Show More
@@ -1,543 +1,544 | |||
|
1 | 1 | # hgweb/hgwebdir_mod.py - Web interface for a directory of repositories. |
|
2 | 2 | # |
|
3 | 3 | # Copyright 21 May 2005 - (c) 2005 Jake Edge <jake@edge2.net> |
|
4 | 4 | # Copyright 2005, 2006 Matt Mackall <mpm@selenic.com> |
|
5 | 5 | # |
|
6 | 6 | # This software may be used and distributed according to the terms of the |
|
7 | 7 | # GNU General Public License version 2 or any later version. |
|
8 | 8 | |
|
9 | 9 | from __future__ import absolute_import |
|
10 | 10 | |
|
11 | 11 | import os |
|
12 | 12 | import re |
|
13 | 13 | import time |
|
14 | 14 | |
|
15 | 15 | from ..i18n import _ |
|
16 | 16 | |
|
17 | 17 | from .common import ( |
|
18 | 18 | ErrorResponse, |
|
19 | 19 | HTTP_NOT_FOUND, |
|
20 | 20 | HTTP_OK, |
|
21 | 21 | HTTP_SERVER_ERROR, |
|
22 | 22 | cspvalues, |
|
23 | 23 | get_contact, |
|
24 | 24 | get_mtime, |
|
25 | 25 | ismember, |
|
26 | 26 | paritygen, |
|
27 | 27 | staticfile, |
|
28 | 28 | ) |
|
29 | 29 | |
|
30 | 30 | from .. import ( |
|
31 | 31 | configitems, |
|
32 | 32 | encoding, |
|
33 | 33 | error, |
|
34 | 34 | hg, |
|
35 | 35 | profiling, |
|
36 | 36 | pycompat, |
|
37 | 37 | scmutil, |
|
38 | 38 | templater, |
|
39 | 39 | ui as uimod, |
|
40 | 40 | util, |
|
41 | 41 | ) |
|
42 | 42 | |
|
43 | 43 | from . import ( |
|
44 | 44 | hgweb_mod, |
|
45 | 45 | request as requestmod, |
|
46 | 46 | webutil, |
|
47 | 47 | wsgicgi, |
|
48 | 48 | ) |
|
49 | 49 | from ..utils import dateutil |
|
50 | 50 | |
|
51 | 51 | def cleannames(items): |
|
52 | 52 | return [(util.pconvert(name).strip('/'), path) for name, path in items] |
|
53 | 53 | |
|
54 | 54 | def findrepos(paths): |
|
55 | 55 | repos = [] |
|
56 | 56 | for prefix, root in cleannames(paths): |
|
57 | 57 | roothead, roottail = os.path.split(root) |
|
58 | 58 | # "foo = /bar/*" or "foo = /bar/**" lets every repo /bar/N in or below |
|
59 | 59 | # /bar/ be served as as foo/N . |
|
60 | 60 | # '*' will not search inside dirs with .hg (except .hg/patches), |
|
61 | 61 | # '**' will search inside dirs with .hg (and thus also find subrepos). |
|
62 | 62 | try: |
|
63 | 63 | recurse = {'*': False, '**': True}[roottail] |
|
64 | 64 | except KeyError: |
|
65 | 65 | repos.append((prefix, root)) |
|
66 | 66 | continue |
|
67 | 67 | roothead = os.path.normpath(os.path.abspath(roothead)) |
|
68 | 68 | paths = scmutil.walkrepos(roothead, followsym=True, recurse=recurse) |
|
69 | 69 | repos.extend(urlrepos(prefix, roothead, paths)) |
|
70 | 70 | return repos |
|
71 | 71 | |
|
72 | 72 | def urlrepos(prefix, roothead, paths): |
|
73 | 73 | """yield url paths and filesystem paths from a list of repo paths |
|
74 | 74 | |
|
75 | 75 | >>> conv = lambda seq: [(v, util.pconvert(p)) for v,p in seq] |
|
76 | 76 | >>> conv(urlrepos(b'hg', b'/opt', [b'/opt/r', b'/opt/r/r', b'/opt'])) |
|
77 | 77 | [('hg/r', '/opt/r'), ('hg/r/r', '/opt/r/r'), ('hg', '/opt')] |
|
78 | 78 | >>> conv(urlrepos(b'', b'/opt', [b'/opt/r', b'/opt/r/r', b'/opt'])) |
|
79 | 79 | [('r', '/opt/r'), ('r/r', '/opt/r/r'), ('', '/opt')] |
|
80 | 80 | """ |
|
81 | 81 | for path in paths: |
|
82 | 82 | path = os.path.normpath(path) |
|
83 | 83 | yield (prefix + '/' + |
|
84 | 84 | util.pconvert(path[len(roothead):]).lstrip('/')).strip('/'), path |
|
85 | 85 | |
|
86 | 86 | def geturlcgivars(baseurl, port): |
|
87 | 87 | """ |
|
88 | 88 | Extract CGI variables from baseurl |
|
89 | 89 | |
|
90 | 90 | >>> geturlcgivars(b"http://host.org/base", b"80") |
|
91 | 91 | ('host.org', '80', '/base') |
|
92 | 92 | >>> geturlcgivars(b"http://host.org:8000/base", b"80") |
|
93 | 93 | ('host.org', '8000', '/base') |
|
94 | 94 | >>> geturlcgivars(b'/base', 8000) |
|
95 | 95 | ('', '8000', '/base') |
|
96 | 96 | >>> geturlcgivars(b"base", b'8000') |
|
97 | 97 | ('', '8000', '/base') |
|
98 | 98 | >>> geturlcgivars(b"http://host", b'8000') |
|
99 | 99 | ('host', '8000', '/') |
|
100 | 100 | >>> geturlcgivars(b"http://host/", b'8000') |
|
101 | 101 | ('host', '8000', '/') |
|
102 | 102 | """ |
|
103 | 103 | u = util.url(baseurl) |
|
104 | 104 | name = u.host or '' |
|
105 | 105 | if u.port: |
|
106 | 106 | port = u.port |
|
107 | 107 | path = u.path or "" |
|
108 | 108 | if not path.startswith('/'): |
|
109 | 109 | path = '/' + path |
|
110 | 110 | |
|
111 | 111 | return name, pycompat.bytestr(port), path |
|
112 | 112 | |
|
113 | 113 | class hgwebdir(object): |
|
114 | 114 | """HTTP server for multiple repositories. |
|
115 | 115 | |
|
116 | 116 | Given a configuration, different repositories will be served depending |
|
117 | 117 | on the request path. |
|
118 | 118 | |
|
119 | 119 | Instances are typically used as WSGI applications. |
|
120 | 120 | """ |
|
121 | 121 | def __init__(self, conf, baseui=None): |
|
122 | 122 | self.conf = conf |
|
123 | 123 | self.baseui = baseui |
|
124 | 124 | self.ui = None |
|
125 | 125 | self.lastrefresh = 0 |
|
126 | 126 | self.motd = None |
|
127 | 127 | self.refresh() |
|
128 | 128 | |
|
129 | 129 | def refresh(self): |
|
130 | 130 | if self.ui: |
|
131 | 131 | refreshinterval = self.ui.configint('web', 'refreshinterval') |
|
132 | 132 | else: |
|
133 | 133 | item = configitems.coreitems['web']['refreshinterval'] |
|
134 | 134 | refreshinterval = item.default |
|
135 | 135 | |
|
136 | 136 | # refreshinterval <= 0 means to always refresh. |
|
137 | 137 | if (refreshinterval > 0 and |
|
138 | 138 | self.lastrefresh + refreshinterval > time.time()): |
|
139 | 139 | return |
|
140 | 140 | |
|
141 | 141 | if self.baseui: |
|
142 | 142 | u = self.baseui.copy() |
|
143 | 143 | else: |
|
144 | 144 | u = uimod.ui.load() |
|
145 | 145 | u.setconfig('ui', 'report_untrusted', 'off', 'hgwebdir') |
|
146 | 146 | u.setconfig('ui', 'nontty', 'true', 'hgwebdir') |
|
147 | 147 | # displaying bundling progress bar while serving feels wrong and may |
|
148 | 148 | # break some wsgi implementations. |
|
149 | 149 | u.setconfig('progress', 'disable', 'true', 'hgweb') |
|
150 | 150 | |
|
151 | 151 | if not isinstance(self.conf, (dict, list, tuple)): |
|
152 | 152 | map = {'paths': 'hgweb-paths'} |
|
153 | 153 | if not os.path.exists(self.conf): |
|
154 | 154 | raise error.Abort(_('config file %s not found!') % self.conf) |
|
155 | 155 | u.readconfig(self.conf, remap=map, trust=True) |
|
156 | 156 | paths = [] |
|
157 | 157 | for name, ignored in u.configitems('hgweb-paths'): |
|
158 | 158 | for path in u.configlist('hgweb-paths', name): |
|
159 | 159 | paths.append((name, path)) |
|
160 | 160 | elif isinstance(self.conf, (list, tuple)): |
|
161 | 161 | paths = self.conf |
|
162 | 162 | elif isinstance(self.conf, dict): |
|
163 | 163 | paths = self.conf.items() |
|
164 | 164 | |
|
165 | 165 | repos = findrepos(paths) |
|
166 | 166 | for prefix, root in u.configitems('collections'): |
|
167 | 167 | prefix = util.pconvert(prefix) |
|
168 | 168 | for path in scmutil.walkrepos(root, followsym=True): |
|
169 | 169 | repo = os.path.normpath(path) |
|
170 | 170 | name = util.pconvert(repo) |
|
171 | 171 | if name.startswith(prefix): |
|
172 | 172 | name = name[len(prefix):] |
|
173 | 173 | repos.append((name.lstrip('/'), repo)) |
|
174 | 174 | |
|
175 | 175 | self.repos = repos |
|
176 | 176 | self.ui = u |
|
177 | 177 | encoding.encoding = self.ui.config('web', 'encoding') |
|
178 | 178 | self.style = self.ui.config('web', 'style') |
|
179 | 179 | self.templatepath = self.ui.config('web', 'templates', untrusted=False) |
|
180 | 180 | self.stripecount = self.ui.config('web', 'stripes') |
|
181 | 181 | if self.stripecount: |
|
182 | 182 | self.stripecount = int(self.stripecount) |
|
183 | 183 | self._baseurl = self.ui.config('web', 'baseurl') |
|
184 | 184 | prefix = self.ui.config('web', 'prefix') |
|
185 | 185 | if prefix.startswith('/'): |
|
186 | 186 | prefix = prefix[1:] |
|
187 | 187 | if prefix.endswith('/'): |
|
188 | 188 | prefix = prefix[:-1] |
|
189 | 189 | self.prefix = prefix |
|
190 | 190 | self.lastrefresh = time.time() |
|
191 | 191 | |
|
192 | 192 | def run(self): |
|
193 | 193 | if not encoding.environ.get('GATEWAY_INTERFACE', |
|
194 | 194 | '').startswith("CGI/1."): |
|
195 | 195 | raise RuntimeError("This function is only intended to be " |
|
196 | 196 | "called while running as a CGI script.") |
|
197 | 197 | wsgicgi.launch(self) |
|
198 | 198 | |
|
199 | 199 | def __call__(self, env, respond): |
|
200 | 200 | wsgireq = requestmod.wsgirequest(env, respond) |
|
201 | 201 | return self.run_wsgi(wsgireq) |
|
202 | 202 | |
|
203 | 203 | def read_allowed(self, ui, wsgireq): |
|
204 | 204 | """Check allow_read and deny_read config options of a repo's ui object |
|
