Show More
@@ -1,458 +1,461 b'' | |||
|
1 | 1 | from __future__ import absolute_import |
|
2 | 2 | |
|
3 | 3 | import errno |
|
4 | 4 | import hashlib |
|
5 | 5 | import os |
|
6 | 6 | import shutil |
|
7 | 7 | import stat |
|
8 | 8 | import time |
|
9 | 9 | |
|
10 | 10 | from mercurial.i18n import _ |
|
11 | 11 | from mercurial.node import bin, hex |
|
12 | 12 | from mercurial.pycompat import open |
|
13 | 13 | from mercurial import ( |
|
14 | 14 | error, |
|
15 | 15 | pycompat, |
|
16 | 16 | util, |
|
17 | 17 | ) |
|
18 | 18 | from . import ( |
|
19 | 19 | constants, |
|
20 | 20 | shallowutil, |
|
21 | 21 | ) |
|
22 | 22 | |
|
23 | 23 | |
|
24 | 24 | class basestore(object): |
|
25 | 25 | def __init__(self, repo, path, reponame, shared=False): |
|
26 | 26 | """Creates a remotefilelog store object for the given repo name. |
|
27 | 27 | |
|
28 | 28 | `path` - The file path where this store keeps its data |
|
29 | 29 | `reponame` - The name of the repo. This is used to partition data from |
|
30 | 30 | many repos. |
|
31 | 31 | `shared` - True if this store is a shared cache of data from the central |
|
32 | 32 | server, for many repos on this machine. False means this store is for |
|
33 | 33 | the local data for one repo. |
|
34 | 34 | """ |
|
35 | 35 | self.repo = repo |
|
36 | 36 | self.ui = repo.ui |
|
37 | 37 | self._path = path |
|
38 | 38 | self._reponame = reponame |
|
39 | 39 | self._shared = shared |
|
40 | 40 | self._uid = os.getuid() if not pycompat.iswindows else None |
|
41 | 41 | |
|
42 | 42 | self._validatecachelog = self.ui.config( |
|
43 | 43 | b"remotefilelog", b"validatecachelog" |
|
44 | 44 | ) |
|
45 | 45 | self._validatecache = self.ui.config( |
|
46 | 46 | b"remotefilelog", b"validatecache", b'on' |
|
47 | 47 | ) |
|
48 | 48 | if self._validatecache not in (b'on', b'strict', b'off'): |
|
49 | 49 | self._validatecache = b'on' |
|
50 | 50 | if self._validatecache == b'off': |
|
51 | 51 | self._validatecache = False |
|
52 | 52 | |
|
53 | 53 | if shared: |
|
54 | 54 | shallowutil.mkstickygroupdir(self.ui, path) |
|
55 | 55 | |
|
56 | 56 | def getmissing(self, keys): |
|
57 | 57 | missing = [] |
|
58 | 58 | for name, node in keys: |
|
59 | 59 | filepath = self._getfilepath(name, node) |
|
60 | 60 | exists = os.path.exists(filepath) |
|
61 | 61 | if ( |
|
62 | 62 | exists |
|
63 | 63 | and self._validatecache == b'strict' |
|
64 | 64 | and not self._validatekey(filepath, b'contains') |
|
65 | 65 | ): |
|
66 | 66 | exists = False |
|
67 | 67 | if not exists: |
|
68 | 68 | missing.append((name, node)) |
|
69 | 69 | |
|
70 | 70 | return missing |
|
71 | 71 | |
|
72 | 72 | # BELOW THIS ARE IMPLEMENTATIONS OF REPACK SOURCE |
|
73 | 73 | |
|
74 | 74 | def markledger(self, ledger, options=None): |
|
75 | 75 | if options and options.get(constants.OPTION_PACKSONLY): |
|
76 | 76 | return |
|
77 | 77 | if self._shared: |
|
78 | 78 | for filename, nodes in self._getfiles(): |
|
79 | 79 | for node in nodes: |
|
80 | 80 | ledger.markdataentry(self, filename, node) |
|
81 | 81 | ledger.markhistoryentry(self, filename, node) |
|
82 | 82 | |
|
83 | 83 | def cleanup(self, ledger): |
