Show More
@@ -1,378 +1,374 b'' | |||
|
1 | 1 | # streamclone.py - producing and consuming streaming repository data |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2015 Gregory Szorc <gregory.szorc@gmail.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import struct |
|
11 | 11 | import time |
|
12 | 12 | |
|
13 | 13 | from .i18n import _ |
|
14 | 14 | from . import ( |
|
15 | 15 | branchmap, |
|
16 | 16 | error, |
|
17 | 17 | store, |
|
18 | 18 | util, |
|
19 | 19 | ) |
|
20 | 20 | |
|
21 | 21 | def canperformstreamclone(pullop, bailifbundle2supported=False): |
|
22 | 22 | """Whether it is possible to perform a streaming clone as part of pull. |
|
23 | 23 | |
|
24 | 24 | ``bailifbundle2supported`` will cause the function to return False if |
|
25 | 25 | bundle2 stream clones are supported. It should only be called by the |
|
26 | 26 | legacy stream clone code path. |
|
27 | 27 | |
|
28 | 28 | Returns a tuple of (supported, requirements). ``supported`` is True if |
|
29 | 29 | streaming clone is supported and False otherwise. ``requirements`` is |
|
30 | 30 | a set of repo requirements from the remote, or ``None`` if stream clone |
|
31 | 31 | isn't supported. |
|
32 | 32 | """ |
|
33 | 33 | repo = pullop.repo |
|
34 | 34 | remote = pullop.remote |
|
35 | 35 | |
|
36 | 36 | bundle2supported = False |
|
37 | 37 | if pullop.canusebundle2: |
|
38 | 38 | if 'v1' in pullop.remotebundle2caps.get('stream', []): |
|
39 | 39 | bundle2supported = True |
|
40 | 40 | # else |
|
41 | 41 | # Server doesn't support bundle2 stream clone or doesn't support |
|
42 | 42 | # the versions we support. Fall back and possibly allow legacy. |
|
43 | 43 | |
|
44 | 44 | # Ensures legacy code path uses available bundle2. |
|
45 | 45 | if bailifbundle2supported and bundle2supported: |
|
46 | 46 | return False, None |
|
47 | 47 | # Ensures bundle2 doesn't try to do a stream clone if it isn't supported. |
|
48 | 48 | #elif not bailifbundle2supported and not bundle2supported: |
|
49 | 49 | # return False, None |
|
50 | 50 | |
|
51 | 51 | # Streaming clone only works on empty repositories. |
|
52 | 52 | if len(repo): |
|
53 | 53 | return False, None |
|
54 | 54 | |
|
55 | 55 | # Streaming clone only works if all data is being requested. |
|
56 | 56 | if pullop.heads: |
|
57 | 57 | return False, None |
|
58 | 58 | |
|
59 | 59 | streamrequested = pullop.streamclonerequested |
|
60 | 60 | |
|
61 | 61 | # If we don't have a preference, let the server decide for us. This |
|
62 | 62 | # likely only comes into play in LANs. |
|
63 | 63 | if streamrequested is None: |
|
64 | 64 | # The server can advertise whether to prefer streaming clone. |
|
65 | 65 | streamrequested = remote.capable('stream-preferred') |
|
66 | 66 | |
|
67 | 67 | if not streamrequested: |
|
68 | 68 | return False, None |
|
69 | 69 | |
|
70 | 70 | # In order for stream clone to work, the client has to support all the |
|
71 | 71 | # requirements advertised by the server. |
|
72 | 72 | # |
|
73 | 73 | # The server advertises its requirements via the "stream" and "streamreqs" |
|
74 | 74 | # capability. "stream" (a value-less capability) is advertised if and only |
|
75 | 75 | # if the only requirement is "revlogv1." Else, the "streamreqs" capability |
|
76 | 76 | # is advertised and contains a comma-delimited list of requirements. |
|