205 | 205 | to determine user permissions. By default, with neither option set (or |
|
206 | 206 | both empty), allow all users to read the repo. There are two ways a |
|
207 | 207 | user can be denied read access: (1) deny_read is not empty, and the |
|
208 | 208 | user is unauthenticated or deny_read contains user (or *), and (2) |
|
209 | 209 | allow_read is not empty and the user is not in allow_read. Return True |
|
210 | 210 | if user is allowed to read the repo, else return False.""" |
|
211 | 211 | |
|
212 | 212 | user = wsgireq.env.get('REMOTE_USER') |
|
213 | 213 | |
|
214 | 214 | deny_read = ui.configlist('web', 'deny_read', untrusted=True) |
|
215 | 215 | if deny_read and (not user or ismember(ui, user, deny_read)): |
|
216 | 216 | return False |
|
217 | 217 | |
|
218 | 218 | allow_read = ui.configlist('web', 'allow_read', untrusted=True) |
|
219 | 219 | # by default, allow reading if no allow_read option has been set |
|
220 | 220 | if (not allow_read) or ismember(ui, user, allow_read): |
|
221 | 221 | return True |
|
222 | 222 | |
|
223 | 223 | return False |
|
224 | 224 | |
|
225 | 225 | def run_wsgi(self, wsgireq): |
|
226 | 226 | profile = self.ui.configbool('profiling', 'enabled') |
|
227 | 227 | with profiling.profile(self.ui, enabled=profile): |
|
228 | 228 | for r in self._runwsgi(wsgireq): |
|
229 | 229 | yield r |
|
230 | 230 | |
|
231 | 231 | def _runwsgi(self, wsgireq): |
|
232 | 232 | try: |
|
233 | 233 | self.refresh() |
|
234 | 234 | |
|
235 | 235 | csp, nonce = cspvalues(self.ui) |
|
236 | 236 | if csp: |
|
237 | 237 | wsgireq.headers.append(('Content-Security-Policy', csp)) |
|
238 | 238 | |
|
239 | 239 | virtual = wsgireq.env.get("PATH_INFO", "").strip('/') |
|
240 | 240 | tmpl = self.templater(wsgireq, nonce) |
|
241 | 241 | ctype = tmpl('mimetype', encoding=encoding.encoding) |
|
242 | 242 | ctype = templater.stringify(ctype) |
|
243 | 243 | |
|
244 | 244 | # a static file |
|
245 | 245 | if virtual.startswith('static/') or 'static' in wsgireq.form: |
|
246 | 246 | if virtual.startswith('static/'): |
|
247 | 247 | fname = virtual[7:] |
|
248 | 248 | else: |
|
249 | 249 | fname = wsgireq.form['static'][0] |
|
250 | 250 | static = self.ui.config("web", "static", None, |
|
251 | 251 | untrusted=False) |
|
252 | 252 | if not static: |
|
253 | 253 | tp = self.templatepath or templater.templatepaths() |
|
254 | 254 | if isinstance(tp, str): |
|
255 | 255 | tp = [tp] |
|
256 | 256 | static = [os.path.join(p, 'static') for p in tp] |
|
257 | 257 | staticfile(static, fname, wsgireq) |
|
258 | 258 | return [] |
|
259 | 259 | |
|
260 | 260 | # top-level index |
|
261 | 261 | |
|
262 | 262 | repos = dict(self.repos) |
|
263 | 263 | |
|
264 | 264 | if (not virtual or virtual == 'index') and virtual not in repos: |
|
265 | 265 | wsgireq.respond(HTTP_OK, ctype) |
|
266 | 266 | return self.makeindex(wsgireq, tmpl) |
|
267 | 267 | |
|
268 | 268 | # nested indexes and hgwebs |
|
269 | 269 | |
|
270 | 270 | if virtual.endswith('/index') and virtual not in repos: |
|
271 | 271 | subdir = virtual[:-len('index')] |
|
272 | 272 | if any(r.startswith(subdir) for r in repos): |
|
273 | 273 | wsgireq.respond(HTTP_OK, ctype) |
|
274 | 274 | return self.makeindex(wsgireq, tmpl, subdir) |
|
275 | 275 | |
|
276 | 276 | def _virtualdirs(): |
|
277 | 277 | # Check the full virtual path, each parent, and the root ('') |
|
278 | 278 | if virtual != '': |
|
279 | 279 | yield virtual |
|
280 | 280 | |
|
281 | 281 | for p in util.finddirs(virtual): |
|
282 | 282 | yield p |
|
283 | 283 | |
|
284 | 284 | yield '' |
|
285 | 285 | |
|
286 | 286 | for virtualrepo in _virtualdirs(): |
|
287 | 287 | real = repos.get(virtualrepo) |
|
288 | 288 | if real: |
|
289 | 289 | wsgireq.env['REPO_NAME'] = virtualrepo |
|
290 | 290 | # We have to re-parse because of updated environment |
|
291 | 291 | # variable. |
|
292 | 292 | # TODO this is kind of hacky and we should have a better |
|
293 | 293 | # way of doing this than with REPO_NAME side-effects. |
|
294 |
wsgireq.req = requestmod.parserequestfromenv( |
|
|
294 | wsgireq.req = requestmod.parserequestfromenv( | |
|
295 | wsgireq.env, wsgireq.req.bodyfh) | |
|
295 | 296 | try: |
|
296 | 297 | # ensure caller gets private copy of ui |
|
297 | 298 | repo = hg.repository(self.ui.copy(), real) |
|
298 | 299 | return hgweb_mod.hgweb(repo).run_wsgi(wsgireq) |
|
299 | 300 | except IOError as inst: |
|
300 | 301 | msg = encoding.strtolocal(inst.strerror) |
|
301 | 302 | raise ErrorResponse(HTTP_SERVER_ERROR, msg) |
|
302 | 303 | except error.RepoError as inst: |
|
303 | 304 | raise ErrorResponse(HTTP_SERVER_ERROR, bytes(inst)) |
|
304 | 305 | |
|
305 | 306 | # browse subdirectories |
|
306 | 307 | subdir = virtual + '/' |
|
307 | 308 | if [r for r in repos if r.startswith(subdir)]: |
|
308 | 309 | wsgireq.respond(HTTP_OK, ctype) |
|
309 | 310 | return self.makeindex(wsgireq, tmpl, subdir) |
|
310 | 311 | |
|
311 | 312 | # prefixes not found |
|
312 | 313 | wsgireq.respond(HTTP_NOT_FOUND, ctype) |
|
313 | 314 | return tmpl("notfound", repo=virtual) |
|
314 | 315 | |
|
315 | 316 | except ErrorResponse as err: |
|
316 | 317 | wsgireq.respond(err, ctype) |
|
317 | 318 | return tmpl('error', error=err.message or '') |
|
318 | 319 | finally: |
|
319 | 320 | tmpl = None |
|
320 | 321 | |
|
321 | 322 | def makeindex(self, wsgireq, tmpl, subdir=""): |
|
322 | 323 | |
|
323 | 324 | def archivelist(ui, nodeid, url): |
|
324 | 325 | allowed = ui.configlist("web", "allow_archive", untrusted=True) |
|
325 | 326 | archives = [] |
|
326 | 327 | for typ, spec in hgweb_mod.archivespecs.iteritems(): |
|
327 | 328 | if typ in allowed or ui.configbool("web", "allow" + typ, |
|
328 | 329 | untrusted=True): |
|
329 | 330 | archives.append({"type": typ, "extension": spec[2], |
|
330 | 331 | "node": nodeid, "url": url}) |
|
331 | 332 | return archives |
|
332 | 333 | |
|
333 | 334 | def rawentries(subdir="", **map): |
|
334 | 335 | |
|
335 | 336 | descend = self.ui.configbool('web', 'descend') |
|
336 | 337 | collapse = self.ui.configbool('web', 'collapse') |
|
337 | 338 | seenrepos = set() |
|
338 | 339 | seendirs = set() |
|
339 | 340 | for name, path in self.repos: |
|
340 | 341 | |
|
341 | 342 | if not name.startswith(subdir): |
|
342 | 343 | continue |
|
343 | 344 | name = name[len(subdir):] |
|
344 | 345 | directory = False |
|
345 | 346 | |
|
346 | 347 | if '/' in name: |
|
347 | 348 | if not descend: |
|
348 | 349 | continue |
|
349 | 350 | |
|
350 | 351 | nameparts = name.split('/') |
|
351 | 352 | rootname = nameparts[0] |
|
352 | 353 | |
|
353 | 354 | if not collapse: |
|
354 | 355 | pass |
|
355 | 356 | elif rootname in seendirs: |
|
356 | 357 | continue |
|
357 | 358 | elif rootname in seenrepos: |
|
358 | 359 | pass |
|
359 | 360 | else: |
|
360 | 361 | directory = True |
|
361 | 362 | name = rootname |
|
362 | 363 | |
|
363 | 364 | # redefine the path to refer to the directory |
|
364 | 365 | discarded = '/'.join(nameparts[1:]) |
|
365 | 366 | |
|
366 | 367 | # remove name parts plus accompanying slash |
|
367 | 368 | path = path[:-len(discarded) - 1] |
|
368 | 369 | |
|
369 | 370 | try: |
|
370 | 371 | r = hg.repository(self.ui, path) |
|
371 | 372 | directory = False |
|
372 | 373 | except (IOError, error.RepoError): |
|
373 | 374 | pass |
|
374 | 375 | |
|
375 | 376 | parts = [name] |
|
376 | 377 | parts.insert(0, '/' + subdir.rstrip('/')) |
|
377 | 378 | if wsgireq.env['SCRIPT_NAME']: |
|
378 | 379 | parts.insert(0, wsgireq.env['SCRIPT_NAME']) |
|
379 | 380 | url = re.sub(r'/+', '/', '/'.join(parts) + '/') |
|
380 | 381 | |
|
381 | 382 | # show either a directory entry or a repository |
|
382 | 383 | if directory: |
|
383 | 384 | # get the directory's time information |
|
384 | 385 | try: |
|
385 | 386 | d = (get_mtime(path), dateutil.makedate()[1]) |
|
386 | 387 | except OSError: |
|
387 | 388 | continue |
|
388 | 389 | |
|
389 | 390 | # add '/' to the name to make it obvious that |
|
390 | 391 | # the entry is a directory, not a regular repository |
|
391 | 392 | row = {'contact': "", |
|
392 | 393 | 'contact_sort': "", |
|
393 | 394 | 'name': name + '/', |
|
394 | 395 | 'name_sort': name, |
|
395 | 396 | 'url': url, |
|
396 | 397 | 'description': "", |
|
397 | 398 | 'description_sort': "", |
|
398 | 399 | 'lastchange': d, |
|
399 | 400 | 'lastchange_sort': d[1]-d[0], |
|
400 | 401 | 'archives': [], |
|
401 | 402 | 'isdirectory': True, |
|
402 | 403 | 'labels': [], |
|
403 | 404 | } |
|
404 | 405 | |
|
405 | 406 | seendirs.add(name) |
|
406 | 407 | yield row |
|
407 | 408 | continue |
|
408 | 409 | |
|
409 | 410 | u = self.ui.copy() |
|
