Branch data Line data Source code
1 : : /* Debuginfo-over-http server.
2 : : Copyright (C) 2019-2024 Red Hat, Inc.
3 : : Copyright (C) 2021, 2022 Mark J. Wielaard <mark@klomp.org>
4 : : This file is part of elfutils.
5 : :
6 : : This file is free software; you can redistribute it and/or modify
7 : : it under the terms of the GNU General Public License as published by
8 : : the Free Software Foundation; either version 3 of the License, or
9 : : (at your option) any later version.
10 : :
11 : : elfutils is distributed in the hope that it will be useful, but
12 : : WITHOUT ANY WARRANTY; without even the implied warranty of
13 : : MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 : : GNU General Public License for more details.
15 : :
16 : : You should have received a copy of the GNU General Public License
17 : : along with this program. If not, see <http://www.gnu.org/licenses/>. */
18 : :
19 : :
20 : : /* cargo-cult from libdwfl linux-kernel-modules.c */
21 : : /* In case we have a bad fts we include this before config.h because it
22 : : can't handle _FILE_OFFSET_BITS.
23 : : Everything we need here is fine if its declarations just come first.
24 : : Also, include sys/types.h before fts. On some systems fts.h is not self
25 : : contained. */
26 : : #ifdef BAD_FTS
27 : : #include <sys/types.h>
28 : : #include <fts.h>
29 : : #endif
30 : :
31 : : #ifdef HAVE_CONFIG_H
32 : : #include "config.h"
33 : : #endif
34 : :
35 : : // #define _GNU_SOURCE
36 : : #ifdef HAVE_SCHED_H
37 : : extern "C" {
38 : : #include <sched.h>
39 : : }
40 : : #endif
41 : : #ifdef HAVE_SYS_RESOURCE_H
42 : : extern "C" {
43 : : #include <sys/resource.h>
44 : : }
45 : : #endif
46 : :
47 : : #ifdef HAVE_EXECINFO_H
48 : : extern "C" {
49 : : #include <execinfo.h>
50 : : }
51 : : #endif
52 : : #ifdef HAVE_MALLOC_H
53 : : extern "C" {
54 : : #include <malloc.h>
55 : : }
56 : : #endif
57 : :
58 : : #include "debuginfod.h"
59 : : #include <dwarf.h>
60 : :
61 : : #include <argp.h>
62 : : #ifdef __GNUC__
63 : : #undef __attribute__ /* glibc bug - rhbz 1763325 */
64 : : #endif
65 : :
66 : : #ifdef USE_LZMA
67 : : #include <lzma.h>
68 : : #endif
69 : :
70 : : #include <unistd.h>
71 : : #include <stdlib.h>
72 : : #include <locale.h>
73 : : #include <pthread.h>
74 : : #include <signal.h>
75 : : #include <sys/stat.h>
76 : : #include <sys/time.h>
77 : : #include <sys/vfs.h>
78 : : #include <unistd.h>
79 : : #include <fcntl.h>
80 : : #include <netdb.h>
81 : : #include <math.h>
82 : : #include <float.h>
83 : : #include <fnmatch.h>
84 : : #include <arpa/inet.h>
85 : :
86 : :
87 : : /* If fts.h is included before config.h, its indirect inclusions may not
88 : : give us the right LFS aliases of these functions, so map them manually. */
89 : : #ifdef BAD_FTS
90 : : #ifdef _FILE_OFFSET_BITS
91 : : #define open open64
92 : : #define fopen fopen64
93 : : #endif
94 : : #else
95 : : #include <sys/types.h>
96 : : #include <fts.h>
97 : : #endif
98 : :
99 : : #include <cstring>
100 : : #include <vector>
101 : : #include <set>
102 : : #include <unordered_set>
103 : : #include <map>
104 : : #include <string>
105 : : #include <iostream>
106 : : #include <iomanip>
107 : : #include <ostream>
108 : : #include <sstream>
109 : : #include <mutex>
110 : : #include <deque>
111 : : #include <condition_variable>
112 : : #include <exception>
113 : : #include <thread>
114 : : // #include <regex> // on rhel7 gcc 4.8, not competent
115 : : #include <regex.h>
116 : : // #include <algorithm>
117 : : using namespace std;
118 : :
119 : : #include <gelf.h>
120 : : #include <libdwelf.h>
121 : :
122 : : #include <microhttpd.h>
123 : :
124 : : #if MHD_VERSION >= 0x00097002
125 : : // libmicrohttpd 0.9.71 broke API
126 : : #define MHD_RESULT enum MHD_Result
127 : : #else
128 : : #define MHD_RESULT int
129 : : #endif
130 : :
131 : : #ifdef ENABLE_IMA_VERIFICATION
132 : : #include <rpm/rpmlib.h>
133 : : #include <rpm/rpmfi.h>
134 : : #include <rpm/header.h>
135 : : #include <glob.h>
136 : : #endif
137 : :
138 : : #include <curl/curl.h>
139 : : #include <archive.h>
140 : : #include <archive_entry.h>
141 : : #include <sqlite3.h>
142 : :
143 : : #ifdef __linux__
144 : : #include <sys/syscall.h>
145 : : #endif
146 : :
147 : : #ifdef __linux__
148 : : #define tid() syscall(SYS_gettid)
149 : : #else
150 : : #define tid() pthread_self()
151 : : #endif
152 : :
153 : : extern "C" {
154 : : #include "printversion.h"
155 : : #include "system.h"
156 : : }
157 : : #include <json-c/json.h>
158 : :
159 : :
160 : : inline bool
161 : 141643 : string_endswith(const string& haystack, const string& needle)
162 : : {
163 [ + + ]: 141643 : return (haystack.size() >= needle.size() &&
164 : 138551 : equal(haystack.end()-needle.size(), haystack.end(),
165 : 141643 : needle.begin()));
166 : : }
167 : :
168 : :
169 : : // Roll this identifier for every sqlite schema incompatibility.
170 : : #define BUILDIDS "buildids10"
171 : :
172 : : #if SQLITE_VERSION_NUMBER >= 3008000
173 : : #define WITHOUT_ROWID "without rowid"
174 : : #else
175 : : #define WITHOUT_ROWID ""
176 : : #endif
177 : :
178 : : static const char DEBUGINFOD_SQLITE_DDL[] =
179 : : "pragma foreign_keys = on;\n"
180 : : "pragma synchronous = 0;\n" // disable fsync()s - this cache is disposable across a machine crash
181 : : "pragma journal_mode = wal;\n" // https://sqlite.org/wal.html
182 : : "pragma wal_checkpoint = truncate;\n" // clean out any preexisting wal file
183 : : "pragma journal_size_limit = 0;\n" // limit steady state file (between grooming, which also =truncate's)
184 : : "pragma auto_vacuum = incremental;\n" // https://sqlite.org/pragma.html
185 : : "pragma busy_timeout = 1000;\n" // https://sqlite.org/pragma.html
186 : : // NB: all these are overridable with -D option
187 : :
188 : : // Normalization table for interning file names
189 : : "create table if not exists " BUILDIDS "_fileparts (\n"
190 : : " id integer primary key not null,\n"
191 : : " name text unique not null\n"
192 : : " );\n"
193 : : "create table if not exists " BUILDIDS "_files (\n"
194 : : " id integer primary key not null,\n"
195 : : " dirname integer not null,\n"
196 : : " basename integer not null,\n"
197 : : " unique (dirname, basename),\n"
198 : : " foreign key (dirname) references " BUILDIDS "_fileparts(id) on delete cascade,\n"
199 : : " foreign key (basename) references " BUILDIDS "_fileparts(id) on delete cascade\n"
200 : : " );\n"
201 : : "create view if not exists " BUILDIDS "_files_v as\n" // a
202 : : " select f.id, n1.name || '/' || n2.name as name\n"
203 : : " from " BUILDIDS "_files f, " BUILDIDS "_fileparts n1, " BUILDIDS "_fileparts n2\n"
204 : : " where f.dirname = n1.id and f.basename = n2.id;\n"
205 : :
206 : : // Normalization table for interning buildids
207 : : "create table if not exists " BUILDIDS "_buildids (\n"
208 : : " id integer primary key not null,\n"
209 : : " hex text unique not null);\n"
210 : : // Track the completion of scanning of a given file & sourcetype at given time
211 : : "create table if not exists " BUILDIDS "_file_mtime_scanned (\n"
212 : : " mtime integer not null,\n"
213 : : " file integer not null,\n"
214 : : " size integer not null,\n" // in bytes
215 : : " sourcetype text(1) not null\n"
216 : : " check (sourcetype IN ('F', 'R')),\n"
217 : : " foreign key (file) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
218 : : " primary key (file, mtime, sourcetype)\n"
219 : : " ) " WITHOUT_ROWID ";\n"
220 : : "create table if not exists " BUILDIDS "_f_de (\n"
221 : : " buildid integer not null,\n"
222 : : " debuginfo_p integer not null,\n"
223 : : " executable_p integer not null,\n"
224 : : " file integer not null,\n"
225 : : " mtime integer not null,\n"
226 : : " foreign key (file) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
227 : : " foreign key (buildid) references " BUILDIDS "_buildids(id) on update cascade on delete cascade,\n"
228 : : " primary key (buildid, file, mtime)\n"
229 : : " ) " WITHOUT_ROWID ";\n"
230 : : // Index for faster delete by file identifier and metadata searches
231 : : "create index if not exists " BUILDIDS "_f_de_idx on " BUILDIDS "_f_de (file, mtime);\n"
232 : : "create table if not exists " BUILDIDS "_f_s (\n"
233 : : " buildid integer not null,\n"
234 : : " artifactsrc integer not null,\n"
235 : : " file integer not null,\n" // NB: not necessarily entered into _mtime_scanned
236 : : " mtime integer not null,\n"
237 : : " foreign key (file) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
238 : : " foreign key (artifactsrc) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
239 : : " foreign key (buildid) references " BUILDIDS "_buildids(id) on update cascade on delete cascade,\n"
240 : : " primary key (buildid, artifactsrc, file, mtime)\n"
241 : : " ) " WITHOUT_ROWID ";\n"
242 : : "create table if not exists " BUILDIDS "_r_de (\n"
243 : : " buildid integer not null,\n"
244 : : " debuginfo_p integer not null,\n"
245 : : " executable_p integer not null,\n"
246 : : " file integer not null,\n"
247 : : " mtime integer not null,\n"
248 : : " content integer not null,\n"
249 : : " foreign key (file) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
250 : : " foreign key (content) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
251 : : " foreign key (buildid) references " BUILDIDS "_buildids(id) on update cascade on delete cascade,\n"
252 : : " primary key (buildid, debuginfo_p, executable_p, file, content, mtime)\n"
253 : : " ) " WITHOUT_ROWID ";\n"
254 : : // Index for faster delete by archive file identifier
255 : : "create index if not exists " BUILDIDS "_r_de_idx on " BUILDIDS "_r_de (file, mtime);\n"
256 : : // Index for metadata searches
257 : : "create index if not exists " BUILDIDS "_r_de_idx2 on " BUILDIDS "_r_de (content);\n"
258 : : "create table if not exists " BUILDIDS "_r_sref (\n" // outgoing dwarf sourcefile references from rpm
259 : : " buildid integer not null,\n"
260 : : " artifactsrc integer not null,\n"
261 : : " foreign key (artifactsrc) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
262 : : " foreign key (buildid) references " BUILDIDS "_buildids(id) on update cascade on delete cascade,\n"
263 : : " primary key (buildid, artifactsrc)\n"
264 : : " ) " WITHOUT_ROWID ";\n"
265 : : "create table if not exists " BUILDIDS "_r_sdef (\n" // rpm contents that may satisfy sref
266 : : " file integer not null,\n"
267 : : " mtime integer not null,\n"
268 : : " content integer not null,\n"
269 : : " foreign key (file) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
270 : : " foreign key (content) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
271 : : " primary key (content, file, mtime)\n"
272 : : " ) " WITHOUT_ROWID ";\n"
273 : : "create table if not exists " BUILDIDS "_r_seekable (\n" // seekable rpm contents
274 : : " file integer not null,\n"
275 : : " content integer not null,\n"
276 : : " type text not null,\n"
277 : : " size integer not null,\n"
278 : : " offset integer not null,\n"
279 : : " mtime integer not null,\n"
280 : : " foreign key (file) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
281 : : " foreign key (content) references " BUILDIDS "_files(id) on update cascade on delete cascade,\n"
282 : : " primary key (file, content)\n"
283 : : " ) " WITHOUT_ROWID ";\n"
284 : : // create views to glue together some of the above tables, for webapi D queries
285 : : // NB: _query_d2 and _query_e2 were added to replace _query_d and _query_e
286 : : // without updating BUILDIDS. They can be renamed back the next time BUILDIDS
287 : : // is updated.
288 : : "create view if not exists " BUILDIDS "_query_d2 as \n"
289 : : "select\n"
290 : : " b.hex as buildid, 'F' as sourcetype, n.file as id0, f0.name as source0, n.mtime as mtime, null as id1, null as source1\n"
291 : : " from " BUILDIDS "_buildids b, " BUILDIDS "_files_v f0, " BUILDIDS "_f_de n\n"
292 : : " where b.id = n.buildid and f0.id = n.file and n.debuginfo_p = 1\n"
293 : : "union all select\n"
294 : : " b.hex as buildid, 'R' as sourcetype, n.file as id0, f0.name as source0, n.mtime as mtime, n.content as id1, f1.name as source1\n"
295 : : " from " BUILDIDS "_buildids b, " BUILDIDS "_files_v f0, " BUILDIDS "_files_v f1, " BUILDIDS "_r_de n\n"
296 : : " where b.id = n.buildid and f0.id = n.file and f1.id = n.content and n.debuginfo_p = 1\n"
297 : : ";"
298 : : // ... and for E queries
299 : : "create view if not exists " BUILDIDS "_query_e2 as \n"
300 : : "select\n"
301 : : " b.hex as buildid, 'F' as sourcetype, n.file as id0, f0.name as source0, n.mtime as mtime, null as id1, null as source1\n"
302 : : " from " BUILDIDS "_buildids b, " BUILDIDS "_files_v f0, " BUILDIDS "_f_de n\n"
303 : : " where b.id = n.buildid and f0.id = n.file and n.executable_p = 1\n"
304 : : "union all select\n"
305 : : " b.hex as buildid, 'R' as sourcetype, n.file as id0, f0.name as source0, n.mtime as mtime, n.content as id1, f1.name as source1\n"
306 : : " from " BUILDIDS "_buildids b, " BUILDIDS "_files_v f0, " BUILDIDS "_files_v f1, " BUILDIDS "_r_de n\n"
307 : : " where b.id = n.buildid and f0.id = n.file and f1.id = n.content and n.executable_p = 1\n"
308 : : ";"
309 : : // ... and for S queries
310 : : "create view if not exists " BUILDIDS "_query_s as \n"
311 : : "select\n"
312 : : " b.hex as buildid, fs.name as artifactsrc, 'F' as sourcetype, f0.name as source0, n.mtime as mtime, null as source1, null as source0ref\n"
313 : : " from " BUILDIDS "_buildids b, " BUILDIDS "_files_v f0, " BUILDIDS "_files_v fs, " BUILDIDS "_f_s n\n"
314 : : " where b.id = n.buildid and f0.id = n.file and fs.id = n.artifactsrc\n"
315 : : "union all select\n"
316 : : " b.hex as buildid, f1.name as artifactsrc, 'R' as sourcetype, f0.name as source0, sd.mtime as mtime, f1.name as source1, fsref.name as source0ref\n"
317 : : " from " BUILDIDS "_buildids b, " BUILDIDS "_files_v f0, " BUILDIDS "_files_v f1, " BUILDIDS "_files_v fsref, "
318 : : " " BUILDIDS "_r_sdef sd, " BUILDIDS "_r_sref sr, " BUILDIDS "_r_de sde\n"
319 : : " where b.id = sr.buildid and f0.id = sd.file and fsref.id = sde.file and f1.id = sd.content\n"
320 : : " and sr.artifactsrc = sd.content and sde.buildid = sr.buildid\n"
321 : : ";"
322 : : // and for startup overview counts
323 : : "drop view if exists " BUILDIDS "_stats;\n"
324 : : "create view if not exists " BUILDIDS "_stats as\n"
325 : : " select 'file d/e' as label,count(*) as quantity from " BUILDIDS "_f_de\n"
326 : : "union all select 'file s',count(*) from " BUILDIDS "_f_s\n"
327 : : "union all select 'archive d/e',count(*) from " BUILDIDS "_r_de\n"
328 : : "union all select 'archive sref',count(*) from " BUILDIDS "_r_sref\n"
329 : : "union all select 'archive sdef',count(*) from " BUILDIDS "_r_sdef\n"
330 : : "union all select 'buildids',count(*) from " BUILDIDS "_buildids\n"
331 : : "union all select 'filenames',count(*) from " BUILDIDS "_files\n"
332 : : "union all select 'fileparts',count(*) from " BUILDIDS "_fileparts\n"
333 : : "union all select 'files scanned (#)',count(*) from " BUILDIDS "_file_mtime_scanned\n"
334 : : "union all select 'files scanned (mb)',coalesce(sum(size)/1024/1024,0) from " BUILDIDS "_file_mtime_scanned\n"
335 : : #if SQLITE_VERSION_NUMBER >= 3016000
336 : : "union all select 'index db size (mb)',page_count*page_size/1024/1024 as size FROM pragma_page_count(), pragma_page_size()\n"
337 : : #endif
338 : : ";\n"
339 : :
340 : : // schema change history & garbage collection
341 : : //
342 : : // XXX: we could have migration queries here to bring prior-schema
343 : : // data over instead of just dropping it. But that could incur
344 : : // doubled storage costs.
345 : : //
346 : : // buildids10: split the _files table into _parts
347 : : "" // <<< we are here
348 : : // buildids9: widen the mtime_scanned table
349 : : "DROP VIEW IF EXISTS buildids9_stats;\n"
350 : : "DROP INDEX IF EXISTS buildids9_r_de_idx;\n"
351 : : "DROP INDEX IF EXISTS buildids9_f_de_idx;\n"
352 : : "DROP VIEW IF EXISTS buildids9_query_s;\n"
353 : : "DROP VIEW IF EXISTS buildids9_query_e;\n"
354 : : "DROP VIEW IF EXISTS buildids9_query_d;\n"
355 : : "DROP TABLE IF EXISTS buildids9_r_sdef;\n"
356 : : "DROP TABLE IF EXISTS buildids9_r_sref;\n"
357 : : "DROP TABLE IF EXISTS buildids9_r_de;\n"
358 : : "DROP TABLE IF EXISTS buildids9_f_s;\n"
359 : : "DROP TABLE IF EXISTS buildids9_f_de;\n"
360 : : "DROP TABLE IF EXISTS buildids9_file_mtime_scanned;\n"
361 : : "DROP TABLE IF EXISTS buildids9_buildids;\n"
362 : : "DROP TABLE IF EXISTS buildids9_files;\n"
363 : : // buildids8: slim the sref table
364 : : "drop table if exists buildids8_f_de;\n"
365 : : "drop table if exists buildids8_f_s;\n"
366 : : "drop table if exists buildids8_r_de;\n"
367 : : "drop table if exists buildids8_r_sref;\n"
368 : : "drop table if exists buildids8_r_sdef;\n"
369 : : "drop table if exists buildids8_file_mtime_scanned;\n"
370 : : "drop table if exists buildids8_files;\n"
371 : : "drop table if exists buildids8_buildids;\n"
372 : : // buildids7: separate _norm table into dense subtype tables
373 : : "drop table if exists buildids7_f_de;\n"
374 : : "drop table if exists buildids7_f_s;\n"
375 : : "drop table if exists buildids7_r_de;\n"
376 : : "drop table if exists buildids7_r_sref;\n"
377 : : "drop table if exists buildids7_r_sdef;\n"
378 : : "drop table if exists buildids7_file_mtime_scanned;\n"
379 : : "drop table if exists buildids7_files;\n"
380 : : "drop table if exists buildids7_buildids;\n"
381 : : // buildids6: drop bolo/rfolo again, represent sources / rpmcontents in main table
382 : : "drop table if exists buildids6_norm;\n"
383 : : "drop table if exists buildids6_files;\n"
384 : : "drop table if exists buildids6_buildids;\n"
385 : : "drop view if exists buildids6;\n"
386 : : // buildids5: redefine srcfile1 column to be '.'-less (for rpms)
387 : : "drop table if exists buildids5_norm;\n"
388 : : "drop table if exists buildids5_files;\n"
389 : : "drop table if exists buildids5_buildids;\n"
390 : : "drop table if exists buildids5_bolo;\n"
391 : : "drop table if exists buildids5_rfolo;\n"
392 : : "drop view if exists buildids5;\n"
393 : : // buildids4: introduce rpmfile RFOLO
394 : : "drop table if exists buildids4_norm;\n"
395 : : "drop table if exists buildids4_files;\n"
396 : : "drop table if exists buildids4_buildids;\n"
397 : : "drop table if exists buildids4_bolo;\n"
398 : : "drop table if exists buildids4_rfolo;\n"
399 : : "drop view if exists buildids4;\n"
400 : : // buildids3*: split out srcfile BOLO
401 : : "drop table if exists buildids3_norm;\n"
402 : : "drop table if exists buildids3_files;\n"
403 : : "drop table if exists buildids3_buildids;\n"
404 : : "drop table if exists buildids3_bolo;\n"
405 : : "drop view if exists buildids3;\n"
406 : : // buildids2: normalized buildid and filenames into interning tables;
407 : : "drop table if exists buildids2_norm;\n"
408 : : "drop table if exists buildids2_files;\n"
409 : : "drop table if exists buildids2_buildids;\n"
410 : : "drop view if exists buildids2;\n"
411 : : // buildids1: made buildid and artifacttype NULLable, to represent cached-negative
412 : : // lookups from sources, e.g. files or rpms that contain no buildid-indexable content
413 : : "drop table if exists buildids1;\n"
414 : : // buildids: original
415 : : "drop table if exists buildids;\n"
416 : : ;
417 : :
418 : : static const char DEBUGINFOD_SQLITE_CLEANUP_DDL[] =
419 : : "pragma wal_checkpoint = truncate;\n" // clean out any preexisting wal file
420 : : ;
421 : :
422 : :
423 : :
424 : :
425 : : /* Name and version of program. */
426 : : ARGP_PROGRAM_VERSION_HOOK_DEF = print_version;
427 : :
428 : : /* Bug report address. */
429 : : ARGP_PROGRAM_BUG_ADDRESS_DEF = PACKAGE_BUGREPORT;
430 : :
431 : : /* Definitions of arguments for argp functions. */
432 : : static const struct argp_option options[] =
433 : : {
434 : : { NULL, 0, NULL, 0, "Scanners:", 1 },
435 : : { "scan-file-dir", 'F', NULL, 0, "Enable ELF/DWARF file scanning.", 0 },
436 : : { "scan-rpm-dir", 'R', NULL, 0, "Enable RPM scanning.", 0 },
437 : : { "scan-deb-dir", 'U', NULL, 0, "Enable DEB scanning.", 0 },
438 : : { "scan-archive", 'Z', "EXT=CMD", 0, "Enable arbitrary archive scanning.", 0 },
439 : : // "source-oci-imageregistry" ...
440 : :
441 : : { NULL, 0, NULL, 0, "Options:", 2 },
442 : : { "logical", 'L', NULL, 0, "Follow symlinks, default=ignore.", 0 },
443 : : { "rescan-time", 't', "SECONDS", 0, "Number of seconds to wait between rescans, 0=disable.", 0 },
444 : : { "groom-time", 'g', "SECONDS", 0, "Number of seconds to wait between database grooming, 0=disable.", 0 },
445 : : { "maxigroom", 'G', NULL, 0, "Run a complete database groom/shrink pass at startup.", 0 },
446 : : { "concurrency", 'c', "NUM", 0, "Limit scanning thread concurrency to NUM, default=#CPUs.", 0 },
447 : : { "connection-pool", 'C', "NUM", OPTION_ARG_OPTIONAL,
448 : : "Use webapi connection pool with NUM threads, default=unlim.", 0 },
449 : : { "include", 'I', "REGEX", 0, "Include files matching REGEX, default=all.", 0 },
450 : : { "exclude", 'X', "REGEX", 0, "Exclude files matching REGEX, default=none.", 0 },
451 : : { "port", 'p', "NUM", 0, "HTTP port to listen on, default 8002.", 0 },
452 : : #define ARGP_KEY_CORS 0x1000
453 : : { "cors", ARGP_KEY_CORS, NULL, 0, "Add CORS response headers to HTTP queries, default no.", 0 },
454 : : { "database", 'd', "FILE", 0, "Path to sqlite database.", 0 },
455 : : { "ddl", 'D', "SQL", 0, "Apply extra sqlite ddl/pragma to connection.", 0 },
456 : : { "verbose", 'v', NULL, 0, "Increase verbosity.", 0 },
457 : : { "regex-groom", 'r', NULL, 0,"Uses regexes from -I and -X arguments to groom the database.",0},
458 : : #define ARGP_KEY_FDCACHE_FDS 0x1001
459 : : { "fdcache-fds", ARGP_KEY_FDCACHE_FDS, "NUM", OPTION_HIDDEN, NULL, 0 },
460 : : #define ARGP_KEY_FDCACHE_MBS 0x1002
461 : : { "fdcache-mbs", ARGP_KEY_FDCACHE_MBS, "MB", 0, "Maximum total size of archive file fdcache.", 0 },
462 : : #define ARGP_KEY_FDCACHE_PREFETCH 0x1003
463 : : { "fdcache-prefetch", ARGP_KEY_FDCACHE_PREFETCH, "NUM", 0, "Number of archive files to prefetch into fdcache.", 0 },
464 : : #define ARGP_KEY_FDCACHE_MINTMP 0x1004
465 : : { "fdcache-mintmp", ARGP_KEY_FDCACHE_MINTMP, "NUM", 0, "Minimum free space% on tmpdir.", 0 },
466 : : #define ARGP_KEY_FDCACHE_PREFETCH_MBS 0x1005
467 : : { "fdcache-prefetch-mbs", ARGP_KEY_FDCACHE_PREFETCH_MBS, "MB", OPTION_HIDDEN, NULL, 0},
468 : : #define ARGP_KEY_FDCACHE_PREFETCH_FDS 0x1006
469 : : { "fdcache-prefetch-fds", ARGP_KEY_FDCACHE_PREFETCH_FDS, "NUM", OPTION_HIDDEN, NULL, 0},
470 : : #define ARGP_KEY_FORWARDED_TTL_LIMIT 0x1007
471 : : {"forwarded-ttl-limit", ARGP_KEY_FORWARDED_TTL_LIMIT, "NUM", 0, "Limit of X-Forwarded-For hops, default 8.", 0},
472 : : #define ARGP_KEY_PASSIVE 0x1008
473 : : { "passive", ARGP_KEY_PASSIVE, NULL, 0, "Do not scan or groom, read-only database.", 0 },
474 : : #define ARGP_KEY_DISABLE_SOURCE_SCAN 0x1009
475 : : { "disable-source-scan", ARGP_KEY_DISABLE_SOURCE_SCAN, NULL, 0, "Do not scan dwarf source info.", 0 },
476 : : #define ARGP_SCAN_CHECKPOINT 0x100A
477 : : { "scan-checkpoint", ARGP_SCAN_CHECKPOINT, "NUM", 0, "Number of files scanned before a WAL checkpoint.", 0 },
478 : : #ifdef ENABLE_IMA_VERIFICATION
479 : : #define ARGP_KEY_KOJI_SIGCACHE 0x100B
480 : : { "koji-sigcache", ARGP_KEY_KOJI_SIGCACHE, NULL, 0, "Do a koji specific mapping of rpm paths to get IMA signatures.", 0 },
481 : : #endif
482 : : #define ARGP_KEY_METADATA_MAXTIME 0x100C
483 : : { "metadata-maxtime", ARGP_KEY_METADATA_MAXTIME, "SECONDS", 0,
484 : : "Number of seconds to limit metadata query run time, 0=unlimited.", 0 },
485 : : #define ARGP_KEY_HTTP_ADDR 0x100D
486 : : { "listen-address", ARGP_KEY_HTTP_ADDR, "ADDR", 0, "HTTP address to listen on.", 0 },
487 : : { "home-redirect", 'h', "URL", 0, "Custom homepage - redirect.", 0 },
488 : : { "home-html", 'H', "FILE", 0, "Custom homepage - htmlfile.", 0 },
489 : : { NULL, 0, NULL, 0, NULL, 0 },
490 : : };
491 : :
492 : : /* Short description of program. */
493 : : static const char doc[] = "Serve debuginfo-related content across HTTP from files under PATHs.";
494 : :
495 : : /* Strings for arguments in help texts. */
496 : : static const char args_doc[] = "[PATH ...]";
497 : :
498 : : /* Prototype for option handler. */
499 : : static error_t parse_opt (int key, char *arg, struct argp_state *state);
500 : :
501 : : static unsigned default_concurrency();
502 : :
503 : : /* Data structure to communicate with argp functions. */
504 : : static struct argp argp =
505 : : {
506 : : options, parse_opt, args_doc, doc, NULL, NULL, NULL
507 : : };
508 : :
509 : :
510 : : static string db_path;
511 : : static sqlite3 *db; // single connection, serialized across all our threads!
512 : : static sqlite3 *dbq; // webapi query-servicing readonly connection, serialized ditto!
513 : : static unsigned verbose;
514 : : static volatile sig_atomic_t interrupted = 0;
515 : : static volatile sig_atomic_t forced_rescan_count = 0;
516 : : static volatile sig_atomic_t sigusr1 = 0;
517 : : static volatile sig_atomic_t forced_groom_count = 0;
518 : : static volatile sig_atomic_t sigusr2 = 0;
519 : : static unsigned http_port = 8002;
520 : : static struct sockaddr_in6 http_sockaddr;
521 : : static string addr_info = "";
522 : : static bool webapi_cors = false;
523 : : static unsigned rescan_s = 300;
524 : : static unsigned groom_s = 86400;
525 : : static bool maxigroom = false;
526 : : static unsigned concurrency = default_concurrency();
527 : : static int connection_pool = 0;
528 : : static set<string> source_paths;
529 : : static bool scan_files = false;
530 : : static map<string,string> scan_archives;
531 : : static vector<string> extra_ddl;
532 : : static regex_t file_include_regex;
533 : : static regex_t file_exclude_regex;
534 : : static bool regex_groom = false;
535 : : static bool traverse_logical;
536 : : static long fdcache_mbs;
537 : : static long fdcache_prefetch;
538 : : static long fdcache_mintmp;
539 : : static unsigned forwarded_ttl_limit = 8;
540 : : static bool scan_source_info = true;
541 : : static string tmpdir;
542 : : static bool passive_p = false;
543 : : static long scan_checkpoint = 256;
544 : : #ifdef ENABLE_IMA_VERIFICATION
545 : : static bool requires_koji_sigcache_mapping = false;
546 : : #endif
547 : : static unsigned metadata_maxtime_s = 5;
548 : : static string cust_homepage_redirect = "";
549 : : static string cust_homepage_file = "";
550 : :
551 : : static void set_metric(const string& key, double value);
552 : : static void inc_metric(const string& key);
553 : : static void add_metric(const string& metric,
554 : : double value);
555 : : static void set_metric(const string& metric,
556 : : const string& lname, const string& lvalue,
557 : : double value);
558 : : static void inc_metric(const string& metric,
559 : : const string& lname, const string& lvalue);
560 : : static void add_metric(const string& metric,
561 : : const string& lname, const string& lvalue,
562 : : double value);
563 : : static void inc_metric(const string& metric,
564 : : const string& lname, const string& lvalue,
565 : : const string& rname, const string& rvalue);
566 : : static void add_metric(const string& metric,
567 : : const string& lname, const string& lvalue,
568 : : const string& rname, const string& rvalue,
569 : : double value);
570 : :
571 : :
572 : : class tmp_inc_metric { // a RAII style wrapper for exception-safe scoped increment & decrement
573 : : string m, n, v;
574 : : public:
575 : 3373 : tmp_inc_metric(const string& mname, const string& lname, const string& lvalue):
576 [ + - + - ]: 3373 : m(mname), n(lname), v(lvalue)
577 : : {
578 [ + - ]: 3373 : add_metric (m, n, v, 1);
579 : 3373 : }
580 : 3373 : ~tmp_inc_metric()
581 : : {
582 : 3373 : add_metric (m, n, v, -1);
583 : 3373 : }
584 : : };
585 : :
586 : : class tmp_ms_metric { // a RAII style wrapper for exception-safe scoped timing
587 : : string m, n, v;
588 : : struct timespec ts_start;
589 : : public:
590 : 359881 : tmp_ms_metric(const string& mname, const string& lname, const string& lvalue):
591 [ + - + - ]: 359881 : m(mname), n(lname), v(lvalue)
592 : : {
593 : 359846 : clock_gettime (CLOCK_MONOTONIC, & ts_start);
594 : 360255 : }
595 : 360914 : ~tmp_ms_metric()
596 : : {
597 : 360914 : struct timespec ts_end;
598 : 360914 : clock_gettime (CLOCK_MONOTONIC, & ts_end);
599 : 360990 : double deltas = (ts_end.tv_sec - ts_start.tv_sec)
600 : 360990 : + (ts_end.tv_nsec - ts_start.tv_nsec)/1.e9;
601 : :
602 : 360990 : add_metric (m + "_milliseconds_sum", n, v, (deltas*1000.0));
603 : 361104 : inc_metric (m + "_milliseconds_count", n, v);
604 : 361079 : }
605 : : };
606 : :
607 : :
608 : : /* Handle program arguments. */
609 : : static error_t
610 : 1276 : parse_opt (int key, char *arg,
611 : : struct argp_state *state __attribute__ ((unused)))
612 : : {
613 : 1276 : int rc;
614 [ + + + + : 1276 : switch (key)
+ + + + +
+ - + + -
- + + + +
+ + + + +
+ - + - -
+ + + ]
615 : : {
616 : 308 : case 'v': verbose ++; break;
617 : 80 : case 'd':
618 : : /* When using the in-memory database make sure it is shareable,
619 : : so we can open it twice as read/write and read-only. */
620 [ + + ]: 80 : if (strcmp (arg, ":memory:") == 0)
621 : 14 : db_path = "file::memory:?cache=shared";
622 : : else
623 : 66 : db_path = string(arg);
624 : : break;
625 : 84 : case 'p': http_port = (unsigned) atoi(arg);
626 [ + - ]: 84 : if (http_port == 0 || http_port > 65535)
627 : 0 : argp_failure(state, 1, EINVAL, "port number");
628 : : break;
629 : 6 : case ARGP_KEY_CORS:
630 : 6 : webapi_cors = true;
631 : 6 : break;
632 : 50 : case 'F': scan_files = true; break;
633 : 24 : case 'R':
634 [ + - + - ]: 24 : scan_archives[".rpm"]="cat"; // libarchive groks rpm natively
635 : 24 : break;
636 : 18 : case 'U':
637 [ + - + - ]: 18 : scan_archives[".deb"]="(bsdtar -O -x -f - data.tar\\*)<";
638 [ + - + - ]: 18 : scan_archives[".ddeb"]="(bsdtar -O -x -f - data.tar\\*)<";
639 [ + - + - ]: 18 : scan_archives[".ipk"]="(bsdtar -O -x -f - data.tar\\*)<";
640 : : // .udeb too?
