2
# Copyright (C) 2007-2008 The University of Melbourne
4
# This program is free software; you can redistribute it and/or modify
5
# it under the terms of the GNU General Public License as published by
6
# the Free Software Foundation; either version 2 of the License, or
7
# (at your option) any later version.
9
# This program is distributed in the hope that it will be useful,
10
# but WITHOUT ANY WARRANTY; without even the implied warranty of
11
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12
# GNU General Public License for more details.
14
# You should have received a copy of the GNU General Public License
15
# along with this program; if not, write to the Free Software
16
# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
22
# Runs a student script in a safe execution environment.
24
# NOTE: This script currently disables cookies. This means students will be
25
# unable to write session-based or stateful web applications. This is done for
26
# security reasons (we do not want the students to see the IVLE cookie of
27
# whoever is visiting their pages).
28
# This can be resolved but needs careful sanitisation. See fixup_environ.
30
from common import studpath
39
# TODO: Make progressive output work
40
# Question: Will having a large buffer size stop progressive output from
41
# working on smaller output
43
CGI_BLOCK_SIZE = 65535
45
def interpret_file(req, owner, jail_dir, filename, interpreter):
46
"""Serves a file by interpreting it using one of IVLE's builtin
47
interpreters. All interpreters are intended to run in the user's jail. The
48
jail location is provided as an argument to the interpreter but it is up
49
to the individual interpreters to create the jail.
51
req: An IVLE request object.
52
owner: Username of the user who owns the file being served.
53
jail_dir: Absolute path to the user's jail.
54
filename: Absolute filename within the user's jail.
55
interpreter: A function object to call.
57
# Make sure the file exists (otherwise some interpreters may not actually
59
# Don't test for execute permission, that will only be required for
60
# certain interpreters.
61
if filename.startswith(os.sep):
62
filename_abs = filename
63
filename_rel = filename[1:]
65
filename_abs = os.path.join(os.sep, filename)
66
filename_rel = filename
68
if not os.access(os.path.join(jail_dir, filename_rel), os.R_OK):
69
req.throw_error(req.HTTP_NOT_FOUND)
71
# Get the UID of the owner of the file
72
# (Note: files are executed by their owners, not the logged in user.
73
# This ensures users are responsible for their own programs and also
74
# allows them to be executed by the public).
75
uid = req.get_session()['unixid']
77
# Split up req.path again, this time with respect to the jail
78
(working_dir, _) = os.path.split(filename_abs)
79
# jail_dir is the absolute jail directory.
80
# path is the filename relative to the user's jail.
81
# working_dir is the directory containing the file relative to the user's
83
# (Note that paths "relative" to the jail actually begin with a '/' as
84
# they are absolute in the jailspace)
86
return interpreter(uid, jail_dir, working_dir, filename_abs, req)
89
"""Stores flags regarding the state of reading CGI output."""
91
self.started_cgi_body = False
92
self.got_cgi_headers = False
93
self.wrote_html_warning = False
95
self.headers = {} # Header names : values
97
def execute_cgi(interpreter, trampoline, uid, jail_dir, working_dir,
100
trampoline: Full path on the local system to the CGI wrapper program
102
uid: User ID of the owner of the file.
103
jail_dir: Absolute path of owner's jail directory.
104
working_dir: Directory containing the script file relative to owner's
106
script_path: CGI script relative to the owner's jail.
107
req: IVLE request object.
109
The called CGI wrapper application shall be called using popen and receive
110
the HTTP body on stdin. It shall receive the CGI environment variables to
114
# Get the student program's directory and execute it from that context.
115
(tramp_dir, _) = os.path.split(trampoline)
117
# TODO: Don't create a file if the body length is known to be 0
118
# Write the HTTP body to a temporary file so it can be passed as a *real*
125
f.seek(0) # Rewind, for reading
127
# Set up the environment
128
# This automatically asks mod_python to load up the CGI variables into the
129
# environment (which is a good first approximation)
130
old_env = os.environ.copy()
131
for k in os.environ.keys():
133
for (k,v) in req.get_cgi_environ().items():
137
# usage: tramp uid jail_dir working_dir script_path
138
pid = subprocess.Popen(
139
[trampoline, str(uid), jail_dir, working_dir, interpreter,
141
stdin=f, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
144
# Restore the environment
145
for k in os.environ.keys():
147
for (k,v) in old_env.items():
