*/
/*
- * Copyright (C) 2012 Eric Biggers
+ * Copyright (C) 2012, 2013, 2014 Eric Biggers
*
* This file is part of wimlib, a library for working with WIM files.
*
* wimlib; if not, see http://www.gnu.org/licenses/.
*/
-#include "buffer_io.h"
-#include "dentry.h"
-#include "lookup_table.h"
-#include "timestamp.h"
-#include "wimlib_internal.h"
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#include "wimlib.h"
+#include "wimlib/case.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/paths.h"
+#include "wimlib/resource.h"
+#include "wimlib/security.h"
+#include "wimlib/sha1.h"
+#include "wimlib/timestamp.h"
+
#include <errno.h>
+/* On-disk format of a WIM dentry (directory entry), located in the metadata
+ * resource for a WIM image. */
+struct wim_dentry_on_disk {
+
+ /* Length of this directory entry in bytes, not including any alternate
+ * data stream entries. Should be a multiple of 8 so that the following
+ * dentry or alternate data stream entry is aligned on an 8-byte
+ * boundary. (If not, wimlib will round it up.) It must be at least as
+ * long as the fixed-length fields of the dentry (WIM_DENTRY_DISK_SIZE),
+ * plus the lengths of the file name and/or short name if present.
+ *
+ * It is also possible for this field to be 0. This situation, which is
+ * undocumented, indicates the end of a list of sibling nodes in a
+ * directory. It also means the real length is 8, because the dentry
+ * included only the length field, but that takes up 8 bytes. */
+ le64 length;
+
+ /* Attributes of the file or directory. This is a bitwise OR of the
+ * FILE_ATTRIBUTE_* constants and should correspond to the value
+ * retrieved by GetFileAttributes() on Windows. */
+ le32 attributes;
+
+ /* A value that specifies the security descriptor for this file or
+ * directory. If -1, the file or directory has no security descriptor.
+ * Otherwise, it is a 0-based index into the WIM image's table of
+ * security descriptors (see: `struct wim_security_data') */
+ sle32 security_id;
+
+ /* Offset, in bytes, from the start of the uncompressed metadata
+ * resource of this directory's child directory entries, or 0 if this
+ * directory entry does not correspond to a directory or otherwise does
+ * not have any children. */
+ le64 subdir_offset;
+
+ /* Reserved fields */
+ le64 unused_1;
+ le64 unused_2;
+
+
+ /* Creation time, last access time, and last write time, in
+ * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601. They
+ * should correspond to the times gotten by calling GetFileTime() on
+ * Windows. */
+ le64 creation_time;
+ le64 last_access_time;
+ le64 last_write_time;
+
+ /* Vaguely, the SHA-1 message digest ("hash") of the file's contents.
+ * More specifically, this is for the "unnamed data stream" rather than
+ * any "alternate data streams". This hash value is used to look up the
+ * corresponding entry in the WIM's stream lookup table to actually find
+ * the file contents within the WIM.
+ *
+ * If the file has no unnamed data stream (e.g. is a directory), then
+ * this field will be all zeroes. If the unnamed data stream is empty
+ * (i.e. an "empty file"), then this field is also expected to be all
+ * zeroes. (It will be if wimlib created the WIM image, at least;
+ * otherwise it can't be ruled out that the SHA-1 message digest of 0
+ * bytes of data is given explicitly.)
+ *
+ * If the file has reparse data, then this field will instead specify
+ * the SHA-1 message digest of the reparse data. If it is somehow
+ * possible for a file to have both an unnamed data stream and reparse
+ * data, then this is not handled by wimlib.
+ *
+ * As a further special case, if this field is all zeroes but there is
+ * an alternate data stream entry with no name and a nonzero SHA-1
+ * message digest field, then that hash must be used instead of this
+ * one. In fact, when named data streams are present, some versions of
+ * Windows PE contain a bug where they only look in the alternate data
+ * stream entries for the unnamed data stream, not here.
+ */
+ u8 unnamed_stream_hash[SHA1_HASH_SIZE];
+
+ /* The format of the following data is not yet completely known and they
+ * do not correspond to Microsoft's documentation.
+ *
+ * If this directory entry is for a reparse point (has
+ * FILE_ATTRIBUTE_REPARSE_POINT set in the attributes field), then the
+ * version of the following fields containing the reparse tag is valid.
+ * Furthermore, the field notated as not_rpfixed, as far as I can tell,
+ * is supposed to be set to 1 if reparse point fixups (a.k.a. fixing the
+ * targets of absolute symbolic links) were *not* done, and otherwise 0.
+ *
+ * If this directory entry is not for a reparse point, then the version
+ * of the following fields containing the hard_link_group_id is valid.
+ * All MS says about this field is that "If this file is part of a hard
+ * link set, all the directory entries in the set will share the same
+ * value in this field.". However, more specifically I have observed
+ * the following:
+ * - If the file is part of a hard link set of size 1, then the
+ * hard_link_group_id should be set to either 0, which is treated
+ * specially as indicating "not hardlinked", or any unique value.
+ * - The specific nonzero values used to identity hard link sets do
+ * not matter, as long as they are unique.
+ * - However, due to bugs in Microsoft's software, it is actually NOT
+ * guaranteed that directory entries that share the same hard link
+ * group ID are actually hard linked to each either. We have to
+ * handle this by using special code to use distinguishing features
+ * (which is possible because some information about the underlying
+ * inode is repeated in each dentry) to split up these fake hard link
+ * groups into what they actually are supposed to be.
+ */
+ 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;
+ };
+
+ /* Number of alternate data stream entries that directly follow this
+ * dentry on-disk. */
+ le16 num_alternate_data_streams;
+
+ /* Length of this file's UTF-16LE encoded short name (8.3 DOS-compatible
+ * name), if present, in bytes, excluding the null terminator. If this
+ * file has no short name, then this field should be 0. */
+ le16 short_name_nbytes;
+
+ /* Length of this file's UTF-16LE encoded "long" name, excluding the
+ * null terminator. If this file has no short name, then this field
+ * should be 0. It's expected that only the root dentry has this field
+ * set to 0. */
+ le16 file_name_nbytes;
+
+ /* Followed by variable length file name, in UTF16-LE, if
+ * file_name_nbytes != 0. Includes null terminator. */
+ /*utf16lechar file_name[];*/
+
+ /* Followed by variable length short name, in UTF16-LE, if
+ * short_name_nbytes != 0. Includes null terminator. */
+ /*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_len,
- u16 short_name_len)
+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;
}
* 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)
+static u64
+dentry_correct_length_aligned(const struct wim_dentry *dentry)
{
- return __dentry_correct_length_unaligned(dentry->file_name_len,
- dentry->short_name_len);
-}
+ u64 len;
-/* Return the "correct" value to write in the length field of a WIM dentry,
- * based on the file name length and short name length. */
-static u64 dentry_correct_length(const struct wim_dentry *dentry)
-{
- return (dentry_correct_length_unaligned(dentry) + 7) & ~7;
+ len = dentry_correct_length_unaligned(dentry->file_name_nbytes,
+ dentry->short_name_nbytes);
+ return (len + 7) & ~7;
}
-/* Return %true iff the alternate data stream entry @entry has the UTF-8 stream
- * name @name that has length @name_len bytes. */
-static inline bool ads_entry_has_name(const struct ads_entry *entry,
- const char *name, size_t name_len)
+static int
+dentry_clear_short_name(struct wim_dentry *dentry)
{
- if (entry->stream_name_utf8_len != name_len)
- return false;
- return memcmp(entry->stream_name_utf8, name, name_len) == 0;
+ if (dentry_has_short_name(dentry)) {
+ FREE(dentry->short_name);
+ dentry->short_name = NULL;
+ dentry->short_name_nbytes = 0;
+ }
+ return 0;
}
-/* Duplicates a UTF-8 string into UTF-8 and UTF-16 strings and returns the
- * strings and their lengths in the pointer arguments. (Frees existing strings
- * first.) */
-static int get_names(char **name_utf16_ret, char **name_utf8_ret,
- u16 *name_utf16_len_ret, u16 *name_utf8_len_ret,
- const char *name)
+/* Sets the name of a WIM dentry from a multibyte string.
+ * Only use this on dentries not inserted into the tree. Use rename_wim_path()
+ * to do a real rename. */
+int
+dentry_set_name(struct wim_dentry *dentry, const tchar *new_name)
{
- size_t utf8_len;
- size_t utf16_len;
- char *name_utf16, *name_utf8;
int ret;
- utf8_len = strlen(name);
- ret = utf8_to_utf16(name, utf8_len, &name_utf16, &utf16_len);
- if (ret != 0)
+ ret = get_utf16le_string(new_name, &dentry->file_name,
+ &dentry->file_name_nbytes);
+ if (ret)
return ret;
- name_utf8 = MALLOC(utf8_len + 1);
- if (!name_utf8) {
- FREE(name_utf16);
- 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;
+ return dentry_clear_short_name(dentry);
}
-/* Sets the name of a WIM dentry. */
-int set_dentry_name(struct wim_dentry *dentry, const char *new_name)
+/* Sets the name of a WIM dentry from a UTF-16LE string.
+ * Only use this on dentries not inserted into the tree. Use rename_wim_path()
+ * to do a real rename. */
+int
+dentry_set_name_utf16le(struct wim_dentry *dentry, const utf16lechar *new_name)
{
- int ret;
+ utf16lechar *name = NULL;
+ size_t name_nbytes = 0;
- ret = get_names(&dentry->file_name, &dentry->file_name_utf8,
- &dentry->file_name_len, &dentry->file_name_utf8_len,
- new_name);
- if (ret == 0) {
- if (dentry->short_name_len) {
- FREE(dentry->short_name);
- dentry->short_name_len = 0;
- }
- dentry->length = dentry_correct_length(dentry);
+ if (new_name && *new_name) {
+ const utf16lechar *tmp;
+
+ tmp = new_name;
+ do {
+ name_nbytes += sizeof(utf16lechar);
+ } while (*++tmp);
+
+ name = memdup(new_name, name_nbytes + sizeof(utf16lechar));
+ if (!name)
+ return WIMLIB_ERR_NOMEM;
}
- return ret;
-}
-/*
- * Changes the name of an alternate data stream */
-static 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);
+ FREE(dentry->file_name);
+ dentry->file_name = name;
+ dentry->file_name_nbytes = name_nbytes;
+
+ return dentry_clear_short_name(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 ads_entry *entry)
+static u64
+ads_entry_total_length(const struct wim_ads_entry *entry)
{
- u64 len = WIM_ADS_ENTRY_DISK_SIZE;
- if (entry->stream_name_len)
- len += entry->stream_name_len + 2;
+ u64 len = sizeof(struct wim_ads_entry_on_disk);
+ if (entry->stream_name_nbytes)
+ len += entry->stream_name_nbytes + 2;
return (len + 7) & ~7;
}
+/*
+ * Determine whether to include a "dummy" stream when writing a WIM dentry:
+ *
+ * Some versions of Microsoft's WIM software (the boot driver(s) in WinPE 3.0,
+ * for example) contain a bug where they assume the first alternate data stream
+ * (ADS) entry of a dentry with a nonzero ADS count specifies the unnamed
+ * stream, even if it has a name and the unnamed stream is already specified in
+ * the hash field of the dentry itself.