641 : 18 : break;
642 : 40 : case 'Z':
643 : 40 : {
644 [ - + ]: 40 : char* extension = strchr(arg, '=');
645 [ - + ]: 40 : if (arg[0] == '\0')
646 : 0 : argp_failure(state, 1, EINVAL, "missing EXT");
647 [ + + ]: 40 : else if (extension)
648 [ + - + - ]: 20 : scan_archives[string(arg, (extension-arg))]=string(extension+1);
649 : : else
650 [ + - + - ]: 20 : scan_archives[string(arg)]=string("cat");
651 : : }
652 : : break;
653 : 8 : case 'L':
654 [ - + ]: 8 : if (passive_p)
655 : 0 : argp_failure(state, 1, EINVAL, "-L option inconsistent with passive mode");
656 : 8 : traverse_logical = true;
657 : 8 : break;
658 : 0 : case 'D':
659 [ # # ]: 0 : if (passive_p)
660 : 0 : argp_failure(state, 1, EINVAL, "-D option inconsistent with passive mode");
661 : 0 : extra_ddl.push_back(string(arg));
662 : 0 : break;
663 : 64 : case 't':
664 [ - + ]: 64 : if (passive_p)
665 : 0 : argp_failure(state, 1, EINVAL, "-t option inconsistent with passive mode");
666 : 64 : rescan_s = (unsigned) atoi(arg);
667 : 64 : break;
668 : 64 : case 'g':
669 [ - + ]: 64 : if (passive_p)
670 : 0 : argp_failure(state, 1, EINVAL, "-g option inconsistent with passive mode");
671 : 64 : groom_s = (unsigned) atoi(arg);
672 : 64 : break;
673 : 0 : case 'G':
674 [ # # ]: 0 : if (passive_p)
675 : 0 : argp_failure(state, 1, EINVAL, "-G option inconsistent with passive mode");
676 : 0 : maxigroom = true;
677 : 0 : break;
678 : 0 : case 'c':
679 [ # # ]: 0 : if (passive_p)
680 : 0 : argp_failure(state, 1, EINVAL, "-c option inconsistent with passive mode");
681 : 0 : concurrency = (unsigned) atoi(arg);
682 [ # # ]: 0 : if (concurrency < 1) concurrency = 1;
683 : : break;
684 : 6 : case 'C':
685 [ + + ]: 6 : if (arg)
686 : : {
687 : 4 : connection_pool = atoi(arg);
688 [ + - ]: 4 : if (connection_pool < 2)
689 : 0 : argp_failure(state, 1, EINVAL, "-C NUM minimum 2");
690 : : }
691 : : break;
692 : 4 : case 'I':
693 : : // NB: no problem with unconditional free here - an earlier failed regcomp would exit program
694 [ - + ]: 4 : if (passive_p)
695 : 0 : argp_failure(state, 1, EINVAL, "-I option inconsistent with passive mode");
696 : 4 : regfree (&file_include_regex);
697 : 4 : rc = regcomp (&file_include_regex, arg, REG_EXTENDED|REG_NOSUB);
698 [ + - ]: 4 : if (rc != 0)
699 : 0 : argp_failure(state, 1, EINVAL, "regular expression");
700 : : break;
701 : 6 : case 'X':
702 [ - + ]: 6 : if (passive_p)
703 : 0 : argp_failure(state, 1, EINVAL, "-X option inconsistent with passive mode");
704 : 6 : regfree (&file_exclude_regex);
705 : 6 : rc = regcomp (&file_exclude_regex, arg, REG_EXTENDED|REG_NOSUB);
706 [ + - ]: 6 : if (rc != 0)
707 : 0 : argp_failure(state, 1, EINVAL, "regular expression");
708 : : break;
709 : 4 : case 'r':
710 [ - + ]: 4 : if (passive_p)
711 : 0 : argp_failure(state, 1, EINVAL, "-r option inconsistent with passive mode");
712 : 4 : regex_groom = true;
713 : 4 : break;
714 : : case ARGP_KEY_FDCACHE_FDS:
715 : : // deprecated
716 : : break;
717 : 4 : case ARGP_KEY_FDCACHE_MBS:
718 : 4 : fdcache_mbs = atol (arg);
719 : 4 : break;
720 : 4 : case ARGP_KEY_FDCACHE_PREFETCH:
721 : 4 : fdcache_prefetch = atol (arg);
722 : 4 : break;
723 : 4 : case ARGP_KEY_FDCACHE_MINTMP:
724 : 4 : fdcache_mintmp = atol (arg);
725 [ + - ]: 4 : if( fdcache_mintmp > 100 || fdcache_mintmp < 0 )
726 : 0 : argp_failure(state, 1, EINVAL, "fdcache mintmp percent");
727 : : break;
728 : 4 : case ARGP_KEY_FORWARDED_TTL_LIMIT:
729 : 4 : forwarded_ttl_limit = (unsigned) atoi(arg);
730 : 4 : break;
731 : 110 : case ARGP_KEY_ARG:
732 [ + - ]: 110 : source_paths.insert(string(arg));
733 : 110 : break;
734 : : case ARGP_KEY_FDCACHE_PREFETCH_FDS:
735 : : // deprecated
736 : : break;
737 : : case ARGP_KEY_FDCACHE_PREFETCH_MBS:
738 : : // deprecated
739 : : break;
740 : 2 : case ARGP_KEY_PASSIVE:
741 : 2 : passive_p = true;
742 [ + - ]: 2 : if (source_paths.size() > 0
743 [ + - ]: 2 : || maxigroom
744 [ + - ]: 2 : || extra_ddl.size() > 0
745 [ + - + - ]: 4 : || traverse_logical)
746 : : // other conflicting options tricky to check
747 : 0 : argp_failure(state, 1, EINVAL, "inconsistent options with passive mode");
748 : : break;
749 : 0 : case ARGP_KEY_DISABLE_SOURCE_SCAN:
750 : 0 : scan_source_info = false;
751 : 0 : break;
752 : 2 : case ARGP_SCAN_CHECKPOINT:
753 : 2 : scan_checkpoint = atol (arg);
754 [ + - ]: 2 : if (scan_checkpoint < 0)
755 : 0 : argp_failure(state, 1, EINVAL, "scan checkpoint");
756 : : break;
757 : 0 : case ARGP_KEY_METADATA_MAXTIME:
758 : 0 : metadata_maxtime_s = (unsigned) atoi(arg);
759 : 0 : break;
760 : : #ifdef ENABLE_IMA_VERIFICATION
761 : : case ARGP_KEY_KOJI_SIGCACHE:
762 : : requires_koji_sigcache_mapping = true;
763 : : break;
764 : : #endif
765 : 0 : case ARGP_KEY_HTTP_ADDR:
766 [ # # ]: 0 : if (inet_pton(AF_INET, arg, &(((sockaddr_in*)&http_sockaddr)->sin_addr)) == 1)
767 : 0 : http_sockaddr.sin6_family = AF_INET;
768 : : else
769 [ # # ]: 0 : if (inet_pton(AF_INET6, arg, &http_sockaddr.sin6_addr) == 1)
770 : 0 : http_sockaddr.sin6_family = AF_INET6;
771 : : else
772 : 0 : argp_failure(state, 1, EINVAL, "listen-address");
773 : 0 : addr_info = arg;
774 : 0 : break;
775 : : // case 'h': argp_state_help (state, stderr, ARGP_HELP_LONG|ARGP_HELP_EXIT_OK);
776 : 4 : case 'h':
777 : 4 : cust_homepage_redirect = arg;
778 : 4 : break;
779 : 4 : case 'H':
780 : 4 : cust_homepage_file = arg;
781 : 4 : break;
782 : : default: return ARGP_ERR_UNKNOWN;
783 : : }
784 : :
785 : : return 0;
786 : : }
787 : :
788 : :
789 : : ////////////////////////////////////////////////////////////////////////
790 : :
791 : :
792 : : static void add_mhd_response_header (struct MHD_Response *r,
793 : : const char *h, const char *v);
794 : :
795 : : // represent errors that may get reported to an ostream and/or a libmicrohttpd connection
796 : :
797 : 8 : struct reportable_exception
798 : : {
799 : : int code;
800 : : string message;
801 : :
802 [ - - + - : 106 : reportable_exception(int c, const string& m): code(c), message(m) {}
- - + - +
- ]
803 [ - - - - : 603 : reportable_exception(const string& m): code(503), message(m) {}
- - - - -
- - - + -
- - - - +
- - - - -
- - - - -
- - - - -
- - - - -
- - - - -
+ - - - ]
804 : : reportable_exception(): code(503), message() {}
805 : :
806 : : void report(ostream& o) const; // defined under obatched() class below
807 : :
808 : 631 : MHD_RESULT mhd_send_response(MHD_Connection* c) const {
809 : 1262 : MHD_Response* r = MHD_create_response_from_buffer (message.size(),
810 : 631 : (void*) message.c_str(),
811 : : MHD_RESPMEM_MUST_COPY);
812 : 631 : add_mhd_response_header (r, "Content-Type", "text/plain");
813 : 631 : MHD_RESULT rc = MHD_queue_response (c, code, r);
814 : 631 : MHD_destroy_response (r);
815 : 631 : return rc;
816 : : }
817 : : };
818 : :
819 : :
820 : : struct sqlite_exception: public reportable_exception
821 : : {
822 : 0 : sqlite_exception(int rc, const string& msg):
823 [ # # # # : 0 : reportable_exception(string("sqlite3 error: ") + msg + ": " + string(sqlite3_errstr(rc) ?: "?")) {
# # # # ]
824 [ # # # # : 0 : inc_metric("error_count","sqlite3",sqlite3_errstr(rc));
# # # # #
# ]
825 : 0 : }
826 : : };
827 : :
828 : 4 : struct libc_exception: public reportable_exception
829 : : {
830 : 596 : libc_exception(int rc, const string& msg):
831 [ - + + - : 1788 : reportable_exception(string("libc error: ") + msg + ": " + string(strerror(rc) ?: "?")) {
+ - + - ]
832 [ + - + - : 1192 : inc_metric("error_count","libc",strerror(rc));
+ - + - ]
833 : 596 : }
834 : : };
835 : :
836 : :
837 : : struct archive_exception: public reportable_exception
838 : : {
839 : 0 : archive_exception(const string& msg):
840 [ # # ]: 0 : reportable_exception(string("libarchive error: ") + msg) {
841 [ # # # # : 0 : inc_metric("error_count","libarchive",msg);
# # ]
842 : 0 : }
843 : 0 : archive_exception(struct archive* a, const string& msg):
844 [ # # # # : 0 : reportable_exception(string("libarchive error: ") + msg + ": " + string(archive_error_string(a) ?: "?")) {
# # # # ]
845 [ # # # # : 0 : inc_metric("error_count","libarchive",msg + ": " + string(archive_error_string(a) ?: "?"));
# # # # #
# # # # #
# # ]
846 : 0 : }
847 : 0 : archive_exception(struct archive* a, const string& fname, const string& msg):
848 [ # # # # : 0 : reportable_exception(string("libarchive error: ") + fname + string(" ") + msg + ": " +
# # # # #
# ]
849 [ # # # # ]: 0 : string(archive_error_string(a) ?: "?")) {
850 [ # # # # : 0 : inc_metric("error_count","libarchive",msg + ": " + string(archive_error_string(a) ?: "?"));
# # # # #
# # # # #
# # ]
851 : 0 : }
852 : : };
853 : :
854 : :
855 : : struct elfutils_exception: public reportable_exception
856 : : {
857 : 0 : elfutils_exception(int rc, const string& msg):
858 [ # # # # : 0 : reportable_exception(string("elfutils error: ") + msg + ": " + string(elf_errmsg(rc) ?: "?")) {
# # # # ]
859 [ # # # # : 0 : inc_metric("error_count","elfutils",elf_errmsg(rc));
# # # # #
# ]
860 : 0 : }
861 : : };
862 : :
863 : :
864 : : ////////////////////////////////////////////////////////////////////////
865 : :
866 : : template <typename Payload>
867 : : class workq
868 : : {
869 : : unordered_set<Payload> q; // eliminate duplicates
870 : : mutex mtx;
871 : : condition_variable cv;
872 : : bool dead;
873 : : unsigned idlers; // number of threads busy with wait_idle / done_idle
874 : : unsigned fronters; // number of threads busy with wait_front / done_front
875 : :
876 : : public:
877 : 84 : workq() { dead = false; idlers = 0; fronters = 0; }
878 : 84 : ~workq() {}
879 : :
880 : 1314 : void push_back(const Payload& p)
881 : : {
882 : 1314 : unique_lock<mutex> lock(mtx);
883 [ + - ]: 1314 : q.insert (p);
884 [ + - + - : 2628 : set_metric("thread_work_pending","role","scan", q.size());
+ - + - ]
885 : 1314 : cv.notify_all();
886 : 1314 : }
887 : :
888 : : // kill this workqueue, wake up all idlers / scanners
889 : 84 : void nuke() {
890 : 84 : unique_lock<mutex> lock(mtx);
891 : : // optional: q.clear();
892 : 84 : dead = true;
893 : 84 : cv.notify_all();
894 : 84 : }
895 : :
896 : : // clear the workqueue, when scanning is interrupted with USR2
897 : 0 : void clear() {
898 : 0 : unique_lock<mutex> lock(mtx);
899 : 0 : q.clear();
900 [ # # # # : 0 : set_metric("thread_work_pending","role","scan", q.size());
# # # # ]
901 : : // NB: there may still be some live fronters
902 : 0 : cv.notify_all(); // maybe wake up waiting idlers
903 : 0 : }
904 : :
905 : : // block this scanner thread until there is work to do and no active idler
906 : 1601 : bool wait_front (Payload& p)
907 : : {
908 : 1601 : unique_lock<mutex> lock(mtx);
909 [ + + + + : 5601 : while (!dead && (q.size() == 0 || idlers > 0))
+ + ]
910 [ + - ]: 3999 : cv.wait(lock);
911 [ + + ]: 1602 : if (dead)
912 : : return false;
913 : : else
914 : : {
915 [ + - ]: 1314 : p = * q.begin();
916 : 1314 : q.erase (q.begin());
917 : 1314 : fronters ++; // prevent idlers from starting awhile, even if empty q
918 [ + - + - : 2628 : set_metric("thread_work_pending","role","scan", q.size());
+ - + - -
- ]
919 : : // NB: don't wake up idlers yet! The consumer is busy
920 : : // processing this element until it calls done_front().
921 : 1314 : return true;
922 : : }
923 : 1602 : }
924 : :
925 : : // notify waitq that scanner thread is done with that last item
926 : 1314 : void done_front ()
927 : : {
928 : 1314 : unique_lock<mutex> lock(mtx);
929 : 1314 : fronters --;
930 [ + + + + ]: 1314 : if (q.size() == 0 && fronters == 0)
931 : 100 : cv.notify_all(); // maybe wake up waiting idlers
932 : 1314 : }
933 : :
934 : : // block this idler thread until there is no work to do
935 : 600 : void wait_idle ()
936 : : {
937 : 600 : unique_lock<mutex> lock(mtx);
938 : 600 : cv.notify_all(); // maybe wake up waiting scanners
939 [ + + + + : 648 : while (!dead && ((q.size() != 0) || fronters > 0))
+ + ]
940 [ + - ]: 48 : cv.wait(lock);
941 [ + - ]: 600 : idlers ++;
942 : 600 : }
943 : :
944 : 518 : void done_idle ()
945 : : {
946 : 518 : unique_lock<mutex> lock(mtx);
947 : 518 : idlers --;
948 : 518 : cv.notify_all(); // maybe wake up waiting scanners, but probably not (shutting down)
949 : 518 : }
950 : : };
951 : :
952 : : typedef struct stat stat_t;
953 : : typedef pair<string,stat_t> scan_payload;
954 : : inline bool operator< (const scan_payload& a, const scan_payload& b)
955 : : {
956 : : return a.first < b.first; // don't bother compare the stat fields
957 : : }
958 : :
959 : : namespace std { // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=56480
960 : : template<> struct hash<::scan_payload>
961 : : {
962 : 5372 : std::size_t operator() (const ::scan_payload& p) const noexcept
963 : : {
964 [ + + + + ]: 5372 : return hash<string>()(p.first);
965 : : }
966 : : };
967 : : template<> struct equal_to<::scan_payload>
968 : : {
969 : 578 : std::size_t operator() (const ::scan_payload& a, const ::scan_payload& b) const noexcept
970 : : {
971 [ + - - - ]: 578 : return a.first == b.first;
972 : : }
973 : : };
974 : : }
975 : :
976 : : static workq<scan_payload> scanq; // just a single one
977 : : // producer & idler: thread_main_fts_source_paths()
978 : : // consumer: thread_main_scanner()
979 : : // idler: thread_main_groom()
980 : :
981 : :
982 : : ////////////////////////////////////////////////////////////////////////
983 : :
984 : : // Unique set is a thread-safe structure that lends 'ownership' of a value
985 : : // to a thread. Other threads requesting the same thing are made to wait.
986 : : // It's like a semaphore-on-demand.
987 : : template <typename T>
988 : : class unique_set
989 : : {
990 : : private:
991 : : set<T> values;
992 : : mutex mtx;
993 : : condition_variable cv;
994 : : public:
995 : 60 : unique_set() {}
996 : 60 : ~unique_set() {}
997 : :
998 : 2595 : void acquire(const T& value)
999 : : {
1000 : 2595 : unique_lock<mutex> lock(mtx);
1001 [ + + ]: 6679 : while (values.find(value) != values.end())
1002 [ + - ]: 4084 : cv.wait(lock);
1003 [ + - ]: 2595 : values.insert(value);
1004 : 2595 : }
1005 : :
1006 : 2595 : void release(const T& value)
1007 : : {
1008 : 2595 : unique_lock<mutex> lock(mtx);
1009 : : // assert (values.find(value) != values.end());
1010 : 2595 : values.erase(value);
1011 : 2595 : cv.notify_all();
1012 : 2595 : }
1013 : : };
1014 : :
1015 : :
1016 : : // This is the object that's instantiate to uniquely hold a value in a
1017 : : // RAII-pattern way.
1018 : : template <typename T>
1019 : : class unique_set_reserver
1020 : : {
1021 : : private:
1022 : : unique_set<T>& please_hold;
1023 : : T mine;
1024 : : public:
1025 : 2595 : unique_set_reserver(unique_set<T>& t, const T& value):
1026 [ + - ]: 2595 : please_hold(t), mine(value) { please_hold.acquire(mine); }
1027 : 2595 : ~unique_set_reserver() { please_hold.release(mine); }
1028 : : };
1029 : :
1030 : :
1031 : : ////////////////////////////////////////////////////////////////////////
1032 : :
1033 : : // periodic_barrier is a concurrency control object that lets N threads
1034 : : // periodically (based on counter value) agree to wait at a barrier,
1035 : : // let one of them carry out some work, then be set free
1036 : :
1037 : : class periodic_barrier
1038 : : {
1039 : : private:
1040 : : unsigned period; // number of count() reports to trigger barrier activation
1041 : : unsigned threads; // number of threads participating
1042 : : mutex mtx; // protects all the following fields
1043 : : unsigned counter; // count of count() reports in the current generation
1044 : : unsigned generation; // barrier activation generation
1045 : : unsigned waiting; // number of threads waiting for barrier
1046 : : bool dead; // bring out your
1047 : : condition_variable cv;
1048 : : public:
1049 : 72 : periodic_barrier(unsigned t, unsigned p):
1050 : 72 : period(p), threads(t), counter(0), generation(0), waiting(0), dead(false) { }
1051 : : virtual ~periodic_barrier() {}
1052 : :
1053 : : virtual void periodic_barrier_work() noexcept = 0;
1054 : 72 : void nuke() {
1055 : 72 : unique_lock<mutex> lock(mtx);
1056 : 72 : dead = true;
1057 : 72 : cv.notify_all();
1058 : 72 : }
1059 : :
1060 : 1602 : void count()
1061 : : {
1062 : 1602 : unique_lock<mutex> lock(mtx);
1063 : 1602 : unsigned prev_generation = this->generation;
1064 [ + + ]: 1602 : if (counter < period-1) // normal case: counter just freely running
1065 : : {
1066 : 1458 : counter ++;
1067 : 1458 : return;
1068 : : }
1069 [ + + ]: 144 : else if (counter == period-1) // we're the doer
1070 : : {
1071 : 36 : counter = period; // entering barrier holding phase
1072 : 36 : cv.notify_all();
1073 [ + + + - ]: 175 : while (waiting < threads-1 && !dead)
1074 [ + - ]: 103 : cv.wait(lock);
1075 : : // all other threads are now stuck in the barrier
1076 : 36 : this->periodic_barrier_work(); // NB: we're holding the mutex the whole time
1077 : : // reset for next barrier, releasing other waiters
1078 : 36 : counter = 0;
1079 : 36 : generation ++;
1080 : 36 : cv.notify_all();
1081 : 36 : return;
1082 : : }
1083 [ + - ]: 108 : else if (counter == period) // we're a waiter, in holding phase
1084 : : {
1085 : 108 : waiting ++;
1086 : 108 : cv.notify_all();
1087 [ + + + + : 388 : while (counter == period && generation == prev_generation && !dead)
+ - ]
1088 [ + - ]: 172 : cv.wait(lock);
1089 : 108 : waiting --;
1090 : 108 : return;
1091 : : }
1092 : 1602 : }
1093 : : };
1094 : :
1095 : :
1096 : :
1097 : : ////////////////////////////////////////////////////////////////////////
1098 : :
1099 : :
1100 : : // Print a standard timestamp.
1101 : : static ostream&
1102 : 51256 : timestamp (ostream &o)
1103 : : {
1104 : 51256 : char datebuf[80];
1105 : 51256 : char *now2 = NULL;
1106 : 51256 : time_t now_t = time(NULL);
1107 : 51257 : struct tm now;
1108 : 51257 : struct tm *nowp = gmtime_r (&now_t, &now);
1109 [ + - ]: 51258 : if (nowp)
1110 : : {
1111 : 51258 : (void) strftime (datebuf, sizeof (datebuf), "%c", nowp);
1112 : 51258 : now2 = datebuf;
1113 : : }
1114 : :
1115 : 51258 : return o << "[" << (now2 ? now2 : "") << "] "
1116 [ - + ]: 51258 : << "(" << getpid () << "/" << tid() << "): ";
1117 : : }
1118 : :
1119 : :
1120 : : // A little class that impersonates an ostream to the extent that it can
1121 : : // take << streaming operations. It batches up the bits into an internal
1122 : : // stringstream until it is destroyed; then flushes to the original ostream.
1123 : : // It adds a timestamp
1124 : : class obatched
1125 : : {
1126 : : private:
1127 : : ostream& o;
1128 : : stringstream stro;
1129 : : static mutex lock;
1130 : : public:
1131 : 51254 : obatched(ostream& oo, bool timestamp_p = true): o(oo)
1132 : : {
1133 [ + - ]: 51258 : if (timestamp_p)
1134 [ + - ]: 51258 : timestamp(stro);
1135 : 51249 : }
1136 : 51253 : ~obatched()
1137 : : {
1138 : 51253 : unique_lock<mutex> do_not_cross_the_streams(obatched::lock);
1139 : 51264 : o << stro.str();
1140 : 51264 : o.flush();
1141 : 51264 : }
1142 : : operator ostream& () { return stro; }
1143 [ - - + - : 41131 : template <typename T> ostream& operator << (const T& t) { stro << t; return stro; }
+ - + - +
- + - + -
- - - - -
- + - + -
+ - + - +
- + - + -
+ - + - +
- + - + -
- - + - +
- + - + -
+ - + - +
- + - - -
+ - + - +
- + - + -
+ - + - -
- + - - -
+ - - - +
- + - + -
- - - - -
- - - - -
- - + - -
- + - + -
- - + - +
- - - + -
- - + - -
- - - + -
+ - + - +
- + - + -
+ - - - ]
1144 : : };
1145 : : mutex obatched::lock; // just the one, since cout/cerr iostreams are not thread-safe
1146 : :
1147 : :
1148 : 695 : void reportable_exception::report(ostream& o) const {
1149 [ + - + - ]: 695 : obatched(o) << message << endl;
1150 : 695 : }
1151 : :
1152 : :
1153 : : ////////////////////////////////////////////////////////////////////////
1154 : :
1155 : :
1156 : : // RAII style sqlite prepared-statement holder that matches { } block lifetime
1157 : :
1158 : : struct sqlite_ps
1159 : : {
1160 : : private:
1161 : : sqlite3* db;
1162 : : const string nickname;
1163 : : const string sql;
1164 : : sqlite3_stmt *pp;
1165 : : // for step_timeout()/callback
1166 : : struct timespec ts_start;
1167 : : double ts_timeout;
1168 : :
1169 : : sqlite_ps(const sqlite_ps&); // make uncopyable
1170 : : sqlite_ps& operator=(const sqlite_ps &); // make unassignable
1171 : :
1172 : : public:
1173 [ + - ]: 9836 : sqlite_ps (sqlite3* d, const string& n, const string& s): db(d), nickname(n), sql(s) {
1174 : : // tmp_ms_metric tick("sqlite3","prep",nickname);
1175 [ + + ]: 9837 : if (verbose > 4)
1176 [ + - + - : 174 : obatched(clog) << nickname << " prep " << sql << endl;
+ - + - +
- ]
1177 [ + - ]: 9837 : int rc = sqlite3_prepare_v2 (db, sql.c_str(), -1 /* to \0 */, & this->pp, NULL);
1178 [ - + ]: 9838 : if (rc != SQLITE_OK)
1179 [ # # # # ]: 0 : throw sqlite_exception(rc, "prepare " + sql);
1180 : 19676 : this->reset_timeout(0.0);
1181 : 0 : }
1182 : :
1183 : 198254 : sqlite_ps& reset()
1184 : : {
1185 [ + - + - ]: 396478 : tmp_ms_metric tick("sqlite3","reset",nickname);
1186 [ + - ]: 198225 : sqlite3_reset(this->pp);
1187 : 198300 : return *this;
1188 : 198228 : }
1189 : :
1190 : 226753 : sqlite_ps& bind(int parameter, const string& str)
1191 : : {
1192 [ + + ]: 226753 : if (verbose > 4)
1193 [ + - + - : 196 : obatched(clog) << nickname << " bind " << parameter << "=" << str << endl;
+ - + - +
- + - ]
1194 : 226753 : int rc = sqlite3_bind_text (this->pp, parameter, str.c_str(), -1, SQLITE_TRANSIENT);
1195 [ - + ]: 226597 : if (rc != SQLITE_OK)
1196 [ # # # # ]: 0 : throw sqlite_exception(rc, "sqlite3 bind");
1197 : 226597 : return *this;
1198 : : }
1199 : :
1200 : 61788 : sqlite_ps& bind(int parameter, int64_t value)
1201 : : {
1202 [ + + ]: 61788 : if (verbose > 4)
1203 [ + - + - : 64 : obatched(clog) << nickname << " bind " << parameter << "=" << value << endl;
+ - + - +
- + - ]
1204 : 61788 : int rc = sqlite3_bind_int64 (this->pp, parameter, value);
1205 [ - + ]: 61799 : if (rc != SQLITE_OK)
1206 [ # # # # ]: 0 : throw sqlite_exception(rc, "sqlite3 bind");
1207 : 61799 : return *this;
1208 : : }
1209 : :
1210 : : sqlite_ps& bind(int parameter)
1211 : : {
1212 : : if (verbose > 4)
1213 : : obatched(clog) << nickname << " bind " << parameter << "=" << "NULL" << endl;
1214 : : int rc = sqlite3_bind_null (this->pp, parameter);
1215 : : if (rc != SQLITE_OK)
1216 : : throw sqlite_exception(rc, "sqlite3 bind");
1217 : : return *this;
1218 : : }
1219 : :
1220 : :
1221 : 121703 : void step_ok_done() {
1222 [ + - + - ]: 243293 : tmp_ms_metric tick("sqlite3","step_done",nickname);
1223 [ + - ]: 121592 : int rc = sqlite3_step (this->pp);
1224 [ + + ]: 121915 : if (verbose > 4)
1225 [ + - + - : 128 : obatched(clog) << nickname << " step-ok-done(" << sqlite3_errstr(rc) << ") " << sql << endl;
+ - + - +
- + - + -
+ - ]
1226 [ + + - + ]: 121915 : if (rc != SQLITE_OK && rc != SQLITE_DONE && rc != SQLITE_ROW)
1227 [ # # # # ]: 0 : throw sqlite_exception(rc, "sqlite3 step");
1228 [ + - ]: 121915 : (void) sqlite3_reset (this->pp);
1229 : 121913 : }
1230 : :
1231 : :
1232 : 40869 : int step() {
1233 [ + - + - ]: 81742 : tmp_ms_metric tick("sqlite3","step",nickname);
1234 [ + - ]: 40873 : int rc = sqlite3_step (this->pp);
1235 [ + + ]: 40874 : if (verbose > 4)
1236 [ + - + - : 62 : obatched(clog) << nickname << " step(" << sqlite3_errstr(rc) << ") " << sql << endl;
+ - + - +
- + - + -
+ - ]
1237 : 40874 : return rc;
1238 : 40874 : }
1239 : :
1240 : :
1241 : 9868 : void reset_timeout(double s) // set starting point for maximum elapsed time in step_timeouts()
1242 : : {
1243 : 9838 : clock_gettime (CLOCK_MONOTONIC, &this->ts_start);
1244 : 9838 : this->ts_timeout = s;
1245 : : }
1246 : :
1247 : :
1248 : 0 : static int sqlite3_progress_handler_cb (void *param)
1249 : : {
1250 : 0 : sqlite_ps *pp = (sqlite_ps*) param;
1251 : 0 : struct timespec ts_end;
1252 : 0 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
1253 : 0 : double deltas = (ts_end.tv_sec - pp->ts_start.tv_sec) + (ts_end.tv_nsec - pp->ts_start.tv_nsec)/1.e9;
1254 [ # # # # ]: 0 : return (interrupted || (deltas > pp->ts_timeout)); // non-zero => interrupt sqlite operation in progress
1255 : : }
1256 : :
1257 : :
1258 : 48 : int step_timeout() {
1259 : : // Do the same thing as step(), except wrapping it into a timeout
1260 : : // relative to the last reset_timeout() invocation.
1261 : : //
1262 : : // Do this by attaching a progress_handler to the database
1263 : : // connection, for the duration of this operation. It should be a
1264 : : // private connection to the calling thread, so other operations
1265 : : // cannot begin concurrently.
1266 : :
1267 : 48 : sqlite3_progress_handler(this->db, 10000 /* bytecode insns */,
1268 : : & sqlite3_progress_handler_cb, (void*) this);
1269 : 48 : int rc = this->step();
1270 : 48 : sqlite3_progress_handler(this->db, 0, 0, 0); // disable
1271 : 48 : struct timespec ts_end;
1272 : 48 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
1273 : 48 : double deltas = (ts_end.tv_sec - this->ts_start.tv_sec) + (ts_end.tv_nsec - this->ts_start.tv_nsec)/1.e9;
1274 [ - + ]: 48 : if (verbose > 3)
1275 [ # # # # : 0 : obatched(clog) << this->nickname << " progress-delta-final " << deltas << endl;
# # # # ]
1276 : 48 : return rc;
1277 : : }
1278 : :
1279 : :
1280 : 9824 : ~sqlite_ps () { sqlite3_finalize (this->pp); }
1281 [ + - + - : 7088 : operator sqlite3_stmt* () { return this->pp; }
+ - + - +
- + - + -
+ - + - +
- ]
1282 : : };
1283 : :
1284 : :
1285 : : ////////////////////////////////////////////////////////////////////////
1286 : :
1287 : :
1288 : : struct sqlite_checkpoint_pb: public periodic_barrier
1289 : : {
1290 : : // NB: don't use sqlite_ps since it can throw exceptions during ctor etc.
1291 : 72 : sqlite_checkpoint_pb(unsigned t, unsigned p):
1292 : 144 : periodic_barrier(t, p) { }
1293 : :
1294 : 36 : void periodic_barrier_work() noexcept
1295 : : {
1296 : 36 : (void) sqlite3_exec (db, "pragma wal_checkpoint(truncate);", NULL, NULL, NULL);
1297 : 36 : }
1298 : : };
1299 : :
1300 : : static periodic_barrier* scan_barrier = 0; // initialized in main()
1301 : :
1302 : :
1303 : : ////////////////////////////////////////////////////////////////////////
1304 : :
1305 : : // RAII style templated autocloser
1306 : :
1307 : : template <class Payload, class Ignore>
1308 : : struct defer_dtor
1309 : : {
1310 : : public:
1311 : : typedef Ignore (*dtor_fn) (Payload);
1312 : :
1313 : : private:
1314 : : Payload p;
1315 : : dtor_fn fn;
1316 : :
1317 : : public:
1318 : 8022 : defer_dtor(Payload _p, dtor_fn _fn): p(_p), fn(_fn) {}
1319 : 414 : ~defer_dtor() { (void) (*fn)(p); }
1320 : :
1321 : : private:
1322 : : defer_dtor(const defer_dtor<Payload,Ignore>&); // make uncopyable
1323 : : defer_dtor& operator=(const defer_dtor<Payload,Ignore> &); // make unassignable
1324 : : };
1325 : :
1326 : :
1327 : :
1328 : : ////////////////////////////////////////////////////////////////////////
1329 : :
1330 : :
1331 : : static string
1332 : 6766 : header_censor(const string& str)
1333 : : {
1334 : 6766 : string y;
1335 [ + + ]: 83809 : for (auto&& x : str)
1336 : : {
1337 [ + + ]: 77043 : if (isalnum(x) || x == '/' || x == '.' || x == ',' || x == '_' || x == ':')
1338 [ + - ]: 154080 : y += x;
1339 : : }
1340 : 6766 : return y;
1341 : 0 : }
1342 : :
1343 : :
1344 : : static string
1345 : 3383 : conninfo (struct MHD_Connection * conn)
1346 : : {
1347 : 3383 : char hostname[256]; // RFC1035
1348 : 3383 : char servname[256];
1349 : 3383 : int sts = -1;
1350 : :
1351 [ - + ]: 3383 : if (conn == 0)
1352 : 0 : return "internal";
1353 : :
1354 : : /* Look up client address data. */
1355 : 3383 : const union MHD_ConnectionInfo *u = MHD_get_connection_info (conn,
1356 : : MHD_CONNECTION_INFO_CLIENT_ADDRESS);
1357 [ + - ]: 3383 : struct sockaddr *so = u ? u->client_addr : 0;
1358 : :
1359 [ + - - + ]: 3383 : if (so && so->sa_family == AF_INET) {
1360 : 0 : sts = getnameinfo (so, sizeof (struct sockaddr_in),
1361 : : hostname, sizeof (hostname),
1362 : : servname, sizeof (servname),
1363 : : NI_NUMERICHOST | NI_NUMERICSERV);
1364 [ + - ]: 3383 : } else if (so && so->sa_family == AF_INET6) {
1365 : 3383 : struct sockaddr_in6* addr6 = (struct sockaddr_in6*) so;
1366 [ + - + - : 3383 : if (IN6_IS_ADDR_V4MAPPED(&addr6->sin6_addr)) {
+ + ]
1367 : 1621 : struct sockaddr_in addr4;
1368 : 1621 : memset (&addr4, 0, sizeof(addr4));
1369 : 1621 : addr4.sin_family = AF_INET;
1370 : 1621 : addr4.sin_port = addr6->sin6_port;
1371 : 1621 : memcpy (&addr4.sin_addr.s_addr, addr6->sin6_addr.s6_addr+12, sizeof(addr4.sin_addr.s_addr));
1372 : 1621 : sts = getnameinfo ((struct sockaddr*) &addr4, sizeof (addr4),
1373 : : hostname, sizeof (hostname),
1374 : : servname, sizeof (servname),
1375 : : NI_NUMERICHOST | NI_NUMERICSERV);
1376 : : } else {
1377 : 1762 : sts = getnameinfo (so, sizeof (struct sockaddr_in6),
1378 : : hostname, sizeof (hostname),
1379 : : servname, sizeof (servname),
1380 : : NI_NUMERICHOST | NI_NUMERICSERV);
1381 : : }
1382 : : }
1383 : :
1384 [ - + ]: 3383 : if (sts != 0) {
1385 : 0 : hostname[0] = servname[0] = '\0';
1386 : : }
1387 : :
1388 : : // extract headers relevant to administration
1389 [ + + ]: 3383 : const char* user_agent = MHD_lookup_connection_value (conn, MHD_HEADER_KIND, "User-Agent") ?: "";
1390 [ + + ]: 3383 : const char* x_forwarded_for = MHD_lookup_connection_value (conn, MHD_HEADER_KIND, "X-Forwarded-For") ?: "";
1391 : : // NB: these are untrustworthy, beware if machine-processing log files
1392 : :
1393 [ + - + - : 10149 : return string(hostname) + string(":") + string(servname) +
+ - + - +
- ]
1394 [ + - + - : 13532 : string(" UA:") + header_censor(string(user_agent)) +
+ - + - +
- ]
1395 [ + - + - : 10149 : string(" XFF:") + header_censor(string(x_forwarded_for));
+ - + - ]
1396 : : }
1397 : :
1398 : :
1399 : :
1400 : : ////////////////////////////////////////////////////////////////////////
1401 : :
1402 : : /* Wrapper for MHD_add_response_header that logs an error if we
1403 : : couldn't add the specified header. */
1404 : : static void
1405 : 13309 : add_mhd_response_header (struct MHD_Response *r,
1406 : : const char *h, const char *v)
1407 : : {
1408 [ - + ]: 13309 : if (MHD_add_response_header (r, h, v) == MHD_NO)
1409 [ # # # # : 0 : obatched(clog) << "Error: couldn't add '" << h << "' header" << endl;
# # ]
1410 : 13309 : }
1411 : :
1412 : : static void
1413 : 2152 : add_mhd_last_modified (struct MHD_Response *resp, time_t mtime)
1414 : : {
1415 : 2152 : struct tm now;
1416 : 2152 : struct tm *nowp = gmtime_r (&mtime, &now);
1417 [ + - ]: 2152 : if (nowp != NULL)
1418 : : {
1419 : 2152 : char datebuf[80];
1420 : 2152 : size_t rc = strftime (datebuf, sizeof (datebuf), "%a, %d %b %Y %T GMT",
1421 : : nowp);
1422 [ + - ]: 2152 : if (rc > 0 && rc < sizeof (datebuf))
1423 : 2152 : add_mhd_response_header (resp, "Last-Modified", datebuf);
1424 : : }
1425 : :
1426 : 2152 : add_mhd_response_header (resp, "Cache-Control", "public");
1427 : 2152 : }
1428 : :
1429 : : // quote all questionable characters of str for safe passage through a sh -c expansion.
1430 : : static string
1431 : 568 : shell_escape(const string& str)
1432 : : {
1433 : 568 : string y;
1434 [ + + ]: 72928 : for (auto&& x : str)
1435 : : {
1436 [ + + + + ]: 72360 : if (! isalnum(x) && x != '/')
1437 [ + - ]: 8966 : y += "\\";
1438 [ + - ]: 144720 : y += x;
1439 : : }
1440 : 568 : return y;
1441 : 0 : }
1442 : :
1443 : :
1444 : : // PR25548: Perform POSIX / RFC3986 style path canonicalization on the input string.
