From: Eric Biggers <ebiggers3@gmail.com>
Date: Tue, 28 Aug 2012 05:24:28 +0000 (-0500)
Subject: dentry_first_lte() -> dentry_unnamed_lte()
X-Git-Tag: v1.0.0~57
X-Git-Url: https://wimlib.net/git/?p=wimlib;a=commitdiff_plain;h=c6127ff151b2963c5b90e4018f08f53051a97dfd

dentry_first_lte() -> dentry_unnamed_lte()

This probably doesn't make a difference, but when we were looking for the
"first" lookup table entry what we really wanted was to find the lookup table
for the un-named data stream, even if it's in an alternate data stream rather
	than the expected location in the dentry.
---

diff --git a/src/extract.c b/src/extract.c
index ec6842dc..b27c4432 100644
--- a/src/extract.c
+++ b/src/extract.c
@@ -200,7 +200,7 @@ static int extract_regular_file(WIMStruct *w,
 {
 	struct lookup_table_entry *lte;
 
-	lte = dentry_first_lte(dentry, w->lookup_table);
+	lte = dentry_unnamed_lte(dentry, w->lookup_table);
 
 	if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
 			      WIMLIB_EXTRACT_FLAG_HARDLINK)) && lte) {
diff --git a/src/lookup_table.c b/src/lookup_table.c
index d15a3088..92fb8a9f 100644
--- a/src/lookup_table.c
+++ b/src/lookup_table.c
@@ -221,6 +221,14 @@ int read_lookup_table(WIMStruct *w)
 		p = get_u32(p, &cur_entry->refcnt);
 		p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
 
+		if (is_zero_hash(cur_entry->hash)) {
+			ERROR("The WIM lookup table contains an entry with a "
+			      "SHA1 message digest of all 0's");
+			ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
+			FREE(cur_entry);
+			goto out;
+		}
+
 		duplicate_entry = __lookup_resource(table, cur_entry->hash);
 		if (duplicate_entry) {
 			ERROR("The WIM lookup table contains two entries with the "
@@ -230,6 +238,7 @@ int read_lookup_table(WIMStruct *w)
 			ERROR("The second entry is:");
 			print_lookup_table_entry(cur_entry);
 			ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
+			FREE(cur_entry);
 			goto out;
 		}
 		lookup_table_insert(table, cur_entry);
@@ -481,12 +490,29 @@ int dentry_resolve_ltes(struct dentry *dentry, void *__table)
 	return 0;
 }
 
+/* Return the lookup table entry for the unnamed data stream of a dentry, or
+ * NULL if there is none.
+ *
+ * You'd think this would be easier than it actually is, since the unnamed data
+ * stream should be the one referenced from the dentry itself.  Alas, if there
+ * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
+ * data stream in one of the alternate data streams instead of inside the
+ * dentry.  So we need to check the alternate data streams too.
+ *
+ * Also, note that a dentry may appear to have than one unnamed stream, but if
+ * the SHA1 message digest is all 0's then the corresponding stream does not
+ * really "count" (this is the case for the dentry'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
+ * missing entry as an empty file or empty named data stream.
+ */
 struct lookup_table_entry *
-dentry_first_lte(const struct dentry *dentry, const struct lookup_table *table)
+dentry_unnamed_lte(const struct dentry *dentry,
+		   const struct lookup_table *table)
 {
 	if (dentry->resolved)
-		return dentry_first_lte_resolved(dentry);
+		return dentry_unnamed_lte_resolved(dentry);
 	else
-		return dentry_first_lte_unresolved(dentry, table);
+		return dentry_unnamed_lte_unresolved(dentry, table);
 }
 
diff --git a/src/lookup_table.h b/src/lookup_table.h
index 43494efe..3a653bfb 100644
--- a/src/lookup_table.h
+++ b/src/lookup_table.h
@@ -299,7 +299,7 @@ static inline u16 dentry_stream_name_len(const struct dentry *dentry,
 {
 	wimlib_assert(stream_idx <= dentry->num_ads);
 	if (stream_idx == 0)
-		return dentry->file_name_len;
+		return 0;
 	else
 		return dentry->ads_entries[stream_idx - 1].stream_name_len;
 }
@@ -332,35 +332,34 @@ static inline const u8 *dentry_stream_hash(const struct dentry *dentry,
 }
 
 static inline struct lookup_table_entry *
-dentry_first_lte_resolved(const struct dentry *dentry)
+dentry_unnamed_lte_resolved(const struct dentry *dentry)
 {
 	struct lookup_table_entry *lte;
 	wimlib_assert(dentry->resolved);
 