|
84 | 84 | ui = self.ui |
|
85 | 85 | entries = ledger.sources.get(self, []) |
|
86 | 86 | count = 0 |
|
87 | 87 | progress = ui.makeprogress( |
|
88 | 88 | _(b"cleaning up"), unit=b"files", total=len(entries) |
|
89 | 89 | ) |
|
90 | 90 | for entry in entries: |
|
91 | 91 | if entry.gced or (entry.datarepacked and entry.historyrepacked): |
|
92 | 92 | progress.update(count) |
|
93 | 93 | path = self._getfilepath(entry.filename, entry.node) |
|
94 | 94 | util.tryunlink(path) |
|
95 | 95 | count += 1 |
|
96 | 96 | progress.complete() |
|
97 | 97 | |
|
98 | 98 | # Clean up the repo cache directory. |
|
99 | 99 | self._cleanupdirectory(self._getrepocachepath()) |
|
100 | 100 | |
|
101 | 101 | # BELOW THIS ARE NON-STANDARD APIS |
|
102 | 102 | |
|
103 | 103 | def _cleanupdirectory(self, rootdir): |
|
104 | 104 | """Removes the empty directories and unnecessary files within the root |
|
105 | 105 | directory recursively. Note that this method does not remove the root |
|
106 | 106 | directory itself. """ |
|
107 | 107 | |
|
108 | 108 | oldfiles = set() |
|
109 | 109 | otherfiles = set() |
|
110 | 110 | # osutil.listdir returns stat information which saves some rmdir/listdir |
|
111 | 111 | # syscalls. |
|
112 | 112 | for name, mode in util.osutil.listdir(rootdir): |
|
113 | 113 | if stat.S_ISDIR(mode): |
|
114 | 114 | dirpath = os.path.join(rootdir, name) |
|
115 | 115 | self._cleanupdirectory(dirpath) |
|
116 | 116 | |
|
117 | 117 | # Now that the directory specified by dirpath is potentially |
|
118 | 118 | # empty, try and remove it. |
|
119 | 119 | try: |
|
120 | 120 | os.rmdir(dirpath) |
|
121 | 121 | except OSError: |
|
122 | 122 | pass |
|
123 | 123 | |
|
124 | 124 | elif stat.S_ISREG(mode): |
|
125 | 125 | if name.endswith(b'_old'): |
|
126 | 126 | oldfiles.add(name[:-4]) |
|
127 | 127 | else: |
|
128 | 128 | otherfiles.add(name) |
|
129 | 129 | |
|
130 | 130 | # Remove the files which end with suffix '_old' and have no |
|
131 | 131 | # corresponding file without the suffix '_old'. See addremotefilelognode |
|
132 | 132 | # method for the generation/purpose of files with '_old' suffix. |
|
133 | 133 | for filename in oldfiles - otherfiles: |
|
134 | 134 | filepath = os.path.join(rootdir, filename + b'_old') |
|
135 | 135 | util.tryunlink(filepath) |
|
136 | 136 | |
|
137 | 137 | def _getfiles(self): |
|
138 | 138 | """Return a list of (filename, [node,...]) for all the revisions that |
|
139 | 139 | exist in the store. |
|
140 | 140 | |
|
141 | 141 | This is useful for obtaining a list of all the contents of the store |
|
142 | 142 | when performing a repack to another store, since the store API requires |
|
143 | 143 | name+node keys and not namehash+node keys. |
|
144 | 144 | """ |
|
145 | 145 | existing = {} |
|
146 | 146 | for filenamehash, node in self._listkeys(): |
|
147 | 147 | existing.setdefault(filenamehash, []).append(node) |
|
148 | 148 | |
|
149 | 149 | filenamemap = self._resolvefilenames(existing.keys()) |
|
150 | 150 | |
|
151 | 151 | for filename, sha in pycompat.iteritems(filenamemap): |
|
152 | 152 | yield (filename, existing[sha]) |
|
153 | 153 | |
|
