busybox-1.27.2: Fix CVE-2011-5325

[No upstream tracking] -- https://bugs.busybox.net/show_bug.cgi?id=8411

libarchive: do not extract unsafe symlinks

Prevent unsafe links extracting unless env variable $EXTRACT_UNSAFE_SYMLINKS=1
is not set. Untarring file with -C DESTDIR parameter could be extracted with
unwanted symlinks. This doesn't feel right, and IIRC GNU tar doesn't do that.
Include necessary changes from previous commits.

Upstream-Status: Backport [https://git.busybox.net/busybox/commit/?id=bc9bbeb2b81001e8731cd2ae501c8fccc8d87cc7]
CVE: CVE-2011-5325
bug: 8411
Signed-off-by: Radovan Scasny <radovan.scasny@siemens.com>
Signed-off-by: Andrej Valek <andrej.valek@siemens.com>

diff --git a/archival/libarchive/Kbuild.src b/archival/libarchive/Kbuild.src
index 942e755..e1a8a75 100644
--- a/archival/libarchive/Kbuild.src
+++ b/archival/libarchive/Kbuild.src
@@ -12,6 +12,8 @@ COMMON_FILES:= \
 	data_extract_all.o \
 	data_extract_to_stdout.o \
 \
+	unsafe_symlink_target.o \
+\
 	filter_accept_all.o \
 	filter_accept_list.o \
 	filter_accept_reject_list.o \
diff --git a/archival/libarchive/data_extract_all.c b/archival/libarchive/data_extract_all.c
index 1830ffb..b828b65 100644
--- a/archival/libarchive/data_extract_all.c
+++ b/archival/libarchive/data_extract_all.c
@@ -128,10 +128,9 @@ void FAST_FUNC data_extract_all(archive_handle_t *archive_handle)
 		res = link(hard_link, dst_name);
 		if (res != 0 && !(archive_handle->ah_flags & ARCHIVE_EXTRACT_QUIET)) {
 			/* shared message */
-			bb_perror_msg("can't create %slink "
-					"%s to %s", "hard",
-					dst_name,
-					hard_link);
+			bb_perror_msg("can't create %slink '%s' to '%s'",
+					 "hard", dst_name, hard_link
+			);
 		}
 		/* Hardlinks have no separate mode/ownership, skip chown/chmod */
 		goto ret;
@@ -178,15 +177,17 @@ void FAST_FUNC data_extract_all(archive_handle_t *archive_handle)
 	case S_IFLNK:
 		/* Symlink */
 //TODO: what if file_header->link_target == NULL (say, corrupted tarball?)
-		res = symlink(file_header->link_target, dst_name);
-		if (res != 0
-		 && !(archive_handle->ah_flags & ARCHIVE_EXTRACT_QUIET)
-		) {
-			/* shared message */
-			bb_perror_msg("can't create %slink "
-				"%s to %s", "sym",
-				dst_name,
-				file_header->link_target);
+		if (!unsafe_symlink_target(file_header->link_target)) {
+			res = symlink(file_header->link_target, dst_name);
+			if (res != 0
+				&& !(archive_handle->ah_flags & ARCHIVE_EXTRACT_QUIET)
+			) {
+						/* shared message */
+						bb_perror_msg("can't create %slink '%s' to '%s'",
+							"sym",
+							dst_name, file_header->link_target
+						);
+			}
 		}
 		break;
 	case S_IFSOCK:
diff --git a/archival/libarchive/unsafe_symlink_target.c b/archival/libarchive/unsafe_symlink_target.c
new file mode 100644
index 0000000..ee46e28
--- /dev/null
+++ b/archival/libarchive/unsafe_symlink_target.c
@@ -0,0 +1,48 @@
+/* vi: set sw=4 ts=4: */
+/*
+ * Licensed under GPLv2 or later, see file LICENSE in this source tree.
+ */
+#include "libbb.h"
+#include "bb_archive.h"
+
+int FAST_FUNC unsafe_symlink_target(const char *target)
+{
+	const char *dot;
+
+	if (target[0] == '/') {
+		const char *var;
+unsafe:
+		var = getenv("EXTRACT_UNSAFE_SYMLINKS");
+		if (var) {
+			if (LONE_CHAR(var, '1'))
+				return 0; /* pretend it's safe */
+			return 1; /* "UNSAFE!" */
+		}
+		bb_error_msg("skipping unsafe symlink to '%s' in archive,"
+			" set %s=1 to extract",
+			target,
+			"EXTRACT_UNSAFE_SYMLINKS"
+		);
+		/* Prevent further messages */
+		setenv("EXTRACT_UNSAFE_SYMLINKS", "0", 0);
+		return 1; /* "UNSAFE!" */
+	}
+
+	dot = target;
+	for (;;) {
+		dot = strchr(dot, '.');
+			if (!dot)
+				return 0; /* safe target */
+
+			/* Is it a path component starting with ".."? */
+			if ((dot[1] == '.')
+				&& (dot == target || dot[-1] == '/')
+					/* Is it exactly ".."? */
+				&& (dot[2] == '/' || dot[2] == '\0')
+			) {
+				goto unsafe;
+			}
+			/* NB: it can even be trailing ".", should only add 1 */
+			dot += 1;
+	}
+}
\ No newline at end of file
diff --git a/archival/unzip.c b/archival/unzip.c
index 9037262..270e261 100644
--- a/archival/unzip.c
+++ b/archival/unzip.c
@@ -335,6 +335,44 @@ static void unzip_create_leading_dirs(const char *fn)
 	free(name);
 }
 