410 | 411 | try: |
|
411 | 412 | u.readconfig(os.path.join(path, '.hg', 'hgrc')) |
|
412 | 413 | except Exception as e: |
|
413 | 414 | u.warn(_('error reading %s/.hg/hgrc: %s\n') % (path, e)) |
|
414 | 415 | continue |
|
415 | 416 | def get(section, name, default=uimod._unset): |
|
416 | 417 | return u.config(section, name, default, untrusted=True) |
|
417 | 418 | |
|
418 | 419 | if u.configbool("web", "hidden", untrusted=True): |
|
419 | 420 | continue |
|
420 | 421 | |
|
421 | 422 | if not self.read_allowed(u, wsgireq): |
|
422 | 423 | continue |
|
423 | 424 | |
|
424 | 425 | # update time with local timezone |
|
425 | 426 | try: |
|
426 | 427 | r = hg.repository(self.ui, path) |
|
427 | 428 | except IOError: |
|
428 | 429 | u.warn(_('error accessing repository at %s\n') % path) |
|
429 | 430 | continue |
|
430 | 431 | except error.RepoError: |
|
431 | 432 | u.warn(_('error accessing repository at %s\n') % path) |
|
432 | 433 | continue |
|
433 | 434 | try: |
|
434 | 435 | d = (get_mtime(r.spath), dateutil.makedate()[1]) |
|
435 | 436 | except OSError: |
|
436 | 437 | continue |
|
437 | 438 | |
|
438 | 439 | contact = get_contact(get) |
|
439 | 440 | description = get("web", "description") |
|
440 | 441 | seenrepos.add(name) |
|
441 | 442 | name = get("web", "name", name) |
|
442 | 443 | row = {'contact': contact or "unknown", |
|
443 | 444 | 'contact_sort': contact.upper() or "unknown", |
|
444 | 445 | 'name': name, |
|
445 | 446 | 'name_sort': name, |
|
446 | 447 | 'url': url, |
|
447 | 448 | 'description': description or "unknown", |
|
448 | 449 | 'description_sort': description.upper() or "unknown", |
|
449 | 450 | 'lastchange': d, |
|
450 | 451 | 'lastchange_sort': d[1]-d[0], |
|
451 | 452 | 'archives': archivelist(u, "tip", url), |
|
452 | 453 | 'isdirectory': None, |
|
453 | 454 | 'labels': u.configlist('web', 'labels', untrusted=True), |
|
454 | 455 | } |
|
455 | 456 | |
|
456 | 457 | yield row |
|
457 | 458 | |
|
458 | 459 | sortdefault = None, False |
|
459 | 460 | def entries(sortcolumn="", descending=False, subdir="", **map): |
|
460 | 461 | rows = rawentries(subdir=subdir, **map) |
|
461 | 462 | |
|
462 | 463 | if sortcolumn and sortdefault != (sortcolumn, descending): |
|
463 | 464 | sortkey = '%s_sort' % sortcolumn |
|
464 | 465 | rows = sorted(rows, key=lambda x: x[sortkey], |
|
465 | 466 | reverse=descending) |
|
466 | 467 | for row, parity in zip(rows, paritygen(self.stripecount)): |
|
467 | 468 | row['parity'] = parity |
|
468 | 469 | yield row |
|
469 | 470 | |
|
470 | 471 | self.refresh() |
|
471 | 472 | sortable = ["name", "description", "contact", "lastchange"] |
|
472 | 473 | sortcolumn, descending = sortdefault |
|
473 | 474 | if 'sort' in wsgireq.form: |
|
474 | 475 | sortcolumn = wsgireq.form['sort'][0] |
|
475 | 476 | descending = sortcolumn.startswith('-') |
|
476 | 477 | if descending: |
|
477 | 478 | sortcolumn = sortcolumn[1:] |
|
478 | 479 | if sortcolumn not in sortable: |
|
479 | 480 | sortcolumn = "" |
|
480 | 481 | |
|
481 | 482 | sort = [("sort_%s" % column, |
|
482 | 483 | "%s%s" % ((not descending and column == sortcolumn) |
|
483 | 484 | and "-" or "", column)) |
|
484 | 485 | for column in sortable] |
|
485 | 486 | |
|
486 | 487 | self.refresh() |
|
487 | 488 | self.updatereqenv(wsgireq.env) |
|
488 | 489 | |
|
489 | 490 | return tmpl("index", entries=entries, subdir=subdir, |
|
490 | 491 | pathdef=hgweb_mod.makebreadcrumb('/' + subdir, self.prefix), |
|
491 | 492 | sortcolumn=sortcolumn, descending=descending, |
|
492 | 493 | **dict(sort)) |
|
493 | 494 | |
|
494 | 495 | def templater(self, wsgireq, nonce): |
|
495 | 496 | |
|
496 | 497 | def motd(**map): |
|
497 | 498 | if self.motd is not None: |
|
498 | 499 | yield self.motd |
|
499 | 500 | else: |
|
500 | 501 | yield config('web', 'motd') |
|
501 | 502 | |
|
502 | 503 | def config(section, name, default=uimod._unset, untrusted=True): |
|
503 | 504 | return self.ui.config(section, name, default, untrusted) |
|
504 | 505 | |
|
505 | 506 | self.updatereqenv(wsgireq.env) |
|
506 | 507 | |
|
507 | 508 | url = wsgireq.env.get('SCRIPT_NAME', '') |
|
508 | 509 | if not url.endswith('/'): |
|
509 | 510 | url += '/' |
|
510 | 511 | |
|
511 | 512 | vars = {} |
|
512 | 513 | styles, (style, mapfile) = hgweb_mod.getstyle(wsgireq, config, |
|
513 | 514 | self.templatepath) |
|
514 | 515 | if style == styles[0]: |
|
515 | 516 | vars['style'] = style |
|
516 | 517 | |
|
517 | 518 | sessionvars = webutil.sessionvars(vars, r'?') |
|
518 | 519 | logourl = config('web', 'logourl') |
|
519 | 520 | logoimg = config('web', 'logoimg') |
|
520 | 521 | staticurl = config('web', 'staticurl') or url + 'static/' |
|
521 | 522 | if not staticurl.endswith('/'): |
|
522 | 523 | staticurl += '/' |
|
523 | 524 | |
|
524 | 525 | defaults = { |
|
525 | 526 | "encoding": encoding.encoding, |
|
526 | 527 | "motd": motd, |
|
527 | 528 | "url": url, |
|
528 | 529 | "logourl": logourl, |
|
529 | 530 | "logoimg": logoimg, |
|
530 | 531 | "staticurl": staticurl, |
|
531 | 532 | "sessionvars": sessionvars, |
|
532 | 533 | "style": style, |
|
533 | 534 | "nonce": nonce, |
|
534 | 535 | } |
|
535 | 536 | tmpl = templater.templater.frommapfile(mapfile, defaults=defaults) |
|
536 | 537 | return tmpl |
|
537 | 538 | |
|
538 | 539 | def updatereqenv(self, env): |
|
539 | 540 | if self._baseurl is not None: |
|
540 | 541 | name, port, path = geturlcgivars(self._baseurl, env['SERVER_PORT']) |
|
541 | 542 | env['SERVER_NAME'] = name |
|
542 | 543 | env['SERVER_PORT'] = port |
|
543 | 544 | env['SCRIPT_NAME'] = path |
@@ -1,363 +1,374 | |||
|
1 | 1 | # hgweb/request.py - An http request from either CGI or the standalone server. |
|
2 | 2 | # |
|
3 | 3 | # Copyright 21 May 2005 - (c) 2005 Jake Edge <jake@edge2.net> |
|
4 | 4 | # Copyright 2005, 2006 Matt Mackall <mpm@selenic.com> |
|
5 | 5 | # |
|
6 | 6 | # This software may be used and distributed according to the terms of the |
|
7 | 7 | # GNU General Public License version 2 or any later version. |
|
8 | 8 | |
|
9 | 9 | from __future__ import absolute_import |
|
10 | 10 | |
|
11 | 11 | import cgi |
|
12 | 12 | import errno |
|
13 | 13 | import socket |
|
14 | 14 | import wsgiref.headers as wsgiheaders |
|
15 | 15 | #import wsgiref.validate |
|
16 | 16 | |
|
17 | 17 | from .common import ( |
|
18 | 18 | ErrorResponse, |
|
19 | 19 | HTTP_NOT_MODIFIED, |
|
20 | 20 | statusmessage, |
|
21 | 21 | ) |
|
22 | 22 | |
|
23 | 23 | from ..thirdparty import ( |
|
24 | 24 | attr, |
|
25 | 25 | ) |
|
26 | 26 | from .. import ( |
|
27 | 27 | pycompat, |
|
28 | 28 | util, |
|
29 | 29 | ) |
|
30 | 30 | |
|
31 | 31 | shortcuts = { |
|
32 | 32 | 'cl': [('cmd', ['changelog']), ('rev', None)], |
|
33 | 33 | 'sl': [('cmd', ['shortlog']), ('rev', None)], |
|
34 | 34 | 'cs': [('cmd', ['changeset']), ('node', None)], |
|
35 | 35 | 'f': [('cmd', ['file']), ('filenode', None)], |
|
36 | 36 | 'fl': [('cmd', ['filelog']), ('filenode', None)], |
|
37 | 37 | 'fd': [('cmd', ['filediff']), ('node', None)], |
|
38 | 38 | 'fa': [('cmd', ['annotate']), ('filenode', None)], |
|
39 | 39 | 'mf': [('cmd', ['manifest']), ('manifest', None)], |
|
40 | 40 | 'ca': [('cmd', ['archive']), ('node', None)], |
|
41 | 41 | 'tags': [('cmd', ['tags'])], |
|
42 | 42 | 'tip': [('cmd', ['changeset']), ('node', ['tip'])], |
|
43 | 43 | 'static': [('cmd', ['static']), ('file', None)] |
|
44 | 44 | } |
|
45 | 45 | |
|
46 | 46 | def normalize(form): |
|
47 | 47 | # first expand the shortcuts |
|
48 | 48 | for k in shortcuts: |
|
49 | 49 | if k in form: |
|
50 | 50 | for name, value in shortcuts[k]: |
|
51 | 51 | if value is None: |
|
52 | 52 | value = form[k] |
|
53 | 53 | form[name] = value |
|
54 | 54 | del form[k] |
|
55 | 55 | # And strip the values |
|
56 | 56 | bytesform = {} |
|
57 | 57 | for k, v in form.iteritems(): |
|
58 | 58 | bytesform[pycompat.bytesurl(k)] = [ |
|
59 | 59 | pycompat.bytesurl(i.strip()) for i in v] |
|
60 | 60 | return bytesform |
|
61 | 61 | |
|
62 | 62 | @attr.s(frozen=True) |
|
63 | 63 | class parsedrequest(object): |
|
64 |
"""Represents a parsed WSGI request |
|
|
64 | """Represents a parsed WSGI request. | |
|
65 | ||
|
66 | Contains both parsed parameters as well as a handle on the input stream. | |
|
67 | """ | |
|
65 | 68 | |
|
66 | 69 | # Request method. |
|
67 | 70 | method = attr.ib() |
|
68 | 71 | # Full URL for this request. |
|
69 | 72 | url = attr.ib() |
|
70 | 73 | # URL without any path components. Just <proto>://<host><port>. |
|
71 | 74 | baseurl = attr.ib() |
|
72 | 75 | # Advertised URL. Like ``url`` and ``baseurl`` but uses SERVER_NAME instead |
|
