean_unused does not remove it. + */ static void name_cache_used(struct send_ctx *sctx, struct name_cache_entry *nce) { list_del(&nce->list); list_add_tail(&nce->list, &sctx->name_cache_list); } +/* + * Remove some entries from the beginning of name_cache_list. + */ static void name_cache_clean_unused(struct send_ctx *sctx) { struct name_cache_entry *nce; @@ -1814,13 +1867,23 @@ static void name_cache_clean_unused(struct send_ctx *sctx) static void name_cache_free(struct send_ctx *sctx) { struct name_cache_entry *nce; - struct name_cache_entry *tmp; - list_for_each_entry_safe(nce, tmp, &sctx->name_cache_list, list) { + while (!list_empty(&sctx->name_cache_list)) { + nce = list_entry(sctx->name_cache_list.next, + struct name_cache_entry, list); name_cache_delete(sctx, nce); + kfree(nce); } } +/* + * Used by get_cur_path for each ref up to the root. + * Returns 0 if it succeeded. + * Returns 1 if the inode is not existent or got overwritten. In that case, the + * name is an orphan name. This instructs get_cur_path to stop iterating. If 1 + * is returned, parent_ino/parent_gen are not guaranteed to be valid. + * Returns <0 in case of error. + */ static int __get_cur_name_and_parent(struct send_ctx *sctx, u64 ino, u64 gen, u64 *parent_ino, @@ -1832,6 +1895,11 @@ static int __get_cur_name_and_parent(struct send_ctx *sctx, struct btrfs_path *path = NULL; struct name_cache_entry *nce = NULL; + /* + * First check if we already did a call to this function with the same + * ino/gen. If yes, check if the cache entry is still up-to-date. If yes + * return the cached result. + */ nce = name_cache_search(sctx, ino, gen); if (nce) { if (ino < sctx->send_progress && nce->need_later_update) { @@ -1854,6 +1922,11 @@ static int __get_cur_name_and_parent(struct send_ctx *sctx, if (!path) return -ENOMEM; + /* + * If the inode is not existent yet, add the orphan name and return 1. + * This should only happen for the parent dir that we determine in + * __record_new_ref + */ ret = is_inode_existent(sctx, ino, gen); if (ret < 0) goto out; @@ -1866,6 +1939,10 @@ static int __get_cur_name_and_parent(struct send_ctx *sctx, goto out_cache; } + /* + * Depending on whether the inode was already processed or not, use + * send_root or parent_root for ref lookup. + */ if (ino < sctx->send_progress) ret = get_first_ref(sctx, sctx->send_root, ino, parent_ino, parent_gen, dest); @@ -1875,6 +1952,10 @@ static int __get_cur_name_and_parent(struct send_ctx *sctx, if (ret < 0) goto out; + /* + * Check if the ref was overwritten by an inode's ref that was processed + * earlier. If yes, treat as orphan and return 1. + */ ret = did_overwrite_ref(sctx, *parent_ino, *parent_gen, ino, gen, dest->start, dest->end - dest->start); if (ret < 0) @@ -1888,6 +1969,9 @@ static int __get_cur_name_and_parent(struct send_ctx *sctx, } out_cache: + /* + * Store the result of the lookup in the name cache. + */ nce = kmalloc(sizeof(*nce) + fs_path_len(dest) + 1, GFP_NOFS); if (!nce) { ret = -ENOMEM; @@ -1901,7 +1985,6 @@ out_cache: nce->name_len = fs_path_len(dest); nce->ret = ret; strcpy(nce->name, dest->start); - memset(&nce->use_list, 0, sizeof(nce->use_list)); if (ino < sctx->send_progress) nce->need_later_update = 0; @@ -2107,9 +2190,6 @@ static int send_subvol_begin(struct send_ctx *sctx) read_extent_buffer(leaf, name, (unsigned long)(ref + 1), namelen); btrfs_release_path(path); - if (ret < 0) - goto out; - if (parent_root) { ret = begin_cmd(sctx, BTRFS_SEND_C_SNAPSHOT); if (ret < 0) @@ -2276,7 +2356,7 @@ verbose_printk("btrfs: send_utimes %llu\n", ino); btrfs_inode_mtime(ii)); TLV_PUT_BTRFS_TIMESPEC(sctx, BTRFS_SEND_A_CTIME, eb, btrfs_inode_ctime(ii)); - /* TODO otime? */ + /* TODO Add otime support when the otime patches get into upstream */ ret = send_cmd(sctx); @@ -2292,39 +2372,39 @@ out: * a valid path yet because we did not process the refs yet. So, the inode * is created as orphan. */ -static int send_create_inode(struct