-	for (unsigned i = 0; i <= dentry->num_ads; i++) {
-		lte = dentry_stream_lte_resolved(dentry, i);
-		if (lte)
-			return lte;
-	}
+	for (unsigned i = 0; i <= dentry->num_ads; i++)
+		if (dentry_stream_name_len(dentry, i) == 0 &&
+		     !is_zero_hash(dentry_stream_hash_resolved(dentry, i)))
+			return dentry_stream_lte_resolved(dentry, i);
 	return NULL;
 }
 
 static inline struct lookup_table_entry *
-dentry_first_lte_unresolved(const struct dentry *dentry,
-			    const struct lookup_table *table)
+dentry_unnamed_lte_unresolved(const struct dentry *dentry,
+			      const struct lookup_table *table)
 {
 	struct lookup_table_entry *lte;
 	wimlib_assert(!dentry->resolved);
 
-	for (unsigned i = 0; i <= dentry->num_ads; i++) {
-		lte = dentry_stream_lte_unresolved(dentry, i, table);
-		if (lte)
-			return lte;
-	}
+	for (unsigned i = 0; i <= dentry->num_ads; i++)
+		if (dentry_stream_name_len(dentry, i) == 0 &&
+		     !is_zero_hash(dentry_stream_hash_unresolved(dentry, i)))
+			return dentry_stream_lte_unresolved(dentry, i, table);
 	return NULL;
 }
 
 extern struct lookup_table_entry *
-dentry_first_lte(const struct dentry *dentry, const struct lookup_table *table);
+dentry_unnamed_lte(const struct dentry *dentry,
+		   const struct lookup_table *table);
 
 #endif
diff --git a/src/mount.c b/src/mount.c
index 1510fbfb..f6c178b4 100644
--- a/src/mount.c
+++ b/src/mount.c
@@ -221,7 +221,7 @@ int dentry_to_stbuf(const struct dentry *dentry, struct stat *stbuf)
 	stbuf->st_gid   = getgid();
 
 	/* Use the size of the unnamed (default) file stream. */
-	lte = dentry_first_lte_resolved(dentry);
+	lte = dentry_unnamed_lte_resolved(dentry);
 	if (lte) {
 		if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
 			wimlib_assert(lte->staging_file_name);
diff --git a/src/ntfs-apply.c b/src/ntfs-apply.c
index 4118a30a..3ec0ae75 100644
--- a/src/ntfs-apply.c
+++ b/src/ntfs-apply.c
@@ -268,7 +268,7 @@ static int apply_reparse_data(ntfs_inode *ni, const struct dentry *dentry,
 
 	wimlib_assert(dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT);
 
-	lte = dentry_first_lte(dentry, w->lookup_table);
+	lte = dentry_unnamed_lte(dentry, w->lookup_table);
 
 	DEBUG("Applying reparse data to `%s'", dentry->full_path_utf8);
 
diff --git a/src/resource.c b/src/resource.c
index 3c573f1c..0c7a4cf9 100644
--- a/src/resource.c
+++ b/src/resource.c
@@ -882,7 +882,9 @@ static int write_wim_resource(struct lookup_table_entry *lte,
 		}
 	}
 
-	if (new_compressed_size > original_size) {
+	if (new_compressed_size >= original_size &&
+	    out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw)
+	{
 		/* Oops!  We compressed the resource to larger than the original
 		 * size.  Write the resource uncompressed instead. */
 		if (fseeko(out_fp, file_offset, SEEK_SET) != 0) {
diff --git a/src/sha1.h b/src/sha1.h
index a5d695d6..e153f533 100644
--- a/src/sha1.h
+++ b/src/sha1.h
@@ -39,9 +39,10 @@ static inline void print_hash(const u8 hash[SHA1_HASH_SIZE])
 
 static inline bool is_zero_hash(const u8 hash[SHA1_HASH_SIZE])
 {
-	for (u8 i = 0; i < SHA1_HASH_SIZE / 4; i++)
-		if (((u32*)hash)[i])
-			return false;
+	if (hash)
+		for (u8 i = 0; i < SHA1_HASH_SIZE / 4; i++)
+			if (((u32*)hash)[i])
+				return false;
 	return true;
 }
 
diff --git a/src/symlink.c b/src/symlink.c
index ca2e422f..3773fb05 100644
--- a/src/symlink.c
+++ b/src/symlink.c
@@ -168,7 +168,7 @@ ssize_t dentry_readlink(const struct dentry *dentry, char *buf, size_t buf_len,
 
 	wimlib_assert(dentry_is_symlink(dentry));
 
-	lte = dentry_first_lte(dentry, w->lookup_table);
+	lte = dentry_unnamed_lte(dentry, w->lookup_table);
 	if (!lte)
 		return -EIO;