X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fdentry.c;h=9411f85dba11f4e9964a10f1e932147e4e85dcbc;hp=20fd34912efa0d9533952b89116026a3d2841ba1;hb=b8df27e5fdd4dae472ababcec11d04eafea8830b;hpb=67f45cecd793345416d5d85fbe37ec54b1bb6ef8 diff --git a/src/dentry.c b/src/dentry.c index 20fd3491..9411f85d 100644 --- a/src/dentry.c +++ b/src/dentry.c @@ -1,16 +1,15 @@ /* * dentry.c * - * A dentry (directory entry) contains the metadata for a file. In the WIM file - * format, the dentries are stored in the "metadata resource" section right - * after the security data. Each image in the WIM file has its own metadata - * resource with its own security data and dentry tree. Dentries in different - * images may share file resources by referring to the same lookup table - * entries. + * In the WIM file format, the dentries are stored in the "metadata resource" + * section right after the security data. Each image in the WIM file has its + * own metadata resource with its own security data and dentry tree. Dentries + * in different images may share file resources by referring to the same lookup + * table entries. */ /* - * Copyright (C) 2012 Eric Biggers + * Copyright (C) 2012, 2013 Eric Biggers * * This file is part of wimlib, a library for working with WIM files. * @@ -27,311 +26,454 @@ * wimlib; if not, see http://www.gnu.org/licenses/. */ -#include -#include -#include -#include +#ifdef HAVE_CONFIG_H +# include "config.h" +#endif -#include "dentry.h" -#include "io.h" -#include "lookup_table.h" -#include "sha1.h" -#include "timestamp.h" -#include "wimlib_internal.h" +#include "wimlib.h" +#include "wimlib/dentry.h" +#include "wimlib/encoding.h" +#include "wimlib/endianness.h" +#include "wimlib/error.h" +#include "wimlib/lookup_table.h" +#include "wimlib/metadata.h" +#include "wimlib/resource.h" +#include "wimlib/sha1.h" +#include "wimlib/timestamp.h" -/* - * Returns true if @dentry has the UTF-8 file name @name that has length - * @name_len. - */ -static bool dentry_has_name(const struct dentry *dentry, const char *name, - size_t name_len) -{ - if (dentry->file_name_utf8_len != name_len) - return false; - return memcmp(dentry->file_name_utf8, name, name_len) == 0; -} +#include -static u64 __dentry_correct_length_unaligned(u16 file_name_len, - u16 short_name_len) +/* WIM alternate data stream entry (on-disk format) */ +struct wim_ads_entry_on_disk { + /* Length of the entry, in bytes. This apparently includes all + * fixed-length fields, plus the stream name and null terminator if + * present, and the padding up to an 8 byte boundary. wimlib is a + * little less strict when reading the entries, and only requires that + * the number of bytes from this field is at least as large as the size + * of the fixed length fields and stream name without null terminator. + * */ + le64 length; + + le64 reserved; + + /* SHA1 message digest of the uncompressed stream; or, alternatively, + * can be all zeroes if the stream has zero length. */ + u8 hash[SHA1_HASH_SIZE]; + + /* Length of the stream name, in bytes. 0 if the stream is unnamed. */ + le16 stream_name_nbytes; + + /* Stream name in UTF-16LE. It is @stream_name_nbytes bytes long, + * excluding the the null terminator. There is a null terminator + * character if @stream_name_nbytes != 0; i.e., if this stream is named. + * */ + utf16lechar stream_name[]; +} _packed_attribute; + +/* WIM directory entry (on-disk format) */ +struct wim_dentry_on_disk { + le64 length; + le32 attributes; + sle32 security_id; + le64 subdir_offset; + le64 unused_1; + le64 unused_2; + le64 creation_time; + le64 last_access_time; + le64 last_write_time; + u8 unnamed_stream_hash[SHA1_HASH_SIZE]; + union { + struct { + le32 rp_unknown_1; + le32 reparse_tag; + le16 rp_unknown_2; + le16 not_rpfixed; + } _packed_attribute reparse; + struct { + le32 rp_unknown_1; + le64 hard_link_group_id; + } _packed_attribute nonreparse; + }; + le16 num_alternate_data_streams; + le16 short_name_nbytes; + le16 file_name_nbytes; + + /* Follewed by variable length file name, if file_name_nbytes != 0 */ + utf16lechar file_name[]; + + /* Followed by variable length short name, if short_name_nbytes != 0 */ + /*utf16lechar short_name[];*/ +} _packed_attribute; + +/* Calculates the unaligned length, in bytes, of an on-disk WIM dentry that has + * a file name and short name that take the specified numbers of bytes. This + * excludes any alternate data stream entries that may follow the dentry. */ +static u64 +_dentry_correct_length_unaligned(u16 file_name_nbytes, u16 short_name_nbytes) { - u64 length = WIM_DENTRY_DISK_SIZE; - if (file_name_len) - length += file_name_len + 2; - if (short_name_len) - length += short_name_len + 2; + u64 length = sizeof(struct wim_dentry_on_disk); + if (file_name_nbytes) + length += file_name_nbytes + 2; + if (short_name_nbytes) + length += short_name_nbytes + 2; return length; } -static u64 dentry_correct_length_unaligned(const struct dentry *dentry) +/* Calculates the unaligned length, in bytes, of an on-disk WIM dentry, based on + * the file name length and short name length. Note that dentry->length is + * ignored; also, this excludes any alternate data stream entries that may + * follow the dentry. */ +static u64 +dentry_correct_length_unaligned(const struct wim_dentry *dentry) { - return __dentry_correct_length_unaligned(dentry->file_name_len, - dentry->short_name_len); + return _dentry_correct_length_unaligned(dentry->file_name_nbytes, + dentry->short_name_nbytes); } -/* Return the "correct" value to write in the length field of the dentry, based - * on the file name length and short name length */ -static u64 dentry_correct_length(const struct dentry *dentry) +/* Return %true iff the alternate data stream entry @entry has the UTF-16LE + * stream name @name that has length @name_nbytes bytes. */ +static inline bool +ads_entry_has_name(const struct wim_ads_entry *entry, + const utf16lechar *name, size_t name_nbytes) { - return (dentry_correct_length_unaligned(dentry) + 7) & ~7; + return entry->stream_name_nbytes == name_nbytes && + memcmp(entry->stream_name, name, name_nbytes) == 0; } -static u64 __dentry_total_length(const struct dentry *dentry, u64 length) +/* Duplicates a string of system-dependent encoding into a UTF-16LE string and + * returns the string and its length, in bytes, in the pointer arguments. Frees + * any existing string at the return location before overwriting it. */ +static int +get_utf16le_name(const tchar *name, utf16lechar **name_utf16le_ret, + u16 *name_utf16le_nbytes_ret) { - const struct inode *inode = dentry->inode; - for (u16 i = 0; i < inode->num_ads; i++) - length += ads_entry_total_length(inode->ads_entries[i]); - return (length + 7) & ~7; + utf16lechar *name_utf16le; + size_t name_utf16le_nbytes; + int ret; +#if TCHAR_IS_UTF16LE + name_utf16le_nbytes = tstrlen(name) * sizeof(utf16lechar); + name_utf16le = MALLOC(name_utf16le_nbytes + sizeof(utf16lechar)); + if (!name_utf16le) + return WIMLIB_ERR_NOMEM; + memcpy(name_utf16le, name, name_utf16le_nbytes + sizeof(utf16lechar)); + ret = 0; +#else + + ret = tstr_to_utf16le(name, tstrlen(name), &name_utf16le, + &name_utf16le_nbytes); + if (ret == 0) { + if (name_utf16le_nbytes > 0xffff) { + FREE(name_utf16le); + ERROR("Multibyte string \"%"TS"\" is too long!", name); + ret = WIMLIB_ERR_INVALID_UTF8_STRING; + } + } +#endif + if (ret == 0) { + FREE(*name_utf16le_ret); + *name_utf16le_ret = name_utf16le; + *name_utf16le_nbytes_ret = name_utf16le_nbytes; + } + return ret; } -u64 dentry_correct_total_length(const struct dentry *dentry) +/* Sets the name of a WIM dentry from a multibyte string. */ +int +set_dentry_name(struct wim_dentry *dentry, const tchar *new_name) { - return __dentry_total_length(dentry, - dentry_correct_length_unaligned(dentry)); + int ret; + ret = get_utf16le_name(new_name, &dentry->file_name, + &dentry->file_name_nbytes); + if (ret == 0) { + /* Clear the short name and recalculate the dentry length */ + if (dentry_has_short_name(dentry)) { + FREE(dentry->short_name); + dentry->short_name = NULL; + dentry->short_name_nbytes = 0; + } + } + return ret; } -/* Real length of a dentry, including the alternate data stream entries, which - * are not included in the dentry->length field... */ -u64 dentry_total_length(const struct dentry *dentry) +/* Returns the total length of a WIM alternate data stream entry on-disk, + * including the stream name, the null terminator, AND the padding after the + * entry to align the next ADS entry or dentry on an 8-byte boundary. */ +static u64 +ads_entry_total_length(const struct wim_ads_entry *entry) { - return __dentry_total_length(dentry, dentry->length); + u64 len = sizeof(struct wim_ads_entry_on_disk); + if (entry->stream_name_nbytes) + len += entry->stream_name_nbytes + 2; + return (len + 7) & ~7; } -/* Transfers file attributes from a `stat' buffer to a struct dentry. */ -void stbuf_to_dentry(const struct stat *stbuf, struct dentry *dentry) + +static u64 +_dentry_total_length(const struct wim_dentry *dentry, u64 length) { - struct inode *inode = dentry->inode; - if (S_ISLNK(stbuf->st_mode)) { - inode->attributes = FILE_ATTRIBUTE_REPARSE_POINT; - inode->reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK; - } else if (S_ISDIR(stbuf->st_mode)) { - inode->attributes = FILE_ATTRIBUTE_DIRECTORY; - } else { - inode->attributes = FILE_ATTRIBUTE_NORMAL; - } - if (sizeof(ino_t) >= 8) - dentry->link_group_id = (u64)stbuf->st_ino; - else - dentry->link_group_id = (u64)stbuf->st_ino | - ((u64)stbuf->st_dev << (sizeof(ino_t) * 8)); - /* Set timestamps */ - inode->creation_time = timespec_to_wim_timestamp(&stbuf->st_mtim); - inode->last_write_time = timespec_to_wim_timestamp(&stbuf->st_mtim); - inode->last_access_time = timespec_to_wim_timestamp(&stbuf->st_atim); + const struct wim_inode *inode = dentry->d_inode; + for (u16 i = 0; i < inode->i_num_ads; i++) + length += ads_entry_total_length(&inode->i_ads_entries[i]); + return (length + 7) & ~7; } - -/* Sets all the timestamp fields of the dentry to the current time. */ -void inode_update_all_timestamps(struct inode *inode) +/* Calculate the aligned *total* length of an on-disk WIM dentry. This includes + * all alternate data streams. */ +u64 +dentry_correct_total_length(const struct wim_dentry *dentry) { - u64 now = get_wim_timestamp(); - inode->creation_time = now; - inode->last_access_time = now; - inode->last_write_time = now; + return _dentry_total_length(dentry, + dentry_correct_length_unaligned(dentry)); } -/* Returns the alternate data stream entry belonging to @inode that has the - * stream name @stream_name. */ -struct ads_entry *inode_get_ads_entry(struct inode *inode, - const char *stream_name) +/* Like dentry_correct_total_length(), but use the existing dentry->length field + * instead of calculating its "correct" value. */ +static u64 +dentry_total_length(const struct wim_dentry *dentry) { - size_t stream_name_len; - if (!stream_name) - return NULL; - if (inode->num_ads) { - u16 i = 0; - stream_name_len = strlen(stream_name); - do { - if (ads_entry_has_name(inode->ads_entries[i], - stream_name, stream_name_len)) - return inode->ads_entries[i]; - } while (++i != inode->num_ads); - } - return NULL; + return _dentry_total_length(dentry, dentry->length); } - -static struct ads_entry *new_ads_entry(const char *name) +int +for_dentry_in_rbtree(struct rb_node *root, + int (*visitor)(struct wim_dentry *, void *), + void *arg) { - struct ads_entry *ads_entry = CALLOC(1, sizeof(struct ads_entry)); - if (!ads_entry) - return NULL; - INIT_LIST_HEAD(&ads_entry->lte_group_list.list); - ads_entry->lte_group_list.type = STREAM_TYPE_ADS; - if (name && *name) { - if (change_ads_name(ads_entry, name)) { - FREE(ads_entry); - return NULL; + int ret; + struct rb_node *node = root; + LIST_HEAD(stack); + while (1) { + if (node) { + list_add(&rbnode_dentry(node)->tmp_list, &stack); + node = node->rb_left; + } else { + struct list_head *next; + struct wim_dentry *dentry; + + next = stack.next; + if (next == &stack) + return 0; + dentry = container_of(next, struct wim_dentry, tmp_list); + list_del(next); + ret = visitor(dentry, arg); + if (ret != 0) + return ret; + node = dentry->rb_node.rb_right; } } - return ads_entry; } -/* - * Add an alternate stream entry to an inode and return a pointer to it, or NULL - * if memory could not be allocated. - */ -struct ads_entry *inode_add_ads(struct inode *inode, const char *stream_name) +static int +for_dentry_tree_in_rbtree_depth(struct rb_node *node, + int (*visitor)(struct wim_dentry*, void*), + void *arg) { - u16 num_ads; - struct ads_entry **ads_entries; - struct ads_entry *new_entry; - - if (inode->num_ads == 0xffff) { - ERROR("Too many alternate data streams in one inode!"); - return NULL; - } - num_ads = inode->num_ads + 1; - ads_entries = REALLOC(inode->ads_entries, - num_ads * sizeof(inode->ads_entries[0])); - if (!ads_entries) { - ERROR("Failed to allocate memory for new alternate data stream"); - return NULL; + int ret; + if (node) { + ret = for_dentry_tree_in_rbtree_depth(node->rb_left, + visitor, arg); + if (ret != 0) + return ret; + ret = for_dentry_tree_in_rbtree_depth(node->rb_right, + visitor, arg); + if (ret != 0) + return ret; + ret = for_dentry_in_tree_depth(rbnode_dentry(node), visitor, arg); + if (ret != 0) + return ret; } - inode->ads_entries = ads_entries; - - new_entry = new_ads_entry(stream_name); - if (new_entry) - return NULL; - inode->num_ads = num_ads; - ads_entries[num_ads - 1] = new_entry; - return new_entry; + return 0; } - -/* - * Calls a function on all directory entries in a directory tree. It is called - * on a parent before its children. - */ -int for_dentry_in_tree(struct dentry *root, - int (*visitor)(struct dentry*, void*), void *arg) +static int +for_dentry_tree_in_rbtree(struct rb_node *node, + int (*visitor)(struct wim_dentry*, void*), + void *arg) { int ret; - struct dentry *child; + if (node) { + ret = for_dentry_tree_in_rbtree(node->rb_left, visitor, arg); + if (ret) + return ret; + ret = for_dentry_in_tree(rbnode_dentry(node), visitor, arg); + if (ret) + return ret; + ret = for_dentry_tree_in_rbtree(node->rb_right, visitor, arg); + if (ret) + return ret; + } + return 0; +} - ret = visitor(root, arg); +/* Calls a function on all directory entries in a WIM dentry tree. Logically, + * this is a pre-order traversal (the function is called on a parent dentry + * before its children), but sibling dentries will be visited in order as well. + * */ +int +for_dentry_in_tree(struct wim_dentry *root, + int (*visitor)(struct wim_dentry*, void*), void *arg) +{ + int ret; - if (ret != 0) + if (!root) + return 0; + ret = (*visitor)(root, arg); + if (ret) return ret; + return for_dentry_tree_in_rbtree(root->d_inode->i_children.rb_node, + visitor, + arg); +} - child = root->children; +/* Like for_dentry_in_tree(), but the visitor function is always called on a + * dentry's children before on itself. */ +int +for_dentry_in_tree_depth(struct wim_dentry *root, + int (*visitor)(struct wim_dentry*, void*), void *arg) +{ + int ret; - if (!child) + if (!root) return 0; - - do { - ret = for_dentry_in_tree(child, visitor, arg); - if (ret != 0) - return ret; - child = child->next; - } while (child != root->children); - return 0; + ret = for_dentry_tree_in_rbtree_depth(root->d_inode->i_children.rb_node, + visitor, arg); + if (ret) + return ret; + return (*visitor)(root, arg); } -/* - * Like for_dentry_in_tree(), but the visitor function is always called on a - * dentry's children before on itself. - */ -int for_dentry_in_tree_depth(struct dentry *root, - int (*visitor)(struct dentry*, void*), void *arg) +/* Calculate the full path of @dentry. The full path of its parent must have + * already been calculated, or it must be the root dentry. */ +static int +calculate_dentry_full_path(struct wim_dentry *dentry) { + tchar *full_path; + u32 full_path_nbytes; int ret; - struct dentry *child; - struct dentry *next; - child = root->children; - if (child) { - do { - next = child->next; - ret = for_dentry_in_tree_depth(child, visitor, arg); - if (ret != 0) - return ret; - child = next; - } while (child != root->children); - } - return visitor(root, arg); -} + if (dentry->_full_path) + return 0; -/* - * Calculate the full path of @dentry, based on its parent's full path and on - * its UTF-8 file name. - */ -int calculate_dentry_full_path(struct dentry *dentry, void *ignore) -{ - char *full_path; - u32 full_path_len; if (dentry_is_root(dentry)) { - full_path = MALLOC(2); + full_path = TSTRDUP(T("/")); if (!full_path) - goto oom; - full_path[0] = '/'; - full_path[1] = '\0'; - full_path_len = 1; + return WIMLIB_ERR_NOMEM; + full_path_nbytes = 1 * sizeof(tchar); } else { - char *parent_full_path; - u32 parent_full_path_len; - const struct dentry *parent = dentry->parent; + struct wim_dentry *parent; + tchar *parent_full_path; + u32 parent_full_path_nbytes; + size_t filename_nbytes; + parent = dentry->parent; if (dentry_is_root(parent)) { - parent_full_path = ""; - parent_full_path_len = 0; + parent_full_path = T(""); + parent_full_path_nbytes = 0; } else { - parent_full_path = parent->full_path_utf8; - parent_full_path_len = parent->full_path_utf8_len; + if (!parent->_full_path) { + ret = calculate_dentry_full_path(parent); + if (ret) + return ret; + } + parent_full_path = parent->_full_path; + parent_full_path_nbytes = parent->full_path_nbytes; + } + + /* Append this dentry's name as a tchar string to the full path + * of the parent followed by the path separator */ + #if TCHAR_IS_UTF16LE + filename_nbytes = dentry->file_name_nbytes; + #else + { + int ret = utf16le_to_tstr_nbytes(dentry->file_name, + dentry->file_name_nbytes, + &filename_nbytes); + if (ret) + return ret; } + #endif - full_path_len = parent_full_path_len + 1 + - dentry->file_name_utf8_len; - full_path = MALLOC(full_path_len + 1); + full_path_nbytes = parent_full_path_nbytes + sizeof(tchar) + + filename_nbytes; + full_path = MALLOC(full_path_nbytes + sizeof(tchar)); if (!full_path) - goto oom; - - memcpy(full_path, parent_full_path, parent_full_path_len); - full_path[parent_full_path_len] = '/'; - memcpy(full_path + parent_full_path_len + 1, - dentry->file_name_utf8, - dentry->file_name_utf8_len); - full_path[full_path_len] = '\0'; + return WIMLIB_ERR_NOMEM; + memcpy(full_path, parent_full_path, parent_full_path_nbytes); + full_path[parent_full_path_nbytes / sizeof(tchar)] = T('/'); + #if TCHAR_IS_UTF16LE + memcpy(&full_path[parent_full_path_nbytes / sizeof(tchar) + 1], + dentry->file_name, + filename_nbytes + sizeof(tchar)); + #else + utf16le_to_tstr_buf(dentry->file_name, + dentry->file_name_nbytes, + &full_path[parent_full_path_nbytes / + sizeof(tchar) + 1]); + #endif } - FREE(dentry->full_path_utf8); - dentry->full_path_utf8 = full_path; - dentry->full_path_utf8_len = full_path_len; + dentry->_full_path = full_path; + dentry->full_path_nbytes= full_path_nbytes; + return 0; +} + +static int +do_calculate_dentry_full_path(struct wim_dentry *dentry, void *_ignore) +{ + return calculate_dentry_full_path(dentry); +} + +int +calculate_dentry_tree_full_paths(struct wim_dentry *root) +{ + return for_dentry_in_tree(root, do_calculate_dentry_full_path, NULL); +} + +tchar * +dentry_full_path(struct wim_dentry *dentry) +{ + calculate_dentry_full_path(dentry); + return dentry->_full_path; +} + +static int +increment_subdir_offset(struct wim_dentry *dentry, void *subdir_offset_p) +{ + *(u64*)subdir_offset_p += dentry_correct_total_length(dentry); + return 0; +} + +static int +call_calculate_subdir_offsets(struct wim_dentry *dentry, void *subdir_offset_p) +{ + calculate_subdir_offsets(dentry, subdir_offset_p); return 0; -oom: - ERROR("Out of memory while calculating dentry full path"); - return WIMLIB_ERR_NOMEM; } -/* - * Recursively calculates the subdir offsets for a directory tree. +/* + * Recursively calculates the subdir offsets for a directory tree. * * @dentry: The root of the directory tree. * @subdir_offset_p: The current subdirectory offset; i.e., the subdirectory - * offset for @dentry. + * offset for @dentry. */ -void calculate_subdir_offsets(struct dentry *dentry, u64 *subdir_offset_p) +void +calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p) { - struct dentry *child; + struct rb_node *node; - child = dentry->children; dentry->subdir_offset = *subdir_offset_p; - - if (child) { + node = dentry->d_inode->i_children.rb_node; + if (node) { /* Advance the subdir offset by the amount of space the children * of this dentry take up. */ - do { - *subdir_offset_p += dentry_correct_total_length(child); - child = child->next; - } while (child != dentry->children); + for_dentry_in_rbtree(node, increment_subdir_offset, subdir_offset_p); /* End-of-directory dentry on disk. */ *subdir_offset_p += 8; /* Recursively call calculate_subdir_offsets() on all the * children. */ - do { - calculate_subdir_offsets(child, subdir_offset_p); - child = child->next; - } while (child != dentry->children); + for_dentry_in_rbtree(node, call_calculate_subdir_offsets, subdir_offset_p); } else { /* On disk, childless directories have a valid subdir_offset * that points to an 8-byte end-of-directory dentry. Regular @@ -343,80 +485,182 @@ void calculate_subdir_offsets(struct dentry *dentry, u64 *subdir_offset_p) } } +static int +compare_utf16le_names(const utf16lechar *name1, size_t nbytes1, + const utf16lechar *name2, size_t nbytes2) +{ + int result = memcmp(name1, name2, min(nbytes1, nbytes2)); + if (result) + return result; + else + return (int)nbytes1 - (int)nbytes2; +} -/* Returns the child of @dentry that has the file name @name. - * Returns NULL if no child has the name. */ -struct dentry *get_dentry_child_with_name(const struct dentry *dentry, - const char *name) +static int +dentry_compare_names(const struct wim_dentry *d1, const struct wim_dentry *d2) { - struct dentry *child; - size_t name_len; - - child = dentry->children; - if (child) { - name_len = strlen(name); - do { - if (dentry_has_name(child, name, name_len)) - return child; - child = child->next; - } while (child != dentry->children); + return compare_utf16le_names(d1->file_name, d1->file_name_nbytes, + d2->file_name, d2->file_name_nbytes); +} + + +struct wim_dentry * +get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry, + const utf16lechar *name, + size_t name_nbytes) +{ + struct rb_node *node = dentry->d_inode->i_children.rb_node; + struct wim_dentry *child; + while (node) { + child = rbnode_dentry(node); + int result = compare_utf16le_names(name, name_nbytes, + child->file_name, + child->file_name_nbytes); + if (result < 0) + node = node->rb_left; + else if (result > 0) + node = node->rb_right; + else + return child; } return NULL; } -/* Retrieves the dentry that has the UTF-8 @path relative to the dentry - * @cur_dir. Returns NULL if no dentry having the path is found. */ -static struct dentry *get_dentry_relative_path(struct dentry *cur_dir, - const char *path) +/* Returns the child of @dentry that has the file name @name. Returns NULL if + * no child has the name. */ +struct wim_dentry * +get_dentry_child_with_name(const struct wim_dentry *dentry, const tchar *name) +{ +#if TCHAR_IS_UTF16LE + return get_dentry_child_with_utf16le_name(dentry, name, + tstrlen(name) * sizeof(tchar)); +#else + utf16lechar *utf16le_name; + size_t utf16le_name_nbytes; + int ret; + struct wim_dentry *child; + + ret = tstr_to_utf16le(name, tstrlen(name) * sizeof(tchar), + &utf16le_name, &utf16le_name_nbytes); + if (ret) { + child = NULL; + } else { + child = get_dentry_child_with_utf16le_name(dentry, + utf16le_name, + utf16le_name_nbytes); + FREE(utf16le_name); + } + return child; +#endif +} + +static struct wim_dentry * +get_dentry_utf16le(WIMStruct *wim, const utf16lechar *path) { - struct dentry *child; - size_t base_len; - const char *new_path; + struct wim_dentry *cur_dentry, *parent_dentry; + const utf16lechar *p, *pp; - if (*path == '\0') - return cur_dir; + cur_dentry = parent_dentry = wim_root_dentry(wim); + if (!cur_dentry) { + errno = ENOENT; + return NULL; + } + p = path; + while (1) { + while (*p == cpu_to_le16('/')) + p++; + if (*p == cpu_to_le16('\0')) + break; + pp = p; + while (*pp != cpu_to_le16('/') && *pp != cpu_to_le16('\0')) + pp++; - child = cur_dir->children; - if (child) { - new_path = path_next_part(path, &base_len); - do { - if (dentry_has_name(child, path, base_len)) - return get_dentry_relative_path(child, new_path); - child = child->next; - } while (child != cur_dir->children); + cur_dentry = get_dentry_child_with_utf16le_name(parent_dentry, p, + (void*)pp - (void*)p); + if (cur_dentry == NULL) + break; + p = pp; + parent_dentry = cur_dentry; } - return NULL; + if (cur_dentry == NULL) { + if (dentry_is_directory(parent_dentry)) + errno = ENOENT; + else + errno = ENOTDIR; + } + return cur_dentry; +} + +/* Returns the dentry corresponding to the @path, or NULL if there is no such + * dentry. */ +struct wim_dentry * +get_dentry(WIMStruct *wim, const tchar *path) +{ +#if TCHAR_IS_UTF16LE + return get_dentry_utf16le(wim, path); +#else + utf16lechar *path_utf16le; + size_t path_utf16le_nbytes; + int ret; + struct wim_dentry *dentry; + + ret = tstr_to_utf16le(path, tstrlen(path) * sizeof(tchar), + &path_utf16le, &path_utf16le_nbytes); + if (ret) + return NULL; + dentry = get_dentry_utf16le(wim, path_utf16le); + FREE(path_utf16le); + return dentry; +#endif +} + +struct wim_inode * +wim_pathname_to_inode(WIMStruct *wim, const tchar *path) +{ + struct wim_dentry *dentry; + dentry = get_dentry(wim, path); + if (dentry) + return dentry->d_inode; + else + return NULL; } -/* Returns the dentry corresponding to the UTF-8 @path, or NULL if there is no - * such dentry. */ -struct dentry *get_dentry(WIMStruct *w, const char *path) +/* Takes in a path of length @len in @buf, and transforms it into a string for + * the path of its parent directory. */ +static void +to_parent_name(tchar *buf, size_t len) { - struct dentry *root = wim_root_dentry(w); - while (*path == '/') - path++; - return get_dentry_relative_path(root, path); + ssize_t i = (ssize_t)len - 1; + while (i >= 0 && buf[i] == T('/')) + i--; + while (i >= 0 && buf[i] != T('/')) + i--; + while (i >= 0 && buf[i] == T('/')) + i--; + buf[i + 1] = T('\0'); } /* Returns the dentry that corresponds to the parent directory of @path, or NULL * if the dentry is not found. */ -struct dentry *get_parent_dentry(WIMStruct *w, const char *path) +struct wim_dentry * +get_parent_dentry(WIMStruct *wim, const tchar *path) { - size_t path_len = strlen(path); - char buf[path_len + 1]; - - memcpy(buf, path, path_len + 1); + size_t path_len = tstrlen(path); + tchar buf[path_len + 1]; + tmemcpy(buf, path, path_len + 1); to_parent_name(buf, path_len); - - return get_dentry(w, buf); + return get_dentry(wim, buf); } /* Prints the full path of a dentry. */ -int print_dentry_full_path(struct dentry *dentry, void *ignore) +int +print_dentry_full_path(struct wim_dentry *dentry, void *_ignore) { - if (dentry->full_path_utf8) - puts(dentry->full_path_utf8); + int ret = calculate_dentry_full_path(dentry); + if (ret) + return ret; + tprintf(T("%"TS"\n"), dentry->_full_path); return 0; } @@ -424,323 +668,336 @@ int print_dentry_full_path(struct dentry *dentry, void *ignore) * set. */ struct file_attr_flag { u32 flag; - const char *name; + const tchar *name; }; struct file_attr_flag file_attr_flags[] = { - {FILE_ATTRIBUTE_READONLY, "READONLY"}, - {FILE_ATTRIBUTE_HIDDEN, "HIDDEN"}, - {FILE_ATTRIBUTE_SYSTEM, "SYSTEM"}, - {FILE_ATTRIBUTE_DIRECTORY, "DIRECTORY"}, - {FILE_ATTRIBUTE_ARCHIVE, "ARCHIVE"}, - {FILE_ATTRIBUTE_DEVICE, "DEVICE"}, - {FILE_ATTRIBUTE_NORMAL, "NORMAL"}, - {FILE_ATTRIBUTE_TEMPORARY, "TEMPORARY"}, - {FILE_ATTRIBUTE_SPARSE_FILE, "SPARSE_FILE"}, - {FILE_ATTRIBUTE_REPARSE_POINT, "REPARSE_POINT"}, - {FILE_ATTRIBUTE_COMPRESSED, "COMPRESSED"}, - {FILE_ATTRIBUTE_OFFLINE, "OFFLINE"}, - {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,"NOT_CONTENT_INDEXED"}, - {FILE_ATTRIBUTE_ENCRYPTED, "ENCRYPTED"}, - {FILE_ATTRIBUTE_VIRTUAL, "VIRTUAL"}, + {FILE_ATTRIBUTE_READONLY, T("READONLY")}, + {FILE_ATTRIBUTE_HIDDEN, T("HIDDEN")}, + {FILE_ATTRIBUTE_SYSTEM, T("SYSTEM")}, + {FILE_ATTRIBUTE_DIRECTORY, T("DIRECTORY")}, + {FILE_ATTRIBUTE_ARCHIVE, T("ARCHIVE")}, + {FILE_ATTRIBUTE_DEVICE, T("DEVICE")}, + {FILE_ATTRIBUTE_NORMAL, T("NORMAL")}, + {FILE_ATTRIBUTE_TEMPORARY, T("TEMPORARY")}, + {FILE_ATTRIBUTE_SPARSE_FILE, T("SPARSE_FILE")}, + {FILE_ATTRIBUTE_REPARSE_POINT, T("REPARSE_POINT")}, + {FILE_ATTRIBUTE_COMPRESSED, T("COMPRESSED")}, + {FILE_ATTRIBUTE_OFFLINE, T("OFFLINE")}, + {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,T("NOT_CONTENT_INDEXED")}, + {FILE_ATTRIBUTE_ENCRYPTED, T("ENCRYPTED")}, + {FILE_ATTRIBUTE_VIRTUAL, T("VIRTUAL")}, }; /* Prints a directory entry. @lookup_table is a pointer to the lookup table, if * available. If the dentry is unresolved and the lookup table is NULL, the * lookup table entries will not be printed. Otherwise, they will be. */ -int print_dentry(struct dentry *dentry, void *lookup_table) +int +print_dentry(struct wim_dentry *dentry, void *lookup_table) { const u8 *hash; - struct lookup_table_entry *lte; - const struct inode *inode = dentry->inode; - time_t time; - char *p; - - printf("[DENTRY]\n"); - printf("Length = %"PRIu64"\n", dentry->length); - printf("Attributes = 0x%x\n", inode->attributes); - for (unsigned i = 0; i < ARRAY_LEN(file_attr_flags); i++) - if (file_attr_flags[i].flag & inode->attributes) - printf(" FILE_ATTRIBUTE_%s is set\n", + struct wim_lookup_table_entry *lte; + const struct wim_inode *inode = dentry->d_inode; + tchar buf[50]; + + tprintf(T("[DENTRY]\n")); + tprintf(T("Length = %"PRIu64"\n"), dentry->length); + tprintf(T("Attributes = 0x%x\n"), inode->i_attributes); + for (size_t i = 0; i < ARRAY_LEN(file_attr_flags); i++) + if (file_attr_flags[i].flag & inode->i_attributes) + tprintf(T(" FILE_ATTRIBUTE_%"TS" is set\n"), file_attr_flags[i].name); - printf("Security ID = %d\n", inode->security_id); - printf("Subdir offset = %"PRIu64"\n", dentry->subdir_offset); -#if 0 - printf("Unused1 = 0x%"PRIu64"\n", dentry->unused1); - printf("Unused2 = %"PRIu64"\n", dentry->unused2); -#endif -#if 0 - printf("Creation Time = 0x%"PRIx64"\n"); - printf("Last Access Time = 0x%"PRIx64"\n"); - printf("Last Write Time = 0x%"PRIx64"\n"); -#endif + tprintf(T("Security ID = %d\n"), inode->i_security_id); + tprintf(T("Subdir offset = %"PRIu64"\n"), dentry->subdir_offset); + + wim_timestamp_to_str(inode->i_creation_time, buf, sizeof(buf)); + tprintf(T("Creation Time = %"TS"\n"), buf); + + wim_timestamp_to_str(inode->i_last_access_time, buf, sizeof(buf)); + tprintf(T("Last Access Time = %"TS"\n"), buf); - /* Translate the timestamps into something readable */ - time = wim_timestamp_to_unix(inode->creation_time); - p = asctime(gmtime(&time)); - *(strrchr(p, '\n')) = '\0'; - printf("Creation Time = %s UTC\n", p); - - time = wim_timestamp_to_unix(inode->last_access_time); - p = asctime(gmtime(&time)); - *(strrchr(p, '\n')) = '\0'; - printf("Last Access Time = %s UTC\n", p); - - time = wim_timestamp_to_unix(inode->last_write_time); - p = asctime(gmtime(&time)); - *(strrchr(p, '\n')) = '\0'; - printf("Last Write Time = %s UTC\n", p); - - printf("Reparse Tag = 0x%"PRIx32"\n", inode->reparse_tag); - printf("Hard Link Group = 0x%"PRIx64"\n", dentry->link_group_id); - printf("Number of Alternate Data Streams = %hu\n", inode->num_ads); - printf("Filename = \""); - print_string(dentry->file_name, dentry->file_name_len); - puts("\""); - printf("Filename Length = %hu\n", dentry->file_name_len); - printf("Filename (UTF-8) = \"%s\"\n", dentry->file_name_utf8); - printf("Filename (UTF-8) Length = %hu\n", dentry->file_name_utf8_len); - printf("Short Name = \""); - print_string(dentry->short_name, dentry->short_name_len); - puts("\""); - printf("Short Name Length = %hu\n", dentry->short_name_len); - printf("Full Path (UTF-8) = \"%s\"\n", dentry->full_path_utf8); - lte = inode_stream_lte(dentry->inode, 0, lookup_table); + wim_timestamp_to_str(inode->i_last_write_time, buf, sizeof(buf)); + tprintf(T("Last Write Time = %"TS"\n"), buf); + + if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { + tprintf(T("Reparse Tag = 0x%"PRIx32"\n"), inode->i_reparse_tag); + tprintf(T("Reparse Point Flags = 0x%"PRIx16"\n"), + inode->i_not_rpfixed); + tprintf(T("Reparse Point Unknown 2 = 0x%"PRIx32"\n"), + inode->i_rp_unknown_2); + } + tprintf(T("Reparse Point Unknown 1 = 0x%"PRIx32"\n"), + inode->i_rp_unknown_1); + tprintf(T("Hard Link Group = 0x%"PRIx64"\n"), inode->i_ino); + tprintf(T("Hard Link Group Size = %"PRIu32"\n"), inode->i_nlink); + tprintf(T("Number of Alternate Data Streams = %hu\n"), inode->i_num_ads); + if (dentry_has_long_name(dentry)) + wimlib_printf(T("Filename = \"%"WS"\"\n"), dentry->file_name); + if (dentry_has_short_name(dentry)) + wimlib_printf(T("Short Name \"%"WS"\"\n"), dentry->short_name); + if (dentry->_full_path) + tprintf(T("Full Path = \"%"TS"\"\n"), dentry->_full_path); + + lte = inode_stream_lte(dentry->d_inode, 0, lookup_table); if (lte) { - print_lookup_table_entry(lte); + print_lookup_table_entry(lte, stdout); } else { hash = inode_stream_hash(inode, 0); if (hash) { - printf("Hash = 0x"); - print_hash(hash); - putchar('\n'); - putchar('\n'); + tprintf(T("Hash = 0x")); + print_hash(hash, stdout); + tputchar(T('\n')); + tputchar(T('\n')); } } - for (u16 i = 0; i < inode->num_ads; i++) { - printf("[Alternate Stream Entry %u]\n", i); - printf("Name = \"%s\"\n", inode->ads_entries[i]->stream_name_utf8); - printf("Name Length (UTF-16) = %u\n", - inode->ads_entries[i]->stream_name_len); + for (u16 i = 0; i < inode->i_num_ads; i++) { + tprintf(T("[Alternate Stream Entry %u]\n"), i); + wimlib_printf(T("Name = \"%"WS"\"\n"), + inode->i_ads_entries[i].stream_name); + tprintf(T("Name Length (UTF16 bytes) = %hu\n"), + inode->i_ads_entries[i].stream_name_nbytes); hash = inode_stream_hash(inode, i + 1); if (hash) { - printf("Hash = 0x"); - print_hash(hash); - putchar('\n'); + tprintf(T("Hash = 0x")); + print_hash(hash, stdout); + tputchar(T('\n')); } - print_lookup_table_entry(inode_stream_lte(inode, i + 1, - lookup_table)); + print_lookup_table_entry(inode_stream_lte(inode, i + 1, lookup_table), + stdout); } return 0; } -/* Initializations done on every `struct dentry'. */ -static void dentry_common_init(struct dentry *dentry) +/* Initializations done on every `struct wim_dentry'. */ +static void +dentry_common_init(struct wim_dentry *dentry) { - memset(dentry, 0, sizeof(struct dentry)); - dentry->refcnt = 1; + memset(dentry, 0, sizeof(struct wim_dentry)); } -struct inode *new_inode() +struct wim_inode * +new_timeless_inode(void) { - struct inode *inode = CALLOC(1, sizeof(struct inode)); - u64 now = get_wim_timestamp(); - if (!inode) - return NULL; - inode->security_id = -1; - inode->link_count = 1; - inode->creation_time = now; - inode->last_access_time = now; - inode->last_write_time = now; - INIT_LIST_HEAD(&inode->dentry_list); + struct wim_inode *inode = CALLOC(1, sizeof(struct wim_inode)); + if (inode) { + inode->i_security_id = -1; + inode->i_nlink = 1; + inode->i_next_stream_id = 1; + inode->i_not_rpfixed = 1; + INIT_LIST_HEAD(&inode->i_list); + #ifdef WITH_FUSE + if (pthread_mutex_init(&inode->i_mutex, NULL) != 0) { + ERROR_WITH_ERRNO("Error initializing mutex"); + FREE(inode); + return NULL; + } + #endif + INIT_LIST_HEAD(&inode->i_dentry); + } return inode; } -/* - * Creates an unlinked directory entry. - * - * @name: The UTF-8 filename of the new dentry. - * - * Returns a pointer to the new dentry, or NULL if out of memory. - */ -struct dentry *new_dentry(const char *name) +static struct wim_inode * +new_inode(void) { - struct dentry *dentry; - - dentry = MALLOC(sizeof(struct dentry)); - if (!dentry) - goto err; + struct wim_inode *inode = new_timeless_inode(); + if (inode) { + u64 now = get_wim_timestamp(); + inode->i_creation_time = now; + inode->i_last_access_time = now; + inode->i_last_write_time = now; + } + return inode; +} - dentry_common_init(dentry); - if (change_dentry_name(dentry, name) != 0) - goto err; +/* Creates an unlinked directory entry. */ +int +new_dentry(const tchar *name, struct wim_dentry **dentry_ret) +{ + struct wim_dentry *dentry; + int ret; - dentry->next = dentry; - dentry->prev = dentry; - dentry->parent = dentry; + dentry = MALLOC(sizeof(struct