154 | 154 | def _resolvefilenames(self, hashes): |
|
155 | 155 | """Given a list of filename hashes that are present in the |
|
156 | 156 | remotefilelog store, return a mapping from filename->hash. |
|
157 | 157 | |
|
158 | 158 | This is useful when converting remotefilelog blobs into other storage |
|
159 | 159 | formats. |
|
160 | 160 | """ |
|
161 | 161 | if not hashes: |
|
162 | 162 | return {} |
|
163 | 163 | |
|
164 | 164 | filenames = {} |
|
165 | 165 | missingfilename = set(hashes) |
|
166 | 166 | |
|
167 | 167 | # Start with a full manifest, since it'll cover the majority of files |
|
168 | 168 | for filename in self.repo[b'tip'].manifest(): |
|
169 | 169 | sha = hashlib.sha1(filename).digest() |
|
170 | 170 | if sha in missingfilename: |
|
171 | 171 | filenames[filename] = sha |
|
172 | 172 | missingfilename.discard(sha) |
|
173 | 173 | |
|
174 | 174 | # Scan the changelog until we've found every file name |
|
175 | 175 | cl = self.repo.unfiltered().changelog |
|
176 | 176 | for rev in pycompat.xrange(len(cl) - 1, -1, -1): |
|
177 | 177 | if not missingfilename: |
|
178 | 178 | break |
|
179 | 179 | files = cl.readfiles(cl.node(rev)) |
|
180 | 180 | for filename in files: |
|
181 | 181 | sha = hashlib.sha1(filename).digest() |
|
182 | 182 | if sha in missingfilename: |
|
183 | 183 | filenames[filename] = sha |
|
184 | 184 | missingfilename.discard(sha) |
|
185 | 185 | |
|
186 | 186 | return filenames |
|
187 | 187 | |
|
188 | 188 | def _getrepocachepath(self): |
|
189 | 189 | return ( |
|
190 | 190 | os.path.join(self._path, self._reponame) |
|
191 | 191 | if self._shared |
|
192 | 192 | else self._path |
|
193 | 193 | ) |
|
194 | 194 | |
|
195 | 195 | def _listkeys(self): |
|
196 | 196 | """List all the remotefilelog keys that exist in the store. |
|
197 | 197 | |
|
198 | 198 | Returns a iterator of (filename hash, filecontent hash) tuples. |
|
199 | 199 | """ |
|
200 | 200 | |
|
201 | 201 | for root, dirs, files in os.walk(self._getrepocachepath()): |
|
202 | 202 | for filename in files: |
|
203 | 203 | if len(filename) != 40: |
|
204 | 204 | continue |
|
205 | 205 | node = filename |
|
206 | 206 | if self._shared: |
|
207 | 207 | # .../1a/85ffda..be21 |
|
208 | 208 | filenamehash = root[-41:-39] + root[-38:] |
|
209 | 209 | else: |
|
210 | 210 | filenamehash = root[-40:] |
|
211 | 211 | yield (bin(filenamehash), bin(node)) |
|
212 | 212 | |
|
213 | 213 | def _getfilepath(self, name, node): |
|
214 | 214 | node = hex(node) |
|
215 | 215 | if self._shared: |
|
216 | 216 | key = shallowutil.getcachekey(self._reponame, name, node) |
|
217 | 217 | else: |
|
218 | 218 | key = shallowutil.getlocalkey(name, node) |
|
219 | 219 | |
|
220 | 220 | return os.path.join(self._path, key) |
|
221 | 221 | |
|
222 | 222 | def _getdata(self, name, node): |
|
223 | 223 | filepath = self._getfilepath(name, node) |
|
224 | 224 | try: |
|
225 | 225 | data = shallowutil.readfile(filepath) |
|
226 | 226 | if self._validatecache and not self._validatedata(data, filepath): |
|
227 | 227 | if self._validatecachelog: |
|
228 | 228 | with open(self._validatecachelog, b'a+') as f: |
|
