X-Git-Url: https://wimlib.net/git/?a=blobdiff_plain;f=src%2Finode.c;h=7242550eac03247a80ed2f637437e9bc58aa0715;hb=93110bb18090d4d2c00294a56f819c7edeef318f;hp=34508659fa6724d56142552960a659df273a69ea;hpb=a1d2684c03d066f103ae50485f48512613b84575;p=wimlib

diff --git a/src/inode.c b/src/inode.c
index 34508659..7242550e 100644
--- a/src/inode.c
+++ b/src/inode.c
@@ -1,9 +1,14 @@
 /*
  * inode.c
+ *
+ * Functions that operate on WIM inodes.
+ *
+ * See dentry.c for a description of the relationship between WIM dentries and
+ * WIM inodes.
  */
 
 /*
- * Copyright (C) 2012, 2013 Eric Biggers
+ * Copyright (C) 2012, 2013, 2014 Eric Biggers
  *
  * This file is part of wimlib, a library for working with WIM files.
  *
@@ -26,7 +31,6 @@
 #endif
 
 #include "wimlib/assert.h"
-#include "wimlib/case.h"
 #include "wimlib/dentry.h" /* Only for dentry_full_path().  Otherwise the code
 			      in this file doesn't care about file names/paths.
 			    */
@@ -39,6 +43,8 @@
 #include "wimlib/security.h"
 #include "wimlib/timestamp.h"
 
+#include <errno.h>
+
 /* Allocate a new inode.  Set the timestamps to the current time.  */
 struct wim_inode *
 new_inode(void)
@@ -53,7 +59,6 @@ new_inode(void)
 	return inode;
 }
 
-
 /* Allocate a new inode.  Leave the timestamps zeroed out.  */
 struct wim_inode *
 new_timeless_inode(void)
@@ -78,9 +83,7 @@ put_inode(struct wim_inode *inode)
 	wimlib_assert(inode->i_nlink != 0);
 	if (--inode->i_nlink == 0) {
 		/* If FUSE mounts are enabled, we must keep a unlinked inode
-		 * around until all file descriptors to it have been closed.
-		 * inode_put_fd() in mount_image.c handles dropping a file
-		 * descriptor.  */
+		 * around until all file descriptors to it have been closed.  */
 	#ifdef WITH_FUSE
 		if (inode->i_num_opened_fds == 0)
 	#endif
@@ -88,7 +91,7 @@ put_inode(struct wim_inode *inode)
 	}
 }
 
-/* De-allocate memory for an alternate data stream entry.  */
+/* Free memory allocated within an alternate data stream entry.  */
 static void
 destroy_ads_entry(struct wim_ads_entry *ads_entry)
 {
@@ -100,14 +103,16 @@ destroy_ads_entry(struct wim_ads_entry *ads_entry)
 void
 free_inode(struct wim_inode *inode)
 {
-	if (inode == NULL)
+	if (unlikely(!inode))
 		return;
 
-	if (inode->i_ads_entries) {
+	if (unlikely(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);
 	}
+	if (unlikely(inode->i_extra))
+		FREE(inode->i_extra);
 	/* HACK: This may instead delete the inode from i_list, but hlist_del()
 	 * behaves the same as list_del(). */
 	if (!hlist_unhashed(&inode->i_hlist))
@@ -136,162 +141,169 @@ ads_entry_has_name(const struct wim_ads_entry *entry,
  *
  * If @p stream_name is the empty string, NULL is returned --- that is, this
  * function will not return "unnamed" alternate data stream entries.
+ *
+ * If NULL is returned, errno is set.
  */
 struct wim_ads_entry *
-inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name,
-		    u16 *idx_ret)
+inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name)
 {
+	int ret;
+	const utf16lechar *stream_name_utf16le;
+	size_t stream_name_utf16le_nbytes;
+	u16 i;
+	struct wim_ads_entry *result;
+
 	if (inode->i_num_ads == 0) {
+		errno = ENOENT;
 		return NULL;
-	} else {
-		size_t stream_name_utf16le_nbytes;
-		u16 i;
-		struct wim_ads_entry *result;
-
-		if (stream_name[0] == T('\0'))
-			return NULL;
+	}
 
-	#if TCHAR_IS_UTF16LE
-		const utf16lechar *stream_name_utf16le;
+	if (stream_name[0] == T('\0')) {
+		errno = ENOENT;
+		return NULL;
+	}
 
-		stream_name_utf16le = stream_name;
-		stream_name_utf16le_nbytes = tstrlen(stream_name) * sizeof(tchar);
-	#else
-		utf16lechar *stream_name_utf16le;
+	ret = tstr_get_utf16le_and_len(stream_name, &stream_name_utf16le,
+				       &stream_name_utf16le_nbytes);
+	if (ret)
+		return NULL;
 