+static void unzip_extract_symlink(zip_header_t *zip, const char *dst_fn)
+{
+	char *target;
+
+	if (zip->fmt.ucmpsize > 0xfff) /* no funny business please */
+		bb_error_msg_and_die("bad archive");
+
+	if (zip->fmt.method == 0) {
+		/* Method 0 - stored (not compressed) */
+		target = xzalloc(zip->fmt.ucmpsize + 1);
+		xread(zip_fd, target, zip->fmt.ucmpsize);
+	} else {
+#if 1
+		bb_error_msg_and_die("compressed symlink is not supported");
+#else
+		transformer_state_t xstate;
+		init_transformer_state(&xstate);
+		xstate.mem_output_size_max = zip->fmt.ucmpsize;
+		/* ...unpack... */
+		if (!xstate.mem_output_buf)
+			WTF();
+		target = xstate.mem_output_buf;
+		target = xrealloc(target, xstate.mem_output_size + 1);
+		target[xstate.mem_output_size] = '\0';
+#endif
+	}
+	if (!unsafe_symlink_target(target)) {
+//TODO: libbb candidate
+		if (symlink(target, dst_fn)) {
+			/* shared message */
+			bb_perror_msg_and_die("can't create %slink '%s' to '%s'",
+				"sym", dst_fn, target
+			);
+		}
+	}
+	free(target);
+}
+
 static void unzip_extract(zip_header_t *zip, int dst_fd)
 {
 	transformer_state_t xstate;
@@ -813,7 +851,7 @@ int unzip_main(int argc, char **argv)
 		}
  check_file:
 		/* Extract file */
-		if (stat(dst_fn, &stat_buf) == -1) {
+		if (lstat(dst_fn, &stat_buf) == -1) {
 			/* File does not exist */
 			if (errno != ENOENT) {
 				bb_perror_msg_and_die("can't stat '%s'", dst_fn);
@@ -834,6 +872,7 @@ int unzip_main(int argc, char **argv)
 			goto do_open_and_extract;
 		printf("replace %s? [y]es, [n]o, [A]ll, [N]one, [r]ename: ", dst_fn);
 		my_fgets80(key_buf);
+//TODO: redo lstat + ISREG check! user input could have taken a long time!
 