+ *
+ * wimlib has to work around this behavior by carefully emulating the behavior
+ * of (most versions of) ImageX/WIMGAPI, which move the unnamed stream reference
+ * into the alternate stream entries whenever there are named data streams, even
+ * though there is already a field in the dentry itself for the unnamed stream
+ * reference, which then goes to waste.
+ */
+static inline bool
+inode_needs_dummy_stream(const struct wim_inode *inode)
+{
+ return (inode->i_num_ads > 0 &&
+ inode->i_num_ads < 0xffff && /* overflow check */
+ inode->i_canonical_streams); /* assume the dentry is okay if it
+ already had an unnamed ADS entry
+ when it was read in */
+}
-static u64 __dentry_total_length(const struct wim_dentry *dentry, u64 length)
+/* Calculate the total number of bytes that will be consumed when a WIM dentry
+ * is written. This includes base dentry and name fields as well as all
+ * alternate data stream entries and alignment bytes. */
+u64
+dentry_out_total_length(const struct wim_dentry *dentry)
{
+ u64 length = dentry_correct_length_aligned(dentry);
const struct wim_inode *inode = dentry->d_inode;
+
+ if (inode_needs_dummy_stream(inode))
+ length += ads_entry_total_length(&(struct wim_ads_entry){});
+
for (u16 i = 0; i < inode->i_num_ads; i++)
length += ads_entry_total_length(&inode->i_ads_entries[i]);
- return (length + 7) & ~7;
-}
-/* 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)
-{
- return __dentry_total_length(dentry,
- dentry_correct_length_unaligned(dentry));
+ return length;
}
-/* 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)
+/* Calculate the aligned, total length of a dentry, including all alternate data
+ * stream entries. Uses dentry->length. */
+static u64
+dentry_in_total_length(const struct wim_dentry *dentry)
{
- return __dentry_total_length(dentry, dentry->length);
+ u64 length = dentry->length;
+ 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;
}
-int for_dentry_in_rbtree(struct rb_node *root,
- int (*visitor)(struct wim_dentry *, void *),
- void *arg)
+static int
+do_for_dentry_in_tree(struct wim_dentry *dentry,
+ int (*visitor)(struct wim_dentry *, void *), void *arg)
{
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;
- }
- }
-}
+ struct wim_dentry *child;
-static int for_dentry_tree_in_rbtree_depth(struct rb_node *node,
- int (*visitor)(struct wim_dentry*, void*),
- void *arg)
-{
- 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)
+ ret = (*visitor)(dentry, arg);
+ if (unlikely(ret))
+ return ret;
+
+ for_dentry_child(child, dentry) {
+ ret = do_for_dentry_in_tree(child, visitor, arg);
+ if (unlikely(ret))
return ret;
}
return 0;
}
-/*#define RECURSIVE_FOR_DENTRY_IN_TREE*/
-#ifdef RECURSIVE_FOR_DENTRY_IN_TREE
-static int for_dentry_tree_in_rbtree(struct rb_node *node,
- int (*visitor)(struct wim_dentry*, void*),
- void *arg)
+static int
+do_for_dentry_in_tree_depth(struct wim_dentry *dentry,
+ int (*visitor)(struct wim_dentry *, void *), void *arg)
{
int ret;
- if (node) {
- ret = for_dentry_tree_in_rbtree(node->rb_left, visitor, arg);
- if (ret != 0)
- return ret;
- ret = for_dentry_in_tree(rbnode_dentry(node), visitor, arg);
- if (ret != 0)
- return ret;
- ret = for_dentry_tree_in_rbtree(node->rb_right, visitor, arg);
- if (ret != 0)
+ struct wim_dentry *child;
+
+ for_dentry_child_postorder(child, dentry) {
+ ret = do_for_dentry_in_tree_depth(child, visitor, arg);
+ if (unlikely(ret))
return ret;
}
- return 0;
+ return unlikely((*visitor)(dentry, arg));
}
-#endif
-/*
- * Calls a function on all directory entries in a WIM dentry tree. Logically,
+/* 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.
- *
- * In reality, the data structures are more complicated than the above might
- * suggest because there is a separate red-black tree for each dentry that
- * contains its direct children.
- */
-int for_dentry_in_tree(struct wim_dentry *root,
- int (*visitor)(struct wim_dentry*, void*), void *arg)
+ * */
+int
+for_dentry_in_tree(struct wim_dentry *root,
+ int (*visitor)(struct wim_dentry *, void *), void *arg)
{
-#ifdef RECURSIVE_FOR_DENTRY_IN_TREE
- int ret = visitor(root, arg);
- if (ret != 0)
- return ret;
- return for_dentry_tree_in_rbtree(root->d_inode->i_children.rb_node, visitor, arg);
-#else
- int ret;
- struct list_head main_stack;
- struct list_head sibling_stack;
- struct list_head *sibling_stack_bottom;
- struct wim_dentry *main_dentry;
- struct rb_node *node;
- struct list_head *next_sibling;
- struct wim_dentry *dentry;
-
- ret = visitor(root, arg);
- if (ret != 0)
- return ret;
-
- main_dentry = root;
- sibling_stack_bottom = &sibling_stack;
- INIT_LIST_HEAD(&main_stack);
- INIT_LIST_HEAD(&sibling_stack);
-
- list_add(&root->tmp_list, &main_stack);
- node = root->d_inode->i_children.rb_node;
-
- while (1) {
- // Prepare for non-recursive in-order traversal of the red-black
- // tree of this dentry's children
-
- while (node) {
- // Push this node to the sibling stack and examine the
- // left neighbor, if any
- list_add(&rbnode_dentry(node)->tmp_list, &sibling_stack);
- node = node->rb_left;
- }
-
- next_sibling = sibling_stack.next;
- if (next_sibling == sibling_stack_bottom) {
- // Done with all siblings. Pop the main dentry to move
- // back up one level.
- main_dentry = container_of(main_stack.next,
- struct wim_dentry,
- tmp_list);
- list_del(&main_dentry->tmp_list);
-
- if (main_dentry == root)
- goto out;
-
- // Restore sibling stack bottom from the previous level
- sibling_stack_bottom = (void*)main_dentry->parent;
-
- // Restore the just-popped main dentry's parent
- main_dentry->parent = container_of(main_stack.next,
- struct wim_dentry,
- tmp_list);
-
- // The next sibling to traverse in the previous level,
- // in the in-order traversal of the red-black tree, is
- // the one to the right.
- node = main_dentry->rb_node.rb_right;
- } else {
- // The sibling stack is not empty, so there are more to
- // go!
-
- // Pop a sibling from the stack.
- list_del(next_sibling);
- dentry = container_of(next_sibling, struct wim_dentry, tmp_list);
-
- // Visit the sibling.
- ret = visitor(dentry, arg);
- if (ret != 0) {
- // Failed. Restore parent pointers for the
- // dentries in the main stack
- list_for_each_entry(dentry, &main_stack, tmp_list) {
- dentry->parent = container_of(dentry->tmp_list.next,
- struct wim_dentry,
- tmp_list);
- }
- goto out;
- }
-
- // We'd like to recursively visit the dentry tree rooted
- // at this sibling. To do this, add it to the main
- // stack, save the bottom of this level's sibling stack
- // in the dentry->parent field, re-set the bottom of the
- // sibling stack to be its current height, and set
- // main_dentry to the sibling so it becomes the parent
- // dentry in the next iteration through the outer loop.