+	i = 0;
+	result = NULL;
+	do {
+		if (ads_entry_has_name(&inode->i_ads_entries[i],
+				       stream_name_utf16le,
+				       stream_name_utf16le_nbytes,
+				       default_ignore_case))
 		{
-			int ret = tstr_to_utf16le(stream_name,
-						  tstrlen(stream_name) *
-						      sizeof(tchar),
-						  &stream_name_utf16le,
-						  &stream_name_utf16le_nbytes);
-			if (ret)
-				return NULL;
+			result = &inode->i_ads_entries[i];
+			break;
 		}
-	#endif
-		i = 0;
-		result = NULL;
-		do {
-			if (ads_entry_has_name(&inode->i_ads_entries[i],
-					       stream_name_utf16le,
-					       stream_name_utf16le_nbytes,
-					       default_ignore_case))
-			{
-				if (idx_ret)
-					*idx_ret = i;
-				result = &inode->i_ads_entries[i];
-				break;
-			}
-		} while (++i != inode->i_num_ads);
-	#if !TCHAR_IS_UTF16LE
-		FREE(stream_name_utf16le);
-	#endif
-		return result;
-	}
-}
+	} while (++i != inode->i_num_ads);
 
-static int
-init_ads_entry(struct wim_ads_entry *ads_entry, const void *name,
-	       size_t name_nbytes, bool is_utf16le)
-{
-	int ret = 0;
-	memset(ads_entry, 0, sizeof(*ads_entry));
+	tstr_put_utf16le(stream_name_utf16le);
 
-	if (is_utf16le) {
-		utf16lechar *p = MALLOC(name_nbytes + sizeof(utf16lechar));
-		if (p == NULL)
-			return WIMLIB_ERR_NOMEM;
-		memcpy(p, name, name_nbytes);
-		p[name_nbytes / 2] = cpu_to_le16(0);
-		ads_entry->stream_name = p;
-		ads_entry->stream_name_nbytes = name_nbytes;
-	} else {
-		if (name && *(const tchar*)name != T('\0')) {
-			ret = get_utf16le_string(name, &ads_entry->stream_name,
-						 &ads_entry->stream_name_nbytes);
-		}
-	}
-	return ret;
+	if (!result)
+		errno = ENOENT;
+	return result;
 }
 
 static struct wim_ads_entry *
-do_inode_add_ads(struct wim_inode *inode, const void *stream_name,
-		 size_t stream_name_nbytes, bool is_utf16le)
+do_inode_add_ads(struct wim_inode *inode,
+		 utf16lechar *stream_name, size_t stream_name_nbytes)
 {
 	u16 num_ads;
 	struct wim_ads_entry *ads_entries;
 	struct wim_ads_entry *new_entry;
 
-	wimlib_assert(stream_name_nbytes != 0);
-
 	if (inode->i_num_ads >= 0xfffe) {
-		ERROR("Too many alternate data streams in one inode!");
+		ERROR("File \"%"TS"\" has too many alternate data streams!",
+		      inode_first_full_path(inode));
+		errno = EFBIG;
 		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 == NULL) {
-		ERROR("Failed to allocate memory for new alternate data stream");
+	if (!ads_entries)
 		return NULL;
-	}
+
 	inode->i_ads_entries = ads_entries;
 
 	new_entry = &inode->i_ads_entries[num_ads - 1];
-	if (init_ads_entry(new_entry, stream_name, stream_name_nbytes, is_utf16le))
-		return NULL;
+
+	memset(new_entry, 0, sizeof(struct wim_ads_entry));
+	new_entry->stream_name = stream_name;
+	new_entry->stream_name_nbytes = stream_name_nbytes;
 	new_entry->stream_id = inode->i_next_stream_id++;
 	inode->i_num_ads = num_ads;
 	return new_entry;
 }
 
+/*
+ * Add an alternate data stream entry to a WIM inode (UTF-16LE version).  On
+ * success, returns a pointer to the new entry.  Note that this pointer might
+ * become invalid if another ADS entry is added to the inode.  On failure,
+ * returns NULL and sets errno.
+ */
 struct wim_ads_entry *
 inode_add_ads_utf16le(struct wim_inode *inode,
-		      const utf16lechar *stream_name,
-		      size_t stream_name_nbytes)
+		      const utf16lechar *stream_name, size_t stream_name_nbytes)
 {
-	DEBUG("Add alternate data stream \"%"WS"\"", stream_name);
-	return do_inode_add_ads(inode, stream_name, stream_name_nbytes, true);
+	utf16lechar *dup = NULL;
+	struct wim_ads_entry *result;
+
+	if (stream_name_nbytes) {
+		dup = utf16le_dupz(stream_name, stream_name_nbytes);
+		if (!dup)
+			return NULL;
+	}
+
+	result = do_inode_add_ads(inode, dup, stream_name_nbytes);
+	if (!result)
+		FREE(dup);
+	return result;
 }
 
