Avoid cgi FieldStorage for uploads

httpuploader.py

@@ -1,5 +1,7 @@
 import argparse
 import base64
+from email.parser import BytesHeaderParser
+from email.policy import default as email_policy
 import gzip
 import hashlib
 import io
@@ -16,7 +18,6 @@
 from wsgiref.util import FileWrapper
 from urllib.parse import parse_qs
 from tempfile import TemporaryFile
-from cgi import FieldStorage
 from pprint import pformat
 from datetime import datetime, timedelta, timezone
 from configparser import ConfigParser, ExtendedInterpolation
@@ -31,6 +32,90 @@
 allow = list, download, mkdir, upload
 """
 
+
+def _get_multipart_boundary(content_type):
+    headers = BytesHeaderParser(policy=email_policy).parsebytes(
+        "Content-Type: {}\r\n\r\n".format(content_type).encode("latin-1")
+    )
+    boundary = headers.get_param("boundary", header="content-type")
+    if boundary is None:
+        return None
+    return boundary.encode("latin-1")
+
+
+class _LimitedReader:
+    def __init__(self, fp, content_length):
+        self.fp = fp
+        self.remaining = content_length
+
+    def readline(self):
+        if self.remaining <= 0:
+            return b""
+        line = self.fp.readline(min(CHUNKSIZE, self.remaining))
+        self.remaining -= len(line)
+        return line
+
+
+def _multipart_boundary_type(line, boundary):
+    marker = b"--" + boundary
+    stripped = line.rstrip(b"\r\n")
+    if stripped == marker:
+        return "part"
+    if stripped == marker + b"--":
+        return "final"
+    return None
+
+
+def _strip_multipart_line_break(line):
+    if line.endswith(b"\r\n"):
+        return line[:-2]
+    if line.endswith(b"\n"):
+        return line[:-1]
+    return line
+
+
+def _skip_to_first_multipart_boundary(reader, boundary):
+    while True:
+        line = reader.readline()
+        if not line:
+            return True
+        boundary_type = _multipart_boundary_type(line, boundary)
+        if boundary_type == "final":
+            return True
+        if boundary_type == "part":
+            return False
+
+
+def _read_multipart_headers(reader):
+    lines = []
+    while True:
+        line = reader.readline()
+        if not line or line in (b"\r\n", b"\n"):
+            break
+        lines.append(line)
+    return BytesHeaderParser(policy=email_policy).parsebytes(b"".join(lines))
+
+
+def _consume_multipart_part(reader, boundary, output=None):
+    pending = None
+    while True:
+        line = reader.readline()
+        if not line:
+            if pending and output is not None:
+                output.write(pending)
+            return True
+
+        boundary_type = _multipart_boundary_type(line, boundary)
+        if boundary_type:
+            if pending and output is not None:
+                output.write(_strip_multipart_line_break(pending))
+            return boundary_type == "final"
+
+        if pending and output is not None:
+            output.write(pending)
+        pending = line
+
+
 LOCAL_TZ = datetime.now(timezone(timedelta(0))).astimezone().tzinfo
 
 
@@ -447,18 +532,28 @@ def checksum(self, path, args):
 
     def upload(self, path, args):
         content_type = self.env.get("CONTENT_TYPE", "")
-        if content_type.startswith("multipart/form-data"):
-            fs = FieldStorage(fp=self.env["wsgi.input"], environ=self.env)
-            for key in fs:
-                if fs[key].file:
-                    pn = path / fs[key].filename
-                    with pn.open("wb") as saved:
-                        while 1:
-                            chunk = fs[key].file.read(CHUNKSIZE)
-                            if len(chunk) > 0:
-                                saved.write(chunk)
-                            else:
-                                break
+        if content_type.lower().startswith("multipart/form-data"):
+            boundary = _get_multipart_boundary(content_type)
+            if boundary is None:
+                raise HTUPLError(400, "Bad request", "No multipart boundary")
+
+            content_length = int(self.env.get("CONTENT_LENGTH", "0"))
+            if not content_length:
+                raise HTUPLError(400, "Bad request", "No Content-length header")
+
+            reader = _LimitedReader(self.env["wsgi.input"], content_length)
+            final_boundary = _skip_to_first_multipart_boundary(reader, boundary)
+            while not final_boundary:
+                headers = _read_multipart_headers(reader)
+                filename = headers.get_filename()
+                if filename:
+                    pn = path / filename
+                    with pn.open("wb") as saved:
+                        final_boundary = _consume_multipart_part(
+                            reader, boundary, saved
+                        )
+                else:
+                    final_boundary = _consume_multipart_part(reader, boundary)
         else:
             fp = self.env["wsgi.input"]
             content_length = int(self.env.get("CONTENT_LENGTH", "0"))