229 | 229 | f.write(b"corrupt %s during read\n" % filepath) |
|
230 | 230 | os.rename(filepath, filepath + b".corrupt") |
|
231 | 231 | raise KeyError(b"corrupt local cache file %s" % filepath) |
|
232 | 232 | except IOError: |
|
233 | 233 | raise KeyError( |
|
234 | 234 | b"no file found at %s for %s:%s" % (filepath, name, hex(node)) |
|
235 | 235 | ) |
|
236 | 236 | |
|
237 | 237 | return data |
|
238 | 238 | |
|
239 | 239 | def addremotefilelognode(self, name, node, data): |
|
240 | 240 | filepath = self._getfilepath(name, node) |
|
241 | 241 | |
|
242 | 242 | oldumask = os.umask(0o002) |
|
243 | 243 | try: |
|
244 | 244 | # if this node already exists, save the old version for |
|
245 | 245 | # recovery/debugging purposes. |
|
246 | 246 | if os.path.exists(filepath): |
|
247 | 247 | newfilename = filepath + b'_old' |
|
248 | 248 | # newfilename can be read-only and shutil.copy will fail. |
|
249 | 249 | # Delete newfilename to avoid it |
|
250 | 250 | if os.path.exists(newfilename): |
|
251 | 251 | shallowutil.unlinkfile(newfilename) |
|
252 | 252 | shutil.copy(filepath, newfilename) |
|
253 | 253 | |
|
254 | 254 | shallowutil.mkstickygroupdir(self.ui, os.path.dirname(filepath)) |
|
255 | 255 | shallowutil.writefile(filepath, data, readonly=True) |
|
256 | 256 | |
|
257 | 257 | if self._validatecache: |
|
258 | 258 | if not self._validatekey(filepath, b'write'): |
|
259 | 259 | raise error.Abort( |
|
260 | 260 | _(b"local cache write was corrupted %s") % filepath |
|
261 | 261 | ) |
|
262 | 262 | finally: |
|
263 | 263 | os.umask(oldumask) |
|
264 | 264 | |
|
265 | 265 | def markrepo(self, path): |
|
266 | 266 | """Call this to add the given repo path to the store's list of |
|
267 | 267 | repositories that are using it. This is useful later when doing garbage |
|
268 | 268 | collection, since it allows us to insecpt the repos to see what nodes |
|
269 | 269 | they want to be kept alive in the store. |
|
270 | 270 | """ |
|
271 | 271 | repospath = os.path.join(self._path, b"repos") |
|
272 | 272 | with open(repospath, b'ab') as reposfile: |
|
273 | 273 | reposfile.write(os.path.dirname(path) + b"\n") |
|
274 | 274 | |
|
275 | 275 | repospathstat = os.stat(repospath) |
|
276 | 276 | if repospathstat.st_uid == self._uid: |
|
277 | 277 | os.chmod(repospath, 0o0664) |
|
278 | 278 | |
|
279 | 279 | def _validatekey(self, path, action): |
|
280 | 280 | with open(path, b'rb') as f: |
|
281 | 281 | data = f.read() |
|
282 | 282 | |
|
283 | 283 | if self._validatedata(data, path): |
|
284 | 284 | return True |
|
285 | 285 | |
|
286 | 286 | if self._validatecachelog: |
|
287 | 287 | with open(self._validatecachelog, b'ab+') as f: |
|
288 | 288 | f.write(b"corrupt %s during %s\n" % (path, action)) |
|
289 | 289 | |
|
290 | 290 | os.rename(path, path + b".corrupt") |
|
291 | 291 | return False |
|
292 | 292 | |
|
293 | 293 | def _validatedata(self, data, path): |
|
294 | 294 | try: |
|
295 | 295 | if len(data) > 0: |
|
296 | 296 | # see remotefilelogserver.createfileblob for the format |
|
297 | 297 | offset, size, flags = shallowutil.parsesizeflags(data) |
|
298 | 298 | if len(data) <= size: |
|