 /*
- * Add an alternate stream entry to a WIM inode.  On success, returns a pointer
- * to the new entry; on failure, returns NULL.
- *
- * @stream_name must be a nonempty string.
+ * Add an alternate data stream entry to a WIM inode (tchar version).  On
+ * success, returns a pointer to the new entry.  Note that this pointer might
+ * become invalid if another ADS entry is added to the inode.  On failure,
+ * returns NULL and sets errno.
  */
 struct wim_ads_entry *
 inode_add_ads(struct wim_inode *inode, const tchar *stream_name)
 {
-	DEBUG("Add alternate data stream \"%"TS"\"", stream_name);
-	return do_inode_add_ads(inode, stream_name,
-				tstrlen(stream_name) * sizeof(tchar),
-				TCHAR_IS_UTF16LE);
+	utf16lechar *stream_name_utf16le = NULL;
+	size_t stream_name_utf16le_nbytes = 0;
+	struct wim_ads_entry *result;
+
+	if (stream_name && *stream_name)
+		if (tstr_to_utf16le(stream_name,
+				    tstrlen(stream_name) * sizeof(tchar),
+				    &stream_name_utf16le,
+				    &stream_name_utf16le_nbytes))
+			return NULL;
+
+	result = do_inode_add_ads(inode, stream_name_utf16le,
+				  stream_name_utf16le_nbytes);
+	if (!result)
+		FREE(stream_name_utf16le);
+	return result;
 }
 
-int
+/*
+ * Add an data alternate stream entry to a WIM inode, where the contents of the
+ * new stream are specified in a data buffer.  The inode must be resolved.
+ *
+ * On success, returns a pointer to the new alternate data stream entry.  Note
+ * that this pointer might become invalid if another ADS entry is added to the
+ * inode.  On failure, returns NULL and sets errno.
+ */
+struct wim_ads_entry *
 inode_add_ads_with_data(struct wim_inode *inode, const tchar *name,
 			const void *value, size_t size,
 			struct wim_lookup_table *lookup_table)
 {
-	struct wim_ads_entry *new_ads_entry;
+	struct wim_ads_entry *new_entry;
 
 	wimlib_assert(inode->i_resolved);
 
-	new_ads_entry = inode_add_ads(inode, name);
-	if (new_ads_entry == NULL)
-		return WIMLIB_ERR_NOMEM;
+	new_entry = inode_add_ads(inode, name);
+	if (!new_entry)
+		return NULL;
 
-	new_ads_entry->lte = new_stream_from_data_buffer(value, size,
-							 lookup_table);
-	if (new_ads_entry->lte == NULL) {
-		inode_remove_ads(inode, new_ads_entry - inode->i_ads_entries,
-				 lookup_table);
-		return WIMLIB_ERR_NOMEM;
+	new_entry->lte = new_stream_from_data_buffer(value, size, lookup_table);
+	if (!new_entry->lte) {
+		inode_remove_ads(inode, new_entry, NULL);
+		return NULL;
 	}
-	return 0;
+	return new_entry;
 }
 
+/*
+ * Does the inode have any named data streams?
+ */
 bool
 inode_has_named_stream(const struct wim_inode *inode)
 {
@@ -302,38 +314,36 @@ inode_has_named_stream(const struct wim_inode *inode)
 }
 
 /* Set the unnamed stream of a WIM inode, given a data buffer containing the
- * stream contents. */
+ * stream contents.  The inode must be resolved and cannot have an unnamed
+ * stream specified already.  */
 int
 inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len,
 			 struct wim_lookup_table *lookup_table)
 {
+	wimlib_assert(inode->i_resolved);
+	wimlib_assert(!inode->i_lte);
 	inode->i_lte = new_stream_from_data_buffer(data, len, lookup_table);
-	if (inode->i_lte == NULL)
+	if (!inode->i_lte)
 		return WIMLIB_ERR_NOMEM;
-	inode->i_resolved = 1;
 	return 0;
 }
 
-/* Remove an alternate data stream from a WIM inode  */
+/* Remove an alternate data stream from a WIM inode.  */
 void
-inode_remove_ads(struct wim_inode *inode, u16 idx,
+inode_remove_ads(struct wim_inode *inode, struct wim_ads_entry *entry,
 		 struct wim_lookup_table *lookup_table)
 {
-	struct wim_ads_entry *ads_entry;
 	struct wim_lookup_table_entry *lte;
+	unsigned idx = entry - inode->i_ads_entries;
 
 	wimlib_assert(idx < inode->i_num_ads);
 	wimlib_assert(inode->i_resolved);
 