73 | 76 | # of HTTP: Host header for hostname. This is likely what clients used. |
|
74 | 77 | advertisedurl = attr.ib() |
|
75 | 78 | advertisedbaseurl = attr.ib() |
|
76 | 79 | # WSGI application path. |
|
77 | 80 | apppath = attr.ib() |
|
78 | 81 | # List of path parts to be used for dispatch. |
|
79 | 82 | dispatchparts = attr.ib() |
|
80 | 83 | # URL path component (no query string) used for dispatch. |
|
81 | 84 | dispatchpath = attr.ib() |
|
82 | 85 | # Whether there is a path component to this request. This can be true |
|
83 | 86 | # when ``dispatchpath`` is empty due to REPO_NAME muckery. |
|
84 | 87 | havepathinfo = attr.ib() |
|
85 | 88 | # Raw query string (part after "?" in URL). |
|
86 | 89 | querystring = attr.ib() |
|
87 | 90 | # List of 2-tuples of query string arguments. |
|
88 | 91 | querystringlist = attr.ib() |
|
89 | 92 | # Dict of query string arguments. Values are lists with at least 1 item. |
|
90 | 93 | querystringdict = attr.ib() |
|
91 | 94 | # wsgiref.headers.Headers instance. Operates like a dict with case |
|
92 | 95 | # insensitive keys. |
|
93 | 96 | headers = attr.ib() |
|
97 | # Request body input stream. | |
|
98 | bodyfh = attr.ib() | |
|
94 | 99 | |
|
95 | def parserequestfromenv(env): | |
|
100 | def parserequestfromenv(env, bodyfh): | |
|
96 | 101 | """Parse URL components from environment variables. |
|
97 | 102 | |
|
98 | 103 | WSGI defines request attributes via environment variables. This function |
|
99 | 104 | parses the environment variables into a data structure. |
|
100 | 105 | """ |
|
101 | 106 | # PEP-0333 defines the WSGI spec and is a useful reference for this code. |
|
102 | 107 | |
|
103 | 108 | # We first validate that the incoming object conforms with the WSGI spec. |
|
104 | 109 | # We only want to be dealing with spec-conforming WSGI implementations. |
|
105 | 110 | # TODO enable this once we fix internal violations. |
|
106 | 111 | #wsgiref.validate.check_environ(env) |
|
107 | 112 | |
|
108 | 113 | # PEP-0333 states that environment keys and values are native strings |
|
109 | 114 | # (bytes on Python 2 and str on Python 3). The code points for the Unicode |
|
110 | 115 | # strings on Python 3 must be between \00000-\000FF. We deal with bytes |
|
111 | 116 | # in Mercurial, so mass convert string keys and values to bytes. |
|
112 | 117 | if pycompat.ispy3: |
|
113 | 118 | env = {k.encode('latin-1'): v for k, v in env.iteritems()} |
|
114 | 119 | env = {k: v.encode('latin-1') if isinstance(v, str) else v |
|
115 | 120 | for k, v in env.iteritems()} |
|
116 | 121 | |
|
117 | 122 | # https://www.python.org/dev/peps/pep-0333/#environ-variables defines |
|
118 | 123 | # the environment variables. |
|
119 | 124 | # https://www.python.org/dev/peps/pep-0333/#url-reconstruction defines |
|
120 | 125 | # how URLs are reconstructed. |
|
121 | 126 | fullurl = env['wsgi.url_scheme'] + '://' |
|
122 | 127 | advertisedfullurl = fullurl |
|
123 | 128 | |
|
124 | 129 | def addport(s): |
|
125 | 130 | if env['wsgi.url_scheme'] == 'https': |
|
126 | 131 | if env['SERVER_PORT'] != '443': |
|
127 | 132 | s += ':' + env['SERVER_PORT'] |
|
128 | 133 | else: |
|
129 | 134 | if env['SERVER_PORT'] != '80': |
|
130 | 135 | s += ':' + env['SERVER_PORT'] |
|
131 | 136 | |
|
132 | 137 | return s |
|
133 | 138 | |
|
134 | 139 | if env.get('HTTP_HOST'): |
|
135 | 140 | fullurl += env['HTTP_HOST'] |
|
136 | 141 | else: |
|
137 | 142 | fullurl += env['SERVER_NAME'] |
|
138 | 143 | fullurl = addport(fullurl) |
|
139 | 144 | |
|
140 | 145 | advertisedfullurl += env['SERVER_NAME'] |
|
141 | 146 | advertisedfullurl = addport(advertisedfullurl) |
|
142 | 147 | |
|
143 | 148 | baseurl = fullurl |
|
144 | 149 | advertisedbaseurl = advertisedfullurl |
|
145 | 150 | |
|
146 | 151 | fullurl += util.urlreq.quote(env.get('SCRIPT_NAME', '')) |
|
147 | 152 | advertisedfullurl += util.urlreq.quote(env.get('SCRIPT_NAME', '')) |
|
148 | 153 | fullurl += util.urlreq.quote(env.get('PATH_INFO', '')) |
|
149 | 154 | advertisedfullurl += util.urlreq.quote(env.get('PATH_INFO', '')) |
|
150 | 155 | |
|
151 | 156 | if env.get('QUERY_STRING'): |
|
152 | 157 | fullurl += '?' + env['QUERY_STRING'] |
|
153 | 158 | advertisedfullurl += '?' + env['QUERY_STRING'] |
|
154 | 159 | |
|
155 | 160 | # When dispatching requests, we look at the URL components (PATH_INFO |
|
156 | 161 | # and QUERY_STRING) after the application root (SCRIPT_NAME). But hgwebdir |
|
157 | 162 | # has the concept of "virtual" repositories. This is defined via REPO_NAME. |
|
158 | 163 | # If REPO_NAME is defined, we append it to SCRIPT_NAME to form a new app |
|
159 | 164 | # root. We also exclude its path components from PATH_INFO when resolving |
|
160 | 165 | # the dispatch path. |
|
161 | 166 | |
|
162 | 167 | apppath = env['SCRIPT_NAME'] |
|
163 | 168 | |
|
164 | 169 | if env.get('REPO_NAME'): |
|
165 | 170 | if not apppath.endswith('/'): |
|
166 | 171 | apppath += '/' |
|
167 | 172 | |
|
168 | 173 | apppath += env.get('REPO_NAME') |
|
169 | 174 | |
|
170 | 175 | if 'PATH_INFO' in env: |
|
171 | 176 | dispatchparts = env['PATH_INFO'].strip('/').split('/') |
|
172 | 177 | |
|
173 | 178 | # Strip out repo parts. |
|
174 | 179 | repoparts = env.get('REPO_NAME', '').split('/') |
|
175 | 180 | if dispatchparts[:len(repoparts)] == repoparts: |
|
176 | 181 | dispatchparts = dispatchparts[len(repoparts):] |
|
177 | 182 | else: |
|
178 | 183 | dispatchparts = [] |
|
179 | 184 | |
|
180 | 185 | dispatchpath = '/'.join(dispatchparts) |
|
181 | 186 | |
|
182 | 187 | querystring = env.get('QUERY_STRING', '') |
|
183 | 188 | |
|
184 | 189 | # We store as a list so we have ordering information. We also store as |
|
185 | 190 | # a dict to facilitate fast lookup. |
|
186 | 191 | querystringlist = util.urlreq.parseqsl(querystring, keep_blank_values=True) |
|
187 | 192 | |
|
188 | 193 | querystringdict = {} |
|
189 | 194 | for k, v in querystringlist: |
|
190 | 195 | if k in querystringdict: |
|
191 | 196 | querystringdict[k].append(v) |
|
192 | 197 | else: |
|
193 | 198 | querystringdict[k] = [v] |
|
194 | 199 | |
|
195 | 200 | # HTTP_* keys contain HTTP request headers. The Headers structure should |
|
196 | 201 | # perform case normalization for us. We just rewrite underscore to dash |
|
197 | 202 | # so keys match what likely went over the wire. |
|
198 | 203 | headers = [] |
|
199 | 204 | for k, v in env.iteritems(): |
|
200 | 205 | if k.startswith('HTTP_'): |
|
201 | 206 | headers.append((k[len('HTTP_'):].replace('_', '-'), v)) |
|
202 | 207 | |
|
203 | 208 | headers = wsgiheaders.Headers(headers) |
|
204 | 209 | |
|
205 | 210 | # This is kind of a lie because the HTTP header wasn't explicitly |
|
206 | 211 | # sent. But for all intents and purposes it should be OK to lie about |
|
207 | 212 | # this, since a consumer will either either value to determine how many |
|
208 | 213 | # bytes are available to read. |
|
209 | 214 | if 'CONTENT_LENGTH' in env and 'HTTP_CONTENT_LENGTH' not in env: |
|
210 | 215 | headers['Content-Length'] = env['CONTENT_LENGTH'] |
|
211 | 216 | |
|
217 | # TODO do this once we remove wsgirequest.inp, otherwise we could have | |
|
218 | # multiple readers from the underlying input stream. | |
|
219 | #bodyfh = env['wsgi.input'] | |
|
220 | #if 'Content-Length' in headers: | |
|
221 | # bodyfh = util.cappedreader(bodyfh, int(headers['Content-Length'])) | |
|
222 | ||
|
212 | 223 | return parsedrequest(method=env['REQUEST_METHOD'], |
|
213 | 224 | url=fullurl, baseurl=baseurl, |
|
214 | 225 | advertisedurl=advertisedfullurl, |
|
215 | 226 | advertisedbaseurl=advertisedbaseurl, |
|
216 | 227 | apppath=apppath, |
|
217 | 228 | dispatchparts=dispatchparts, dispatchpath=dispatchpath, |
|
218 | 229 | havepathinfo='PATH_INFO' in env, |
|
219 | 230 | querystring=querystring, |
|
220 | 231 | querystringlist=querystringlist, |
|
221 | 232 | querystringdict=querystringdict, |
|
222 |
headers=headers |
|
|
233 | headers=headers, | |
|
234 | bodyfh=bodyfh) | |
|
223 | 235 | |
|
224 | 236 | class wsgirequest(object): |
|
225 | 237 | """Higher-level API for a WSGI request. |
|
226 | 238 | |
|
227 | 239 | WSGI applications are invoked with 2 arguments. They are used to |
|
228 | 240 | instantiate instances of this class, which provides higher-level APIs |
|
229 | 241 | for obtaining request parameters, writing HTTP output, etc. |
|
230 | 242 | """ |
|
231 | 243 | def __init__(self, wsgienv, start_response): |
|
232 | 244 | version = wsgienv[r'wsgi.version'] |
|
233 | 245 | if (version < (1, 0)) or (version >= (2, 0)): |
|
234 | 246 | raise RuntimeError("Unknown and unsupported WSGI version %d.%d" |