send_ctx *sctx, struct btrfs_path *path, - struct btrfs_key *key) +static int send_create_inode(struct send_ctx *sctx, u64 ino) { int ret = 0; - struct extent_buffer *eb = path->nodes[0]; - struct btrfs_inode_item *ii; struct fs_path *p; - int slot = path->slots[0]; int cmd; + u64 gen; u64 mode; + u64 rdev; -verbose_printk("btrfs: send_create_inode %llu\n", sctx->cur_ino); +verbose_printk("btrfs: send_create_inode %llu\n", ino); p = fs_path_alloc(sctx); if (!p) return -ENOMEM; - ii = btrfs_item_ptr(eb, slot, struct btrfs_inode_item); - mode = btrfs_inode_mode(eb, ii); + ret = get_inode_info(sctx->send_root, ino, NULL, &gen, &mode, NULL, + NULL, &rdev); + if (ret < 0) + goto out; - if (S_ISREG(mode)) + if (S_ISREG(mode)) { cmd = BTRFS_SEND_C_MKFILE; - else if (S_ISDIR(mode)) + } else if (S_ISDIR(mode)) { cmd = BTRFS_SEND_C_MKDIR; - else if (S_ISLNK(mode)) + } else if (S_ISLNK(mode)) { cmd = BTRFS_SEND_C_SYMLINK; - else if (S_ISCHR(mode) || S_ISBLK(mode)) + } else if (S_ISCHR(mode) || S_ISBLK(mode)) { cmd = BTRFS_SEND_C_MKNOD; - else if (S_ISFIFO(mode)) + } else if (S_ISFIFO(mode)) { cmd = BTRFS_SEND_C_MKFIFO; - else if (S_ISSOCK(mode)) + } else if (S_ISSOCK(mode)) { cmd = BTRFS_SEND_C_MKSOCK; - else { + } else { printk(KERN_WARNING "btrfs: unexpected inode type %o", (int)(mode & S_IFMT)); ret = -ENOTSUPP; @@ -2335,22 +2415,22 @@ verbose_printk("btrfs: send_create_inode %llu\n", sctx->cur_ino); if (ret < 0) goto out; - ret = gen_unique_name(sctx, sctx->cur_ino, sctx->cur_inode_gen, p); + ret = gen_unique_name(sctx, ino, gen, p); if (ret < 0) goto out; TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p); - TLV_PUT_U64(sctx, BTRFS_SEND_A_INO, sctx->cur_ino); + TLV_PUT_U64(sctx, BTRFS_SEND_A_INO, ino); if (S_ISLNK(mode)) { fs_path_reset(p); - ret = read_symlink(sctx, sctx->send_root, sctx->cur_ino, p); + ret = read_symlink(sctx, sctx->send_root, ino, p); if (ret < 0) goto out; TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH_LINK, p); } else if (S_ISCHR(mode) || S_ISBLK(mode) || S_ISFIFO(mode) || S_ISSOCK(mode)) { - TLV_PUT_U64(sctx, BTRFS_SEND_A_RDEV, btrfs_inode_rdev(eb, ii)); + TLV_PUT_U64(sctx, BTRFS_SEND_A_RDEV, rdev); } ret = send_cmd(sctx); @@ -2364,6 +2444,92 @@ out: return ret; } +/* + * We need some special handling for inodes that get processed before the parent + * directory got created. See process_recorded_refs for details. + * This function does the check if we already created the dir out of order. + */ +static int did_create_dir(struct send_ctx *sctx, u64 dir) +{ + int ret = 0; + struct btrfs_path *path = NULL; + struct btrfs_key key; + struct btrfs_key found_key; + struct btrfs_key di_key; + struct extent_buffer *eb; + struct btrfs_dir_item *di; + int slot; + + path = alloc_path_for_send(); + if (!path) { + ret = -ENOMEM; + goto out; + } + + key.objectid = dir; + key.type = BTRFS_DIR_INDEX_KEY; + key.offset = 0; + while (1) { + ret = btrfs_search_slot_for_read(sctx->send_root, &key, path, + 1, 0); + if (ret < 0) + goto out; + if (!ret) { + eb = path->nodes[0]; + slot = path->slots[0]; + btrfs_item_key_to_cpu(eb, &found_key, slot); + } + if (ret || found_key.objectid != key.objectid || + found_key.type != key.type) { + ret = 0; + goto out; + } + + di = btrfs_item_ptr(eb, slot, struct btrfs_dir_item); + btrfs_dir_item_key_to_cpu(eb, di, &di_key); + + if (di_key.objectid < sctx->send_progress) { + ret = 1; + goto out; + } + + key.offset = found_key.offset + 1; + btrfs_release_path(path); + } + +out: + btrfs_free_path(path); + return ret; +} + +/* + * Only creates the inode if it is: + * 1. Not a directory + * 2. Or a directory which was not created already due to out of order + * directories. See did_create_dir and process_recorded_refs for details. + */ +static int send_create_inode_if_needed(struct send_ctx *sctx) +{ + int ret; + + if (S_ISDIR(sctx->cur_inode_mode)) { + ret = did_create_dir(sctx, sctx->cur_ino); + if (ret < 