1445 : : //
1446 : : // Namely:
1447 : : // // -> /
1448 : : // /foo/../ -> /
1449 : : // /./ -> /
1450 : : //
1451 : : // This mapping is done on dwarf-side source path names, which may
1452 : : // include these constructs, so we can deal with debuginfod clients
1453 : : // that accidentally canonicalize the paths.
1454 : : //
1455 : : // realpath(3) is close but not quite right, because it also resolves
1456 : : // symbolic links. Symlinks at the debuginfod server have nothing to
1457 : : // do with the build-time symlinks, thus they must not be considered.
1458 : : //
1459 : : // see also curl Curl_dedotdotify() aka RFC3986, which we mostly follow here
1460 : : // see also libc __realpath()
1461 : : // see also llvm llvm::sys::path::remove_dots()
1462 : : static string
1463 : 16942 : canon_pathname (const string& input)
1464 : : {
1465 : 16942 : string i = input; // 5.2.4 (1)
1466 : 16942 : string o;
1467 : :
1468 : 16942 : while (i.size() != 0)
1469 : : {
1470 : : // 5.2.4 (2) A
1471 [ + - - + ]: 145114 : if (i.substr(0,3) == "../")
1472 [ # # ]: 0 : i = i.substr(3);
1473 [ + - - + ]: 145113 : else if(i.substr(0,2) == "./")
1474 [ # # ]: 0 : i = i.substr(2);
1475 : :
1476 : : // 5.2.4 (2) B
1477 [ + - + + ]: 145114 : else if (i.substr(0,3) == "/./")
1478 [ + - ]: 1872 : i = i.substr(2);
1479 [ - + ]: 143242 : else if (i == "/.")
1480 [ # # ]: 0 : i = ""; // no need to handle "/." complete-path-segment case; we're dealing with file names
1481 : :
1482 : : // 5.2.4 (2) C
1483 [ + - + + ]: 143242 : else if (i.substr(0,4) == "/../") {
1484 [ + - ]: 2518 : i = i.substr(3);
1485 : 2518 : string::size_type sl = o.rfind("/");
1486 [ + - ]: 2518 : if (sl != string::npos)
1487 [ + - ]: 2518 : o = o.substr(0, sl);
1488 : : else
1489 [ # # ]: 0 : o = "";
1490 [ - + ]: 140723 : } else if (i == "/..")
1491 [ # # ]: 0 : i = ""; // no need to handle "/.." complete-path-segment case; we're dealing with file names
1492 : :
1493 : : // 5.2.4 (2) D
1494 : : // no need to handle these cases; we're dealing with file names
1495 [ - + ]: 140723 : else if (i == ".")
1496 [ # # ]: 0 : i = "";
1497 [ - + ]: 140723 : else if (i == "..")
1498 [ # # ]: 0 : i = "";
1499 : :
1500 : : // POSIX special: map // to /
1501 [ + - + + ]: 140723 : else if (i.substr(0,2) == "//")
1502 [ + - ]: 128 : i = i.substr(1);
1503 : :
1504 : : // 5.2.4 (2) E
1505 : : else {
1506 [ - + ]: 140596 : string::size_type next_slash = i.find("/", (i[0]=='/' ? 1 : 0)); // skip first slash
1507 [ + - ]: 281192 : o += i.substr(0, next_slash);
1508 [ + + ]: 140596 : if (next_slash == string::npos)
1509 [ + - + + ]: 178998 : i = "";
1510 : : else
1511 [ + - ]: 123654 : i = i.substr(next_slash);
1512 : : }
1513 : : }
1514 : :
1515 : 16942 : return o;
1516 : 16942 : }
1517 : :
1518 : :
1519 : : // Estimate available free space for a given filesystem via statfs(2).
1520 : : // Return true if the free fraction is known to be smaller than the
1521 : : // given minimum percentage. Also update a related metric.
1522 : 4210 : bool statfs_free_enough_p(const string& path, const string& label, long minfree = 0)
1523 : : {
1524 : 4210 : struct statfs sfs;
1525 : 4210 : int rc = statfs(path.c_str(), &sfs);
1526 [ + + ]: 4209 : if (rc == 0)
1527 : : {
1528 : 4135 : double s = (double) sfs.f_bavail / (double) sfs.f_blocks;
1529 [ + - + - ]: 8271 : set_metric("filesys_free_ratio","purpose",label, s);
1530 : 4136 : return ((s * 100.0) < minfree);
1531 : : }
1532 : : return false;
1533 : : }
1534 : :
1535 : :
1536 : :
1537 : : // A map-like class that owns a cache of file descriptors (indexed by
1538 : : // file / content names).
1539 : : //
1540 : : // If only it could use fd's instead of file names ... but we can't
1541 : : // dup(2) to create independent descriptors for the same unlinked
1542 : : // files, so would have to use some goofy linux /proc/self/fd/%d
1543 : : // hack such as the following
1544 : :
1545 : : #if 0
1546 : : int superdup(int fd)
1547 : : {
1548 : : #ifdef __linux__
1549 : : char *fdpath = NULL;
1550 : : int rc = asprintf(& fdpath, "/proc/self/fd/%d", fd);
1551 : : int newfd;
1552 : : if (rc >= 0)
1553 : : newfd = open(fdpath, O_RDONLY);
1554 : : else
1555 : : newfd = -1;
1556 : : free (fdpath);
1557 : : return newfd;
1558 : : #else
1559 : : return -1;
1560 : : #endif
1561 : : }
1562 : : #endif
1563 : :
1564 : : class libarchive_fdcache
1565 : : {
1566 : : private:
1567 : : mutex fdcache_lock;
1568 : :
1569 : : typedef pair<string,string> key; // archive, entry
1570 : 1504 : struct fdcache_entry
1571 : : {
1572 : : string fd; // file name (probably in $TMPDIR), not an actual open fd (EMFILE)
1573 : : double fd_size_mb; // slightly rounded up megabytes
1574 : : time_t freshness; // when was this entry created or requested last
1575 : : unsigned request_count; // how many requests were made; or 0=prefetch only
1576 : : double latency; // how many seconds it took to extract the file
1577 : : };
1578 : :
1579 : : map<key,fdcache_entry> entries; // optimized for lookup
1580 : : time_t last_cleaning;
1581 : : long max_mbs;
1582 : :
1583 : : public:
1584 : 3006 : void set_metrics()
1585 : : {
1586 : 3006 : double fdcache_mb = 0.0;
1587 : 3006 : double prefetch_mb = 0.0;
1588 : 3006 : unsigned fdcache_count = 0;
1589 : 3006 : unsigned prefetch_count = 0;
1590 [ + + ]: 6152 : for (auto &i : entries) {
1591 [ + + ]: 3146 : if (i.second.request_count) {
1592 : 2550 : fdcache_mb += i.second.fd_size_mb;
1593 : 2550 : fdcache_count ++;
1594 : : } else {
1595 : 596 : prefetch_mb += i.second.fd_size_mb;
1596 : 596 : prefetch_count ++;
1597 : : }
1598 : : }
1599 [ + - ]: 3006 : set_metric("fdcache_bytes", fdcache_mb*1024.0*1024.0);
1600 [ + - ]: 3006 : set_metric("fdcache_count", fdcache_count);
1601 [ + - ]: 3006 : set_metric("fdcache_prefetch_bytes", prefetch_mb*1024.0*1024.0);
1602 [ + - ]: 3006 : set_metric("fdcache_prefetch_count", prefetch_count);
1603 : 3006 : }
1604 : :
1605 : 1504 : void intern(const string& a, const string& b, string fd, off_t sz,
1606 : : bool requested_p, double lat)
1607 : : {
1608 : 1504 : {
1609 : 1504 : unique_lock<mutex> lock(fdcache_lock);
1610 : 1504 : time_t now = time(NULL);
1611 : : // there is a chance it's already in here, just wasn't found last time
1612 : : // if so, there's nothing to do but count our luck
1613 [ + - ]: 1504 : auto i = entries.find(make_pair(a,b));
1614 [ - + ]: 1504 : if (i != entries.end())
1615 : : {
1616 [ # # # # : 0 : inc_metric("fdcache_op_count","op","redundant_intern");
# # # # ]
1617 [ # # ]: 0 : if (requested_p) i->second.request_count ++; // repeat prefetch doesn't count
1618 : 0 : i->second.freshness = now;
1619 : : // We need to nuke the temp file, since interning passes
1620 : : // responsibility over the path to this structure. It is
1621 : : // possible that the caller still has an fd open, but that's
1622 : : // OK.
1623 : 0 : unlink (fd.c_str());
1624 : 0 : return;
1625 : : }
1626 : 1504 : double mb = (sz+65535)/1048576.0; // round up to 64K block
1627 : 1504 : fdcache_entry n = { .fd=fd, .fd_size_mb=mb,
1628 : 1504 : .freshness=now, .request_count = requested_p?1U:0U,
1629 [ + - + + ]: 1504 : .latency=lat};
1630 [ + - + - : 1504 : entries.insert(make_pair(make_pair(a,b),n));
+ - ]
1631 : :
1632 [ + + ]: 1504 : if (requested_p)
1633 [ + - + - : 1816 : inc_metric("fdcache_op_count","op","enqueue");
+ - + - ]
1634 : : else
1635 [ + - + - : 1192 : inc_metric("fdcache_op_count","op","prefetch_enqueue");
+ - + - ]
1636 : :
1637 [ + + ]: 1504 : if (verbose > 3)
1638 [ + - + - ]: 4164 : obatched(clog) << "fdcache interned a=" << a << " b=" << b
1639 [ + - + - : 1388 : << " fd=" << fd << " mb=" << mb << " front=" << requested_p
+ - + - +
- + - + -
+ - ]
1640 [ + - + - : 1388 : << " latency=" << lat << endl;
+ - ]
1641 : :
1642 [ + - ]: 1504 : set_metrics();
1643 : 1504 : }
1644 : :
1645 : : // NB: we age the cache at lookup time too
1646 [ + - + + ]: 1504 : if (statfs_free_enough_p(tmpdir, "tmpdir", fdcache_mintmp))
1647 : : {
1648 [ + - + - : 2836 : inc_metric("fdcache_op_count","op","emerg-flush");
+ - ]
1649 [ + - ]: 2836 : obatched(clog) << "fdcache emergency flush for filling tmpdir" << endl;
1650 : 1418 : this->limit(0); // emergency flush
1651 : : }
1652 : : else // age cache normally
1653 : 86 : this->limit(max_mbs);
1654 : : }
1655 : :
1656 : 1012 : int lookup(const string& a, const string& b)
1657 : : {
1658 : 1012 : int fd = -1;
1659 : 1012 : {
1660 : 1012 : unique_lock<mutex> lock(fdcache_lock);
1661 [ + - ]: 1012 : auto i = entries.find(make_pair(a,b));
1662 [ + + ]: 1012 : if (i != entries.end())
1663 : : {
1664 [ + + ]: 100 : if (i->second.request_count == 0) // was a prefetch!
1665 : : {
1666 [ + - + - ]: 2 : inc_metric("fdcache_prefetch_saved_milliseconds_count");
1667 [ + - + - ]: 4 : add_metric("fdcache_prefetch_saved_milliseconds_sum", i->second.latency*1000.);
1668 : : }
1669 : 100 : i->second.request_count ++;
1670 : 100 : i->second.freshness = time(NULL);
1671 : : // brag about our success
1672 [ + - + - : 200 : inc_metric("fdcache_op_count","op","prefetch_access"); // backward compat
+ - + - ]
1673 [ + - + - ]: 100 : inc_metric("fdcache_saved_milliseconds_count");
1674 [ + - + - ]: 100 : add_metric("fdcache_saved_milliseconds_sum", i->second.latency*1000.);
1675 [ + - ]: 1012 : fd = open(i->second.fd.c_str(), O_RDONLY);
1676 : : }
1677 : 0 : }
1678 : :
1679 [ + + ]: 1012 : if (fd >= 0)
1680 [ + - + - : 200 : inc_metric("fdcache_op_count","op","lookup_hit");
+ - ]
1681 : : else
1682 [ + - + - : 1824 : inc_metric("fdcache_op_count","op","lookup_miss");
+ - ]
1683 : :
1684 : : // NB: no need to age the cache after just a lookup
1685 : :
1686 : 1012 : return fd;
1687 : : }
1688 : :
1689 : 1296 : int probe(const string& a, const string& b) // just a cache residency check - don't modify state, don't open
1690 : : {
1691 : 1296 : unique_lock<mutex> lock(fdcache_lock);
1692 [ + - ]: 1296 : auto i = entries.find(make_pair(a,b));
1693 [ - + ]: 1296 : if (i != entries.end()) {
1694 [ # # # # : 0 : inc_metric("fdcache_op_count","op","probe_hit");
# # # # ]
1695 : 0 : return true;
1696 : : } else {
1697 [ + - + - : 2592 : inc_metric("fdcache_op_count","op","probe_miss");
+ - + - ]
1698 : 1296 : return false;
1699 : : }
1700 : 1296 : }
1701 : :
1702 : 0 : void clear(const string& a, const string& b)
1703 : : {
1704 : 0 : unique_lock<mutex> lock(fdcache_lock);
1705 [ # # ]: 0 : auto i = entries.find(make_pair(a,b));
1706 [ # # ]: 0 : if (i != entries.end()) {
1707 [ # # # # : 0 : inc_metric("fdcache_op_count","op",
# # # # #
# ]
1708 [ # # ]: 0 : i->second.request_count > 0 ? "clear" : "prefetch_clear");
1709 : 0 : unlink (i->second.fd.c_str());
1710 : 0 : entries.erase(i);
1711 [ # # ]: 0 : set_metrics();
1712 : 0 : return;
1713 : : }
1714 : 0 : }
1715 : :
1716 : 1672 : void limit(long maxmbs, bool metrics_p = true)
1717 : : {
1718 : 1672 : time_t now = time(NULL);
1719 : :
1720 : : // avoid overly frequent limit operations
1721 [ + + + + ]: 1672 : if (maxmbs > 0 && (now - this->last_cleaning) < 10) // probably not worth parametrizing
1722 : 86 : return;
1723 : 1586 : this->last_cleaning = now;
1724 : :
1725 [ + + + + ]: 1586 : if (verbose > 3 && (this->max_mbs != maxmbs))
1726 [ + - + - ]: 192 : obatched(clog) << "fdcache limited to maxmbs=" << maxmbs << endl;
1727 : :
1728 : 1586 : unique_lock<mutex> lock(fdcache_lock);
1729 : :
1730 : 1586 : this->max_mbs = maxmbs;
1731 : 1586 : double total_mb = 0.0;
1732 : :
1733 : 1586 : map<double, pair<string,string>> sorted_entries;
1734 [ + + ]: 3090 : for (auto &i: entries)
1735 : : {
1736 : 1504 : total_mb += i.second.fd_size_mb;
1737 : :
1738 : : // need a scalar quantity that combines these inputs in a sensible way:
1739 : : //
1740 : : // 1) freshness of this entry (last time it was accessed)
1741 : : // 2) size of this entry
1742 : : // 3) number of times it has been accessed (or if just prefetched with 0 accesses)
1743 : : // 4) latency it required to extract
1744 : : //
1745 : : // The lower the "score", the earlier garbage collection will
1746 : : // nuke it, so to prioritize entries for preservation, the
1747 : : // score should be higher, and vice versa.
1748 : 1504 : time_t factor_1_freshness = (now - i.second.freshness); // seconds
1749 : 1504 : double factor_2_size = i.second.fd_size_mb; // megabytes
1750 : 1504 : unsigned factor_3_accesscount = i.second.request_count; // units
1751 : 1504 : double factor_4_latency = i.second.latency; // seconds
1752 : :
1753 : : #if 0
1754 : : double score = - factor_1_freshness; // simple LRU
1755 : : #endif
1756 : :
1757 [ + + ]: 1504 : double score = 0.
1758 : 1504 : - log1p(factor_1_freshness) // penalize old file
1759 : 1504 : - log1p(factor_2_size) // penalize large file
1760 : 1504 : + factor_4_latency * factor_3_accesscount; // reward slow + repeatedly read files
1761 : :
1762 [ + + ]: 1504 : if (verbose > 4)
1763 [ + - ]: 8 : obatched(clog) << "fdcache scored score=" << score
1764 [ + - + - ]: 8 : << " a=" << i.first.first << " b=" << i.first.second
1765 [ + - + - : 12 : << " f1=" << factor_1_freshness << " f2=" << factor_2_size
+ - + - +
- + - +
- ]
1766 [ + - + - : 4 : << " f3=" << factor_3_accesscount << " f4=" << factor_4_latency
+ - + - +
- ]
1767 : 4 : << endl;
1768 : :
1769 [ + - + - ]: 4512 : sorted_entries.insert(make_pair(score, i.first));
1770 : : }
1771 : :
1772 : 1586 : unsigned cleaned = 0;
1773 : 1586 : unsigned entries_original = entries.size();
1774 : 1586 : double cleaned_score_min = DBL_MAX;
1775 : 1586 : double cleaned_score_max = DBL_MIN;
1776 : :
1777 : : // drop as many entries[] as needed to bring total mb down to the threshold
1778 [ + + ]: 3090 : for (auto &i: sorted_entries) // in increasing score order!
1779 : : {
1780 [ - + ]: 1504 : if (this->max_mbs > 0 // if this is not a "clear entire table"
1781 [ # # ]: 0 : && total_mb < this->max_mbs) // we've cleared enough to meet threshold
1782 : : break; // stop clearing
1783 : :
1784 [ - + ]: 1504 : auto j = entries.find(i.second);
1785 [ - + ]: 1504 : if (j == entries.end())
1786 : 0 : continue; // should not happen
1787 : :
1788 [ + + ]: 1504 : if (cleaned == 0)
1789 : 1422 : cleaned_score_min = i.first;
1790 : 1504 : cleaned++;
1791 : 1504 : cleaned_score_max = i.first;
1792 : :
1793 [ + + ]: 1504 : if (verbose > 3)
1794 [ + - + - ]: 4164 : obatched(clog) << "fdcache evicted score=" << i.first
1795 [ + - + - ]: 2776 : << " a=" << i.second.first << " b=" << i.second.second
1796 [ + - + - : 4164 : << " fd=" << j->second.fd << " mb=" << j->second.fd_size_mb
+ - + - +
- + - ]
1797 [ + - + - : 1388 : << " rq=" << j->second.request_count << " lat=" << j->second.latency
+ - + - ]
1798 [ + - + - : 1388 : << " fr=" << (now - j->second.freshness)
+ - ]
1799 : 1388 : << endl;
1800 [ + + ]: 1504 : if (metrics_p)
1801 [ + - + - : 2836 : inc_metric("fdcache_op_count","op","evict");
+ - + - ]
1802 : :
1803 : 1504 : total_mb -= j->second.fd_size_mb;
1804 : 1504 : unlink (j->second.fd.c_str());
1805 : 1504 : entries.erase(j);
1806 : : }
1807 : :
1808 [ + + ]: 1586 : if (metrics_p)
1809 [ + - + - : 3004 : inc_metric("fdcache_op_count","op","evict_cycle");
+ - + - ]
1810 : :
1811 [ + - + + ]: 1586 : if (verbose > 1 && cleaned > 0)
1812 : : {
1813 [ + - + - : 4266 : obatched(clog) << "fdcache evicted num=" << cleaned << " of=" << entries_original
+ - + - ]
1814 [ + - + - : 1422 : << " min=" << cleaned_score_min << " max=" << cleaned_score_max
+ - + - +
- ]
1815 : 1422 : << endl;
1816 : : }
1817 : :
1818 [ + + + - ]: 1586 : if (metrics_p) set_metrics();
1819 : 1586 : }
1820 : :
1821 : :
1822 : 84 : ~libarchive_fdcache()
1823 : : {
1824 : : // unlink any fdcache entries in $TMPDIR
1825 : : // don't update metrics; those globals may be already destroyed
1826 : 84 : limit(0, false);
1827 : 84 : }
1828 : : };
1829 : : static libarchive_fdcache fdcache;
1830 : :
1831 : : /* Search ELF_FD for an ELF/DWARF section with name SECTION.
1832 : : If found copy the section to a temporary file and return
1833 : : its file descriptor, otherwise return -1.
1834 : :
1835 : : The temporary file's mtime will be set to PARENT_MTIME.
1836 : : B_SOURCE should be a description of the parent file suitable
1837 : : for printing to the log. */
1838 : :
1839 : : static int
1840 : 12 : extract_section (int elf_fd, int64_t parent_mtime,
1841 : : const string& b_source, const string& section,
1842 : : const timespec& extract_begin)
1843 : : {
1844 : : /* Search the fdcache. */
1845 : 12 : struct stat fs;
1846 : 12 : int fd = fdcache.lookup (b_source, section);
1847 [ - + ]: 12 : if (fd >= 0)
1848 : : {
1849 [ # # ]: 0 : if (fstat (fd, &fs) != 0)
1850 : : {
1851 [ # # ]: 0 : if (verbose)
1852 [ # # ]: 0 : obatched (clog) << "cannot fstate fdcache "
1853 [ # # # # : 0 : << b_source << " " << section << endl;
# # ]
1854 : 0 : close (fd);
1855 : 0 : return -1;
1856 : : }
1857 [ # # ]: 0 : if ((int64_t) fs.st_mtime != parent_mtime)
1858 : : {
1859 [ # # ]: 0 : if (verbose)
1860 [ # # ]: 0 : obatched(clog) << "mtime mismatch for "
1861 [ # # # # : 0 : << b_source << " " << section << endl;
# # ]
1862 : 0 : close (fd);
1863 : 0 : return -1;
1864 : : }
1865 : : /* Success. */
1866 : : return fd;
1867 : : }
1868 : :
1869 : 12 : Elf *elf = elf_begin (elf_fd, ELF_C_READ_MMAP_PRIVATE, NULL);
1870 [ - + ]: 12 : if (elf == NULL)
1871 : : return -1;
1872 : :
1873 : : /* Try to find the section and copy the contents into a separate file. */
1874 : 12 : try
1875 : : {
1876 : 12 : size_t shstrndx;
1877 [ + - ]: 12 : int rc = elf_getshdrstrndx (elf, &shstrndx);
1878 [ - + ]: 12 : if (rc < 0)
1879 [ # # # # ]: 0 : throw elfutils_exception (rc, "getshdrstrndx");
1880 : :
1881 : : Elf_Scn *scn = NULL;
1882 : 424 : while (true)
1883 : : {
1884 [ + - ]: 218 : scn = elf_nextscn (elf, scn);
1885 [ + - ]: 218 : if (scn == NULL)
1886 : : break;
1887 : 218 : GElf_Shdr shdr_storage;
1888 [ + - ]: 218 : GElf_Shdr *shdr = gelf_getshdr (scn, &shdr_storage);
1889 [ + - ]: 218 : if (shdr == NULL)
1890 : : break;
1891 : :
1892 [ + - ]: 218 : const char *scn_name = elf_strptr (elf, shstrndx, shdr->sh_name);
1893 [ + - ]: 218 : if (scn_name == NULL)
1894 : : break;
1895 [ + + ]: 218 : if (scn_name == section)
1896 : : {
1897 : 12 : Elf_Data *data = NULL;
1898 : :
1899 : : /* We found the desired section. */
1900 [ + - ]: 12 : data = elf_rawdata (scn, NULL);
1901 [ - + ]: 12 : if (data == NULL)
1902 [ # # # # : 0 : throw elfutils_exception (elf_errno (), "elfraw_data");
# # ]
1903 [ + + ]: 12 : if (data->d_buf == NULL)
1904 : : {
1905 [ + - + - ]: 8 : obatched(clog) << "section " << section
1906 [ + - + - ]: 4 : << " is empty" << endl;
1907 : 4 : break;
1908 : : }
1909 : :
1910 : : /* Create temporary file containing the section. */
1911 : 8 : char *tmppath = NULL;
1912 : 8 : rc = asprintf (&tmppath, "%s/debuginfod-section.XXXXXX", tmpdir.c_str());
1913 [ - + ]: 8 : if (rc < 0)
1914 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate tmppath");
1915 : 8 : defer_dtor<void*,void> tmmpath_freer (tmppath, free);
1916 [ + - ]: 8 : fd = mkstemp (tmppath);
1917 [ - + ]: 8 : if (fd < 0)
1918 [ # # # # ]: 0 : throw libc_exception (errno, "cannot create temporary file");
1919 : :
1920 [ + - ]: 8 : ssize_t res = write_retry (fd, data->d_buf, data->d_size);
1921 [ + - - + ]: 8 : if (res < 0 || (size_t) res != data->d_size) {
1922 [ # # ]: 0 : close (fd);
1923 : 0 : unlink (tmppath);
1924 [ # # # # ]: 0 : throw libc_exception (errno, "cannot write to temporary file");
1925 : : }
1926 : :
1927 : : /* Set mtime to be the same as the parent file's mtime. */
1928 : 8 : struct timespec tvs[2];
1929 [ - + ]: 8 : if (fstat (elf_fd, &fs) != 0) {
1930 [ # # ]: 0 : close (fd);
1931 : 0 : unlink (tmppath);
1932 [ # # # # ]: 0 : throw libc_exception (errno, "cannot fstat file");
1933 : : }
1934 : :
1935 : 8 : tvs[0].tv_sec = 0;
1936 : 8 : tvs[0].tv_nsec = UTIME_OMIT;
1937 : 8 : tvs[1] = fs.st_mtim;
1938 : 8 : (void) futimens (fd, tvs);
1939 : :
1940 : 8 : struct timespec extract_end;
1941 : 8 : clock_gettime (CLOCK_MONOTONIC, &extract_end);
1942 : 8 : double extract_time = (extract_end.tv_sec - extract_begin.tv_sec)
1943 : 8 : + (extract_end.tv_nsec - extract_begin.tv_nsec)/1.e9;
1944 : :
1945 : : /* Add to fdcache. */
1946 [ + - + - ]: 8 : fdcache.intern (b_source, section, tmppath, data->d_size, true, extract_time);
1947 : 8 : break;
1948 : 8 : }
1949 : 206 : }
1950 : : }
1951 [ - - ]: 0 : catch (const reportable_exception &e)
1952 : : {
1953 [ - - ]: 0 : e.report (clog);
1954 [ - - ]: 0 : close (fd);
1955 : 0 : fd = -1;
1956 : 0 : }
1957 : :
1958 : 12 : elf_end (elf);
1959 : : return fd;
1960 : : }
1961 : :
1962 : : static struct MHD_Response*
1963 : 1152 : handle_buildid_f_match (bool internal_req_t,
1964 : : int64_t b_mtime,
1965 : : const string& b_source0,
1966 : : const string& section,
1967 : : int *result_fd)
1968 : : {
1969 : 1152 : (void) internal_req_t; // ignored
1970 : :
1971 : 1152 : struct timespec extract_begin;
1972 : 1152 : clock_gettime (CLOCK_MONOTONIC, &extract_begin);
1973 : :
1974 : 1152 : int fd = open(b_source0.c_str(), O_RDONLY);
1975 [ - + ]: 1152 : if (fd < 0)
1976 [ # # # # : 0 : throw libc_exception (errno, string("open ") + b_source0);
# # ]
1977 : :
1978 : : // NB: use manual close(2) in error case instead of defer_dtor, because
1979 : : // in the normal case, we want to hand the fd over to libmicrohttpd for
1980 : : // file transfer.
1981 : :
1982 : 1152 : struct stat s;
1983 : 1152 : int rc = fstat(fd, &s);
1984 [ - + ]: 1152 : if (rc < 0)
1985 : : {
1986 : 0 : close(fd);
1987 [ # # # # : 0 : throw libc_exception (errno, string("fstat ") + b_source0);
# # ]
1988 : : }
1989 : :
1990 [ - + ]: 1152 : if ((int64_t) s.st_mtime != b_mtime)
1991 : : {
1992 [ # # ]: 0 : if (verbose)
1993 [ # # # # ]: 0 : obatched(clog) << "mtime mismatch for " << b_source0 << endl;
1994 : 0 : close(fd);
1995 : 0 : return 0;
1996 : : }
1997 : :
1998 [ + + ]: 1152 : if (!section.empty ())
1999 : : {
2000 : 6 : int scn_fd = extract_section (fd, s.st_mtime, b_source0, section, extract_begin);
2001 : 6 : close (fd);
2002 : :
2003 [ + + ]: 6 : if (scn_fd >= 0)
2004 : 4 : fd = scn_fd;
2005 : : else
2006 : : {
2007 [ + - ]: 2 : if (verbose)
2008 [ + - ]: 6 : obatched (clog) << "cannot find section " << section
2009 [ + - + - : 2 : << " for " << b_source0 << endl;
+ - ]
2010 : 2 : return 0;
2011 : : }
2012 : :
2013 : 4 : rc = fstat(fd, &s);
2014 [ - + ]: 4 : if (rc < 0)
2015 : : {
2016 : 0 : close (fd);
2017 [ # # # # ]: 0 : throw libc_exception (errno, string ("fstat ") + b_source0
2018 [ # # # # : 0 : + string (" ") + section);
# # # # ]
2019 : : }
2020 : : }
2021 : :
2022 : 1150 : struct MHD_Response* r = MHD_create_response_from_fd ((uint64_t) s.st_size, fd);
2023 [ + - + - : 2300 : inc_metric ("http_responses_total","result","file");
+ - ]
2024 [ - + ]: 1150 : if (r == 0)
2025 : : {
2026 [ # # ]: 0 : if (verbose)
2027 [ # # ]: 0 : obatched(clog) << "cannot create fd-response for " << b_source0
2028 [ # # # # : 0 : << " section=" << section << endl;
# # ]
2029 : 0 : close(fd);
2030 : : }
2031 : : else
2032 : : {
2033 : 1150 : add_mhd_response_header (r, "Content-Type", "application/octet-stream");
2034 [ + - ]: 1150 : add_mhd_response_header (r, "X-DEBUGINFOD-SIZE",
2035 : 1150 : to_string(s.st_size).c_str());
2036 : 1150 : add_mhd_response_header (r, "X-DEBUGINFOD-FILE", b_source0.c_str());
2037 : 1150 : add_mhd_last_modified (r, s.st_mtime);
2038 [ + - ]: 1150 : if (verbose > 1)
2039 [ + - + - : 2300 : obatched(clog) << "serving file " << b_source0 << " section=" << section << endl;
+ - + - ]
2040 : : /* libmicrohttpd will close it. */
2041 [ - + ]: 1150 : if (result_fd)
2042 : 1150 : *result_fd = fd;
2043 : : }
2044 : :
2045 : : return r;
2046 : : }
2047 : :
2048 : :
2049 : : #ifdef USE_LZMA
2050 : : struct lzma_exception: public reportable_exception
2051 : : {
2052 : 0 : lzma_exception(int rc, const string& msg):
2053 : : // liblzma doesn't have a lzma_ret -> string conversion function, so just
2054 : : // report the value.
2055 [ # # # # : 0 : reportable_exception(string ("lzma error: ") + msg + ": error " + to_string(rc)) {
# # ]
2056 [ # # # # : 0 : inc_metric("error_count","lzma",to_string(rc));
# # ]
2057 : 0 : }
2058 : : };
2059 : :
2060 : : // Neither RPM nor deb files support seeking to a specific file in the package.
2061 : : // Instead, to extract a specific file, we normally need to read the archive
2062 : : // sequentially until we find the file. This is very slow for files at the end
2063 : : // of a large package with lots of files, like kernel debuginfo.
2064 : : //
2065 : : // However, if the compression format used in the archive supports seeking, we
2066 : : // can accelerate this. As of July 2024, xz is the only widely-used format that
2067 : : // supports seeking, and usually only in multi-threaded mode. Luckily, the
2068 : : // kernel-debuginfo package in Fedora and its downstreams, and the
2069 : : // linux-image-*-dbg package in Debian and its downstreams, all happen to use
2070 : : // this.
2071 : : //
2072 : : // The xz format [1] ends with an index of independently compressed blocks in
2073 : : // the stream. In RPM and deb files, the xz stream is the last thing in the
2074 : : // file, so we assume that the xz Stream Footer is at the end of the package
2075 : : // file and do everything relative to that. For each file in the archive, we
2076 : : // remember the size and offset of the file data in the uncompressed xz stream,
2077 : : // then we use the index to seek to that offset when we need that file.
2078 : : //
2079 : : // 1: https://xz.tukaani.org/format/xz-file-format.txt
2080 : :
2081 : : // Return whether an archive supports seeking.
2082 : : static bool
2083 : 1092 : is_seekable_archive (const string& rps, struct archive* a)
2084 : : {
2085 : : // Only xz supports seeking.
2086 [ + + ]: 1092 : if (archive_filter_code (a, 0) != ARCHIVE_FILTER_XZ)
2087 : : return false;
2088 : :
2089 : 390 : int fd = open (rps.c_str(), O_RDONLY);
2090 [ - + ]: 390 : if (fd < 0)
2091 : : return false;
2092 : 390 : defer_dtor<int,int> fd_closer (fd, close);
2093 : :
2094 : : // Seek to the xz Stream Footer. We assume that it's the last thing in the
2095 : : // file, which is true for RPM and deb files.
2096 : 390 : off_t footer_pos = -LZMA_STREAM_HEADER_SIZE;
2097 [ - + ]: 390 : if (lseek (fd, footer_pos, SEEK_END) == -1)
2098 : : return false;
2099 : :
2100 : : // Decode the Stream Footer.
2101 : : uint8_t footer[LZMA_STREAM_HEADER_SIZE];
2102 : : size_t footer_read = 0;
2103 [ + + ]: 780 : while (footer_read < sizeof (footer))
2104 : : {
2105 [ - + ]: 390 : ssize_t bytes_read = read (fd, footer + footer_read,
2106 : : sizeof (footer) - footer_read);
2107 [ - + ]: 390 : if (bytes_read < 0)
2108 : : {
2109 [ # # ]: 0 : if (errno == EINTR)
2110 : 0 : continue;
2111 : : return false;
2112 : : }
2113 [ - + ]: 390 : if (bytes_read == 0)
2114 : : return false;
2115 : 390 : footer_read += bytes_read;
2116 : : }
2117 : :
2118 : 390 : lzma_stream_flags stream_flags;
2119 : 390 : lzma_ret ret = lzma_stream_footer_decode (&stream_flags, footer);
2120 [ - + ]: 390 : if (ret != LZMA_OK)
2121 : : return false;
2122 : :
2123 : : // Seek to the xz Index.
2124 [ - + ]: 390 : if (lseek (fd, footer_pos - stream_flags.backward_size, SEEK_END) == -1)
2125 : : return false;
2126 : :
2127 : : // Decode the Number of Records in the Index. liblzma doesn't have an API for
2128 : : // this if you don't want to decode the whole Index, so we have to do it
2129 : : // ourselves.
2130 : : //
2131 : : // We need 1 byte for the Index Indicator plus 1-9 bytes for the
2132 : : // variable-length integer Number of Records.
2133 : : uint8_t index[10];
2134 : : size_t index_read = 0;
2135 [ + + ]: 780 : while (index_read == 0) {
2136 [ + - ]: 390 : ssize_t bytes_read = read (fd, index, sizeof (index));
2137 [ - + ]: 390 : if (bytes_read < 0)
2138 : : {
2139 [ # # ]: 0 : if (errno == EINTR)
2140 : 0 : continue;
2141 : : return false;
2142 : : }
2143 [ - + ]: 390 : if (bytes_read == 0)
2144 : : return false;
2145 : 390 : index_read += bytes_read;
2146 : : }
2147 : : // The Index Indicator must be 0.
2148 [ - + ]: 390 : if (index[0] != 0)
2149 : : return false;
2150 : :
2151 : 390 : lzma_vli num_records;
2152 : 390 : size_t pos = 0;
2153 : 390 : size_t in_pos = 1;
2154 : 390 : while (true)
2155 : : {
2156 [ - + ]: 390 : if (in_pos >= index_read)
2157 : : {
2158 [ # # ]: 0 : ssize_t bytes_read = read (fd, index, sizeof (index));
2159 [ # # ]: 0 : if (bytes_read < 0)
2160 : : {
2161 [ # # ]: 0 : if (errno == EINTR)
2162 : 0 : continue;
2163 : : return false;
2164 : : }
2165 [ # # ]: 0 : if (bytes_read == 0)
2166 : : return false;
2167 : 0 : index_read = bytes_read;
2168 : 0 : in_pos = 0;
2169 : : }
2170 : 390 : ret = lzma_vli_decode (&num_records, &pos, index, &in_pos, index_read);
2171 [ - + ]: 390 : if (ret == LZMA_STREAM_END)
2172 : : break;
2173 [ # # ]: 0 : else if (ret != LZMA_OK)
2174 : : return false;
2175 : : }
2176 : :
2177 [ + + ]: 390 : if (verbose > 3)
2178 [ + - + - : 676 : obatched(clog) << rps << " has " << num_records << " xz Blocks" << endl;
+ - + - +
- ]
2179 : :
2180 : : // The file is only seekable if it has more than one Block.
2181 : 390 : return num_records > 1;
2182 : 390 : }
2183 : :
2184 : : // Read the Index at the end of an xz file.
