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
from ivle import studpath
26
from ivle.util import IVLEError, IVLEJailError
36
# TODO: Make progressive output work
37
# Question: Will having a large buffer size stop progressive output from
38
# working on smaller output
40
CGI_BLOCK_SIZE = 65535
45
"""Get the unix uid corresponding to the given login name.
46
If it is not in the dictionary of uids, then consult the
47
database and retrieve an update of the user table."""
53
res = conn.get_all('login', ['login', 'unixid'])
55
return (flds['login'], flds['unixid'])
56
uids = dict(map(repack,res))
60
def interpret_file(req, owner, jail_dir, filename, interpreter, gentle=True):
61
"""Serves a file by interpreting it using one of IVLE's builtin
62
interpreters. All interpreters are intended to run in the user's jail. The
63
jail location is provided as an argument to the interpreter but it is up
64
to the individual interpreters to create the jail.
66
req: An IVLE request object.
67
owner: Username of the user who owns the file being served.
68
jail_dir: Absolute path to the user's jail.
69
filename: Absolute filename within the user's jail.
70
interpreter: A function object to call.
72
# We can't test here whether or not the target file actually exists,
73
# because the apache user may not have permission. Instead we have to
74
# rely on the interpreter generating an error.
75
if filename.startswith(os.sep):
76
filename_abs = filename
77
filename_rel = filename[1:]
79
filename_abs = os.path.join(os.sep, filename)
80
filename_rel = filename
82
# Get the UID of the owner of the file
83
# (Note: files are executed by their owners, not the logged in user.
84
# This ensures users are responsible for their own programs and also
85
# allows them to be executed by the public).
88
# Split up req.path again, this time with respect to the jail
89
(working_dir, _) = os.path.split(filename_abs)
90
# jail_dir is the absolute jail directory.
91
# path is the filename relative to the user's jail.
92
# working_dir is the directory containing the file relative to the user's
94
# (Note that paths "relative" to the jail actually begin with a '/' as
95
# they are absolute in the jailspace)
97
return interpreter(uid, jail_dir, working_dir, filename_abs, req,
101
"""Stores flags regarding the state of reading CGI output.
102
If this is to be gentle, detection of invalid headers will result in an
104
def __init__(self, begentle=True):
105
self.gentle = begentle
106
self.started_cgi_body = False
107
self.got_cgi_headers = False
108
self.wrote_html_warning = False
110
self.headers = {} # Header names : values
112
def execute_cgi(interpreter, trampoline, uid, jail_dir, working_dir,
113
script_path, req, gentle):
115
trampoline: Full path on the local system to the CGI wrapper program
117
uid: User ID of the owner of the file.
118
jail_dir: Absolute path of owner's jail directory.
119
working_dir: Directory containing the script file relative to owner's
121
script_path: CGI script relative to the owner's jail.
122
req: IVLE request object.
124
The called CGI wrapper application shall be called using popen and receive
125
the HTTP body on stdin. It shall receive the CGI environment variables to
129
# Support no-op trampoline runs.
130
if interpreter is None:
131
interpreter = '/bin/true'
137
# Get the student program's directory and execute it from that context.
138
(tramp_dir, _) = os.path.split(trampoline)
140
# TODO: Don't create a file if the body length is known to be 0
141
# Write the HTTP body to a temporary file so it can be passed as a *real*
144
body = req.read() if not noop else None
148
f.seek(0) # Rewind, for reading
150
# Set up the environment
151
# This automatically asks mod_python to load up the CGI variables into the
152
# environment (which is a good first approximation)
153
old_env = os.environ.copy()
154
for k in os.environ.keys():
156
for (k,v) in req.get_cgi_environ().items():
160
# usage: tramp uid jail_dir working_dir script_path
161
pid = subprocess.Popen(
162
[trampoline, str(uid), jail_dir, working_dir, interpreter,
164
stdin=f, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
167
# Restore the environment
168
for k in os.environ.keys():
170
for (k,v) in old_env.items():
173
# We don't want any output! Bail out after the process terminates.
178
# process_cgi_line: Reads a single line of CGI output and processes it.
179
# Prints to req, and also does fancy HTML warnings if Content-Type
181
cgiflags = CGIFlags(gentle)
183
# Read from the process's stdout into req
184
data = pid.stdout.read(CGI_BLOCK_SIZE)
186
process_cgi_output(req, data, cgiflags)
187
data = pid.stdout.read(CGI_BLOCK_SIZE)
189
# If we haven't processed headers yet, now is a good time
190
if not cgiflags.started_cgi_body:
191
process_cgi_output(req, '\n', cgiflags)
193
# If we wrote an HTML warning header, write the footer
194
if cgiflags.wrote_html_warning:
200
def process_cgi_output(req, data, cgiflags):
201
"""Processes a chunk of CGI output. data is a string of arbitrary length;
202
some arbitrary chunk of output written by the CGI script."""
