X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fntfs-3g_capture.c;h=80abca02c8143864bb3493ed6eeea15930c4add7;hp=10d4bcc5a02e709f68ac44aa3bf1083053932687;hb=b5b9681794d1f5f13350e3567f6f6e74f5c779cf;hpb=52a908ca68284a4dfa89a884f547fd78d1543772 diff --git a/src/ntfs-3g_capture.c b/src/ntfs-3g_capture.c index 10d4bcc5..80abca02 100644 --- a/src/ntfs-3g_capture.c +++ b/src/ntfs-3g_capture.c @@ -1,12 +1,12 @@ /* * ntfs-3g_capture.c * - * Capture a WIM image directly from a NTFS volume using libntfs-3g. We capture + * Capture a WIM image directly from an NTFS volume using libntfs-3g. We capture * everything we can, including security data and alternate data streams. */ /* - * Copyright (C) 2012, 2013 Eric Biggers + * Copyright (C) 2012, 2013, 2014 Eric Biggers * * This file is part of wimlib, a library for working with WIM files. * @@ -75,11 +75,8 @@ open_ntfs_attr(ntfs_inode *ni, struct ntfs_location *loc) } int -read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte, - u64 size, - consume_data_callback_t cb, - void *ctx_or_buf, - int _ignored_flags) +read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte, u64 size, + consume_data_callback_t cb, void *cb_ctx) { struct ntfs_location *loc = lte->ntfs_loc; ntfs_volume *vol = loc->ntfs_vol; @@ -87,10 +84,10 @@ read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte, ntfs_attr *na; s64 pos; s64 bytes_remaining; - void *out_buf; int ret; + u8 buf[BUFFER_SIZE]; - ni = ntfs_pathname_to_inode(vol, NULL, loc->path); + ni = ntfs_pathname_to_inode(vol, NULL, loc->path); if (!ni) { ERROR_WITH_ERRNO("Can't find NTFS inode for \"%"TS"\"", loc->path); ret = WIMLIB_ERR_NTFS_3G; @@ -103,28 +100,20 @@ read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte, goto out_close_ntfs_inode; } - if (cb) - out_buf = alloca(WIM_CHUNK_SIZE); - else - out_buf = ctx_or_buf; pos = (loc->is_reparse_point) ? 8 : 0; bytes_remaining = size; while (bytes_remaining) { - s64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE); - if (ntfs_attr_pread(na, pos, to_read, out_buf) != to_read) { + s64 to_read = min(bytes_remaining, sizeof(buf)); + if (ntfs_attr_pread(na, pos, to_read, buf) != to_read) { ERROR_WITH_ERRNO("Error reading \"%"TS"\"", loc->path); ret = WIMLIB_ERR_NTFS_3G; goto out_close_ntfs_attr; } pos += to_read; bytes_remaining -= to_read; - if (cb) { - ret = cb(out_buf, to_read, ctx_or_buf); - if (ret) - goto out_close_ntfs_attr; - } else { - out_buf += to_read; - } + ret = cb(buf, to_read, cb_ctx); + if (ret) + goto out_close_ntfs_attr; } ret = 0; out_close_ntfs_attr: @@ -166,14 +155,13 @@ out: } -/* Load the streams from a file or reparse point in the NTFS volume into the WIM - * lookup table */ +/* Load the streams from a file or reparse point in the NTFS volume */ static int capture_ntfs_streams(struct wim_inode *inode, ntfs_inode *ni, char *path, size_t path_len, - struct wim_lookup_table *lookup_table, + struct list_head *unhashed_streams, ntfs_volume *vol, ATTR_TYPES type) { @@ -182,7 +170,7 @@ capture_ntfs_streams(struct wim_inode *inode, int ret; struct wim_lookup_table_entry *lte; - DEBUG2("Capturing NTFS data streams from `%s'", path); + DEBUG("Capturing NTFS data streams from `%s'", path); /* Get context to search the streams of the NTFS file. */ actx = ntfs_attr_get_search_ctx(ni, NULL); @@ -242,14 +230,14 @@ capture_ntfs_streams(struct wim_inode *inode, goto out_free_lte; } lte->ntfs_loc->is_reparse_point = true; - lte->resource_entry.original_size = data_size - 8; + lte->size = data_size - 8; ret = read_reparse_tag(ni, lte->ntfs_loc, &inode->i_reparse_tag); if (ret) goto out_free_lte; } else { lte->ntfs_loc->is_reparse_point = false; - lte->resource_entry.original_size = data_size; + lte->size = data_size; } } if (name_length == 0) { @@ -257,12 +245,19 @@ capture_ntfs_streams(struct wim_inode *inode, * dentry's inode. */ if (inode->i_lte) { if (lte) { - ERROR("Found two un-named data streams for \"%s\" " - "(sizes = %"PRIu64", %"PRIu64")", - path, wim_resource_size(inode->i_lte), - wim_resource_size(lte)); - ret = WIMLIB_ERR_NTFS_3G; - goto out_free_lte; + if (!(inode->i_attributes & + FILE_ATTRIBUTE_REPARSE_POINT)) + { + WARNING("Found two un-named " + "data streams for \"%s\" " + "(sizes = %"PRIu64", " + "%"PRIu64")", + path, + inode->i_lte->size, + lte->size); + } + free_lookup_table_entry(lte); + continue; } } else { stream_id = 0; @@ -285,8 +280,8 @@ capture_ntfs_streams(struct wim_inode *inode, new_ads_entry->lte = lte; } if (lte) { - lookup_table_insert_unhashed(lookup_table, lte, - inode, stream_id); + add_unhashed_stream(lte, inode, + stream_id, unhashed_streams); } } if (errno == ENOENT) { @@ -307,33 +302,41 @@ out_free_ntfs_loc: out_put_actx: ntfs_attr_put_search_ctx(actx); if (ret == 0) - DEBUG2("Successfully captured NTFS streams from \"%s\"", path); + DEBUG("Successfully captured NTFS streams from \"%s\"", path); else ERROR("Failed to capture NTFS streams from \"%s\"", path); return ret; } -/* Red-black tree that maps NTFS inode numbers to DOS names */ +/* Binary tree that maps NTFS inode numbers to DOS names */ struct dos_name_map { - struct rb_root rb_root; + struct avl_tree_node *root; }; struct dos_name_node { - struct rb_node rb_node; + struct avl_tree_node index_node; char dos_name[24]; int name_nbytes; le64 ntfs_ino; }; +#define DOS_NAME_NODE(avl_node) \ + avl_tree_entry(avl_node, struct dos_name_node, index_node) + +static int +_avl_cmp_by_ntfs_ino(const struct avl_tree_node *n1, + const struct avl_tree_node *n2) +{ + return cmp_u64(DOS_NAME_NODE(n1)->ntfs_ino, + DOS_NAME_NODE(n2)->ntfs_ino); +} + /* Inserts a new DOS name into the map */ static int insert_dos_name(struct dos_name_map *map, const ntfschar *dos_name, size_t name_nbytes, le64 ntfs_ino) { struct dos_name_node *new_node; - struct rb_node **p; - struct rb_root *root; - struct rb_node *rb_parent; DEBUG("DOS name_len = %zu", name_nbytes); new_node = MALLOC(sizeof(struct dos_name_node)); @@ -345,73 +348,56 @@ insert_dos_name(struct dos_name_map *map, const ntfschar *dos_name, wimlib_assert(name_nbytes <= sizeof(new_node->dos_name)); /* Initialize the DOS name, DOS name length, and NTFS inode number of - * the red-black tree node */ + * the search tree node */ memcpy(new_node->dos_name, dos_name, name_nbytes); new_node->name_nbytes = name_nbytes; new_node->ntfs_ino = ntfs_ino; - /* Insert the red-black tree node */ - root = &map->rb_root; - p = &root->rb_node; - rb_parent = NULL; - while (*p) { - struct dos_name_node *this; - - this = container_of(*p, struct dos_name_node, rb_node); - rb_parent = *p; - if (new_node->ntfs_ino < this->ntfs_ino) - p = &((*p)->rb_left); - else if (new_node->ntfs_ino > this->ntfs_ino) - p = &((*p)->rb_right); - else { - /* This should be impossible since a NTFS inode cannot - * have multiple DOS names, and we only should get each - * DOS name entry once from the ntfs_readdir() calls. */ - ERROR("NTFS inode %"PRIu64" has multiple DOS names", - le64_to_cpu(ntfs_ino)); - return -1; - } + /* Insert the search tree node */ + if (avl_tree_insert(&map->root, &new_node->index_node, + _avl_cmp_by_ntfs_ino)) + { + /* This should be impossible since an NTFS inode cannot + * have multiple DOS names, and we only should get each + * DOS name entry once from the ntfs_readdir() calls. */ + ERROR("NTFS inode %"PRIu64" has multiple DOS names", + le64_to_cpu(ntfs_ino)); + FREE(new_node); + return -1; } - rb_link_node(&new_node->rb_node, rb_parent, p); - rb_insert_color(&new_node->rb_node, root); DEBUG("Inserted DOS name for inode %"PRIu64, le64_to_cpu(ntfs_ino)); return 0; } -/* Returns a structure that contains the DOS name and its length for a NTFS +/* Returns a structure that contains the DOS name and its length for an NTFS * inode, or NULL if the inode has no DOS name. */ static struct dos_name_node * lookup_dos_name(const struct dos_name_map *map, u64 ntfs_ino) { - struct rb_node *node = map->rb_root.rb_node; - while (node) { - struct dos_name_node *this; - this = container_of(node, struct dos_name_node, rb_node); - if (ntfs_ino < this->ntfs_ino) - node = node->rb_left; - else if (ntfs_ino > this->ntfs_ino) - node = node->rb_right; - else - return this; - } - return NULL; + struct dos_name_node dummy; + struct avl_tree_node *res; + + dummy.ntfs_ino = cpu_to_le64(ntfs_ino); + + res = avl_tree_lookup_node(map->root, &dummy.index_node, + _avl_cmp_by_ntfs_ino); + if (!res) + return NULL; + return DOS_NAME_NODE(res); } static int -set_dentry_dos_name(struct wim_dentry *dentry, void *arg) +set_dentry_dos_name(struct wim_dentry *dentry, const struct dos_name_map *map) { - const struct dos_name_map *map = arg; const struct dos_name_node *node; if (dentry->is_win32_name) { node = lookup_dos_name(map, dentry->d_inode->i_ino); if (node) { - dentry->short_name = MALLOC(node->name_nbytes + 2); + dentry->short_name = utf16le_dupz((const utf16lechar *)node->dos_name, + node->name_nbytes); if (!dentry->short_name) return WIMLIB_ERR_NOMEM; - memcpy(dentry->short_name, node->dos_name, - node->name_nbytes); - dentry->short_name[node->name_nbytes / 2] = 0; dentry->short_name_nbytes = node->name_nbytes; DEBUG("Assigned DOS name to ino %"PRIu64, dentry->d_inode->i_ino); @@ -425,18 +411,18 @@ set_dentry_dos_name(struct wim_dentry *dentry, void *arg) } static void -free_dos_name_tree(struct rb_node *node) { +free_dos_name_tree(struct avl_tree_node *node) { if (node) { - free_dos_name_tree(node->rb_left); - free_dos_name_tree(node->rb_right); - FREE(container_of(node, struct dos_name_node, rb_node)); + free_dos_name_tree(node->left); + free_dos_name_tree(node->right); + FREE(DOS_NAME_NODE(node)); } } static