2185 : : static lzma_index*
2186 : 202 : read_xz_index (int fd)
2187 : : {
2188 : 202 : off_t footer_pos = -LZMA_STREAM_HEADER_SIZE;
2189 [ - + ]: 202 : if (lseek (fd, footer_pos, SEEK_END) == -1)
2190 [ # # # # ]: 0 : throw libc_exception (errno, "lseek");
2191 : :
2192 : : uint8_t footer[LZMA_STREAM_HEADER_SIZE];
2193 : : size_t footer_read = 0;
2194 [ + + ]: 404 : while (footer_read < sizeof (footer))
2195 : : {
2196 [ - + ]: 202 : ssize_t bytes_read = read (fd, footer + footer_read,
2197 : : sizeof (footer) - footer_read);
2198 [ - + ]: 202 : if (bytes_read < 0)
2199 : : {
2200 [ # # ]: 0 : if (errno == EINTR)
2201 : 0 : continue;
2202 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2203 : : }
2204 [ - + ]: 202 : if (bytes_read == 0)
2205 [ # # ]: 0 : throw reportable_exception ("truncated file");
2206 : 202 : footer_read += bytes_read;
2207 : : }
2208 : :
2209 : 202 : lzma_stream_flags stream_flags;
2210 : 202 : lzma_ret ret = lzma_stream_footer_decode (&stream_flags, footer);
2211 [ - + ]: 202 : if (ret != LZMA_OK)
2212 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_stream_footer_decode");
2213 : :
2214 [ - + ]: 202 : if (lseek (fd, footer_pos - stream_flags.backward_size, SEEK_END) == -1)
2215 [ # # # # ]: 0 : throw libc_exception (errno, "lseek");
2216 : :
2217 : 202 : lzma_stream strm = LZMA_STREAM_INIT;
2218 : 202 : lzma_index* index = NULL;
2219 : 202 : ret = lzma_index_decoder (&strm, &index, UINT64_MAX);
2220 [ - + ]: 202 : if (ret != LZMA_OK)
2221 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_index_decoder");
2222 : 202 : defer_dtor<lzma_stream*,void> strm_ender (&strm, lzma_end);
2223 : :
2224 : 202 : uint8_t in_buf[4096];
2225 : 202 : while (true)
2226 : : {
2227 [ + - ]: 202 : if (strm.avail_in == 0)
2228 : : {
2229 [ + - ]: 202 : ssize_t bytes_read = read (fd, in_buf, sizeof (in_buf));
2230 [ - + ]: 202 : if (bytes_read < 0)
2231 : : {
2232 [ # # ]: 0 : if (errno == EINTR)
2233 : 0 : continue;
2234 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2235 : : }
2236 [ - + ]: 202 : if (bytes_read == 0)
2237 [ # # ]: 0 : throw reportable_exception ("truncated file");
2238 : 202 : strm.avail_in = bytes_read;
2239 : 202 : strm.next_in = in_buf;
2240 : : }
2241 : :
2242 : 202 : ret = lzma_code (&strm, LZMA_RUN);
2243 [ - + ]: 202 : if (ret == LZMA_STREAM_END)
2244 : : break;
2245 [ # # ]: 0 : else if (ret != LZMA_OK)
2246 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_code index");
2247 : : }
2248 : :
2249 : 202 : ret = lzma_index_stream_flags (index, &stream_flags);
2250 [ - + ]: 202 : if (ret != LZMA_OK)
2251 : : {
2252 : 0 : lzma_index_end (index, NULL);
2253 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_index_stream_flags");
2254 : : }
2255 : 202 : return index;
2256 : 202 : }
2257 : :
2258 : : static void
2259 : 202 : my_lzma_index_end (lzma_index* index)
2260 : : {
2261 : 202 : lzma_index_end (index, NULL);
2262 : 202 : }
2263 : :
2264 : : static void
2265 : 210 : free_lzma_block_filter_options (lzma_block* block)
2266 : : {
2267 [ + + ]: 1050 : for (int i = 0; i < LZMA_FILTERS_MAX; i++)
2268 : : {
2269 : 840 : free (block->filters[i].options);
2270 : 840 : block->filters[i].options = NULL;
2271 : : }
2272 : 210 : }
2273 : :
2274 : : static void
2275 : 202 : free_lzma_block_filters (lzma_block* block)
2276 : : {
2277 [ + - ]: 202 : if (block->filters != NULL)
2278 : : {
2279 : 202 : free_lzma_block_filter_options (block);
2280 : 202 : free (block->filters);
2281 : : }
2282 : 202 : }
2283 : :
2284 : : static void
2285 : 202 : extract_xz_blocks_into_fd (const string& srcpath,
2286 : : int src,
2287 : : int dst,
2288 : : lzma_index_iter* iter,
2289 : : uint64_t offset,
2290 : : uint64_t size)
2291 : : {
2292 : : // Seek to the Block. Seeking from the end using the compressed size from the
2293 : : // footer means we don't need to know where the xz stream starts in the
2294 : : // archive.
2295 [ - + ]: 202 : if (lseek (src,
2296 : 202 : (off_t) iter->block.compressed_stream_offset
2297 : 202 : - (off_t) iter->stream.compressed_size,
2298 : : SEEK_END) == -1)
2299 [ # # # # ]: 0 : throw libc_exception (errno, "lseek");
2300 : :
2301 : 202 : offset -= iter->block.uncompressed_file_offset;
2302 : :
2303 : 202 : lzma_block block{};
2304 : 202 : block.filters = (lzma_filter*) calloc (LZMA_FILTERS_MAX + 1,
2305 : : sizeof (lzma_filter));
2306 [ - + ]: 202 : if (block.filters == NULL)
2307 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate lzma_block filters");
2308 : 202 : defer_dtor<lzma_block*,void> filters_freer (&block, free_lzma_block_filters);
2309 : :
2310 : 202 : uint8_t in_buf[4096];
2311 : 202 : uint8_t out_buf[4096];
2312 : 202 : size_t header_read = 0;
2313 : 202 : bool need_log_extracting = verbose > 3;
2314 : 8 : while (true)
2315 : : {
2316 : : // The first byte of the Block is the encoded Block Header Size. Read the
2317 : : // first byte and whatever extra fits in the buffer.
2318 [ + + ]: 412 : while (header_read == 0)
2319 : : {
2320 [ + - ]: 202 : ssize_t bytes_read = read (src, in_buf, sizeof (in_buf));
2321 [ - + ]: 202 : if (bytes_read < 0)
2322 : : {
2323 [ # # ]: 0 : if (errno == EINTR)
2324 : 0 : continue;
2325 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2326 : : }
2327 [ - + ]: 202 : if (bytes_read == 0)
2328 [ # # ]: 0 : throw reportable_exception ("truncated file");
2329 : 202 : header_read += bytes_read;
2330 : : }
2331 : :
2332 : 210 : block.header_size = lzma_block_header_size_decode (in_buf[0]);
2333 : :
2334 : : // If we didn't buffer the whole Block Header earlier, get the rest.
2335 : 210 : eu_static_assert (sizeof (in_buf)
2336 : : >= lzma_block_header_size_decode (UINT8_MAX));
2337 [ - + ]: 210 : while (header_read < block.header_size)
2338 : : {
2339 [ # # ]: 0 : ssize_t bytes_read = read (src, in_buf + header_read,
2340 : : sizeof (in_buf) - header_read);
2341 [ # # ]: 0 : if (bytes_read < 0)
2342 : : {
2343 [ # # ]: 0 : if (errno == EINTR)
2344 : 0 : continue;
2345 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2346 : : }
2347 [ # # ]: 0 : if (bytes_read == 0)
2348 [ # # ]: 0 : throw reportable_exception ("truncated file");
2349 : 0 : header_read += bytes_read;
2350 : : }
2351 : :
2352 : : // Decode the Block Header.
2353 : 210 : block.check = iter->stream.flags->check;
2354 : 210 : lzma_ret ret = lzma_block_header_decode (&block, NULL, in_buf);
2355 [ - + ]: 210 : if (ret != LZMA_OK)
2356 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_block_header_decode");
2357 : 210 : ret = lzma_block_compressed_size (&block, iter->block.unpadded_size);
2358 [ - + ]: 210 : if (ret != LZMA_OK)
2359 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_block_compressed_size");
2360 : :
2361 : : // Start decoding the Block data.
2362 : 210 : lzma_stream strm = LZMA_STREAM_INIT;
2363 : 210 : ret = lzma_block_decoder (&strm, &block);
2364 [ - + ]: 210 : if (ret != LZMA_OK)
2365 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_block_decoder");
2366 : 210 : defer_dtor<lzma_stream*,void> strm_ender (&strm, lzma_end);
2367 : :
2368 : : // We might still have some input buffered from when we read the header.
2369 : 210 : strm.avail_in = header_read - block.header_size;
2370 : 210 : strm.next_in = in_buf + block.header_size;
2371 : 210 : strm.avail_out = sizeof (out_buf);
2372 : 210 : strm.next_out = out_buf;
2373 : 11734 : while (true)
2374 : : {
2375 [ + + ]: 11734 : if (strm.avail_in == 0)
2376 : : {
2377 [ + - ]: 14 : ssize_t bytes_read = read (src, in_buf, sizeof (in_buf));
2378 [ - + ]: 14 : if (bytes_read < 0)
2379 : : {
2380 [ # # ]: 0 : if (errno == EINTR)
2381 : 0 : continue;
2382 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2383 : : }
2384 [ - + ]: 14 : if (bytes_read == 0)
2385 [ # # ]: 0 : throw reportable_exception ("truncated file");
2386 : 14 : strm.avail_in = bytes_read;
2387 : 14 : strm.next_in = in_buf;
2388 : : }
2389 : :
2390 : 11734 : ret = lzma_code (&strm, LZMA_RUN);
2391 [ - + ]: 11734 : if (ret != LZMA_OK && ret != LZMA_STREAM_END)
2392 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_code block");
2393 : :
2394 : : // Throw away anything we decode until we reach the offset, then
2395 : : // start writing to the destination.
2396 [ + + ]: 11734 : if (strm.total_out > offset)
2397 : : {
2398 : 3006 : size_t bytes_to_write = strm.next_out - out_buf;
2399 : 3006 : uint8_t* buf_to_write = out_buf;
2400 : :
2401 : : // Ignore anything in the buffer before the offset.
2402 [ + + ]: 3006 : if (bytes_to_write > strm.total_out - offset)
2403 : : {
2404 : 196 : buf_to_write += bytes_to_write - (strm.total_out - offset);
2405 : 196 : bytes_to_write = strm.total_out - offset;
2406 : : }
2407 : :
2408 : : // Ignore anything after the size.
2409 [ + + ]: 3006 : if (strm.total_out - offset >= size)
2410 : 202 : bytes_to_write -= strm.total_out - offset - size;
2411 : :
2412 [ + + ]: 3006 : if (need_log_extracting)
2413 : : {
2414 [ + - + - ]: 240 : obatched(clog) << "extracting from xz archive " << srcpath
2415 [ + - + - : 120 : << " size=" << size << endl;
+ - ]
2416 : 120 : need_log_extracting = false;
2417 : : }
2418 : :
2419 [ + + ]: 6012 : while (bytes_to_write > 0)
2420 : : {
2421 [ + - ]: 3006 : ssize_t written = write (dst, buf_to_write, bytes_to_write);
2422 [ - + ]: 3006 : if (written < 0)
2423 : : {
2424 [ # # ]: 0 : if (errno == EAGAIN)
2425 : 0 : continue;
2426 [ # # # # ]: 0 : throw libc_exception (errno, "write");
2427 : : }
2428 : 3006 : bytes_to_write -= written;
2429 : 3006 : buf_to_write += written;
2430 : : }
2431 : :
2432 : : // If we reached the size, we're done.
2433 [ + + ]: 3006 : if (strm.total_out - offset >= size)
2434 : 202 : return;
2435 : : }
2436 : :
2437 : 11532 : strm.avail_out = sizeof (out_buf);
2438 : 11532 : strm.next_out = out_buf;
2439 : :
2440 [ + + ]: 11532 : if (ret == LZMA_STREAM_END)
2441 : : break;
2442 : : }
2443 : :
2444 : : // This Block didn't have enough data. Go to the next one.
2445 [ - + ]: 8 : if (lzma_index_iter_next (iter, LZMA_INDEX_ITER_BLOCK))
2446 [ # # ]: 0 : throw reportable_exception ("no more blocks");
2447 [ + - ]: 8 : if (strm.total_out > offset)
2448 : 8 : size -= strm.total_out - offset;
2449 : 8 : offset = 0;
2450 : : // If we had any buffered input left, move it to the beginning of the
2451 : : // buffer to decode the next Block Header.
2452 [ + - ]: 8 : if (strm.avail_in > 0)
2453 : : {
2454 : 8 : memmove (in_buf, strm.next_in, strm.avail_in);
2455 : 8 : header_read = strm.avail_in;
2456 : : }
2457 : : else
2458 : : header_read = 0;
2459 : 8 : free_lzma_block_filter_options (&block);
2460 : 210 : }
2461 : 202 : }
2462 : :
2463 : : static int
2464 : 202 : extract_from_seekable_archive (const string& srcpath,
2465 : : char* tmppath,
2466 : : uint64_t offset,
2467 : : uint64_t size)
2468 : : {
2469 [ + - + - : 404 : inc_metric ("seekable_archive_extraction_attempts","type","xz");
+ - ]
2470 : 202 : try
2471 : : {
2472 [ + - ]: 202 : int src = open (srcpath.c_str(), O_RDONLY);
2473 [ - + ]: 202 : if (src < 0)
2474 [ # # # # : 0 : throw libc_exception (errno, string("open ") + srcpath);
# # ]
2475 : 202 : defer_dtor<int,int> src_closer (src, close);
2476 : :
2477 [ + - ]: 202 : lzma_index* index = read_xz_index (src);
2478 : 202 : defer_dtor<lzma_index*,void> index_ender (index, my_lzma_index_end);
2479 : :
2480 : : // Find the Block containing the offset.
2481 : 202 : lzma_index_iter iter;
2482 : 202 : lzma_index_iter_init (&iter, index);
2483 [ - + ]: 202 : if (lzma_index_iter_locate (&iter, offset))
2484 [ # # ]: 0 : throw reportable_exception ("offset not found");
2485 : :
2486 [ + + ]: 202 : if (verbose > 3)
2487 [ + - + - ]: 360 : obatched(clog) << "seeking in xz archive " << srcpath
2488 [ + - + - : 120 : << " offset=" << offset << " block_offset="
+ - ]
2489 [ + - + - ]: 120 : << iter.block.uncompressed_file_offset << endl;
2490 : :
2491 [ + - ]: 202 : int dst = mkstemp (tmppath);
2492 [ - + ]: 202 : if (dst < 0)
2493 [ # # # # ]: 0 : throw libc_exception (errno, "cannot create temporary file");
2494 : :
2495 : 202 : try
2496 : : {
2497 [ + - ]: 202 : extract_xz_blocks_into_fd (srcpath, src, dst, &iter, offset, size);
2498 : : }
2499 : 0 : catch (...)
2500 : : {
2501 : 0 : unlink (tmppath);
2502 [ - - ]: 0 : close (dst);
2503 : 0 : throw;
2504 : 0 : }
2505 : :
2506 [ + - + - : 404 : inc_metric ("seekable_archive_extraction_successes","type","xz");
+ - + - ]
2507 : 202 : return dst;
2508 : 202 : }
2509 [ - - ]: 0 : catch (const reportable_exception &e)
2510 : : {
2511 [ - - - - : 0 : inc_metric ("seekable_archive_extraction_failures","type","xz");
- - - - ]
2512 [ - - ]: 0 : if (verbose)
2513 [ - - - - ]: 0 : obatched(clog) << "failed to extract from seekable xz archive "
2514 [ - - - - : 0 : << srcpath << ": " << e.message << endl;
- - ]
2515 : 0 : return -1;
2516 : 0 : }
2517 : : }
2518 : : #else
2519 : : static bool
2520 : : is_seekable_archive (const string& rps __attribute__ ((unused)),
2521 : : struct archive* a __attribute__ ((unused)))
2522 : : {
2523 : : return false;
2524 : : }
2525 : : static int
2526 : : extract_from_seekable_archive (const string& srcpath __attribute__ ((unused)),
2527 : : char* tmppath __attribute__ ((unused)),
2528 : : uint64_t offset __attribute__ ((unused)),
2529 : : uint64_t size __attribute__ ((unused)))
2530 : : {
2531 : : return -1;
2532 : : }
2533 : : #endif
2534 : :
2535 : :
2536 : : // For security/portability reasons, many distro-package archives have
2537 : : // a "./" in front of path names; others have nothing, others have
2538 : : // "/". Canonicalize them all to a single leading "/", with the
2539 : : // assumption that this matches the dwarf-derived file names too.
2540 : 3966 : string canonicalized_archive_entry_pathname(struct archive_entry *e)
2541 : : {
2542 : 3966 : string fn = archive_entry_pathname(e);
2543 [ - + ]: 3966 : if (fn.size() == 0)
2544 : 0 : return fn;
2545 [ - + ]: 3966 : if (fn[0] == '/')
2546 : 0 : return fn;
2547 [ + + ]: 3966 : if (fn[0] == '.')
2548 [ + - ]: 2726 : return fn.substr(1);
2549 : : else
2550 [ + - + - ]: 2480 : return string("/")+fn;
2551 : 3966 : }
2552 : :
2553 : :
2554 : : // NB: takes ownership of, and may reassign, fd.
2555 : : static struct MHD_Response*
2556 : 1000 : create_buildid_r_response (int64_t b_mtime0,
2557 : : const string& b_source0,
2558 : : const string& b_source1,
2559 : : const string& section,
2560 : : const string& ima_sig,
2561 : : const char* tmppath,
2562 : : int& fd,
2563 : : off_t size,
2564 : : time_t mtime,
2565 : : const string& metric,
2566 : : const struct timespec& extract_begin)
2567 : : {
2568 [ + + ]: 1000 : if (tmppath != NULL)
2569 : : {
2570 : 900 : struct timespec extract_end;
2571 : 900 : clock_gettime (CLOCK_MONOTONIC, &extract_end);
2572 : 900 : double extract_time = (extract_end.tv_sec - extract_begin.tv_sec)
2573 : 900 : + (extract_end.tv_nsec - extract_begin.tv_nsec)/1.e9;
2574 [ + - ]: 1800 : fdcache.intern(b_source0, b_source1, tmppath, size, true, extract_time);
2575 : : }
2576 : :
2577 [ + + ]: 1000 : if (!section.empty ())
2578 : : {
2579 [ + - ]: 6 : int scn_fd = extract_section (fd, b_mtime0,
2580 [ + - ]: 12 : b_source0 + ":" + b_source1,
2581 : : section, extract_begin);
2582 : 6 : close (fd);
2583 [ + + ]: 6 : if (scn_fd >= 0)
2584 : 4 : fd = scn_fd;
2585 : : else
2586 : : {
2587 [ + - ]: 2 : if (verbose)
2588 [ + - ]: 6 : obatched (clog) << "cannot find section " << section
2589 : : << " for archive " << b_source0
2590 [ + - + - : 2 : << " file " << b_source1 << endl;
+ - + - +
- ]
2591 : 2 : return 0;
2592 : : }
2593 : :
2594 : 4 : struct stat fs;
2595 [ - + ]: 4 : if (fstat (fd, &fs) < 0)
2596 : : {
2597 : 0 : close (fd);
2598 : 0 : throw libc_exception (errno,
2599 [ # # # # : 0 : string ("fstat ") + b_source0 + string (" ") + section);
# # # # #
# # # ]
2600 : : }
2601 : 4 : size = fs.st_size;
2602 : : }
2603 : :
2604 : 998 : struct MHD_Response* r = MHD_create_response_from_fd (size, fd);
2605 [ - + ]: 998 : if (r == 0)
2606 : : {
2607 [ # # ]: 0 : if (verbose)
2608 [ # # # # ]: 0 : obatched(clog) << "cannot create fd-response for " << b_source0 << endl;
2609 : 0 : close(fd);
2610 : : }
2611 : : else
2612 : : {
2613 [ + - + - ]: 1996 : inc_metric ("http_responses_total","result",metric);
2614 : 998 : add_mhd_response_header (r, "Content-Type", "application/octet-stream");
2615 [ + - ]: 998 : add_mhd_response_header (r, "X-DEBUGINFOD-SIZE", to_string(size).c_str());
2616 : 998 : add_mhd_response_header (r, "X-DEBUGINFOD-ARCHIVE", b_source0.c_str());
2617 : 998 : add_mhd_response_header (r, "X-DEBUGINFOD-FILE", b_source1.c_str());
2618 [ - + ]: 998 : if(!ima_sig.empty()) add_mhd_response_header(r, "X-DEBUGINFOD-IMASIGNATURE", ima_sig.c_str());
2619 : 998 : add_mhd_last_modified (r, mtime);
2620 [ - + ]: 998 : if (verbose > 1)
2621 [ + - ]: 2994 : obatched(clog) << "serving " << metric << " " << b_source0
2622 : : << " file " << b_source1
2623 : : << " section=" << section
2624 [ + - + - : 998 : << " IMA signature=" << ima_sig << endl;
+ - + - +
- + - + -
+ - + - ]
2625 : : /* libmicrohttpd will close fd. */
2626 : : }
2627 : : return r;
2628 : : }
2629 : :
2630 : : static struct MHD_Response*
2631 : 1058 : handle_buildid_r_match (bool internal_req_p,
2632 : : int64_t b_mtime,
2633 : : const string& b_source0,
2634 : : const string& b_source1,
2635 : : int64_t b_id0,
2636 : : int64_t b_id1,
2637 : : const string& section,
2638 : : int *result_fd)
2639 : : {
2640 : 1058 : struct timespec extract_begin;
2641 : 1058 : clock_gettime (CLOCK_MONOTONIC, &extract_begin);
2642 : :
2643 : 1058 : struct stat fs;
2644 : 1058 : int rc = stat (b_source0.c_str(), &fs);
2645 [ + + ]: 1058 : if (rc != 0)
2646 [ + - + - : 116 : throw libc_exception (errno, string("stat ") + b_source0);
+ - ]
2647 : :
2648 [ - + ]: 1000 : if ((int64_t) fs.st_mtime != b_mtime)
2649 : : {
2650 [ # # ]: 0 : if (verbose)
2651 [ # # # # ]: 0 : obatched(clog) << "mtime mismatch for " << b_source0 << endl;
2652 : 0 : return 0;
2653 : : }
2654 : :
2655 : : // Extract the IMA per-file signature (if it exists)
2656 : 1000 : string ima_sig = "";
2657 : : #ifdef ENABLE_IMA_VERIFICATION
2658 : : do
2659 : : {
2660 : : FD_t rpm_fd;
2661 : : if(!(rpm_fd = Fopen(b_source0.c_str(), "r.ufdio"))) // read, uncompressed, rpm/rpmio.h
2662 : : {
2663 : : if (verbose) obatched(clog) << "There was an error while opening " << b_source0 << endl;
2664 : : break; // Exit IMA extraction
2665 : : }
2666 : :
2667 : : Header rpm_hdr;
2668 : : if(RPMRC_FAIL == rpmReadPackageFile(NULL, rpm_fd, b_source0.c_str(), &rpm_hdr))
2669 : : {
2670 : : if (verbose) obatched(clog) << "There was an error while reading the header of " << b_source0 << endl;
2671 : : Fclose(rpm_fd);
2672 : : break; // Exit IMA extraction
2673 : : }
2674 : :
2675 : : // Fill sig_tag_data with an alloc'd copy of the array of IMA signatures (if they exist)
2676 : : struct rpmtd_s sig_tag_data;
2677 : : rpmtdReset(&sig_tag_data);
2678 : : do{ /* A do-while so we can break out of the koji sigcache checking on failure */
2679 : : if(requires_koji_sigcache_mapping)
2680 : : {
2681 : : /* NB: Koji builds result in a directory structure like the following
2682 : : - PACKAGE/VERSION/RELEASE
2683 : : - ARCH1
2684 : : - foo.rpm // The rpm known by debuginfod
2685 : : - ...
2686 : : - ARCHN
2687 : : - data
2688 : : - signed // Periodically purged (and not scanned by debuginfod)
2689 : : - sigcache
2690 : : - ARCH1
2691 : : - foo.rpm.sig // An empty rpm header
2692 : : - ...
2693 : : - ARCHN
2694 : : - PACKAGE_KEYID1
2695 : : - ARCH1
2696 : : - foo.rpm.sig // The header of the signed rpm. This is the file we need to extract the IMA signatures
2697 : : - ...
2698 : : - ARCHN
2699 : : - ...
2700 : : - PACKAGE_KEYIDn
2701 : :
2702 : : We therefore need to do a mapping:
2703 : :
2704 : : P/V/R/A/N-V-R.A.rpm ->
2705 : : P/V/R/data/sigcache/KEYID/A/N-V-R.A.rpm.sig
2706 : :
2707 : : There are 2 key insights here
2708 : :
2709 : : 1. We need to go 2 directories down from sigcache to get to the
2710 : : rpm header. So to distinguish ARCH1/foo.rpm.sig and
2711 : : PACKAGE_KEYID1/ARCH1/foo.rpm.sig we can look 2 directories down
2712 : :
2713 : : 2. It's safe to assume that the user will have all of the
2714 : : required verification certs. So we can pick from any of the
2715 : : PACKAGE_KEYID* directories. For simplicity we choose first we
2716 : : match against
2717 : :
2718 : : See: https://pagure.io/koji/issue/3670
2719 : : */
2720 : :
2721 : : // Do the mapping from b_source0 to the koji path for the signed rpm header
2722 : : string signed_rpm_path = b_source0;
2723 : : size_t insert_pos = string::npos;
2724 : : for(int i = 0; i < 2; i++) insert_pos = signed_rpm_path.rfind("/", insert_pos) - 1;
2725 : : string globbed_path = signed_rpm_path.insert(insert_pos + 1, "/data/sigcache/*").append(".sig"); // The globbed path we're seeking
2726 : : glob_t pglob;
2727 : : int grc;
2728 : : if(0 != (grc = glob(globbed_path.c_str(), GLOB_NOSORT, NULL, &pglob)))
2729 : : {
2730 : : // Break out, but only report real errors
2731 : : if (verbose && grc != GLOB_NOMATCH) obatched(clog) << "There was an error (" << strerror(errno) << ") globbing " << globbed_path << endl;
2732 : : break; // Exit koji sigcache check
2733 : : }
2734 : : signed_rpm_path = pglob.gl_pathv[0]; // See insight 2 above
2735 : : globfree(&pglob);
2736 : :
2737 : : if (verbose > 2) obatched(clog) << "attempting IMA signature extraction from koji header " << signed_rpm_path << endl;
2738 : :
2739 : : FD_t sig_rpm_fd;
2740 : : if(NULL == (sig_rpm_fd = Fopen(signed_rpm_path.c_str(), "r")))
2741 : : {
2742 : : if (verbose) obatched(clog) << "There was an error while opening " << signed_rpm_path << endl;
2743 : : break; // Exit koji sigcache check
2744 : : }
2745 : :
2746 : : Header sig_hdr = headerRead(sig_rpm_fd, HEADER_MAGIC_YES /* Validate magic too */ );
2747 : : if (!sig_hdr || 1 != headerGet(sig_hdr, RPMSIGTAG_FILESIGNATURES, &sig_tag_data, HEADERGET_ALLOC))
2748 : : {
2749 : : if (verbose) obatched(clog) << "Unable to extract RPMSIGTAG_FILESIGNATURES from " << signed_rpm_path << endl;
2750 : : }
2751 : : headerFree(sig_hdr); // We can free here since sig_tag_data has an alloc'd copy of the data
2752 : : Fclose(sig_rpm_fd);
2753 : : }
2754 : : }while(false);
2755 : :
2756 : : if(0 == sig_tag_data.count)
2757 : : {
2758 : : // In the general case (or a fallback from the koji sigcache mapping not finding signatures)
2759 : : // we can just (try) extract the signatures from the rpm header
2760 : : if (1 != headerGet(rpm_hdr, RPMTAG_FILESIGNATURES, &sig_tag_data, HEADERGET_ALLOC))
2761 : : {
2762 : : if (verbose) obatched(clog) << "Unable to extract RPMTAG_FILESIGNATURES from " << b_source0 << endl;
2763 : : }
2764 : : }
2765 : : // Search the array for the signature coresponding to b_source1
2766 : : int idx = -1;
2767 : : char *sig = NULL;
2768 : : rpmfi hdr_fi = rpmfiNew(NULL, rpm_hdr, RPMTAG_BASENAMES, RPMFI_FLAGS_QUERY);
2769 : : do
2770 : : {
2771 : : sig = (char*)rpmtdNextString(&sig_tag_data);
2772 : : idx = rpmfiNext(hdr_fi);
2773 : : }
2774 : : while (idx != -1 && 0 != strcmp(b_source1.c_str(), rpmfiFN(hdr_fi)));
2775 : : rpmfiFree(hdr_fi);
2776 : :
2777 : : if(sig && 0 != strlen(sig) && idx != -1)
2778 : : {
2779 : : if (verbose > 2) obatched(clog) << "Found IMA signature for " << b_source1 << ":\n" << sig << endl;
2780 : : ima_sig = sig;
2781 : : inc_metric("http_responses_total","extra","ima-sigs-extracted");
2782 : : }
2783 : : else
2784 : : {
2785 : : if (verbose > 2) obatched(clog) << "Could not find IMA signature for " << b_source1 << endl;
2786 : : }
2787 : :
2788 : : rpmtdFreeData (&sig_tag_data);
2789 : : headerFree(rpm_hdr);
2790 : : Fclose(rpm_fd);
2791 : : } while(false);
2792 : : #endif
2793 : :
2794 : : // check for a match in the fdcache first
2795 [ + - ]: 1000 : int fd = fdcache.lookup(b_source0, b_source1);
2796 [ + + ]: 1000 : while (fd >= 0) // got one!; NB: this is really an if() with a possible branch out to the end
2797 : : {
2798 : 100 : rc = fstat(fd, &fs);
2799 [ - + ]: 100 : if (rc < 0) // disappeared?
2800 : : {
2801 [ # # ]: 0 : if (verbose)
2802 [ # # # # : 0 : obatched(clog) << "cannot fstat fdcache " << b_source0 << endl;
# # ]
2803 [ # # ]: 0 : close(fd);
2804 [ # # ]: 0 : fdcache.clear(b_source0, b_source1);
2805 : : break; // branch out of if "loop", to try new libarchive fetch attempt
2806 : : }
2807 : :
2808 [ + - + - ]: 100 : struct MHD_Response* r = create_buildid_r_response (b_mtime, b_source0,
2809 : : b_source1, section,
2810 : : ima_sig, NULL, fd,
2811 : : fs.st_size,
2812 : : fs.st_mtime,
2813 : : "archive fdcache",
2814 : : extract_begin);
2815 [ + - ]: 100 : if (r == 0)
2816 : : break; // branch out of if "loop", to try new libarchive fetch attempt
2817 [ + - ]: 100 : if (result_fd)
2818 : 100 : *result_fd = fd;
2819 : : return r;
2820 : : // NB: see, we never go around the 'loop' more than once
2821 : : }
2822 : :
2823 : : // no match ... look for a seekable entry
2824 : 900 : bool populate_seekable = ! passive_p;
2825 : 900 : unique_ptr<sqlite_ps> pp (new sqlite_ps (internal_req_p ? db : dbq,
2826 : : "rpm-seekable-query",
2827 : : "select type, size, offset, mtime from " BUILDIDS "_r_seekable "
2828 [ + - + - : 900 : "where file = ? and content = ?"));
+ - + + +
- ]
2829 [ + - + - : 900 : rc = pp->reset().bind(1, b_id0).bind(2, b_id1).step();
+ - + - ]
2830 [ + + ]: 900 : if (rc != SQLITE_DONE)
2831 : : {
2832 [ - + ]: 202 : if (rc != SQLITE_ROW)
2833 [ # # # # ]: 0 : throw sqlite_exception(rc, "step");
2834 : : // if we found a match in _r_seekable but we fail to extract it, don't
2835 : : // bother populating it again
2836 : 202 : populate_seekable = false;
2837 [ + - ]: 202 : const char* seekable_type = (const char*) sqlite3_column_text (*pp, 0);
2838 [ + - - + ]: 202 : if (seekable_type != NULL && strcmp (seekable_type, "xz") == 0)
2839 : : {
2840 [ + - ]: 202 : int64_t seekable_size = sqlite3_column_int64 (*pp, 1);
2841 [ + - ]: 202 : int64_t seekable_offset = sqlite3_column_int64 (*pp, 2);
2842 [ + - ]: 202 : int64_t seekable_mtime = sqlite3_column_int64 (*pp, 3);
2843 : :
2844 : 202 : char* tmppath = NULL;
2845 [ - + ]: 202 : if (asprintf (&tmppath, "%s/debuginfod-fdcache.XXXXXX", tmpdir.c_str()) < 0)
2846 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate tmppath");
2847 : 202 : defer_dtor<void*,void> tmmpath_freer (tmppath, free);
2848 : :
2849 [ + - ]: 202 : fd = extract_from_seekable_archive (b_source0, tmppath,
2850 : : seekable_offset, seekable_size);
2851 [ + - ]: 202 : if (fd >= 0)
2852 : : {
2853 : : // Set the mtime so the fdcache file mtimes propagate to future webapi
2854 : : // clients.
2855 : 202 : struct timespec tvs[2];
2856 : 202 : tvs[0].tv_sec = 0;
2857 : 202 : tvs[0].tv_nsec = UTIME_OMIT;
2858 : 202 : tvs[1].tv_sec = seekable_mtime;
2859 : 202 : tvs[1].tv_nsec = 0;
2860 : 202 : (void) futimens (fd, tvs); /* best effort */
2861 [ + - + - ]: 202 : struct MHD_Response* r = create_buildid_r_response (b_mtime,
2862 : : b_source0,
2863 : : b_source1,
2864 : : section,
2865 : : ima_sig,
2866 : : tmppath, fd,
2867 : : seekable_size,
2868 : : seekable_mtime,
2869 : : "seekable xz archive",
2870 : : extract_begin);
2871 [ + - ]: 202 : if (r != 0 && result_fd)
2872 : 202 : *result_fd = fd;
2873 : 202 : return r;
2874 : : }
2875 : 202 : }
2876 : : }
2877 : 698 : pp.reset();
2878 : :
2879 : : // still no match ... grumble, must process the archive
2880 [ + - ]: 698 : string archive_decoder = "/dev/null";
2881 [ + - ]: 698 : string archive_extension = "";
2882 [ + + ]: 1942 : for (auto&& arch : scan_archives)
2883 [ + + ]: 1244 : if (string_endswith(b_source0, arch.first))
2884 : : {
2885 [ + - ]: 698 : archive_extension = arch.first;
2886 [ + - ]: 1942 : archive_decoder = arch.second;
2887 : : }
2888 : 698 : FILE* fp;
2889 : :
2890 : 698 : defer_dtor<FILE*,int>::dtor_fn dfn;
2891 [ + + ]: 698 : if (archive_decoder != "cat")
2892 : : {
2893 [ + - + - : 1056 : string popen_cmd = archive_decoder + " " + shell_escape(b_source0);
+ - ]
2894 [ + - ]: 528 : fp = popen (popen_cmd.c_str(), "r"); // "e" O_CLOEXEC?
2895 : 528 : dfn = pclose;
2896 [ - + ]: 528 : if (fp == NULL)
2897 [ # # # # : 0 : throw libc_exception (errno, string("popen ") + popen_cmd);
# # ]
2898 : 528 : }
2899 : : else
2900 : : {
2901 [ + - ]: 170 : fp = fopen (b_source0.c_str(), "r");
2902 : 170 : dfn = fclose;
2903 [ - + ]: 170 : if (fp == NULL)
2904 [ # # # # : 0 : throw libc_exception (errno, string("fopen ") + b_source0);
# # ]
2905 : : }
2906 : 698 : defer_dtor<FILE*,int> fp_closer (fp, dfn);
2907 : :
2908 : 698 : struct archive *a;
2909 [ + - ]: 698 : a = archive_read_new();
2910 [ - + ]: 698 : if (a == NULL)
2911 [ # # # # ]: 0 : throw archive_exception("cannot create archive reader");
2912 : 698 : defer_dtor<struct archive*,int> archive_closer (a, archive_read_free);
2913 : :
2914 [ + - ]: 698 : rc = archive_read_support_format_all(a);
2915 [ - + ]: 698 : if (rc != ARCHIVE_OK)
2916 [ # # # # ]: 0 : throw archive_exception(a, "cannot select all format");
2917 [ + - ]: 698 : rc = archive_read_support_filter_all(a);
2918 [ - + ]: 698 : if (rc != ARCHIVE_OK)
2919 [ # # # # ]: 0 : throw archive_exception(a, "cannot select all filters");
2920 : :
2921 [ + - ]: 698 : rc = archive_read_open_FILE (a, fp);
2922 [ - + ]: 698 : if (rc != ARCHIVE_OK)
2923 : : {
2924 [ # # # # : 0 : obatched(clog) << "cannot open archive from pipe " << b_source0 << endl;
# # ]
2925 [ # # # # ]: 0 : throw archive_exception(a, "cannot open archive from pipe");
2926 : : }
2927 : :
2928 : : // If the archive was scanned in a version without _r_seekable, then we may
2929 : : // need to populate _r_seekable now. This can be removed the next time
2930 : : // BUILDIDS is updated.