- if (inode_has_children(dentry->d_inode)) {
- list_add(&dentry->tmp_list, &main_stack);
- dentry->parent = (void*)sibling_stack_bottom;
- sibling_stack_bottom = sibling_stack.next;
-
- main_dentry = dentry;
- node = main_dentry->d_inode->i_children.rb_node;
- } else {
- node = dentry->rb_node.rb_right;
- }
- }
- }
-out:
- root->parent = root;
- return ret;
-#endif
+ if (unlikely(!root))
+ return 0;
+ return do_for_dentry_in_tree(root, visitor, 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 wim_dentry *root,
- int (*visitor)(struct wim_dentry*, void*), void *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 wim_dentry *root,
+ int (*visitor)(struct wim_dentry *, void *), void *arg)
{
-#if 1
- int ret;
- ret = for_dentry_tree_in_rbtree_depth(root->d_inode->i_children.rb_node,
- visitor, arg);
- if (ret != 0)
- return ret;
- return visitor(root, arg);
+ if (unlikely(!root))
+ return 0;
+ return do_for_dentry_in_tree_depth(root, visitor, arg);
+}
-#else
+/* Calculate the full path of @dentry. */
+int
+calculate_dentry_full_path(struct wim_dentry *dentry)
+{
+ tchar *full_path;
+ u32 full_path_nbytes;
int ret;
- struct list_head main_stack;
- struct list_head sibling_stack;
- struct list_head *sibling_stack_bottom;
- struct wim_dentry *main_dentry;
- struct rb_node *node;
- struct list_head *next_sibling;
- struct wim_dentry *dentry;
- main_dentry = root;
- sibling_stack_bottom = &sibling_stack;
- INIT_LIST_HEAD(&main_stack);
- INIT_LIST_HEAD(&sibling_stack);
-
- list_add(&main_dentry->tmp_list, &main_stack);
+ if (dentry->_full_path)
+ return 0;
- while (1) {
- node = main_dentry->d_inode->i_children.rb_node;
+ if (dentry_is_root(dentry)) {
+ static const tchar _root_path[] = {WIM_PATH_SEPARATOR, T('\0')};
+ full_path = TSTRDUP(_root_path);
+ if (full_path == NULL)
+ return WIMLIB_ERR_NOMEM;
+ full_path_nbytes = 1 * sizeof(tchar);
+ } else {
+ struct wim_dentry *parent;
+ tchar *parent_full_path;
+ u32 parent_full_path_nbytes;
+ size_t filename_nbytes;
- while (1) {
- if (node->rb_left) {
- list_add(&rbnode_dentry(node)->tmp_list, &sibling_stack);
- node = node->rb_left;
- continue;
- }
- if (node->rb_right) {
- list_add(&rbnode_dentry(node)->tmp_list, &sibling_stack);
- node = node->rb_right;
- continue;
+ parent = dentry->parent;
+ if (dentry_is_root(parent)) {
+ parent_full_path = T("");
+ parent_full_path_nbytes = 0;
+ } else {
+ if (parent->_full_path == NULL) {
+ ret = calculate_dentry_full_path(parent);
+ if (ret)
+ return ret;
}
- list_add(&rbnode_dentry(node)->tmp_list, &sibling_stack);
+ parent_full_path = parent->_full_path;
+ parent_full_path_nbytes = parent->full_path_nbytes;
}
- pop_sibling:
- next_sibling = sibling_stack.next;
- if (next_sibling == sibling_stack_bottom) {
- main_dentry = container_of(main_stack.next,
- struct wim_dentry,
- tmp_list);
- list_del(&main_dentry->tmp_list);
+ /* 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_nbytes = parent_full_path_nbytes + sizeof(tchar) +
+ filename_nbytes;
+ full_path = MALLOC(full_path_nbytes + sizeof(tchar));
+ if (full_path == NULL)
+ return WIMLIB_ERR_NOMEM;
+ memcpy(full_path, parent_full_path, parent_full_path_nbytes);
+ full_path[parent_full_path_nbytes / sizeof(tchar)] = WIM_PATH_SEPARATOR;
+ #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
+ }
+ dentry->_full_path = full_path;
+ dentry->full_path_nbytes= full_path_nbytes;
+ return 0;
+}
- sibling_stack_bottom = (void*)main_dentry->parent;
+static int
+do_calculate_dentry_full_path(struct wim_dentry *dentry, void *_ignore)
+{
+ return calculate_dentry_full_path(dentry);
+}
- if (main_dentry == root) {
- main_dentry->parent = main_dentry;
- ret = visitor(dentry, arg);
- return ret;
- } else {
- main_dentry->parent = container_of(main_stack.next,
- struct wim_dentry,
- tmp_list);
- }
+int
+calculate_dentry_tree_full_paths(struct wim_dentry *root)
+{
+ return for_dentry_in_tree(root, do_calculate_dentry_full_path, NULL);
+}
- ret = visitor(main_dentry, arg);
+tchar *
+dentry_full_path(struct wim_dentry *dentry)
+{
+ calculate_dentry_full_path(dentry);
+ return dentry->_full_path;
+}
- if (ret != 0) {
- list_del(&root->tmp_list);
- list_for_each_entry(dentry, &main_stack, tmp_list) {
- dentry->parent = container_of(dentry->tmp_list.next,
- struct wim_dentry,
- tmp_list);
- }
- root->parent = root;
- return ret;
- }
- goto pop_sibling;
- } else {
+static int
+dentry_calculate_subdir_offset(struct wim_dentry *dentry, void *_subdir_offset_p)
+{
- list_del(next_sibling);
- dentry = container_of(next_sibling, struct wim_dentry, tmp_list);
+ if (dentry_is_directory(dentry)) {
+ u64 *subdir_offset_p = _subdir_offset_p;
+ struct wim_dentry *child;
+ /* Set offset of directory's child dentries */
+ dentry->subdir_offset = *subdir_offset_p;
- list_add(&dentry->tmp_list, &main_stack);
- dentry->parent = (void*)sibling_stack_bottom;
- sibling_stack_bottom = sibling_stack.next;
+ /* Account for child dentries */
+ for_dentry_child(child, dentry)
+ *subdir_offset_p += dentry_out_total_length(child);
- main_dentry = dentry;
- }
+ /* Account for end-of-directory entry */
+ *subdir_offset_p += 8;
+ } else {
+ /* Not a directory; set subdir_offset to 0 */
+ dentry->subdir_offset = 0;
}
-#endif
+ return 0;
}
/*
- * Calculate the full path of @dentry, based on its parent's full path and on
- * its UTF-8 file name.
+ * Calculates the subdir offsets for a directory tree.
*/
-int calculate_dentry_full_path(struct wim_dentry *dentry, void *ignore)
+void
+calculate_subdir_offsets(struct wim_dentry *root, u64 *subdir_offset_p)
{
- char *full_path;
- u32 full_path_len;
- if (dentry_is_root(dentry)) {
- full_path = MALLOC(2);
- if (!full_path)
- goto oom;
- full_path[0] = '/';
- full_path[1] = '\0';
- full_path_len = 1;
- } else {
- char *parent_full_path;
- u32 parent_full_path_len;
- const struct wim_dentry *parent = dentry->parent;
-
- if (dentry_is_root(parent)) {
- parent_full_path = "";
- parent_full_path_len = 0;
- } else {
- parent_full_path = parent->full_path_utf8;
- parent_full_path_len = parent->full_path_utf8_len;
- }
-
- full_path_len = parent_full_path_len + 1 +
- dentry->file_name_utf8_len;
- full_path = MALLOC(full_path_len + 1);
- 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';
- }
- FREE(dentry->full_path_utf8);
- dentry->full_path_utf8 = full_path;
- dentry->full_path_utf8_len = full_path_len;
- return 0;
-oom:
- ERROR("Out of memory while calculating dentry full path");
- return WIMLIB_ERR_NOMEM;
+ for_dentry_in_tree(root, dentry_calculate_subdir_offset, subdir_offset_p);
}
-static int increment_subdir_offset(struct wim_dentry *dentry, void *subdir_offset_p)
+/* Compare the UTF-16LE long filenames of two dentries case insensitively. */
+static int
+dentry_compare_names_case_insensitive(const struct wim_dentry *d1,
+ const struct wim_dentry *d2)
{
- *(u64*)subdir_offset_p += dentry_correct_total_length(dentry);
- return 0;
+ return cmp_utf16le_strings(d1->file_name,
+ d1->file_name_nbytes / 2,
+ d2->file_name,
+ d2->file_name_nbytes / 2,
+ true);
}
-static int call_calculate_subdir_offsets(struct wim_dentry *dentry,
- void *subdir_offset_p)
+/* Compare the UTF-16LE long filenames of two dentries case sensitively. */
+static int
+dentry_compare_names_case_sensitive(const struct wim_dentry *d1,
+ const struct wim_dentry *d2)
{
- calculate_subdir_offsets(dentry, subdir_offset_p);
- return 0;
+ return cmp_utf16le_strings(d1->file_name,
+ d1->file_name_nbytes / 2,
+ d2->file_name,
+ d2->file_name_nbytes / 2,
+ false);
}
-/*
- * 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.
- */
-void calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p)
+static int
+_avl_dentry_compare_names_ci(const struct avl_tree_node *n1,
+ const struct avl_tree_node *n2)
{
- struct rb_node *node;
+ const struct wim_dentry *d1, *d2;
- dentry->subdir_offset = *subdir_offset_p;
- 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. */
- 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. */
- 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
- * files or reparse points have a subdir_offset of 0. */
- if (dentry_is_directory(dentry))
- *subdir_offset_p += 8;
- else
- dentry->subdir_offset = 0;
- }
+ d1 = avl_tree_entry(n1, struct wim_dentry, d_index_node_ci);
+ d2 = avl_tree_entry(n2, struct wim_dentry, d_index_node_ci);
+ return dentry_compare_names_case_insensitive(d1, d2);
}
-static int compare_names(const char *name_1, u16 len_1,
- const char *name_2, u16 len_2)
+static int
+_avl_dentry_compare_names(const struct avl_tree_node *n1,
+ const struct avl_tree_node *n2)
{
- int result = strncasecmp(name_1, name_2, min(len_1, len_2));
- if (result) {
- return result;
- } else {
- return (int)len_1 - (int)len_2;
- }
-}
+ const struct wim_dentry *d1, *d2;
-static int dentry_compare_names(const struct wim_dentry *d1, const struct wim_dentry *d2)
-{
- return compare_names(d1->file_name_utf8, d1->file_name_utf8_len,
- d2->file_name_utf8, d2->file_name_utf8_len);
+ d1 = avl_tree_entry(n1, struct wim_dentry, d_index_node);
+ d2 = avl_tree_entry(n2, struct wim_dentry, d_index_node);
+ return dentry_compare_names_case_sensitive(d1, d2);
}
+/* Default case sensitivity behavior for searches with
+ * WIMLIB_CASE_PLATFORM_DEFAULT specified. This can be modified by
+ * wimlib_global_init(). */
+bool default_ignore_case =
+#ifdef __WIN32__
+ true
+#else
+ false
+#endif
+;
+/* Case-sensitive dentry lookup. Only @file_name and @file_name_nbytes of
+ * @dummy must be valid. */
static struct wim_dentry *
-get_rbtree_child_with_name(const struct rb_node *node,
- const char *name, size_t name_len)
+dir_lookup(const struct wim_inode *dir, const struct wim_dentry *dummy)
{
- do {
- struct wim_dentry *child = rbnode_dentry(node);
- int result = compare_names(name, name_len,
- child->file_name_utf8,
- child->file_name_utf8_len);
- if (result < 0)
- node = node->rb_left;
- else if (result > 0)
- node = node->rb_right;
- else
- return child;
- } while (node);
- return NULL;
+ struct avl_tree_node *node;
+
+ node = avl_tree_lookup_node(dir->i_children,
+ &dummy->d_index_node,
+ _avl_dentry_compare_names);
+ if (!node)
+ return NULL;
+ return avl_tree_entry(node, struct wim_dentry, d_index_node);
}
-/* 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 char *name)
+/* Case-insensitive dentry lookup. Only @file_name and @file_name_nbytes of
+ * @dummy must be valid. */
+static struct wim_dentry *
+dir_lookup_ci(const struct wim_inode *dir, const struct wim_dentry *dummy)
{
- struct rb_node *node = dentry->d_inode->i_children.rb_node;
- if (node)
- return get_rbtree_child_with_name(node, name, strlen(name));
- else
+ struct avl_tree_node *node;
+
+ node = avl_tree_lookup_node(dir->i_children_ci,
+ &dummy->d_index_node_ci,
+ _avl_dentry_compare_names_ci);
+ if (!node)
return NULL;
+ return avl_tree_entry(node, struct wim_dentry, d_index_node_ci);
}
-/* Retrieves the dentry that has the UTF-8 @path relative to the dentry
- * @cur_dentry. Returns NULL if no dentry having the path is found. */
-static struct wim_dentry *get_dentry_relative_path(struct wim_dentry *cur_dentry,
- const char *path)
+/* Given a UTF-16LE filename and a directory, look up the dentry for the file.
