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, 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
filename: Filename in the local file system.
54
interpreter: A function object to call.
56
# Make sure the file exists (otherwise some interpreters may not actually
58
# Don't test for execute permission, that will only be required for
59
# certain interpreters.
60
if not os.access(filename, os.R_OK):
61
req.throw_error(req.HTTP_NOT_FOUND)
63
# Get the UID of the owner of the file
64
# (Note: files are executed by their owners, not the logged in user.
65
# This ensures users are responsible for their own programs and also
66
# allows them to be executed by the public).
68
(_,_,uid,_,_,_,_) = pwd.getpwnam(owner)
70
# The user does not exist. This should have already failed the
72
req.throw_error(req.HTTP_INTERNAL_SERVER_ERROR)
74
# Split up req.path again, this time with respect to the jail
75
(_, jail_dir, path) = studpath.url_to_jailpaths(req.path)
76
path = os.path.join('/', path)
77
(working_dir, _) = os.path.split(path)
78
# jail_dir is the absolute jail directory.
79
# path is the filename relative to the user's jail.
80
# working_dir is the directory containing the file relative to the user's
82
# (Note that paths "relative" to the jail actually begin with a '/' as
83
# they are absolute in the jailspace)
85
return interpreter(uid, jail_dir, working_dir, path, req)
88
"""Stores flags regarding the state of reading CGI output."""
90
self.started_cgi_body = False
91
self.got_cgi_headers = False
92
self.wrote_html_warning = False
94
self.headers = {} # Header names : values
96
def execute_cgi(interpreter, trampoline, uid, jail_dir, working_dir,
99
trampoline: Full path on the local system to the CGI wrapper program
101
uid: User ID of the owner of the file.
102
jail_dir: Absolute path of owner's jail directory.
103
working_dir: Directory containing the script file relative to owner's
105
script_path: CGI script relative to the owner's jail.
106
req: IVLE request object.
108
The called CGI wrapper application shall be called using popen and receive
109
the HTTP body on stdin. It shall receive the CGI environment variables to
113
# Get the student program's directory and execute it from that context.
114
(tramp_dir, _) = os.path.split(trampoline)
116
# TODO: Don't create a file if the body length is known to be 0
117
# Write the HTTP body to a temporary file so it can be passed as a *real*
124
f.seek(0) # Rewind, for reading
126
# Set up the environment
127
# This automatically asks mod_python to load up the CGI variables into the
128
# environment (which is a good first approximation)
129
old_env = os.environ.copy()
130
for k in os.environ.keys():
132
for (k,v) in req.get_cgi_environ().items():
136
# usage: tramp uid jail_dir working_dir script_path
137
pid = subprocess.Popen(
138
[trampoline, str(uid), jail_dir, working_dir, interpreter,
140
stdin=f, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
143
# Restore the environment
144
for k in os.environ.keys():
146
for (k,v) in old_env.items():
149
# process_cgi_line: Reads a single line of CGI output and processes it.
150
# Prints to req, and also does fancy HTML warnings if Content-Type
152
cgiflags = CGIFlags()
154
# Read from the process's stdout into req
155
data = pid.stdout.read(CGI_BLOCK_SIZE)
157
process_cgi_output(req, data, cgiflags)
158
data = pid.stdout.read(CGI_BLOCK_SIZE)
160
# If we haven't processed headers yet, now is a good time
161
if not cgiflags.started_cgi_body:
162
process_cgi_output(req, '\n', cgiflags)
164
# If we wrote an HTML warning header, write the footer
165
if cgiflags.wrote_html_warning:
171
def process_cgi_output(req, data, cgiflags):
172
"""Processes a chunk of CGI output. data is a string of arbitrary length;
173
some arbitrary chunk of output written by the CGI script."""
174
if cgiflags.started_cgi_body:
175
if cgiflags.wrote_html_warning:
176
# HTML escape text if wrote_html_warning
177
req.write(cgi.escape(data))
181
# Break data into lines of CGI header data.
182
linebuf = cgiflags.linebuf + data
183
# First see if we can split all header data
184
split = linebuf.split('\r\n\r\n', 1)
186
# Allow UNIX newlines instead
187
split = linebuf.split('\n\n', 1)
189
# Haven't seen all headers yet. Buffer and come back later.
190
cgiflags.linebuf = linebuf
195
cgiflags.linebuf = ""
196
cgiflags.started_cgi_body = True
197
# Process all the header lines
198
split = headers.split('\r\n', 1)
200
split = headers.split('\n', 1)
202
process_cgi_header_line(req, split[0], cgiflags)
203
if len(split) == 1: break
205
if cgiflags.wrote_html_warning:
206
# We're done with headers. Treat the rest as data.
207
data = headers + '\n' + data
209
split = headers.split('\r\n', 1)
211
split = headers.split('\n', 1)
213
# Check to make sure the required headers were written
214
if cgiflags.wrote_html_warning:
215
# We already reported an error, that's enough
217
elif "Content-Type" in cgiflags.headers:
219
elif "Location" in cgiflags.headers:
220
if ("Status" in cgiflags.headers and req.status >= 300
221
and req.status < 400):
224
message = """You did not write a valid status code for
225
the given location. To make a redirect, you may wish to try:</p>
226
<pre style="margin-left: 1em">Status: 302 Found
227
Location: <redirect address></pre>"""
228
write_html_warning(req, message)
229
cgiflags.wrote_html_warning = True
231
message = """You did not print a Content-Type header.