2931 [ + + ]: 698 : if (populate_seekable)
2932 : : {
2933 [ + - ]: 696 : populate_seekable = is_seekable_archive (b_source0, a);
2934 [ + - ]: 696 : if (populate_seekable)
2935 : : {
2936 : : // NB: the names are already interned
2937 : 0 : pp.reset(new sqlite_ps (db, "rpm-seekable-insert2",
2938 : : "insert or ignore into " BUILDIDS "_r_seekable (file, content, type, size, offset, mtime) "
2939 : : "values (?, "
2940 : : "(select id from " BUILDIDS "_files "
2941 : : "where dirname = (select id from " BUILDIDS "_fileparts where name = ?) "
2942 : : "and basename = (select id from " BUILDIDS "_fileparts where name = ?) "
2943 [ # # # # : 0 : "), 'xz', ?, ?, ?)"));
# # # # ]
2944 : : }
2945 : : }
2946 : :
2947 : : // archive traversal is in five stages:
2948 : : // 1) before we find a matching entry, insert it into _r_seekable if needed or
2949 : : // skip it otherwise
2950 : : // 2) extract the matching entry (set r = result). Also insert it into
2951 : : // _r_seekable if needed
2952 : : // 3) extract some number of prefetched entries (just into fdcache). Also
2953 : : // insert them into _r_seekable if needed
2954 : : // 4) if needed, insert all of the remaining entries into _r_seekable
2955 : : // 5) abort any further processing
2956 : 698 : struct MHD_Response* r = 0; // will set in stage 2
2957 [ + + ]: 698 : unsigned prefetch_count =
2958 : : internal_req_p ? 0 : fdcache_prefetch; // will decrement in stage 3
2959 : :
2960 [ + + - + ]: 10822 : while(r == 0 || prefetch_count > 0 || populate_seekable) // stage 1-4
2961 : : {
2962 [ + - ]: 10782 : if (interrupted)
2963 : : break;
2964 : :
2965 : 10782 : struct archive_entry *e;
2966 [ + - ]: 10782 : rc = archive_read_next_header (a, &e);
2967 [ + + ]: 10782 : if (rc != ARCHIVE_OK)
2968 : : break;
2969 : :
2970 [ + - + + ]: 10126 : if (! S_ISREG(archive_entry_mode (e))) // skip non-files completely
2971 : 9428 : continue;
2972 : :
2973 [ + - ]: 2926 : string fn = canonicalized_archive_entry_pathname (e);
2974 : :
2975 [ - + ]: 2926 : if (populate_seekable)
2976 : : {
2977 : 0 : string dn, bn;
2978 : 0 : size_t slash = fn.rfind('/');
2979 [ # # ]: 0 : if (slash == std::string::npos) {
2980 [ # # ]: 0 : dn = "";
2981 [ # # ]: 0 : bn = fn;
2982 : : } else {
2983 [ # # ]: 0 : dn = fn.substr(0, slash);
2984 [ # # ]: 0 : bn = fn.substr(slash + 1);
2985 : : }
2986 : :
2987 [ # # ]: 0 : int64_t seekable_size = archive_entry_size (e);
2988 [ # # ]: 0 : int64_t seekable_offset = archive_filter_bytes (a, 0);
2989 [ # # ]: 0 : time_t seekable_mtime = archive_entry_mtime (e);
2990 : :
2991 [ # # ]: 0 : pp->reset();
2992 [ # # ]: 0 : pp->bind(1, b_id0);
2993 [ # # ]: 0 : pp->bind(2, dn);
2994 [ # # ]: 0 : pp->bind(3, bn);
2995 [ # # ]: 0 : pp->bind(4, seekable_size);
2996 [ # # ]: 0 : pp->bind(5, seekable_offset);
2997 [ # # ]: 0 : pp->bind(6, seekable_mtime);
2998 [ # # ]: 0 : rc = pp->step();
2999 [ # # ]: 0 : if (rc != SQLITE_DONE)
3000 [ # # # # ]: 0 : obatched(clog) << "recording seekable file=" << fn
3001 [ # # # # : 0 : << " sqlite3 error: " << (sqlite3_errstr(rc) ?: "?") << endl;
# # # # #
# ]
3002 [ # # ]: 0 : else if (verbose > 2)
3003 [ # # # # ]: 0 : obatched(clog) << "recorded seekable file=" << fn
3004 [ # # # # ]: 0 : << " size=" << seekable_size
3005 [ # # # # ]: 0 : << " offset=" << seekable_offset
3006 [ # # # # : 0 : << " mtime=" << seekable_mtime << endl;
# # ]
3007 [ # # ]: 0 : if (r != 0 && prefetch_count == 0) // stage 4
3008 : 0 : continue;
3009 : 0 : }
3010 : :
3011 [ + + + + ]: 2926 : if ((r == 0) && (fn != b_source1)) // stage 1
3012 : 1632 : continue;
3013 : :
3014 [ + - - + ]: 1294 : if (fdcache.probe (b_source0, fn) && // skip if already interned
3015 [ # # ]: 0 : fn != b_source1) // but only if we'd just be prefetching, PR29474
3016 : 0 : continue;
3017 : :
3018 : : // extract this file to a temporary file
3019 : 1294 : char* tmppath = NULL;
3020 : 1294 : rc = asprintf (&tmppath, "%s/debuginfod-fdcache.XXXXXX", tmpdir.c_str());
3021 [ - + ]: 1294 : if (rc < 0)
3022 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate tmppath");
3023 : 1294 : defer_dtor<void*,void> tmmpath_freer (tmppath, free);
3024 [ + - ]: 1294 : fd = mkstemp (tmppath);
3025 [ - + ]: 1294 : if (fd < 0)
3026 [ # # # # ]: 0 : throw libc_exception (errno, "cannot create temporary file");
3027 : : // NB: don't unlink (tmppath), as fdcache will take charge of it.
3028 : :
3029 : : // NB: this can take many uninterruptible seconds for a huge file
3030 [ + - ]: 1294 : rc = archive_read_data_into_fd (a, fd);
3031 [ - + ]: 1294 : if (rc != ARCHIVE_OK) // e.g. ENOSPC!
3032 : : {
3033 [ # # ]: 0 : close (fd);
3034 : 0 : unlink (tmppath);
3035 [ # # # # ]: 0 : throw archive_exception(a, b_source0, "cannot extract file");
3036 : : }
3037 : :
3038 : : // Set the mtime so the fdcache file mtimes, even prefetched ones,
3039 : : // propagate to future webapi clients.
3040 : 1294 : struct timespec tvs[2];
3041 : 1294 : tvs[0].tv_sec = 0;
3042 : 1294 : tvs[0].tv_nsec = UTIME_OMIT;
3043 [ + - ]: 1294 : tvs[1].tv_sec = archive_entry_mtime(e);
3044 [ + - ]: 1294 : tvs[1].tv_nsec = archive_entry_mtime_nsec(e);
3045 : 1294 : (void) futimens (fd, tvs); /* best effort */
3046 : :
3047 [ + + ]: 1294 : if (r != 0) // stage 3
3048 : : {
3049 : 596 : struct timespec extract_end;
3050 : 596 : clock_gettime (CLOCK_MONOTONIC, &extract_end);
3051 : 596 : double extract_time = (extract_end.tv_sec - extract_begin.tv_sec)
3052 : 596 : + (extract_end.tv_nsec - extract_begin.tv_nsec)/1.e9;
3053 : : // NB: now we know we have a complete reusable file; make fdcache
3054 : : // responsible for unlinking it later.
3055 [ + - + - : 596 : fdcache.intern(b_source0, fn,
+ - ]
3056 : : tmppath, archive_entry_size(e),
3057 : : false, extract_time); // prefetched ones go to the prefetch cache
3058 : 596 : prefetch_count --;
3059 [ + - ]: 596 : close (fd); // we're not saving this fd to make a mhd-response from!
3060 : 596 : continue;
3061 : 596 : }
3062 : :
3063 [ + - + - : 698 : r = create_buildid_r_response (b_mtime, b_source0, b_source1, section,
+ - ]
3064 : : ima_sig, tmppath, fd,
3065 : : archive_entry_size(e),
3066 : : archive_entry_mtime(e),
3067 [ + - ]: 698 : archive_extension + " archive",
3068 : : extract_begin);
3069 [ + + ]: 698 : if (r == 0)
3070 : : break; // assume no chance of better luck around another iteration; no other copies of same file
3071 [ + - ]: 696 : if (result_fd)
3072 : 696 : *result_fd = fd;
3073 : 3522 : }
3074 : :
3075 : : // XXX: rpm/file not found: delete this R entry?
3076 : 698 : return r;
3077 : 1900 : }
3078 : :
3079 : : void
3080 : 650 : add_client_federation_headers(debuginfod_client *client, MHD_Connection* conn){
3081 : : // Transcribe incoming User-Agent:
3082 [ - + ]: 650 : string ua = MHD_lookup_connection_value (conn, MHD_HEADER_KIND, "User-Agent") ?: "";
3083 [ + - + - ]: 654 : string ua_complete = string("User-Agent: ") + ua;
3084 [ + - ]: 650 : debuginfod_add_http_header (client, ua_complete.c_str());
3085 : :
3086 : : // Compute larger XFF:, for avoiding info loss during
3087 : : // federation, and for future cyclicity detection.
3088 [ + - + + : 1278 : string xff = MHD_lookup_connection_value (conn, MHD_HEADER_KIND, "X-Forwarded-For") ?: "";
+ - ]
3089 [ + + ]: 650 : if (xff != "")
3090 [ + - ]: 52 : xff += string(", "); // comma separated list
3091 : :
3092 : 650 : unsigned int xff_count = 0;
3093 [ + + ]: 978 : for (auto&& i : xff){
3094 [ + + ]: 328 : if (i == ',') xff_count++;
3095 : : }
3096 : :
3097 : : // if X-Forwarded-For: exceeds N hops,
3098 : : // do not delegate a local lookup miss to upstream debuginfods.
3099 [ + + ]: 650 : if (xff_count >= forwarded_ttl_limit)
3100 : 4 : throw reportable_exception(MHD_HTTP_NOT_FOUND, "not found, --forwared-ttl-limit reached \
3101 [ + - ]: 8 : and will not query the upstream servers");
3102 : :
3103 : : // Compute the client's numeric IP address only - so can't merge with conninfo()
3104 [ + - ]: 646 : const union MHD_ConnectionInfo *u = MHD_get_connection_info (conn,
3105 : : MHD_CONNECTION_INFO_CLIENT_ADDRESS);
3106 [ + - ]: 646 : struct sockaddr *so = u ? u->client_addr : 0;
3107 : 646 : char hostname[256] = ""; // RFC1035
3108 [ + - - + ]: 646 : if (so && so->sa_family == AF_INET) {
3109 [ # # ]: 0 : (void) getnameinfo (so, sizeof (struct sockaddr_in), hostname, sizeof (hostname), NULL, 0,
3110 : : NI_NUMERICHOST);
3111 [ + - ]: 646 : } else if (so && so->sa_family == AF_INET6) {
3112 : 646 : struct sockaddr_in6* addr6 = (struct sockaddr_in6*) so;
3113 [ + - + - : 646 : if (IN6_IS_ADDR_V4MAPPED(&addr6->sin6_addr)) {
- + ]
3114 : 646 : struct sockaddr_in addr4;
3115 [ + - ]: 646 : memset (&addr4, 0, sizeof(addr4));
3116 : 646 : addr4.sin_family = AF_INET;
3117 : 646 : addr4.sin_port = addr6->sin6_port;
3118 [ + - ]: 646 : memcpy (&addr4.sin_addr.s_addr, addr6->sin6_addr.s6_addr+12, sizeof(addr4.sin_addr.s_addr));
3119 [ + - ]: 646 : (void) getnameinfo ((struct sockaddr*) &addr4, sizeof (addr4),
3120 : : hostname, sizeof (hostname), NULL, 0,
3121 : : NI_NUMERICHOST);
3122 : : } else {
3123 [ # # ]: 0 : (void) getnameinfo (so, sizeof (struct sockaddr_in6), hostname, sizeof (hostname), NULL, 0,
3124 : : NI_NUMERICHOST);
3125 : : }
3126 : : }
3127 : :
3128 [ + - + - : 1296 : string xff_complete = string("X-Forwarded-For: ")+xff+string(hostname);
+ - + - ]
3129 [ + - ]: 646 : debuginfod_add_http_header (client, xff_complete.c_str());
3130 : 654 : }
3131 : :
3132 : : static struct MHD_Response*
3133 : 2210 : handle_buildid_match (bool internal_req_p,
3134 : : int64_t b_mtime,
3135 : : const string& b_stype,
3136 : : const string& b_source0,
3137 : : const string& b_source1,
3138 : : int64_t b_id0,
3139 : : int64_t b_id1,
3140 : : const string& section,
3141 : : int *result_fd)
3142 : : {
3143 : 2210 : try
3144 : : {
3145 [ + + ]: 2210 : if (b_stype == "F")
3146 [ + - ]: 1152 : return handle_buildid_f_match(internal_req_p, b_mtime, b_source0,
3147 : : section, result_fd);
3148 [ + - ]: 1058 : else if (b_stype == "R")
3149 [ + + ]: 1058 : return handle_buildid_r_match(internal_req_p, b_mtime, b_source0,
3150 : : b_source1, b_id0, b_id1, section,
3151 : : result_fd);
3152 : : }
3153 [ - + ]: 58 : catch (const reportable_exception &e)
3154 : : {
3155 [ + - ]: 58 : e.report(clog);
3156 : : // Report but swallow libc etc. errors here; let the caller
3157 : : // iterate to other matches of the content.
3158 : 58 : }
3159 : :
3160 : : return 0;
3161 : : }
3162 : :
3163 : :
3164 : : static int
3165 : 4 : debuginfod_find_progress (debuginfod_client *, long a, long b)
3166 : : {
3167 [ - + ]: 4 : if (verbose > 4)
3168 [ # # # # : 0 : obatched(clog) << "federated debuginfod progress=" << a << "/" << b << endl;
# # # # ]
3169 : :
3170 : 4 : return interrupted;
3171 : : }
3172 : :
3173 : :
3174 : : // a little lru pool of debuginfod_client*s for reuse between query threads
3175 : :
3176 : : mutex dc_pool_lock;
3177 : : deque<debuginfod_client*> dc_pool;
3178 : :
3179 : 670 : debuginfod_client* debuginfod_pool_begin()
3180 : : {
3181 : 670 : unique_lock<mutex> lock(dc_pool_lock);
3182 [ + + ]: 670 : if (dc_pool.size() > 0)
3183 : : {
3184 [ + - + - : 1276 : inc_metric("dc_pool_op_count","op","begin-reuse");
+ - + - ]
3185 : 638 : debuginfod_client *c = dc_pool.front();
3186 : 638 : dc_pool.pop_front();
3187 : 638 : return c;
3188 : : }
3189 [ + - + - : 64 : inc_metric("dc_pool_op_count","op","begin-new");
+ - + - ]
3190 [ + - ]: 32 : return debuginfod_begin();
3191 : 670 : }
3192 : :
3193 : :
3194 : 162 : void debuginfod_pool_groom()
3195 : : {
3196 : 162 : unique_lock<mutex> lock(dc_pool_lock);
3197 [ + + ]: 194 : while (dc_pool.size() > 0)
3198 : : {
3199 [ + - + - : 64 : inc_metric("dc_pool_op_count","op","end");
+ - + - ]
3200 [ + - ]: 32 : debuginfod_end(dc_pool.front());
3201 : 32 : dc_pool.pop_front();
3202 : : }
3203 : 162 : }
3204 : :
3205 : :
3206 : 670 : void debuginfod_pool_end(debuginfod_client* c)
3207 : : {
3208 : 670 : unique_lock<mutex> lock(dc_pool_lock);
3209 [ + - + - : 1340 : inc_metric("dc_pool_op_count","op","end-save");
+ - + - ]
3210 [ + - ]: 670 : dc_pool.push_front(c); // accelerate reuse, vs. push_back
3211 : 670 : }
3212 : :
3213 : :
3214 : : static struct MHD_Response*
3215 : 2795 : handle_buildid (MHD_Connection* conn,
3216 : : const string& buildid /* unsafe */,
3217 : : string& artifacttype /* unsafe, cleanse on exception/return */,
3218 : : const string& suffix /* unsafe */,
3219 : : int *result_fd)
3220 : : {
3221 : : // validate artifacttype
3222 [ + + ]: 2795 : string atype_code;
3223 [ + + + - ]: 2795 : if (artifacttype == "debuginfo") atype_code = "D";
3224 [ + + + - ]: 1827 : else if (artifacttype == "executable") atype_code = "E";
3225 [ + + + - ]: 1138 : else if (artifacttype == "source") atype_code = "S";
3226 [ + + + - ]: 12 : else if (artifacttype == "section") atype_code = "I";
3227 : : else {
3228 [ + - ]: 4 : artifacttype = "invalid"; // PR28242 ensure http_resposes metrics don't propagate unclean user data
3229 [ + - ]: 8 : throw reportable_exception("invalid artifacttype");
3230 : : }
3231 : :
3232 [ + + ]: 2791 : if (conn != 0)
3233 [ + - + - : 5825 : inc_metric("http_requests_total", "type", artifacttype);
+ - ]
3234 : :
3235 [ + + ]: 2791 : string section;
3236 [ + + ]: 2791 : if (atype_code == "I")
3237 : : {
3238 [ - + ]: 8 : if (suffix.size () < 2)
3239 [ # # ]: 0 : throw reportable_exception ("invalid section suffix");
3240 : :
3241 : : // Remove leading '/'
3242 [ + - ]: 8 : section = suffix.substr(1);
3243 : : }
3244 : :
3245 [ + + - + ]: 2791 : if (atype_code == "S" && suffix == "")
3246 [ # # ]: 0 : throw reportable_exception("invalid source suffix");
3247 : :
3248 : : // validate buildid
3249 [ + + ]: 2791 : if ((buildid.size() < 2) || // not empty
3250 [ + + + - : 5579 : (buildid.size() % 2) || // even number
+ - ]
3251 : 2788 : (buildid.find_first_not_of("0123456789abcdef") != string::npos)) // pure tasty lowercase hex
3252 [ + - ]: 6 : throw reportable_exception("invalid buildid");
3253 : :
3254 [ + - ]: 2788 : if (verbose > 1)
3255 [ + - + - ]: 8364 : obatched(clog) << "searching for buildid=" << buildid << " artifacttype=" << artifacttype
3256 [ + - + - : 2788 : << " suffix=" << suffix << endl;
+ - + - +
- ]
3257 : :
3258 : : // If invoked from the scanner threads, use the scanners' read-write
3259 : : // connection. Otherwise use the web query threads' read-only connection.
3260 [ + + ]: 2788 : sqlite3 *thisdb = (conn == 0) ? db : dbq;
3261 : :
3262 : 2788 : sqlite_ps *pp = 0;
3263 : :
3264 [ + + ]: 2788 : if (atype_code == "D")
3265 : : {
3266 : 968 : pp = new sqlite_ps (thisdb, "mhd-query-d",
3267 : : "select mtime, sourcetype, source0, source1, id0, id1 from " BUILDIDS "_query_d2 where buildid = ? "
3268 [ + - + - : 1936 : "order by mtime desc");
+ - + - ]
3269 [ + - ]: 968 : pp->reset();
3270 [ + - ]: 968 : pp->bind(1, buildid);
3271 : : }
3272 [ + + ]: 1820 : else if (atype_code == "E")
3273 : : {
3274 : 686 : pp = new sqlite_ps (thisdb, "mhd-query-e",
3275 : : "select mtime, sourcetype, source0, source1, id0, id1 from " BUILDIDS "_query_e2 where buildid = ? "
3276 [ + - + - : 1372 : "order by mtime desc");
+ - + - ]
3277 [ + - ]: 686 : pp->reset();
3278 [ + - ]: 686 : pp->bind(1, buildid);
3279 : : }
3280 [ + + ]: 1134 : else if (atype_code == "S")
3281 : : {
3282 : : // PR25548
3283 : : // Incoming source queries may come in with either dwarf-level OR canonicalized paths.
3284 : : // We let the query pass with either one.
3285 : :
3286 : 1126 : pp = new sqlite_ps (thisdb, "mhd-query-s",
3287 : : "select mtime, sourcetype, source0, source1 from " BUILDIDS "_query_s where buildid = ? and artifactsrc in (?,?) "
3288 [ + - + - : 2252 : "order by sharedprefix(source0,source0ref) desc, mtime desc");
+ - + - ]
3289 [ + - ]: 1126 : pp->reset();
3290 [ + - ]: 1126 : pp->bind(1, buildid);
3291 : : // NB: we don't store the non-canonicalized path names any more, but old databases
3292 : : // might have them (and no canon ones), so we keep searching for both.
3293 [ + - ]: 1126 : pp->bind(2, suffix);
3294 [ + - + - ]: 2891 : pp->bind(3, canon_pathname(suffix));
3295 : : }
3296 [ + - ]: 8 : else if (atype_code == "I")
3297 : : {
3298 : 8 : pp = new sqlite_ps (thisdb, "mhd-query-i",
3299 : : "select mtime, sourcetype, source0, source1, 1 as debug_p from " BUILDIDS "_query_d2 where buildid = ? "
3300 : : "union all "
3301 : : "select mtime, sourcetype, source0, source1, 0 as debug_p from " BUILDIDS "_query_e2 where buildid = ? "
3302 [ + - + - : 16 : "order by debug_p desc, mtime desc");
+ - + - ]
3303 [ + - ]: 8 : pp->reset();
3304 [ + - ]: 8 : pp->bind(1, buildid);
3305 [ + - ]: 8 : pp->bind(2, buildid);
3306 : : }
3307 : 2788 : unique_ptr<sqlite_ps> ps_closer(pp); // release pp if exception or return
3308 : :
3309 : 2788 : bool do_upstream_section_query = true;
3310 : :
3311 : : // consume all the rows
3312 : 2912 : while (1)
3313 : : {
3314 [ + - ]: 2850 : int rc = pp->step();
3315 [ + + ]: 2850 : if (rc == SQLITE_DONE) break;
3316 [ - + ]: 2210 : if (rc != SQLITE_ROW)
3317 [ # # # # ]: 0 : throw sqlite_exception(rc, "step");
3318 : :
3319 [ + - ]: 2210 : int64_t b_mtime = sqlite3_column_int64 (*pp, 0);
3320 [ + - - + : 2210 : string b_stype = string((const char*) sqlite3_column_text (*pp, 1) ?: ""); /* by DDL may not be NULL */
+ - ]
3321 [ + - - + : 2210 : string b_source0 = string((const char*) sqlite3_column_text (*pp, 2) ?: ""); /* may be NULL */
+ - ]
3322 [ + - + + : 3362 : string b_source1 = string((const char*) sqlite3_column_text (*pp, 3) ?: ""); /* may be NULL */
+ - ]
3323 : 2210 : int64_t b_id0 = 0, b_id1 = 0;
3324 [ + + + + ]: 2210 : if (atype_code == "D" || atype_code == "E")
3325 : : {
3326 [ + - ]: 1070 : b_id0 = sqlite3_column_int64 (*pp, 4);
3327 [ + - ]: 1070 : b_id1 = sqlite3_column_int64 (*pp, 5);
3328 : : }
3329 : :
3330 [ + - ]: 2210 : if (verbose > 1)
3331 [ + - + - ]: 6630 : obatched(clog) << "found mtime=" << b_mtime << " stype=" << b_stype
3332 [ + - + - : 2210 : << " source0=" << b_source0 << " source1=" << b_source1 << endl;
+ - + - +
- + - +
- ]
3333 : :
3334 : : // Try accessing the located match.
3335 : : // XXX: in case of multiple matches, attempt them in parallel?
3336 [ + - ]: 2210 : auto r = handle_buildid_match (conn ? false : true,
3337 : : b_mtime, b_stype, b_source0, b_source1,
3338 : : b_id0, b_id1, section, result_fd);
3339 [ + + ]: 2210 : if (r)
3340 : 2148 : return r;
3341 : :
3342 : : // If a debuginfo file matching BUILDID was found but didn't contain
3343 : : // the desired section, then the section should not exist. Don't
3344 : : // bother querying upstream servers.
3345 [ + + + - : 62 : if (!section.empty () && (sqlite3_column_int (*pp, 4) == 1))
- + ]
3346 : : {
3347 : 4 : struct stat st;
3348 : :
3349 : : // For "F" sourcetype, check if the debuginfo exists. For "R"
3350 : : // sourcetype, check if the debuginfo was interned into the fdcache.
3351 [ - + ]: 2 : if ((b_stype == "F" && (stat (b_source0.c_str (), &st) == 0))
3352 [ + + + - : 4 : || (b_stype == "R" && fdcache.probe (b_source0, b_source1)))
+ - + - ]
3353 : : do_upstream_section_query = false;
3354 : : }
3355 : 2210 : }
3356 [ + - ]: 640 : pp->reset();
3357 : :
3358 [ - + ]: 640 : if (!do_upstream_section_query)
3359 [ # # ]: 0 : throw reportable_exception(MHD_HTTP_NOT_FOUND, "not found");
3360 : :
3361 : : // We couldn't find it in the database. Last ditch effort
3362 : : // is to defer to other debuginfo servers.
3363 : :
3364 : 640 : int fd = -1;
3365 [ + - ]: 640 : debuginfod_client *client = debuginfod_pool_begin ();
3366 [ - + ]: 640 : if (client == NULL)
3367 [ # # # # ]: 0 : throw libc_exception(errno, "debuginfod client pool alloc");
3368 : 640 : defer_dtor<debuginfod_client*,void> client_closer (client, debuginfod_pool_end);
3369 : :
3370 [ + - ]: 640 : debuginfod_set_progressfn (client, & debuginfod_find_progress);
3371 : :
3372 [ + + ]: 640 : if (conn)
3373 [ + + ]: 620 : add_client_federation_headers(client, conn);
3374 : :
3375 [ + + ]: 636 : if (artifacttype == "debuginfo")
3376 [ + - ]: 88 : fd = debuginfod_find_debuginfo (client,
3377 [ + - ]: 88 : (const unsigned char*) buildid.c_str(),
3378 : : 0, NULL);
3379 [ + + ]: 548 : else if (artifacttype == "executable")
3380 [ + - ]: 546 : fd = debuginfod_find_executable (client,
3381 [ + - ]: 546 : (const unsigned char*) buildid.c_str(),
3382 : : 0, NULL);
3383 [ + - ]: 2 : else if (artifacttype == "source")
3384 [ + - ]: 2 : fd = debuginfod_find_source (client,
3385 [ + - ]: 2 : (const unsigned char*) buildid.c_str(),
3386 : : 0, suffix.c_str(), NULL);
3387 [ # # ]: 0 : else if (artifacttype == "section")
3388 [ # # ]: 0 : fd = debuginfod_find_section (client,
3389 [ # # ]: 0 : (const unsigned char*) buildid.c_str(),
3390 : : 0, section.c_str(), NULL);
3391 : :
3392 [ + + ]: 636 : if (fd >= 0)
3393 : : {
3394 [ + - ]: 4 : if (conn != 0)
3395 [ + - + - : 644 : inc_metric ("http_responses_total","result","upstream");
+ - + - ]
3396 : 4 : struct stat s;
3397 : 4 : int rc = fstat (fd, &s);
3398 [ + - ]: 4 : if (rc == 0)
3399 : : {
3400 [ + - ]: 4 : auto r = MHD_create_response_from_fd ((uint64_t) s.st_size, fd);
3401 [ + - ]: 4 : if (r)
3402 : : {
3403 [ + - ]: 4 : add_mhd_response_header (r, "Content-Type",
3404 : : "application/octet-stream");
3405 : : // Copy the incoming headers
3406 [ + - ]: 4 : const char * hdrs = debuginfod_get_headers(client);
3407 [ + - ]: 4 : string header_dup;
3408 [ + - ]: 4 : if (hdrs)
3409 [ + - ]: 4 : header_dup = string(hdrs);
3410 : : // Parse the "header: value\n" lines into (h,v) tuples and pass on
3411 : 20 : while(1)
3412 : : {
3413 : 12 : size_t newline = header_dup.find('\n');
3414 [ + + ]: 12 : if (newline == string::npos) break;
3415 : 8 : size_t colon = header_dup.find(':');
3416 [ + - ]: 8 : if (colon == string::npos) break;
3417 [ + - ]: 8 : string header = header_dup.substr(0,colon);
3418 [ + - ]: 8 : string value = header_dup.substr(colon+1,newline-colon-1);
3419 : : // strip leading spaces from value
3420 : 8 : size_t nonspace = value.find_first_not_of(" ");
3421 [ + - ]: 8 : if (nonspace != string::npos)
3422 [ + - ]: 8 : value = value.substr(nonspace);
3423 [ + - ]: 8 : add_mhd_response_header(r, header.c_str(), value.c_str());
3424 [ + - ]: 8 : header_dup = header_dup.substr(newline+1);
3425 : 8 : }
3426 : :
3427 [ + - ]: 4 : add_mhd_last_modified (r, s.st_mtime);
3428 [ + - ]: 4 : if (verbose > 1)
3429 [ + - + - ]: 8 : obatched(clog) << "serving file from upstream debuginfod/cache" << endl;
3430 [ + - ]: 4 : if (result_fd)
3431 : 4 : *result_fd = fd;
3432 : 4 : return r; // NB: don't close fd; libmicrohttpd will
3433 : 0 : }
3434 : : }
3435 [ # # ]: 0 : close (fd);
3436 : : }
3437 : : else
3438 [ + + ]: 632 : switch(fd)
3439 : : {
3440 : : case -ENOSYS:
3441 : : break;
3442 : : case -ENOENT:
3443 : : break;
3444 : 532 : default: // some more tricky error
3445 [ + - + - ]: 1064 : throw libc_exception(-fd, "upstream debuginfod query failed");
3446 : : }
3447 : :
3448 [ + - ]: 200 : throw reportable_exception(MHD_HTTP_NOT_FOUND, "not found");
3449 : 3427 : }
3450 : :
3451 : :
3452 : : ////////////////////////////////////////////////////////////////////////
3453 : :
3454 : : static map<string,double> metrics; // arbitrary data for /metrics query
3455 : : // NB: store int64_t since all our metrics are integers; prometheus accepts double
3456 : : static mutex metrics_lock;
3457 : : // NB: these objects get released during the process exit via global dtors
3458 : : // do not call them from within other global dtors
3459 : :
3460 : : // utility function for assembling prometheus-compatible
3461 : : // name="escaped-value" strings
3462 : : // https://prometheus.io/docs/instrumenting/exposition_formats/
3463 : : static string
3464 : 825507 : metric_label(const string& name, const string& value)
3465 : : {
3466 : 825507 : string x = name + "=\"";
3467 [ + + ]: 14199056 : for (auto&& c : value)
3468 [ - - - + ]: 13374429 : switch(c)
3469 : : {
3470 [ # # ]: 0 : case '\\': x += "\\\\"; break;
3471 [ # # ]: 0 : case '\"': x += "\\\""; break;
3472 [ # # ]: 0 : case '\n': x += "\\n"; break;
3473 [ + - ]: 26748332 : default: x += c; break;
3474 : : }
3475 [ + - ]: 824627 : x += "\"";
3476 : 824811 : return x;
3477 : 0 : }
3478 : :
3479 : :
3480 : : // add prometheus-format metric name + label tuple (if any) + value
3481 : :
3482 : : static void
3483 : 12192 : set_metric(const string& metric, double value)
3484 : : {
3485 : 12192 : unique_lock<mutex> lock(metrics_lock);
3486 [ + - ]: 12192 : metrics[metric] = value;
3487 : 12192 : }
3488 : : static void
3489 : 102 : inc_metric(const string& metric)
3490 : : {
3491 : 102 : unique_lock<mutex> lock(metrics_lock);
3492 [ + - ]: 102 : metrics[metric] ++;
3493 : 102 : }
3494 : : static void
3495 : 8307 : set_metric(const string& metric,
3496 : : const string& lname, const string& lvalue,
3497 : : double value)
3498 : : {
3499 [ + - + - ]: 16617 : string key = (metric + "{" + metric_label(lname, lvalue) + "}");
3500 [ + - ]: 8310 : unique_lock<mutex> lock(metrics_lock);
3501 [ + - ]: 8310 : metrics[key] = value;
3502 : 8310 : }
3503 : :
3504 : : static void
3505 : 398442 : inc_metric(const string& metric,
3506 : : const string& lname, const string& lvalue)
3507 : : {
3508 [ + - + - ]: 796878 : string key = (metric + "{" + metric_label(lname, lvalue) + "}");
3509 [ + - ]: 398438 : unique_lock<mutex> lock(metrics_lock);
3510 [ + - ]: 398506 : metrics[key] ++;
3511 : 398481 : }
3512 : : static void
3513 : 379147 : add_metric(const string& metric,
3514 : : const string& lname, const string& lvalue,
3515 : : double value)
3516 : : {
3517 [ + - + - ]: 758374 : string key = (metric + "{" + metric_label(lname, lvalue) + "}");
3518 [ + - ]: 379254 : unique_lock<mutex> lock(metrics_lock);
3519 [ + - ]: 379520 : metrics[key] += value;
3520 : 379513 : }
3521 : : static void
3522 : 102 : add_metric(const string& metric,
3523 : : double value)
3524 : : {
3525 : 102 : unique_lock<mutex> lock(metrics_lock);
3526 [ + - ]: 102 : metrics[metric] += value;
3527 : 102 : }
3528 : :
3529 : :
3530 : : // and more for higher arity labels if needed
3531 : :
3532 : : static void
3533 : 10149 : inc_metric(const string& metric,
3534 : : const string& lname, const string& lvalue,
3535 : : const string& rname, const string& rvalue)
3536 : : {
3537 [ + - ]: 20298 : string key = (metric + "{"
3538 [ + - + - ]: 40596 : + metric_label(lname, lvalue) + ","
3539 [ + - ]: 30447 : + metric_label(rname, rvalue) + "}");
3540 [ + - ]: 10149 : unique_lock<mutex> lock(metrics_lock);
3541 [ + - ]: 10149 : metrics[key] ++;
3542 : 10149 : }
3543 : : static void
3544 : 10149 : add_metric(const string& metric,
3545 : : const string& lname, const string& lvalue,
3546 : : const string& rname, const string& rvalue,
3547 : : double value)
3548 : : {
3549 [ + - ]: 20298 : string key = (metric + "{"
3550 [ + - + - ]: 40596 : + metric_label(lname, lvalue) + ","
3551 [ + - ]: 30447 : + metric_label(rname, rvalue) + "}");
3552 [ + - ]: 10149 : unique_lock<mutex> lock(metrics_lock);
3553 [ + - ]: 10149 : metrics[key] += value;
3554 : 10149 : }
3555 : :
3556 : : static struct MHD_Response*
3557 : 748 : handle_metrics (off_t* size)
3558 : : {
3559 : 748 : stringstream o;
3560 : 748 : {
3561 [ + - ]: 748 : unique_lock<mutex> lock(metrics_lock);
3562 [ + + ]: 79517 : for (auto&& i : metrics)
3563 [ + - ]: 78769 : o << i.first
3564 : : << " "
3565 [ + - + - ]: 78769 : << std::setprecision(std::numeric_limits<double>::digits10 + 1)
3566 [ + - + - ]: 78769 : << i.second
3567 : 78769 : << endl;
3568 : 0 : }
3569 [ + - ]: 748 : const string& os = o.str();
3570 [ + - ]: 748 : MHD_Response* r = MHD_create_response_from_buffer (os.size(),
3571 [ + - ]: 748 : (void*) os.c_str(),
3572 : : MHD_RESPMEM_MUST_COPY);
3573 [ + - ]: 748 : if (r != NULL)
3574 : : {
3575 [ + - ]: 748 : *size = os.size();
3576 [ + - ]: 748 : add_mhd_response_header (r, "Content-Type", "text/plain");
3577 : : }
3578 : 1496 : return r;
3579 : 748 : }
3580 : :
3581 : :
3582 : : static struct MHD_Response*
3583 : 30 : handle_metadata (MHD_Connection* conn,
3584 : : string key, string value, off_t* size)
3585 : : {
3586 : 30 : MHD_Response* r;
3587 : : // Because this query can take on the order of many seconds, we need
3588 : : // to prevent DoS against the other normal quick queries, so we use
3589 : : // a dedicated database connection.