 		switch (key_buf[0]) {
 		case 'A':
@@ -842,7 +881,8 @@ int unzip_main(int argc, char **argv)
  do_open_and_extract:
 			unzip_create_leading_dirs(dst_fn);
 #if ENABLE_FEATURE_UNZIP_CDF
-			dst_fd = xopen3(dst_fn, O_WRONLY | O_CREAT | O_TRUNC, file_mode);
+			if (!S_ISLNK(file_mode))
+				dst_fd = xopen3(dst_fn, O_WRONLY | O_CREAT | O_TRUNC, file_mode);
 #else
 			dst_fd = xopen(dst_fn, O_WRONLY | O_CREAT | O_TRUNC);
 #endif
@@ -852,10 +892,18 @@ int unzip_main(int argc, char **argv)
 					? " extracting: %s\n"
 					: */ "  inflating: %s\n", dst_fn);
 			}
-			unzip_extract(&zip, dst_fd);
-			if (dst_fd != STDOUT_FILENO) {
-				/* closing STDOUT is potentially bad for future business */
-				close(dst_fd);
+#if ENABLE_FEATURE_UNZIP_CDF
+			if (S_ISLNK(file_mode)) {
+				if (dst_fd != STDOUT_FILENO) /* no -p */
+					unzip_extract_symlink(&zip, dst_fn);
+			} else
+#endif
+			{
+				unzip_extract(&zip, dst_fd);
+				if (dst_fd != STDOUT_FILENO) {
+					/* closing STDOUT is potentially bad for future business */
+					close(dst_fd);
+				};
 			}
 			break;
 
diff --git a/coreutils/link.c b/coreutils/link.c
index ac3ef85..aab249d 100644
--- a/coreutils/link.c
+++ b/coreutils/link.c
@@ -32,9 +32,8 @@ int link_main(int argc UNUSED_PARAM, char **argv)
 	argv += optind;
 	if (link(argv[0], argv[1]) != 0) {
 		/* shared message */
-		bb_perror_msg_and_die("can't create %slink "
-					"%s to %s", "hard",
-					argv[1], argv[0]
+		bb_perror_msg_and_die("can't create %slink '%s' to '%s'",
+					"hard",	argv[1], argv[0]
 		);
 	}
 	return EXIT_SUCCESS;
diff --git a/include/bb_archive.h b/include/bb_archive.h
index 2b9c5f0..1e4da3c 100644
--- a/include/bb_archive.h
+++ b/include/bb_archive.h
@@ -196,6 +196,7 @@ void seek_by_jump(int fd, off_t amount) FAST_FUNC;
 void seek_by_read(int fd, off_t amount) FAST_FUNC;
 
 const char *strip_unsafe_prefix(const char *str) FAST_FUNC;
+int unsafe_symlink_target(const char *target) FAST_FUNC;
 
 void data_align(archive_handle_t *archive_handle, unsigned boundary) FAST_FUNC;
 const llist_t *find_list_entry(const llist_t *list, const char *filename) FAST_FUNC;
diff --git a/libbb/copy_file.c b/libbb/copy_file.c
index 23c0f83..be90066 100644
--- a/libbb/copy_file.c
+++ b/libbb/copy_file.c
@@ -371,7 +371,10 @@ int FAST_FUNC copy_file(const char *source, const char *dest, int flags)
 			int r = symlink(lpath, dest);
 			free(lpath);
 			if (r < 0) {
-				bb_perror_msg("can't create symlink '%s'", dest);
+				/* shared message */
+				bb_perror_msg("can't create %slink '%s' to '%s'",
+					"sym", dest, lpath
+				);
 				return -1;
 			}
 			if (flags & FILEUTILS_PRESERVE_STATUS)
diff --git a/testsuite/tar.tests b/testsuite/tar.tests
index 9f7ce15..b7cd74c 100755
--- a/testsuite/tar.tests
+++ b/testsuite/tar.tests
@@ -10,9 +10,6 @@ unset LC_COLLATE
 unset LC_ALL
 umask 022
 