wim_dentry)); + if (!dentry) + return WIMLIB_ERR_NOMEM; - return dentry; -err: - FREE(dentry); - ERROR("Failed to allocate new dentry"); - return NULL; + dentry_common_init(dentry); + ret = set_dentry_name(dentry, name); + if (ret == 0) { + dentry->parent = dentry; + *dentry_ret = dentry; + } else { + FREE(dentry); + ERROR("Failed to set name on new dentry with name \"%"TS"\"", + name); + } + return ret; } -struct dentry *new_dentry_with_inode(const char *name) + +static int +_new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret, + bool timeless) { - struct dentry *dentry; - dentry = new_dentry(name); - if (dentry) { - dentry->inode = new_inode(); - if (dentry->inode) { - list_add(&dentry->inode_dentry_list, - &dentry->inode->dentry_list); - } else { - free_dentry(dentry); - dentry = NULL; - } + struct wim_dentry *dentry; + int ret; + + ret = new_dentry(name, &dentry); + if (ret) + return ret; + + if (timeless) + dentry->d_inode = new_timeless_inode(); + else + dentry->d_inode = new_inode(); + if (!dentry->d_inode) { + free_dentry(dentry); + return WIMLIB_ERR_NOMEM; } - return dentry; + + inode_add_dentry(dentry, dentry->d_inode); + *dentry_ret = dentry; + return 0; } -static void free_ads_entry(struct ads_entry *entry) +int +new_dentry_with_timeless_inode(const tchar *name, struct wim_dentry **dentry_ret) { - if (entry) { - FREE(entry->stream_name); - FREE(entry->stream_name_utf8); - FREE(entry); - } + return _new_dentry_with_inode(name, dentry_ret, true); } - -#ifdef WITH_FUSE -/* Remove an alternate data stream from a dentry. - * - * The corresponding lookup table entry for the stream is NOT changed. - * - * @dentry: The dentry - * @ads_entry: The alternate data stream entry (it MUST be one of the - * ads_entry's in the array dentry->ads_entries). - */ -void dentry_remove_ads(struct dentry *dentry, struct ads_entry *ads_entry) +int +new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret) { - u16 idx; - u16 following; + return _new_dentry_with_inode(name, dentry_ret, false); +} - wimlib_assert(dentry->num_ads); - idx = ads_entry - dentry->ads_entries; - wimlib_assert(idx < dentry->num_ads); - following = dentry->num_ads - idx - 1; +int +new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret) +{ + int ret; + struct wim_dentry *dentry; - destroy_ads_entry(ads_entry); - memcpy(ads_entry, ads_entry + 1, following * sizeof(struct ads_entry)); - - /* We moved the ADS entries. Adjust the stream lists. */ - for (u16 i = 0; i < following; i++) { - struct list_head *cur = &ads_entry[i].lte_group_list.list; - struct list_head *prev = cur->prev; - struct list_head *next; - if ((u8*)prev >= (u8*)(ads_entry + 1) - && (u8*)prev < (u8*)(ads_entry + following + 1)) { - cur->prev = (struct list_head*)((u8*)prev - sizeof(struct ads_entry)); - } else { - prev->next = cur; - } - next = cur->next; - if ((u8*)next >= (u8*)(ads_entry + 1) - && (u8*)next < (u8*)(ads_entry + following + 1)) { - cur->next = (struct list_head*)((u8*)next - sizeof(struct ads_entry)); - } else { - next->prev = cur; - } - } - dentry->num_ads--; + DEBUG("Creating filler directory \"%"TS"\"", name); + ret = new_dentry_with_inode(name, &dentry); + if (ret) + return ret; + /* Leave the inode number as 0; this is allowed for non + * hard-linked files. */ + dentry->d_inode->i_resolved = 1; + dentry->d_inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY; + *dentry_ret = dentry; + return 0; } -#endif -static void inode_free_ads_entries(struct inode *inode) +static int +init_ads_entry(struct wim_ads_entry *ads_entry, const void *name, + size_t name_nbytes, bool is_utf16le) { - if (inode->ads_entries) { - for (u16 i = 0; i < inode->num_ads; i++) - free_ads_entry(inode->ads_entries[i]); - FREE(inode->ads_entries); + int ret = 0; + memset(ads_entry, 0, sizeof(*ads_entry)); + + if (is_utf16le) { + utf16lechar *p = MALLOC(name_nbytes + sizeof(utf16lechar)); + if (!p) + return WIMLIB_ERR_NOMEM; + memcpy(p, name, name_nbytes); + p[name_nbytes / 2] = cpu_to_le16(0); + ads_entry->stream_name = p; + ads_entry->stream_name_nbytes = name_nbytes; + } else { + if (name && *(const tchar*)name != T('\0')) { + ret = get_utf16le_name(name, &ads_entry->stream_name, + &ads_entry->stream_name_nbytes); + } } + return ret; } -void free_inode(struct inode *inode) +static void +destroy_ads_entry(struct wim_ads_entry *ads_entry) { - wimlib_assert(inode); - inode_free_ads_entries(inode); - FREE(inode); + FREE(ads_entry->stream_name); } -void put_inode(struct inode *inode) +/* Frees an inode. */ +void +free_inode(struct wim_inode *inode) { if (inode) { - wimlib_assert(inode->link_count); - if (--inode->link_count) - free_inode(inode); + if (inode->i_ads_entries) { + for (u16 i = 0; i < inode->i_num_ads; i++) + destroy_ads_entry(&inode->i_ads_entries[i]); + FREE(inode->i_ads_entries); + } + #ifdef WITH_FUSE + wimlib_assert(inode->i_num_opened_fds == 0); + FREE(inode->i_fds); + pthread_mutex_destroy(&inode->i_mutex); + #endif + /* HACK: This may instead delete the inode from i_list, but the + * hlist_del() behaves the same as list_del(). */ + hlist_del(&inode->i_hlist); + FREE(inode->i_extracted_file); + FREE(inode); } } -/* Frees a WIM dentry. */ -void free_dentry(struct dentry *dentry) +/* Decrements link count on an inode and frees it if the link count reaches 0. + * */ +static void +put_inode(struct wim_inode *inode) { - wimlib_assert(dentry); - FREE(dentry->file_name); - FREE(dentry->file_name_utf8); - FREE(dentry->short_name); - FREE(dentry->full_path_utf8); - put_inode(dentry->inode); - FREE(dentry); + wimlib_assert(inode->i_nlink != 0); + if (--inode->i_nlink == 0) { + #ifdef WITH_FUSE + if (inode->i_num_opened_fds == 0) + #endif + { + free_inode(inode); + } + } } -/* Partically clones a dentry. +/* Frees a WIM dentry. * - * Beware: - * - memory for file names is not cloned (the pointers are all set to NULL - * and the lengths are set to zero) - * - next, prev, and children pointers and not touched + * The corresponding inode (if any) is freed only if its link count is + * decremented to 0. */ -struct dentry *clone_dentry(struct dentry *old) +void +free_dentry(struct wim_dentry *dentry) { - struct dentry *new = MALLOC(sizeof(struct dentry)); - if (!new) - return NULL; - memcpy(new, old, sizeof(struct dentry)); - new->file_name = NULL; - new->file_name_len = 0; - new->file_name_utf8 = NULL; - new->file_name_utf8_len = 0; - new->short_name = NULL; - new->short_name_len = 0; - return new; -} - -/* - * This function is passed as an argument to for_dentry_in_tree_depth() in order - * to free a directory tree. __args is a pointer to a `struct free_dentry_args'. - */ -static int do_free_dentry(struct dentry *dentry, void *__lookup_table) + if (dentry) { + FREE(dentry->file_name); + FREE(dentry->short_name); + FREE(dentry->_full_path); + if (dentry->d_inode) + put_inode(dentry->d_inode); + FREE(dentry); + } +} + +/* This function is passed as an argument to for_dentry_in_tree_depth() in order + * to free a directory tree. */ +static int +do_free_dentry(struct wim_dentry *dentry, void *_lookup_table) { - struct lookup_table *lookup_table = __lookup_table; - struct lookup_table_entry *lte; - struct inode *inode = dentry->inode; - unsigned i; + struct wim_lookup_table *lookup_table = _lookup_table; if (lookup_table) { - for (i = 0; i <= inode->num_ads; i++) { + struct wim_inode *inode = dentry->d_inode; + for (unsigned i = 0; i <= inode->i_num_ads; i++) { + struct wim_lookup_table_entry *lte; + lte = inode_stream_lte(inode, i, lookup_table); - lte_decrement_refcnt(lte, lookup_table); + if (lte) + lte_decrement_refcnt(lte, lookup_table); } } - - wimlib_assert(dentry->refcnt != 0); - if (--dentry->refcnt == 0) - free_dentry(dentry); + free_dentry(dentry); return 0; } -/* +/* * Unlinks and frees a dentry tree. * * @root: The root of the tree. @@ -749,395 +1006,561 @@ static int do_free_dentry(struct dentry *dentry, void *__lookup_table) * table entries corresponding to the dentries will be * decremented. */ -void free_dentry_tree(struct dentry *root, struct lookup_table *lookup_table) +void +free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table) { - if (!root || !root->parent) - return; for_dentry_in_tree_depth(root, do_free_dentry, lookup_table); } -int increment_dentry_refcnt(struct dentry *dentry, void *ignore) -{ - dentry->refcnt++; - return 0; -} - -/* +/* * Links a dentry into the directory tree. * - * @dentry: The dentry to link. - * @parent: The dentry that will be the parent of @dentry. + * @parent: The dentry that will be the parent of @child. + * @child: The dentry to link. + * + * Returns non-NULL if a duplicate dentry was detected. */ -void link_dentry(struct dentry *dentry, struct dentry *parent) +struct wim_dentry * +dentry_add_child(struct wim_dentry * restrict parent, + struct wim_dentry * restrict child) { wimlib_assert(dentry_is_directory(parent)); - dentry->parent = parent; - if (parent->children) { - /* Not an only child; link to siblings. */ - dentry->next = parent->children; - dentry->prev = parent->children->prev; - dentry->next->prev = dentry; - dentry->prev->next = dentry; - } else { - /* Only child; link to parent. */ - parent->children = dentry; - dentry->next = dentry; - dentry->prev = dentry; - } -} + struct rb_root *root = &parent->d_inode->i_children; + struct rb_node **new = &(root->rb_node); + struct rb_node *rb_parent = NULL; -/* Unlink a dentry from the directory tree. - * - * Note: This merely removes it from the in-memory tree structure. See - * remove_dentry() in mount.c for a function implemented on top of this one that - * frees the dentry and implements reference counting for the lookup table - * entries. */ -void unlink_dentry(struct dentry *dentry) -{ - if (dentry_is_root(dentry)) - return; - if (dentry_is_only_child(dentry)) { - dentry->parent->children = NULL; - } else { - if (dentry_is_first_sibling(dentry)) - dentry->parent->children = dentry->next; - dentry->next->prev = dentry->prev; - dentry->prev->next = dentry->next; + while (*new) { + struct wim_dentry *this = rbnode_dentry(*new); + int result = dentry_compare_names(child, this); + + rb_parent = *new; + + if (result < 0) + new = &((*new)->rb_left); + else if (result > 0) + new = &((*new)->rb_right); + else + return this; } + child->parent = parent; + rb_link_node(&child->rb_node, rb_parent, new); + rb_insert_color(&child->rb_node, root); + return NULL; } -/* Duplicates a UTF-8 name into UTF-8 and UTF-16 strings and returns the strings - * and their lengths in the pointer arguments */ -int get_names(char **name_utf16_ret, char **name_utf8_ret, - u16 *name_utf16_len_ret, u16 *name_utf8_len_ret, - const char *name) +/* Unlink a WIM dentry from the directory entry tree. */ +void +unlink_dentry(struct wim_dentry *dentry) { - size_t utf8_len; - size_t utf16_len; - char *name_utf16, *name_utf8; - - utf8_len = strlen(name); + if (!dentry_is_root(dentry)) + rb_erase(&dentry->rb_node, &dentry->parent->d_inode->i_children); +} - name_utf16 = utf8_to_utf16(name, utf8_len, &utf16_len); +/* + * Returns the alternate data stream entry belonging to @inode that has the + * stream name @stream_name. + */ +struct wim_ads_entry * +inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name, + u16 *idx_ret) +{ + if (inode->i_num_ads == 0) { + return NULL; + } else { + size_t stream_name_utf16le_nbytes; + u16 i; + struct wim_ads_entry *result; + + #if TCHAR_IS_UTF16LE + const utf16lechar *stream_name_utf16le; + + stream_name_utf16le = stream_name; + stream_name_utf16le_nbytes = tstrlen(stream_name) * sizeof(tchar); + #else + utf16lechar *stream_name_utf16le; + + { + int ret = tstr_to_utf16le(stream_name, + tstrlen(stream_name) * + sizeof(tchar), + &stream_name_utf16le, + &stream_name_utf16le_nbytes); + if (ret) + return NULL; + } + #endif + i = 0; + result = NULL; + do { + if (ads_entry_has_name(&inode->i_ads_entries[i], + stream_name_utf16le, + stream_name_utf16le_nbytes)) + { + if (idx_ret) + *idx_ret = i; + result = &inode->i_ads_entries[i]; + break; + } + } while (++i != inode->i_num_ads); + #if !TCHAR_IS_UTF16LE + FREE(stream_name_utf16le); + #endif + return