3590 : 30 : sqlite3 *thisdb = 0;
3591 : 30 : int rc = sqlite3_open_v2 (db_path.c_str(), &thisdb, (SQLITE_OPEN_READONLY
3592 : : |SQLITE_OPEN_URI
3593 : : |SQLITE_OPEN_PRIVATECACHE
3594 : : |SQLITE_OPEN_NOMUTEX), /* private to us */
3595 : : NULL);
3596 [ - + ]: 30 : if (rc)
3597 [ # # # # ]: 0 : throw sqlite_exception(rc, "cannot open database for metadata query");
3598 : 30 : defer_dtor<sqlite3*,int> sqlite_db_closer (thisdb, sqlite3_close_v2);
3599 : :
3600 : : // Query locally for matching e, d files
3601 [ + + ]: 30 : string op;
3602 [ + + ]: 30 : if (key == "glob")
3603 [ + - ]: 26 : op = "glob";
3604 [ + - ]: 4 : else if (key == "file")
3605 [ + - ]: 4 : op = "=";
3606 : : else
3607 [ # # ]: 0 : throw reportable_exception("/metadata webapi error, unsupported key");
3608 : :
3609 : : // Since PR30378, the file names are segmented into two tables. We
3610 : : // could do a glob/= search over the _files_v view that combines
3611 : : // them, but that means that the entire _files_v thing has to be
3612 : : // materialized & scanned to do the query. Slow! Instead, we can
3613 : : // segment the incoming file/glob pattern into dirname / basename
3614 : : // parts, and apply them to the corresponding table. This is done
3615 : : // by splitting the value at the last "/". If absent, the same
3616 : : // convention as is used in register_file_name().
3617 : :
3618 : 30 : string dirname, bname; // basename is a "poisoned" identifier on some distros
3619 : 30 : size_t slash = value.rfind('/');
3620 [ - + ]: 30 : if (slash == std::string::npos) {
3621 [ # # ]: 0 : dirname = "";
3622 [ # # ]: 0 : bname = value;
3623 : : } else {
3624 [ + - ]: 30 : dirname = value.substr(0, slash);
3625 [ + - ]: 30 : bname = value.substr(slash+1);
3626 : : }
3627 : :
3628 : : // NB: further optimization is possible: replacing the 'glob' op
3629 : : // with simple equality, if the corresponding value segment lacks
3630 : : // metacharacters. sqlite may or may not be smart enough to do so,
3631 : : // so we help out.
3632 [ + - ]: 30 : string metacharacters = "[]*?";
3633 [ + + + + : 56 : string dop = (op == "glob" && dirname.find_first_of(metacharacters) == string::npos) ? "=" : op;
+ - + - ]
3634 [ + + - + : 56 : string bop = (op == "glob" && bname.find_first_of(metacharacters) == string::npos) ? "=" : op;
- - + - ]
3635 : :
3636 : 30 : string sql = string(
3637 : : // explicit query r_de and f_de once here, rather than the query_d and query_e
3638 : : // separately, because they scan the same tables, so we'd double the work
3639 : : "select d1.executable_p, d1.debuginfo_p, 0 as source_p, "
3640 : : " b1.hex, f1d.name || '/' || f1b.name as file, a1.name as archive "
3641 : : "from " BUILDIDS "_r_de d1, " BUILDIDS "_files f1, " BUILDIDS "_fileparts f1b, " BUILDIDS "_fileparts f1d, "
3642 : : BUILDIDS "_buildids b1, " BUILDIDS "_files_v a1 "
3643 : : "where f1.id = d1.content and a1.id = d1.file and d1.buildid = b1.id "
3644 [ + - + - ]: 90 : " and f1d.name " + dop + " ? and f1b.name " + bop + " ? and f1.dirname = f1d.id and f1.basename = f1b.id "
3645 : : "union all \n"
3646 : : "select d2.executable_p, d2.debuginfo_p, 0, "
3647 : : " b2.hex, f2d.name || '/' || f2b.name, NULL "
3648 : : "from " BUILDIDS "_f_de d2, " BUILDIDS "_files f2, " BUILDIDS "_fileparts f2b, " BUILDIDS "_fileparts f2d, "
3649 : : BUILDIDS "_buildids b2 "
3650 : : "where f2.id = d2.file and d2.buildid = b2.id "
3651 [ + - + - ]: 90 : " and f2d.name " + dop + " ? and f2b.name " + bop + " ? "
3652 : 30 : " and f2.dirname = f2d.id and f2.basename = f2b.id");
3653 : :
3654 : : // NB: we could query source file names too, thusly:
3655 : : //
3656 : : // select * from " BUILDIDS "_buildids b, " BUILDIDS "_files_v f1, " BUILDIDS "_r_sref sr
3657 : : // where b.id = sr.buildid and f1.id = sr.artifactsrc and f1.name " + op + "?"
3658 : : // UNION ALL something with BUILDIDS "_f_s"
3659 : : //
3660 : : // But the first part of this query cannot run fast without the same index temp-created
3661 : : // during "maxigroom":
3662 : : // create index " BUILDIDS "_r_sref_arc on " BUILDIDS "_r_sref(artifactsrc);
3663 : : // and unfortunately this index is HUGE. It's similar to the size of the _r_sref
3664 : : // table, which is already the largest part of a debuginfod index. Adding that index
3665 : : // would nearly double the .sqlite db size.
3666 : :
3667 [ + - + - : 30 : sqlite_ps *pp = new sqlite_ps (thisdb, "mhd-query-meta-glob", sql);
+ - ]
3668 [ + - ]: 30 : pp->reset();
3669 [ + - ]: 30 : pp->bind(1, dirname);
3670 [ + - ]: 30 : pp->bind(2, bname);
3671 [ + - ]: 30 : pp->bind(3, dirname);
3672 [ + - ]: 30 : pp->bind(4, bname);
3673 : 30 : unique_ptr<sqlite_ps> ps_closer(pp); // release pp if exception or return
3674 : 30 : pp->reset_timeout(metadata_maxtime_s);
3675 : :
3676 [ + - ]: 30 : json_object *metadata = json_object_new_object();
3677 [ - + - - : 30 : if (!metadata) throw libc_exception(ENOMEM, "json allocation");
- - ]
3678 : 30 : defer_dtor<json_object*,int> metadata_d(metadata, json_object_put);
3679 [ + - ]: 30 : json_object *metadata_arr = json_object_new_array();
3680 [ - + - - : 30 : if (!metadata_arr) throw libc_exception(ENOMEM, "json allocation");
- - ]
3681 [ + - ]: 30 : json_object_object_add(metadata, "results", metadata_arr);
3682 : : // consume all the rows
3683 : :
3684 : 48 : bool metadata_complete = true;
3685 : 48 : while (1)
3686 : : {
3687 [ + - ]: 48 : rc = pp->step_timeout();
3688 [ + + ]: 48 : if (rc == SQLITE_DONE) // success
3689 : : break;
3690 [ + - ]: 18 : if (rc == SQLITE_ABORT || rc == SQLITE_INTERRUPT) // interrupted such as by timeout
3691 : : {
3692 : : metadata_complete = false;
3693 : : break;
3694 : : }
3695 [ - + ]: 18 : if (rc != SQLITE_ROW) // error
3696 [ # # # # ]: 0 : throw sqlite_exception(rc, "step");
3697 : :
3698 [ + - ]: 18 : int m_executable_p = sqlite3_column_int (*pp, 0);
3699 [ + - ]: 18 : int m_debuginfo_p = sqlite3_column_int (*pp, 1);
3700 [ + - ]: 18 : int m_source_p = sqlite3_column_int (*pp, 2);
3701 [ + - - + : 18 : string m_buildid = (const char*) sqlite3_column_text (*pp, 3) ?: ""; // should always be non-null
+ - ]
3702 [ + - - + : 18 : string m_file = (const char*) sqlite3_column_text (*pp, 4) ?: "";
+ - ]
3703 [ + - - + : 18 : string m_archive = (const char*) sqlite3_column_text (*pp, 5) ?: "";
+ - ]
3704 : :
3705 : : // Confirm that m_file matches in the fnmatch(FNM_PATHNAME)
3706 : : // sense, since sqlite's GLOB operator is a looser filter.
3707 [ + - + - : 18 : if (key == "glob" && fnmatch(value.c_str(), m_file.c_str(), FNM_PATHNAME) != 0)
- + ]
3708 : 0 : continue;
3709 : :
3710 : 36 : auto add_metadata = [metadata_arr, m_buildid, m_file, m_archive](const string& type) {
3711 : 18 : json_object* entry = json_object_new_object();
3712 [ - + - - : 18 : if (NULL == entry) throw libc_exception (ENOMEM, "cannot allocate json");
- - ]
3713 : 18 : defer_dtor<json_object*,int> entry_d(entry, json_object_put);
3714 : :
3715 : 90 : auto add_entry_metadata = [entry](const char* k, string v) {
3716 : 72 : json_object* s;
3717 [ + - ]: 72 : if(v != "") {
3718 : 72 : s = json_object_new_string(v.c_str());
3719 [ - + - - : 72 : if (NULL == s) throw libc_exception (ENOMEM, "cannot allocate json");
- - ]
3720 : 72 : json_object_object_add(entry, k, s);
3721 : : }
3722 : 72 : };
3723 : :
3724 [ + - + - ]: 18 : add_entry_metadata("type", type.c_str());
3725 [ + - + - ]: 18 : add_entry_metadata("buildid", m_buildid);
3726 [ + - + - ]: 18 : add_entry_metadata("file", m_file);
3727 [ + - + - : 36 : if (m_archive != "") add_entry_metadata("archive", m_archive);
+ - ]
3728 [ - + ]: 18 : if (verbose > 3)
3729 [ # # ]: 0 : obatched(clog) << "metadata found local "
3730 : : << json_object_to_json_string_ext(entry,
3731 [ # # # # : 0 : JSON_C_TO_STRING_PRETTY)
# # ]
3732 : 0 : << endl;
3733 : :
3734 : : // Increase ref count to switch its ownership
3735 [ + - + - ]: 18 : json_object_array_add(metadata_arr, json_object_get(entry));
3736 [ + - + - : 36 : };
+ - ]
3737 : :
3738 [ + - + - : 36 : if (m_executable_p) add_metadata("executable");
+ - ]
3739 [ - + - - : 18 : if (m_debuginfo_p) add_metadata("debuginfo");
- - ]
3740 [ - + - - : 18 : if (m_source_p) add_metadata("source");
- - ]
3741 : 18 : }
3742 [ + - ]: 30 : pp->reset();
3743 : :
3744 [ + - ]: 30 : unsigned num_local_results = json_object_array_length(metadata_arr);
3745 : :
3746 : : // Query upstream as well
3747 [ + - ]: 30 : debuginfod_client *client = debuginfod_pool_begin();
3748 [ + - ]: 30 : if (client != NULL)
3749 : : {
3750 [ + - ]: 30 : add_client_federation_headers(client, conn);
3751 : :
3752 : 30 : int upstream_metadata_fd;
3753 : 30 : char *upstream_metadata_file = NULL;
3754 [ + - ]: 30 : upstream_metadata_fd = debuginfod_find_metadata(client, key.c_str(), (char*)value.c_str(),
3755 : : &upstream_metadata_file);
3756 [ + + ]: 30 : if (upstream_metadata_fd >= 0) {
3757 : : /* json-c >= 0.13 has json_object_from_fd(). */
3758 [ + - ]: 18 : json_object *upstream_metadata_json = json_object_from_file(upstream_metadata_file);
3759 : 18 : free (upstream_metadata_file);
3760 : 18 : json_object *upstream_metadata_json_arr;
3761 : 18 : json_object *upstream_complete;
3762 [ - + ]: 18 : if (NULL != upstream_metadata_json &&
3763 [ + - + - : 36 : json_object_object_get_ex(upstream_metadata_json, "results", &upstream_metadata_json_arr) &&
- + ]
3764 [ + - ]: 18 : json_object_object_get_ex(upstream_metadata_json, "complete", &upstream_complete))
3765 : : {
3766 [ + - ]: 18 : metadata_complete &= json_object_get_boolean(upstream_complete);
3767 [ + - + + ]: 22 : for (int i = 0, n = json_object_array_length(upstream_metadata_json_arr); i < n; i++)
3768 : : {
3769 [ + - ]: 4 : json_object *entry = json_object_array_get_idx(upstream_metadata_json_arr, i);
3770 [ - + ]: 4 : if (verbose > 3)
3771 [ # # ]: 0 : obatched(clog) << "metadata found remote "
3772 : : << json_object_to_json_string_ext(entry,
3773 [ # # # # : 0 : JSON_C_TO_STRING_PRETTY)
# # ]
3774 : 0 : << endl;
3775 : :
3776 [ + - ]: 4 : json_object_get(entry); // increment reference count
3777 [ + - ]: 4 : json_object_array_add(metadata_arr, entry);
3778 : : }
3779 [ + - ]: 18 : json_object_put(upstream_metadata_json);
3780 : : }
3781 [ + - ]: 18 : close(upstream_metadata_fd);
3782 : : }
3783 [ + - ]: 30 : debuginfod_pool_end (client);
3784 : : }
3785 : :
3786 [ + - ]: 30 : unsigned num_total_results = json_object_array_length(metadata_arr);
3787 : :
3788 [ + - ]: 30 : if (verbose > 2)
3789 [ + - + - ]: 90 : obatched(clog) << "metadata found local=" << num_local_results
3790 [ + - + - ]: 30 : << " remote=" << (num_total_results-num_local_results)
3791 [ + - + - : 30 : << " total=" << num_total_results
+ - ]
3792 : 30 : << endl;
3793 : :
3794 [ + - + - ]: 30 : json_object_object_add(metadata, "complete", json_object_new_boolean(metadata_complete));
3795 [ + - ]: 30 : const char* metadata_str = json_object_to_json_string(metadata);
3796 [ - + ]: 30 : if (!metadata_str)
3797 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate json");
3798 [ + - ]: 30 : r = MHD_create_response_from_buffer (strlen(metadata_str),
3799 : : (void*) metadata_str,
3800 : : MHD_RESPMEM_MUST_COPY);
3801 : 30 : *size = strlen(metadata_str);
3802 [ + - ]: 30 : if (r)
3803 [ + - ]: 30 : add_mhd_response_header(r, "Content-Type", "application/json");
3804 : 30 : return r;
3805 : 30 : }
3806 : :
3807 : :
3808 : : static struct MHD_Response*
3809 : 4 : handle_root (off_t* size)
3810 : : {
3811 : 4 : MHD_Response* r;
3812 [ + - ]: 4 : if (cust_homepage_file != "")
3813 : 4 : try
3814 : : {
3815 [ + - ]: 4 : int fd = open (cust_homepage_file.c_str(), O_RDONLY);
3816 [ + + ]: 4 : if (fd != -1) {
3817 : 2 : struct stat buf;
3818 : 2 : stat (cust_homepage_file.c_str(), &buf);
3819 [ + - ]: 2 : r = MHD_create_response_from_fd(buf.st_size, fd);
3820 : : // NB: MHD owns and handles the fd from now. Must not close()!
3821 [ + - ]: 2 : if (r != NULL)
3822 : : {
3823 : 2 : *size = buf.st_size;
3824 [ + - ]: 2 : add_mhd_response_header (r, "Content-Type", "text/html");
3825 : : }
3826 : : } else {
3827 [ + - + - ]: 4 : throw libc_exception (errno, "cannot open file " + cust_homepage_file);
3828 : : }
3829 : 2 : return r;
3830 : : }
3831 [ - + ]: 2 : catch (const reportable_exception& e)
3832 : : {
3833 [ + - ]: 2 : e.report(clog);
3834 : 2 : }
3835 : :
3836 [ + - + - ]: 4 : static string version = "debuginfod (" + string (PACKAGE_NAME) + ") "
3837 [ + - + - : 66 : + string (PACKAGE_VERSION);
+ - + - ]
3838 : 2 : r = MHD_create_response_from_buffer (version.size (),
3839 : 2 : (void *) version.c_str (),
3840 : : MHD_RESPMEM_PERSISTENT);
3841 [ + - ]: 2 : if (r != NULL)
3842 : : {
3843 : 2 : *size = version.size ();
3844 : 2 : add_mhd_response_header (r, "Content-Type", "text/plain");
3845 : : }
3846 : : return r;
3847 : : }
3848 : :
3849 : :
3850 : : static struct MHD_Response*
3851 : 2 : handle_options (off_t* size)
3852 : : {
3853 : 2 : static char empty_body[] = " ";
3854 : 2 : MHD_Response* r = MHD_create_response_from_buffer (1, empty_body,
3855 : : MHD_RESPMEM_PERSISTENT);
3856 [ + - ]: 2 : if (r != NULL)
3857 : : {
3858 : 2 : *size = 1;
3859 : 2 : add_mhd_response_header (r, "Access-Control-Allow-Origin", "*");
3860 : 2 : add_mhd_response_header (r, "Access-Control-Allow-Methods", "GET, OPTIONS");
3861 : 2 : add_mhd_response_header (r, "Access-Control-Allow-Headers", "cache-control");
3862 : : }
3863 : 2 : return r;
3864 : : }
3865 : :
3866 : :
3867 : : ////////////////////////////////////////////////////////////////////////
3868 : :
3869 : :
3870 : : /* libmicrohttpd callback */
3871 : : static MHD_RESULT
3872 : 6769 : handler_cb (void * /*cls*/,
3873 : : struct MHD_Connection *connection,
3874 : : const char *url,
3875 : : const char *method,
3876 : : const char * /*version*/,
3877 : : const char * /*upload_data*/,
3878 : : size_t * /*upload_data_size*/,
3879 : : void ** ptr)
3880 : : {
3881 : 6769 : struct MHD_Response *r = NULL;
3882 : 6769 : string url_copy = url;
3883 : :
3884 : : /* libmicrohttpd always makes (at least) two callbacks: once just
3885 : : past the headers, and one after the request body is finished
3886 : : being received. If we process things early (first callback) and
3887 : : queue a response, libmicrohttpd would suppress http keep-alive
3888 : : (via connection->read_closed = true). */
3889 : 6770 : static int aptr; /* just some random object to use as a flag */
3890 [ + + ]: 6770 : if (&aptr != *ptr)
3891 : : {
3892 : : /* do never respond on first call */
3893 : 3385 : *ptr = &aptr;
3894 : 3385 : return MHD_YES;
3895 : : }
3896 : 3385 : *ptr = NULL; /* reset when done */
3897 : :
3898 [ + - ]: 3385 : const char *maxsize_string = MHD_lookup_connection_value(connection, MHD_HEADER_KIND, "X-DEBUGINFOD-MAXSIZE");
3899 : 3385 : long maxsize = 0;
3900 [ + + + - ]: 3385 : if (maxsize_string != NULL && maxsize_string[0] != '\0')
3901 : 2 : maxsize = atol(maxsize_string);
3902 : : else
3903 : : maxsize = 0;
3904 : :
3905 : : #if MHD_VERSION >= 0x00097002
3906 : 3385 : enum MHD_Result rc;
3907 : : #else
3908 : : int rc = MHD_NO; // mhd
3909 : : #endif
3910 : 3385 : int http_code = 500;
3911 : 3385 : off_t http_size = -1;
3912 : 3385 : struct timespec ts_start, ts_end;
3913 : 3385 : clock_gettime (CLOCK_MONOTONIC, &ts_start);
3914 : 3385 : double afteryou = 0.0;
3915 [ + + ]: 3385 : string artifacttype, suffix;
3916 : 3385 : string urlargs; // for logging
3917 : :
3918 : 3385 : try
3919 : : {
3920 [ + + + - : 3385 : if (webapi_cors && method == string("OPTIONS"))
+ + ]
3921 : : {
3922 [ + - + - : 4 : inc_metric("http_requests_total", "type", method);
+ - + - ]
3923 [ + - ]: 2 : r = handle_options(& http_size);
3924 [ + - ]: 2 : rc = MHD_queue_response (connection, MHD_HTTP_OK, r);
3925 : 2 : http_code = MHD_HTTP_OK;
3926 [ + - ]: 2 : MHD_destroy_response (r);
3927 : 2 : return rc;
3928 : : }
3929 [ + - - + ]: 4014 : else if (string(method) != "GET")
3930 [ # # ]: 0 : throw reportable_exception(400, "we support OPTIONS+GET only");
3931 : :
3932 : : /* Start decoding the URL. */
3933 : 3383 : size_t slash1 = url_copy.find('/', 1);
3934 [ + - ]: 3383 : string url1 = url_copy.substr(0, slash1); // ok even if slash1 not found
3935 : :
3936 [ + + - + ]: 3383 : if (slash1 != string::npos && url1 == "/buildid")
3937 : : {
3938 : : // PR27863: block this thread awhile if another thread is already busy
3939 : : // fetching the exact same thing. This is better for Everyone.
3940 : : // The latecomer says "... after you!" and waits.
3941 [ + - + - : 5813 : add_metric ("thread_busy", "role", "http-buildid-after-you", 1);
+ - + - ]
3942 : : #ifdef HAVE_PTHREAD_SETNAME_NP
3943 : 2595 : (void) pthread_setname_np (pthread_self(), "mhd-buildid-after-you");
3944 : : #endif
3945 : 2595 : struct timespec tsay_start, tsay_end;
3946 : 2595 : clock_gettime (CLOCK_MONOTONIC, &tsay_start);
3947 [ + + + - ]: 2655 : static unique_set<string> busy_urls;
3948 [ + - ]: 2595 : unique_set_reserver<string> after_you(busy_urls, url_copy);
3949 : 2595 : clock_gettime (CLOCK_MONOTONIC, &tsay_end);
3950 : 2595 : afteryou = (tsay_end.tv_sec - tsay_start.tv_sec) + (tsay_end.tv_nsec - tsay_start.tv_nsec)/1.e9;
3951 [ + - + - : 5190 : add_metric ("thread_busy", "role", "http-buildid-after-you", -1);
+ - + - ]
3952 : :
3953 [ + - + - : 5190 : tmp_inc_metric m ("thread_busy", "role", "http-buildid");
+ - + - ]
3954 : : #ifdef HAVE_PTHREAD_SETNAME_NP
3955 : 2595 : (void) pthread_setname_np (pthread_self(), "mhd-buildid");
3956 : : #endif
3957 : 2595 : size_t slash2 = url_copy.find('/', slash1+1);
3958 [ - + ]: 2595 : if (slash2 == string::npos)
3959 [ # # ]: 0 : throw reportable_exception("/buildid/ webapi error, need buildid");
3960 : :
3961 [ + - ]: 2595 : string buildid = url_copy.substr(slash1+1, slash2-slash1-1);
3962 : :
3963 : 2595 : size_t slash3 = url_copy.find('/', slash2+1);
3964 : :
3965 [ + + ]: 2595 : if (slash3 == string::npos)
3966 : : {
3967 [ + - ]: 1461 : artifacttype = url_copy.substr(slash2+1);
3968 [ + - ]: 1461 : suffix = "";
3969 : : }
3970 : : else
3971 : : {
3972 [ + - ]: 1134 : artifacttype = url_copy.substr(slash2+1, slash3-slash2-1);
3973 [ + - ]: 1757 : suffix = url_copy.substr(slash3); // include the slash in the suffix
3974 : : }
3975 : :
3976 : : // get the resulting fd so we can report its size
3977 : 2595 : int fd;
3978 [ + + ]: 2595 : r = handle_buildid (connection, buildid, artifacttype, suffix, &fd);
3979 : 1972 : if (r)
3980 : : {
3981 : 1972 : struct stat fs;
3982 [ + - ]: 1972 : if (fstat(fd, &fs) == 0)
3983 : 1972 : http_size = fs.st_size;
3984 : : // libmicrohttpd will close (fd);
3985 : : }
3986 : 3218 : }
3987 [ + + ]: 788 : else if (url1 == "/metrics")
3988 : : {
3989 [ + - + - : 1496 : tmp_inc_metric m ("thread_busy", "role", "http-metrics");
+ - + - ]
3990 [ + - ]: 748 : artifacttype = "metrics";
3991 [ + - + - : 1496 : inc_metric("http_requests_total", "type", artifacttype);
+ - ]
3992 [ + - ]: 748 : r = handle_metrics(& http_size);
3993 : 748 : }
3994 [ + + ]: 40 : else if (url1 == "/metadata")
3995 : : {
3996 [ + - + - : 60 : tmp_inc_metric m ("thread_busy", "role", "http-metadata");
+ - + - ]
3997 [ + - ]: 30 : const char* key = MHD_lookup_connection_value(connection, MHD_GET_ARGUMENT_KIND, "key");
3998 [ + - ]: 30 : const char* value = MHD_lookup_connection_value(connection, MHD_GET_ARGUMENT_KIND, "value");
3999 [ - + ]: 30 : if (NULL == value || NULL == key)
4000 [ # # ]: 0 : throw reportable_exception("/metadata webapi error, need key and value");
4001 : :
4002 [ + - + - : 30 : urlargs = string("?key=") + string(key) + string("&value=") + string(value); // apprx., for logging
+ - + - +
- + - +
- ]
4003 [ + - ]: 30 : artifacttype = "metadata";
4004 [ + - + - : 60 : inc_metric("http_requests_total", "type", artifacttype);
+ - ]
4005 [ + - + - : 30 : r = handle_metadata(connection, key, value, &http_size);
+ - ]
4006 : 30 : }
4007 [ + + ]: 10 : else if (url1 == "/")
4008 : : {
4009 [ + - ]: 4 : artifacttype = "/";
4010 [ + - + - : 639 : inc_metric("http_requests_total", "type", artifacttype);
+ - ]
4011 [ + - ]: 4 : r = handle_root(& http_size);
4012 : : }
4013 : : else
4014 [ + - + - ]: 12 : throw reportable_exception("webapi error, unrecognized '" + url1 + "'");
4015 : :
4016 [ - + ]: 2754 : if (r == 0)
4017 [ # # ]: 0 : throw reportable_exception("internal error, missing response");
4018 : :
4019 [ + + + - ]: 2754 : if (maxsize > 0 && http_size > maxsize)
4020 : : {
4021 [ + - ]: 2 : MHD_destroy_response(r);
4022 [ + - + - ]: 4 : throw reportable_exception(406, "File too large, max size=" + std::to_string(maxsize));
4023 : : }
4024 : :
4025 [ + + ]: 2752 : if (webapi_cors)
4026 : : // add ACAO header for all successful requests
4027 [ + - ]: 132 : add_mhd_response_header (r, "Access-Control-Allow-Origin", "*");
4028 [ + + + + ]: 2752 : if ((cust_homepage_redirect) != "" && (url1 == "/"))
4029 : : {
4030 : : // redirect to given custom --homepage
4031 [ + - ]: 4 : MHD_add_response_header(r, "Location", cust_homepage_redirect.c_str());
4032 [ + - ]: 4 : rc = MHD_queue_response (connection, MHD_HTTP_FOUND, r);
4033 : : http_code = MHD_HTTP_FOUND;
4034 : : }
4035 : : else
4036 : : {
4037 [ + - ]: 2748 : rc = MHD_queue_response (connection, MHD_HTTP_OK, r);
4038 : : http_code = MHD_HTTP_OK;
4039 : : }
4040 [ + - ]: 2752 : MHD_destroy_response (r);
4041 : 3383 : }
4042 [ - + ]: 631 : catch (const reportable_exception& e)
4043 : : {
4044 [ + - + - : 1262 : inc_metric("http_responses_total","result","error");
+ - + - ]
4045 [ + - ]: 631 : e.report(clog);
4046 : 631 : http_code = e.code;
4047 [ + - ]: 631 : http_size = e.message.size();
4048 [ + - ]: 631 : rc = e.mhd_send_response (connection);
4049 : 631 : }
4050 : :
4051 : 3383 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
4052 : 3383 : double deltas = (ts_end.tv_sec - ts_start.tv_sec) + (ts_end.tv_nsec - ts_start.tv_nsec)/1.e9;
4053 : : // afteryou: delay waiting for other client's identical query to complete
4054 : : // deltas: total latency, including afteryou waiting
4055 [ + - + - ]: 6766 : obatched(clog) << conninfo(connection)
4056 : : << ' ' << method << ' ' << url << urlargs
4057 [ + - + - : 3383 : << ' ' << http_code << ' ' << http_size
+ - + - +
- + - + -
+ - ]
4058 [ + - + - : 3383 : << ' ' << (int)(afteryou*1000) << '+' << (int)((deltas-afteryou)*1000) << "ms"
+ - + - +
- + - +
- ]
4059 : 3383 : << endl;
4060 : :
4061 : : // related prometheus metrics
4062 : 3383 : string http_code_str = to_string(http_code);
4063 [ + - + - : 6766 : add_metric("http_responses_transfer_bytes_sum",
+ - + - ]
4064 : : "code", http_code_str, "type", artifacttype, http_size);
4065 [ + - + - : 6766 : inc_metric("http_responses_transfer_bytes_count",
+ - + - ]
4066 : : "code", http_code_str, "type", artifacttype);
4067 : :
4068 [ + - + - : 6766 : add_metric("http_responses_duration_milliseconds_sum",
+ - + - ]
4069 : : "code", http_code_str, "type", artifacttype, deltas*1000); // prometheus prefers _seconds and floating point
4070 [ + - + - : 6766 : inc_metric("http_responses_duration_milliseconds_count",
+ - + - ]
4071 : : "code", http_code_str, "type", artifacttype);
4072 : :
4073 [ + - + - : 6766 : add_metric("http_responses_after_you_milliseconds_sum",
+ - + - ]
4074 : : "code", http_code_str, "type", artifacttype, afteryou*1000);
4075 [ + - + - : 6766 : inc_metric("http_responses_after_you_milliseconds_count",
+ - + - ]
4076 : : "code", http_code_str, "type", artifacttype);
4077 : :
4078 : 3383 : return rc;
4079 : 13538 : }
4080 : :
4081 : :
4082 : : ////////////////////////////////////////////////////////////////////////
4083 : : // borrowed originally from src/nm.c get_local_names()
4084 : :
4085 : : static void
4086 : 412 : dwarf_extract_source_paths (Elf *elf, set<string>& debug_sourcefiles)
4087 : : noexcept // no exceptions - so we can simplify the altdbg resource release at end
4088 : : {
4089 : 412 : Dwarf* dbg = dwarf_begin_elf (elf, DWARF_C_READ, NULL);
4090 [ - + ]: 412 : if (dbg == NULL)
4091 : 0 : return;
4092 : :
4093 : 412 : Dwarf* altdbg = NULL;
4094 : 412 : int altdbg_fd = -1;
4095 : :
4096 : : // DWZ handling: if we have an unsatisfied debug-alt-link, add an
4097 : : // empty string into the outgoing sourcefiles set, so the caller
4098 : : // should know that our data is incomplete.
4099 : 412 : const char *alt_name_p;
4100 : 412 : const void *alt_build_id; // elfutils-owned memory
4101 : 412 : ssize_t sz = dwelf_dwarf_gnu_debugaltlink (dbg, &alt_name_p, &alt_build_id);
4102 [ + + ]: 412 : if (sz > 0) // got one!
4103 : : {
4104 : 200 : string buildid;
4105 : 200 : unsigned char* build_id_bytes = (unsigned char*) alt_build_id;
4106 [ + + ]: 4200 : for (ssize_t idx=0; idx<sz; idx++)
4107 : : {
4108 : 4000 : buildid += "0123456789abcdef"[build_id_bytes[idx] >> 4];
4109 : 4000 : buildid += "0123456789abcdef"[build_id_bytes[idx] & 0xf];
4110 : : }
4111 : :
4112 [ + + ]: 200 : if (verbose > 3)
4113 : 156 : obatched(clog) << "Need altdebug buildid=" << buildid << endl;
4114 : :
4115 : : // but is it unsatisfied the normal elfutils ways?
4116 : 200 : Dwarf* alt = dwarf_getalt (dbg);
4117 [ + - ]: 200 : if (alt == NULL)
4118 : : {
4119 : : // Yup, unsatisfied the normal way. Maybe we can satisfy it
4120 : : // from our own debuginfod database.
4121 : 200 : int alt_fd;
4122 : 200 : struct MHD_Response *r = 0;
4123 : 200 : try
4124 : : {
4125 [ + - ]: 200 : string artifacttype = "debuginfo";
4126 [ + - + + ]: 220 : r = handle_buildid (0, buildid, artifacttype, "", &alt_fd);
4127 : : // NB: no need for ACAO etc. headers; this is not getting sent to a client
4128 : 20 : }
4129 [ - + ]: 20 : catch (const reportable_exception& e)
4130 : : {
4131 : : // swallow exceptions
4132 : 20 : }
4133 : :
4134 : : // NB: this is not actually recursive! This invokes the web-query
4135 : : // path, which cannot get back into the scan code paths.
4136 : 200 : if (r)
4137 : : {
4138 : : // Found it!
4139 : 180 : altdbg_fd = dup(alt_fd); // ok if this fails, downstream failures ok
4140 : 180 : alt = altdbg = dwarf_begin (altdbg_fd, DWARF_C_READ);
4141 : : // NB: must close this dwarf and this fd at the bottom of the function!
4142 : 180 : MHD_destroy_response (r); // will close alt_fd
4143 [ - + ]: 180 : if (alt)
4144 : 180 : dwarf_setalt (dbg, alt);
4145 : : }
4146 : : }
4147 : : else
4148 : : {
4149 : : // NB: dwarf_setalt(alt) inappropriate - already done!
4150 : : // NB: altdbg will stay 0 so nothing tries to redundantly dealloc.
4151 : : }
4152 : :
4153 [ + + ]: 200 : if (alt)
4154 : : {
4155 [ + + ]: 180 : if (verbose > 3)
4156 : 156 : obatched(clog) << "Resolved altdebug buildid=" << buildid << endl;
4157 : : }
4158 : : else // (alt == NULL) - signal possible presence of poor debuginfo
4159 : : {
4160 : 20 : debug_sourcefiles.insert("");
4161 [ + - ]: 20 : if (verbose > 3)
4162 : 0 : obatched(clog) << "Unresolved altdebug buildid=" << buildid << endl;
4163 : : }
4164 : 200 : }
4165 : :
4166 : 412 : Dwarf_Off offset = 0;
4167 : 412 : Dwarf_Off old_offset;
4168 : 412 : size_t hsize;
4169 : :
4170 [ + + ]: 9436 : while (dwarf_nextcu (dbg, old_offset = offset, &offset, &hsize, NULL, NULL, NULL) == 0)
4171 : : {
4172 : 9024 : Dwarf_Die cudie_mem;
4173 : 9024 : Dwarf_Die *cudie = dwarf_offdie (dbg, old_offset + hsize, &cudie_mem);
4174 : :
4175 [ - + ]: 9023 : if (cudie == NULL)
4176 : 36 : continue;
4177 [ + + ]: 9023 : if (dwarf_tag (cudie) != DW_TAG_compile_unit)
4178 : 36 : continue;
4179 : :
4180 [ - + ]: 8988 : const char *cuname = dwarf_diename(cudie) ?: "unknown";
4181 : :
4182 : 8988 : Dwarf_Files *files;
4183 : 8988 : size_t nfiles;
4184 [ - + ]: 8988 : if (dwarf_getsrcfiles (cudie, &files, &nfiles) != 0)
4185 : 0 : continue;
4186 : :
4187 : : // extract DW_AT_comp_dir to resolve relative file names
4188 : 8988 : const char *comp_dir = "";
4189 : 8988 : const char *const *dirs;
4190 : 8988 : size_t ndirs;
4191 [ - + ]: 8988 : if (dwarf_getsrcdirs (files, &dirs, &ndirs) == 0 &&
4192 [ - + ]: 8988 : dirs[0] != NULL)
4193 : : comp_dir = dirs[0];
4194 : : if (comp_dir == NULL)
4195 : : comp_dir = "";
4196 : :
4197 [ + + ]: 8988 : if (verbose > 3)
4198 : 13812 : obatched(clog) << "searching for sources for cu=" << cuname << " comp_dir=" << comp_dir
4199 : 6906 : << " #files=" << nfiles << " #dirs=" << ndirs << endl;
4200 : :
4201 [ - + - - ]: 8988 : if (comp_dir[0] == '\0' && cuname[0] != '/')
4202 : : {
4203 [ # # ]: 0 : if (verbose > 3)
4204 : 0 : obatched(clog) << "skipping cu=" << cuname << " due to empty comp_dir" << endl;
4205 : 0 : continue;
4206 : : }
4207 : :
4208 [ + + ]: 147201 : for (size_t f = 1; f < nfiles; f++)
4209 : : {
4210 : 138213 : const char *hat = dwarf_filesrc (files, f, NULL, NULL);
4211 [ - + ]: 138244 : if (hat == NULL)
4212 : 1690 : continue;
4213 : :
4214 [ + + ]: 138265 : if (string(hat) == "<built-in>"
4215 [ + + + + ]: 276501 : || string_endswith(hat, "<built-in>")) // gcc intrinsics, don't bother record
4216 : 1690 : continue;
4217 : :
4218 [ + + ]: 136575 : string waldo;
4219 [ + + ]: 136575 : if (hat[0] == '/') // absolute
4220 : 85815 : waldo = (string (hat));
4221 [ + - ]: 50760 : else if (comp_dir[0] != '\0') // comp_dir relative
4222 : 50760 : waldo = (string (comp_dir) + string("/") + string (hat));
4223 : : else
4224 : : {
4225 [ # # ]: 0 : if (verbose > 3)
4226 : 0 : obatched(clog) << "skipping hat=" << hat << " due to empty comp_dir" << endl;
4227 : 0 : continue;
4228 : : }
4229 : :
4230 : : // NB: this is the 'waldo' that a dbginfo client will have
4231 : : // to supply for us to give them the file The comp_dir
4232 : : // prefixing is a definite complication. Otherwise we'd
4233 : : // have to return a setof comp_dirs (one per CU!) with
4234 : : // corresponding filesrc[] names, instead of one absolute
4235 : : // resoved set. Maybe we'll have to do that anyway. XXX
4236 : :
4237 [ + + ]: 136586 : if (verbose > 4)
4238 [ - + ]: 32 : obatched(clog) << waldo
4239 [ - + ]: 16 : << (debug_sourcefiles.find(waldo)==debug_sourcefiles.end() ? " new" : " dup") << endl;
4240 : :
4241 : 136586 : debug_sourcefiles.insert (waldo);
4242 : 136523 : }
4243 : : }
4244 : :
4245 : 412 : dwarf_end(dbg);
4246 [ + + ]: 412 : if (altdbg)
4247 : 180 : dwarf_end(altdbg);
4248 [ + + ]: 412 : if (altdbg_fd >= 0)
4249 : 180 : close(altdbg_fd);
4250 : : }
4251 : :
4252 : :
4253 : :
4254 : : static void
4255 : 1708 : elf_classify (int fd, bool &executable_p, bool &debuginfo_p, string &buildid, set<string>& debug_sourcefiles)
4256 : : {
4257 : 1708 : Elf *elf = elf_begin (fd, ELF_C_READ_MMAP_PRIVATE, NULL);
4258 [ + - ]: 1708 : if (elf == NULL)
4259 : : return;
4260 : :
4261 : 1708 : try // catch our types of errors and clean up the Elf* object
4262 : : {
4263 [ + - + + ]: 1708 : if (elf_kind (elf) != ELF_K_ELF)
4264 : : {
4265 [ + - ]: 906 : elf_end (elf);
4266 : 946 : return;
4267 : : }
4268 : :
4269 : 802 : GElf_Ehdr ehdr_storage;
4270 [ + - ]: 802 : GElf_Ehdr *ehdr = gelf_getehdr (elf, &ehdr_storage);
4271 [ - + ]: 802 : if (ehdr == NULL)
4272 : : {
4273 [ # # ]: 0 : elf_end (elf);
4274 : : return;
4275 : : }
4276 : 802 : auto elf_type = ehdr->e_type;
4277 : :
4278 : 802 : const void *build_id; // elfutils-owned memory
4279 [ + - ]: 802 : ssize_t sz = dwelf_elf_gnu_build_id (elf, & build_id);
4280 [ + + ]: 799 : if (sz <= 0)
4281 : : {
4282 : : // It's not a diagnostic-worthy error for an elf file to lack build-id.