+ * Return it if found, otherwise NULL. This is case-sensitive on UNIX and
+ * case-insensitive on Windows. */
+struct wim_dentry *
+get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry,
+ const utf16lechar *name,
+ size_t name_nbytes,
+ CASE_SENSITIVITY_TYPE case_ctype)
{
- if (*path == '\0')
- return cur_dentry;
+ const struct wim_inode *dir = dentry->d_inode;
+ bool ignore_case = will_ignore_case(case_ctype);
+ struct wim_dentry dummy;
+ struct wim_dentry *child;
- struct rb_node *node = cur_dentry->d_inode->i_children.rb_node;
- if (node) {
- struct wim_dentry *child;
- size_t base_len;
- const char *new_path;
+ dummy.file_name = (utf16lechar*)name;
+ dummy.file_name_nbytes = name_nbytes;
+
+ if (!ignore_case)
+ /* Case-sensitive lookup. */
+ return dir_lookup(dir, &dummy);
+
+ /* Case-insensitive lookup. */
+
+ child = dir_lookup_ci(dir, &dummy);
+ if (!child)
+ return NULL;
+
+ if (likely(list_empty(&child->d_ci_conflict_list)))
+ /* Only one dentry has this case-insensitive name; return it */
+ return child;
+
+ /* Multiple dentries have the same case-insensitive name. Choose the
+ * dentry with the same case-sensitive name, if one exists; otherwise
+ * print a warning and choose one of the possible dentries arbitrarily.
+ */
+ struct wim_dentry *alt = child;
+ size_t num_alts = 0;
- new_path = path_next_part(path, &base_len);
+ do {
+ num_alts++;
+ if (!dentry_compare_names_case_sensitive(&dummy, alt))
+ return alt;
+ alt = list_entry(alt->d_ci_conflict_list.next,
+ struct wim_dentry, d_ci_conflict_list);
+ } while (alt != child);
+
+ WARNING("Result of case-insensitive lookup is ambiguous\n"
+ " (returning \"%"TS"\" of %zu "
+ "possible files, including \"%"TS"\")",
+ dentry_full_path(child),
+ num_alts,
+ dentry_full_path(list_entry(child->d_ci_conflict_list.next,
+ struct wim_dentry,
+ d_ci_conflict_list)));
+ return child;
+}
- child = get_rbtree_child_with_name(node, path, base_len);
- if (child)
- return get_dentry_relative_path(child, new_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,
+ CASE_SENSITIVITY_TYPE case_type)
+{
+#if TCHAR_IS_UTF16LE
+ return get_dentry_child_with_utf16le_name(dentry, name,
+ tstrlen(name) * sizeof(tchar),
+ case_type);
+#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,
+ case_type);
+ FREE(utf16le_name);
}
- /* errno is set to ENOTDIR if the lookup failed due to reaching a
- * non-directory, or ENOENT if the lookup failed otherwise. This maybe
- * should be factored out somehow. */
- if (dentry_is_directory(cur_dentry))
- errno = ENOENT;
- else
- errno = ENOTDIR;
- return NULL;
+ return child;
+#endif
}
-/* Returns the dentry corresponding to the UTF-8 @path, or NULL if there is no
- * such dentry. */
-struct wim_dentry *get_dentry(WIMStruct *w, const char *path)
+static struct wim_dentry *
+get_dentry_utf16le(WIMStruct *wim, const utf16lechar *path,
+ CASE_SENSITIVITY_TYPE case_type)
{
- struct wim_dentry *root = wim_root_dentry(w);
- while (*path == '/')
- path++;
- return get_dentry_relative_path(root, path);
+ struct wim_dentry *cur_dentry;
+ const utf16lechar *name_start, *name_end;
+
+ /* Start with the root directory of the image. Note: this will be NULL
+ * if an image has been added directly with wimlib_add_empty_image() but
+ * no files have been added yet; in that case we fail with ENOENT. */
+ cur_dentry = wim_root_dentry(wim);
+
+ name_start = path;
+ for (;;) {
+ if (cur_dentry == NULL) {
+ errno = ENOENT;
+ return NULL;
+ }
+
+ if (*name_start && !dentry_is_directory(cur_dentry)) {
+ errno = ENOTDIR;
+ return NULL;
+ }
+
+ while (*name_start == cpu_to_le16(WIM_PATH_SEPARATOR))
+ name_start++;
+
+ if (!*name_start)
+ return cur_dentry;
+
+ name_end = name_start;
+ do {
+ ++name_end;
+ } while (*name_end != cpu_to_le16(WIM_PATH_SEPARATOR) && *name_end);
+
+ cur_dentry = get_dentry_child_with_utf16le_name(cur_dentry,
+ name_start,
+ (u8*)name_end - (u8*)name_start,
+ case_type);
+ name_start = name_end;
+ }
}
-struct wim_inode *wim_pathname_to_inode(WIMStruct *w, const char *path)
+/*
+ * WIM path lookup: translate a path in the currently selected WIM image to the
+ * corresponding dentry, if it exists.
+ *
+ * @wim
+ * The WIMStruct for the WIM. The search takes place in the currently
+ * selected image.
+ *
+ * @path
+ * The path to look up, given relative to the root of the WIM image.
+ * Characters with value WIM_PATH_SEPARATOR are taken to be path
+ * separators. Leading path separators are ignored, whereas one or more
+ * trailing path separators cause the path to only match a directory.
+ *
+ * @case_type
+ * The case-sensitivity behavior of this function, as one of the following
+ * constants:
+ *
+ * - WIMLIB_CASE_SENSITIVE: Perform the search case sensitively. This means
+ * that names must match exactly.
+ *
+ * - WIMLIB_CASE_INSENSITIVE: Perform the search case insensitively. This
+ * means that names are considered to match if they are equal when
+ * transformed to upper case. If a path component matches multiple names
+ * case-insensitively, the name that matches the path component
+ * case-sensitively is chosen, if existent; otherwise one
+ * case-insensitively matching name is chosen arbitrarily.
+ *
+ * - WIMLIB_CASE_PLATFORM_DEFAULT: Perform either case-sensitive or
+ * case-insensitive search, depending on the value of the global variable
+ * default_ignore_case.
+ *
+ * In any case, no Unicode normalization is done before comparing strings.
+ *
+ * Returns a pointer to the dentry that is the result of the lookup, or NULL if
+ * no such dentry exists. If NULL is returned, errno is set to one of the
+ * following values:
+ *
+ * ENOTDIR if one of the path components used as a directory existed but
+ * was not, in fact, a directory.
+ *
+ * ENOENT otherwise.
+ *
+ * Additional notes:
+ *
+ * - This function does not consider a reparse point to be a directory, even
+ * if it has FILE_ATTRIBUTE_DIRECTORY set.
+ *
+ * - This function does not dereference symbolic links or junction points
+ * when performing the search.
+ *
+ * - Since this function ignores leading slashes, the empty path is valid and
+ * names the root directory of the WIM image.
+ *
+ * - An image added with wimlib_add_empty_image() does not have a root
+ * directory yet, and this function will fail with ENOENT for any path on
+ * such an image.
+ */
+struct wim_dentry *
+get_dentry(WIMStruct *wim, const tchar *path, CASE_SENSITIVITY_TYPE case_type)
{
+#if TCHAR_IS_UTF16LE
+ return get_dentry_utf16le(wim, path, case_type);
+#else
+ utf16lechar *path_utf16le;
+ size_t path_utf16le_nbytes;
+ int ret;
struct wim_dentry *dentry;
- dentry = get_dentry(w, path);
- if (dentry)
- return dentry->d_inode;
- else
+
+ 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, case_type);
+ FREE(path_utf16le);
+ return dentry;
+#endif
}
-/* Returns the dentry that corresponds to the parent directory of @path, or NULL
- * if the dentry is not found. */
-struct wim_dentry *get_parent_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)
{
- size_t path_len = strlen(path);
- char buf[path_len + 1];
-
- memcpy(buf, path, path_len + 1);
-
- to_parent_name(buf, path_len);
-
- return get_dentry(w, buf);
+ ssize_t i = (ssize_t)len - 1;
+ while (i >= 0 && buf[i] == WIM_PATH_SEPARATOR)
+ i--;
+ while (i >= 0 && buf[i] != WIM_PATH_SEPARATOR)
+ i--;
+ while (i >= 0 && buf[i] == WIM_PATH_SEPARATOR)
+ i--;
+ buf[i + 1] = T('\0');
}
-/* Prints the full path of a dentry. */
-int print_dentry_full_path(struct wim_dentry *dentry, void *ignore)
+/* Similar to get_dentry(), but returns the dentry named by @path with the last
+ * component stripped off.
+ *
+ * Note: The returned dentry is NOT guaranteed to be a directory. */
+struct wim_dentry *
+get_parent_dentry(WIMStruct *wim, const tchar *path,
+ CASE_SENSITIVITY_TYPE case_type)
{
- if (dentry->full_path_utf8)
- puts(dentry->full_path_utf8);
- return 0;
+ 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(wim, buf, case_type);
}
-/* We want to be able to show the names of the file attribute flags that are
- * set. */
-struct file_attr_flag {
- u32 flag;
- const char *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"},
-};
-
-/* 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 wim_dentry *dentry, void *lookup_table)
+#ifdef WITH_FUSE
+/* Finds the dentry, lookup table entry, and stream index for a WIM file stream,
+ * given a path name.