|
235 | 247 | % version) |
|
236 | self.inp = wsgienv[r'wsgi.input'] | |
|
248 | ||
|
249 | inp = wsgienv[r'wsgi.input'] | |
|
237 | 250 | |
|
238 | 251 | if r'HTTP_CONTENT_LENGTH' in wsgienv: |
|
239 |
|
|
|
240 | int(wsgienv[r'HTTP_CONTENT_LENGTH'])) | |
|
252 | inp = util.cappedreader(inp, int(wsgienv[r'HTTP_CONTENT_LENGTH'])) | |
|
241 | 253 | elif r'CONTENT_LENGTH' in wsgienv: |
|
242 |
|
|
|
243 | int(wsgienv[r'CONTENT_LENGTH'])) | |
|
254 | inp = util.cappedreader(inp, int(wsgienv[r'CONTENT_LENGTH'])) | |
|
244 | 255 | |
|
245 | 256 | self.err = wsgienv[r'wsgi.errors'] |
|
246 | 257 | self.threaded = wsgienv[r'wsgi.multithread'] |
|
247 | 258 | self.multiprocess = wsgienv[r'wsgi.multiprocess'] |
|
248 | 259 | self.run_once = wsgienv[r'wsgi.run_once'] |
|
249 | 260 | self.env = wsgienv |
|
250 |
self.form = normalize(cgi.parse( |
|
|
261 | self.form = normalize(cgi.parse(inp, | |
|
251 | 262 | self.env, |
|
252 | 263 | keep_blank_values=1)) |
|
253 | 264 | self._start_response = start_response |
|
254 | 265 | self.server_write = None |
|
255 | 266 | self.headers = [] |
|
256 | 267 | |
|
257 | self.req = parserequestfromenv(wsgienv) | |
|
268 | self.req = parserequestfromenv(wsgienv, inp) | |
|
258 | 269 | |
|
259 | 270 | def respond(self, status, type, filename=None, body=None): |
|
260 | 271 | if not isinstance(type, str): |
|
261 | 272 | type = pycompat.sysstr(type) |
|
262 | 273 | if self._start_response is not None: |
|
263 | 274 | self.headers.append((r'Content-Type', type)) |
|
264 | 275 | if filename: |
|
265 | 276 | filename = (filename.rpartition('/')[-1] |
|
266 | 277 | .replace('\\', '\\\\').replace('"', '\\"')) |
|
267 | 278 | self.headers.append(('Content-Disposition', |
|
268 | 279 | 'inline; filename="%s"' % filename)) |
|
269 | 280 | if body is not None: |
|
270 | 281 | self.headers.append((r'Content-Length', str(len(body)))) |
|
271 | 282 | |
|
272 | 283 | for k, v in self.headers: |
|
273 | 284 | if not isinstance(v, str): |
|
274 | 285 | raise TypeError('header value must be string: %r' % (v,)) |
|
275 | 286 | |
|
276 | 287 | if isinstance(status, ErrorResponse): |
|
277 | 288 | self.headers.extend(status.headers) |
|
278 | 289 | if status.code == HTTP_NOT_MODIFIED: |
|
279 | 290 | # RFC 2616 Section 10.3.5: 304 Not Modified has cases where |
|
280 | 291 | # it MUST NOT include any headers other than these and no |
|
281 | 292 | # body |
|
282 | 293 | self.headers = [(k, v) for (k, v) in self.headers if |
|
283 | 294 | k in ('Date', 'ETag', 'Expires', |
|
284 | 295 | 'Cache-Control', 'Vary')] |
|
285 | 296 | status = statusmessage(status.code, pycompat.bytestr(status)) |
|
286 | 297 | elif status == 200: |
|
287 | 298 | status = '200 Script output follows' |
|
288 | 299 | elif isinstance(status, int): |
|
289 | 300 | status = statusmessage(status) |
|
290 | 301 | |
|
291 | 302 | # Various HTTP clients (notably httplib) won't read the HTTP |
|
292 | 303 | # response until the HTTP request has been sent in full. If servers |
|
293 | 304 | # (us) send a response before the HTTP request has been fully sent, |
|
294 | 305 | # the connection may deadlock because neither end is reading. |
|
295 | 306 | # |
|
296 | 307 | # We work around this by "draining" the request data before |
|
297 | 308 | # sending any response in some conditions. |
|
298 | 309 | drain = False |
|
299 | 310 | close = False |
|
300 | 311 | |
|
301 | 312 | # If the client sent Expect: 100-continue, we assume it is smart |
|
302 | 313 | # enough to deal with the server sending a response before reading |
|
303 | 314 | # the request. (httplib doesn't do this.) |
|
304 | 315 | if self.env.get(r'HTTP_EXPECT', r'').lower() == r'100-continue': |
|
305 | 316 | pass |
|
306 | 317 | # Only tend to request methods that have bodies. Strictly speaking, |
|
307 | 318 | # we should sniff for a body. But this is fine for our existing |
|
308 | 319 | # WSGI applications. |
|
309 | 320 | elif self.env[r'REQUEST_METHOD'] not in (r'POST', r'PUT'): |
|
310 | 321 | pass |
|
311 | 322 | else: |
|
312 | 323 | # If we don't know how much data to read, there's no guarantee |
|
313 | 324 | # that we can drain the request responsibly. The WSGI |
|
314 | 325 | # specification only says that servers *should* ensure the |
|
315 | 326 | # input stream doesn't overrun the actual request. So there's |
|
316 | 327 | # no guarantee that reading until EOF won't corrupt the stream |
|
317 | 328 | # state. |
|
318 |
if not isinstance(self. |
|
|
329 | if not isinstance(self.req.bodyfh, util.cappedreader): | |
|
319 | 330 | close = True |
|
320 | 331 | else: |
|
321 | 332 | # We /could/ only drain certain HTTP response codes. But 200 |
|
322 | 333 | # and non-200 wire protocol responses both require draining. |
|
323 | 334 | # Since we have a capped reader in place for all situations |
|
324 | 335 | # where we drain, it is safe to read from that stream. We'll |
|
325 | 336 | # either do a drain or no-op if we're already at EOF. |
|
326 | 337 | drain = True |
|
327 | 338 | |
|
328 | 339 | if close: |
|
329 | 340 | self.headers.append((r'Connection', r'Close')) |
|
330 | 341 | |
|
331 | 342 | if drain: |
|
332 |
assert isinstance(self. |
|
|
343 | assert isinstance(self.req.bodyfh, util.cappedreader) | |
|
333 | 344 | while True: |
|
334 |
chunk = self. |
|
|
345 | chunk = self.req.bodyfh.read(32768) | |
|
335 | 346 | if not chunk: |
|
336 | 347 | break |
|
337 | 348 | |
|
338 | 349 | self.server_write = self._start_response( |
|
339 | 350 | pycompat.sysstr(status), self.headers) |
|
340 | 351 | self._start_response = None |
|
341 | 352 | self.headers = [] |
|
342 | 353 | if body is not None: |
|
343 | 354 | self.write(body) |
|
344 | 355 | self.server_write = None |
|
345 | 356 | |
|
346 | 357 | def write(self, thing): |
|
347 | 358 | if thing: |
|
348 | 359 | try: |
|
349 | 360 | self.server_write(thing) |
|
350 | 361 | except socket.error as inst: |
|
351 | 362 | if inst[0] != errno.ECONNRESET: |
|
352 | 363 | raise |
|
353 | 364 | |
|
354 | 365 | def flush(self): |
|
355 | 366 | return None |
|
356 | 367 | |
|
357 | 368 | def wsgiapplication(app_maker): |
|
358 | 369 | '''For compatibility with old CGI scripts. A plain hgweb() or hgwebdir() |
|
359 | 370 | can and should now be used as a WSGI application.''' |
|
360 | 371 | application = app_maker() |
|
361 | 372 | def run_wsgi(env, respond): |
|
362 | 373 | return application(env, respond) |
|
363 | 374 | return run_wsgi |
@@ -1,649 +1,649 | |||
|
1 | 1 | # Copyright 21 May 2005 - (c) 2005 Jake Edge <jake@edge2.net> |
|
2 | 2 | # Copyright 2005-2007 Matt Mackall <mpm@selenic.com> |
|
3 | 3 | # |
|
4 | 4 | # This software may be used and distributed according to the terms of the |
|
5 | 5 | # GNU General Public License version 2 or any later version. |
|
6 | 6 | |
|
7 | 7 | from __future__ import absolute_import |
|
8 | 8 | |
|
9 | 9 | import contextlib |
|
10 | 10 | import struct |
|
11 | 11 | import sys |
|
12 | 12 | import threading |
|
13 | 13 | |
|
14 | 14 | from .i18n import _ |
|
15 | 15 | from . import ( |
|
16 | 16 | encoding, |
|
17 | 17 | error, |
|
18 | 18 | hook, |
|
19 | 19 | pycompat, |
|
20 | 20 | util, |
|
21 | 21 | wireproto, |
|
22 | 22 | wireprototypes, |
|
23 | 23 | ) |
|
24 | 24 | |
|
25 | 25 | stringio = util.stringio |
|
26 | 26 | |
|
27 | 27 | urlerr = util.urlerr |
|
28 | 28 | urlreq = util.urlreq |
|
29 | 29 | |
|
30 | 30 | HTTP_OK = 200 |
|
31 | 31 | |
|
32 | 32 | HGTYPE = 'application/mercurial-0.1' |
|
33 | 33 | HGTYPE2 = 'application/mercurial-0.2' |
|
34 | 34 | HGERRTYPE = 'application/hg-error' |
|
35 | 35 | |
|
36 | 36 | SSHV1 = wireprototypes.SSHV1 |
|
37 | 37 | SSHV2 = wireprototypes.SSHV2 |
|
38 | 38 | |
|
39 | 39 | def decodevaluefromheaders(req, headerprefix): |
|
40 | 40 | """Decode a long value from multiple HTTP request headers. |
|
41 | 41 | |
|
42 | 42 | Returns the value as a bytes, not a str. |
|
43 | 43 | """ |
|
44 | 44 | chunks = [] |
|
45 | 45 | i = 1 |
|
46 | 46 | while True: |
|
47 | 47 | v = req.headers.get(b'%s-%d' % (headerprefix, i)) |
|
48 | 48 | if v is None: |
|
49 | 49 | break |
|
50 | 50 | chunks.append(pycompat.bytesurl(v)) |
|
51 | 51 | i += 1 |
|
52 | 52 | |
|
53 | 53 | return ''.join(chunks) |
|
54 | 54 | |
|
55 | 55 | class httpv1protocolhandler(wireprototypes.baseprotocolhandler): |
|
56 | 56 | def __init__(self, wsgireq, req, ui, checkperm): |
|
57 | 57 | self._wsgireq = wsgireq |
|
58 | 58 | self._req = req |
|
59 | 59 | self._ui = ui |
|
60 | 60 | self._checkperm = checkperm |
|
61 | 61 | |
|
62 | 62 | @property |
|
63 | 63 | def name(self): |
|
64 | 64 | return 'http-v1' |