77 | 77 | requirements = set() |
|
78 | 78 | if remote.capable('stream'): |
|
79 | 79 | requirements.add('revlogv1') |
|
80 | 80 | else: |
|
81 | 81 | streamreqs = remote.capable('streamreqs') |
|
82 | 82 | # This is weird and shouldn't happen with modern servers. |
|
83 | 83 | if not streamreqs: |
|
84 | 84 | return False, None |
|
85 | 85 | |
|
86 | 86 | streamreqs = set(streamreqs.split(',')) |
|
87 | 87 | # Server requires something we don't support. Bail. |
|
88 | 88 | if streamreqs - repo.supportedformats: |
|
89 | 89 | return False, None |
|
90 | 90 | requirements = streamreqs |
|
91 | 91 | |
|
92 | 92 | return True, requirements |
|
93 | 93 | |
|
94 | 94 | def maybeperformlegacystreamclone(pullop): |
|
95 | 95 | """Possibly perform a legacy stream clone operation. |
|
96 | 96 | |
|
97 | 97 | Legacy stream clones are performed as part of pull but before all other |
|
98 | 98 | operations. |
|
99 | 99 | |
|
100 | 100 | A legacy stream clone will not be performed if a bundle2 stream clone is |
|
101 | 101 | supported. |
|
102 | 102 | """ |
|
103 | 103 | supported, requirements = canperformstreamclone(pullop) |
|
104 | 104 | |
|
105 | 105 | if not supported: |
|
106 | 106 | return |
|
107 | 107 | |
|
108 | 108 | repo = pullop.repo |
|
109 | 109 | remote = pullop.remote |
|
110 | 110 | |
|
111 | 111 | # Save remote branchmap. We will use it later to speed up branchcache |
|
112 | 112 | # creation. |
|
113 | 113 | rbranchmap = None |
|
114 | 114 | if remote.capable('branchmap'): |
|
115 | 115 | rbranchmap = remote.branchmap() |
|
116 | 116 | |
|
117 | 117 | repo.ui.status(_('streaming all changes\n')) |
|
118 | 118 | |
|
119 | 119 | fp = remote.stream_out() |
|
120 | 120 | l = fp.readline() |
|
121 | 121 | try: |
|
122 | 122 | resp = int(l) |
|
123 | 123 | except ValueError: |
|
124 | 124 | raise error.ResponseError( |
|
125 | 125 | _('unexpected response from remote server:'), l) |
|
126 | 126 | if resp == 1: |
|
127 | 127 | raise error.Abort(_('operation forbidden by server')) |
|
128 | 128 | elif resp == 2: |
|
129 | 129 | raise error.Abort(_('locking the remote repository failed')) |
|
130 | 130 | elif resp != 0: |
|
131 | 131 | raise error.Abort(_('the server sent an unknown error code')) |
|
132 | 132 | |
|
133 | 133 | l = fp.readline() |
|
134 | 134 | try: |
|
135 | 135 | filecount, bytecount = map(int, l.split(' ', 1)) |
|
136 | 136 | except (ValueError, TypeError): |
|
137 | 137 | raise error.ResponseError( |
|
138 | 138 | _('unexpected response from remote server:'), l) |
|
139 | 139 | |
|
140 | 140 | with repo.lock(): |
|
141 | 141 | consumev1(repo, fp, filecount, bytecount) |
|
142 | 142 | |
|
143 | 143 | # new requirements = old non-format requirements + |
|
144 | 144 | # new format-related remote requirements |
|
145 | 145 | # requirements from the streamed-in repository |
|
146 | 146 | repo.requirements = requirements | ( |
|
147 | 147 | repo.requirements - repo.supportedformats) |
|
148 | 148 | repo._applyopenerreqs() |
|
149 | 149 | repo._writerequirements() |
|
150 | 150 | |
|
151 | 151 | if rbranchmap: |
|
152 | 152 | branchmap.replacecache(repo, rbranchmap) |
|
153 | 153 | |
|
154 | 154 | repo.invalidate() |
|
155 | 155 | |
|
156 | 156 | def allowservergeneration(ui): |
|
157 | 157 | """Whether streaming clones are allowed from the server.""" |
|