+ *
+ * Currently, lookups of this type are only needed if FUSE is enabled. */
+int
+wim_pathname_to_stream(WIMStruct *wim,
+ const tchar *path,
+ int lookup_flags,
+ struct wim_dentry **dentry_ret,
+ struct wim_lookup_table_entry **lte_ret,
+ u16 *stream_idx_ret)
{
- const u8 *hash;
+ struct wim_dentry *dentry;
struct wim_lookup_table_entry *lte;
- const struct wim_inode *inode = dentry->d_inode;
- char buf[50];
-
- printf("[DENTRY]\n");
- printf("Length = %"PRIu64"\n", dentry->length);
- printf("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)
- printf(" FILE_ATTRIBUTE_%s is set\n",
- file_attr_flags[i].name);
- printf("Security ID = %d\n", inode->i_security_id);
- printf("Subdir offset = %"PRIu64"\n", dentry->subdir_offset);
-
- wim_timestamp_to_str(inode->i_creation_time, buf, sizeof(buf));
- printf("Creation Time = %s\n", buf);
-
- wim_timestamp_to_str(inode->i_last_access_time, buf, sizeof(buf));
- printf("Last Access Time = %s\n", buf);
-
- wim_timestamp_to_str(inode->i_last_write_time, buf, sizeof(buf));
- printf("Last Write Time = %s\n", buf);
-
- printf("Reparse Tag = 0x%"PRIx32"\n", inode->i_reparse_tag);
- printf("Hard Link Group = 0x%"PRIx64"\n", inode->i_ino);
- printf("Hard Link Group Size = %"PRIu32"\n", inode->i_nlink);
- printf("Number of Alternate Data Streams = %hu\n", inode->i_num_ads);
- 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 (UTF-16LE) = \"");
- 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->d_inode, 0, lookup_table);
- if (lte) {
- print_lookup_table_entry(lte);
- } else {
- hash = inode_stream_hash(inode, 0);
- if (hash) {
- printf("Hash = 0x");
- print_hash(hash);
- putchar('\n');
- putchar('\n');
+ u16 stream_idx;
+ const tchar *stream_name = NULL;
+ struct wim_inode *inode;
+ tchar *p = NULL;
+
+ if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
+ stream_name = path_stream_name(path);
+ if (stream_name) {
+ p = (tchar*)stream_name - 1;
+ *p = T('\0');
}
}
- for (u16 i = 0; i < inode->i_num_ads; i++) {
- printf("[Alternate Stream Entry %u]\n", i);
- printf("Name = \"%s\"\n", inode->i_ads_entries[i].stream_name_utf8);
- printf("Name Length (UTF-16) = %u\n",
- inode->i_ads_entries[i].stream_name_len);
- hash = inode_stream_hash(inode, i + 1);
- if (hash) {
- printf("Hash = 0x");
- print_hash(hash);
- putchar('\n');
+
+ dentry = get_dentry(wim, path, WIMLIB_CASE_SENSITIVE);
+ if (p)
+ *p = T(':');
+ if (!dentry)
+ return -errno;
+
+ inode = dentry->d_inode;
+
+ if (!inode->i_resolved)
+ if (inode_resolve_streams(inode, wim->lookup_table, false))
+ return -EIO;
+
+ if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
+ && inode_is_directory(inode))
+ return -EISDIR;
+
+ if (stream_name) {
+ struct wim_ads_entry *ads_entry;
+ u16 ads_idx;
+ ads_entry = inode_get_ads_entry(inode, stream_name,
+ &ads_idx);
+ if (ads_entry) {
+ stream_idx = ads_idx + 1;
+ lte = ads_entry->lte;
+ goto out;
+ } else {
+ return -ENOENT;
}
- print_lookup_table_entry(inode_stream_lte(inode, i + 1,
- lookup_table));
+ } else {
+ lte = inode_unnamed_stream_resolved(inode, &stream_idx);
}
+out:
+ if (dentry_ret)
+ *dentry_ret = dentry;
+ if (lte_ret)
+ *lte_ret = lte;
+ if (stream_idx_ret)
+ *stream_idx_ret = stream_idx;
return 0;
}
+#endif /* WITH_FUSE */
/* Initializations done on every `struct wim_dentry'. */
-static void dentry_common_init(struct wim_dentry *dentry)
+static void
+dentry_common_init(struct wim_dentry *dentry)
{
memset(dentry, 0, sizeof(struct wim_dentry));
- dentry->refcnt = 1;
}
-static struct wim_inode *new_timeless_inode()
-{
- struct wim_inode *inode = CALLOC(1, sizeof(struct wim_inode));
- if (inode) {
- inode->i_security_id = -1;
- inode->i_nlink = 1;
- #ifdef WITH_FUSE
- inode->i_next_stream_id = 1;
- 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;
-}
-
-static struct wim_inode *new_inode()
-{
- 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;
-}
-
-/*
- * 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 wim_dentry *new_dentry(const char *name)
+/* Creates an unlinked directory entry. */
+int
+new_dentry(const tchar *name, struct wim_dentry **dentry_ret)
{
struct wim_dentry *dentry;
+ int ret;
dentry = MALLOC(sizeof(struct wim_dentry));
- if (!dentry)
- goto err;
+ if (dentry == NULL)
+ return WIMLIB_ERR_NOMEM;
dentry_common_init(dentry);
- if (set_dentry_name(dentry, name) != 0)
- goto err;
-
+ if (*name) {
+ ret = dentry_set_name(dentry, name);
+ if (ret) {
+ FREE(dentry);
+ ERROR("Failed to set name on new dentry with name \"%"TS"\"",
+ name);
+ return ret;
+ }
+ }
dentry->parent = dentry;
-
- return dentry;
-err:
- FREE(dentry);
- ERROR_WITH_ERRNO("Failed to create new dentry with name \"%s\"", name);
- return NULL;
+ *dentry_ret = dentry;
+ return 0;
}
-
-static struct wim_dentry *
-__new_dentry_with_inode(const char *name, bool timeless)
+static int
+_new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret,
+ bool timeless)
{
struct wim_dentry *dentry;
- dentry = new_dentry(name);
- if (dentry) {
- if (timeless)
- dentry->d_inode = new_timeless_inode();
- else
- dentry->d_inode = new_inode();
- if (dentry->d_inode) {
- inode_add_dentry(dentry, dentry->d_inode);
- } else {
- free_dentry(dentry);
- dentry = NULL;
- }
+ 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 == NULL) {
+ free_dentry(dentry);
+ return WIMLIB_ERR_NOMEM;
}
- return dentry;
-}
-struct wim_dentry *new_dentry_with_timeless_inode(const char *name)
-{
- return __new_dentry_with_inode(name, true);
+ inode_add_dentry(dentry, dentry->d_inode);
+ *dentry_ret = dentry;
+ return 0;
}
-struct wim_dentry *new_dentry_with_inode(const char *name)
+int
+new_dentry_with_timeless_inode(const tchar *name, struct wim_dentry **dentry_ret)
{
- return __new_dentry_with_inode(name, false);
+ return _new_dentry_with_inode(name, dentry_ret, true);
}
-
-static int init_ads_entry(struct ads_entry *ads_entry, const char *name)
+int
+new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret)
{
- int ret = 0;
- memset(ads_entry, 0, sizeof(*ads_entry));
- if (name && *name)
- ret = change_ads_name(ads_entry, name);
- return ret;
+ return _new_dentry_with_inode(name, dentry_ret, false);
}
-static void destroy_ads_entry(struct ads_entry *ads_entry)
+int
+new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret)
{
- FREE(ads_entry->stream_name);
- FREE(ads_entry->stream_name_utf8);
-}
+ int ret;
+ struct wim_dentry *dentry;
+ 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;
+}
-/* Frees an inode. */
-void free_inode(struct wim_inode *inode)
+static int
+dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore)
{
- if (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);
- if (inode->i_hlist.pprev)
- hlist_del(&inode->i_hlist);
- #endif
- FREE(inode->i_extracted_file);
- FREE(inode);
- }
+ dentry->d_inode->i_visited = 0;
+ return 0;
}
-/* Decrements link count on an inode and frees it if the link count reaches 0.
- * */
-static void put_inode(struct wim_inode *inode)
+void
+dentry_tree_clear_inode_visited(struct wim_dentry *root)
{
- 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);
- }
- }
+ for_dentry_in_tree(root, dentry_clear_inode_visited, NULL);
}
/* Frees a WIM dentry.
*
* The corresponding inode (if any) is freed only if its link count is
- * decremented to 0.
- */
-void free_dentry(struct wim_dentry *dentry)
-{
- FREE(dentry->file_name);
- FREE(dentry->file_name_utf8);
- FREE(dentry->short_name);
- FREE(dentry->full_path_utf8);
- if (dentry->d_inode)
- put_inode(dentry->d_inode);
- FREE(dentry);
-}
-
-void put_dentry(struct wim_dentry *dentry)
+ * decremented to 0. */
+void
+free_dentry(struct wim_dentry *dentry)
{
- wimlib_assert(dentry->refcnt != 0);
- if (--dentry->refcnt == 0)
- free_dentry(dentry);
+ 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)
+static int
+do_free_dentry(struct wim_dentry *dentry, void *_lookup_table)
{
- struct wim_lookup_table *lookup_table = __lookup_table;
- unsigned i;
+ struct wim_lookup_table *lookup_table = _lookup_table;
if (lookup_table) {
- struct wim_lookup_table_entry *lte;
struct wim_inode *inode = dentry->d_inode;
- wimlib_assert(inode->i_nlink != 0);
- for (i = 0; i <= inode->i_num_ads; i++) {
+ for (unsigned i = 0; i <= inode->i_num_ads; i++) {
+ struct wim_lookup_table_entry *lte;
+
lte = inode_stream_lte(inode, i, lookup_table);
if (lte)
lte_decrement_refcnt(lte, lookup_table);
}
}
-
- put_dentry(dentry);
+ free_dentry(dentry);
return 0;
}
/*
- * Unlinks and frees a dentry tree.
+ * Recursively frees all directory entries in the specified tree.
+ *
+ * @root:
+ * The root of the tree.
+ *
+ * @lookup_table:
+ * The lookup table for dentries. If non-NULL, the reference counts in the
+ * lookup table for the lookup table entries corresponding to the dentries
+ * will be decremented.
+ *
+ * This also puts references to the corresponding inodes.
*
- * @root: The root of the tree.
- * @lookup_table: The lookup table for dentries. If non-NULL, the
- * reference counts in the lookup table for the lookup
- * table entries corresponding to the dentries will be
- * decremented.
+ * This does *not* unlink @root from its parent directory (if it has one).
*/
-void free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table)
+void
+free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table)
{
- if (root)
- for_dentry_in_tree_depth(root, do_free_dentry, lookup_table);
+ for_dentry_in_tree_depth(root, do_free_dentry, lookup_table);
}
-int increment_dentry_refcnt(struct wim_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.
- */
-bool dentry_add_child(struct wim_dentry * restrict parent,
- struct wim_dentry * restrict child)
+/* Insert the @child dentry into the case sensitive index of the @dir directory.
+ * Return NULL if successfully inserted, otherwise a pointer to the
+ * already-inserted duplicate. */
+static struct wim_dentry *
+dir_index_child(struct wim_inode *dir, struct wim_dentry *child)
{
- wimlib_assert(dentry_is_directory(parent));
-
- struct rb_root *root = &parent->d_inode->i_children;
- struct rb_node **new = &(root->rb_node);
- struct rb_node *rb_parent = NULL;
-
- while (*new) {
- struct wim_dentry *this = rbnode_dentry(*new);
- int result = dentry_compare_names(child, this);
-
- rb_parent = *new;
+ struct avl_tree_node *duplicate;
- if (result < 0)
- new = &((*new)->rb_left);
- else if (result > 0)
- new = &((*new)->rb_right);
- else
- return false;
- }
- child->parent = parent;
- rb_link_node(&child->rb_node, rb_parent, new);
- rb_insert_color(&child->rb_node, root);
- return true;
+ duplicate = avl_tree_insert(&dir->i_children,
+ &child->d_index_node,
+ _avl_dentry_compare_names);
+ if (!duplicate)
+ return NULL;
+ return avl_tree_entry(duplicate, struct wim_dentry, d_index_node);
}
-#ifdef WITH_FUSE
-/*
- * Unlink a dentry from the directory tree.