result; + } +} + +static struct wim_ads_entry * +do_inode_add_ads(struct wim_inode *inode, const void *stream_name, + size_t stream_name_nbytes, bool is_utf16le) +{ + u16 num_ads; + struct wim_ads_entry *ads_entries; + struct wim_ads_entry *new_entry; + + if (inode->i_num_ads >= 0xfffe) { + ERROR("Too many alternate data streams in one inode!"); + return NULL; + } + num_ads = inode->i_num_ads + 1; + ads_entries = REALLOC(inode->i_ads_entries, + num_ads * sizeof(inode->i_ads_entries[0])); + if (!ads_entries) { + ERROR("Failed to allocate memory for new alternate data stream"); + return NULL; + } + inode->i_ads_entries = ads_entries; + + new_entry = &inode->i_ads_entries[num_ads - 1]; + if (init_ads_entry(new_entry, stream_name, stream_name_nbytes, is_utf16le)) + return NULL; + new_entry->stream_id = inode->i_next_stream_id++; + inode->i_num_ads = num_ads; + return new_entry; +} + +struct wim_ads_entry * +inode_add_ads_utf16le(struct wim_inode *inode, + const utf16lechar *stream_name, + size_t stream_name_nbytes) +{ + DEBUG("Add alternate data stream \"%"WS"\"", stream_name); + return do_inode_add_ads(inode, stream_name, stream_name_nbytes, true); +} + +/* + * Add an alternate stream entry to a WIM inode and return a pointer to it, or + * NULL if memory could not be allocated. + */ +struct wim_ads_entry * +inode_add_ads(struct wim_inode *inode, const tchar *stream_name) +{ + DEBUG("Add alternate data stream \"%"TS"\"", stream_name); + return do_inode_add_ads(inode, stream_name, + tstrlen(stream_name) * sizeof(tchar), + TCHAR_IS_UTF16LE); +} + +static struct wim_lookup_table_entry * +add_stream_from_data_buffer(const void *buffer, size_t size, + struct wim_lookup_table *lookup_table) +{ + u8 hash[SHA1_HASH_SIZE]; + struct wim_lookup_table_entry *lte, *existing_lte; + + sha1_buffer(buffer, size, hash); + existing_lte = __lookup_resource(lookup_table, hash); + if (existing_lte) { + wimlib_assert(wim_resource_size(existing_lte) == size); + lte = existing_lte; + lte->refcnt++; + } else { + void *buffer_copy; + lte = new_lookup_table_entry(); + if (!lte) + return NULL; + buffer_copy = MALLOC(size); + if (!buffer_copy) { + free_lookup_table_entry(lte); + return NULL; + } + memcpy(buffer_copy, buffer, size); + lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER; + lte->attached_buffer = buffer_copy; + lte->resource_entry.original_size = size; + copy_hash(lte->hash, hash); + lookup_table_insert(lookup_table, lte); + } + return lte; +} + +int +inode_add_ads_with_data(struct wim_inode *inode, const tchar *name, + const void *value, size_t size, + struct wim_lookup_table *lookup_table) +{ + struct wim_ads_entry *new_ads_entry; + + wimlib_assert(inode->i_resolved); - if (!name_utf16) + new_ads_entry = inode_add_ads(inode, name); + if (!new_ads_entry) return WIMLIB_ERR_NOMEM; - name_utf8 = MALLOC(utf8_len + 1); - if (!name_utf8) { - FREE(name_utf8); + new_ads_entry->lte = add_stream_from_data_buffer(value, size, + lookup_table); + if (!new_ads_entry->lte) { + inode_remove_ads(inode, new_ads_entry - inode->i_ads_entries, + lookup_table); return WIMLIB_ERR_NOMEM; } - memcpy(name_utf8, name, utf8_len + 1); - FREE(*name_utf8_ret); - FREE(*name_utf16_ret); - *name_utf8_ret = name_utf8; - *name_utf16_ret = name_utf16; - *name_utf8_len_ret = utf8_len; - *name_utf16_len_ret = utf16_len; return 0; } -/* Changes the name of a dentry to @new_name. Only changes the file_name and - * file_name_utf8 fields; does not change the short_name, short_name_utf8, or - * full_path_utf8 fields. Also recalculates its length. */ -int change_dentry_name(struct dentry *dentry, const char *new_name) +/* Set the unnamed stream of a WIM inode, given a data buffer containing the + * stream contents. */ +int +inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len, + struct wim_lookup_table *lookup_table) { - int ret; - - ret = get_names(&dentry->file_name, &dentry->file_name_utf8, - &dentry->file_name_len, &dentry->file_name_utf8_len, - new_name); - FREE(dentry->short_name); - dentry->short_name_len = 0; - if (ret == 0) - dentry->length = dentry_correct_length(dentry); - return ret; + inode->i_lte = add_stream_from_data_buffer(data, len, lookup_table); + if (!inode->i_lte) + return WIMLIB_ERR_NOMEM; + inode->i_resolved = 1; + return 0; } -/* - * Changes the name of an alternate data stream */ -int change_ads_name(struct ads_entry *entry, const char *new_name) -{ - return get_names(&entry->stream_name, &entry->stream_name_utf8, - &entry->stream_name_len, - &entry->stream_name_utf8_len, - new_name); -} - -/* Parameters for calculate_dentry_statistics(). */ -struct image_statistics { - struct lookup_table *lookup_table; - u64 *dir_count; - u64 *file_count; - u64 *total_bytes; - u64 *hard_link_bytes; -}; +/* Remove an alternate data stream from a WIM inode */ +void +inode_remove_ads(struct wim_inode *inode, u16 idx, + struct wim_lookup_table *lookup_table) +{ + struct wim_ads_entry *ads_entry; + struct wim_lookup_table_entry *lte; + + wimlib_assert(idx < inode->i_num_ads); + wimlib_assert(inode->i_resolved); + + ads_entry = &inode->i_ads_entries[idx]; + + DEBUG("Remove alternate data stream \"%"WS"\"", ads_entry->stream_name); + + lte = ads_entry->lte; + if (lte) + lte_decrement_refcnt(lte, lookup_table); + + destroy_ads_entry(ads_entry); + + memmove(&inode->i_ads_entries[idx], + &inode->i_ads_entries[idx + 1], + (inode->i_num_ads - idx - 1) * sizeof(inode->i_ads_entries[0])); + inode->i_num_ads--; +} -static int calculate_dentry_statistics(struct dentry *dentry, void *arg) +#ifndef __WIN32__ +int +inode_get_unix_data(const struct wim_inode *inode, + struct wimlib_unix_data *unix_data, + u16 *stream_idx_ret) { - struct image_statistics *stats; - struct lookup_table_entry *lte; - - stats = arg; + const struct wim_ads_entry *ads_entry; + const struct wim_lookup_table_entry *lte; + size_t size; + int ret; - if (dentry_is_directory(dentry) && !dentry_is_root(dentry)) - ++*stats->dir_count; - else - ++*stats->file_count; - - for (unsigned i = 0; i <= dentry->inode->num_ads; i++) { - lte = inode_stream_lte(dentry->inode, i, stats->lookup_table); - if (lte) { - *stats->total_bytes += wim_resource_size(lte); - if (++lte->out_refcnt == 1) - *stats->hard_link_bytes += wim_resource_size(lte); - } - } + wimlib_assert(inode->i_resolved); + + ads_entry = inode_get_ads_entry((struct wim_inode*)inode, + WIMLIB_UNIX_DATA_TAG, NULL); + if (!ads_entry) + return NO_UNIX_DATA; + + if (stream_idx_ret) + *stream_idx_ret = ads_entry - inode->i_ads_entries; + + lte = ads_entry->lte; + if (!lte) + return NO_UNIX_DATA; + + size = wim_resource_size(lte); + if (size != sizeof(struct wimlib_unix_data)) + return BAD_UNIX_DATA; + + ret = read_full_resource_into_buf(lte, unix_data); + if (ret) + return ret; + + if (unix_data->version != 0) + return BAD_UNIX_DATA; return 0; } -/* Calculates some statistics about a dentry tree. */ -void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *table, - u64 *dir_count_ret, u64 *file_count_ret, - u64 *total_bytes_ret, - u64 *hard_link_bytes_ret) +int +inode_set_unix_data(struct wim_inode *inode, uid_t uid, gid_t gid, mode_t mode, + struct wim_lookup_table *lookup_table, int which) { - struct image_statistics stats; - *dir_count_ret = 0; - *file_count_ret = 0; - *total_bytes_ret = 0; - *hard_link_bytes_ret = 0; - stats.lookup_table = table; - stats.dir_count = dir_count_ret; - stats.file_count = file_count_ret; - stats.total_bytes = total_bytes_ret; - stats.hard_link_bytes = hard_link_bytes_ret; - for_lookup_table_entry(table, zero_out_refcnts, NULL); - for_dentry_in_tree(root, calculate_dentry_statistics, &stats); + struct wimlib_unix_data unix_data; + int ret; + bool have_good_unix_data = false; + bool have_unix_data = false; + u16 stream_idx; + + if (!(which & UNIX_DATA_CREATE)) { + ret = inode_get_unix_data(inode, &unix_data, &stream_idx); + if (ret == 0 || ret == BAD_UNIX_DATA || ret > 0) + have_unix_data = true; + if (ret == 0) + have_good_unix_data = true; + } + unix_data.version = 0; + if (which & UNIX_DATA_UID || !have_good_unix_data) + unix_data.uid = uid; + if (which & UNIX_DATA_GID || !have_good_unix_data) + unix_data.gid = gid; + if (which & UNIX_DATA_MODE || !have_good_unix_data) + unix_data.mode = mode; + ret = inode_add_ads_with_data(inode, WIMLIB_UNIX_DATA_TAG, + &unix_data, + sizeof(struct wimlib_unix_data), + lookup_table); + if (ret == 0 && have_unix_data) + inode_remove_ads(inode, stream_idx, lookup_table); + return ret; } +#endif /* !__WIN32__ */ +/* Replace weird characters in filenames and alternate data stream names. + * + * In particular we do not want the path separator to appear in any names, as + * that would make it possible for a "malicious" WIM to extract itself to any + * location it wanted to. */ +static void +replace_forbidden_characters(utf16lechar *name) +{ + utf16lechar *p; + + for (p = name; *p; p++) { + #ifdef __WIN32__ + if (wcschr(L"<>:\"/\\|?*", (wchar_t)*p)) + #else + if (*p == cpu_to_le16('/')) + #endif + { + if (name) { + WARNING("File, directory, or stream name \"%"WS"\"\n" + " contains forbidden characters; " + "substituting replacement characters.", + name); + name = NULL; + } + #ifdef __WIN32__ + *p = cpu_to_le16(0xfffd); + #else + *p = cpu_to_le16('?'); + #endif + } + } +} -/* - * Reads the alternate data stream entries for a dentry. +/* + * Reads the alternate data stream entries of a WIM dentry. * * @p: Pointer to buffer that starts with the first alternate stream entry. * * @inode: Inode to load the alternate data streams into. - * @inode->num_ads must have been set to the number of + * @inode->i_num_ads must have been set to the number of * alternate data streams that are expected. * * @remaining_size: Number of bytes of data remaining in the buffer pointed * to by @p. * - * The format of the on-disk alternate stream entries is as follows: - * - * struct ads_entry_on_disk { - * u64 length; // Length of the entry, in bytes. This includes - * all fields (including the stream name and - * null terminator if present, AND the padding!). - * u64 reserved; // Seems to be unused - * u8 hash[20]; // SHA1 message digest of the uncompressed stream - * u16 stream_name_len; // Length of the stream name, in bytes - * char stream_name[]; // Stream name in UTF-16LE, @stream_name_len bytes long, - * not including null terminator - * u16 zero; // UTF-16 null terminator for the stream name, NOT - * included in @stream_name_len. Based on what - * I've observed from filenames in dentries, - * this field should not exist when - * (@stream_name_len == 0), but you can't - * actually tell because of the padding anyway - * (provided that the padding is zeroed, which - * it always seems to be). - * char padding[]; // Padding to make the size a multiple of 8 bytes. - * }; - * - * In addition, the entries are 8-byte aligned. * - * Return 0 on success or nonzero on failure. On success, inode->ads_entries - * is set to an array of `struct ads_entry's of length inode->num_ads. On + * Return 0 on success or nonzero on failure. On success, inode->i_ads_entries + * is set to an array of `struct wim_ads_entry's of length inode->i_num_ads. On * failure, @inode is not modified. */ -static int read_ads_entries(const u8 *p, struct inode *inode, - u64 remaining_size) +static int +read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode, + size_t nbytes_remaining) { u16 num_ads; - struct ads_entry **ads_entries; + struct wim_ads_entry *ads_entries; int ret; - num_ads = inode->num_ads; - ads_entries = CALLOC(num_ads, sizeof(inode->ads_entries[0])); - if (!ads_entries) { - ERROR("Could not allocate memory for %"PRIu16" " - "alternate data stream entries", num_ads); - return WIMLIB_ERR_NOMEM; - } + num_ads = inode->i_num_ads; + ads_entries = CALLOC(num_ads, sizeof(inode->i_ads_entries[0])); + if (!ads_entries) + goto out_of_memory; for (u16 i = 0; i < num_ads; i++) { - struct ads_entry *cur_entry; u64 length; - u64 length_no_padding; - u64 total_length; - size_t utf8_len; - const u8 *p_save = p; - - cur_entry = new_ads_entry(NULL); - if (!cur_entry) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_ads_entries; - } - - ads_entries[i] = cur_entry; - - /* Read