-	ads_entry = &inode->i_ads_entries[idx];
-
-	DEBUG("Remove alternate data stream \"%"WS"\"", ads_entry->stream_name);
-
-	lte = ads_entry->lte;
+	lte = entry->lte;
 	if (lte)
 		lte_decrement_refcnt(lte, lookup_table);
 
-	destroy_ads_entry(ads_entry);
+	destroy_ads_entry(entry);
 
 	memmove(&inode->i_ads_entries[idx],
 		&inode->i_ads_entries[idx + 1],
@@ -341,185 +351,136 @@ inode_remove_ads(struct wim_inode *inode, u16 idx,
 	inode->i_num_ads--;
 }
 
-bool
-inode_has_unix_data(const struct wim_inode *inode)
-{
-	for (u16 i = 0; i < inode->i_num_ads; i++)
-		if (ads_entry_is_unix_data(&inode->i_ads_entries[i]))
-			return true;
-	return false;
-}
-
-#ifndef __WIN32__
-int
-inode_get_unix_data(const struct wim_inode *inode,
-		    struct wimlib_unix_data *unix_data,
-		    u16 *stream_idx_ret)
-{
-	const struct wim_ads_entry *ads_entry;
-	const struct wim_lookup_table_entry *lte;
-	size_t size;
-	int ret;
-
-	wimlib_assert(inode->i_resolved);
-
-	ads_entry = inode_get_ads_entry((struct wim_inode*)inode,
-					WIMLIB_UNIX_DATA_TAG, NULL);
-	if (ads_entry == NULL)
-		return NO_UNIX_DATA;
-
-	if (stream_idx_ret)
-		*stream_idx_ret = ads_entry - inode->i_ads_entries;
-
-	lte = ads_entry->lte;
-	if (lte == NULL)
-		return NO_UNIX_DATA;
-
-	size = lte->size;
-	if (size != sizeof(struct wimlib_unix_data))
-		return BAD_UNIX_DATA;
-
-	ret = read_full_stream_into_buf(lte, unix_data);
-	if (ret)
-		return ret;
-
-	if (unix_data->version != 0)
-		return BAD_UNIX_DATA;
-	return 0;
-}
-
-int
-inode_set_unix_data(struct wim_inode *inode, u16 uid, u16 gid, u16 mode,
-		    struct wim_lookup_table *lookup_table, int which)
-{
-	struct wimlib_unix_data unix_data;
-	int ret;
-	bool have_good_unix_data = false;
-	bool have_unix_data = false;
-	u16 stream_idx;
-
-	if (!(which & UNIX_DATA_CREATE)) {
-		ret = inode_get_unix_data(inode, &unix_data, &stream_idx);
-		if (ret == 0 || ret == BAD_UNIX_DATA || ret > 0)
-			have_unix_data = true;
-		if (ret == 0)
-			have_good_unix_data = true;
-	}
-	unix_data.version = 0;
-	if (which & UNIX_DATA_UID || !have_good_unix_data)
-		unix_data.uid = uid;
-	if (which & UNIX_DATA_GID || !have_good_unix_data)
-		unix_data.gid = gid;
-	if (which & UNIX_DATA_MODE || !have_good_unix_data)
-		unix_data.mode = mode;
-	ret = inode_add_ads_with_data(inode, WIMLIB_UNIX_DATA_TAG,
-				      &unix_data,
-				      sizeof(struct wimlib_unix_data),
-				      lookup_table);
-	if (ret == 0 && have_unix_data)
-		inode_remove_ads(inode, stream_idx, lookup_table);
-	return ret;
-}
-#endif /* __WIN32__  */
-
 /*
- * Resolve an inode's lookup table entries.
+ * Resolve an inode's single-instance streams.
  *
- * This replaces the SHA1 hash fields (which are used to lookup an entry in the
- * lookup table) with pointers directly to the lookup table entries.
+ * This takes each SHA-1 message digest stored in the inode or one of its ADS
+ * entries and replaces it with a pointer directly to the appropriate 'struct
+ * wim_lookup_table_entry' currently inserted into @table to represent the
+ * single-instance stream having that SHA-1 message digest.
  *
  * If @force is %false:
- *	If any needed SHA1 message digests are not found in the lookup table,
- *	WIMLIB_ERR_RESOURCE_NOT_FOUND is returned and the inode is left
- *	unmodified.
+ *	If any of the needed single-instance streams do not exist in @table,
+ *	return WIMLIB_ERR_RESOURCE_NOT_FOUND and leave the inode unmodified.
  * If @force is %true:
- *	If any needed SHA1 message digests are not found in the lookup table,
- *	new entries are allocated and inserted into the lookup table.
+ *	If any of the needed single-instance streams do not exist in @table,
+ *	allocate new entries for them and insert them into @table.  This does
+ *	not, of course, cause these streams to magically exist, but this is
+ *	needed by the code for extraction from a pipe.
+ *
+ * If the inode is already resolved, this function does nothing.
+ *
+ * Returns 0 on success; WIMLIB_ERR_NOMEM if out of memory; or
+ * WIMLIB_ERR_RESOURCE_NOT_FOUND if @force is %false and at least one
+ * single-instance stream referenced by the inode was missing.
  */
 int
 inode_resolve_streams(struct wim_inode *inode, struct wim_lookup_table *table,
 		      bool force)
 {
 	const u8 *hash;
+	struct wim_lookup_table_entry *lte, *ads_lte;
+
+	if (inode->i_resolved)
+		return 0;
+
+	struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
+
+	/* Resolve the default data stream */
+	lte = NULL;
+	hash = inode->i_hash;
+	if (!is_zero_hash(hash)) {
+		lte = lookup_stream(table, hash);
+		if (!lte) {
+			if (force) {
+				lte = new_lookup_table_entry();
+				if (!lte)
+					return WIMLIB_ERR_NOMEM;
+				copy_hash(lte->hash, hash);
+				lookup_table_insert(table, lte);
+			} else {
+				goto stream_not_found;
+			}
+		}
+	}
 