- *
- * Note: This merely removes it from the in-memory tree structure.
- */
-void unlink_dentry(struct wim_dentry *dentry)
+/* Insert the @child dentry into the case insensitive index of the @dir
+ * directory. Return NULL if successfully inserted, otherwise a pointer to the
+ * already-inserted duplicate. */
+static struct wim_dentry *
+dir_index_child_ci(struct wim_inode *dir, struct wim_dentry *child)
{
- struct wim_dentry *parent = dentry->parent;
- if (parent == dentry)
- return;
- rb_erase(&dentry->rb_node, &parent->d_inode->i_children);
+ struct avl_tree_node *duplicate;
+
+ duplicate = avl_tree_insert(&dir->i_children_ci,
+ &child->d_index_node_ci,
+ _avl_dentry_compare_names_ci);
+ if (!duplicate)
+ return NULL;
+ return avl_tree_entry(duplicate, struct wim_dentry, d_index_node_ci);
}
-#endif
-#ifdef WITH_FUSE
-/* Returns the alternate data stream entry belonging to @inode that has the
- * stream name @stream_name. */
-struct ads_entry *inode_get_ads_entry(struct wim_inode *inode,
- const char *stream_name,
- u16 *idx_ret)
+/* Removes the specified dentry from its directory's case-sensitive index. */
+static void
+dir_unindex_child(struct wim_inode *dir, struct wim_dentry *child)
{
- if (inode->i_num_ads != 0) {
- u16 i = 0;
- size_t stream_name_len = strlen(stream_name);
- do {
- if (ads_entry_has_name(&inode->i_ads_entries[i],
- stream_name, stream_name_len))
- {
- if (idx_ret)
- *idx_ret = i;
- return &inode->i_ads_entries[i];
- }
- } while (++i != inode->i_num_ads);
- }
- return NULL;
+ avl_tree_remove(&dir->i_children, &child->d_index_node);
}
-#endif
-#if defined(WITH_FUSE) || defined(WITH_NTFS_3G)
-/*
- * 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 wim_inode *inode, const char *stream_name)
+/* Removes the specified dentry from its directory's case-insensitive index. */
+static void
+dir_unindex_child_ci(struct wim_inode *dir, struct wim_dentry *child)
{
- u16 num_ads;
- struct ads_entry *ads_entries;
- struct ads_entry *new_entry;
-
- DEBUG("Add alternate data stream \"%s\"", stream_name);
-
- 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) != 0)
- return NULL;
-#ifdef WITH_FUSE
- new_entry->stream_id = inode->i_next_stream_id++;
-#endif
- inode->i_num_ads = num_ads;
- return new_entry;
+ avl_tree_remove(&dir->i_children_ci, &child->d_index_node_ci);
}
-#endif
-#ifdef WITH_FUSE
-/* Remove an alternate data stream from the inode */
-void inode_remove_ads(struct wim_inode *inode, u16 idx,
- struct wim_lookup_table *lookup_table)
+/* Returns true iff the specified dentry is in its parent directory's
+ * case-insensitive index. */
+static bool
+dentry_in_ci_index(const struct wim_dentry *dentry)
{
- struct 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 \"%s\"", ads_entry->stream_name_utf8);
-
- lte = ads_entry->lte;
- if (lte)
- lte_decrement_refcnt(lte, lookup_table);
-
- destroy_ads_entry(ads_entry);
-
- memcpy(&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--;
+ return !avl_tree_node_is_unlinked(&dentry->d_index_node_ci);
}
-#endif
-
-
/*
- * Reads the alternate data stream entries for a dentry.
- *
- * @p: Pointer to buffer that starts with the first alternate stream entry.
- *
- * @inode: Inode to load the alternate data streams into.
- * @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.
- * };
+ * Links a dentry into the directory tree.
*
- * In addition, the entries are 8-byte aligned.
+ * @parent: The dentry that will be the parent of @child.
+ * @child: The dentry to link.
*
- * Return 0 on success or nonzero on failure. On success, inode->i_ads_entries
- * is set to an array of `struct ads_entry's of length inode->i_num_ads. On
- * failure, @inode is not modified.
+ * Returns NULL if successful. If @parent already contains a dentry with the
+ * same case-sensitive name as @child, returns a pointer to this duplicate
+ * dentry.
*/
-static int read_ads_entries(const u8 *p, struct wim_inode *inode,
- u64 remaining_size)
+struct wim_dentry *
+dentry_add_child(struct wim_dentry *parent, struct wim_dentry *child)
{
- u16 num_ads;
- struct ads_entry *ads_entries;
- int ret;
+ struct wim_dentry *duplicate;
+ struct wim_inode *dir;
- num_ads = inode->i_num_ads;
- ads_entries = CALLOC(num_ads, sizeof(inode->i_ads_entries[0]));
- if (!ads_entries) {
- ERROR("Could not allocate memory for %"PRIu16" "
- "alternate data stream entries", num_ads);
- return WIMLIB_ERR_NOMEM;
- }
+ wimlib_assert(parent != child);
- 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;
+ dir = parent->d_inode;
- cur_entry = &ads_entries[i];
+ wimlib_assert(inode_is_directory(dir));
- #ifdef WITH_FUSE
- ads_entries[i].stream_id = i + 1;
- #endif
+ duplicate = dir_index_child(dir, child);
+ if (duplicate)
+ return duplicate;
- /* 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;
- }
+ duplicate = dir_index_child_ci(dir, child);
+ if (duplicate) {
+ list_add(&child->d_ci_conflict_list, &duplicate->d_ci_conflict_list);
+ avl_tree_node_set_unlinked(&child->d_index_node_ci);
+ } else {
+ INIT_LIST_HEAD(&child->d_ci_conflict_list);
+ }
+ child->parent = parent;
+ return NULL;
+}
- 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);
+/* Unlink a WIM dentry from the directory entry tree. */
+void
+unlink_dentry(struct wim_dentry *dentry)
+{
+ struct wim_inode *dir;
- cur_entry->stream_name = NULL;
- cur_entry->stream_name_utf8 = NULL;
+ if (dentry_is_root(dentry))
+ return;
- /* Length including neither the null terminator nor the padding
- * */
- length_no_padding = WIM_ADS_ENTRY_DISK_SIZE +
- cur_entry->stream_name_len;
+ dir = dentry->parent->d_inode;
- /* Length including the null terminator and the padding */
- total_length = ((length_no_padding + 2) + 7) & ~7;
+ dir_unindex_child(dir, dentry);
- wimlib_assert(total_length == ads_entry_total_length(cur_entry));
+ if (dentry_in_ci_index(dentry)) {
- 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;
- }
+ dir_unindex_child_ci(dir, dentry);
- /* 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 (!list_empty(&dentry->d_ci_conflict_list)) {
+ /* Make a different case-insensitively-the-same dentry
+ * be the "representative" in the search index. */
+ struct list_head *next;
+ struct wim_dentry *other;
+ struct wim_dentry *existing;
- 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);
-
- ret = utf16_to_utf8(cur_entry->stream_name,
- cur_entry->stream_name_len,
- &cur_entry->stream_name_utf8,
- &utf8_len);
- if (ret != 0)
- goto out_free_ads_entries;
- cur_entry->stream_name_utf8_len = utf8_len;
+ next = dentry->d_ci_conflict_list.next;
+ other = list_entry(next, struct wim_dentry, d_ci_conflict_list);
+ existing = dir_index_child_ci(dir, other);
+ wimlib_assert(existing == NULL);
}
- /* 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;
- else
- remaining_size -= total_length;
}
- inode->i_ads_entries = ads_entries;
-#ifdef WITH_FUSE
- inode->i_next_stream_id = inode->i_num_ads + 1;
-#endif
- return 0;
-out_free_ads_entries:
- for (u16 i = 0; i < num_ads; i++)
- destroy_ads_entry(&ads_entries[i]);
- FREE(ads_entries);
- return ret;
+ list_del(&dentry->d_ci_conflict_list);
}
-/*
- * Reads a 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 wim_dentry' that will be filled in by this function.
- *
- * 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 wim_dentry *dentry)
+/* Reads a WIM directory entry, including all alternate data stream entries that
+ * follow it, from the WIM image's metadata resource. */
+static int
+read_dentry(const u8 * restrict buf, size_t buf_len,
+ u64 offset, struct wim_dentry **dentry_ret)
{
+ u64 length;
const u8 *p;
+ const struct wim_dentry_on_disk *disk_dentry;
+ struct wim_dentry *dentry;
+ struct wim_inode *inode;
+ u16 short_name_nbytes;
+ u16 file_name_nbytes;
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;
int ret;
- struct wim_inode *inode = NULL;
- dentry_common_init(dentry);
-
- /*Make sure the dentry really fits into the metadata resource.*/
- if (offset + 8 > metadata_resource_len || offset + 8 < offset) {
- 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;
- }
+ BUILD_BUG_ON(sizeof(struct wim_dentry_on_disk) != WIM_DENTRY_DISK_SIZE);
/* 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);
+ /* Check for buffer overrun. */
+ if (unlikely(offset + sizeof(u64) > buf_len ||
+ offset + sizeof(u64) < offset))
+ {
+ ERROR("Directory entry starting at %"PRIu64" ends past the "
+ "end of the metadata resource (size %zu)",
+ offset, buf_len);
+ return WIMLIB_ERR_INVALID_METADATA_RESOURCE;
+ }
- /* 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 == 0)
- return 0;
+ /* Get pointer to the dentry data. */
+ p = &buf[offset];
+ disk_dentry = (const struct wim_dentry_on_disk*)p;
- /* 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. */
- if (offset + dentry->length >= metadata_resource_len
- || offset + dentry->length < offset)
- {
- ERROR("Directory entry at offset %"PRIu64" and with size "
- "%"PRIu64" ends past the end of the metadata resource "
- "(size %"PRIu64")",
- offset, dentry->length, metadata_resource_len);
- return WIMLIB_ERR_INVALID_DENTRY;
+ if (unlikely((uintptr_t)p & 7))
+ WARNING("WIM dentry is not 8-byte aligned");
+
+ /* Get dentry length. */
+ length = le64_to_cpu(disk_dentry->length);
+
+ /* Check for end-of-directory. */
+ if (length <= 8) {
+ *dentry_ret = NULL;
+ return 0;
}
- if (dentry->length < WIM_DENTRY_DISK_SIZE) {
+ /* Validate dentry length. */
+ if (unlikely(length < sizeof(struct wim_dentry_on_disk))) {
ERROR("Directory entry has invalid length of %"PRIu64" bytes",
- dentry->length);
- return WIMLIB_ERR_INVALID_DENTRY;
+ length);
+ return WIMLIB_ERR_INVALID_METADATA_RESOURCE;
}
- inode = new_timeless_inode();
- if (!inode)
- return WIMLIB_ERR_NOMEM;
-
- p = get_u32(p, &inode->i_attributes);
- p = get_u32(p, (u32*)&inode->i_security_id);
- p = get_u64(p, &dentry->subdir_offset);
+ /* Check for buffer overrun. */
+ if (unlikely(offset + length > buf_len ||
+ offset + length < offset))
+ {
+ ERROR("Directory entry at offset %"PRIu64" and with size "
+ "%"PRIu64" ends past the end of the metadata resource "
+ "(size %zu)", offset, length, buf_len);
+ return WIMLIB_ERR_INVALID_METADATA_RESOURCE;
+ }
- /* 2 unused fields */
- p += 2 * sizeof(u64);
- /*p = get_u64(p, &dentry->unused1);*/
- /*p = get_u64(p, &dentry->unused2);*/
+ /* Allocate new dentry structure, along with a preliminary inode. */
+ ret = new_dentry_with_timeless_inode(T(""), &dentry);
+ if (ret)
+ return ret;
- p = get_u64(p, &inode->i_creation_time);
- p = get_u64(p, &inode->i_last_access_time);
- p = get_u64(p, &inode->i_last_write_time);
+ dentry->length = length;
+ inode = dentry->d_inode;
- p = get_bytes(p, SHA1_HASH_SIZE, inode->i_hash);
+ /* Read more fields: some into the dentry, and some into the inode. */
+ 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);
+ 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);
- /*
- * 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.