0) + goto out; + if (ret) { + ret = 0; + goto out; + } + } + + ret = send_create_inode(sctx, sctx->cur_ino); + if (ret < 0) + goto out; + +out: + return ret; +} + struct recorded_ref { struct list_head list; char *dir_path; @@ -2416,13 +2582,13 @@ static int record_ref(struct list_head *head, u64 dir, static void __free_recorded_refs(struct send_ctx *sctx, struct list_head *head) { struct recorded_ref *cur; - struct recorded_ref *tmp; - list_for_each_entry_safe(cur, tmp, head, list) { + while (!list_empty(head)) { + cur = list_entry(head->next, struct recorded_ref, list); fs_path_free(sctx, cur->full_path); + list_del(&cur->list); kfree(cur); } - INIT_LIST_HEAD(head); } static void free_recorded_refs(struct send_ctx *sctx) @@ -2432,7 +2598,7 @@ static void free_recorded_refs(struct send_ctx *sctx) } /* - * Renames/moves a file/dir to it's orphan name. Used when the first + * Renames/moves a file/dir to its orphan name. Used when the first * ref of an unprocessed inode gets overwritten and for all non empty * directories. */ @@ -2472,6 +2638,12 @@ static int can_rmdir(struct send_ctx *sctx, u64 dir, u64 send_progress) struct btrfs_key loc; struct btrfs_dir_item *di; + /* + * Don't try to rmdir the top/root subvolume dir. + */ + if (dir == BTRFS_FIRST_FREE_OBJECTID) + return 0; + path = alloc_path_for_send(); if (!path) return -ENOMEM; @@ -2513,160 +2685,6 @@ out: return ret; } -struct finish_unordered_dir_ctx { - struct send_ctx *sctx; - struct fs_path *cur_path; - struct fs_path *dir_path; - u64 dir_ino; - int need_delete; - int delete_pass; -}; - -int __finish_unordered_dir(int num, struct btrfs_key *di_key, - const char *name, int name_len, - const char *data, int data_len, - u8 type, void *ctx) -{ - int ret = 0; - struct finish_unordered_dir_ctx *fctx = ctx; - struct send_ctx *sctx = fctx->sctx; - u64 di_gen; - u64 di_mode; - int is_orphan = 0; - - if (di_key->objectid >= fctx->dir_ino) - goto out; - - fs_path_reset(fctx->cur_path); - - ret = get_inode_info(sctx->send_root, di_key->objectid, - NULL, &di_gen, &di_mode, NULL, NULL); - if (ret < 0) - goto out; - - ret = is_first_ref(sctx, sctx->send_root, di_key->objectid, - fctx->dir_ino, name, name_len); - if (ret < 0) - goto out; - if (ret) { - is_orphan = 1; - ret = gen_unique_name(sctx, di_key->objectid, di_gen, - fctx->cur_path); - } else { - ret = get_cur_path(sctx, di_key->objectid, di_gen, - fctx->cur_path); - } - if (ret < 0) - goto out; - - ret = fs_path_add(fctx->dir_path, name, name_len); - if (ret < 0) - goto out; - - if (!fctx->delete_pass) { - if (S_ISDIR(di_mode)) { - ret = send_rename(sctx, fctx->cur_path, - fctx->dir_path); - } else { - ret = send_link(sctx, fctx->dir_path, - fctx->cur_path); - if (is_orphan) - fctx->need_delete = 1; - } - } else if (!S_ISDIR(di_mode)) { - ret = send_unlink(sctx, fctx->cur_path); - } else { - ret = 0; - } - - fs_path_remove(fctx->dir_path); - -out: - return ret; -} - -/* - * Go through all dir items and see if we find refs which could not be created - * in the past because the dir did not exist at that time. - */ -static int finish_outoforder_dir(struct send_ctx *sctx, u64 dir, u64 dir_gen) -{ - int ret = 0; - struct btrfs_path *path = NULL; - struct btrfs_key key; - struct btrfs_key found_key; - struct extent_buffer *eb; - struct finish_unordered_dir_ctx fctx; - int slot; - - path = alloc_path_for_send(); - if (!path) { - ret = -ENOMEM; - goto out; - } - - memset(&fctx, 0, sizeof(fctx)); - fctx.sctx = sctx; - fctx.cur_path = fs_path_alloc(sctx); - fctx.dir_path = fs_path_alloc(sctx); - if (!fctx.cur_path || !fctx.dir_path) { - ret = -ENOMEM; - goto out; - } - fctx.dir_ino = dir; - - ret = get_cur_path(sctx, dir, dir_gen, fctx.dir_path); - if (ret < 0) - goto out; - - /* - * We do two passes. The first links in the new refs and the second - * deletes orphans if required. Deletion of orphans is not required for - * directory inodes, as we always have only one ref