-	if (!inode->i_resolved) {
-		struct wim_lookup_table_entry *lte, *ads_lte;
+	/* Resolve the alternate data streams */
+	for (u16 i = 0; i < inode->i_num_ads; i++) {
+		struct wim_ads_entry *cur_entry;
 
-		/* Resolve the default file stream */
-		lte = NULL;
-		hash = inode->i_hash;
+		ads_lte = NULL;
+		cur_entry = &inode->i_ads_entries[i];
+		hash = cur_entry->hash;
 		if (!is_zero_hash(hash)) {
-			lte = lookup_stream(table, hash);
-			if (!lte) {
+			ads_lte = lookup_stream(table, hash);
+			if (!ads_lte) {
 				if (force) {
-					lte = new_lookup_table_entry();
-					if (!lte)
+					ads_lte = new_lookup_table_entry();
+					if (!ads_lte)
 						return WIMLIB_ERR_NOMEM;
-					copy_hash(lte->hash, hash);
-					lookup_table_insert(table, lte);
+					copy_hash(ads_lte->hash, hash);
+					lookup_table_insert(table, ads_lte);
 				} else {
 					goto stream_not_found;
 				}
 			}
 		}
-
-		/* Resolve the alternate data streams */
-		struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
-		for (u16 i = 0; i < inode->i_num_ads; i++) {
-			struct wim_ads_entry *cur_entry;
-
-			ads_lte = NULL;
-			cur_entry = &inode->i_ads_entries[i];
-			hash = cur_entry->hash;
-			if (!is_zero_hash(hash)) {
-				ads_lte = lookup_stream(table, hash);
-				if (!ads_lte) {
-					if (force) {
-						ads_lte = new_lookup_table_entry();
-						if (!ads_lte)
-							return WIMLIB_ERR_NOMEM;
-						copy_hash(ads_lte->hash, hash);
-						lookup_table_insert(table, ads_lte);
-					} else {
-						goto stream_not_found;
-					}
-				}
-			}
-			ads_ltes[i] = ads_lte;
-		}
-		inode->i_lte = lte;
-		for (u16 i = 0; i < inode->i_num_ads; i++)
-			inode->i_ads_entries[i].lte = ads_ltes[i];
-		inode->i_resolved = 1;
+		ads_ltes[i] = ads_lte;
 	}
+	inode->i_lte = lte;
+	for (u16 i = 0; i < inode->i_num_ads; i++)
+		inode->i_ads_entries[i].lte = ads_ltes[i];
+	inode->i_resolved = 1;
 	return 0;
 
 stream_not_found:
 	return stream_not_found_error(inode, hash);
 }
 
+/*
+ * Undo the effects of inode_resolve_streams().
+ *
+ * If the inode is not resolved, this function does nothing.
+ */
 void
 inode_unresolve_streams(struct wim_inode *inode)
 {
-	if (inode->i_resolved) {
-		if (inode->i_lte)
-			copy_hash(inode->i_hash, inode->i_lte->hash);
+	if (!inode->i_resolved)
+		return;
+
+	if (inode->i_lte)
+		copy_hash(inode->i_hash, inode->i_lte->hash);
+	else
+		zero_out_hash(inode->i_hash);
+
+	for (u16 i = 0; i < inode->i_num_ads; i++) {
+		if (inode->i_ads_entries[i].lte)
+			copy_hash(inode->i_ads_entries[i].hash,
+				  inode->i_ads_entries[i].lte->hash);
 		else
-			zero_out_hash(inode->i_hash);
-
-		for (u16 i = 0; i < inode->i_num_ads; i++) {
-			if (inode->i_ads_entries[i].lte)
-				copy_hash(inode->i_ads_entries[i].hash,
-					  inode->i_ads_entries[i].lte->hash);
-			else
-				zero_out_hash(inode->i_ads_entries[i].hash);
-		}
-		inode->i_resolved = 0;
+			zero_out_hash(inode->i_ads_entries[i].hash);
 	}
+	inode->i_resolved = 0;
+}
+
+int
+stream_not_found_error(const struct wim_inode *inode, const u8 *hash)
+{
+	if (wimlib_print_errors) {
+		tchar hashstr[SHA1_HASH_SIZE * 2 + 1];
+
+		sprint_hash(hash, hashstr);
+
+		ERROR("\"%"TS"\": stream not found\n"
+		      "        SHA-1 message digest of missing stream:\n"
+		      "        %"TS"",
+		      inode_first_full_path(inode), hashstr);
+	}
+	return WIMLIB_ERR_RESOURCE_NOT_FOUND;
 }
 
