Show More
@@ -1,287 +1,283 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 time |
|
11 | 11 | |
|
12 | 12 | from .i18n import _ |
|
13 | 13 | from . import ( |
|
14 | 14 | branchmap, |
|
15 | 15 | error, |
|
16 | 16 | store, |
|
17 | 17 | util, |
|
18 | 18 | ) |
|
19 | 19 | |
|
20 | 20 | def canperformstreamclone(repo, remote, heads, streamrequested=None): |
|
21 | 21 | """Whether it is possible to perform a streaming clone as part of pull. |
|
22 | 22 | |
|
23 | 23 | Returns a tuple of (supported, requirements). ``supported`` is True if |
|
24 | 24 | streaming clone is supported and False otherwise. ``requirements`` is |
|
25 | 25 | a set of repo requirements from the remote, or ``None`` if stream clone |
|
26 | 26 | isn't supported. |
|
27 | 27 | """ |
|
28 | 28 | # Streaming clone only works on empty repositories. |
|
29 | 29 | if len(repo): |
|
30 | 30 | return False, None |
|
31 | 31 | |
|
32 | 32 | # Streaming clone only works if all data is being requested. |
|
33 | 33 | if heads: |
|
34 | 34 | return False, None |
|
35 | 35 | |
|
36 | 36 | # If we don't have a preference, let the server decide for us. This |
|
37 | 37 | # likely only comes into play in LANs. |
|
38 | 38 | if streamrequested is None: |
|
39 | 39 | # The server can advertise whether to prefer streaming clone. |
|
40 | 40 | streamrequested = remote.capable('stream-preferred') |
|
41 | 41 | |
|
42 | 42 | if not streamrequested: |
|
43 | 43 | return False, None |
|
44 | 44 | |
|
45 | 45 | # In order for stream clone to work, the client has to support all the |
|
46 | 46 | # requirements advertised by the server. |
|
47 | 47 | # |
|
48 | 48 | # The server advertises its requirements via the "stream" and "streamreqs" |
|
49 | 49 | # capability. "stream" (a value-less capability) is advertised if and only |
|
50 | 50 | # if the only requirement is "revlogv1." Else, the "streamreqs" capability |
|
51 | 51 | # is advertised and contains a comma-delimited list of requirements. |
|
52 | 52 | requirements = set() |
|
53 | 53 | if remote.capable('stream'): |
|
54 | 54 | requirements.add('revlogv1') |
|
55 | 55 | else: |
|
56 | 56 | streamreqs = remote.capable('streamreqs') |
|
57 | 57 | # This is weird and shouldn't happen with modern servers. |
|
58 | 58 | if not streamreqs: |
|
59 | 59 | return False, None |
|
60 | 60 | |
|
61 | 61 | streamreqs = set(streamreqs.split(',')) |
|
62 | 62 | # Server requires something we don't support. Bail. |
|
63 | 63 | if streamreqs - repo.supportedformats: |
|
64 | 64 | return False, None |
|
65 | 65 | requirements = streamreqs |
|
66 | 66 | |
|
67 | 67 | return True, requirements |
|
68 | 68 | |
|
69 | 69 | def maybeperformstreamclone(pullop): |
|
70 | 70 | repo = pullop.repo |
|
71 | 71 | remote = pullop.remote |
|
72 | 72 | |
|
73 | 73 | r = canperformstreamclone(repo, remote, pullop.heads, |
|
74 | 74 | streamrequested=pullop.streamclonerequested) |
|
75 | 75 | supported, requirements = r |
|
76 | 76 | |
|
77 | 77 | if not supported: |
|
78 | 78 | return |
|
79 | 79 | |
|
80 | streamin(repo, remote, requirements) | |
|
80 | # Save remote branchmap. We will use it later to speed up branchcache | |
|
81 | # creation. | |
|
82 | rbranchmap = None | |
|