150
# process_cgi_line: Reads a single line of CGI output and processes it.
151
# Prints to req, and also does fancy HTML warnings if Content-Type
153
cgiflags = CGIFlags()
155
# Read from the process's stdout into req
156
data = pid.stdout.read(CGI_BLOCK_SIZE)
158
process_cgi_output(req, data, cgiflags)
159
data = pid.stdout.read(CGI_BLOCK_SIZE)
161
# If we haven't processed headers yet, now is a good time
162
if not cgiflags.started_cgi_body:
163
process_cgi_output(req, '\n', cgiflags)
165
# If we wrote an HTML warning header, write the footer
166
if cgiflags.wrote_html_warning:
172
def process_cgi_output(req, data, cgiflags):
173
"""Processes a chunk of CGI output. data is a string of arbitrary length;
174
some arbitrary chunk of output written by the CGI script."""
175
if cgiflags.started_cgi_body:
176
if cgiflags.wrote_html_warning:
177
# HTML escape text if wrote_html_warning
178
req.write(cgi.escape(data))
182
# Break data into lines of CGI header data.
183
linebuf = cgiflags.linebuf + data
184
# First see if we can split all header data
185
split = linebuf.split('\r\n\r\n', 1)
187
# Allow UNIX newlines instead
188
split = linebuf.split('\n\n', 1)
190
# Haven't seen all headers yet. Buffer and come back later.
191
cgiflags.linebuf = linebuf
196
cgiflags.linebuf = ""
197
cgiflags.started_cgi_body = True
198
# Process all the header lines
199
split = headers.split('\r\n', 1)
201
split = headers.split('\n', 1)
203
process_cgi_header_line(req, split[0], cgiflags)
204
if len(split) == 1: break
206
if cgiflags.wrote_html_warning:
207
# We're done with headers. Treat the rest as data.
208
data = headers + '\n' + data
210
split = headers.split('\r\n', 1)
212
split = headers.split('\n', 1)
214
# Check to make sure the required headers were written
215
if cgiflags.wrote_html_warning:
216
# We already reported an error, that's enough
218
elif "Content-Type" in cgiflags.headers:
220
elif "Location" in cgiflags.headers:
221
if ("Status" in cgiflags.headers and req.status >= 300
222
and req.status < 400):
225
message = """You did not write a valid status code for
226
the given location. To make a redirect, you may wish to try:</p>
227
<pre style="margin-left: 1em">Status: 302 Found
228
Location: <redirect address></pre>"""
229
write_html_warning(req, message)
230
cgiflags.wrote_html_warning = True
232
message = """You did not print a Content-Type header.
233
CGI requires that you print a "Content-Type". You may wish to try:</p>
234
<pre style="margin-left: 1em">Content-Type: text/html</pre>"""
235
write_html_warning(req, message)
236
cgiflags.wrote_html_warning = True
238
# Call myself to flush out the extra bit of data we read
239
process_cgi_output(req, data, cgiflags)
241
def process_cgi_header_line(req, line, cgiflags):
242
"""Process a line of CGI header data. line is a string representing a
243
complete line of text, stripped and without the newline.
246
name, value = line.split(':', 1)
248
# No colon. The user did not write valid headers.
249
if len(cgiflags.headers) == 0:
250
# First line was not a header line. We can assume this is not
252
message = """You did not print a CGI header.
253
CGI requires that you print a "Content-Type". You may wish to try:</p>
254
<pre style="margin-left: 1em">Content-Type: text/html</pre>"""
256
# They printed some header at least, but there was an invalid
258
message = """You printed an invalid CGI header. You need to leave
259
a blank line after the headers, before writing the page contents."""