void destroy_dos_name_map(struct dos_name_map *map) { - free_dos_name_tree(map->rb_root.rb_node); + free_dos_name_tree(map->root); } struct readdir_ctx { @@ -530,7 +516,7 @@ out: return ret; } -/* Recursively build a WIM dentry tree corresponding to a NTFS volume. +/* Recursively build a WIM dentry tree corresponding to an NTFS volume. * At the same time, update the WIM lookup table with lookup table entries for * the NTFS streams, and build an array of security descriptors. */ @@ -545,56 +531,55 @@ build_dentry_tree_ntfs_recursive(struct wim_dentry **root_ret, { le32 attributes; int ret; - struct wim_dentry *root; - struct wim_inode *inode; - ATTR_TYPES stream_type; + struct wim_dentry *root = NULL; + struct wim_inode *inode = NULL; - if (exclude_path(path, path_len, params->config, false)) { + if (should_exclude_path(path, path_len, params->config)) { /* Exclude a file or directory tree based on the capture - * configuration file */ - if ((params->add_flags & WIMLIB_ADD_FLAG_EXCLUDE_VERBOSE) - && params->progress_func) - { - union wimlib_progress_info info; - info.scan.cur_path = path; - info.scan.excluded = true; - params->progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info); - } - root = NULL; + * configuration file. */ ret = 0; - goto out; + goto out_progress; } /* Get file attributes */ ret = ntfs_get_ntfs_attrib(ni, (char*)&attributes, sizeof(attributes)); if (ret != sizeof(attributes)) { ERROR_WITH_ERRNO("Failed to get NTFS attributes from \"%s\"", path); - return WIMLIB_ERR_NTFS_3G; + ret = WIMLIB_ERR_NTFS_3G; + goto out; } - if ((params->add_flags & WIMLIB_ADD_FLAG_VERBOSE) - && params->progress_func) + if ((attributes & (FILE_ATTRIBUTE_DIRECTORY | + FILE_ATTRIBUTE_ENCRYPTED)) == FILE_ATTRIBUTE_ENCRYPTED) { - union wimlib_progress_info info; - info.scan.cur_path = path; - info.scan.excluded = false; - params->progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info); + if (params->add_flags & WIMLIB_ADD_FLAG_NO_UNSUPPORTED_EXCLUDE) + { + ERROR("Can't archive unsupported encrypted file \"%s\"", path); + ret = WIMLIB_ERR_UNSUPPORTED_FILE; + goto out; + } + params->progress.scan.cur_path = path; + ret = do_capture_progress(params, WIMLIB_SCAN_DENTRY_UNSUPPORTED, NULL); + goto out; } /* Create a WIM dentry with an associated inode, which may be shared */ - ret = inode_table_new_dentry(¶ms->inode_table, + ret = inode_table_new_dentry(params->inode_table, path_basename_with_len(path, path_len), ni->mft_no, 0, false, &root); if (ret) - return ret; + goto out; if (name_type & FILE_NAME_WIN32) /* Win32 or Win32+DOS name (rather than POSIX) */ root->is_win32_name = 1; inode = root->d_inode; - if (inode->i_nlink > 1) /* Shared inode; nothing more to do */ - goto out; + if (inode->i_nlink > 1) { + /* Shared inode; nothing more to do */ + ret = 0; + goto out_progress; + } inode->i_creation_time = le64_to_cpu(ni->creation_time); inode->i_last_write_time = le64_to_cpu(ni->last_data_change_time); @@ -602,20 +587,30 @@ build_dentry_tree_ntfs_recursive(struct wim_dentry **root_ret, inode->i_attributes = le32_to_cpu(attributes); inode->i_resolved = 1; - if (attributes & FILE_ATTR_REPARSE_POINT) - stream_type = AT_REPARSE_POINT; - else - stream_type = AT_DATA; + /* Capture streams. */ - /* Capture the file's streams; more specifically, this is supposed to: + if (attributes & FILE_ATTR_REPARSE_POINT) { + /* Capture reparse data stream. */ + ret = capture_ntfs_streams(inode, ni, path, path_len, + params->unhashed_streams, + vol, AT_REPARSE_POINT); + if (ret) + goto out; + } + + /* Capture data streams. + * + * Directories should not have an unnamed data stream, but they may have + * named data streams. + * + * Reparse points may have an unnamed data stream (which will be ignored + * in favor of the reparse data stream), and they also may have named + * data streams. * - * - Regular files: capture unnamed data stream and any named data - * streams - * - Directories: capture any named data streams - * - Reparse points: capture reparse data only - */ + * Regular files can have an unnamed data stream as well as named data + * streams. */ ret = capture_ntfs_streams(inode, ni, path, path_len, - params->lookup_table, vol, stream_type); + params->unhashed_streams, vol, AT_DATA); if (ret) goto out; @@ -623,7 +618,7 @@ build_dentry_tree_ntfs_recursive(struct wim_dentry **root_ret, /* Recurse to directory children */ s64 pos = 0; - struct dos_name_map dos_name_map = { .rb_root = {.rb_node = NULL} }; + struct dos_name_map dos_name_map = { .root = NULL }; struct readdir_ctx ctx = { .parent = root, .path = path, @@ -637,13 +632,20 @@ build_dentry_tree_ntfs_recursive(struct wim_dentry **root_ret, ERROR_WITH_ERRNO("Error reading directory \"%s\"", path); ret = WIMLIB_ERR_NTFS_3G; } else { - ret = for_dentry_child(root, set_dentry_dos_name, - &dos_name_map); + struct wim_dentry *child; + + ret = 0; + for_dentry_child(child, root) { + ret = set_dentry_dos_name(child, &dos_name_map); + if (ret) + break; + } } destroy_dos_name_map(&dos_name_map); if (ret) goto out; } + path[path_len] = '\0'; /* Reparse-point fixups are a no-op because in NTFS-3g capture mode we * only allow capturing an entire volume. */ @@ -653,7 +655,7 @@ build_dentry_tree_ntfs_recursive(struct wim_dentry **root_ret, if (!(params->add_flags & WIMLIB_ADD_FLAG_NO_ACLS)) { struct SECURITY_CONTEXT sec_ctx; - char _sd[1]; + char _sd[4096]; char *sd; /* Get security descriptor */ @@ -661,13 +663,14 @@ build_dentry_tree_ntfs_recursive(struct wim_dentry **root_ret, sec_ctx.vol = vol; errno = 0; - ret = ntfs_get_ntfs_acl(&sec_ctx, ni, _sd, sizeof(_sd)); + sd = _sd; + ret = ntfs_get_ntfs_acl(&sec_ctx, ni, sd, sizeof(_sd)); if (ret > sizeof(_sd)) { sd = alloca(ret); ret = ntfs_get_ntfs_acl(&sec_ctx, ni, sd, ret); } if (ret > 0) { - inode->i_security_id = sd_set_add_sd(¶ms->sd_set, + inode->i_security_id = sd_set_add_sd(params->sd_set, sd, ret); if (inode->i_security_id == -1) { ERROR("Out of memory"); @@ -686,6 +689,15 @@ build_dentry_tree_ntfs_recursive(struct wim_dentry **root_ret, DEBUG("No security ID for `%s'", path); } } + if (ret) + goto out; + +out_progress: + params->progress.scan.cur_path = path; + if (root == NULL) + ret = do_capture_progress(params, WIMLIB_SCAN_DENTRY_EXCLUDED, NULL); + else + ret = do_capture_progress(params, WIMLIB_SCAN_DENTRY_OK, inode); out: if (ret == 0) *root_ret = root;