|
65 | 65 | |
|
66 | 66 | def getargs(self, args): |
|
67 | 67 | knownargs = self._args() |
|
68 | 68 | data = {} |
|
69 | 69 | keys = args.split() |
|
70 | 70 | for k in keys: |
|
71 | 71 | if k == '*': |
|
72 | 72 | star = {} |
|
73 | 73 | for key in knownargs.keys(): |
|
74 | 74 | if key != 'cmd' and key not in keys: |
|
75 | 75 | star[key] = knownargs[key][0] |
|
76 | 76 | data['*'] = star |
|
77 | 77 | else: |
|
78 | 78 | data[k] = knownargs[k][0] |
|
79 | 79 | return [data[k] for k in keys] |
|
80 | 80 | |
|
81 | 81 | def _args(self): |
|
82 | 82 | args = util.rapply(pycompat.bytesurl, self._wsgireq.form.copy()) |
|
83 | 83 | postlen = int(self._req.headers.get(b'X-HgArgs-Post', 0)) |
|
84 | 84 | if postlen: |
|
85 | 85 | args.update(urlreq.parseqs( |
|
86 |
self._ |
|
|
86 | self._req.bodyfh.read(postlen), keep_blank_values=True)) | |
|
87 | 87 | return args |
|
88 | 88 | |
|
89 | 89 | argvalue = decodevaluefromheaders(self._req, b'X-HgArg') |
|
90 | 90 | args.update(urlreq.parseqs(argvalue, keep_blank_values=True)) |
|
91 | 91 | return args |
|
92 | 92 | |
|
93 | 93 | def forwardpayload(self, fp): |
|
94 | 94 | # Existing clients *always* send Content-Length. |
|
95 | 95 | length = int(self._req.headers[b'Content-Length']) |
|
96 | 96 | |
|
97 | 97 | # If httppostargs is used, we need to read Content-Length |
|
98 | 98 | # minus the amount that was consumed by args. |
|
99 | 99 | length -= int(self._req.headers.get(b'X-HgArgs-Post', 0)) |
|
100 |
for s in util.filechunkiter(self._ |
|
|
100 | for s in util.filechunkiter(self._req.bodyfh, limit=length): | |
|
101 | 101 | fp.write(s) |
|
102 | 102 | |
|
103 | 103 | @contextlib.contextmanager |
|
104 | 104 | def mayberedirectstdio(self): |
|
105 | 105 | oldout = self._ui.fout |
|
106 | 106 | olderr = self._ui.ferr |
|
107 | 107 | |
|
108 | 108 | out = util.stringio() |
|
109 | 109 | |
|
110 | 110 | try: |
|
111 | 111 | self._ui.fout = out |
|
112 | 112 | self._ui.ferr = out |
|
113 | 113 | yield out |
|
114 | 114 | finally: |
|
115 | 115 | self._ui.fout = oldout |
|
116 | 116 | self._ui.ferr = olderr |
|
117 | 117 | |
|
118 | 118 | def client(self): |
|
119 | 119 | return 'remote:%s:%s:%s' % ( |
|
120 | 120 | self._wsgireq.env.get('wsgi.url_scheme') or 'http', |
|
121 | 121 | urlreq.quote(self._wsgireq.env.get('REMOTE_HOST', '')), |
|
122 | 122 | urlreq.quote(self._wsgireq.env.get('REMOTE_USER', ''))) |
|
123 | 123 | |
|
124 | 124 | def addcapabilities(self, repo, caps): |
|
125 | 125 | caps.append('httpheader=%d' % |
|
126 | 126 | repo.ui.configint('server', 'maxhttpheaderlen')) |
|
127 | 127 | if repo.ui.configbool('experimental', 'httppostargs'): |
|
128 | 128 | caps.append('httppostargs') |
|
129 | 129 | |
|
130 | 130 | # FUTURE advertise 0.2rx once support is implemented |
|
131 | 131 | # FUTURE advertise minrx and mintx after consulting config option |
|
132 | 132 | caps.append('httpmediatype=0.1rx,0.1tx,0.2tx') |
|
133 | 133 | |
|
134 | 134 | compengines = wireproto.supportedcompengines(repo.ui, util.SERVERROLE) |
|
135 | 135 | if compengines: |
|
136 | 136 | comptypes = ','.join(urlreq.quote(e.wireprotosupport().name) |
|
137 | 137 | for e in compengines) |
|
138 | 138 | caps.append('compression=%s' % comptypes) |
|
139 | 139 | |
|
140 | 140 | return caps |
|
141 | 141 | |
|
142 | 142 | def checkperm(self, perm): |
|
143 | 143 | return self._checkperm(perm) |
|
144 | 144 | |
|
145 | 145 | # This method exists mostly so that extensions like remotefilelog can |
|
146 | 146 | # disable a kludgey legacy method only over http. As of early 2018, |
|
147 | 147 | # there are no other known users, so with any luck we can discard this |
|
148 | 148 | # hook if remotefilelog becomes a first-party extension. |
|
149 | 149 | def iscmd(cmd): |
|
150 | 150 | return cmd in wireproto.commands |
|
151 | 151 | |
|
152 | 152 | def handlewsgirequest(rctx, wsgireq, req, checkperm): |
|
153 | 153 | """Possibly process a wire protocol request. |
|
154 | 154 | |
|
155 | 155 | If the current request is a wire protocol request, the request is |
|
156 | 156 | processed by this function. |
|
157 | 157 | |
|
158 | 158 | ``wsgireq`` is a ``wsgirequest`` instance. |
|
159 | 159 | ``req`` is a ``parsedrequest`` instance. |
|
160 | 160 | |
|
161 | 161 | Returns a 2-tuple of (bool, response) where the 1st element indicates |
|
162 | 162 | whether the request was handled and the 2nd element is a return |
|
163 | 163 | value for a WSGI application (often a generator of bytes). |
|
164 | 164 | """ |
|
165 | 165 | # Avoid cycle involving hg module. |
|
166 | 166 | from .hgweb import common as hgwebcommon |
|
167 | 167 | |
|
168 | 168 | repo = rctx.repo |
|
169 | 169 | |
|
170 | 170 | # HTTP version 1 wire protocol requests are denoted by a "cmd" query |
|
171 | 171 | # string parameter. If it isn't present, this isn't a wire protocol |
|
172 | 172 | # request. |
|
173 | 173 | if 'cmd' not in req.querystringdict: |
|
174 | 174 | return False, None |
|
175 | 175 | |
|
176 | 176 | cmd = req.querystringdict['cmd'][0] |
|
177 | 177 | |
|
178 | 178 | # The "cmd" request parameter is used by both the wire protocol and hgweb. |
|
179 | 179 | # While not all wire protocol commands are available for all transports, |
|
180 | 180 | # if we see a "cmd" value that resembles a known wire protocol command, we |
|
181 | 181 | # route it to a protocol handler. This is better than routing possible |
|
182 | 182 | # wire protocol requests to hgweb because it prevents hgweb from using |
|
183 | 183 | # known wire protocol commands and it is less confusing for machine |
|
184 | 184 | # clients. |
|
185 | 185 | if not iscmd(cmd): |
|
186 | 186 | return False, None |
|
187 | 187 | |
|
188 | 188 | # The "cmd" query string argument is only valid on the root path of the |
|
189 | 189 | # repo. e.g. ``/?cmd=foo``, ``/repo?cmd=foo``. URL paths within the repo |
|
190 | 190 | # like ``/blah?cmd=foo`` are not allowed. So don't recognize the request |
|
191 | 191 | # in this case. We send an HTTP 404 for backwards compatibility reasons. |
|
192 | 192 | if req.dispatchpath: |
|
193 | 193 | res = _handlehttperror( |
|
194 | 194 | hgwebcommon.ErrorResponse(hgwebcommon.HTTP_NOT_FOUND), wsgireq, |
|
195 | 195 | req) |
|
196 | 196 | |
|
197 | 197 | return True, res |
|
198 | 198 | |
|
199 | 199 | proto = httpv1protocolhandler(wsgireq, req, repo.ui, |
|
200 | 200 | lambda perm: checkperm(rctx, wsgireq, perm)) |
|
201 | 201 | |
|
202 | 202 | # The permissions checker should be the only thing that can raise an |
|
203 | 203 | # ErrorResponse. It is kind of a layer violation to catch an hgweb |
|
204 | 204 | # exception here. So consider refactoring into a exception type that |
|
205 | 205 | # is associated with the wire protocol. |
|
206 | 206 | try: |
|
207 | 207 | res = _callhttp(repo, wsgireq, req, proto, cmd) |
|
208 | 208 | except hgwebcommon.ErrorResponse as e: |
|
209 | 209 | res = _handlehttperror(e, wsgireq, req) |
|
210 | 210 | |
|
211 | 211 | return True, res |
|
212 | 212 | |
|
213 | 213 | def _httpresponsetype(ui, req, prefer_uncompressed): |
|
214 | 214 | """Determine the appropriate response type and compression settings. |
|
215 | 215 | |
|
216 | 216 | Returns a tuple of (mediatype, compengine, engineopts). |
|
217 | 217 | """ |
|
218 | 218 | # Determine the response media type and compression engine based |
|
219 | 219 | # on the request parameters. |
|
220 | 220 | protocaps = decodevaluefromheaders(req, 'X-HgProto').split(' ') |
|
221 | 221 | |
|
222 | 222 | if '0.2' in protocaps: |
|
223 | 223 | # All clients are expected to support uncompressed data. |
|
224 | 224 | if prefer_uncompressed: |
|
225 | 225 | return HGTYPE2, util._noopengine(), {} |
|
226 | 226 | |
|
227 | 227 | # Default as defined by wire protocol spec. |
|
228 | 228 | compformats = ['zlib', 'none'] |
|
229 | 229 | for cap in protocaps: |
|
230 | 230 | if cap.startswith('comp='): |
|
231 | 231 | compformats = cap[5:].split(',') |
|
232 | 232 | break |
|
233 | 233 | |
|
234 | 234 | # Now find an agreed upon compression format. |
|
235 | 235 | for engine in wireproto.supportedcompengines(ui, util.SERVERROLE): |
|
236 | 236 | if engine.wireprotosupport().name in compformats: |
|
237 | 237 | opts = {} |
|
238 | 238 | level = ui.configint('server', '%slevel' % engine.name()) |
|
239 | 239 | if level is not None: |
|
240 | 240 | opts['level'] = level |
|
241 | 241 | |
|
242 | 242 | return HGTYPE2, engine, opts |
|
243 | 243 | |
|
244 | 244 | # No mutually supported compression format. Fall back to the |
|
245 | 245 | # legacy protocol. |
|
246 | 246 | |
|