 /*
- * Returns the lookup table entry for stream @stream_idx of the inode, where
+ * Return the lookup table entry for stream @stream_idx of the inode, where
  * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
  * corresponds to an alternate data stream.
  *
@@ -535,6 +496,11 @@ inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
 		return inode_stream_lte_unresolved(inode, stream_idx, table);
 }
 
+/*
+ * Return the lookup table entry for the unnamed data stream of a *resolved*
+ * inode, or NULL if there is none.  Also return the 0-based index of the
+ * corresponding stream in *stream_idx_ret.
+ */
 struct wim_lookup_table_entry *
 inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret)
 {
@@ -551,6 +517,10 @@ inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret
 	return NULL;
 }
 
+/*
+ * Return the lookup table entry for the unnamed data stream of a *resolved*
+ * inode, or NULL if there is none.
+ */
 struct wim_lookup_table_entry *
 inode_unnamed_lte_resolved(const struct wim_inode *inode)
 {
@@ -558,6 +528,10 @@ inode_unnamed_lte_resolved(const struct wim_inode *inode)
 	return inode_unnamed_stream_resolved(inode, &stream_idx);
 }
 
+/*
+ * Return the lookup table entry for the unnamed data stream of an *unresolved*
+ * inode, or NULL if there is none.
+ */
 struct wim_lookup_table_entry *
 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
 			     const struct wim_lookup_table *table)
@@ -573,7 +547,8 @@ inode_unnamed_lte_unresolved(const struct wim_inode *inode,
 	return NULL;
 }
 
-/* Return the lookup table entry for the unnamed data stream of an inode, or
+/*
+ * Return the lookup table entry for the unnamed data stream of an inode, or
  * NULL if there is none.
  *
  * You'd think this would be easier than it actually is, since the unnamed data
@@ -583,7 +558,7 @@ inode_unnamed_lte_unresolved(const struct wim_inode *inode,
  * dentry itself.  So we need to check the alternate data streams too.
  *
  * Also, note that a dentry may appear to have more than one unnamed stream, but
- * if the SHA1 message digest is all 0's then the corresponding stream does not
+ * if the SHA-1 message digest is all 0's then the corresponding stream does not
  * really "count" (this is the case for the inode's own file stream when the
  * file stream that should be there is actually in one of the alternate stream
  * entries.).  This is despite the fact that we may need to extract such a
@@ -599,9 +574,12 @@ inode_unnamed_lte(const struct wim_inode *inode,
 		return inode_unnamed_lte_unresolved(inode, table);
 }
 
-/* Returns the SHA1 message digest of the unnamed data stream of a WIM inode, or
- * 'zero_hash' if the unnamed data stream is missing has all zeroes in its SHA1
- * message digest field.  */
+/*
+ * Return the SHA-1 message digest of the unnamed data stream of a WIM inode.
+ *
+ * If inode does not have an unnamed data stream, this returns a void SHA-1
+ * message digest containing all zero bytes.
+ */
 const u8 *
 inode_unnamed_stream_hash(const struct wim_inode *inode)
 {
@@ -617,8 +595,15 @@ inode_unnamed_stream_hash(const struct wim_inode *inode)
 	return zero_hash;
 }
 
-/* Given an unhashed stream, get the pointer to it in an inode.
- * As this is only for unhashed streams, there can only be one such pointer.  */
+/*
+ * Translate a single-instance stream entry into the pointer contained in the
+ * inode (or ads entry of an inode) that references it.
+ *
+ * This is only possible for "unhashed" streams, which are guaranteed to have
+ * only one reference, and that reference is guaranteed to be in a resolved
+ * inode.  (It can't be in an unresolved inode, since that would imply the hash
+ * is known!)
+ */
 struct wim_lookup_table_entry **
 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
 {
@@ -635,20 +620,8 @@ retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
 	return NULL;
 }
 