4283 : : // It might just be very old.
4284 [ + - ]: 40 : elf_end (elf);
4285 : : return;
4286 : : }
4287 : :
4288 : : // build_id is a raw byte array; convert to hexadecimal *lowercase*
4289 : 759 : unsigned char* build_id_bytes = (unsigned char*) build_id;
4290 [ + + ]: 15948 : for (ssize_t idx=0; idx<sz; idx++)
4291 : : {
4292 [ + - ]: 15186 : buildid += "0123456789abcdef"[build_id_bytes[idx] >> 4];
4293 [ + - ]: 30380 : buildid += "0123456789abcdef"[build_id_bytes[idx] & 0xf];
4294 : : }
4295 : :
4296 : : // now decide whether it's an executable - namely, any allocatable section has
4297 : : // PROGBITS;
4298 [ + + ]: 762 : if (elf_type == ET_EXEC || elf_type == ET_DYN)
4299 : : {
4300 : 700 : size_t shnum;
4301 [ + - ]: 700 : int rc = elf_getshdrnum (elf, &shnum);
4302 [ - + ]: 699 : if (rc < 0)
4303 [ # # # # ]: 0 : throw elfutils_exception(rc, "getshdrnum");
4304 : :
4305 : 699 : executable_p = false;
4306 [ + + ]: 13338 : for (size_t sc = 0; sc < shnum; sc++)
4307 : : {
4308 [ + - ]: 13012 : Elf_Scn *scn = elf_getscn (elf, sc);
4309 [ - + ]: 13011 : if (scn == NULL)
4310 : 0 : continue;
4311 : :
4312 : 13011 : GElf_Shdr shdr_mem;
4313 [ + - ]: 13011 : GElf_Shdr *shdr = gelf_getshdr (scn, &shdr_mem);
4314 [ - + ]: 13013 : if (shdr == NULL)
4315 : 0 : continue;
4316 : :
4317 : : // allocated (loadable / vm-addr-assigned) section with available content?
4318 [ + + + + ]: 13013 : if ((shdr->sh_type == SHT_PROGBITS) && (shdr->sh_flags & SHF_ALLOC))
4319 : : {
4320 [ - + ]: 374 : if (verbose > 4)
4321 [ # # # # : 0 : obatched(clog) << "executable due to SHF_ALLOC SHT_PROGBITS sc=" << sc << endl;
# # ]
4322 : 374 : executable_p = true;
4323 : 374 : break; // no need to keep looking for others
4324 : : }
4325 : : } // iterate over sections
4326 : : } // executable_p classification
4327 : :
4328 : : // now decide whether it's a debuginfo - namely, if it has any .debug* or .zdebug* sections
4329 : : // logic mostly stolen from fweimer@redhat.com's elfclassify drafts
4330 : 762 : size_t shstrndx;
4331 [ + - ]: 762 : int rc = elf_getshdrstrndx (elf, &shstrndx);
4332 [ - + ]: 762 : if (rc < 0)
4333 [ # # # # ]: 0 : throw elfutils_exception(rc, "getshdrstrndx");
4334 : :
4335 : : Elf_Scn *scn = NULL;
4336 : : bool symtab_p = false;
4337 : : bool bits_alloc_p = false;
4338 : 40866 : while (true)
4339 : : {
4340 [ + - ]: 20814 : scn = elf_nextscn (elf, scn);
4341 [ + + ]: 20684 : if (scn == NULL)
4342 : : break;
4343 : 20334 : GElf_Shdr shdr_storage;
4344 [ + - ]: 20334 : GElf_Shdr *shdr = gelf_getshdr (scn, &shdr_storage);
4345 [ + - ]: 20331 : if (shdr == NULL)
4346 : : break;
4347 [ + - ]: 20331 : const char *section_name = elf_strptr (elf, shstrndx, shdr->sh_name);
4348 [ + - ]: 20464 : if (section_name == NULL)
4349 : : break;
4350 [ + + ]: 20464 : if (startswith (section_name, ".debug_line") ||
4351 [ - + ]: 20052 : startswith (section_name, ".zdebug_line"))
4352 : : {
4353 : 412 : debuginfo_p = true;
4354 [ + - ]: 412 : if (scan_source_info)
4355 : 412 : dwarf_extract_source_paths (elf, debug_sourcefiles);
4356 : : break; // expecting only one .*debug_line, so no need to look for others
4357 : : }
4358 [ + + ]: 20052 : else if (startswith (section_name, ".debug_") ||
4359 [ + - ]: 18688 : startswith (section_name, ".zdebug_"))
4360 : : {
4361 : 1250 : debuginfo_p = true;
4362 : : // NB: don't break; need to parse .debug_line for sources
4363 : : }
4364 [ + + ]: 18802 : else if (shdr->sh_type == SHT_SYMTAB)
4365 : : {
4366 : : symtab_p = true;
4367 : : }
4368 : 18778 : else if (shdr->sh_type != SHT_NOBITS
4369 [ + + ]: 18778 : && shdr->sh_type != SHT_NOTE
4370 [ + + ]: 9188 : && (shdr->sh_flags & SHF_ALLOC) != 0)
4371 : : {
4372 : 20052 : bits_alloc_p = true;
4373 : : }
4374 : 20052 : }
4375 : :
4376 : : // For more expansive elf/split-debuginfo classification, we
4377 : : // want to identify as debuginfo "strip -s"-produced files
4378 : : // without .debug_info* (like libicudata), but we don't want to
4379 : : // identify "strip -g" executables (with .symtab left there).
4380 [ - + ]: 762 : if (symtab_p && !bits_alloc_p)
4381 : 0 : debuginfo_p = true;
4382 : : }
4383 [ # # ]: 0 : catch (const reportable_exception& e)
4384 : : {
4385 [ # # ]: 0 : e.report(clog);
4386 : 0 : }
4387 : 762 : elf_end (elf);
4388 : : }
4389 : :
4390 : :
4391 : : // Intern the given file name in two parts (dirname & basename) and
4392 : : // return the resulting file's id.
4393 : : static int64_t
4394 : 34057 : register_file_name(sqlite_ps& ps_upsert_fileparts,
4395 : : sqlite_ps& ps_upsert_file,
4396 : : sqlite_ps& ps_lookup_file,
4397 : : const string& name)
4398 : : {
4399 : 34057 : std::size_t slash = name.rfind('/');
4400 [ + + ]: 34058 : string dirname, filename;
4401 [ + + ]: 34058 : if (slash == std::string::npos)
4402 : : {
4403 [ + - ]: 90 : dirname = "";
4404 [ + - ]: 90 : filename = name;
4405 : : }
4406 : : else
4407 : : {
4408 [ + - ]: 33968 : dirname = name.substr(0, slash);
4409 [ + - ]: 33972 : filename = name.substr(slash+1);
4410 : : }
4411 : : // NB: see also handle_metadata()
4412 : :
4413 : : // intern the two substrings
4414 : 34061 : ps_upsert_fileparts
4415 [ + - ]: 34061 : .reset()
4416 [ + - ]: 34059 : .bind(1, dirname)
4417 [ + - ]: 34045 : .step_ok_done();
4418 : 34062 : ps_upsert_fileparts
4419 [ + - ]: 34062 : .reset()
4420 [ + - ]: 34062 : .bind(1, filename)
4421 [ + - ]: 34060 : .step_ok_done();
4422 : :
4423 : : // intern the tuple
4424 : 34062 : ps_upsert_file
4425 [ + - ]: 34062 : .reset()
4426 [ + - ]: 34062 : .bind(1, dirname)
4427 [ + - ]: 34059 : .bind(2, filename)
4428 [ + - ]: 34056 : .step_ok_done();
4429 : :
4430 : : // look up the tuple's id
4431 : 34062 : ps_lookup_file
4432 [ + - ]: 34062 : .reset()
4433 [ + - ]: 34062 : .bind(1, dirname)
4434 [ + - ]: 34056 : .bind(2, filename);
4435 [ + - ]: 34057 : int rc = ps_lookup_file.step();
4436 [ - + - - : 34062 : if (rc != SQLITE_ROW) throw sqlite_exception(rc, "step");
- - ]
4437 : :
4438 [ + - ]: 34062 : int64_t id = sqlite3_column_int64 (ps_lookup_file, 0);
4439 [ + - ]: 34060 : ps_lookup_file.reset();
4440 : 34062 : return id;
4441 : 34062 : }
4442 : :
4443 : :
4444 : :
4445 : : static void
4446 : 1106 : scan_source_file (const string& rps, const stat_t& st,
4447 : : sqlite_ps& ps_upsert_buildids,
4448 : : sqlite_ps& ps_upsert_fileparts,
4449 : : sqlite_ps& ps_upsert_file,
4450 : : sqlite_ps& ps_lookup_file,
4451 : : sqlite_ps& ps_upsert_de,
4452 : : sqlite_ps& ps_upsert_s,
4453 : : sqlite_ps& ps_query,
4454 : : sqlite_ps& ps_scan_done,
4455 : : unsigned& fts_cached,
4456 : : unsigned& fts_executable,
4457 : : unsigned& fts_debuginfo,
4458 : : unsigned& fts_sourcefiles)
4459 : : {
4460 : 1106 : int64_t fileid = register_file_name(ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, rps);
4461 : :
4462 : : /* See if we know of it already. */
4463 : 1106 : int rc = ps_query
4464 : 1106 : .reset()
4465 : 1106 : .bind(1, fileid)
4466 : 1105 : .bind(2, st.st_mtime)
4467 : 1106 : .step();
4468 : 1106 : ps_query.reset();
4469 [ + + ]: 1106 : if (rc == SQLITE_ROW) // i.e., a result, as opposed to DONE (no results)
4470 : : // no need to recheck a file/version we already know
4471 : : // specifically, no need to elf-begin a file we already determined is non-elf
4472 : : // (so is stored with buildid=NULL)
4473 : : {
4474 : 438 : fts_cached++;
4475 : 438 : return;
4476 : : }
4477 : :
4478 : 668 : bool executable_p = false, debuginfo_p = false; // E and/or D
4479 [ + - ]: 668 : string buildid;
4480 [ + - ]: 668 : set<string> sourcefiles;
4481 : :
4482 [ + - ]: 668 : int fd = open (rps.c_str(), O_RDONLY);
4483 : 668 : try
4484 : : {
4485 [ + - ]: 668 : if (fd >= 0)
4486 [ + - ]: 668 : elf_classify (fd, executable_p, debuginfo_p, buildid, sourcefiles);
4487 : : else
4488 [ # # # # : 0 : throw libc_exception(errno, string("open ") + rps);
# # ]
4489 [ + - + - : 1336 : add_metric ("scanned_bytes_total","source","file",
+ - ]
4490 [ + - ]: 668 : st.st_size);
4491 [ + - + - : 1336 : inc_metric ("scanned_files_total","source","file");
+ - + - ]
4492 : : }
4493 : : // NB: we catch exceptions here too, so that we can
4494 : : // cache the corrupt-elf case (!executable_p &&
4495 : : // !debuginfo_p) just below, just as if we had an
4496 : : // EPERM error from open(2).
4497 [ - - ]: 0 : catch (const reportable_exception& e)
4498 : : {
4499 [ - - ]: 0 : e.report(clog);
4500 : 0 : }
4501 : :
4502 [ + - ]: 668 : if (fd >= 0)
4503 [ + - ]: 668 : close (fd);
4504 : :
4505 [ + + ]: 668 : if (buildid == "")
4506 : : {
4507 : : // no point storing an elf file without buildid
4508 : 572 : executable_p = false;
4509 : 572 : debuginfo_p = false;
4510 : : }
4511 : : else
4512 : : {
4513 : : // register this build-id in the interning table
4514 : 96 : ps_upsert_buildids
4515 [ + - ]: 96 : .reset()
4516 [ + - ]: 96 : .bind(1, buildid)
4517 [ + - ]: 96 : .step_ok_done();
4518 : : }
4519 : :
4520 [ + + ]: 668 : if (executable_p)
4521 : 72 : fts_executable ++;
4522 [ + + ]: 668 : if (debuginfo_p)
4523 : 72 : fts_debuginfo ++;
4524 [ + + + + ]: 668 : if (executable_p || debuginfo_p)
4525 : : {
4526 : 96 : ps_upsert_de
4527 [ + - ]: 96 : .reset()
4528 [ + - ]: 96 : .bind(1, buildid)
4529 [ + + + - ]: 120 : .bind(2, debuginfo_p ? 1 : 0)
4530 [ + + + - ]: 120 : .bind(3, executable_p ? 1 : 0)
4531 [ + - ]: 96 : .bind(4, fileid)
4532 [ + - ]: 96 : .bind(5, st.st_mtime)
4533 [ + - ]: 96 : .step_ok_done();
4534 : : }
4535 [ + + ]: 668 : if (executable_p)
4536 [ + - + - : 144 : inc_metric("found_executable_total","source","files");
+ - + - ]
4537 [ + + ]: 668 : if (debuginfo_p)
4538 [ + - + - : 144 : inc_metric("found_debuginfo_total","source","files");
+ - + - ]
4539 : :
4540 [ + + + - ]: 668 : if (sourcefiles.size() && buildid != "")
4541 : : {
4542 : 72 : fts_sourcefiles += sourcefiles.size();
4543 : :
4544 [ + + ]: 15410 : for (auto&& dwarfsrc : sourcefiles)
4545 : : {
4546 [ + - ]: 15338 : char *srp = realpath(dwarfsrc.c_str(), NULL);
4547 [ - + ]: 15338 : if (srp == NULL) // also if DWZ unresolved dwarfsrc=""
4548 : 0 : continue; // unresolvable files are not a serious problem
4549 : : // throw libc_exception(errno, "fts/file realpath " + srcpath);
4550 [ + - ]: 15338 : string srps = string(srp);
4551 : 15338 : free (srp);
4552 : :
4553 : 15338 : struct stat sfs;
4554 : 15338 : rc = stat(srps.c_str(), &sfs);
4555 [ - + ]: 15338 : if (rc != 0)
4556 : 0 : continue;
4557 : :
4558 [ + - ]: 15338 : if (verbose > 2)
4559 [ + - + - ]: 46014 : obatched(clog) << "recorded buildid=" << buildid << " file=" << srps
4560 [ + - + - : 15338 : << " mtime=" << sfs.st_mtime
+ - + - ]
4561 [ + - + - : 15338 : << " as source " << dwarfsrc << endl;
+ - ]
4562 : :
4563 : : // PR25548: store canonicalized dwarfsrc path
4564 [ + - ]: 15338 : string dwarfsrc_canon = canon_pathname (dwarfsrc);
4565 [ + + ]: 15338 : if (dwarfsrc_canon != dwarfsrc)
4566 : : {
4567 [ + + ]: 2938 : if (verbose > 3)
4568 [ + - + - : 4568 : obatched(clog) << "canonicalized src=" << dwarfsrc << " alias=" << dwarfsrc_canon << endl;
+ - + - +
- ]
4569 : : }
4570 : :
4571 [ + - ]: 15338 : int64_t fileid1 = register_file_name (ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, dwarfsrc_canon);
4572 [ + - ]: 15338 : int64_t fileid2 = register_file_name (ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, srps);
4573 : :
4574 : 15338 : ps_upsert_s
4575 [ + - ]: 15338 : .reset()
4576 [ + - ]: 15338 : .bind(1, buildid)
4577 [ + - ]: 15332 : .bind(2, fileid1)
4578 [ + - ]: 15327 : .bind(3, fileid2)
4579 [ + - ]: 15334 : .bind(4, sfs.st_mtime)
4580 [ + - ]: 15332 : .step_ok_done();
4581 : :
4582 [ + - + - : 30676 : inc_metric("found_sourcerefs_total","source","files");
+ - + - ]
4583 : 15338 : }
4584 : : }
4585 : :
4586 : 668 : ps_scan_done
4587 [ + - ]: 668 : .reset()
4588 [ + - ]: 668 : .bind(1, fileid)
4589 [ + - ]: 668 : .bind(2, st.st_mtime)
4590 [ + - ]: 668 : .bind(3, st.st_size)
4591 [ + - ]: 668 : .step_ok_done();
4592 : :
4593 [ + - ]: 668 : if (verbose > 2)
4594 [ + - + - ]: 2004 : obatched(clog) << "recorded buildid=" << buildid << " file=" << rps
4595 [ + - + - : 668 : << " mtime=" << st.st_mtime << " atype="
+ - + - ]
4596 : : << (executable_p ? "E" : "")
4597 [ + - + + : 1860 : << (debuginfo_p ? "D" : "") << endl;
+ - + + +
- + - ]
4598 : 668 : }
4599 : :
4600 : :
4601 : :
4602 : :
4603 : :
4604 : : // Analyze given archive file of given age; record buildids / exec/debuginfo-ness of its
4605 : : // constituent files with given upsert statements.
4606 : : static void
4607 : 396 : archive_classify (const string& rps, string& archive_extension, int64_t archiveid,
4608 : : sqlite_ps& ps_upsert_buildids, sqlite_ps& ps_upsert_fileparts, sqlite_ps& ps_upsert_file,
4609 : : sqlite_ps& ps_lookup_file,
4610 : : sqlite_ps& ps_upsert_de, sqlite_ps& ps_upsert_sref, sqlite_ps& ps_upsert_sdef,
4611 : : sqlite_ps& ps_upsert_seekable,
4612 : : time_t mtime,
4613 : : unsigned& fts_executable, unsigned& fts_debuginfo, unsigned& fts_sref, unsigned& fts_sdef,
4614 : : bool& fts_sref_complete_p)
4615 : : {
4616 : 396 : string archive_decoder = "/dev/null";
4617 [ + + ]: 1024 : for (auto&& arch : scan_archives)
4618 [ + + ]: 628 : if (string_endswith(rps, arch.first))
4619 : : {
4620 [ + - ]: 396 : archive_extension = arch.first;
4621 [ + - ]: 1024 : archive_decoder = arch.second;
4622 : : }
4623 : :
4624 : 396 : FILE* fp;
4625 : 396 : defer_dtor<FILE*,int>::dtor_fn dfn;
4626 [ + + ]: 396 : if (archive_decoder != "cat")
4627 : : {
4628 [ + - + - : 80 : string popen_cmd = archive_decoder + " " + shell_escape(rps);
+ - ]
4629 [ + - ]: 40 : fp = popen (popen_cmd.c_str(), "r"); // "e" O_CLOEXEC?
4630 : 40 : dfn = pclose;
4631 [ - + ]: 40 : if (fp == NULL)
4632 [ # # # # : 0 : throw libc_exception (errno, string("popen ") + popen_cmd);
# # ]
4633 : 40 : }
4634 : : else
4635 : : {
4636 [ + - ]: 356 : fp = fopen (rps.c_str(), "r");
4637 : 356 : dfn = fclose;
4638 [ - + ]: 356 : if (fp == NULL)
4639 [ # # # # : 0 : throw libc_exception (errno, string("fopen ") + rps);
# # ]
4640 : : }
4641 : 396 : defer_dtor<FILE*,int> fp_closer (fp, dfn);
4642 : :
4643 : 396 : struct archive *a;
4644 [ + - ]: 396 : a = archive_read_new();
4645 [ - + ]: 396 : if (a == NULL)
4646 [ # # # # ]: 0 : throw archive_exception("cannot create archive reader");
4647 : 396 : defer_dtor<struct archive*,int> archive_closer (a, archive_read_free);
4648 : :
4649 [ + - ]: 396 : int rc = archive_read_support_format_all(a);
4650 [ - + ]: 396 : if (rc != ARCHIVE_OK)
4651 [ # # # # ]: 0 : throw archive_exception(a, "cannot select all formats");
4652 [ + - ]: 396 : rc = archive_read_support_filter_all(a);
4653 [ - + ]: 396 : if (rc != ARCHIVE_OK)
4654 [ # # # # ]: 0 : throw archive_exception(a, "cannot select all filters");
4655 : :
4656 [ + - ]: 396 : rc = archive_read_open_FILE (a, fp);
4657 [ - + ]: 396 : if (rc != ARCHIVE_OK)
4658 : : {
4659 [ # # # # : 0 : obatched(clog) << "cannot open archive from pipe " << rps << endl;
# # ]
4660 [ # # # # ]: 0 : throw archive_exception(a, "cannot open archive from pipe");
4661 : : }
4662 : :
4663 [ + + ]: 396 : if (verbose > 3)
4664 [ + - + - : 712 : obatched(clog) << "libarchive scanning " << rps << " id " << archiveid << endl;
+ - + - +
- ]
4665 : :
4666 [ + - ]: 396 : bool seekable = is_seekable_archive (rps, a);
4667 [ + - + + ]: 396 : if (verbose> 2 && seekable)
4668 [ + - + - : 64 : obatched(clog) << rps << " is seekable" << endl;
+ - ]
4669 : :
4670 : : bool any_exceptions = false;
4671 : 3774 : while(1) // parse archive entries
4672 : : {
4673 [ + - ]: 3774 : if (interrupted)
4674 : : break;
4675 : :
4676 : 3774 : try
4677 : : {
4678 : 3774 : struct archive_entry *e;
4679 [ + - ]: 3774 : rc = archive_read_next_header (a, &e);
4680 [ + + ]: 3774 : if (rc != ARCHIVE_OK)
4681 : : break;
4682 : :
4683 [ + - + + ]: 3378 : if (! S_ISREG(archive_entry_mode (e))) // skip non-files completely
4684 : 2338 : continue;
4685 : :
4686 [ + - ]: 1040 : string fn = canonicalized_archive_entry_pathname (e);
4687 : :
4688 [ + + ]: 1040 : if (verbose > 3)
4689 [ + - + - : 1756 : obatched(clog) << "libarchive checking " << fn << endl;
+ - ]
4690 : :
4691 [ + - ]: 1040 : int64_t seekable_size = archive_entry_size (e);
4692 [ + - ]: 1040 : int64_t seekable_offset = archive_filter_bytes (a, 0);
4693 [ + - ]: 1040 : time_t seekable_mtime = archive_entry_mtime (e);
4694 : :
4695 : : // extract this file to a temporary file
4696 : 1040 : char* tmppath = NULL;
4697 : 1040 : rc = asprintf (&tmppath, "%s/debuginfod-classify.XXXXXX", tmpdir.c_str());
4698 [ - + ]: 1040 : if (rc < 0)
4699 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate tmppath");
4700 : 1040 : defer_dtor<void*,void> tmmpath_freer (tmppath, free);
4701 [ + - ]: 1040 : int fd = mkstemp (tmppath);
4702 [ - + ]: 1040 : if (fd < 0)
4703 [ # # # # ]: 0 : throw libc_exception (errno, "cannot create temporary file");
4704 : 1040 : unlink (tmppath); // unlink now so OS will release the file as soon as we close the fd
4705 : 1040 : defer_dtor<int,int> minifd_closer (fd, close);
4706 : :
4707 [ + - ]: 1040 : rc = archive_read_data_into_fd (a, fd);
4708 [ - + ]: 1040 : if (rc != ARCHIVE_OK) {
4709 [ # # ]: 0 : close (fd);
4710 [ # # # # ]: 0 : throw archive_exception(a, rps, "cannot extract file");
4711 : : }
4712 : :
4713 : : // finally ... time to run elf_classify on this bad boy and update the database
4714 : 1040 : bool executable_p = false, debuginfo_p = false;
4715 [ + - ]: 1040 : string buildid;
4716 [ + - ]: 1040 : set<string> sourcefiles;
4717 [ + - ]: 1040 : elf_classify (fd, executable_p, debuginfo_p, buildid, sourcefiles);
4718 : : // NB: might throw
4719 : :
4720 [ + + ]: 1040 : if (buildid != "") // intern buildid
4721 : : {
4722 : 666 : ps_upsert_buildids
4723 [ + - ]: 666 : .reset()
4724 [ + - ]: 666 : .bind(1, buildid)
4725 [ + - ]: 666 : .step_ok_done();
4726 : : }
4727 : :
4728 [ + - ]: 1040 : int64_t fileid = register_file_name (ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, fn);
4729 : :
4730 [ + + ]: 1040 : if (sourcefiles.size() > 0) // sref records needed
4731 : : {
4732 : : // NB: we intern each source file once. Once raw, as it
4733 : : // appears in the DWARF file list coming back from
4734 : : // elf_classify() - because it'll end up in the
4735 : : // _norm.artifactsrc column. We don't also put another
4736 : : // version with a '.' at the front, even though that's
4737 : : // how rpm/cpio packs names, because we hide that from
4738 : : // the database for storage efficiency.
4739 : :
4740 [ + + ]: 802 : for (auto&& s : sourcefiles)
4741 : : {
4742 [ + + ]: 498 : if (s == "")
4743 : : {
4744 : 20 : fts_sref_complete_p = false;
4745 : 20 : continue;
4746 : : }
4747 : :
4748 : : // PR25548: store canonicalized source path
4749 : 478 : const string& dwarfsrc = s;
4750 [ + - ]: 478 : string dwarfsrc_canon = canon_pathname (dwarfsrc);
4751 [ + + ]: 478 : if (dwarfsrc_canon != dwarfsrc)
4752 : : {
4753 [ + - ]: 28 : if (verbose > 3)
4754 [ + - + - : 56 : obatched(clog) << "canonicalized src=" << dwarfsrc << " alias=" << dwarfsrc_canon << endl;
+ - + - +
- ]
4755 : : }
4756 : :
4757 [ + - ]: 478 : int64_t srcfileid = register_file_name(ps_upsert_fileparts, ps_upsert_file, ps_lookup_file,
4758 : : dwarfsrc_canon);
4759 : :
4760 : 478 : ps_upsert_sref
4761 [ + - ]: 478 : .reset()
4762 [ + - ]: 478 : .bind(1, buildid)
4763 [ + - ]: 478 : .bind(2, srcfileid)
4764 [ + - ]: 478 : .step_ok_done();
4765 : :
4766 : 478 : fts_sref ++;
4767 : 478 : }
4768 : : }
4769 : :
4770 [ + + ]: 1040 : if (executable_p)
4771 : 302 : fts_executable ++;
4772 [ + + ]: 1040 : if (debuginfo_p)
4773 : 366 : fts_debuginfo ++;
4774 : :
4775 [ + + + + ]: 1040 : if (executable_p || debuginfo_p)
4776 : : {
4777 : 666 : ps_upsert_de
4778 [ + - ]: 666 : .reset()
4779 [ + - ]: 666 : .bind(1, buildid)
4780 [ + + + - ]: 966 : .bind(2, debuginfo_p ? 1 : 0)
4781 [ + + + - ]: 1030 : .bind(3, executable_p ? 1 : 0)
4782 [ + - ]: 665 : .bind(4, archiveid)
4783 [ + - ]: 666 : .bind(5, mtime)
4784 [ + - ]: 666 : .bind(6, fileid)
4785 [ + - ]: 666 : .step_ok_done();
4786 [ + + ]: 666 : if (seekable)
4787 : 336 : ps_upsert_seekable
4788 [ + - ]: 336 : .reset()
4789 [ + - ]: 336 : .bind(1, archiveid)
4790 [ + - ]: 336 : .bind(2, fileid)
4791 [ + - ]: 336 : .bind(3, seekable_size)
4792 [ + - ]: 336 : .bind(4, seekable_offset)
4793 [ + - ]: 336 : .bind(5, seekable_mtime)
4794 [ + - ]: 336 : .step_ok_done();
4795 : : }
4796 : : else // potential source - sdef record
4797 : : {
4798 : 374 : fts_sdef ++;
4799 : 374 : ps_upsert_sdef
4800 [ + - ]: 374 : .reset()
4801 [ + - ]: 374 : .bind(1, archiveid)
4802 [ + - ]: 374 : .bind(2, mtime)
4803 [ + - ]: 374 : .bind(3, fileid)
4804 [ + - ]: 374 : .step_ok_done();
4805 : : }
4806 : :
4807 [ + - + + : 1040 : if ((verbose > 2) && (executable_p || debuginfo_p))
+ + ]
4808 : : {
4809 [ + - ]: 666 : obatched ob(clog);
4810 [ + - + - ]: 666 : auto& o = ob << "recorded buildid=" << buildid << " rpm=" << rps << " file=" << fn
4811 [ + - + - : 666 : << " mtime=" << mtime << " atype="
+ - + - +
- + - ]
4812 : : << (executable_p ? "E" : "")
4813 : : << (debuginfo_p ? "D" : "")
4814 [ + - + + : 1330 : << " sourcefiles=" << sourcefiles.size();
+ - + + +
- + - +
- ]
4815 [ + + ]: 666 : if (seekable)
4816 [ + - + - ]: 336 : o << " seekable size=" << seekable_size
4817 [ + - + - ]: 336 : << " offset=" << seekable_offset
4818 [ + - + - ]: 336 : << " mtime=" << seekable_mtime;
4819 [ + - ]: 666 : o << endl;
4820 : 666 : }
4821 : :
4822 : 1040 : }
4823 [ - - ]: 0 : catch (const reportable_exception& e)
4824 : : {
4825 [ - - ]: 0 : e.report(clog);
4826 : 0 : any_exceptions = true;
4827 : : // NB: but we allow the libarchive iteration to continue, in
4828 : : // case we can still gather some useful information. That
4829 : : // would allow some webapi queries to work, until later when
4830 : : // this archive is rescanned. (Its vitals won't go into the
4831 : : // _file_mtime_scanned table until after a successful scan.)
4832 : 0 : }
4833 : : }
4834 : :
4835 [ - + ]: 396 : if (any_exceptions)
4836 [ # # ]: 0 : throw reportable_exception("exceptions encountered during archive scan");
4837 : 396 : }
4838 : :
4839 : :
4840 : :
4841 : : // scan for archive files such as .rpm
4842 : : static void
4843 : 762 : scan_archive_file (const string& rps, const stat_t& st,
4844 : : sqlite_ps& ps_upsert_buildids,
4845 : : sqlite_ps& ps_upsert_fileparts,
4846 : : sqlite_ps& ps_upsert_file,
4847 : : sqlite_ps& ps_lookup_file,
4848 : : sqlite_ps& ps_upsert_de,
4849 : : sqlite_ps& ps_upsert_sref,
4850 : : sqlite_ps& ps_upsert_sdef,
4851 : : sqlite_ps& ps_upsert_seekable,
4852 : : sqlite_ps& ps_query,
4853 : : sqlite_ps& ps_scan_done,
4854 : : unsigned& fts_cached,
4855 : : unsigned& fts_executable,
4856 : : unsigned& fts_debuginfo,
4857 : : unsigned& fts_sref,
4858 : : unsigned& fts_sdef)
4859 : : {
4860 : : // intern the archive file name
4861 : 762 : int64_t archiveid = register_file_name (ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, rps);
4862 : :
4863 : : /* See if we know of it already. */
4864 : 762 : int rc = ps_query
4865 : 762 : .reset()
4866 : 762 : .bind(1, archiveid)
4867 : 762 : .bind(2, st.st_mtime)
4868 : 762 : .step();
4869 : 762 : ps_query.reset();
4870 [ + + ]: 762 : if (rc == SQLITE_ROW) // i.e., a result, as opposed to DONE (no results)
4871 : : // no need to recheck a file/version we already know
4872 : : // specifically, no need to parse this archive again, since we already have
4873 : : // it as a D or E or S record,
4874 : : // (so is stored with buildid=NULL)
4875 : : {
4876 : 366 : fts_cached ++;
4877 : 366 : return;
4878 : : }
4879 : :
4880 : : // extract the archive contents
4881 : 396 : unsigned my_fts_executable = 0, my_fts_debuginfo = 0, my_fts_sref = 0, my_fts_sdef = 0;
4882 : 396 : bool my_fts_sref_complete_p = true;
4883 : 396 : bool any_exceptions = false;
4884 : 396 : try
4885 : : {
4886 [ + - ]: 396 : string archive_extension;
4887 : 396 : archive_classify (rps, archive_extension, archiveid,
4888 : : ps_upsert_buildids, ps_upsert_fileparts, ps_upsert_file, ps_lookup_file,
4889 : : ps_upsert_de, ps_upsert_sref, ps_upsert_sdef, ps_upsert_seekable, // dalt
4890 [ + - ]: 396 : st.st_mtime,
4891 : : my_fts_executable, my_fts_debuginfo, my_fts_sref, my_fts_sdef,
4892 : : my_fts_sref_complete_p);
4893 [ + - + - : 792 : add_metric ("scanned_bytes_total","source",archive_extension + " archive",
+ - ]
4894 [ + - ]: 396 : st.st_size);
4895 [ + - + - : 792 : inc_metric ("scanned_files_total","source",archive_extension + " archive");
+ - + - ]
4896 [ + - + - : 792 : add_metric("found_debuginfo_total","source",archive_extension + " archive",
+ - + - ]
4897 : : my_fts_debuginfo);
4898 [ + - + - : 792 : add_metric("found_executable_total","source",archive_extension + " archive",
+ - + - ]
4899 : : my_fts_executable);
4900 [ + - + - : 792 : add_metric("found_sourcerefs_total","source",archive_extension + " archive",
+ - + - ]
4901 : : my_fts_sref);
4902 : 396 : }
4903 [ - - ]: 0 : catch (const reportable_exception& e)
4904 : : {
4905 [ - - ]: 0 : e.report(clog);
4906 : 0 : any_exceptions = true;
4907 : 0 : }
4908 : :
4909 [ + - ]: 396 : if (verbose > 2)
4910 [ + - ]: 1188 : obatched(clog) << "scanned archive=" << rps
4911 [ + - + - ]: 396 : << " mtime=" << st.st_mtime
4912 [ + - ]: 396 : << " executables=" << my_fts_executable
4913 [ + - + - ]: 396 : << " debuginfos=" << my_fts_debuginfo
4914 [ + - + - ]: 396 : << " srefs=" << my_fts_sref
4915 [ + - + - ]: 396 : << " sdefs=" << my_fts_sdef
4916 [ + - + - : 396 : << " exceptions=" << any_exceptions
+ - + - ]
4917 : 396 : << endl;
4918 : :
4919 : 396 : fts_executable += my_fts_executable;
4920 : 396 : fts_debuginfo += my_fts_debuginfo;
4921 : 396 : fts_sref += my_fts_sref;
4922 : 396 : fts_sdef += my_fts_sdef;
4923 : :
4924 [ - + ]: 396 : if (any_exceptions)
4925 [ # # ]: 0 : throw reportable_exception("exceptions encountered during archive scan");
4926 : :
4927 [ + + ]: 396 : if (my_fts_sref_complete_p) // leave incomplete?
4928 : 394 : ps_scan_done
4929 : 394 : .reset()
4930 : 394 : .bind(1, archiveid)
4931 : 394 : .bind(2, st.st_mtime)
4932 : 394 : .bind(3, st.st_size)
4933 : 394 : .step_ok_done();
4934 : : }
4935 : :
4936 : :
4937 : :
4938 : : ////////////////////////////////////////////////////////////////////////
4939 : :
4940 : :
4941 : :
4942 : : // The thread that consumes file names off of the scanq. We hold
4943 : : // the persistent sqlite_ps's at this level and delegate file/archive
4944 : : // scanning to other functions.