247 | 247 | # Don't allow untrusted settings because disabling compression or |
|
248 | 248 | # setting a very high compression level could lead to flooding |
|
249 | 249 | # the server's network or CPU. |
|
250 | 250 | opts = {'level': ui.configint('server', 'zliblevel')} |
|
251 | 251 | return HGTYPE, util.compengines['zlib'], opts |
|
252 | 252 | |
|
253 | 253 | def _callhttp(repo, wsgireq, req, proto, cmd): |
|
254 | 254 | def genversion2(gen, engine, engineopts): |
|
255 | 255 | # application/mercurial-0.2 always sends a payload header |
|
256 | 256 | # identifying the compression engine. |
|
257 | 257 | name = engine.wireprotosupport().name |
|
258 | 258 | assert 0 < len(name) < 256 |
|
259 | 259 | yield struct.pack('B', len(name)) |
|
260 | 260 | yield name |
|
261 | 261 | |
|
262 | 262 | for chunk in gen: |
|
263 | 263 | yield chunk |
|
264 | 264 | |
|
265 | 265 | if not wireproto.commands.commandavailable(cmd, proto): |
|
266 | 266 | wsgireq.respond(HTTP_OK, HGERRTYPE, |
|
267 | 267 | body=_('requested wire protocol command is not ' |
|
268 | 268 | 'available over HTTP')) |
|
269 | 269 | return [] |
|
270 | 270 | |
|
271 | 271 | proto.checkperm(wireproto.commands[cmd].permission) |
|
272 | 272 | |
|
273 | 273 | rsp = wireproto.dispatch(repo, proto, cmd) |
|
274 | 274 | |
|
275 | 275 | if isinstance(rsp, bytes): |
|
276 | 276 | wsgireq.respond(HTTP_OK, HGTYPE, body=rsp) |
|
277 | 277 | return [] |
|
278 | 278 | elif isinstance(rsp, wireprototypes.bytesresponse): |
|
279 | 279 | wsgireq.respond(HTTP_OK, HGTYPE, body=rsp.data) |
|
280 | 280 | return [] |
|
281 | 281 | elif isinstance(rsp, wireprototypes.streamreslegacy): |
|
282 | 282 | gen = rsp.gen |
|
283 | 283 | wsgireq.respond(HTTP_OK, HGTYPE) |
|
284 | 284 | return gen |
|
285 | 285 | elif isinstance(rsp, wireprototypes.streamres): |
|
286 | 286 | gen = rsp.gen |
|
287 | 287 | |
|
288 | 288 | # This code for compression should not be streamres specific. It |
|
289 | 289 | # is here because we only compress streamres at the moment. |
|
290 | 290 | mediatype, engine, engineopts = _httpresponsetype( |
|
291 | 291 | repo.ui, req, rsp.prefer_uncompressed) |
|
292 | 292 | gen = engine.compressstream(gen, engineopts) |
|
293 | 293 | |
|
294 | 294 | if mediatype == HGTYPE2: |
|
295 | 295 | gen = genversion2(gen, engine, engineopts) |
|
296 | 296 | |
|
297 | 297 | wsgireq.respond(HTTP_OK, mediatype) |
|
298 | 298 | return gen |
|
299 | 299 | elif isinstance(rsp, wireprototypes.pushres): |
|
300 | 300 | rsp = '%d\n%s' % (rsp.res, rsp.output) |
|
301 | 301 | wsgireq.respond(HTTP_OK, HGTYPE, body=rsp) |
|
302 | 302 | return [] |
|
303 | 303 | elif isinstance(rsp, wireprototypes.pusherr): |
|
304 | 304 | rsp = '0\n%s\n' % rsp.res |
|
305 | 305 | wsgireq.respond(HTTP_OK, HGTYPE, body=rsp) |
|
306 | 306 | return [] |
|
307 | 307 | elif isinstance(rsp, wireprototypes.ooberror): |
|
308 | 308 | rsp = rsp.message |
|
309 | 309 | wsgireq.respond(HTTP_OK, HGERRTYPE, body=rsp) |
|
310 | 310 | return [] |
|
311 | 311 | raise error.ProgrammingError('hgweb.protocol internal failure', rsp) |
|
312 | 312 | |
|
313 | 313 | def _handlehttperror(e, wsgireq, req): |
|
314 | 314 | """Called when an ErrorResponse is raised during HTTP request processing.""" |
|
315 | 315 | |
|
316 | 316 | # TODO This response body assumes the failed command was |
|
317 | 317 | # "unbundle." That assumption is not always valid. |
|
318 | 318 | wsgireq.respond(e, HGTYPE, body='0\n%s\n' % pycompat.bytestr(e)) |
|
319 | 319 | |
|
320 | 320 | return '' |
|
321 | 321 | |
|
322 | 322 | def _sshv1respondbytes(fout, value): |
|
323 | 323 | """Send a bytes response for protocol version 1.""" |
|
324 | 324 | fout.write('%d\n' % len(value)) |
|
325 | 325 | fout.write(value) |
|
326 | 326 | fout.flush() |
|
327 | 327 | |
|
328 | 328 | def _sshv1respondstream(fout, source): |
|
329 | 329 | write = fout.write |
|
330 | 330 | for chunk in source.gen: |
|
331 | 331 | write(chunk) |
|
332 | 332 | fout.flush() |
|
333 | 333 | |
|
334 | 334 | def _sshv1respondooberror(fout, ferr, rsp): |
|
335 | 335 | ferr.write(b'%s\n-\n' % rsp) |
|
336 | 336 | ferr.flush() |
|
337 | 337 | fout.write(b'\n') |
|
338 | 338 | fout.flush() |
|
339 | 339 | |
|
340 | 340 | class sshv1protocolhandler(wireprototypes.baseprotocolhandler): |
|
341 | 341 | """Handler for requests services via version 1 of SSH protocol.""" |
|
342 | 342 | def __init__(self, ui, fin, fout): |
|
343 | 343 | self._ui = ui |
|
344 | 344 | self._fin = fin |
|
345 | 345 | self._fout = fout |
|
346 | 346 | |
|
347 | 347 | @property |
|
348 | 348 | def name(self): |
|
349 | 349 | return wireprototypes.SSHV1 |
|
350 | 350 | |
|
351 | 351 | def getargs(self, args): |
|
352 | 352 | data = {} |
|
353 | 353 | keys = args.split() |
|
354 | 354 | for n in xrange(len(keys)): |
|
355 | 355 | argline = self._fin.readline()[:-1] |
|
356 | 356 | arg, l = argline.split() |
|
357 | 357 | if arg not in keys: |
|
358 | 358 | raise error.Abort(_("unexpected parameter %r") % arg) |
|
359 | 359 | if arg == '*': |
|
360 | 360 | star = {} |
|
361 | 361 | for k in xrange(int(l)): |
|
362 | 362 | argline = self._fin.readline()[:-1] |
|
363 | 363 | arg, l = argline.split() |
|
364 | 364 | val = self._fin.read(int(l)) |
|
365 | 365 | star[arg] = val |
|
366 | 366 | data['*'] = star |
|
367 | 367 | else: |
|
368 | 368 | val = self._fin.read(int(l)) |
|
369 | 369 | data[arg] = val |
|
370 | 370 | return [data[k] for k in keys] |
|
371 | 371 | |
|
372 | 372 | def forwardpayload(self, fpout): |
|
373 | 373 | # We initially send an empty response. This tells the client it is |
|
374 | 374 | # OK to start sending data. If a client sees any other response, it |
|
375 | 375 | # interprets it as an error. |
|
376 | 376 | _sshv1respondbytes(self._fout, b'') |
|
377 | 377 | |
|
378 | 378 | # The file is in the form: |
|
379 | 379 | # |
|
380 | 380 | # <chunk size>\n<chunk> |
|
381 | 381 | # ... |
|
382 | 382 | # 0\n |
|
383 | 383 | count = int(self._fin.readline()) |
|
384 | 384 | while count: |
|
385 | 385 | fpout.write(self._fin.read(count)) |
|
386 | 386 | count = int(self._fin.readline()) |
|
387 | 387 | |
|
388 | 388 | @contextlib.contextmanager |
|
389 | 389 | def mayberedirectstdio(self): |
|
390 | 390 | yield None |
|
391 | 391 | |
|
392 | 392 | def client(self): |
|
393 | 393 | client = encoding.environ.get('SSH_CLIENT', '').split(' ', 1)[0] |
|
394 | 394 | return 'remote:ssh:' + client |
|
395 | 395 | |
|
396 | 396 | def addcapabilities(self, repo, caps): |
|
397 | 397 | return caps |
|
398 | 398 | |
|
399 | 399 | def checkperm(self, perm): |
|
400 | 400 | pass |
|
401 | 401 | |
|
402 | 402 | class sshv2protocolhandler(sshv1protocolhandler): |
|
403 | 403 | """Protocol handler for version 2 of the SSH protocol.""" |
|
404 | 404 | |
|
405 | 405 | @property |
|
406 | 406 | def name(self): |
|
407 | 407 | return wireprototypes.SSHV2 |
|
408 | 408 | |
|
409 | 409 | def _runsshserver(ui, repo, fin, fout, ev): |
|
410 | 410 | # This function operates like a state machine of sorts. The following |
|
411 | 411 | # states are defined: |
|
412 | 412 | # |
|
413 | 413 | # protov1-serving |
|
414 | 414 | # Server is in protocol version 1 serving mode. Commands arrive on |
|
415 | 415 | # new lines. These commands are processed in this state, one command |
|
416 | 416 | # after the other. |
|
417 | 417 | # |
|
418 | 418 | # protov2-serving |
|
419 | 419 | # Server is in protocol version 2 serving mode. |
|
420 | 420 | # |
|
421 | 421 | # upgrade-initial |
|
422 | 422 | # The server is going to process an upgrade request. |
|
423 | 423 | # |
|
424 | 424 | # upgrade-v2-filter-legacy-handshake |
|
425 | 425 | # The protocol is being upgraded to version 2. The server is expecting |
|
426 | 426 | # the legacy handshake from version 1. |
|
427 | 427 | # |
|
428 | 428 | # upgrade-v2-finish |
|
429 | 429 | # The upgrade to version 2 of the protocol is imminent. |
|
430 | 430 | # |
|
431 | 431 | # shutdown |
|
432 | 432 | # The server is shutting down, possibly in reaction to a client event. |
|
433 | 433 | # |
|
434 | 434 | # And here are their transitions: |
|
435 | 435 | # |
|
436 | 436 | # protov1-serving -> shutdown |
|
437 | 437 | # When server receives an empty request or encounters another |
|
438 | 438 | # error. |
|
439 | 439 | # |
|
440 | 440 | # protov1-serving -> upgrade-initial |
|
441 | 441 | # An upgrade request line was seen. |
|
442 | 442 | # |
|
443 | 443 | # upgrade-initial -> upgrade-v2-filter-legacy-handshake |
|
444 | 444 | # Upgrade to version 2 in progress. Server is expecting to |
|
445 | 445 | # process a legacy handshake. |
|
446 | 446 | # |
|