-int
-stream_not_found_error(const struct wim_inode *inode, const u8 *hash)
-{
-	if (wimlib_print_errors) {
-		ERROR("\"%"TS"\": stream not found", inode_first_full_path(inode));
-		tfprintf(stderr, T("        SHA-1 message digest of missing stream:\n        "));
-		print_hash(hash, stderr);
-		tputc(T('\n'), stderr);
-	}
-	return WIMLIB_ERR_RESOURCE_NOT_FOUND;
-}
-
 /*
- * Reads the alternate data stream entries of a WIM dentry.
+ * Read the alternate data stream entries of a WIM dentry.
  *
  * @p:
  *	Pointer to buffer that starts with the first alternate stream entry.
@@ -657,8 +630,9 @@ stream_not_found_error(const struct wim_inode *inode, const u8 *hash)
  *	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:
+ * @nbytes_remaining_p:
  *	Number of bytes of data remaining in the buffer pointed to by @p.
+ *	On success this will be updated to point just past the ADS entries.
  *
  * On success, inode->i_ads_entries is set to an array of `struct
  * wim_ads_entry's of length inode->i_num_ads.  On failure, @inode is not
@@ -671,8 +645,9 @@ stream_not_found_error(const struct wim_inode *inode, const u8 *hash)
  */
 int
 read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode,
-		 size_t nbytes_remaining)
+		 size_t *nbytes_remaining_p)
 {
+	size_t nbytes_remaining = *nbytes_remaining_p;
 	u16 num_ads;
 	struct wim_ads_entry *ads_entries;
 	int ret;
@@ -683,7 +658,7 @@ read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode,
 	 * data stream entries. */
 	num_ads = inode->i_num_ads;
 	ads_entries = CALLOC(num_ads, sizeof(inode->i_ads_entries[0]));
-	if (ads_entries == NULL)
+	if (!ads_entries)
 		goto out_of_memory;
 
 	/* Read the entries into our newly allocated buffer. */
@@ -737,14 +712,10 @@ read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode,
 			    cur_entry->stream_name_nbytes > length)
 				goto out_invalid;
 
-			cur_entry->stream_name = MALLOC(cur_entry->stream_name_nbytes + 2);
-			if (cur_entry->stream_name == NULL)
+			cur_entry->stream_name = utf16le_dupz(disk_entry->stream_name,
+							      cur_entry->stream_name_nbytes);
+			if (!cur_entry->stream_name)
 				goto out_of_memory;
-
-			memcpy(cur_entry->stream_name,
-			       disk_entry->stream_name,
-			       cur_entry->stream_name_nbytes);
-			cur_entry->stream_name[cur_entry->stream_name_nbytes / 2] = cpu_to_le16(0);
 		} else {
 			/* Mark inode as having weird stream entries.  */
 			inode->i_canonical_streams = 0;
@@ -758,7 +729,7 @@ read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode,
 		 * that less than @length is actually remaining in the metadata
 		 * resource. We should set the remaining bytes to 0 if this
 		 * happens. */
-		length = (length + 7) & ~(u64)7;
+		length = (length + 7) & ~7;
 		p += length;
 		if (nbytes_remaining < length)
 			nbytes_remaining = 0;
@@ -767,6 +738,7 @@ read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode,
 	}
 	inode->i_ads_entries = ads_entries;
 	inode->i_next_stream_id = inode->i_num_ads + 1;
+	*nbytes_remaining_p = nbytes_remaining;
 	ret = 0;
 	goto out;
 out_of_memory:
@@ -785,21 +757,10 @@ out:
 	return ret;
 }
 
-/*
- * Verify a WIM inode:
- *
- * - Check to make sure the security ID is valid
- * - Check to make sure there is at most one unnamed stream
- * - Check to make sure there is at most one DOS name.
- *
- * Return values:
- *	WIMLIB_ERR_SUCCESS (0)
- */
-int
-verify_inode(struct wim_inode *inode, const struct wim_security_data *sd)
+/* Check a WIM inode for unusual field values.  */
+void
+check_inode(struct wim_inode *inode, const struct wim_security_data *sd)
 {
-	struct wim_dentry *dentry;
-
 	/* Check the security ID.  -1 is valid and means "no security
 	 * descriptor".  Anything else has to be a valid index into the WIM
 	 * image's security descriptors table. */
@@ -823,36 +784,56 @@ verify_inode(struct wim_inode *inode, const struct wim_security_data *sd)
 	if (num_unnamed_streams > 1) {
 		WARNING("\"%"TS"\" has multiple (%u) un-named streams",
 			inode_first_full_path(inode), num_unnamed_streams);
+		/* We currently don't treat this as an error and will just end
+		 * up using the first unnamed data stream in the inode.  */
 	}
-
-	return 0;
 }
 