299 | 299 | # it is truncated |
|
300 | 300 | return False |
|
301 | 301 | |
|
302 | 302 | # extract the node from the metadata |
|
303 | 303 | offset += size |
|
304 | 304 | datanode = data[offset : offset + 20] |
|
305 | 305 | |
|
306 | 306 | # and compare against the path |
|
307 | 307 | if os.path.basename(path) == hex(datanode): |
|
308 | 308 | # Content matches the intended path |
|
309 | 309 | return True |
|
310 | 310 | return False |
|
311 | 311 | except (ValueError, RuntimeError): |
|
312 | 312 | pass |
|
313 | 313 | |
|
314 | 314 | return False |
|
315 | 315 | |
|
316 | 316 | def gc(self, keepkeys): |
|
317 | 317 | ui = self.ui |
|
318 | 318 | cachepath = self._path |
|
319 | 319 | |
|
320 | 320 | # prune cache |
|
321 | 321 | queue = pycompat.queue.PriorityQueue() |
|
322 | 322 | originalsize = 0 |
|
323 | 323 | size = 0 |
|
324 | 324 | count = 0 |
|
325 | 325 | removed = 0 |
|
326 | 326 | |
|
327 | 327 | # keep files newer than a day even if they aren't needed |
|
328 | 328 | limit = time.time() - (60 * 60 * 24) |
|
329 | 329 | |
|
330 | 330 | progress = ui.makeprogress( |
|
331 | 331 | _(b"removing unnecessary files"), unit=b"files" |
|
332 | 332 | ) |
|
333 | 333 | progress.update(0) |
|
334 | 334 | for root, dirs, files in os.walk(cachepath): |
|
335 | 335 | for file in files: |
|
336 | 336 | if file == b'repos': |
|
337 | 337 | continue |
|
338 | 338 | |
|
339 | 339 | # Don't delete pack files |
|
340 | 340 | if b'/packs/' in root: |
|
341 | 341 | continue |
|
342 | 342 | |
|
343 | 343 | progress.update(count) |
|
344 | 344 | path = os.path.join(root, file) |
|
345 | 345 | key = os.path.relpath(path, cachepath) |
|
346 | 346 | count += 1 |
|
347 | 347 | try: |
|
348 | 348 | pathstat = os.stat(path) |
|
349 | 349 | except OSError as e: |
|
350 | 350 | # errno.ENOENT = no such file or directory |
|
351 | 351 | if e.errno != errno.ENOENT: |
|
352 | 352 | raise |
|
353 | 353 | msg = _( |
|
354 | 354 | b"warning: file %s was removed by another process\n" |
|
355 | 355 | ) |
|
356 | 356 | ui.warn(msg % path) |
|
357 | 357 | continue |
|
358 | 358 | |
|
359 | 359 | originalsize += pathstat.st_size |
|
360 | 360 | |
|
361 | 361 | if key in keepkeys or pathstat.st_atime > limit: |
|
362 | 362 | queue.put((pathstat.st_atime, path, pathstat)) |
|
363 | 363 | size += pathstat.st_size |
|
364 | 364 | else: |
|
365 | 365 | try: |
|
366 | 366 | shallowutil.unlinkfile(path) |
|
367 | 367 | except OSError as e: |
|
368 | 368 | # errno.ENOENT = no such file or directory |
|
369 | 369 | if e.errno != errno.ENOENT: |
|
370 | 370 | raise |
|
371 | 371 | msg = _( |
|
372 | 372 | b"warning: file %s was removed by another " |
|
373 | 373 | b"process\n" |
|
374 | 374 | ) |
|
375 | 375 | ui.warn(msg % path) |
|
376 | 376 | continue |
|
377 | 377 | removed += 1 |
|
378 | 378 | progress.complete() |
|
379 | 379 | |
|
380 | 380 | # remove oldest files until under limit |
|
381 | 381 | limit = ui.configbytes(b"remotefilelog", b"cachelimit") |
|
382 | 382 | if size > limit: |
|
383 | 383 | excess = size - limit |
|
384 | 384 | progress = ui.makeprogress( |
|
385 | 385 | _(b"enforcing cache limit"), unit=b"bytes", total=excess |