- */
+ * document it. So we have some fields we read for reparse points, and
+ * some fields in the same place for non-reparse-points. */
if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
- /* ??? */
- p += 4;
- p = get_u32(p, &inode->i_reparse_tag);
- p += 4;
+ 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->i_reparse_tag);
- p = get_u64(p, &inode->i_ino);
+ 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);
}
+ inode->i_num_ads = le16_to_cpu(disk_dentry->num_alternate_data_streams);
- /* By the way, the reparse_reserved field does not actually exist (at
- * least when the file is not a reparse point) */
+ /* Now onto reading the names. There are two of them: the (long) file
+ * name, and the short name. */
- p = get_u16(p, &inode->i_num_ads);
+ short_name_nbytes = le16_to_cpu(disk_dentry->short_name_nbytes);
+ file_name_nbytes = le16_to_cpu(disk_dentry->file_name_nbytes);
- p = get_u16(p, &short_name_len);
- p = get_u16(p, &file_name_len);
+ if (unlikely((short_name_nbytes & 1) | (file_name_nbytes & 1))) {
+ ERROR("Dentry name is not valid UTF-16 (odd number of bytes)!");
+ ret = WIMLIB_ERR_INVALID_METADATA_RESOURCE;
+ goto err_free_dentry;
+ }
/* 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 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);
+ * would be unexpected. */
+ calculated_size = dentry_correct_length_unaligned(file_name_nbytes,
+ short_name_nbytes);
- if (dentry->length < calculated_size) {
+ if (unlikely(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);
- ret = WIMLIB_ERR_INVALID_DENTRY;
- goto out_free_inode;
+ "at least %"PRIu64" bytes, got %"PRIu64" bytes.)",
+ calculated_size, dentry->length);
+ ret = WIMLIB_ERR_INVALID_METADATA_RESOURCE;
+ goto err_free_dentry;
}
+ /* Advance p to point past the base dentry, to the first name. */
+ 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) {
- ERROR("Failed to allocate %hu bytes for dentry file name",
- file_name_len);
+ * is no null terminator following it. */
+ if (file_name_nbytes) {
+ dentry->file_name = MALLOC(file_name_nbytes + 2);
+ if (dentry->file_name == NULL) {
ret = WIMLIB_ERR_NOMEM;
- goto out_free_inode;
+ goto err_free_dentry;
}
- p = get_bytes(p, file_name_len, file_name);
-
- /* Convert filename to UTF-8. */
- ret = utf16_to_utf8(file_name, file_name_len, &file_name_utf8,
- &file_name_utf8_len);
- if (ret != 0)
- goto out_free_file_name;
- if (*(u16*)p)
- WARNING("Expected two zero bytes following the file name "
- "`%s', but found non-zero bytes", file_name_utf8);
- p += 2;
- }
-
- /* 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);
+ dentry->file_name_nbytes = file_name_nbytes;
+ memcpy(dentry->file_name, p, file_name_nbytes);
+ p += file_name_nbytes + 2;
+ dentry->file_name[file_name_nbytes / 2] = cpu_to_le16(0);
}
/* 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) {
- ERROR("Failed to allocate %hu bytes for short filename",
- short_name_len);
+ if (short_name_nbytes) {
+ dentry->short_name = MALLOC(short_name_nbytes + 2);
+ if (dentry->short_name == NULL) {
ret = WIMLIB_ERR_NOMEM;
- goto out_free_file_name_utf8;
+ goto err_free_dentry;
}
-
- p = get_bytes(p, short_name_len, short_name);
- if (*(u16*)p)
- WARNING("Expected two zero bytes following the short name of "
- "`%s', but found non-zero bytes", file_name_utf8);
- p += 2;
+ dentry->short_name_nbytes = short_name_nbytes;
+ memcpy(dentry->short_name, p, short_name_nbytes);
+ p += short_name_nbytes + 2;
+ dentry->short_name[short_name_nbytes / 2] = cpu_to_le16(0);
}
- /*
- * 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.
+ /* Align the dentry length. */
+ dentry->length = (dentry->length + 7) & ~7;
+
+ /* Read the alternate data streams, if present. inode->i_num_ads tells
+ * us how many they are, and they will directly follow the dentry in the
+ * metadata resource buffer.
*
* 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.
- */
- if (inode->i_num_ads != 0) {
+ * aligned boundary, and the alternate data stream entries seem to NOT
+ * be included in the dentry->length field for some reason. */
+ if (unlikely(inode->i_num_ads != 0)) {
+ ret = WIMLIB_ERR_INVALID_METADATA_RESOURCE;
+ if (offset + dentry->length > buf_len ||
+ (ret = read_ads_entries(&buf[offset + dentry->length],
+ inode,
+ buf_len - offset - dentry->length)))
+ {
+ ERROR("Failed to read alternate data stream "
+ "entries of WIM dentry \"%"WS"\"",
+ dentry->file_name);
+ goto err_free_dentry;
+ }
+ }
- /* Trying different lengths is just a hack to make sure we have
- * a chance of reading the ADS entries correctly despite the
- * poor documentation. */
+ *dentry_ret = dentry;
+ return 0;
+
+err_free_dentry:
+ free_dentry(dentry);
+ return ret;
+}
+
+static const tchar *
+dentry_get_file_type_string(const struct wim_dentry *dentry)
+{
+ const struct wim_inode *inode = dentry->d_inode;
+ if (inode_is_directory(inode))
+ return T("directory");
+ else if (inode_is_symlink(inode))
+ return T("symbolic link");
+ else
+ return T("file");
+}
- if (calculated_size != dentry->length) {
- WARNING("Trying calculated dentry length (%"PRIu64") "
- "instead of dentry->length field (%"PRIu64") "
- "to read ADS entries",
- calculated_size, dentry->length);
+static bool
+dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
+{
+ if (dentry->file_name_nbytes <= 4) {
+ if (dentry->file_name_nbytes == 4) {
+ if (dentry->file_name[0] == cpu_to_le16('.') &&
+ dentry->file_name[1] == cpu_to_le16('.'))
+ return true;
+ } else if (dentry->file_name_nbytes == 2) {
+ if (dentry->file_name[0] == cpu_to_le16('.'))
+ return true;
}
- u64 lengths_to_try[3] = {calculated_size,
- (dentry->length + 7) & ~7,
- dentry->length};
- ret = WIMLIB_ERR_INVALID_DENTRY;
- for (size_t i = 0; i < ARRAY_LEN(lengths_to_try); i++) {
- if (lengths_to_try[i] > metadata_resource_len - offset)
- continue;
- ret = read_ads_entries(&metadata_resource[offset + lengths_to_try[i]],
- inode,
- metadata_resource_len - offset - lengths_to_try[i]);
- if (ret == 0)
- goto out;
+ }
+ return false;
+}
+
+static int
+read_dentry_tree_recursive(const u8 * restrict buf, size_t buf_len,
+ struct wim_dentry * restrict dir)
+{
+ u64 cur_offset = dir->subdir_offset;
+
+ /* Check for cyclic directory structure, which would cause infinite
+ * recursion if not handled. */
+ for (struct wim_dentry *d = dir->parent;
+ !dentry_is_root(d); d = d->parent)
+ {
+ if (unlikely(d->subdir_offset == cur_offset)) {
+ ERROR("Cyclic directory structure detected: children "
+ "of \"%"TS"\" coincide with children of \"%"TS"\"",
+ dentry_full_path(dir), dentry_full_path(d));
+ return WIMLIB_ERR_INVALID_METADATA_RESOURCE;
}
- ERROR("Failed to read alternate data stream "
- "entries of `%s'", dentry->file_name_utf8);
- goto out_free_short_name;
}
-out:
- /* We've read all the data for this dentry. Set the names and their
- * lengths, and we've done. */
- dentry->d_inode = inode;
- 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;
-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);
- return ret;
+ for (;;) {
+ struct wim_dentry *child;
+ struct wim_dentry *duplicate;
+ int ret;
+
+ /* Read next child of @dir. */
+ ret = read_dentry(buf, buf_len, cur_offset, &child);
+ if (ret)
+ return ret;
+
+ /* Check for end of directory. */
+ if (child == NULL)
+ return 0;
+
+ /* 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_in_total_length(child);
+
+ /* All dentries except the root should be named. */
+ if (unlikely(!dentry_has_long_name(child))) {
+ WARNING("Ignoring unnamed dentry in "
+ "directory \"%"TS"\"", dentry_full_path(dir));
+ free_dentry(child);
+ continue;
+ }
+
+ /* Don't allow files named "." or "..". */
+ if (unlikely(dentry_is_dot_or_dotdot(child))) {
+ WARNING("Ignoring file named \".\" or \"..\"; "
+ "potentially malicious archive!!!");
+ free_dentry(child);
+ continue;
+ }
+
+ /* Link the child into the directory. */
+ duplicate = dentry_add_child(dir, child);
+ if (unlikely(duplicate)) {
+ /* We already found a dentry with this same
+ * case-sensitive long name. Only keep the first one.