260
write_html_warning(req, message)
261
cgiflags.wrote_html_warning = True
262
# Handle the rest of this line as normal data
263
process_cgi_output(req, line + '\n', cgiflags)
267
value = value.strip()
268
if name == "Content-Type":
269
req.content_type = value
270
elif name == "Location":
272
elif name == "Status":
273
# Must be an integer, followed by a space, and then the status line
274
# which we ignore (seems like Apache has no way to send a custom
277
req.status = int(value.split(' ', 1)[0])
279
message = """The "Status" CGI header was invalid. You need to
280
print a number followed by a message, such as "302 Found"."""
281
write_html_warning(req, message)
282
cgiflags.wrote_html_warning = True
283
# Handle the rest of this line as normal data
284
process_cgi_output(req, line + '\n', cgiflags)
286
# Generic HTTP header
287
# FIXME: Security risk letting users write arbitrary headers?
288
req.headers_out[name] = value
289
cgiflags.headers[name] = value
291
def write_html_warning(req, text):
292
"""Prints an HTML warning about invalid CGI interaction on the part of the
293
user. text may contain HTML markup."""
294
req.content_type = "text/html"
295
req.write("""<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
296
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
297
<html xmlns="http://www.w3.org/1999/xhtml">
299
<meta http-equiv="Content-Type"
300
content="text/html; charset=utf-8" />
302
<body style="margin: 0; padding: 0; font-family: sans-serif;">
303
<div style="background-color: #faa; border-bottom: 1px solid black;
305
<p><strong>Warning</strong>: %s
307
<div style="margin: 8px;">
311
location_cgi_python = os.path.join(conf.ivle_install_dir,
314
# Mapping of interpreter names (as given in conf/app/server.py) to
315
# interpreter functions.
317
interpreter_objects = {
319
: functools.partial(execute_cgi, "/usr/bin/python",
320
location_cgi_python),
326
def fixup_environ(req):
327
"""Assuming os.environ has been written with the CGI variables from
328
apache, make a few changes for security and correctness.
330
Does not modify req, only reads it.
333
# Comments here are on the heavy side, explained carefully for security
334
# reasons. Please read carefully before making changes.
336
# Remove HTTP_COOKIE. It is a security risk to have students see the IVLE
337
# cookie of their visitors.
339
del env['HTTP_COOKIE']
342
# Remove DOCUMENT_ROOT and SCRIPT_FILENAME. Not part of CGI spec and
343
# exposes unnecessary details about server.
345
del env['DOCUMENT_ROOT']
348
del env['SCRIPT_FILENAME']
351
# Remove PATH. The PATH here is the path on the server machine; not useful
352
# inside the jail. It may be a good idea to add another path, reflecting
353
# the inside of the jail, but not done at this stage.
358
# Remove SCRIPT_FILENAME. Not part of CGI spec (see SCRIPT_NAME).
360
# PATH_INFO is wrong because the script doesn't physically exist.
361
# Apache makes it relative to the "serve" app. It should actually be made
362
# relative to the student's script.
363
# TODO: At this stage, it is not possible to add a path after the script,
364
# so PATH_INFO is always "".
366
env['PATH_INFO'] = path_info
368
# PATH_TRANSLATED currently points to a non-existant location within the
369
# local web server directory. Instead make it represent a path within the
371
(username, _, path_translated) = studpath.url_to_jailpaths(req.path)
372
if len(path_translated) == 0 or path_translated[0] != os.sep:
373
path_translated = os.sep + path_translated
374
env['PATH_TRANSLATED'] = path_translated
376
# CGI specifies that REMOTE_HOST SHOULD be set, and MAY just be set to
377
# REMOTE_ADDR. Since Apache does not appear to set this, set it to
379
if 'REMOTE_HOST' not in env and 'REMOTE_ADDR' in env:
380
env['REMOTE_HOST'] = env['REMOTE_ADDR']
382
# SCRIPT_NAME is the path to the script WITHOUT PATH_INFO.
383
script_name = req.uri
384
if len(path_info) > 0:
385
script_name = script_name[:-len(path_info)]
386
env['SCRIPT_NAME'] = script_name
388
# SERVER_SOFTWARE is actually not Apache but IVLE, since we are
389
# custom-making the CGI request.
390
env['SERVER_SOFTWARE'] = "IVLE/" + str(conf.ivle_version)
392
# Additional environment variables
393
env['HOME'] = os.path.join('/home', username)