203
if cgiflags.started_cgi_body:
204
if cgiflags.wrote_html_warning:
205
# HTML escape text if wrote_html_warning
206
req.write(cgi.escape(data))
210
# Break data into lines of CGI header data.
211
linebuf = cgiflags.linebuf + data
212
# First see if we can split all header data
213
# We need to get the double CRLF- or LF-terminated headers, whichever
214
# is smaller, as either sequence may appear somewhere in the body.
215
usplit = linebuf.split('\n\n', 1)
216
wsplit = linebuf.split('\r\n\r\n', 1)
217
split = len(usplit[0]) > len(wsplit[0]) and wsplit or usplit
219
# Haven't seen all headers yet. Buffer and come back later.
220
cgiflags.linebuf = linebuf
225
cgiflags.linebuf = ""
226
cgiflags.started_cgi_body = True
227
# Process all the header lines
228
split = headers.split('\r\n', 1)
230
split = headers.split('\n', 1)
232
process_cgi_header_line(req, split[0], cgiflags)
233
if len(split) == 1: break
235
if cgiflags.wrote_html_warning:
236
# We're done with headers. Treat the rest as data.
237
data = headers + '\n' + data
239
split = headers.split('\r\n', 1)
241
split = headers.split('\n', 1)
243
# Is this an internal IVLE error condition?
244
hs = cgiflags.headers
245
if 'X-IVLE-Error-Type' in hs:
246
t = hs['X-IVLE-Error-Type']
247
if t == IVLEError.__name__:
248
raise IVLEError(int(hs['X-IVLE-Error-Code']),
249
hs['X-IVLE-Error-Message'])
252
raise IVLEJailError(hs['X-IVLE-Error-Type'],
253
hs['X-IVLE-Error-Message'],
254
hs['X-IVLE-Error-Info'])
256
raise IVLEError(500, 'bad error headers written by CGI')
258
# Check to make sure the required headers were written
259
if cgiflags.wrote_html_warning or not cgiflags.gentle:
260
# We already reported an error, that's enough
262
elif "Content-Type" in cgiflags.headers:
264
elif "Location" in cgiflags.headers:
265
if ("Status" in cgiflags.headers and req.status >= 300
266
and req.status < 400):
269
message = """You did not write a valid status code for
270
the given location. To make a redirect, you may wish to try:</p>
271
<pre style="margin-left: 1em">Status: 302 Found
272
Location: <redirect address></pre>"""
273
write_html_warning(req, message)
274
cgiflags.wrote_html_warning = True
276
message = """You did not print a Content-Type header.
277
CGI requires that you print a "Content-Type". You may wish to try:</p>
278
<pre style="margin-left: 1em">Content-Type: text/html</pre>"""
279
write_html_warning(req, message)
280
cgiflags.wrote_html_warning = True
282
# Call myself to flush out the extra bit of data we read
283
process_cgi_output(req, data, cgiflags)
285
def process_cgi_header_line(req, line, cgiflags):
286
"""Process a line of CGI header data. line is a string representing a
287
complete line of text, stripped and without the newline.
290
name, value = line.split(':', 1)
292
# No colon. The user did not write valid headers.
293
# If we are being gentle, we want to help the user understand what
294
# went wrong. Otherwise, just admit we screwed up.
296
if not cgiflags.gentle:
297
message = """An unexpected server error has occured."""
299
elif len(cgiflags.headers) == 0:
300
# First line was not a header line. We can assume this is not
302
message = """You did not print a CGI header.
303
CGI requires that you print a "Content-Type". You may wish to try:</p>
304
<pre style="margin-left: 1em">Content-Type: text/html</pre>"""
306
# They printed some header at least, but there was an invalid
308
message = """You printed an invalid CGI header. You need to leave
309
a blank line after the headers, before writing the page contents."""