and use rename - * instead of link for those. - */ - -again: - key.objectid = dir; - key.type = BTRFS_DIR_ITEM_KEY; - key.offset = 0; - while (1) { - ret = btrfs_search_slot_for_read(sctx->send_root, &key, path, - 1, 0); - if (ret < 0) - goto out; - eb = path->nodes[0]; - slot = path->slots[0]; - btrfs_item_key_to_cpu(eb, &found_key, slot); - - if (found_key.objectid != key.objectid || - found_key.type != key.type) { - btrfs_release_path(path); - break; - } - - ret = iterate_dir_item(sctx, sctx->send_root, path, - &found_key, __finish_unordered_dir, - &fctx); - if (ret < 0) - goto out; - - key.offset = found_key.offset + 1; - btrfs_release_path(path); - } - - if (!fctx.delete_pass && fctx.need_delete) { - fctx.delete_pass = 1; - goto again; - } - -out: - btrfs_free_path(path); - fs_path_free(sctx, fctx.cur_path); - fs_path_free(sctx, fctx.dir_path); - return ret; -} - /* * This does all the move/link/unlink/rmdir magic. */ @@ -2674,6 +2692,7 @@ static int process_recorded_refs(struct send_ctx *sctx) { int ret = 0; struct recorded_ref *cur; + struct recorded_ref *cur2; struct ulist *check_dirs = NULL; struct ulist_iterator uit; struct ulist_node *un; @@ -2685,6 +2704,12 @@ static int process_recorded_refs(struct send_ctx *sctx) verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino); + /* + * This should never happen as the root dir always has the same ref + * which is always '..' + */ + BUG_ON(sctx->cur_ino <= BTRFS_FIRST_FREE_OBJECTID); + valid_path = fs_path_alloc(sctx); if (!valid_path) { ret = -ENOMEM; @@ -2731,6 +2756,46 @@ verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino); list_for_each_entry(cur, &sctx->new_refs, list) { /* + * We may have refs where the parent directory does not exist + * yet. This happens if the parent directories inum is higher + * the the current inum. To handle this case, we create the + * parent directory out of order. But we need to check if this + * did already happen before due to other refs in the same dir. + */ + ret = get_cur_inode_state(sctx, cur->dir, cur->dir_gen); + if (ret < 0) + goto out; + if (ret == inode_state_will_create) { + ret = 0; + /* + * First check if any of the current inodes refs did + * already create the dir. + */ + list_for_each_entry(cur2, &sctx->new_refs, list) { + if (cur == cur2) + break; + if (cur2->dir == cur->dir) { + ret = 1; + break; + } + } + + /* + * If that did not happen, check if a previous inode + * did already create the dir. + */ + if (!ret) + ret = did_create_dir(sctx, cur->dir); + if (ret < 0) + goto out; + if (!ret) { + ret = send_create_inode(sctx, cur->dir); + if (ret < 0) + goto out; + } + } + + /* * Check if this new ref would overwrite the first ref of * another unprocessed inode. If yes, orphanize the * overwritten inode. If we find an overwritten ref that is @@ -2764,7 +2829,7 @@ verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino); * inode, move it and update valid_path. If not, link or move * it depending on the inode mode. */ - if (is_orphan && !sctx->cur_inode_first_ref_orphan) { + if (is_orphan) { ret = send_rename(sctx, valid_path, cur->full_path); if (ret < 0) goto out; @@ -2827,6 +2892,17 @@ verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino); if (ret < 0) goto out; } + } else if (S_ISDIR(sctx->cur_inode_mode) && + !list_empty(&sctx->deleted_refs)) { + /* + * We have a moved dir. Add the old parent to check_dirs + */ + cur = list_entry(sctx->deleted_refs.next, struct recorded_ref, + list); + ret = ulist_add(check_dirs, cur->dir, cur->dir_gen, + GFP_NOFS); + if (ret < 0) + goto out; } else if (!S_ISDIR(sctx->cur_inode_mode)) { /* * We have a non dir inode. Go through all deleted refs and @@ -2840,35 +2916,9 @@ verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino); if (ret < 0) goto out; if (!ret) { - /* - * In case the inode was moved to a directory - * that was not created yet (see - * __record_new_ref), we can not unlink the ref - * as it will be needed later when the parent - * directory is created, so that we can move in - * the inode to the new dir. - */ - if (!is_orphan && - sctx->cur_inode_first_ref_orphan) { - ret = orphanize_inode(sctx, - sctx->cur_ino, - sctx->cur_inode_gen, - cur->full_path); - if (ret < 0) - goto out; - ret = gen_unique_name(sctx, - sctx->cur_ino, - sctx->cur_inode_gen, - valid_path); - if (ret < 0) - goto out; - is_orphan = 1; - - } else { - ret = send_unlink(sctx, cur->full_path); - if (ret < 0) - goto out; - } + ret = send_unlink(sctx, cur->full_path); + if (ret < 0) + goto out; } ret = ulist_add(check_dirs, cur->dir, cur->dir_gen, GFP_NOFS); @@ -2880,12 +2930,11 @@ verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino); * If the inode is still orphan, unlink the orphan. This may * happen when a previous inode did overwrite the first ref * of this inode and no new refs were added for the current - * inode. - * We can however not delete the orphan in case the inode relies - * in a directory that was not created yet (see - * __record_new_ref) + * inode. Unlinking does not mean that the inode is deleted in + * all cases. There may still be links to this inode in other + * places. */ - if (is_orphan && !sctx->cur_inode_first_ref_orphan) { + if (is_orphan) { ret = send_unlink(sctx, valid_path); if (ret < 0) goto out; @@ -2900,6 +2949,11 @@ verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino); */ ULIST_ITER_INIT(&uit); while ((un = ulist_next(check_dirs, &uit))) { + /* + * In case we had refs into dirs that were not processed yet, + * we don't need to do the utime and rmdir logic for these dirs. + * The dir will be processed later. + */ if (un->val > sctx->cur_ino) continue; @@ -2929,25 +2983,6 @@ verbose_printk("btrfs: process_recorded_refs %llu\n", sctx->cur_ino); } } - /* - * Current inode is now at it's new position, so we must increase - * send_progress - */ - sctx->send_progress = sctx->cur_ino + 1; - - /* - * We may have a directory here that has pending refs which could not - * be created before (because the dir did not exist before, see - * __record_new_ref). finish_outoforder_dir will link/move the pending - * refs. - */ - if (S_ISDIR(sctx->cur_inode_mode) && sctx->cur_inode_new) { - ret = finish_outoforder_dir(sctx, sctx->cur_ino, - sctx->cur_inode_gen); - if (ret < 0) - goto out; - } - ret = 0; out: @@ -2971,34 +3006,9 @@ static int __record_new_ref(int num, u64 dir, int index, return -ENOMEM; ret = get_inode_info(sctx->send_root, dir, NULL, &gen, NULL, NULL, - NULL); - if (ret < 0) - goto out; - - /* - * The parent may be non-existent at this point in time. This happens - * if the ino of the parent dir is higher then the current ino. In this - * case, we can not process this ref until the parent dir is finally - * created. If we reach the parent dir later, process_recorded_refs - * will go through all dir items and process the refs that could not be - * processed before. In case this is the first ref, we set - * cur_inode_first_ref_orphan to 1 to inform process_recorded_refs to - * keep an orphan of the inode so that it later can be used for - * link/move - */ - ret = is_inode_existent(sctx, dir, gen); + NULL, NULL); if (ret < 0) goto out; - if (!ret) { - ret = is_first_ref(sctx, sctx->send_root, sctx->cur_ino, dir, - name->start, fs_path_len(name)); - if (ret < 0) - goto out; - if (ret) - sctx->cur_inode_first_ref_orphan = 1; - ret = 0; - goto out; - } ret = get_cur_path(sctx, dir, gen, p); if (ret < 0) @@ -3029,7 +3039,7 @@ static int __record_deleted_ref(int num, u64 dir, int index, return -ENOMEM; ret = get_inode_info(sctx->parent_root, dir, NULL, &gen, NULL, NULL, - NULL); + NULL, NULL); if (ret < 0) goto out; @@ -3206,33 +3216,28 @@ static int process_all_refs(struct send_ctx *sctx, key.offset = 0; while (1) { ret = btrfs_search_slot_for_read(root, &key, path, 1, 0); - if (ret < 0) { - btrfs_release_path(path); + if (ret < 0) goto