the base stream entry, excluding the stream name. */ - if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) { - ERROR("Stream entries go past end of metadata resource"); - ERROR("(remaining_size = %"PRIu64")", remaining_size); - ret = WIMLIB_ERR_INVALID_DENTRY; - goto out_free_ads_entries; - } - - p = get_u64(p, &length); - p += 8; /* Skip the reserved field */ - p = get_bytes(p, SHA1_HASH_SIZE, (u8*)cur_entry->hash); - p = get_u16(p, &cur_entry->stream_name_len); - - cur_entry->stream_name = NULL; - cur_entry->stream_name_utf8 = NULL; - - /* Length including neither the null terminator nor the padding - * */ - length_no_padding = WIM_ADS_ENTRY_DISK_SIZE + - cur_entry->stream_name_len; - - /* Length including the null terminator and the padding */ - total_length = ((length_no_padding + 2) + 7) & ~7; - - wimlib_assert(total_length == ads_entry_total_length(cur_entry)); - - if (remaining_size < length_no_padding) { - ERROR("Stream entries go past end of metadata resource"); - ERROR("(remaining_size = %"PRIu64" bytes, " - "length_no_padding = %"PRIu64" bytes)", - remaining_size, length_no_padding); - ret = WIMLIB_ERR_INVALID_DENTRY; - goto out_free_ads_entries; + struct wim_ads_entry *cur_entry; + const struct wim_ads_entry_on_disk *disk_entry = + (const struct wim_ads_entry_on_disk*)p; + + cur_entry = &ads_entries[i]; + ads_entries[i].stream_id = i + 1; + + /* Do we have at least the size of the fixed-length data we know + * need? */ + if (nbytes_remaining < sizeof(struct wim_ads_entry_on_disk)) + goto out_invalid; + + /* Read the length field */ + length = le64_to_cpu(disk_entry->length); + + /* Make sure the length field is neither so small it doesn't + * include all the fixed-length data, or so large it overflows + * the metadata resource buffer. */ + if (length < sizeof(struct wim_ads_entry_on_disk) || + length > nbytes_remaining) + goto out_invalid; + + /* Read the rest of the fixed-length data. */ + + cur_entry->reserved = le64_to_cpu(disk_entry->reserved); + + copy_hash(cur_entry->hash, disk_entry->hash); + cur_entry->stream_name_nbytes = le16_to_cpu(cur_entry->stream_name_nbytes); + + /* If stream_name_nbytes != 0, this is a named stream. + * Otherwise this is an unnamed stream, or in some cases (bugs + * in Microsoft's software I guess) a meaningless entry + * distinguished from the real unnamed stream entry, if any, by + * the fact that the real unnamed stream entry has a nonzero + * hash field. */ + if (cur_entry->stream_name_nbytes) { + u64 length_no_padding; + + /* The name is encoded in UTF16-LE, which uses 2-byte + * coding units, so the length of the name had better be + * an even number of bytes... */ + if (cur_entry->stream_name_nbytes & 1) + goto out_invalid; + + /* Add the length of the stream name to get the length + * we actually need to read. Make sure this isn't more + * than the specified length of the entry. */ + if (sizeof(struct wim_ads_entry_on_disk) + + cur_entry->stream_name_nbytes > length) + goto out_invalid; + + cur_entry->stream_name = MALLOC(cur_entry->stream_name_nbytes + 2); + if (!cur_entry->stream_name) + goto out_of_memory; + + memcpy(cur_entry->stream_name, + disk_entry->stream_name, + cur_entry->stream_name_nbytes); + cur_entry->stream_name[cur_entry->stream_name_nbytes / 2] = 0; + replace_forbidden_characters(cur_entry->stream_name); } - /* The @length field in the on-disk ADS entry is expected to be - * equal to @total_length, which includes all of the entry and - * the padding that follows it to align the next ADS entry to an - * 8-byte boundary. However, to be safe, we'll accept the - * length field as long as it's not less than the un-padded - * total length and not more than the padded total length. */ - if (length < length_no_padding || length > total_length) { - ERROR("Stream entry has unexpected length " - "field (length field = %"PRIu64", " - "unpadded total length = %"PRIu64", " - "padded total length = %"PRIu64")", - length, length_no_padding, total_length); - ret = WIMLIB_ERR_INVALID_DENTRY; - goto out_free_ads_entries; - } - - if (cur_entry->stream_name_len) { - cur_entry->stream_name = MALLOC(cur_entry->stream_name_len); - if (!cur_entry->stream_name) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_ads_entries; - } - get_bytes(p, cur_entry->stream_name_len, - (u8*)cur_entry->stream_name); - cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name, - cur_entry->stream_name_len, - &utf8_len); - cur_entry->stream_name_utf8_len = utf8_len; - - if (!cur_entry->stream_name_utf8) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_ads_entries; - } - } /* It's expected that the size of every ADS entry is a multiple * of 8. However, to be safe, I'm allowing the possibility of * an ADS entry at the very end of the metadata resource ending * un-aligned. So although we still need to increment the input - * pointer by @total_length to reach the next ADS entry, it's - * possible that less than @total_length is actually remaining - * in the metadata resource. We should set the remaining size to - * 0 bytes if this happens. */ - p = p_save + total_length; - if (remaining_size < total_length) - remaining_size = 0; + * pointer by @length to reach the next ADS entry, it's possible + * that less than @length is actually remaining in the metadata + * resource. We should set the remaining bytes to 0 if this + * happens. */ + length = (length + 7) & ~(u64)7; + p += length; + if (nbytes_remaining < length) + nbytes_remaining = 0; else - remaining_size -= total_length; + nbytes_remaining -= length; } - inode->ads_entries = ads_entries; - return 0; + inode->i_ads_entries = ads_entries; +#ifdef WITH_FUSE + inode->i_next_stream_id = inode->i_num_ads + 1; +#endif + ret = 0; + goto out; +out_of_memory: + ret = WIMLIB_ERR_NOMEM; + goto out_free_ads_entries; +out_invalid: + ERROR("An alternate data stream entry is invalid"); + ret = WIMLIB_ERR_INVALID_DENTRY; out_free_ads_entries: - for (u16 i = 0; i < num_ads; i++) - free_ads_entry(ads_entries[i]); - FREE(ads_entries); + if (ads_entries) { + for (u16 i = 0; i < num_ads; i++) + destroy_ads_entry(&ads_entries[i]); + FREE(ads_entries); + } +out: return ret; } -/* - * Reads a directory entry, including all alternate data stream entries that +/* + * Reads a WIM directory entry, including all alternate data stream entries that * follow it, from the WIM image's metadata resource. * - * @metadata_resource: Buffer containing the uncompressed metadata resource. - * @metadata_resource_len: Length of the metadata resource. - * @offset: Offset of this directory entry in the metadata resource. - * @dentry: A `struct dentry' that will be filled in by this function. + * @metadata_resource: + * Pointer to the metadata resource buffer. + * + * @metadata_resource_len: + * Length of the metadata resource buffer, in bytes. + * + * @offset: Offset of the dentry within the metadata resource. + * + * @dentry: A `struct wim_dentry' that will be filled in by this function. * - * Return 0 on success or nonzero on failure. On failure, @dentry have been - * modified, bu it will be left with no pointers to any allocated buffers. - * On success, the dentry->length field must be examined. If zero, this was a - * special "end of directory" dentry and not a real dentry. If nonzero, this - * was a real dentry. + * Return 0 on success or nonzero on failure. On failure, @dentry will have + * been modified, but it will not be left with pointers to any allocated + * buffers. On success, the dentry->length field must be examined. If zero, + * this was a special "end of directory" dentry and not a real dentry. If + * nonzero, this was a real dentry. */ -int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len, - u64 offset, struct dentry *dentry) +int +read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, + u64 offset, struct wim_dentry * restrict dentry) { - const u8 *p; + u64 calculated_size; - char *file_name = NULL; - char *file_name_utf8 = NULL; - char *short_name = NULL; - u16 short_name_len; - u16 file_name_len; - size_t file_name_utf8_len = 0; + utf16lechar *file_name; + utf16lechar *short_name; + u16 short_name_nbytes; + u16 file_name_nbytes; int ret; - struct inode *inode = NULL; + struct wim_inode *inode; + const struct wim_dentry_on_disk *disk_dentry; + const u8 *p = &metadata_resource[offset]; + + if ((uintptr_t)p & 7) + WARNING("WIM dentry is not 8-byte aligned"); dentry_common_init(dentry); /*Make sure the dentry really fits into the metadata resource.*/ - if (offset + 8 > metadata_resource_len || offset + 8 < offset) { + if (offset + sizeof(u64) < offset || + offset + sizeof(u64) > metadata_resource_len) + { ERROR("Directory entry starting at %"PRIu64" ends past the " "end of the metadata resource (size %"PRIu64")", offset, metadata_resource_len); return WIMLIB_ERR_INVALID_DENTRY; } + disk_dentry = (const struct wim_dentry_on_disk*)p; + /* Before reading the whole dentry, we need to read just the length. * This is because a dentry of length 8 (that is, just the length field) * terminates the list of sibling directory entries. */ - - p = get_u64(&metadata_resource[offset], &dentry->length); + dentry->length = le64_to_cpu(disk_dentry->length); /* A zero length field (really a length of 8, since that's how big the * directory entry is...) indicates that this is the end of directory * dentry. We do not read it into memory as an actual dentry, so just * return successfully in that case. */ + if (dentry->length == 8) + dentry->length = 0; if (dentry->length == 0) return 0; - /* If the dentry does not overflow the metadata resource buffer and is - * not too short, read the rest of it (excluding the alternate data - * streams, but including the file name and short name variable-length - * fields) into memory. */ + /* Now that we have the actual length provided in the on-disk structure, + * make sure it doesn't overflow the metadata buffer. */ if (offset + dentry->length >= metadata_resource_len || offset + dentry->length < offset) { @@ -1148,394 +1571,377 @@ int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len, return WIMLIB_ERR_INVALID_DENTRY; } - if (dentry->length < WIM_DENTRY_DISK_SIZE) { + /* Make sure the dentry length is at least as large as the number of + * fixed-length fields */ + if (dentry->length < sizeof(struct wim_dentry_on_disk)) { ERROR("Directory entry has invalid length of %"PRIu64" bytes", dentry->length); return WIMLIB_ERR_INVALID_DENTRY; } - inode = new_inode(); + /* Allocate a `struct wim_inode' for this `struct wim_dentry'. */ + inode = new_timeless_inode(); if (!inode) return WIMLIB_ERR_NOMEM; - p = get_u32(p, &inode->attributes); - p = get_u32(p, (u32*)&inode->security_id); - p = get_u64(p, &dentry->subdir_offset); + /* Read more fields; some into the dentry, and some into the inode. */ - /* 2 unused fields */ - p += 2 * sizeof(u64); - /*p = get_u64(p, &dentry->unused1);*/ - /*p = get_u64(p, &dentry->unused2);*/ + inode->i_attributes = le32_to_cpu(disk_dentry->attributes); + inode->i_security_id = le32_to_cpu(disk_dentry->security_id); + dentry->subdir_offset = le64_to_cpu(disk_dentry->subdir_offset); + dentry->d_unused_1 = le64_to_cpu(disk_dentry->unused_1); + dentry->d_unused_2 = le64_to_cpu(disk_dentry->unused_2); + inode->i_creation_time = le64_to_cpu(disk_dentry->creation_time); + inode->i_last_access_time = le64_to_cpu(disk_dentry->last_access_time); + inode->i_last_write_time = le64_to_cpu(disk_dentry->last_write_time); + copy_hash(inode->i_hash, disk_dentry->unnamed_stream_hash); - p = get_u64(p, &inode->creation_time); - p = get_u64(p, &inode->last_access_time); - p = get_u64(p, &inode->last_write_time); - - p = get_bytes(p, SHA1_HASH_SIZE, inode->hash); - - /* - * I don't know what's going on here. It seems like M$ screwed up the + /* I don't know what's going on here. It seems like M$ screwed up the * reparse points, then put the fields in the same place and didn't - * document it. The WIM_HDR_FLAG_RP_FIX flag in the WIM header might - * have something to do with this, but it's not documented. - */ - if (inode->attributes & FILE_ATTRIBUTE_REPARSE_POINT) { - /* ??? */ - p += 4; - p = get_u32(p, &inode->reparse_tag); - p += 4; + * document it. So we have some fields we read for reparse points, and + * some fields in the same place for non-reparse-point.s */ + if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { + inode->i_rp_unknown_1 = le32_to_cpu(disk_dentry->reparse.rp_unknown_1); + inode->i_reparse_tag = le32_to_cpu(disk_dentry->reparse.reparse_tag); + inode->i_rp_unknown_2 = le16_to_cpu(disk_dentry->reparse.rp_unknown_2); + inode->i_not_rpfixed = le16_to_cpu(disk_dentry->reparse.not_rpfixed); + /* Leave inode->i_ino at 0. Note that this means the WIM file + * cannot archive hard-linked reparse points. Such a thing + * doesn't really make sense anyway, although I believe it's + * theoretically possible to have them on NTFS. */ } else { - p = get_u32(p, &inode->reparse_tag); - p = get_u64(p, &dentry->link_group_id); + inode->i_rp_unknown_1 = le32_to_cpu(disk_dentry->nonreparse.rp_unknown_1); + inode->i_ino = le64_to_cpu(disk_dentry->nonreparse.hard_link_group_id); } - /* By the way, the reparse_reserved field does not actually exist (at - * least when the file is not a reparse point) */ - - p = get_u16(p, &inode->num_ads); + inode->i_num_ads = le16_to_cpu(disk_dentry->num_alternate_data_streams); - p = get_u16(p, &short_name_len); - p = get_u16(p, &file_name_len); + short_name_nbytes = le16_to_cpu(disk_dentry->short_name_nbytes); + file_name_nbytes = le16_to_cpu(disk_dentry->file_name_nbytes); + + if ((short_name_nbytes & 1) | (file_name_nbytes & 1)) + { + ERROR("Dentry name is not valid UTF-16LE (odd number of bytes)!"); + ret = WIMLIB_ERR_INVALID_DENTRY; + goto out_free_inode; + } /* We now know the length of the file name and short name. Make sure - * the length of the dentry is large enough to actually hold them. + * the length of the dentry is large enough to actually hold them. * * The calculated length here is unaligned to allow for the possibility * that the dentry->length names an unaligned length, although this * would be unexpected. */ - calculated_size = __dentry_correct_length_unaligned(file_name_len, - short_name_len); + calculated_size = _dentry_correct_length_unaligned(file_name_nbytes, + short_name_nbytes); if (dentry->length < calculated_size) { ERROR("Unexpected end of directory entry! (Expected " - "at least %"PRIu64" bytes, got %"PRIu64" bytes. " - "short_name_len = %hu, file_name_len = %hu)", - calculated_size, dentry->length, - short_name_len, file_name_len); - return WIMLIB_ERR_INVALID_DENTRY; + "at least %"PRIu64" bytes, got %"PRIu64" bytes.)", + calculated_size, dentry->length); + ret = WIMLIB_ERR_INVALID_DENTRY; + goto out_free_inode; } + p += sizeof(struct wim_dentry_on_disk); + /* Read the filename if present. Note: if the filename is empty, there * is no null terminator following it. */ - if (file_name_len) { - file_name = MALLOC(file_name_len); + if (file_name_nbytes) { + file_name = MALLOC(file_name_nbytes + 2); if (!file_name) { - ERROR("Failed to allocate %hu bytes for dentry file name", - file_name_len); - return WIMLIB_ERR_NOMEM; - } - p = get_bytes(p, file_name_len, file_name); - - /* Convert filename to UTF-8. */ - file_name_utf8 = utf16_to_utf8(file_name, file_name_len, - &file_name_utf8_len); - - if (!file_name_utf8) { - ERROR("Failed to allocate memory to convert UTF-16 " - "filename (%hu bytes) to UTF-8", file_name_len); + ERROR("Failed to allocate %d bytes for dentry file name", + file_name_nbytes + 2); ret = WIMLIB_ERR_NOMEM; - goto out_free_file_name; + goto out_free_inode; } - if (*(u16*)p) - WARNING("Expected two zero bytes following the file name " - "`%s', but found non-zero bytes", file_name_utf8); - p += 2; + memcpy(file_name, p, file_name_nbytes); + p += file_name_nbytes + 2; + file_name[file_name_nbytes / 2] = cpu_to_le16(0); + replace_forbidden_characters(file_name); + } else { + file_name = NULL; } - /* Align the calculated size */ - calculated_size = (calculated_size + 7) & ~7; - - if (dentry->length > calculated_size) { - /* Weird; the dentry says it's longer than it should be. Note - * that the length field does NOT include the size of the - * alternate stream entries. */ - - /* Strangely, some directory entries inexplicably have a little - * over 70 bytes of extra data. The exact amount of data seems - * to be 72 bytes, but it is aligned on the next 8-byte - * boundary. It does NOT seem to be alternate data stream - * entries. Here's an example of the aligned data: - * - * 01000000 40000000 6c786bba c58ede11 b0bb0026 1870892a b6adb76f - * e63a3e46 8fca8653 0d2effa1 6c786bba c58ede11 b0bb0026 1870892a - * 00000000 00000000 00000000 00000000 - * - * Here's one interpretation of how the data is laid out. - * - * struct unknown { - * u32 field1; (always 0x00000001) - * u32 field2; (always 0x40000000) - * u8 data[48]; (???) - * u64 reserved1; (always 0) - * u64 reserved2; (always 0) - * };*/ - DEBUG("Dentry for file or directory `%s' has %zu extra " - "bytes of data", - file_name_utf8, dentry->length - calculated_size); - } /* Read the short filename if present. Note: if there is no short * filename, there is no null terminator following it. */ - if (short_name_len) { - short_name = MALLOC(short_name_len); + if (short_name_nbytes) { + short_name = MALLOC(short_name_nbytes + 2); if (!short_name) { - ERROR("Failed to allocate %hu bytes for short filename", - short_name_len); + ERROR("Failed to allocate %d bytes for dentry short name", + short_name_nbytes + 2); ret = WIMLIB_ERR_NOMEM; - goto out_free_file_name_utf8; + goto out_free_file_name; } - - p = get_bytes(p, short_name_len, short_name); - if (*(u16*)p) - WARNING("Expected two zero bytes following the file name " - "`%s', but found non-zero bytes", file_name_utf8); - p += 2; + memcpy(short_name, p, short_name_nbytes); + p += short_name_nbytes + 2; + short_name[short_name_nbytes / 2] = cpu_to_le16(0); + replace_forbidden_characters(short_name); + } else { + short_name = NULL; } - /* + /* Align the dentry length */ + dentry->length = (dentry->length + 7) & ~7; + + /* * Read the alternate data streams, if present. dentry->num_ads tells * us how many they are, and they will directly follow the dentry * on-disk. * * Note that each alternate data stream entry begins on an 8-byte - * aligned boundary, and the alternate data stream entries are NOT - * included in the dentry->length field for some reason. + * aligned boundary, and the alternate data stream entries seem to NOT + * be included in the dentry->length field for some reason. */ - if (inode->num_ads != 0) { - if (calculated_size > metadata_resource_len - offset) { - ERROR("Not enough space in metadata resource for " - "alternate stream entries"); - ret = WIMLIB_ERR_INVALID_DENTRY; + if (inode->i_num_ads != 0) { + ret = WIMLIB_ERR_INVALID_DENTRY; + if (offset + dentry->length > metadata_resource_len || + (ret = read_ads_entries(&metadata_resource[offset + dentry->length], + inode, + metadata_resource_len - offset - dentry->length))) + { + ERROR("Failed to read alternate data stream " + "entries of WIM dentry \"%"WS"\"", file_name); goto out_free_short_name; } - ret = read_ads_entries(&metadata_resource[offset + calculated_size], - inode, - metadata_resource_len - offset - calculated_size); - if (ret != 0) - goto out_free_short_name; } - +out_success: /* We've read all the data for this dentry. Set the names and their * lengths, and we've done. */ - dentry->file_name = file_name; - dentry->file_name_utf8 = file_name_utf8; - dentry->short_name = short_name; - dentry->file_name_len = file_name_len; - dentry->file_name_utf8_len = file_name_utf8_len; - dentry->short_name_len = short_name_len; - return 0; + dentry->d_inode = inode; + dentry->file_name = file_name; + dentry->short_name = short_name; + dentry->file_name_nbytes = file_name_nbytes; + dentry->short_name_nbytes = short_name_nbytes; + ret = 0; + goto out; out_free_short_name: FREE(short_name); -out_free_file_name_utf8: - FREE(file_name_utf8); out_free_file_name: FREE(file_name); out_free_inode: free_inode(inode); +out: return ret; } -/* Run some miscellaneous verifications on a WIM dentry */ -int verify_dentry(struct dentry *dentry, void *wim) +/* Reads the children of a dentry, and all their children, ..., etc. from the + * metadata resource and into the dentry tree. + * + * @metadata_resource: An array that contains the uncompressed metadata + * resource for the WIM file. + * + * @metadata_resource_len: The length of the uncompressed metadata resource, in + * bytes. + * + * @dentry: A pointer to a `struct wim_dentry' that is the root of the directory + * tree and has already been read from the metadata resource. It + * does not need to be the real root because this procedure is + * called recursively. + * + * Returns zero on success; nonzero on failure. + */ +int +read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, + struct wim_dentry *dentry) { - const WIMStruct *w = wim; - const struct lookup_table *table = w->lookup_table; - const struct wim_security_data *sd = wim_const_security_data(w); - const struct inode *inode = dentry->inode; - int ret = WIMLIB_ERR_INVALID_DENTRY; + u64 cur_offset = dentry->subdir_offset; + struct wim_dentry *child; + struct wim_dentry cur_child; + int ret; - /* Check the security ID */ - if (inode->security_id < -1) { - ERROR("Dentry `%s' has an invalid security ID (%d)", - dentry->full_path_utf8, inode->security_id); - goto out; - } - if (inode->security_id >= sd->num_entries) { - ERROR("Dentry `%s' has an invalid security ID (%d) " - "(there are only %u entries in the security table)", - dentry->full_path_utf8, inode->security_id, - sd->num_entries); - goto out; - } + /* + * If @dentry has no child dentries, nothing more needs to be done for + * this branch. This is the case for regular files, symbolic links, and + * *possibly* empty directories (although an empty directory may also + * have one child dentry that is the special end-of-directory dentry) + */ + if (cur_offset == 0) + return 0; - /* Check that lookup table entries for all the resources exist, except - * if the SHA1 message digest is all 0's, which indicates there is - * intentionally no resource there. */ - if (w->hdr.total_parts == 1) { - for (unsigned i = 0; i <= inode->num_ads; i++) { - struct lookup_table_entry *lte; - const u8 *hash; - hash = inode_stream_hash_unresolved(inode, i); - lte = __lookup_resource(table, hash); - if (!lte && !is_zero_hash(hash)) { - ERROR("Could not find lookup table entry for stream " - "%u of dentry `%s'", i, dentry->full_path_utf8); - goto out; - } - } - } + /* Find and read all the children of @dentry. */ + while (1) { - /* Make sure there is only one un-named stream. */ - unsigned num_unnamed_streams = 0; - for (unsigned i = 0; i <= inode->num_ads; i++) { - const u8 *hash; - hash = inode_stream_hash_unresolved(inode, i); - if (!inode_stream_name_len(inode, i) && !is_zero_hash(hash)) - num_unnamed_streams++; - } - if (num_unnamed_streams > 1) { - ERROR("Dentry `%s' has multiple (%u) un-named streams", - dentry->full_path_utf8, num_unnamed_streams); - goto out; - } + /* Read next child of @dentry into @cur_child. */ + ret = read_dentry(metadata_resource, metadata_resource_len, + cur_offset, &cur_child); + if (ret != 0) + break; - /* Cannot have a short name but no long name */ - if (dentry->short_name_len && !dentry->file_name_len) { - ERROR("Dentry `%s' has a short name but no long name", - dentry->full_path_utf8); - goto out; - } + /* Check for end of directory. */ + if (cur_child.length == 0) + break; - /* Make sure root dentry is unnamed */ - if (dentry_is_root(dentry)) { - if (dentry->file_name_len) { - ERROR("The root dentry is named `%s', but it must " - "be unnamed", dentry->file_name_utf8); - goto out; + /* Not end of directory. Allocate this child permanently and + * link it to the parent and previous child. */ + child = MALLOC(sizeof(struct wim_dentry)); + if (!child) { + ERROR("Failed to allocate %zu bytes for new dentry", + sizeof(struct wim_dentry)); + ret = WIMLIB_ERR_NOMEM; + break; } - } + memcpy(child, &cur_child, sizeof(struct