158 | 158 | return ui.configbool('server', 'uncompressed', True, untrusted=True) |
|
159 | 159 | |
|
160 | 160 | # This is it's own function so extensions can override it. |
|
161 | 161 | def _walkstreamfiles(repo): |
|
162 | 162 | return repo.store.walk() |
|
163 | 163 | |
|
164 | 164 | def generatev1(repo): |
|
165 | 165 | """Emit content for version 1 of a streaming clone. |
|
166 | 166 | |
|
167 | 167 | This returns a 3-tuple of (file count, byte size, data iterator). |
|
168 | 168 | |
|
169 | 169 | The data iterator consists of N entries for each file being transferred. |
|
170 | 170 | Each file entry starts as a line with the file name and integer size |
|
171 | 171 | delimited by a null byte. |
|
172 | 172 | |
|
173 | 173 | The raw file data follows. Following the raw file data is the next file |
|
174 | 174 | entry, or EOF. |
|
175 | 175 | |
|
176 | 176 | When used on the wire protocol, an additional line indicating protocol |
|
177 | 177 | success will be prepended to the stream. This function is not responsible |
|
178 | 178 | for adding it. |
|
179 | 179 | |
|
180 | 180 | This function will obtain a repository lock to ensure a consistent view of |
|
181 | 181 | the store is captured. It therefore may raise LockError. |
|
182 | 182 | """ |
|
183 | 183 | entries = [] |
|
184 | 184 | total_bytes = 0 |
|
185 | 185 | # Get consistent snapshot of repo, lock during scan. |
|
186 | 186 | with repo.lock(): |
|
187 | 187 | repo.ui.debug('scanning\n') |
|
188 | 188 | for name, ename, size in _walkstreamfiles(repo): |
|
189 | 189 | if size: |
|
190 | 190 | entries.append((name, size)) |
|
191 | 191 | total_bytes += size |
|
192 | 192 | |
|
193 | 193 | repo.ui.debug('%d files, %d bytes to transfer\n' % |
|
194 | 194 | (len(entries), total_bytes)) |
|
195 | 195 | |
|
196 | 196 | svfs = repo.svfs |
|
197 | 197 | oldaudit = svfs.mustaudit |
|
198 | 198 | debugflag = repo.ui.debugflag |
|
199 | 199 | svfs.mustaudit = False |
|
200 | 200 | |
|
201 | 201 | def emitrevlogdata(): |
|
202 | 202 | try: |
|
203 | 203 | for name, size in entries: |
|
204 | 204 | if debugflag: |
|
205 | 205 | repo.ui.debug('sending %s (%d bytes)\n' % (name, size)) |
|
206 | 206 | # partially encode name over the wire for backwards compat |
|
207 | 207 | yield '%s\0%d\n' % (store.encodedir(name), size) |
|
208 | 208 | if size <= 65536: |
|
209 | 209 | yield svfs.read(name) |
|
210 | 210 | else: |
|
211 | 211 | for chunk in util.filechunkiter(svfs(name), limit=size): |
|
212 | 212 | yield chunk |
|
213 | 213 | finally: |
|
214 | 214 | svfs.mustaudit = oldaudit |
|
215 | 215 | |
|
216 | 216 | return len(entries), total_bytes, emitrevlogdata() |
|
217 | 217 | |
|
218 | 218 | def generatev1wireproto(repo): |
|
219 | 219 | """Emit content for version 1 of streaming clone suitable for the wire. |
|
220 | 220 | |
|
221 | 221 | This is the data output from ``generatev1()`` with a header line |
|
222 | 222 | indicating file count and byte size. |
|
223 | 223 | """ |
|
224 | 224 | filecount, bytecount, it = generatev1(repo) |
|
225 | 225 | yield '%d %d\n' % (filecount, bytecount) |
|
226 | 226 | for chunk in it: |
|
227 | 227 | yield chunk |
|
228 | 228 | |
|
229 | 229 | def generatebundlev1(repo, compression='UN'): |
|
230 | 230 | """Emit content for version 1 of a stream clone bundle. |
|
231 | 231 | |
|