out; - } - if (ret) { - btrfs_release_path(path); + if (ret) break; - } eb = path->nodes[0]; slot = path->slots[0]; btrfs_item_key_to_cpu(eb, &found_key, slot); if (found_key.objectid != key.objectid || - found_key.type != key.type) { - btrfs_release_path(path); + found_key.type != key.type) break; - } - ret = iterate_inode_ref(sctx, sctx->parent_root, path, - &found_key, 0, cb, sctx); + ret = iterate_inode_ref(sctx, root, path, &found_key, 0, cb, + sctx); btrfs_release_path(path); if (ret < 0) goto out; key.offset = found_key.offset + 1; } + btrfs_release_path(path); ret = process_recorded_refs(sctx); @@ -3555,7 +3560,7 @@ static int send_write(struct send_ctx *sctx, u64 offset, u32 len) int ret = 0; struct fs_path *p; loff_t pos = offset; - int readed = 0; + int num_read = 0; mm_segment_t old_fs; p = fs_path_alloc(sctx); @@ -3580,8 +3585,8 @@ verbose_printk("btrfs: send_write offset=%llu, len=%d\n", offset, len); ret = vfs_read(sctx->cur_inode_filp, sctx->read_buf, len, &pos); if (ret < 0) goto out; - readed = ret; - if (!readed) + num_read = ret; + if (!num_read) goto out; ret = begin_cmd(sctx, BTRFS_SEND_C_WRITE); @@ -3594,7 +3599,7 @@ verbose_printk("btrfs: send_write offset=%llu, len=%d\n", offset, len); TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p); TLV_PUT_U64(sctx, BTRFS_SEND_A_FILE_OFFSET, offset); - TLV_PUT(sctx, BTRFS_SEND_A_DATA, sctx->read_buf, readed); + TLV_PUT(sctx, BTRFS_SEND_A_DATA, sctx->read_buf, num_read); ret = send_cmd(sctx); @@ -3604,7 +3609,7 @@ out: set_fs(old_fs); if (ret < 0) return ret; - return readed; + return num_read; } /* @@ -3615,7 +3620,6 @@ static int send_clone(struct send_ctx *sctx, struct clone_root *clone_root) { int ret = 0; - struct btrfs_root *clone_root2 = clone_root->root; struct fs_path *p; u64 gen; @@ -3640,22 +3644,23 @@ verbose_printk("btrfs: send_clone offset=%llu, len=%d, clone_root=%llu, " TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_LEN, len); TLV_PUT_PATH(sctx, BTRFS_SEND_A_PATH, p); - if (clone_root2 == sctx->send_root) { + if (clone_root->root == sctx->send_root) { ret = get_inode_info(sctx->send_root, clone_root->ino, NULL, - &gen, NULL, NULL, NULL); + &gen, NULL, NULL, NULL, NULL); if (ret < 0) goto out; ret = get_cur_path(sctx, clone_root->ino, gen, p); } else { - ret = get_inode_path(sctx, clone_root2, clone_root->ino, p); + ret = get_inode_path(sctx, clone_root->root, + clone_root->ino, p); } if (ret < 0) goto out; TLV_PUT_UUID(sctx, BTRFS_SEND_A_CLONE_UUID, - clone_root2->root_item.uuid); + clone_root->root->root_item.uuid); TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_CTRANSID, - clone_root2->root_item.ctransid); + clone_root->root->root_item.ctransid); TLV_PUT_PATH(sctx, BTRFS_SEND_A_CLONE_PATH, p); TLV_PUT_U64(sctx, BTRFS_SEND_A_CLONE_OFFSET, clone_root->offset); @@ -3684,10 +3689,17 @@ static int send_write_or_clone(struct send_ctx *sctx, ei = btrfs_item_ptr(path->nodes[0], path->slots[0], struct btrfs_file_extent_item); type = btrfs_file_extent_type(path->nodes[0], ei); - if (type == BTRFS_FILE_EXTENT_INLINE) + if (type == BTRFS_FILE_EXTENT_INLINE) { len = btrfs_file_extent_inline_len(path->nodes[0], ei); - else + /* + * it is possible the inline item won't cover the whole page, + * but there may be items after this page. Make + * sure to send the whole thing + */ + len = PAGE_CACHE_ALIGN(len); + } else { len = btrfs_file_extent_num_bytes(path->nodes[0], ei); + } if (offset + len > sctx->cur_inode_size) len = sctx->cur_inode_size - offset; @@ -3735,6 +3747,8 @@ static int is_extent_unchanged(struct send_ctx *sctx, u64 left_offset_fixed; u64 left_len; u64 right_len; + u64 left_gen; + u64 right_gen; u8 left_type; u8 right_type; @@ -3744,17 +3758,17 @@ static int is_extent_unchanged(struct send_ctx *sctx, eb = left_path->nodes[0]; slot = left_path->slots[0]; - ei = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item); left_type = btrfs_file_extent_type(eb, ei); - left_disknr = btrfs_file_extent_disk_bytenr(eb, ei); - left_len = btrfs_file_extent_num_bytes(eb, ei); - left_offset = btrfs_file_extent_offset(eb, ei); if (left_type != BTRFS_FILE_EXTENT_REG) { ret = 0; goto out; } + left_disknr = btrfs_file_extent_disk_bytenr(eb, ei); + left_len = btrfs_file_extent_num_bytes(eb, ei); + left_offset = btrfs_file_extent_offset(eb, ei); + left_gen = btrfs_file_extent_generation(eb, ei); /* * Following comments will refer to these graphics. L is the left @@ -3810,6 +3824,7 @@ static int is_extent_unchanged(struct send_ctx *sctx, right_disknr = btrfs_file_extent_disk_bytenr(eb, ei); right_len = btrfs_file_extent_num_bytes(eb, ei); right_offset = btrfs_file_extent_offset(eb, ei); + right_gen = btrfs_file_extent_generation(eb, ei); if (right_type != BTRFS_FILE_EXTENT_REG) { ret = 0; @@ -3820,7 +3835,7 @@ static int is_extent_unchanged(struct send_ctx *sctx, * Are we at extent 8? If yes, we know the extent is changed. * This may only happen on the first iteration. */ - if (found_key.offset + right_len < ekey->offset) { + if (found_key.offset + right_len <= ekey->offset) { ret = 0; goto out; } @@ -3837,8 +3852,9 @@ static int is_extent_unchanged(struct send_ctx *sctx, /* * Check if we have the same extent. */ - if (left_disknr + left_offset_fixed != - right_disknr + right_offset) { + if (left_disknr != right_disknr || + left_offset_fixed != right_offset || + left_gen != right_gen) { ret = 0; goto out; } @@ -3977,6 +3993,15 @@ static int process_recorded_refs_if_needed(struct send_ctx *sctx, int at_end) goto out; ret = process_recorded_refs(sctx); + if (ret < 0) + goto out; + + /* + * We have processed the refs and thus need to advance send_progress. + * Now, calls to get_cur_xxx will take the updated refs of the current + * inode into account. + */ + sctx->send_progress = sctx->cur_ino + 1; out: return ret; @@ -4004,7 +4029,7 @@ static int finish_inode_if_needed(struct send_ctx *sctx, int at_end) goto out; ret = get_inode_info(sctx->send_root, sctx->cur_ino, NULL, NULL, - &left_mode, &left_uid, &left_gid); + &left_mode, &left_uid, &left_gid, NULL); if (ret < 0) goto out; @@ -4015,7 +4040,7 @@ static int finish_inode_if_needed(struct send_ctx *sctx, int at_end) } else { ret = get_inode_info(sctx->parent_root, sctx->cur_ino, NULL, NULL, &right_mode, &right_uid, - &right_gid); + &right_gid, NULL); if (ret < 0) goto out; @@ -4074,7 +4099,12 @@ static int changed_inode(struct send_ctx *sctx, sctx->cur_ino = key->objectid; sctx->cur_inode_new_gen = 0; - sctx->cur_inode_first_ref_orphan = 0; + + /* + * Set send_progress to current inode. This will tell all get_cur_xxx + * functions that the current inode's refs are not updated yet. Later, + * when process_recorded_refs is finished, it is set to cur_ino + 1. + */ sctx->send_progress = sctx->cur_ino; if (result == BTRFS_COMPARE_TREE_NEW || @@ -4098,7 +4128,14 @@ static int changed_inode(struct send_ctx *sctx, right_gen = btrfs_inode_generation(sctx->right_path->nodes[0], right_ii); - if (left_gen != right_gen) + + /* + * The cur_ino = root dir case is special here. We can't treat + * the inode as deleted+reused because it would generate a + * stream that tries to delete/mkdir the root dir. + */ + if (left_gen != right_gen && + sctx->cur_ino != BTRFS_FIRST_FREE_OBJECTID) sctx->cur_inode_new_gen = 1; } @@ -4111,8 +4148,7 @@ static int changed_inode(struct send_ctx *sctx, sctx->cur_inode_mode = btrfs_inode_mode( sctx->left_path->nodes[0], left_ii); if (sctx->cur_ino != BTRFS_FIRST_FREE_OBJECTID) - ret = send_create_inode(sctx, sctx->left_path, - sctx->cmp_key); + ret = send_create_inode_if_needed(sctx); } else if (result == BTRFS_COMPARE_TREE_DELETED) { sctx->cur_inode_gen = right_gen; sctx->cur_inode_new = 0; @@ -4122,7 +4158,17 @@ static int changed_inode(struct send_ctx *sctx, sctx->cur_inode_mode = btrfs_inode_mode( sctx->right_path->nodes[0], right_ii); } else if (result == BTRFS_COMPARE_TREE_CHANGED) { + /* + * We