wim_dentry)); + dentry_add_child(dentry, child); + inode_add_dentry(child, child->d_inode); -#if 0 - /* Check timestamps */ - if (inode->last_access_time < inode->creation_time || - inode->last_write_time < inode->creation_time) { - WARNING("Dentry `%s' was created after it was last accessed or " - "written to", dentry->full_path_utf8); - } -#endif + /* If there are children of this child, call this procedure + * recursively. */ + if (child->subdir_offset != 0) { + ret = read_dentry_tree(metadata_resource, + metadata_resource_len, child); + if (ret != 0) + break; + } - ret = 0; -out: + /* Advance to the offset of the next child. Note: We need to + * advance by the TOTAL length of the dentry, not by the length + * child->length, which although it does take into account the + * padding, it DOES NOT take into account alternate stream + * entries. */ + cur_offset += dentry_total_length(child); + } return ret; } -/* +/* * Writes a WIM dentry to an output buffer. * * @dentry: The dentry structure. * @p: The memory location to write the data to. - * @return: Pointer to the byte after the last byte we wrote as part of the - * dentry. + * + * Returns the pointer to the byte after the last byte we wrote as part of the + * dentry, including any alternate data streams entry. */ -static u8 *write_dentry(const struct dentry *dentry, u8 *p) +static u8 * +write_dentry(const struct wim_dentry * restrict dentry, u8 * restrict p) { - u8 *orig_p = p; + const struct wim_inode *inode; + struct wim_dentry_on_disk *disk_dentry; + const u8 *orig_p; const u8 *hash; - const struct inode *inode = dentry->inode; + + wimlib_assert(((uintptr_t)p & 7) == 0); /* 8 byte aligned */ + orig_p = p; + + inode = dentry->d_inode; + disk_dentry = (struct wim_dentry_on_disk*)p; + + disk_dentry->attributes = cpu_to_le32(inode->i_attributes); + disk_dentry->security_id = cpu_to_le32(inode->i_security_id); + disk_dentry->subdir_offset = cpu_to_le64(dentry->subdir_offset); + disk_dentry->unused_1 = cpu_to_le64(dentry->d_unused_1); + disk_dentry->unused_2 = cpu_to_le64(dentry->d_unused_2); + disk_dentry->creation_time = cpu_to_le64(inode->i_creation_time); + disk_dentry->last_access_time = cpu_to_le64(inode->i_last_access_time); + disk_dentry->last_write_time = cpu_to_le64(inode->i_last_write_time); + + if (inode->i_resolved) + hash = inode->i_lte->hash; + else + hash = inode->i_hash; + copy_hash(disk_dentry->unnamed_stream_hash, inode_stream_hash(inode, 0)); + if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { + disk_dentry->reparse.rp_unknown_1 = cpu_to_le32(inode->i_rp_unknown_1); + disk_dentry->reparse.reparse_tag = cpu_to_le32(inode->i_reparse_tag); + disk_dentry->reparse.rp_unknown_2 = cpu_to_le16(inode->i_rp_unknown_2); + disk_dentry->reparse.not_rpfixed = cpu_to_le16(inode->i_not_rpfixed); + } else { + disk_dentry->nonreparse.rp_unknown_1 = cpu_to_le32(inode->i_rp_unknown_1); + disk_dentry->nonreparse.hard_link_group_id = + cpu_to_le64((inode->i_nlink == 1) ? 0 : inode->i_ino); + } + disk_dentry->num_alternate_data_streams = cpu_to_le16(inode->i_num_ads); + disk_dentry->short_name_nbytes = cpu_to_le16(dentry->short_name_nbytes); + disk_dentry->file_name_nbytes = cpu_to_le16(dentry->file_name_nbytes); + p += sizeof(struct wim_dentry_on_disk); + + if (dentry_has_long_name(dentry)) + p = mempcpy(p, dentry->file_name, dentry->file_name_nbytes + 2); + + if (dentry_has_short_name(dentry)) + p = mempcpy(p, dentry->short_name, dentry->short_name_nbytes + 2); + + /* Align to 8-byte boundary */ + while ((uintptr_t)p & 7) + *p++ = 0; /* We calculate the correct length of the dentry ourselves because the * dentry->length field may been set to an unexpected value from when we * read the dentry in (for example, there may have been unknown data * appended to the end of the dentry...) */ - u64 length = dentry_correct_length(dentry); - - p = put_u64(p, length); - p = put_u32(p, inode->attributes); - p = put_u32(p, inode->security_id); - p = put_u64(p, dentry->subdir_offset); - p = put_u64(p, 0); /* unused1 */ - p = put_u64(p, 0); /* unused2 */ - p = put_u64(p, inode->creation_time); - p = put_u64(p, inode->last_access_time); - p = put_u64(p, inode->last_write_time); - hash = inode_stream_hash(inode, 0); - p = put_bytes(p, SHA1_HASH_SIZE, hash); - if (inode->attributes & FILE_ATTRIBUTE_REPARSE_POINT) { - p = put_zeroes(p, 4); - p = put_u32(p, inode->reparse_tag); - p = put_zeroes(p, 4); - } else { - u64 link_group_id; - p = put_u32(p, 0); - if (dentry->inode->link_count == 1) - link_group_id = 0; - else - link_group_id = dentry->inode->ino; - p = put_u64(p, link_group_id); - } - p = put_u16(p, inode->num_ads); - p = put_u16(p, dentry->short_name_len); - p = put_u16(p, dentry->file_name_len); - if (dentry->file_name_len) { - p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name); - p = put_u16(p, 0); /* filename padding, 2 bytes. */ - } - if (dentry->short_name) { - p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name); - p = put_u16(p, 0); /* short name padding, 2 bytes */ - } + disk_dentry->length = cpu_to_le64(p - orig_p); + + /* Write the alternate data streams entries, if there are any. */ + for (u16 i = 0; i < inode->i_num_ads; i++) + { + const struct wim_ads_entry *ads_entry = + &inode->i_ads_entries[i]; + struct wim_ads_entry_on_disk *disk_ads_entry = + (struct wim_ads_entry_on_disk*)p; + + disk_ads_entry->reserved = cpu_to_le64(ads_entry->reserved); + orig_p = p; - /* Align to 8-byte boundary */ - wimlib_assert(length >= (p - orig_p) - && length - (p - orig_p) <= 7); - p = put_zeroes(p, length - (p - orig_p)); - - /* Write the alternate data streams, if there are any. Please see - * read_ads_entries() for comments about the format of the on-disk - * alternate data stream entries. */ - for (u16 i = 0; i < inode->num_ads; i++) { - p = put_u64(p, ads_entry_total_length(inode->ads_entries[i])); - p = put_u64(p, 0); /* Unused */ hash = inode_stream_hash(inode, i + 1); - p = put_bytes(p, SHA1_HASH_SIZE, hash); - p = put_u16(p, inode->ads_entries[i]->stream_name_len); - if (inode->ads_entries[i]->stream_name_len) { - p = put_bytes(p, inode->ads_entries[i]->stream_name_len, - (u8*)inode->ads_entries[i]->stream_name); - p = put_u16(p, 0); + copy_hash(disk_ads_entry->hash, hash); + disk_ads_entry->stream_name_nbytes = cpu_to_le16(ads_entry->stream_name_nbytes); + p += sizeof(struct wim_ads_entry_on_disk); + if (ads_entry->stream_name_nbytes) { + p = mempcpy(p, ads_entry->stream_name, + ads_entry->stream_name_nbytes + 2); } - p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8); + /* Align to 8-byte boundary */ + while ((uintptr_t)p & 7) + *p++ = 0; + disk_ads_entry->length = cpu_to_le64(p - orig_p); } -#ifdef ENABLE_ASSERTIONS - wimlib_assert(p - orig_p == __dentry_total_length(dentry, length)); -#endif return p; } +static int +write_dentry_cb(struct wim_dentry *dentry, void *_p) +{ + u8 **p = _p; + *p = write_dentry(dentry, *p); + return 0; +} + +static u8 * +write_dentry_tree_recursive(const struct wim_dentry *parent, u8 *p); + +static int +write_dentry_tree_recursive_cb(struct wim_dentry *dentry, void *_p) +{ + u8 **p = _p; + *p = write_dentry_tree_recursive(dentry, *p); + return 0; +} + /* Recursive function that writes a dentry tree rooted at @parent, not including * @parent itself, which has already been written. */ -static u8 *write_dentry_tree_recursive(const struct dentry *parent, u8 *p) +static u8 * +write_dentry_tree_recursive(const struct wim_dentry *parent, u8 *p) { - const struct dentry *child; - /* Nothing to do if this dentry has no children. */ if (parent->subdir_offset == 0) return p; - /* Write child dentries and end-of-directory entry. + /* Write child dentries and end-of-directory entry. * * Note: we need to write all of this dentry's children before * recursively writing the directory trees rooted at each of the child * dentries, since the on-disk dentries for a dentry's children are * always located at consecutive positions in the metadata resource! */ - child = parent->children; - if (child) { - do { - p = write_dentry(child, p); - child = child->next; - } while (child != parent->children); - } + for_dentry_child(parent, write_dentry_cb, &p); /* write end of directory entry */ - p = put_u64(p, 0); + *(le64*)p = cpu_to_le64(0); + p += 8; /* Recurse on children. */ - if (child) { - do { - p = write_dentry_tree_recursive(child, p); - child = child->next; - } while (child != parent->children); - } + for_dentry_child(parent, write_dentry_tree_recursive_cb, &p); return p; } @@ -1546,8 +1952,10 @@ static u8 *write_dentry_tree_recursive(const struct dentry *parent, u8 *p) * * Returns pointer to the byte after the last byte we wrote. */ -u8 *write_dentry_tree(const struct dentry *root, u8 *p) +u8 * +write_dentry_tree(const struct wim_dentry *root, u8 *p) { + DEBUG("Writing dentry tree."); wimlib_assert(dentry_is_root(root)); /* If we're the root dentry, we have no parent that already @@ -1556,104 +1964,9 @@ u8 *write_dentry_tree(const struct dentry *root, u8 *p) /* Write end of directory entry after the root dentry just to be safe; * however the root dentry obviously cannot have any siblings. */ - p = put_u64(p, 0); + *(le64*)p = cpu_to_le64(0); + p += 8; /* Recursively write the rest of the dentry tree. */ return write_dentry_tree_recursive(root, p); } - -/* Reads the children of a dentry, and all their children, ..., etc. from the - * metadata resource and into the dentry tree. - * - * @metadata_resource: An array that contains the uncompressed metadata - * resource for the WIM file. - * - * @metadata_resource_len: The length of the uncompressed metadata resource, in - * bytes. - * - * @dentry: A pointer to a `struct dentry' that is the root of the directory - * tree and has already been read from the metadata resource. It - * does not need to be the real root because this procedure is - * called recursively. - * - * @return: Zero on success, nonzero on failure. - */ -int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, - struct dentry *dentry) -{ - u64 cur_offset = dentry->subdir_offset; - struct dentry *prev_child = NULL; - struct dentry *first_child = NULL; - struct dentry *child; - struct dentry cur_child; - int ret; - - /* - * If @dentry has no child dentries, nothing more needs to be done for - * this branch. This is the case for regular files, symbolic links, and - * *possibly* empty directories (although an empty directory may also - * have one child dentry that is the special end-of-directory dentry) - */ - if (cur_offset == 0) - return 0; - - /* Find and read all the children of @dentry. */ - while (1) { - - /* Read next child of @dentry into @cur_child. */ - ret = read_dentry(metadata_resource, metadata_resource_len, - cur_offset, &cur_child); - if (ret != 0) - break; - - /* Check for end of directory. */ - if (cur_child.length == 0) - break; - - /* Not end of directory. Allocate this child permanently and - * link it to the parent and previous child. */ - child = MALLOC(sizeof(struct dentry)); - if (!child) { - ERROR("Failed to allocate %zu bytes for new dentry", - sizeof(struct dentry)); - ret = WIMLIB_ERR_NOMEM; - break; - } - memcpy(child, &cur_child, sizeof(struct dentry)); - - if (prev_child) { - prev_child->next = child; - child->prev = prev_child; - } else { - first_child = child; - } - - child->parent = dentry; - prev_child = child; - - /* If there are children of this child, call this procedure - * recursively. */ - if (child->subdir_offset != 0) { - ret = read_dentry_tree(metadata_resource, - metadata_resource_len, child); - if (ret != 0) - break; - } - - /* Advance to the offset of the next child. Note: We need to - * advance by the TOTAL length of the dentry, not by the length - * child->length, which although it does take into account the - * padding, it DOES NOT take into account alternate stream - * entries. */ - cur_offset += dentry_total_length(child); - } - - /* Link last child to first one, and set parent's children pointer to - * the first child. */ - if (prev_child) { - prev_child->next = first_child; - first_child->prev = prev_child; - } - dentry->children = first_child; - return ret; -}