+/* Acquire another reference to each single-instance stream referenced by this
+ * inode.  This is necessary when creating a hard link to this inode.
+ *
+ * The inode must be resolved.  */
 void
 inode_ref_streams(struct wim_inode *inode)
 {
 	for (unsigned i = 0; i <= inode->i_num_ads; i++) {
 		struct wim_lookup_table_entry *lte;
+
 		lte = inode_stream_lte_resolved(inode, i);
 		if (lte)
 			lte->refcnt++;
 	}
 }
 
+/* Drop a reference to each single-instance stream referenced by this inode.
+ * This is necessary when deleting a hard link to this inode.  */
+void
+inode_unref_streams(struct wim_inode *inode,
+		    struct wim_lookup_table *lookup_table)
+{
+	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);
+	}
+}
+
+/* Initialize a hash table for hard link detection.  */
 int
 init_inode_table(struct wim_inode_table *table, size_t capacity)
 {
 	table->array = CALLOC(capacity, sizeof(table->array[0]));
-	if (table->array == NULL) {
-		ERROR("Cannot initalize inode table: out of memory");
+	if (!table->array)
 		return WIMLIB_ERR_NOMEM;
-	}
 	table->num_entries  = 0;
 	table->capacity  = capacity;
 	INIT_LIST_HEAD(&table->extra_inodes);
 	return 0;
 }
 
+/* Free the memory allocated by init_inode_table().  */
 void
 destroy_inode_table(struct wim_inode_table *table)
 {
@@ -867,14 +848,17 @@ inode_table_get_inode(struct wim_inode_table *table, u64 ino, u64 devno)
 	struct wim_inode *inode;
 	struct hlist_node *cur;
 
+	/* Search for an existing inode having the same inode number and device
+	 * number.  */
 	hlist_for_each_entry(inode, cur, &table->array[pos], i_hlist) {
 		if (inode->i_ino == ino && inode->i_devno == devno) {
-			DEBUG("Using existing inode {devno=%"PRIu64", ino=%"PRIu64"}",
-			      devno, ino);
+			/* Found; use the existing inode.  */
 			inode->i_nlink++;
 			return inode;
 		}
 	}
+
+	/* Create a new inode and insert it into the table.  */
 	inode = new_timeless_inode();
 	if (inode) {
 		inode->i_ino = ino;
@@ -885,13 +869,39 @@ inode_table_get_inode(struct wim_inode_table *table, u64 ino, u64 devno)
 	return inode;
 }
 
-
-/* Given a directory entry with the name @name for the file with the inode
- * number @ino and device number @devno, create a new WIM dentry with an
- * associated inode, where the inode is shared if an inode with the same @ino
- * and @devno has already been created.  On success, the new WIM dentry is
- * written to *dentry_ret, and its inode has i_nlink > 1 if a previously
- * existing inode was used.
+/*
+ * Allocate a new dentry, with hard link detection.
+ *
+ * @table
+ *	The inode table being used for the current directory scan operation.  It
+ *	will contain the mapping from (ino, devno) pairs to inodes.
+ *
+ * @name
+ *	The name to give the new dentry.
+ *
+ * @ino
+ *	The inode number of the file, read from the filesystem.
+ *
+ * @devno
+ *	The device number of the file, read from the filesystem.  Proper setting
+ *	of this parameter prevents cross-device hardlinks from being created.
+ *	If this is not a problem (perhaps because the current directory scan
+ *	operation is guaranteed to never traverse a filesystem boundary), then
+ *	this parameter can just be a fixed value such as 0.
+ *
+ * @noshare
+ *	If %true, the new dentry will not be hard linked to any existing inode,
+ *	regardless of the values of @ino and @devno.  If %false, normal hard
+ *	link detection will be done.
+ *
+ * @dentry_ret
+ *	On success, a pointer to the new dentry will be returned in this
+ *	location.  If i_nlink of the dentry's inode is greater than 1, then this
+ *	function created a hard link to an existing inode rather than creating a
+ *	new inode.
+ *
+ * On success, returns 0.  On failure, returns WIMLIB_ERR_NOMEM or an error code
+ * resulting from a failed string conversion.
  */
 int
 inode_table_new_dentry(struct wim_inode_table *table, const tchar *name,
@@ -934,9 +944,12 @@ inode_table_new_dentry(struct wim_inode_table *table, const tchar *name,
 
 
 
-/* Assign consecutive inode numbers to a new set of inodes from the inode table,
- * and append the inodes to a single list @head that contains the inodes already
- * existing in the WIM image.  */
+/*
+ * Following the allocation of dentries with hard link detection using
+ * inode_table_new_dentry(), this function will assign consecutive inode numbers
+ * to the new set of inodes.  It will also append the list of new inodes to the
+ * list @head, which must contain any inodes already existing in the WIM image.
+ */
 void
 inode_table_prepare_inode_list(struct wim_inode_table *table,
 			       struct list_head *head)