232 | 232 | The first 4 bytes of the output ("HGS1") denote this as stream clone |
|
233 | 233 | bundle version 1. |
|
234 | 234 | |
|
235 | 235 | The next 2 bytes indicate the compression type. Only "UN" is currently |
|
236 | 236 | supported. |
|
237 | 237 | |
|
238 | 238 | The next 16 bytes are two 64-bit big endian unsigned integers indicating |
|
239 | 239 | file count and byte count, respectively. |
|
240 | 240 | |
|
241 | 241 | The next 2 bytes is a 16-bit big endian unsigned short declaring the length |
|
242 | 242 | of the requirements string, including a trailing \0. The following N bytes |
|
243 | 243 | are the requirements string, which is ASCII containing a comma-delimited |
|
244 | 244 | list of repo requirements that are needed to support the data. |
|
245 | 245 | |
|
246 | 246 | The remaining content is the output of ``generatev1()`` (which may be |
|
247 | 247 | compressed in the future). |
|
248 | 248 | |
|
249 | 249 | Returns a tuple of (requirements, data generator). |
|
250 | 250 | """ |
|
251 | 251 | if compression != 'UN': |
|
252 | 252 | raise ValueError('we do not support the compression argument yet') |
|
253 | 253 | |
|
254 | 254 | requirements = repo.requirements & repo.supportedformats |
|
255 | 255 | requires = ','.join(sorted(requirements)) |
|
256 | 256 | |
|
257 | 257 | def gen(): |
|
258 | 258 | yield 'HGS1' |
|
259 | 259 | yield compression |
|
260 | 260 | |
|
261 | 261 | filecount, bytecount, it = generatev1(repo) |
|
262 | 262 | repo.ui.status(_('writing %d bytes for %d files\n') % |
|
263 | 263 | (bytecount, filecount)) |
|
264 | 264 | |
|
265 | 265 | yield struct.pack('>QQ', filecount, bytecount) |
|
266 | 266 | yield struct.pack('>H', len(requires) + 1) |
|
267 | 267 | yield requires + '\0' |
|
268 | 268 | |
|
269 | 269 | # This is where we'll add compression in the future. |
|
270 | 270 | assert compression == 'UN' |
|
271 | 271 | |
|
272 | 272 | seen = 0 |
|
273 | 273 | repo.ui.progress(_('bundle'), 0, total=bytecount) |
|
274 | 274 | |
|
275 | 275 | for chunk in it: |
|
276 | 276 | seen += len(chunk) |
|
277 | 277 | repo.ui.progress(_('bundle'), seen, total=bytecount) |
|
278 | 278 | yield chunk |
|
279 | 279 | |
|
280 | 280 | repo.ui.progress(_('bundle'), None) |
|
281 | 281 | |
|
282 | 282 | return requirements, gen() |
|
283 | 283 | |
|
284 | 284 | def consumev1(repo, fp, filecount, bytecount): |
|
285 | 285 | """Apply the contents from version 1 of a streaming clone file handle. |
|
286 | 286 | |
|
287 | 287 | This takes the output from "streamout" and applies it to the specified |
|
288 | 288 | repository. |
|
289 | 289 | |
|
290 | 290 | Like "streamout," the status line added by the wire protocol is not handled |
|
291 | 291 | by this function. |
|
292 | 292 | """ |
|
293 | 293 | with repo.lock(): |
|
294 | 294 | repo.ui.status(_('%d files to transfer, %s of data\n') % |
|
295 | 295 | (filecount, util.bytecount(bytecount))) |
|
296 | 296 | handled_bytes = 0 |
|
297 | 297 | repo.ui.progress(_('clone'), 0, total=bytecount) |
|
298 | 298 | start = time.time() |
|
299 | 299 | |
|
300 |
|
|
|
301 | try: | |
|
300 | with repo.transaction('clone'): | |
|
302 | 301 | for i in xrange(filecount): |
|
303 | 302 | # XXX doesn't support '\n' or '\r' in filenames |
|
304 | 303 | l = fp.readline() |
|
305 | 304 | try: |
|
306 | 305 | name, size = l.split('\0', 1) |