232
CGI requires that you print a "Content-Type". You may wish to try:</p>
233
<pre style="margin-left: 1em">Content-Type: text/html</pre>"""
234
write_html_warning(req, message)
235
cgiflags.wrote_html_warning = True
237
# Call myself to flush out the extra bit of data we read
238
process_cgi_output(req, data, cgiflags)
240
def process_cgi_header_line(req, line, cgiflags):
241
"""Process a line of CGI header data. line is a string representing a
242
complete line of text, stripped and without the newline.
245
name, value = line.split(':', 1)
247
# No colon. The user did not write valid headers.
248
if len(cgiflags.headers) == 0:
249
# First line was not a header line. We can assume this is not
251
message = """You did not print a CGI header.
252
CGI requires that you print a "Content-Type". You may wish to try:</p>
253
<pre style="margin-left: 1em">Content-Type: text/html</pre>"""
255
# They printed some header at least, but there was an invalid
257
message = """You printed an invalid CGI header. You need to leave
258
a blank line after the headers, before writing the page contents."""
259
write_html_warning(req, message)
260
cgiflags.wrote_html_warning = True
261
# Handle the rest of this line as normal data
262
process_cgi_output(req, line + '\n', cgiflags)
266
value = value.strip()
267
if name == "Content-Type":
268
req.content_type = value
269
elif name == "Location":
271
elif name == "Status":
272
# Must be an integer, followed by a space, and then the status line
273
# which we ignore (seems like Apache has no way to send a custom
276
req.status = int(value.split(' ', 1)[0])
278
message = """The "Status" CGI header was invalid. You need to
279
print a number followed by a message, such as "302 Found"."""
280
write_html_warning(req, message)
281
cgiflags.wrote_html_warning = True
282
# Handle the rest of this line as normal data
283
process_cgi_output(req, line + '\n', cgiflags)
285
# Generic HTTP header
286
# FIXME: Security risk letting users write arbitrary headers?
287
req.headers_out[name] = value
288
cgiflags.headers[name] = value
290
def write_html_warning(req, text):
291
"""Prints an HTML warning about invalid CGI interaction on the part of the
292
user. text may contain HTML markup."""
293
req.content_type = "text/html"
294
req.write("""<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
295
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
296
<html xmlns="http://www.w3.org/1999/xhtml">
298
<meta http-equiv="Content-Type"
299
content="text/html; charset=utf-8" />
301
<body style="margin: 0; padding: 0; font-family: sans-serif;">
302
<div style="background-color: #faa; border-bottom: 1px solid black;
304
<p><strong>Warning</strong>: %s
306
<div style="margin: 8px;">
310
location_cgi_python = os.path.join(conf.ivle_install_dir,
313
# Mapping of interpreter names (as given in conf/app/server.py) to
314
# interpreter functions.
316
interpreter_objects = {
318
: functools.partial(execute_cgi, "/usr/bin/python",
319
location_cgi_python),
325
def fixup_environ(req):
326
"""Assuming os.environ has been written with the CGI variables from
327
apache, make a few changes for security and correctness.
329
Does not modify req, only reads it.
332
# Comments here are on the heavy side, explained carefully for security
333
# reasons. Please read carefully before making changes.
335
# Remove HTTP_COOKIE. It is a security risk to have students see the IVLE
336
# cookie of their visitors.
338
del env['HTTP_COOKIE']
341
# Remove DOCUMENT_ROOT and SCRIPT_FILENAME. Not part of CGI spec and
342
# exposes unnecessary details about server.
344
del env['DOCUMENT_ROOT']
347
del env['SCRIPT_FILENAME']
350
# Remove PATH. The PATH here is the path on the server machine; not useful
351
# inside the jail. It may be a good idea to add another path, reflecting
352
# the inside of the jail, but not done at this stage.
357
# Remove SCRIPT_FILENAME. Not part of CGI spec (see SCRIPT_NAME).
359
# PATH_INFO is wrong because the script doesn't physically exist.
360
# Apache makes it relative to the "serve" app. It should actually be made
361
# relative to the student's script.
362
# TODO: At this stage, it is not possible to add a path after the script,
363
# so PATH_INFO is always "".
365
env['PATH_INFO'] = path_info
367
# PATH_TRANSLATED currently points to a non-existant location within the
368
# local web server directory. Instead make it represent a path within the
370
(username, _, path_translated) = studpath.url_to_jailpaths(req.path)
371
if len(path_translated) == 0 or path_translated[0] != os.sep:
372
path_translated = os.sep + path_translated
373
env['PATH_TRANSLATED'] = path_translated
375
# CGI specifies that REMOTE_HOST SHOULD be set, and MAY just be set to
376
# REMOTE_ADDR. Since Apache does not appear to set this, set it to
378
if 'REMOTE_HOST' not in env and 'REMOTE_ADDR' in env:
379
env['REMOTE_HOST'] = env['REMOTE_ADDR']
381
# SCRIPT_NAME is the path to the script WITHOUT PATH_INFO.
382
script_name = req.uri
383
if len(path_info) > 0:
384
script_name = script_name[:-len(path_info)]
385
env['SCRIPT_NAME'] = script_name
387
# SERVER_SOFTWARE is actually not Apache but IVLE, since we are
388
# custom-making the CGI request.
389
env['SERVER_SOFTWARE'] = "IVLE/" + str(conf.ivle_version)
391
# Additional environment variables
392
env['HOME'] = os.path.join('/home', username)
added
removed
www/common/interpret.py