-rm -rf tar.tempdir 2>/dev/null
-mkdir tar.tempdir && cd tar.tempdir || exit 1
-
 # testing "test name" "script" "expected result" "file input" "stdin"
 
 testing "Empty file is not a tarball" '\
@@ -53,6 +50,7 @@ dd if=/dev/zero bs=512 count=20 2>/dev/null | tar xvf - 2>&1; echo $?
 "" ""
 SKIP=
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 # "tar cf test.tar input input_dir/ input_hard1 input_hard2 input_hard1 input_dir/ input":
 # GNU tar 1.26 records as hardlinks:
 #  input_hard2 -> input_hard1
@@ -64,7 +62,6 @@ SKIP=
 # We also don't use "hrw-r--r--" notation for hardlinks in "tar tv" listing.
 optional FEATURE_TAR_CREATE FEATURE_LS_SORTFILES
 testing "tar hardlinks and repeated files" '\
-rm -rf input_* test.tar 2>/dev/null
 >input_hard1
 ln input_hard1 input_hard2
 mkdir input_dir
@@ -95,10 +92,11 @@ drwxr-xr-x input_dir
 " \
 "" ""
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 optional FEATURE_TAR_CREATE FEATURE_LS_SORTFILES
 testing "tar hardlinks mode" '\
-rm -rf input_* test.tar 2>/dev/null
 >input_hard1
 chmod 741 input_hard1
 ln input_hard1 input_hard2
@@ -128,10 +126,11 @@ Ok: 0
 " \
 "" ""
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 optional FEATURE_TAR_CREATE FEATURE_LS_SORTFILES
 testing "tar symlinks mode" '\
-rm -rf input_* test.tar 2>/dev/null
 >input_file
 chmod 741 input_file
 ln -s input_file input_soft
@@ -159,10 +158,11 @@ lrwxrwxrwx input_file
 " \
 "" ""
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 optional FEATURE_TAR_CREATE FEATURE_TAR_LONG_OPTIONS
 testing "tar --overwrite" "\
-rm -rf input_* test.tar 2>/dev/null
 ln input input_hard
 tar cf test.tar input_hard
 echo WRONG >input
@@ -174,12 +174,13 @@ Ok
 " \
 "Ok\n" ""
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 test x"$SKIP_KNOWN_BUGS" = x"" && {
 # Needs to be run under non-root for meaningful test
 optional FEATURE_TAR_CREATE
 testing "tar writing into read-only dir" '\
-rm -rf input_* test.tar 2>/dev/null
 mkdir input_dir
 >input_dir/input_file
 chmod 550 input_dir
@@ -201,7 +202,9 @@ dr-xr-x--- input_dir
 "" ""
 SKIP=
 }
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 # Had a bug where on extract autodetect first "switched off" -z
 # and then failed to recognize .tgz extension
 optional FEATURE_TAR_CREATE FEATURE_SEAMLESS_GZ GUNZIP
@@ -217,7 +220,9 @@ Ok
 " \
 "" ""
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 # Do we detect XZ-compressed data (even w/o .tar.xz or txz extension)?
 # (the uuencoded hello_world.txz contains one empty file named "hello_world")
 optional UUDECODE FEATURE_TAR_AUTODETECT FEATURE_SEAMLESS_XZ
@@ -236,7 +241,9 @@ AAAEWVo=
 ====
 "
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 # On extract, everything up to and including last ".." component is stripped
 optional FEATURE_TAR_CREATE
 testing "tar strips /../ on extract" "\
@@ -255,7 +262,9 @@ Ok
 " \
 "" ""
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 # attack.tar.bz2 has symlink pointing to a system file
 # followed by a regular file with the same name
 # containing "root::0:0::/root:/bin/sh":
@@ -270,6 +279,7 @@ optional UUDECODE FEATURE_TAR_AUTODETECT FEATURE_SEAMLESS_BZ2