+ */
+ const tchar *child_type, *duplicate_type;
+ child_type = dentry_get_file_type_string(child);
+ duplicate_type = dentry_get_file_type_string(duplicate);
+ WARNING("Ignoring duplicate %"TS" \"%"TS"\" "
+ "(the WIM image already contains a %"TS" "
+ "at that path with the exact same name)",
+ child_type, dentry_full_path(duplicate),
+ duplicate_type);
+ free_dentry(child);
+ continue;
+ }
+
+ /* If this child is a directory that itself has children, call
+ * this procedure recursively. */
+ if (child->subdir_offset != 0) {
+ if (likely(dentry_is_directory(child))) {
+ ret = read_dentry_tree_recursive(buf,
+ buf_len,
+ child);
+ if (ret)
+ return ret;
+ } else {
+ WARNING("Ignoring children of "
+ "non-directory file \"%"TS"\"",
+ dentry_full_path(child));
+ }
+ }
+ }
}
-/* Reads the children of a dentry, and all their children, ..., etc. from the
- * metadata resource and into the dentry tree.
+/*
+ * Read a tree of dentries (directory entries) from a WIM metadata resource.
*
- * @metadata_resource: An array that contains the uncompressed metadata
- * resource for the WIM file.
+ * @buf:
+ * Buffer containing an uncompressed WIM metadata resource.
*
- * @metadata_resource_len: The length of the uncompressed metadata resource, in
- * bytes.
+ * @buf_len:
+ * 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.
+ * @root_offset
+ * Offset in the metadata resource of the root of the dentry tree.
*
- * @return: Zero on success, nonzero on failure.
+ * @root_ret:
+ * On success, either NULL or a pointer to the root dentry is written to
+ * this location. The former case only occurs in the unexpected case that
+ * the tree began with an end-of-directory entry.
+ *
+ * Return values:
+ * WIMLIB_ERR_SUCCESS (0)
+ * WIMLIB_ERR_INVALID_METADATA_RESOURCE
+ * WIMLIB_ERR_NOMEM
*/
-int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
- struct wim_dentry *dentry)
+int
+read_dentry_tree(const u8 *buf, size_t buf_len,
+ u64 root_offset, struct wim_dentry **root_ret)
{
- u64 cur_offset = dentry->subdir_offset;
- struct wim_dentry *child;
- struct wim_dentry cur_child;
int ret;
+ struct wim_dentry *root;
- /*
- * 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;
+ DEBUG("Reading dentry tree (root_offset=%"PRIu64")", root_offset);
- /* 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 wim_dentry));
- if (!child) {
- ERROR("Failed to allocate %zu bytes for new dentry",
- sizeof(struct wim_dentry));
- ret = WIMLIB_ERR_NOMEM;
- break;
+ ret = read_dentry(buf, buf_len, root_offset, &root);
+ if (ret)
+ return ret;
+
+ if (likely(root != NULL)) {
+ if (unlikely(dentry_has_long_name(root) ||
+ dentry_has_short_name(root)))
+ {
+ WARNING("The root directory has a nonempty name; "
+ "removing it.");
+ FREE(root->file_name);
+ FREE(root->short_name);
+ root->file_name = NULL;
+ root->short_name = NULL;
+ root->file_name_nbytes = 0;
+ root->short_name_nbytes = 0;
}
- memcpy(child, &cur_child, sizeof(struct wim_dentry));
- dentry_add_child(dentry, child);
- inode_add_dentry(child, child->d_inode);
- /* 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;
+ if (unlikely(!dentry_is_directory(root))) {
+ ERROR("The root of the WIM image is not a directory!");
+ ret = WIMLIB_ERR_INVALID_METADATA_RESOURCE;
+ goto err_free_dentry_tree;
}
- /* 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);
+ if (likely(root->subdir_offset != 0)) {
+ ret = read_dentry_tree_recursive(buf, buf_len, root);
+ if (ret)
+ goto err_free_dentry_tree;
+ }
+ } else {
+ WARNING("The metadata resource has no directory entries; "
+ "treating as an empty image.");
}
+ *root_ret = root;
+ return 0;
+
+err_free_dentry_tree:
+ free_dentry_tree(root, NULL);
return ret;
}
+/*
+ * Writes a WIM alternate data stream (ADS) entry to an output buffer.
+ *
+ * @ads_entry: The ADS entry structure.
+ * @hash: The hash field to use (instead of the one in the ADS entry).
+ * @p: The memory location to write the data to.
+ *
+ * Returns a pointer to the byte after the last byte written.
+ */
+static u8 *
+write_ads_entry(const struct wim_ads_entry *ads_entry,
+ const u8 *hash, u8 * restrict p)
+{
+ struct wim_ads_entry_on_disk *disk_ads_entry =
+ (struct wim_ads_entry_on_disk*)p;
+ u8 *orig_p = p;
+
+ disk_ads_entry->reserved = cpu_to_le64(ads_entry->reserved);
+ 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);
+ }
+ /* Align to 8-byte boundary */
+ while ((uintptr_t)p & 7)
+ *p++ = 0;
+ disk_ads_entry->length = cpu_to_le64(p - orig_p);
+ return p;
+}
+
/*
* 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 stream entries.
*/
-static u8 *write_dentry(const struct wim_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 wim_inode *inode = dentry->d_inode;
+ bool use_dummy_stream;
+ u16 num_ads;
- /* 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->i_attributes);
- p = put_u32(p, inode->i_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->i_creation_time);
- p = put_u64(p, inode->i_last_access_time);
- p = put_u64(p, inode->i_last_write_time);
- hash = inode_stream_hash(inode, 0);
- p = put_bytes(p, SHA1_HASH_SIZE, hash);
+ wimlib_assert(((uintptr_t)p & 7) == 0); /* 8 byte aligned */
+ orig_p = p;
+
+ inode = dentry->d_inode;
+ use_dummy_stream = inode_needs_dummy_stream(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(0);
+ disk_dentry->unused_2 = cpu_to_le64(0);
+ 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 (use_dummy_stream)
+ hash = zero_hash;
+ else
+ hash = inode_stream_hash(inode, 0);
+ copy_hash(disk_dentry->unnamed_stream_hash, hash);
if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
- p = put_zeroes(p, 4);
- p = put_u32(p, inode->i_reparse_tag);
- p = put_zeroes(p, 4);
+ 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 {
- u64 link_group_id;
- p = put_u32(p, 0);
- if (inode->i_nlink == 1)
- link_group_id = 0;
- else
- link_group_id = inode->i_ino;
- p = put_u64(p, link_group_id);
- }
- p = put_u16(p, inode->i_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->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);
}
+ num_ads = inode->i_num_ads;
+ if (use_dummy_stream)
+ num_ads++;
+ disk_dentry->num_alternate_data_streams = cpu_to_le16(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);
+
+ wimlib_assert(dentry_is_root(dentry) != dentry_has_long_name(dentry));
+
+ 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 */
- wimlib_assert(length >= (p - orig_p) && length - (p - orig_p) <= 7);
- p = put_zeroes(p, length - (p - orig_p));
+ 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...). Furthermore, the dentry may
+ * have been renamed, thus changing its needed length. */
+ disk_dentry->length = cpu_to_le64(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. */
+ if (use_dummy_stream) {
+ hash = inode_unnamed_stream_hash(inode);
+ p = write_ads_entry(&(struct wim_ads_entry){}, hash, p);
+ }
+
+ /* Write the alternate data streams entries, if any. */
for (u16 i = 0; i < inode->i_num_ads; i++) {
- p = put_u64(p, ads_entry_total_length(&inode->i_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->i_ads_entries[i].stream_name_len);
- if (inode->i_ads_entries[i].stream_name_len) {
- p = put_bytes(p, inode->i_ads_entries[i].stream_name_len,
- (u8*)inode->i_ads_entries[i].stream_name);
- p = put_u16(p, 0);
- }
- p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
+ p = write_ads_entry(&inode->i_ads_entries[i], hash, p);
}
- wimlib_assert(p - orig_p == __dentry_total_length(dentry, length));
+
return p;
}
-static int write_dentry_cb(struct wim_dentry *dentry, void *_p)
+static int
+write_dir_dentries(struct wim_dentry *dir, void *_pp)
{
- u8 **p = _p;
- *p = write_dentry(dentry, *p);
- return 0;
-}
+ if (dir->subdir_offset != 0) {
+ u8 **pp = _pp;
+ u8 *p = *pp;
+ struct wim_dentry *child;
-static u8 *write_dentry_tree_recursive(const struct wim_dentry *parent, u8 *p);
+ /* write child dentries */
+ for_dentry_child(child, dir)
+ p = write_dentry(child, p);
-static int write_dentry_tree_recursive_cb(struct wim_dentry *dentry, void *_p)
-{
- u8 **p = _p;
- *p = write_dentry_tree_recursive(dentry, *p);
+ /* write end of directory entry */
+ *(u64*)p = 0;
+ p += 8;
+ *pp = 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 wim_dentry *parent, u8 *p)
-{
- /* Nothing to do if this dentry has no children. */
- if (parent->subdir_offset == 0)
- return p;
-
- /* 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! */
- for_dentry_in_rbtree(parent->d_inode->i_children.rb_node, write_dentry_cb, &p);
-
- /* write end of directory entry */
- p = put_u64(p, 0);
-
- /* Recurse on children. */
- for_dentry_in_rbtree(parent->d_inode->i_children.rb_node,
- write_dentry_tree_recursive_cb, &p);
- return p;
-}
-
/* Writes a directory tree to the metadata resource.
*
* @root: Root of the dentry tree.
*
* Returns pointer to the byte after the last byte we wrote.
*/
-u8 *write_dentry_tree(const struct wim_dentry *root, u8 *p)
+u8 *
+write_dentry_tree(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
- * wrote us, so we need to write ourselves. */
+ /* write root dentry and end-of-directory entry following it */
p = write_dentry(root, p);
+ *(u64*)p = 0;
+ p += 8;
- /* 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);
+ /* write the rest of the dentry tree */
+ for_dentry_in_tree(root, write_dir_dentries, &p);
- /* Recursively write the rest of the dentry tree. */
- return write_dentry_tree_recursive(root, p);
+ return p;
}