83 | if remote.capable('branchmap'): | |
|
84 | rbranchmap = remote.branchmap() | |
|
85 | ||
|
86 | fp = remote.stream_out() | |
|
87 | l = fp.readline() | |
|
88 | try: | |
|
89 | resp = int(l) | |
|
90 | except ValueError: | |
|
91 | raise error.ResponseError( | |
|
92 | _('unexpected response from remote server:'), l) | |
|
93 | if resp == 1: | |
|
94 | raise util.Abort(_('operation forbidden by server')) | |
|
95 | elif resp == 2: | |
|
96 | raise util.Abort(_('locking the remote repository failed')) | |
|
97 | elif resp != 0: | |
|
98 | raise util.Abort(_('the server sent an unknown error code')) | |
|
99 | ||
|
100 | applyremotedata(repo, requirements, rbranchmap, fp) | |
|
81 | 101 | |
|
82 | 102 | def allowservergeneration(ui): |
|
83 | 103 | """Whether streaming clones are allowed from the server.""" |
|
84 | 104 | return ui.configbool('server', 'uncompressed', True, untrusted=True) |
|
85 | 105 | |
|
86 | 106 | # This is it's own function so extensions can override it. |
|
87 | 107 | def _walkstreamfiles(repo): |
|
88 | 108 | return repo.store.walk() |
|
89 | 109 | |
|
90 | 110 | def generatev1(repo): |
|
91 | 111 | """Emit content for version 1 of a streaming clone. |
|
92 | 112 | |
|
93 | 113 | This is a generator of raw chunks that constitute a streaming clone. |
|
94 | 114 | |
|
95 | 115 | The stream begins with a line of 2 space-delimited integers containing the |
|
96 | 116 | number of entries and total bytes size. |
|
97 | 117 | |
|
98 | 118 | Next, are N entries for each file being transferred. Each file entry starts |
|
99 | 119 | as a line with the file name and integer size delimited by a null byte. |
|
100 | 120 | The raw file data follows. Following the raw file data is the next file |
|
101 | 121 | entry, or EOF. |
|
102 | 122 | |
|
103 | 123 | When used on the wire protocol, an additional line indicating protocol |
|
104 | 124 | success will be prepended to the stream. This function is not responsible |
|
105 | 125 | for adding it. |
|
106 | 126 | |
|
107 | 127 | This function will obtain a repository lock to ensure a consistent view of |
|
108 | 128 | the store is captured. It therefore may raise LockError. |
|
109 | 129 | """ |
|
110 | 130 | entries = [] |
|
111 | 131 | total_bytes = 0 |
|
112 | 132 | # Get consistent snapshot of repo, lock during scan. |
|
113 | 133 | lock = repo.lock() |
|
114 | 134 | try: |
|
115 | 135 | repo.ui.debug('scanning\n') |
|
116 | 136 | for name, ename, size in _walkstreamfiles(repo): |
|
117 | 137 | if size: |
|
118 | 138 | entries.append((name, size)) |
|
119 | 139 | total_bytes += size |
|
120 | 140 | finally: |
|
121 | 141 | lock.release() |
|
122 | 142 | |
|
123 | 143 | repo.ui.debug('%d files, %d bytes to transfer\n' % |
|
124 | 144 | (len(entries), total_bytes)) |
|
125 | 145 | yield '%d %d\n' % (len(entries), total_bytes) |
|
126 | 146 | |
|
127 | 147 | svfs = repo.svfs |
|
128 | 148 | oldaudit = svfs.mustaudit |
|
129 | 149 | debugflag = repo.ui.debugflag |
|
130 | 150 | svfs.mustaudit = False |
|
131 | 151 | |
|
132 | 152 | try: |
|
133 | 153 | for name, size in entries: |
|
134 | 154 | if debugflag: |
|
135 | 155 | repo.ui.debug('sending %s (%d bytes)\n' % (name, size)) |
|