need to do some special handling in case the inode was + * reported as changed with a changed generation number. This + * means that the original inode was deleted and new inode + * reused the same inum. So we have to treat the old inode as + * deleted and the new one as new. + */ if (sctx->cur_inode_new_gen) { + /* + * First, process the inode as if it was deleted. + */ sctx->cur_inode_gen = right_gen; sctx->cur_inode_new = 0; sctx->cur_inode_deleted = 1; @@ -4135,6 +4181,9 @@ static int changed_inode(struct send_ctx *sctx, if (ret < 0) goto out; + /* + * Now process the inode as if it was new. + */ sctx->cur_inode_gen = left_gen; sctx->cur_inode_new = 1; sctx->cur_inode_deleted = 0; @@ -4142,14 +4191,23 @@ static int changed_inode(struct send_ctx *sctx, sctx->left_path->nodes[0], left_ii); sctx->cur_inode_mode = btrfs_inode_mode( sctx->left_path->nodes[0], left_ii); - ret = send_create_inode(sctx, sctx->left_path, - sctx->cmp_key); + ret = send_create_inode_if_needed(sctx); if (ret < 0) goto out; ret = process_all_refs(sctx, BTRFS_COMPARE_TREE_NEW); if (ret < 0) goto out; + /* + * Advance send_progress now as we did not get into + * process_recorded_refs_if_needed in the new_gen case. + */ + sctx->send_progress = sctx->cur_ino + 1; + + /* + * Now process all extents and xattrs of the inode as if + * they were all new. + */ ret = process_all_extents(sctx); if (ret < 0) goto out; @@ -4172,6 +4230,16 @@ out: return ret; } +/* + * We have to process new refs before deleted refs, but compare_trees gives us + * the new and deleted refs mixed. To fix this, we record the new/deleted refs + * first and later process them in process_recorded_refs. + * For the cur_inode_new_gen case, we skip recording completely because + * changed_inode did already initiate processing of refs. The reason for this is + * that in this case, compare_tree actually compares the refs of 2 different + * inodes. To fix this, process_all_refs is used in changed_inode to handle all + * refs of the right tree as deleted and all refs of the left tree as new. + */ static int changed_ref(struct send_ctx *sctx, enum btrfs_compare_tree_result result) { @@ -4192,6 +4260,11 @@ static int changed_ref(struct send_ctx *sctx, return ret; } +/* + * Process new/deleted/changed xattrs. We skip processing in the + * cur_inode_new_gen case because changed_inode did already initiate processing + * of xattrs. The reason is the same as in changed_ref + */ static int changed_xattr(struct send_ctx *sctx, enum btrfs_compare_tree_result result) { @@ -4211,6 +4284,11 @@ static int changed_xattr(struct send_ctx *sctx, return ret; } +/* + * Process new/deleted/changed extents. We skip processing in the + * cur_inode_new_gen case because changed_inode did already initiate processing + * of extents. The reason is the same as in changed_ref + */ static int changed_extent(struct send_ctx *sctx, enum btrfs_compare_tree_result result) { @@ -4227,7 +4305,10 @@ static int changed_extent(struct send_ctx *sctx, return ret; } - +/* + * Updates compare related fields in sctx and simply forwards to the actual + * changed_xxx functions. + */ static int changed_cb(struct btrfs_root *left_root, struct btrfs_root *right_root, struct btrfs_path *left_path, @@ -4247,6 +4328,11 @@ static int changed_cb(struct btrfs_root *left_root, if (ret < 0) goto out; + /* Ignore non-FS objects */ + if (key->objectid == BTRFS_FREE_INO_OBJECTID || + key->objectid == BTRFS_FREE_SPACE_OBJECTID) + goto out; + if (key->type == BTRFS_INODE_ITEM_KEY) ret = changed_inode(sctx, result); else if (key->type == BTRFS_INODE_REF_KEY) @@ -4299,7 +4385,8 @@ join_trans: } /* - * Make sure the tree has not changed + * Make sure the tree has not changed after re-joining. We detect this + * by comparing start_ctransid and ctransid. They should always match. */ spin_lock(&send_root->root_times_lock); ctransid = btrfs_root_ctransid(&send_root->root_item); |