Fix path traversal security vulnerability by canonicalizing path name…

…s of every inodes and discarding inodes with a path

pointing outside of the extraction directory.

Fix path traversal through symlinks by canonicalizing link target path using the extraction directory as root. If the link
still points outside the extraction root, it gets discarded. This way symlinks reflects the reality of a filesystem on device
by pointing to files within the extracted filesystem instead of files from the host executing jefferson.

src/scripts/jefferson

@@ -6,7 +6,6 @@ import stat
 import os
 import zlib
 import binascii
-
 import cstruct
 
 from jefferson import jffs2_lzma, rtime
@@ -52,6 +51,12 @@ JFFS2_NODETYPE_XREF = JFFS2_FEATURE_INCOMPAT | JFFS2_NODE_ACCURATE | 9
 def mtd_crc(data):
     return (binascii.crc32(data, -1) ^ -1) & 0xFFFFFFFF
 
+def is_safe_path(basedir, path, follow_symlinks=True):
+    if follow_symlinks:
+        matchpath = os.path.realpath(path)
+    else:
+        matchpath = os.path.abspath(path)
+    return basedir == os.path.commonpath((basedir, matchpath))
 
 cstruct.typedef("uint8", "uint8_t")
 cstruct.typedef("uint16", "jint16_t")
@@ -474,20 +479,32 @@ def dump_fs(fs, target):
         node_names.append(dirent.name.decode())
         path = "/".join(node_names)
 
-        target_path = os.path.join(os.getcwd(), target, path)
+        target_path = os.path.realpath(os.path.join(os.getcwd(), target, path))
+
+        if not is_safe_path(target, target_path):
+            print(f"Path traversal attempt to {target_path}, discarding.")
+            continue
+
         for inode in dirent.inodes:
             try:
                 if stat.S_ISDIR(inode.mode):
                     print("writing S_ISDIR", path)
                     if not os.path.isdir(target_path):
                         os.makedirs(target_path)
                 elif stat.S_ISLNK(inode.mode):
+                    link_path = inode.data.decode('utf-8')
+                    if link_path:
+                        link_path = link_path[1:]
+                    link_target = os.path.realpath(os.path.join(target, link_path))
+                    if not is_safe_path(target, link_target):
+                        print(f"Path traversal attempt through symlink to {link_target}, discarding.")
+                        continue
                     print("writing S_ISLNK", path)
                     if not os.path.islink(target_path):
                         if os.path.exists(target_path):
                             print("file already exists as", inode.data)
                             continue
-                        os.symlink(inode.data, target_path)
+                        os.symlink(link_target, target_path)
                 elif stat.S_ISREG(inode.mode):
                     print("writing S_ISREG", path)
                     if not os.path.isfile(target_path):
@@ -561,7 +578,7 @@ def main():
         if not fs[JFFS2_NODETYPE_DIRENT]:
             continue
 
-        dest_path_fs = os.path.join(dest_path, "fs_%i" % fs_index)
+        dest_path_fs = os.path.realpath(os.path.join(dest_path, "fs_%i" % fs_index))
         print("dumping fs #%i to %s" % (fs_index, dest_path_fs))
         for key, value in fs.items():
             if key == "endianness":