136 | 156 | # partially encode name over the wire for backwards compat |
|
137 | 157 | yield '%s\0%d\n' % (store.encodedir(name), size) |
|
138 | 158 | if size <= 65536: |
|
139 | 159 | fp = svfs(name) |
|
140 | 160 | try: |
|
141 | 161 | data = fp.read(size) |
|
142 | 162 | finally: |
|
143 | 163 | fp.close() |
|
144 | 164 | yield data |
|
145 | 165 | else: |
|
146 | 166 | for chunk in util.filechunkiter(svfs(name), limit=size): |
|
147 | 167 | yield chunk |
|
148 | 168 | finally: |
|
149 | 169 | svfs.mustaudit = oldaudit |
|
150 | 170 | |
|
151 | 171 | def consumev1(repo, fp): |
|
152 | 172 | """Apply the contents from version 1 of a streaming clone file handle. |
|
153 | 173 | |
|
154 | 174 | This takes the output from "streamout" and applies it to the specified |
|
155 | 175 | repository. |
|
156 | 176 | |
|
157 | 177 | Like "streamout," the status line added by the wire protocol is not handled |
|
158 | 178 | by this function. |
|
159 | 179 | """ |
|
160 | 180 | lock = repo.lock() |
|
161 | 181 | try: |
|
162 | 182 | repo.ui.status(_('streaming all changes\n')) |
|
163 | 183 | l = fp.readline() |
|
164 | 184 | try: |
|
165 | 185 | total_files, total_bytes = map(int, l.split(' ', 1)) |
|
166 | 186 | except (ValueError, TypeError): |
|
167 | 187 | raise error.ResponseError( |
|
168 | 188 | _('unexpected response from remote server:'), l) |
|
169 | 189 | repo.ui.status(_('%d files to transfer, %s of data\n') % |
|
170 | 190 | (total_files, util.bytecount(total_bytes))) |
|
171 | 191 | handled_bytes = 0 |
|
172 | 192 | repo.ui.progress(_('clone'), 0, total=total_bytes) |
|
173 | 193 | start = time.time() |
|
174 | 194 | |
|
175 | 195 | tr = repo.transaction(_('clone')) |
|
176 | 196 | try: |
|
177 | 197 | for i in xrange(total_files): |
|
178 | 198 | # XXX doesn't support '\n' or '\r' in filenames |
|
179 | 199 | l = fp.readline() |
|
180 | 200 | try: |
|
181 | 201 | name, size = l.split('\0', 1) |
|
182 | 202 | size = int(size) |
|
183 | 203 | except (ValueError, TypeError): |
|
184 | 204 | raise error.ResponseError( |
|
185 | 205 | _('unexpected response from remote server:'), l) |
|
186 | 206 | if repo.ui.debugflag: |
|
187 | 207 | repo.ui.debug('adding %s (%s)\n' % |
|
188 | 208 | (name, util.bytecount(size))) |
|
189 | 209 | # for backwards compat, name was partially encoded |
|
190 | 210 | ofp = repo.svfs(store.decodedir(name), 'w') |
|
191 | 211 | for chunk in util.filechunkiter(fp, limit=size): |
|
192 | 212 | handled_bytes += len(chunk) |
|
193 | 213 | repo.ui.progress(_('clone'), handled_bytes, |
|
194 | 214 | total=total_bytes) |
|
195 | 215 | ofp.write(chunk) |
|
196 | 216 | ofp.close() |
|
197 | 217 | tr.close() |
|
198 | 218 | finally: |
|
199 | 219 | tr.release() |
|
200 | 220 | |
|
201 | 221 | # Writing straight to files circumvented the inmemory caches |
|
202 | 222 | repo.invalidate() |
|
203 | 223 | |
|
204 | 224 | elapsed = time.time() - start |
|
205 | 225 | if elapsed <= 0: |
|
206 | 226 | elapsed = 0.001 |
|
207 | 227 | repo.ui.progress(_('clone'), None) |
|
208 | 228 | repo.ui.status(_('transferred %s in %.1f seconds (%s/sec)\n') % |
|
209 | 229 | (util.bytecount(total_bytes), elapsed, |
|
210 | 230 | util.bytecount(total_bytes / elapsed))) |
|
211 | 231 | finally: |
|