310
write_html_warning(req, message, warning=warning)
311
cgiflags.wrote_html_warning = True
312
# Handle the rest of this line as normal data
313
process_cgi_output(req, line + '\n', cgiflags)
317
value = value.strip()
318
if name == "Content-Type":
319
req.content_type = value
320
elif name == "Location":
322
elif name == "Status":
323
# Must be an integer, followed by a space, and then the status line
324
# which we ignore (seems like Apache has no way to send a custom
327
req.status = int(value.split(' ', 1)[0])
329
if not cgiflags.gentle:
330
# This isn't user code, so it should be good.
331
# Get us out of here!
333
message = """The "Status" CGI header was invalid. You need to
334
print a number followed by a message, such as "302 Found"."""
335
write_html_warning(req, message)
336
cgiflags.wrote_html_warning = True
337
# Handle the rest of this line as normal data
338
process_cgi_output(req, line + '\n', cgiflags)
340
# Generic HTTP header
341
# FIXME: Security risk letting users write arbitrary headers?
342
req.headers_out.add(name, value)
343
cgiflags.headers[name] = value # FIXME: Only the last header will end up here.
345
def write_html_warning(req, text, warning="Warning"):
346
"""Prints an HTML warning about invalid CGI interaction on the part of the
347
user. text may contain HTML markup."""
348
req.content_type = "text/html"
349
req.write("""<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
350
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
351
<html xmlns="http://www.w3.org/1999/xhtml">
353
<meta http-equiv="Content-Type"
354
content="text/html; charset=utf-8" />
356
<body style="margin: 0; padding: 0; font-family: sans-serif;">
357
<div style="background-color: #faa; border-bottom: 1px solid black;
359
<p><strong>%s</strong>: %s
361
<div style="margin: 8px;">
363
""" % (warning, text))
365
location_cgi_python = os.path.join(ivle.conf.lib_path, "trampoline")
367
# Mapping of interpreter names (as given in conf/app/server.py) to
368
# interpreter functions.
370
interpreter_objects = {
372
: functools.partial(execute_cgi, "/usr/bin/python",
373
location_cgi_python),
375
: functools.partial(execute_cgi, None,
376
location_cgi_python),
382
def fixup_environ(req):
383
"""Assuming os.environ has been written with the CGI variables from
384
apache, make a few changes for security and correctness.
386
Does not modify req, only reads it.
389
# Comments here are on the heavy side, explained carefully for security
390
# reasons. Please read carefully before making changes.
392
# Remove DOCUMENT_ROOT and SCRIPT_FILENAME. Not part of CGI spec and
393
# exposes unnecessary details about server.
395
del env['DOCUMENT_ROOT']
398
del env['SCRIPT_FILENAME']
401
# Remove PATH. The PATH here is the path on the server machine; not useful
402
# inside the jail. It may be a good idea to add another path, reflecting
403
# the inside of the jail, but not done at this stage.
408
# Remove SCRIPT_FILENAME. Not part of CGI spec (see SCRIPT_NAME).
410
# PATH_INFO is wrong because the script doesn't physically exist.
411
# Apache makes it relative to the "serve" app. It should actually be made
412
# relative to the student's script. intepretservice does that in the jail,
413
# so here we just clear it.
414
env['PATH_INFO'] = ''
415
env['PATH_TRANSLATED'] = ''
417
# CGI specifies that REMOTE_HOST SHOULD be set, and MAY just be set to
418
# REMOTE_ADDR. Since Apache does not appear to set this, set it to
420
if 'REMOTE_HOST' not in env and 'REMOTE_ADDR' in env:
421
env['REMOTE_HOST'] = env['REMOTE_ADDR']
423
# SCRIPT_NAME is the path to the script WITHOUT PATH_INFO.
424
script_name = req.uri
425
env['SCRIPT_NAME'] = script_name
427
# SERVER_SOFTWARE is actually not Apache but IVLE, since we are
428
# custom-making the CGI request.
429
env['SERVER_SOFTWARE'] = "IVLE/" + str(ivle.conf.ivle_version)
431
# Additional environment variables
432
username = studpath.url_to_jailpaths(req.path)[0]
433
env['HOME'] = os.path.join('/home', username)
added
removed
ivle/interpret.py