447 | 447 | # upgrade-v2-filter-legacy-handshake -> shutdown |
|
448 | 448 | # Client did not fulfill upgrade handshake requirements. |
|
449 | 449 | # |
|
450 | 450 | # upgrade-v2-filter-legacy-handshake -> upgrade-v2-finish |
|
451 | 451 | # Client fulfilled version 2 upgrade requirements. Finishing that |
|
452 | 452 | # upgrade. |
|
453 | 453 | # |
|
454 | 454 | # upgrade-v2-finish -> protov2-serving |
|
455 | 455 | # Protocol upgrade to version 2 complete. Server can now speak protocol |
|
456 | 456 | # version 2. |
|
457 | 457 | # |
|
458 | 458 | # protov2-serving -> protov1-serving |
|
459 | 459 | # Ths happens by default since protocol version 2 is the same as |
|
460 | 460 | # version 1 except for the handshake. |
|
461 | 461 | |
|
462 | 462 | state = 'protov1-serving' |
|
463 | 463 | proto = sshv1protocolhandler(ui, fin, fout) |
|
464 | 464 | protoswitched = False |
|
465 | 465 | |
|
466 | 466 | while not ev.is_set(): |
|
467 | 467 | if state == 'protov1-serving': |
|
468 | 468 | # Commands are issued on new lines. |
|
469 | 469 | request = fin.readline()[:-1] |
|
470 | 470 | |
|
471 | 471 | # Empty lines signal to terminate the connection. |
|
472 | 472 | if not request: |
|
473 | 473 | state = 'shutdown' |
|
474 | 474 | continue |
|
475 | 475 | |
|
476 | 476 | # It looks like a protocol upgrade request. Transition state to |
|
477 | 477 | # handle it. |
|
478 | 478 | if request.startswith(b'upgrade '): |
|
479 | 479 | if protoswitched: |
|
480 | 480 | _sshv1respondooberror(fout, ui.ferr, |
|
481 | 481 | b'cannot upgrade protocols multiple ' |
|
482 | 482 | b'times') |
|
483 | 483 | state = 'shutdown' |
|
484 | 484 | continue |
|
485 | 485 | |
|
486 | 486 | state = 'upgrade-initial' |
|
487 | 487 | continue |
|
488 | 488 | |
|
489 | 489 | available = wireproto.commands.commandavailable(request, proto) |
|
490 | 490 | |
|
491 | 491 | # This command isn't available. Send an empty response and go |
|
492 | 492 | # back to waiting for a new command. |
|
493 | 493 | if not available: |
|
494 | 494 | _sshv1respondbytes(fout, b'') |
|
495 | 495 | continue |
|
496 | 496 | |
|
497 | 497 | rsp = wireproto.dispatch(repo, proto, request) |
|
498 | 498 | |
|
499 | 499 | if isinstance(rsp, bytes): |
|
500 | 500 | _sshv1respondbytes(fout, rsp) |
|
501 | 501 | elif isinstance(rsp, wireprototypes.bytesresponse): |
|
502 | 502 | _sshv1respondbytes(fout, rsp.data) |
|
503 | 503 | elif isinstance(rsp, wireprototypes.streamres): |
|
504 | 504 | _sshv1respondstream(fout, rsp) |
|
505 | 505 | elif isinstance(rsp, wireprototypes.streamreslegacy): |
|
506 | 506 | _sshv1respondstream(fout, rsp) |
|
507 | 507 | elif isinstance(rsp, wireprototypes.pushres): |
|
508 | 508 | _sshv1respondbytes(fout, b'') |
|
509 | 509 | _sshv1respondbytes(fout, b'%d' % rsp.res) |
|
510 | 510 | elif isinstance(rsp, wireprototypes.pusherr): |
|
511 | 511 | _sshv1respondbytes(fout, rsp.res) |
|
512 | 512 | elif isinstance(rsp, wireprototypes.ooberror): |
|
513 | 513 | _sshv1respondooberror(fout, ui.ferr, rsp.message) |
|
514 | 514 | else: |
|
515 | 515 | raise error.ProgrammingError('unhandled response type from ' |
|
516 | 516 | 'wire protocol command: %s' % rsp) |
|
517 | 517 | |
|
518 | 518 | # For now, protocol version 2 serving just goes back to version 1. |
|
519 | 519 | elif state == 'protov2-serving': |
|
520 | 520 | state = 'protov1-serving' |
|
521 | 521 | continue |
|
522 | 522 | |
|
523 | 523 | elif state == 'upgrade-initial': |
|
524 | 524 | # We should never transition into this state if we've switched |
|
525 | 525 | # protocols. |
|
526 | 526 | assert not protoswitched |
|
527 | 527 | assert proto.name == wireprototypes.SSHV1 |
|
528 | 528 | |
|
529 | 529 | # Expected: upgrade <token> <capabilities> |
|
530 | 530 | # If we get something else, the request is malformed. It could be |
|
531 | 531 | # from a future client that has altered the upgrade line content. |
|
532 | 532 | # We treat this as an unknown command. |
|
533 | 533 | try: |
|
534 | 534 | token, caps = request.split(b' ')[1:] |
|
535 | 535 | except ValueError: |
|
536 | 536 | _sshv1respondbytes(fout, b'') |
|
537 | 537 | state = 'protov1-serving' |
|
538 | 538 | continue |
|
539 | 539 | |
|
540 | 540 | # Send empty response if we don't support upgrading protocols. |
|
541 | 541 | if not ui.configbool('experimental', 'sshserver.support-v2'): |
|
542 | 542 | _sshv1respondbytes(fout, b'') |
|
543 | 543 | state = 'protov1-serving' |
|
544 | 544 | continue |
|
545 | 545 | |
|
546 | 546 | try: |
|
547 | 547 | caps = urlreq.parseqs(caps) |
|
548 | 548 | except ValueError: |
|
549 | 549 | _sshv1respondbytes(fout, b'') |
|
550 | 550 | state = 'protov1-serving' |
|
551 | 551 | continue |
|
552 | 552 | |
|
553 | 553 | # We don't see an upgrade request to protocol version 2. Ignore |
|
554 | 554 | # the upgrade request. |
|
555 | 555 | wantedprotos = caps.get(b'proto', [b''])[0] |
|
556 | 556 | if SSHV2 not in wantedprotos: |
|
557 | 557 | _sshv1respondbytes(fout, b'') |
|
558 | 558 | state = 'protov1-serving' |
|
559 | 559 | continue |
|
560 | 560 | |
|
561 | 561 | # It looks like we can honor this upgrade request to protocol 2. |
|
562 | 562 | # Filter the rest of the handshake protocol request lines. |
|
563 | 563 | state = 'upgrade-v2-filter-legacy-handshake' |
|
564 | 564 | continue |
|
565 | 565 | |
|
566 | 566 | elif state == 'upgrade-v2-filter-legacy-handshake': |
|
567 | 567 | # Client should have sent legacy handshake after an ``upgrade`` |
|
568 | 568 | # request. Expected lines: |
|
569 | 569 | # |
|
570 | 570 | # hello |
|
571 | 571 | # between |
|
572 | 572 | # pairs 81 |
|
573 | 573 | # 0000...-0000... |
|
574 | 574 | |
|
575 | 575 | ok = True |
|
576 | 576 | for line in (b'hello', b'between', b'pairs 81'): |
|
577 | 577 | request = fin.readline()[:-1] |
|
578 | 578 | |
|
579 | 579 | if request != line: |
|
580 | 580 | _sshv1respondooberror(fout, ui.ferr, |
|
581 | 581 | b'malformed handshake protocol: ' |
|
582 | 582 | b'missing %s' % line) |
|
583 | 583 | ok = False |
|
584 | 584 | state = 'shutdown' |
|
585 | 585 | break |
|
586 | 586 | |
|
587 | 587 | if not ok: |
|
588 | 588 | continue |
|
589 | 589 | |
|
590 | 590 | request = fin.read(81) |
|
591 | 591 | if request != b'%s-%s' % (b'0' * 40, b'0' * 40): |
|
592 | 592 | _sshv1respondooberror(fout, ui.ferr, |
|
593 | 593 | b'malformed handshake protocol: ' |
|
594 | 594 | b'missing between argument value') |
|
595 | 595 | state = 'shutdown' |
|
596 | 596 | continue |
|
597 | 597 | |
|
598 | 598 | state = 'upgrade-v2-finish' |
|
599 | 599 | continue |
|
600 | 600 | |
|
601 | 601 | elif state == 'upgrade-v2-finish': |
|
602 | 602 | # Send the upgrade response. |
|
603 | 603 | fout.write(b'upgraded %s %s\n' % (token, SSHV2)) |
|
604 | 604 | servercaps = wireproto.capabilities(repo, proto) |
|
605 | 605 | rsp = b'capabilities: %s' % servercaps.data |
|
606 | 606 | fout.write(b'%d\n%s\n' % (len(rsp), rsp)) |
|
607 | 607 | fout.flush() |
|
608 | 608 | |
|
609 | 609 | proto = sshv2protocolhandler(ui, fin, fout) |
|
610 | 610 | protoswitched = True |
|
611 | 611 | |
|
612 | 612 | state = 'protov2-serving' |
|
613 | 613 | continue |
|
614 | 614 | |
|
615 | 615 | elif state == 'shutdown': |
|
616 | 616 | break |
|
617 | 617 | |
|
618 | 618 | else: |
|
619 | 619 | raise error.ProgrammingError('unhandled ssh server state: %s' % |
|
620 | 620 | state) |
|
621 | 621 | |
|
622 | 622 | class sshserver(object): |
|
623 | 623 | def __init__(self, ui, repo, logfh=None): |
|
624 | 624 | self._ui = ui |
|
625 | 625 | self._repo = repo |
|
626 | 626 | self._fin = ui.fin |
|
627 | 627 | self._fout = ui.fout |
|
628 | 628 | |
|
629 | 629 | # Log write I/O to stdout and stderr if configured. |
|
630 | 630 | if logfh: |
|
631 | 631 | self._fout = util.makeloggingfileobject( |
|
632 | 632 | logfh, self._fout, 'o', logdata=True) |
|
633 | 633 | ui.ferr = util.makeloggingfileobject( |
|
634 | 634 | logfh, ui.ferr, 'e', logdata=True) |
|
635 | 635 | |
|
636 | 636 | hook.redirect(True) |
|
637 | 637 | ui.fout = repo.ui.fout = ui.ferr |
|
638 | 638 | |
|
639 | 639 | # Prevent insertion/deletion of CRs |
|
640 | 640 | util.setbinary(self._fin) |
|
641 | 641 | util.setbinary(self._fout) |
|
642 | 642 | |
|
643 | 643 | def serve_forever(self): |
|
644 | 644 | self.serveuntil(threading.Event()) |
|
645 | 645 | sys.exit(0) |
|
646 | 646 | |
|
647 | 647 | def serveuntil(self, ev): |
|
648 | 648 | """Serve until a threading.Event is set.""" |
|
649 | 649 | _runsshserver(self._ui, self._repo, self._fin, self._fout, ev) |
General Comments 0
You need to be logged in to leave comments.
Login now