212 | 232 | lock.release() |
|
213 | 233 | |
|
214 | def streamin(repo, remote, remotereqs): | |
|
215 | # Save remote branchmap. We will use it later | |
|
216 | # to speed up branchcache creation | |
|
217 | rbranchmap = None | |
|
218 | if remote.capable("branchmap"): | |
|
219 | rbranchmap = remote.branchmap() | |
|
220 | ||
|
221 | fp = remote.stream_out() | |
|
222 | l = fp.readline() | |
|
223 | try: | |
|
224 | resp = int(l) | |
|
225 | except ValueError: | |
|
226 | raise error.ResponseError( | |
|
227 | _('unexpected response from remote server:'), l) | |
|
228 | if resp == 1: | |
|
229 | raise util.Abort(_('operation forbidden by server')) | |
|
230 | elif resp == 2: | |
|
231 | raise util.Abort(_('locking the remote repository failed')) | |
|
232 | elif resp != 0: | |
|
233 | raise util.Abort(_('the server sent an unknown error code')) | |
|
234 | ||
|
235 | applyremotedata(repo, remotereqs, rbranchmap, fp) | |
|
236 | return len(repo.heads()) + 1 | |
|
237 | ||
|
238 | 234 | def applyremotedata(repo, remotereqs, remotebranchmap, fp): |
|
239 | 235 | """Apply stream clone data to a repository. |
|
240 | 236 | |
|
241 | 237 | "remotereqs" is a set of requirements to handle the incoming data. |
|
242 | 238 | "remotebranchmap" is the result of a branchmap lookup on the remote. It |
|
243 | 239 | can be None. |
|
244 | 240 | "fp" is a file object containing the raw stream data, suitable for |
|
245 | 241 | feeding into consumev1(). |
|
246 | 242 | """ |
|
247 | 243 | lock = repo.lock() |
|
248 | 244 | try: |
|
249 | 245 | consumev1(repo, fp) |
|
250 | 246 | |
|
251 | 247 | # new requirements = old non-format requirements + |
|
252 | 248 | # new format-related remote requirements |
|
253 | 249 | # requirements from the streamed-in repository |
|
254 | 250 | repo.requirements = remotereqs | ( |
|
255 | 251 | repo.requirements - repo.supportedformats) |
|
256 | 252 | repo._applyopenerreqs() |
|
257 | 253 | repo._writerequirements() |
|
258 | 254 | |
|
259 | 255 | if remotebranchmap: |
|
260 | 256 | rbheads = [] |
|
261 | 257 | closed = [] |
|
262 | 258 | for bheads in remotebranchmap.itervalues(): |
|
263 | 259 | rbheads.extend(bheads) |
|
264 | 260 | for h in bheads: |
|
265 | 261 | r = repo.changelog.rev(h) |
|
266 | 262 | b, c = repo.changelog.branchinfo(r) |
|
267 | 263 | if c: |
|
268 | 264 | closed.append(h) |
|
269 | 265 | |
|
270 | 266 | if rbheads: |
|
271 | 267 | rtiprev = max((int(repo.changelog.rev(node)) |
|
272 | 268 | for node in rbheads)) |
|
273 | 269 | cache = branchmap.branchcache(remotebranchmap, |
|
274 | 270 | repo[rtiprev].node(), |
|
275 | 271 | rtiprev, |
|
276 | 272 | closednodes=closed) |
|
277 | 273 | # Try to stick it as low as possible |
|
278 | 274 | # filter above served are unlikely to be fetch from a clone |
|
279 | 275 | for candidate in ('base', 'immutable', 'served'): |
|
280 | 276 | rview = repo.filtered(candidate) |
|
281 | 277 | if cache.validfor(rview): |
|
282 | 278 | repo._branchcaches[candidate] = cache |
|
283 | 279 | cache.write(rview) |
|
284 | 280 | break |
|
285 | 281 | repo.invalidate() |
|
286 | 282 | finally: |
|
287 | 283 | lock.release() |
General Comments 0
You need to be logged in to leave comments.
Login now