|
386 | 386 | ) |
|
387 | 387 | removedexcess = 0 |
|
388 | 388 | while queue and size > limit and size > 0: |
|
389 | 389 | progress.update(removedexcess) |
|
390 | 390 | atime, oldpath, oldpathstat = queue.get() |
|
391 | 391 | try: |
|
392 | 392 | shallowutil.unlinkfile(oldpath) |
|
393 | 393 | except OSError as e: |
|
394 | 394 | # errno.ENOENT = no such file or directory |
|
395 | 395 | if e.errno != errno.ENOENT: |
|
396 | 396 | raise |
|
397 | 397 | msg = _( |
|
398 | 398 | b"warning: file %s was removed by another process\n" |
|
399 | 399 | ) |
|
400 | 400 | ui.warn(msg % oldpath) |
|
401 | 401 | size -= oldpathstat.st_size |
|
402 | 402 | removed += 1 |
|
403 | 403 | removedexcess += oldpathstat.st_size |
|
404 | 404 | progress.complete() |
|
405 | 405 | |
|
406 | 406 | ui.status( |
|
407 | 407 | _(b"finished: removed %d of %d files (%0.2f GB to %0.2f GB)\n") |
|
408 | 408 | % ( |
|
409 | 409 | removed, |
|
410 | 410 | count, |
|
411 | 411 | float(originalsize) / 1024.0 / 1024.0 / 1024.0, |
|
412 | 412 | float(size) / 1024.0 / 1024.0 / 1024.0, |
|
413 | 413 | ) |
|
414 | 414 | ) |
|
415 | 415 | |
|
416 | 416 | |
|
417 | 417 | class baseunionstore(object): |
|
418 | 418 | def __init__(self, *args, **kwargs): |
|
419 | 419 | # If one of the functions that iterates all of the stores is about to |
|
420 | 420 | # throw a KeyError, try this many times with a full refresh between |
|
421 | 421 | # attempts. A repack operation may have moved data from one store to |
|
422 | 422 | # another while we were running. |
|
423 | 423 | self.numattempts = kwargs.get(r'numretries', 0) + 1 |
|
424 | 424 | # If not-None, call this function on every retry and if the attempts are |
|
425 | 425 | # exhausted. |
|
426 | 426 | self.retrylog = kwargs.get(r'retrylog', None) |
|
427 | 427 | |
|
428 | 428 | def markforrefresh(self): |
|
429 | 429 | for store in self.stores: |
|
430 | 430 | if util.safehasattr(store, b'markforrefresh'): |
|
431 | 431 | store.markforrefresh() |
|
432 | 432 | |
|
433 | 433 | @staticmethod |
|
434 | 434 | def retriable(fn): |
|
435 | 435 | def noop(*args): |
|
436 | 436 | pass |
|
437 | 437 | |
|
438 | 438 | def wrapped(self, *args, **kwargs): |
|
439 | 439 | retrylog = self.retrylog or noop |
|
440 | 440 | funcname = fn.__name__ |
|
441 | 441 | i = 0 |
|
442 | 442 | while i < self.numattempts: |
|
443 | 443 | if i > 0: |
|
444 | retrylog(b're-attempting (n=%d) %s\n' % (i, funcname)) | |
|
444 | retrylog( | |
|
445 | b're-attempting (n=%d) %s\n' | |
|
446 | % (i, pycompat.sysbytes(funcname)) | |
|
447 | ) | |
|
445 | 448 | self.markforrefresh() |
|
446 | 449 | i += 1 |
|
447 | 450 | try: |
|
448 | 451 | return fn(self, *args, **kwargs) |
|
449 | 452 | except KeyError: |
|
450 | 453 | if i == self.numattempts: |
|
451 | 454 | # retries exhausted |
|
452 | 455 | retrylog( |
|
453 | 456 | b'retries exhausted in %s, raising KeyError\n' |
|
454 | 457 | % pycompat.sysbytes(funcname) |
|
455 | 458 | ) |
|
456 | 459 | raise |
|
457 | 460 | |
|
458 | 461 | return wrapped |
General Comments 0
You need to be logged in to leave comments.
Login now