 testing "tar does not extract into symlinks" "\
 >>/tmp/passwd && uudecode -o input && tar xf input 2>&1 && rm passwd; cat /tmp/passwd; echo \$?
 " "\
+tar: skipping unsafe symlink to '/tmp/passwd' in archive, set EXTRACT_UNSAFE_SYMLINKS=1 to extract
 0
 " \
 "" "\
@@ -281,12 +291,15 @@ l4/V8LDoe90yiWJhOJvIypgEfxdyRThQkBVn/bI=
 ====
 "
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
+
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 # And same with -k
 optional UUDECODE FEATURE_TAR_AUTODETECT FEATURE_SEAMLESS_BZ2
 testing "tar -k does not extract into symlinks" "\
 >>/tmp/passwd && uudecode -o input && tar xf input -k 2>&1 && rm passwd; cat /tmp/passwd; echo \$?
 " "\
-tar: can't open 'passwd': File exists
+tar: skipping unsafe symlink to '/tmp/passwd' in archive, set EXTRACT_UNSAFE_SYMLINKS=1 to extract
 0
 " \
 "" "\
@@ -298,7 +311,9 @@ l4/V8LDoe90yiWJhOJvIypgEfxdyRThQkBVn/bI=
 ====
 "
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
+mkdir tar.tempdir && cd tar.tempdir || exit 1
 optional UNICODE_SUPPORT FEATURE_TAR_GNU_EXTENSIONS FEATURE_SEAMLESS_BZ2 FEATURE_TAR_AUTODETECT
 testing "Pax-encoded UTF8 names and symlinks" '\
 tar xvf ../tar.utf8.tar.bz2 2>&1; echo $?
@@ -309,17 +324,45 @@ rm -rf etc usr
 ' "\
 etc/ssl/certs/3b2716e5.0
 etc/ssl/certs/EBG_Elektronik_Sertifika_Hizmet_Sağlayıcısı.pem
+tar: skipping unsafe symlink to '/usr/share/ca-certificates/mozilla/EBG_Elektronik_Sertifika_Hizmet_Sağlayıcısı.crt' in archive, set EXTRACT_UNSAFE_SYMLINKS=1 to extract
 etc/ssl/certs/f80cc7f6.0
 usr/share/ca-certificates/mozilla/EBG_Elektronik_Sertifika_Hizmet_Sağlayıcısı.crt
 0
 etc/ssl/certs/3b2716e5.0 -> EBG_Elektronik_Sertifika_Hizmet_Sağlayıcısı.pem
-etc/ssl/certs/EBG_Elektronik_Sertifika_Hizmet_Sağlayıcısı.pem -> /usr/share/ca-certificates/mozilla/EBG_Elektronik_Sertifika_Hizmet_Sağlayıcısı.crt
 etc/ssl/certs/f80cc7f6.0 -> EBG_Elektronik_Sertifika_Hizmet_Sağlayıcısı.pem
 " \
 "" ""
 SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
-
-cd .. && rm -rf tar.tempdir || exit 1
+mkdir tar.tempdir && cd tar.tempdir || exit 1
+optional UUDECODE FEATURE_SEAMLESS_BZ2 FEATURE_TAR_AUTODETECT
+testing "Symlink attack: create symlink and then write through it" '\
+exec 2>&1
+uudecode -o input && tar xvf input; echo $?
+ls /tmp/bb_test_evilfile
+ls bb_test_evilfile
+ls symlink/bb_test_evilfile
+' "\
+anything.txt
+symlink
+tar: skipping unsafe symlink to '/tmp' in archive, set EXTRACT_UNSAFE_SYMLINKS=1 to extract
+symlink/bb_test_evilfile
+0
+ls: /tmp/bb_test_evilfile: No such file or directory
+ls: bb_test_evilfile: No such file or directory
+symlink/bb_test_evilfile
+" \
+"" "\
+begin-base64 644 tar_symlink_attack.tar.bz2
+QlpoOTFBWSZTWZgs7bQAALT/hMmQAFBAAf+AEMAGJPPv32AAAIAIMAC5thlR
+omAjAmCMADQT1BqNE0AEwAAjAEwElTKeo9NTR6h6gaeoA0DQNLVdwZZ5iNTk
+AQwCAV6S00QFJYhrlfFkVCEDEGtgNVqYrI0uK3ggnt30gqk4e1TTQm5QIAKa
+SJqzRGSFLMmOloHSAcvLiFxxRiQtQZF+qPxbo173ZDISOAoNoPN4PQPhBhKS
+n8fYaKlioCTzL2oXYczyUUIP4u5IpwoSEwWdtoA=
+====
+"
+SKIP=
+cd .. || exit 1; rm -rf tar.tempdir 2>/dev/null
 
 exit $FAILCOUNT