|
307 | 306 | size = int(size) |
|
308 | 307 | except (ValueError, TypeError): |
|
309 | 308 | raise error.ResponseError( |
|
310 | 309 | _('unexpected response from remote server:'), l) |
|
311 | 310 | if repo.ui.debugflag: |
|
312 | 311 | repo.ui.debug('adding %s (%s)\n' % |
|
313 | 312 | (name, util.bytecount(size))) |
|
314 | 313 | # for backwards compat, name was partially encoded |
|
315 | 314 | with repo.svfs(store.decodedir(name), 'w') as ofp: |
|
316 | 315 | for chunk in util.filechunkiter(fp, limit=size): |
|
317 | 316 | handled_bytes += len(chunk) |
|
318 | 317 | repo.ui.progress(_('clone'), handled_bytes, |
|
319 | 318 | total=bytecount) |
|
320 | 319 | ofp.write(chunk) |
|
321 | tr.close() | |
|
322 | finally: | |
|
323 | tr.release() | |
|
324 | 320 | |
|
325 | 321 | # Writing straight to files circumvented the inmemory caches |
|
326 | 322 | repo.invalidate() |
|
327 | 323 | |
|
328 | 324 | elapsed = time.time() - start |
|
329 | 325 | if elapsed <= 0: |
|
330 | 326 | elapsed = 0.001 |
|
331 | 327 | repo.ui.progress(_('clone'), None) |
|
332 | 328 | repo.ui.status(_('transferred %s in %.1f seconds (%s/sec)\n') % |
|
333 | 329 | (util.bytecount(bytecount), elapsed, |
|
334 | 330 | util.bytecount(bytecount / elapsed))) |
|
335 | 331 | |
|
336 | 332 | def applybundlev1(repo, fp): |
|
337 | 333 | """Apply the content from a stream clone bundle version 1. |
|
338 | 334 | |
|
339 | 335 | We assume the 4 byte header has been read and validated and the file handle |
|
340 | 336 | is at the 2 byte compression identifier. |
|
341 | 337 | """ |
|
342 | 338 | if len(repo): |
|
343 | 339 | raise error.Abort(_('cannot apply stream clone bundle on non-empty ' |
|
344 | 340 | 'repo')) |
|
345 | 341 | |
|
346 | 342 | compression = fp.read(2) |
|
347 | 343 | if compression != 'UN': |
|
348 | 344 | raise error.Abort(_('only uncompressed stream clone bundles are ' |
|
349 | 345 | 'supported; got %s') % compression) |
|
350 | 346 | |
|
351 | 347 | filecount, bytecount = struct.unpack('>QQ', fp.read(16)) |
|
352 | 348 | requireslen = struct.unpack('>H', fp.read(2))[0] |
|
353 | 349 | requires = fp.read(requireslen) |
|
354 | 350 | |
|
355 | 351 | if not requires.endswith('\0'): |
|
356 | 352 | raise error.Abort(_('malformed stream clone bundle: ' |
|
357 | 353 | 'requirements not properly encoded')) |
|
358 | 354 | |
|
359 | 355 | requirements = set(requires.rstrip('\0').split(',')) |
|
360 | 356 | missingreqs = requirements - repo.supportedformats |
|
361 | 357 | if missingreqs: |
|
362 | 358 | raise error.Abort(_('unable to apply stream clone: ' |
|
363 | 359 | 'unsupported format: %s') % |
|
364 | 360 | ', '.join(sorted(missingreqs))) |
|
365 | 361 | |
|
366 | 362 | consumev1(repo, fp, filecount, bytecount) |
|
367 | 363 | |
|
368 | 364 | class streamcloneapplier(object): |
|
369 | 365 | """Class to manage applying streaming clone bundles. |
|
370 | 366 | |
|
371 | 367 | We need to wrap ``applybundlev1()`` in a dedicated type to enable bundle |
|
372 | 368 | readers to perform bundle type-specific functionality. |
|
373 | 369 | """ |
|
374 | 370 | def __init__(self, fh): |
|
375 | 371 | self._fh = fh |
|
376 | 372 | |
|
377 | 373 | def apply(self, repo): |
|
378 | 374 | return applybundlev1(repo, self._fh) |
General Comments 0
You need to be logged in to leave comments.
Login now