4945 : : static void
4946 : 288 : scan ()
4947 : : {
4948 : : // all the prepared statements fit to use, the _f_ set:
4949 [ + - + - ]: 576 : sqlite_ps ps_f_upsert_buildids (db, "file-buildids-intern", "insert or ignore into " BUILDIDS "_buildids VALUES (NULL, ?);");
4950 [ + - + - : 576 : sqlite_ps ps_f_upsert_fileparts (db, "file-fileparts-intern", "insert or ignore into " BUILDIDS "_fileparts VALUES (NULL, ?);");
+ - ]
4951 [ + - ]: 288 : sqlite_ps ps_f_upsert_file (db, "file-file-intern", "insert or ignore into " BUILDIDS "_files VALUES (NULL, \n"
4952 : : "(select id from " BUILDIDS "_fileparts where name = ?),\n"
4953 [ + - + - : 576 : "(select id from " BUILDIDS "_fileparts where name = ?));");
+ - ]
4954 [ + - ]: 288 : sqlite_ps ps_f_lookup_file (db, "file-file-lookup",
4955 : : "select f.id\n"
4956 : : " from " BUILDIDS "_files f, " BUILDIDS "_fileparts p1, " BUILDIDS "_fileparts p2 \n"
4957 [ + - + - : 576 : " where f.dirname = p1.id and f.basename = p2.id and p1.name = ? and p2.name = ?;\n");
+ - ]
4958 [ + - ]: 288 : sqlite_ps ps_f_upsert_de (db, "file-de-upsert",
4959 : : "insert or ignore into " BUILDIDS "_f_de "
4960 : : "(buildid, debuginfo_p, executable_p, file, mtime) "
4961 : : "values ((select id from " BUILDIDS "_buildids where hex = ?),"
4962 [ + - + - : 576 : " ?,?,?,?);");
+ - ]
4963 [ + - ]: 288 : sqlite_ps ps_f_upsert_s (db, "file-s-upsert",
4964 : : "insert or ignore into " BUILDIDS "_f_s "
4965 : : "(buildid, artifactsrc, file, mtime) "
4966 : : "values ((select id from " BUILDIDS "_buildids where hex = ?),"
4967 [ + - + - : 576 : " ?,?,?);");
+ - ]
4968 [ + - ]: 288 : sqlite_ps ps_f_query (db, "file-negativehit-find",
4969 : : "select 1 from " BUILDIDS "_file_mtime_scanned where sourcetype = 'F' "
4970 [ + - + - : 576 : "and file = ? and mtime = ?;");
+ - ]
4971 [ + - ]: 288 : sqlite_ps ps_f_scan_done (db, "file-scanned",
4972 : : "insert or ignore into " BUILDIDS "_file_mtime_scanned (sourcetype, file, mtime, size)"
4973 [ + - + - : 576 : "values ('F', ?,?,?);");
+ - ]
4974 : :
4975 : : // and now for the _r_ set
4976 [ + - + - : 576 : sqlite_ps ps_r_upsert_buildids (db, "rpm-buildid-intern", "insert or ignore into " BUILDIDS "_buildids VALUES (NULL, ?);");
+ - ]
4977 [ + - + - : 576 : sqlite_ps ps_r_upsert_fileparts (db, "rpm-fileparts-intern", "insert or ignore into " BUILDIDS "_fileparts VALUES (NULL, ?);");
+ - ]
4978 [ + - ]: 288 : sqlite_ps ps_r_upsert_file (db, "rpm-file-intern", "insert or ignore into " BUILDIDS "_files VALUES (NULL, \n"
4979 : : "(select id from " BUILDIDS "_fileparts where name = ?),\n"
4980 [ + - + - : 576 : "(select id from " BUILDIDS "_fileparts where name = ?));");
+ - ]
4981 [ + - ]: 288 : sqlite_ps ps_r_lookup_file (db, "rpm-file-lookup",
4982 : : "select f.id\n"
4983 : : " from " BUILDIDS "_files f, " BUILDIDS "_fileparts p1, " BUILDIDS "_fileparts p2 \n"
4984 [ + - + - : 576 : " where f.dirname = p1.id and f.basename = p2.id and p1.name = ? and p2.name = ?;\n");
+ - ]
4985 [ + - ]: 288 : sqlite_ps ps_r_upsert_de (db, "rpm-de-insert",
4986 : : "insert or ignore into " BUILDIDS "_r_de (buildid, debuginfo_p, executable_p, file, mtime, content) values ("
4987 [ + - + - : 576 : "(select id from " BUILDIDS "_buildids where hex = ?), ?, ?, ?, ?, ?);");
+ - ]
4988 [ + - ]: 288 : sqlite_ps ps_r_upsert_sref (db, "rpm-sref-insert",
4989 : : "insert or ignore into " BUILDIDS "_r_sref (buildid, artifactsrc) values ("
4990 : : "(select id from " BUILDIDS "_buildids where hex = ?), "
4991 [ + - + - : 576 : "?);");
+ - ]
4992 [ + - ]: 288 : sqlite_ps ps_r_upsert_sdef (db, "rpm-sdef-insert",
4993 : : "insert or ignore into " BUILDIDS "_r_sdef (file, mtime, content) values ("
4994 [ + - + - : 576 : "?, ?, ?);");
+ - ]
4995 [ + - ]: 288 : sqlite_ps ps_r_upsert_seekable (db, "rpm-seekable-insert",
4996 : : "insert or ignore into " BUILDIDS "_r_seekable (file, content, type, size, offset, mtime) "
4997 [ + - + - : 576 : "values (?, ?, 'xz', ?, ?, ?);");
+ - ]
4998 [ + - ]: 288 : sqlite_ps ps_r_query (db, "rpm-negativehit-query",
4999 : : "select 1 from " BUILDIDS "_file_mtime_scanned where "
5000 [ + - + - : 576 : "sourcetype = 'R' and file = ? and mtime = ?;");
+ - ]
5001 [ + - ]: 288 : sqlite_ps ps_r_scan_done (db, "rpm-scanned",
5002 : : "insert or ignore into " BUILDIDS "_file_mtime_scanned (sourcetype, file, mtime, size)"
5003 [ + - + - : 576 : "values ('R', ?, ?, ?);");
+ - ]
5004 : :
5005 : :
5006 : 288 : unsigned fts_cached = 0, fts_executable = 0, fts_debuginfo = 0, fts_sourcefiles = 0;
5007 : 288 : unsigned fts_sref = 0, fts_sdef = 0;
5008 : :
5009 [ + - + - : 576 : add_metric("thread_count", "role", "scan", 1);
+ - + - ]
5010 [ + - + - : 576 : add_metric("thread_busy", "role", "scan", 1);
+ - + - ]
5011 [ + + ]: 1890 : while (! interrupted)
5012 : : {
5013 [ + - ]: 1602 : scan_payload p;
5014 : :
5015 [ + - + - : 3204 : add_metric("thread_busy", "role", "scan", -1);
+ - + - ]
5016 : : // NB: threads may be blocked within either of these two waiting
5017 : : // states, if the work queue happens to run dry. That's OK.
5018 [ + - + - ]: 1602 : if (scan_barrier) scan_barrier->count();
5019 [ + - ]: 1602 : bool gotone = scanq.wait_front(p);
5020 [ + - + - : 3204 : add_metric("thread_busy", "role", "scan", 1);
+ - + - ]
5021 : :
5022 [ + + ]: 1602 : if (! gotone) continue; // go back to waiting
5023 : :
5024 : 1314 : try
5025 : : {
5026 : 1314 : bool scan_archive = false;
5027 [ + + ]: 2886 : for (auto&& arch : scan_archives)
5028 [ + + ]: 1572 : if (string_endswith(p.first, arch.first))
5029 : 762 : scan_archive = true;
5030 : :
5031 [ + + ]: 1314 : if (scan_archive)
5032 [ + - ]: 762 : scan_archive_file (p.first, p.second,
5033 : : ps_r_upsert_buildids,
5034 : : ps_r_upsert_fileparts,
5035 : : ps_r_upsert_file,
5036 : : ps_r_lookup_file,
5037 : : ps_r_upsert_de,
5038 : : ps_r_upsert_sref,
5039 : : ps_r_upsert_sdef,
5040 : : ps_r_upsert_seekable,
5041 : : ps_r_query,
5042 : : ps_r_scan_done,
5043 : : fts_cached,
5044 : : fts_executable,
5045 : : fts_debuginfo,
5046 : : fts_sref,
5047 : : fts_sdef);
5048 : :
5049 [ + + ]: 1314 : if (scan_files) // NB: maybe "else if" ?
5050 [ + - ]: 1106 : scan_source_file (p.first, p.second,
5051 : : ps_f_upsert_buildids,
5052 : : ps_f_upsert_fileparts,
5053 : : ps_f_upsert_file,
5054 : : ps_f_lookup_file,
5055 : : ps_f_upsert_de,
5056 : : ps_f_upsert_s,
5057 : : ps_f_query,
5058 : : ps_f_scan_done,
5059 : : fts_cached, fts_executable, fts_debuginfo, fts_sourcefiles);
5060 : : }
5061 [ - - ]: 0 : catch (const reportable_exception& e)
5062 : : {
5063 [ - - ]: 0 : e.report(cerr);
5064 : 0 : }
5065 : :
5066 [ + - ]: 1314 : scanq.done_front(); // let idlers run
5067 : :
5068 : 1314 : if (fts_cached || fts_executable || fts_debuginfo || fts_sourcefiles || fts_sref || fts_sdef)
5069 : : {} // NB: not just if a successful scan - we might have encountered -ENOSPC & failed
5070 [ + - + - ]: 1314 : (void) statfs_free_enough_p(db_path, "database"); // report sqlite filesystem size
5071 [ + - + - ]: 1314 : (void) statfs_free_enough_p(tmpdir, "tmpdir"); // this too, in case of fdcache/tmpfile usage
5072 : :
5073 : : // finished a scanning step -- not a "loop", because we just
5074 : : // consume the traversal loop's work, whenever
5075 [ + - + - : 2628 : inc_metric("thread_work_total","role","scan");
+ - + - ]
5076 : 1602 : }
5077 : :
5078 [ + - + - : 576 : add_metric("thread_busy", "role", "scan", -1);
+ - + - ]
5079 : 288 : }
5080 : :
5081 : :
5082 : : // Use this function as the thread entry point, so it can catch our
5083 : : // fleet of exceptions (incl. the sqlite_ps ctors) and report.
5084 : : static void*
5085 : 288 : thread_main_scanner (void* arg)
5086 : : {
5087 : 288 : (void) arg;
5088 [ + + ]: 864 : while (! interrupted)
5089 : 288 : try
5090 : : {
5091 [ + - ]: 288 : scan();
5092 : : }
5093 [ - - ]: 0 : catch (const reportable_exception& e)
5094 : : {
5095 [ - - ]: 0 : e.report(cerr);
5096 : 0 : }
5097 : 288 : return 0;
5098 : : }
5099 : :
5100 : :
5101 : :
5102 : : // The thread that traverses all the source_paths and enqueues all the
5103 : : // matching files into the file/archive scan queue.
5104 : : static void
5105 : 126 : scan_source_paths()
5106 : : {
5107 : : // NB: fedora 31 glibc/fts(3) crashes inside fts_read() on empty
5108 : : // path list.
5109 [ + + ]: 126 : if (source_paths.empty())
5110 : 2 : return;
5111 : :
5112 : : // Turn the source_paths into an fts(3)-compatible char**. Since
5113 : : // source_paths[] does not change after argv processing, the
5114 : : // c_str()'s are safe to keep around awile.
5115 : 124 : vector<const char *> sps;
5116 [ + + ]: 332 : for (auto&& sp: source_paths)
5117 [ + - ]: 208 : sps.push_back(sp.c_str());
5118 [ + - - - ]: 124 : sps.push_back(NULL);
5119 : :
5120 [ + + + - ]: 234 : FTS *fts = fts_open ((char * const *)sps.data(),
5121 : : (traverse_logical ? FTS_LOGICAL : FTS_PHYSICAL|FTS_XDEV)
5122 : : | FTS_NOCHDIR /* multithreaded */,
5123 : : NULL);
5124 [ - + ]: 124 : if (fts == NULL)
5125 [ # # # # ]: 0 : throw libc_exception(errno, "cannot fts_open");
5126 : 124 : defer_dtor<FTS*,int> fts_cleanup (fts, fts_close);
5127 : :
5128 : 124 : struct timespec ts_start, ts_end;
5129 : 124 : clock_gettime (CLOCK_MONOTONIC, &ts_start);
5130 : 124 : unsigned fts_scanned = 0, fts_regex = 0;
5131 : :
5132 : 124 : FTSENT *f;
5133 [ + - + + ]: 2644 : while ((f = fts_read (fts)) != NULL)
5134 : : {
5135 [ + - ]: 2396 : if (interrupted) break;
5136 : :
5137 [ - + ]: 2396 : if (sigusr2 != forced_groom_count) // stop early if groom triggered
5138 : : {
5139 [ # # ]: 0 : scanq.clear(); // clear previously issued work for scanner threads
5140 : : break;
5141 : : }
5142 : :
5143 : 2396 : fts_scanned ++;
5144 : :
5145 [ + - ]: 2396 : if (verbose > 2)
5146 [ + - + - : 4792 : obatched(clog) << "fts traversing " << f->fts_path << endl;
+ - ]
5147 : :
5148 [ + + + + : 2396 : switch (f->fts_info)
+ ]
5149 : : {
5150 : 1424 : case FTS_F:
5151 : 1424 : {
5152 : : /* Found a file. Convert it to an absolute path, so
5153 : : the buildid database does not have relative path
5154 : : names that are unresolvable from a subsequent run
5155 : : in a different cwd. */
5156 [ + - ]: 1424 : char *rp = realpath(f->fts_path, NULL);
5157 [ - + ]: 1424 : if (rp == NULL)
5158 : 0 : continue; // ignore dangling symlink or such
5159 [ + - ]: 1424 : string rps = string(rp);
5160 : 1424 : free (rp);
5161 : :
5162 [ + - ]: 1424 : bool ri = !regexec (&file_include_regex, rps.c_str(), 0, 0, 0);
5163 [ + - ]: 1424 : bool rx = !regexec (&file_exclude_regex, rps.c_str(), 0, 0, 0);
5164 [ + + ]: 1424 : if (!ri || rx)
5165 : : {
5166 [ + - ]: 110 : if (verbose > 3)
5167 [ + - ]: 220 : obatched(clog) << "fts skipped by regex "
5168 [ + + + - : 134 : << (!ri ? "I" : "") << (rx ? "X" : "") << endl;
+ + + - +
- ]
5169 : 110 : fts_regex ++;
5170 [ + + ]: 110 : if (!ri)
5171 [ + - + - : 24 : inc_metric("traversed_total","type","file-skipped-I");
+ - + - ]
5172 [ + + ]: 110 : if (rx)
5173 [ + - + - : 196 : inc_metric("traversed_total","type","file-skipped-X");
+ - + - ]
5174 : : }
5175 : : else
5176 : : {
5177 [ + - + - ]: 1314 : scanq.push_back (make_pair(rps, *f->fts_statp));
5178 [ + - + - : 2628 : inc_metric("traversed_total","type","file");
+ - + - ]
5179 : : }
5180 : 0 : }
5181 : 1424 : break;
5182 : :
5183 : 4 : case FTS_ERR:
5184 : 4 : case FTS_NS:
5185 : : // report on some types of errors because they may reflect fixable misconfiguration
5186 : 4 : {
5187 [ + - + - : 8 : auto x = libc_exception(f->fts_errno, string("fts traversal ") + string(f->fts_path));
+ - + - ]
5188 [ + - ]: 4 : x.report(cerr);
5189 : 0 : }
5190 [ + - + - : 8 : inc_metric("traversed_total","type","error");
+ - + - ]
5191 : 4 : break;
5192 : :
5193 : 32 : case FTS_SL: // ignore, but count because debuginfod -L would traverse these
5194 [ + - + - : 64 : inc_metric("traversed_total","type","symlink");
+ - + - ]
5195 : 32 : break;
5196 : :
5197 : 468 : case FTS_D: // ignore
5198 [ + - + - : 936 : inc_metric("traversed_total","type","directory");
+ - + - ]
5199 : 468 : break;
5200 : :
5201 : 468 : default: // ignore
5202 [ + - + - : 936 : inc_metric("traversed_total","type","other");
+ - + - ]
5203 : 468 : break;
5204 : : }
5205 : : }
5206 : 124 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
5207 : 124 : double deltas = (ts_end.tv_sec - ts_start.tv_sec) + (ts_end.tv_nsec - ts_start.tv_nsec)/1.e9;
5208 : :
5209 [ + - + - : 372 : obatched(clog) << "fts traversed source paths in " << deltas << "s, scanned=" << fts_scanned
+ - + - ]
5210 [ + - + - : 124 : << ", regex-skipped=" << fts_regex << endl;
+ - ]
5211 [ + - ]: 248 : }
5212 : :
5213 : :
5214 : : static void*
5215 : 72 : thread_main_fts_source_paths (void* arg)
5216 : : {
5217 : 72 : (void) arg; // ignore; we operate on global data
5218 : :
5219 [ + - + - : 144 : set_metric("thread_tid", "role","traverse", tid());
+ - ]
5220 [ + - + - : 144 : add_metric("thread_count", "role", "traverse", 1);
+ - ]
5221 : :
5222 : 72 : time_t last_rescan = 0;
5223 : :
5224 [ + - ]: 294 : while (! interrupted)
5225 : : {
5226 : 294 : sleep (1);
5227 : 294 : scanq.wait_idle(); // don't start a new traversal while scanners haven't finished the job
5228 : 294 : scanq.done_idle(); // release the hounds
5229 [ + + ]: 294 : if (interrupted) break;
5230 : :
5231 : 222 : time_t now = time(NULL);
5232 : 222 : bool rescan_now = false;
5233 [ + + ]: 222 : if (last_rescan == 0) // at least one initial rescan is documented even for -t0
5234 : 70 : rescan_now = true;
5235 [ + + + + ]: 222 : if (rescan_s > 0 && (long)now > (long)(last_rescan + rescan_s))
5236 : 222 : rescan_now = true;
5237 [ + + ]: 222 : if (sigusr1 != forced_rescan_count)
5238 : : {
5239 : 58 : forced_rescan_count = sigusr1;
5240 : 58 : rescan_now = true;
5241 : : }
5242 [ + + ]: 222 : if (rescan_now)
5243 : : {
5244 [ + - + - : 252 : set_metric("thread_busy", "role","traverse", 1);
+ - ]
5245 : 126 : try
5246 : : {
5247 [ + - ]: 126 : scan_source_paths();
5248 : : }
5249 [ - - ]: 0 : catch (const reportable_exception& e)
5250 : : {
5251 [ - - ]: 0 : e.report(cerr);
5252 : 0 : }
5253 : 126 : last_rescan = time(NULL); // NB: now was before scanning
5254 : : // finished a traversal loop
5255 [ + - + - : 252 : inc_metric("thread_work_total", "role","traverse");
+ - ]
5256 [ + - + - : 252 : set_metric("thread_busy", "role","traverse", 0);
+ - ]
5257 : : }
5258 : : }
5259 : :
5260 : 72 : return 0;
5261 : : }
5262 : :
5263 : :
5264 : :
5265 : : ////////////////////////////////////////////////////////////////////////
5266 : :
5267 : : static void
5268 : 78 : database_stats_report()
5269 : : {
5270 : 78 : sqlite_ps ps_query (db, "database-overview",
5271 [ + - + - ]: 156 : "select label,quantity from " BUILDIDS "_stats");
5272 : :
5273 [ + - + - ]: 156 : obatched(clog) << "database record counts:" << endl;
5274 : 1794 : while (1)
5275 : : {
5276 [ + - ]: 936 : if (interrupted) break;
5277 [ + - ]: 936 : if (sigusr1 != forced_rescan_count) // stop early if scan triggered
5278 : : break;
5279 : :
5280 [ + - ]: 936 : int rc = ps_query.step();
5281 [ + + ]: 936 : if (rc == SQLITE_DONE) break;
5282 [ - + ]: 858 : if (rc != SQLITE_ROW)
5283 [ # # # # ]: 0 : throw sqlite_exception(rc, "step");
5284 : :
5285 [ + - ]: 858 : obatched(clog)
5286 [ + - - + : 858 : << ((const char*) sqlite3_column_text(ps_query, 0) ?: (const char*) "NULL")
+ - ]
5287 : : << " "
5288 [ + - + - : 1716 : << (sqlite3_column_text(ps_query, 1) ?: (const unsigned char*) "NULL")
- + + - ]
5289 : 858 : << endl;
5290 : :
5291 [ + - + - : 1716 : set_metric("groom", "statistic",
- + + - +
- + - +
- ]
5292 [ + - ]: 858 : ((const char*) sqlite3_column_text(ps_query, 0) ?: (const char*) "NULL"),
5293 : : (sqlite3_column_double(ps_query, 1)));
5294 : 858 : }
5295 : 78 : }
5296 : :
5297 : :
5298 : : // Do a round of database grooming that might take many minutes to run.
5299 : 78 : void groom()
5300 : : {
5301 [ + - ]: 156 : obatched(clog) << "grooming database" << endl;
5302 : :
5303 : 78 : struct timespec ts_start, ts_end;
5304 : 78 : clock_gettime (CLOCK_MONOTONIC, &ts_start);
5305 : :
5306 : : // scan for files that have disappeared
5307 : 78 : sqlite_ps files (db, "check old files",
5308 : : "select distinct s.mtime, s.file, f.name from "
5309 : : BUILDIDS "_file_mtime_scanned s, " BUILDIDS "_files_v f "
5310 [ + - + - ]: 156 : "where f.id = s.file");
5311 : : // NB: Because _ftime_mtime_scanned can contain both F and
5312 : : // R records for the same file, this query would return duplicates if the
5313 : : // DISTINCT qualifier were not there.
5314 [ + - ]: 78 : files.reset();
5315 : :
5316 : : // DECISION TIME - we enumerate stale fileids/mtimes
5317 [ + - ]: 78 : deque<pair<int64_t,int64_t> > stale_fileid_mtime;
5318 : :
5319 : 78 : time_t time_start = time(NULL);
5320 : 342 : while(1)
5321 : : {
5322 : : // PR28514: limit grooming iteration to O(rescan time), to avoid
5323 : : // slow filesystem tests over many files locking out rescans for
5324 : : // too long.
5325 [ + + - + ]: 210 : if (rescan_s > 0 && (long)time(NULL) > (long)(time_start + rescan_s))
5326 : : {
5327 [ # # # # : 0 : inc_metric("groomed_total", "decision", "aborted");
# # # # ]
5328 : 0 : break;
5329 : : }
5330 : :
5331 [ + - ]: 210 : if (interrupted) break;
5332 : :
5333 [ + - ]: 210 : int rc = files.step();
5334 [ + + ]: 210 : if (rc != SQLITE_ROW)
5335 : : break;
5336 : :
5337 [ + - ]: 132 : int64_t mtime = sqlite3_column_int64 (files, 0);
5338 [ + - ]: 132 : int64_t fileid = sqlite3_column_int64 (files, 1);
5339 [ + - - + ]: 132 : const char* filename = ((const char*) sqlite3_column_text (files, 2) ?: "");
5340 : 132 : struct stat s;
5341 : 132 : bool regex_file_drop = 0;
5342 : :
5343 [ + + ]: 132 : if (regex_groom)
5344 : : {
5345 [ + - ]: 16 : bool reg_include = !regexec (&file_include_regex, filename, 0, 0, 0);
5346 [ + - ]: 16 : bool reg_exclude = !regexec (&file_exclude_regex, filename, 0, 0, 0);
5347 : 16 : regex_file_drop = !reg_include || reg_exclude; // match logic of scan_source_paths
5348 : : }
5349 : :
5350 : 132 : rc = stat(filename, &s);
5351 [ + + - + ]: 132 : if ( regex_file_drop || rc < 0 || (mtime != (int64_t) s.st_mtime) )
5352 : : {
5353 [ + - ]: 24 : if (verbose > 2)
5354 [ + - + - : 48 : obatched(clog) << "groom: stale file=" << filename << " mtime=" << mtime << endl;
+ - + - +
- ]
5355 [ + - ]: 24 : stale_fileid_mtime.push_back(make_pair(fileid,mtime));
5356 [ + - + - : 48 : inc_metric("groomed_total", "decision", "stale");
+ - + - ]
5357 [ + - + - : 48 : set_metric("thread_work_pending","role","groom", stale_fileid_mtime.size());
+ - + - ]
5358 : : }
5359 : : else
5360 [ + - + - : 216 : inc_metric("groomed_total", "decision", "fresh");
+ - + - ]
5361 : :
5362 [ + - ]: 132 : if (sigusr1 != forced_rescan_count) // stop early if scan triggered
5363 : : break;
5364 : 132 : }
5365 [ + - ]: 78 : files.reset();
5366 : :
5367 : : // ACTION TIME
5368 : :
5369 : : // Now that we know which file/mtime tuples are stale, actually do
5370 : : // the deletion from the database. Doing this during the SELECT
5371 : : // iteration above results in undefined behaviour in sqlite, as per
5372 : : // https://www.sqlite.org/isolation.html
5373 : :
5374 : : // We could shuffle stale_fileid_mtime[] here. It'd let aborted
5375 : : // sequences of nuke operations resume at random locations, instead
5376 : : // of just starting over. But it doesn't matter much either way,
5377 : : // as long as we make progress.
5378 : :
5379 [ + - + - : 156 : sqlite_ps files_del_f_de (db, "nuke f_de", "delete from " BUILDIDS "_f_de where file = ? and mtime = ?");
+ - ]
5380 [ + - + - : 156 : sqlite_ps files_del_r_de (db, "nuke r_de", "delete from " BUILDIDS "_r_de where file = ? and mtime = ?");
+ - ]
5381 [ + - ]: 78 : sqlite_ps files_del_scan (db, "nuke f_m_s", "delete from " BUILDIDS "_file_mtime_scanned "
5382 [ + - + - : 156 : "where file = ? and mtime = ?");
+ - ]
5383 : :
5384 [ + + ]: 102 : while (! stale_fileid_mtime.empty())
5385 : : {
5386 : 24 : auto stale = stale_fileid_mtime.front();
5387 : 24 : stale_fileid_mtime.pop_front();
5388 [ + - + - : 48 : set_metric("thread_work_pending","role","groom", stale_fileid_mtime.size());
+ - + - ]
5389 : :
5390 : : // PR28514: limit grooming iteration to O(rescan time), to avoid
5391 : : // slow nuke_* queries over many files locking out rescans for too
5392 : : // long. We iterate over the files in random() sequence to avoid
5393 : : // partial checks going over the same set.
5394 [ - + - - ]: 24 : if (rescan_s > 0 && (long)time(NULL) > (long)(time_start + rescan_s))
5395 : : {
5396 [ # # # # : 0 : inc_metric("groomed_total", "action", "aborted");
# # # # ]
5397 : 0 : break;
5398 : : }
5399 : :
5400 [ + - ]: 24 : if (interrupted) break;
5401 : :
5402 : 24 : int64_t fileid = stale.first;
5403 : 24 : int64_t mtime = stale.second;
5404 [ + - + - : 24 : files_del_f_de.reset().bind(1,fileid).bind(2,mtime).step_ok_done();
+ - + - ]
5405 [ + - + - : 24 : files_del_r_de.reset().bind(1,fileid).bind(2,mtime).step_ok_done();
+ - + - ]
5406 [ + - + - : 24 : files_del_scan.reset().bind(1,fileid).bind(2,mtime).step_ok_done();
+ - + - ]
5407 [ + - + - : 48 : inc_metric("groomed_total", "action", "cleaned");
+ - + - ]
5408 : :
5409 [ + - ]: 24 : if (sigusr1 != forced_rescan_count) // stop early if scan triggered
5410 : : break;
5411 : : }
5412 : 78 : stale_fileid_mtime.clear(); // no need for this any longer
5413 [ + - + - : 156 : set_metric("thread_work_pending","role","groom", stale_fileid_mtime.size());
+ - + - ]
5414 : :
5415 : : // delete buildids with no references in _r_de or _f_de tables;
5416 : : // cascades to _r_sref & _f_s records
5417 [ + - ]: 78 : sqlite_ps buildids_del (db, "nuke orphan buildids",
5418 : : "delete from " BUILDIDS "_buildids "
5419 : : "where not exists (select 1 from " BUILDIDS "_f_de d where " BUILDIDS "_buildids.id = d.buildid) "
5420 [ + - + - : 156 : "and not exists (select 1 from " BUILDIDS "_r_de d where " BUILDIDS "_buildids.id = d.buildid)");
+ - ]
5421 [ + - + - ]: 78 : buildids_del.reset().step_ok_done();
5422 : :
5423 [ - + ]: 78 : if (interrupted) return;
5424 : :
5425 : : // NB: "vacuum" is too heavy for even daily runs: it rewrites the entire db, so is done as maxigroom -G
5426 [ + - + - : 156 : { sqlite_ps g (db, "incremental vacuum", "pragma incremental_vacuum"); g.reset().step_ok_done(); }
+ - + - +
- ]
5427 : : // https://www.sqlite.org/lang_analyze.html#approx
5428 [ + - + - : 156 : { sqlite_ps g (db, "analyze setup", "pragma analysis_limit = 1000;\n"); g.reset().step_ok_done(); }
+ - + - +
- ]
5429 [ + - + - : 156 : { sqlite_ps g (db, "analyze", "analyze"); g.reset().step_ok_done(); }
+ - + - +
- ]
5430 [ + - + - : 156 : { sqlite_ps g (db, "analyze reload", "analyze sqlite_schema"); g.reset().step_ok_done(); }
+ - + - +
- ]
5431 [ + - + - : 156 : { sqlite_ps g (db, "optimize", "pragma optimize"); g.reset().step_ok_done(); }
+ - + - +
- ]
5432 [ + - + - : 156 : { sqlite_ps g (db, "wal checkpoint", "pragma wal_checkpoint=truncate"); g.reset().step_ok_done(); }
+ - + - +
- ]
5433 : :
5434 [ + - ]: 78 : database_stats_report();
5435 : :
5436 [ + - + - ]: 78 : (void) statfs_free_enough_p(db_path, "database"); // report sqlite filesystem size
5437 : :
5438 [ + - ]: 78 : sqlite3_db_release_memory(db); // shrink the process if possible
5439 [ + - ]: 78 : sqlite3_db_release_memory(dbq); // ... for both connections
5440 [ + - ]: 78 : debuginfod_pool_groom(); // and release any debuginfod_client objects we've been holding onto
5441 : : #if HAVE_MALLOC_TRIM
5442 : 78 : malloc_trim(0); // PR31103: release memory allocated for temporary purposes
5443 : : #endif
5444 : :
5445 : : #if 0 /* PR31265: don't jettison cache unnecessarily */
5446 : : fdcache.limit(0); // release the fdcache contents
5447 : : fdcache.limit(fdcache_mbs); // restore status quo parameters
5448 : : #endif
5449 : :
5450 : 78 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
5451 : 78 : double deltas = (ts_end.tv_sec - ts_start.tv_sec) + (ts_end.tv_nsec - ts_start.tv_nsec)/1.e9;
5452 : :
5453 [ + - + - : 156 : obatched(clog) << "groomed database in " << deltas << "s" << endl;
+ - + - ]
5454 : 78 : }
5455 : :
5456 : :
5457 : : static void*
5458 : 82 : thread_main_groom (void* /*arg*/)
5459 : : {
5460 [ + - + - : 164 : set_metric("thread_tid", "role", "groom", tid());
+ - ]
5461 [ + - + - : 164 : add_metric("thread_count", "role", "groom", 1);
+ - ]
5462 : :
5463 : 82 : time_t last_groom = 0;
5464 : :
5465 : 530 : while (1)
5466 : : {
5467 : 306 : sleep (1);
5468 : 306 : scanq.wait_idle(); // PR25394: block scanners during grooming!
5469 [ + + ]: 306 : if (interrupted) break;
5470 : :
5471 : 224 : time_t now = time(NULL);
5472 : 224 : bool groom_now = false;
5473 [ + + ]: 224 : if (last_groom == 0) // at least one initial groom is documented even for -g0
5474 : 72 : groom_now = true;
5475 [ + + + + ]: 224 : if (groom_s > 0 && (long)now > (long)(last_groom + groom_s))
5476 : 224 : groom_now = true;
5477 [ + + ]: 224 : if (sigusr2 != forced_groom_count)
5478 : : {
5479 : 6 : forced_groom_count = sigusr2;
5480 : 6 : groom_now = true;
5481 : : }
5482 [ + + ]: 224 : if (groom_now)
5483 : : {
5484 [ + - + - : 156 : set_metric("thread_busy", "role", "groom", 1);
+ - ]
5485 : 78 : try
5486 : : {
5487 [ + - ]: 78 : groom ();
5488 : : }
5489 [ - - ]: 0 : catch (const sqlite_exception& e)
5490 : : {
5491 [ - - - - : 0 : obatched(cerr) << e.message << endl;
- - ]
5492 : 0 : }
5493 : 78 : last_groom = time(NULL); // NB: now was before grooming
5494 : : // finished a grooming loop
5495 [ + - + - : 156 : inc_metric("thread_work_total", "role", "groom");
+ - ]
5496 [ + - + - : 156 : set_metric("thread_busy", "role", "groom", 0);
+ - ]
5497 : : }
5498 : :
5499 : 224 : scanq.done_idle();
5500 : 224 : }
5501 : :
5502 : 82 : return 0;
5503 : : }
5504 : :
5505 : :
5506 : : ////////////////////////////////////////////////////////////////////////
5507 : :
5508 : :
5509 : : static void
5510 : 84 : signal_handler (int /* sig */)
5511 : : {
5512 : 84 : interrupted ++;
5513 : :
5514 [ + + ]: 84 : if (db)
5515 : 82 : sqlite3_interrupt (db);
5516 [ + - ]: 84 : if (dbq)
5517 : 84 : sqlite3_interrupt (dbq);
5518 : :
5519 : : // NB: don't do anything else in here
5520 : 84 : }
5521 : :
5522 : : static void
5523 : 58 : sigusr1_handler (int /* sig */)
5524 : : {
5525 : 58 : sigusr1 ++;
5526 : : // NB: don't do anything else in here
5527 : 58 : }
5528 : :
5529 : : static void
5530 : 6 : sigusr2_handler (int /* sig */)
5531 : : {
5532 : 6 : sigusr2 ++;
5533 : : // NB: don't do anything else in here
5534 : 6 : }
5535 : :
5536 : :
5537 : : static void // error logging callback from libmicrohttpd internals
5538 : 0 : error_cb (void *arg, const char *fmt, va_list ap)
5539 : : {
5540 : 0 : (void) arg;
5541 [ # # # # : 0 : inc_metric("error_count","libmicrohttpd",fmt);
# # ]
5542 : 0 : char errmsg[512];
5543 : 0 : (void) vsnprintf (errmsg, sizeof(errmsg), fmt, ap); // ok if slightly truncated
5544 [ # # ]: 0 : obatched(cerr) << "libmicrohttpd error: " << errmsg; // MHD_DLOG calls already include \n
5545 : 0 : }
5546 : :
5547 : :
5548 : : // A user-defined sqlite function, to score the sharedness of the
5549 : : // prefix of two strings. This is used to compare candidate debuginfo
5550 : : // / source-rpm names, so that the closest match
5551 : : // (directory-topology-wise closest) is found. This is important in
5552 : : // case the same sref (source file name) is in many -debuginfo or
5553 : : // -debugsource RPMs, such as when multiple versions/releases of the
5554 : : // same package are in the database.
5555 : :
5556 : 1344 : static void sqlite3_sharedprefix_fn (sqlite3_context* c, int argc, sqlite3_value** argv)
5557 : : {
5558 [ - + ]: 1344 : if (argc != 2)
5559 : 0 : sqlite3_result_error(c, "expect 2 string arguments", -1);
5560 [ + - + + ]: 2688 : else if ((sqlite3_value_type(argv[0]) != SQLITE_TEXT) ||
5561 : 1344 : (sqlite3_value_type(argv[1]) != SQLITE_TEXT))
5562 : 1082 : sqlite3_result_null(c);
5563 : : else
5564 : : {
5565 : 262 : const unsigned char* a = sqlite3_value_text (argv[0]);
5566 : 262 : const unsigned char* b = sqlite3_value_text (argv[1]);
5567 : 262 : int i = 0;
5568 [ + + + - : 27722 : while (*a != '\0' && *b != '\0' && *a++ == *b++)
+ + ]
5569 : 27198 : i++;
5570 : 262 : sqlite3_result_int (c, i);
5571 : : }
5572 : 1344 : }
5573 : :
5574 : :
5575 : : static unsigned
5576 : 164 : default_concurrency() // guaranteed >= 1
5577 : : {
5578 : : // Prior to PR29975 & PR29976, we'd just use this:
5579 : 164 : unsigned sth = std::thread::hardware_concurrency();
5580 : : // ... but on many-CPU boxes, admins or distros may throttle
5581 : : // resources in such a way that debuginfod would mysteriously fail.
5582 : : // So we reduce the defaults:
5583 : :
5584 : 164 : unsigned aff = 0;
5585 : : #ifdef HAVE_SCHED_GETAFFINITY
5586 : 164 : {
5587 : 164 : int ret;
5588 : 164 : cpu_set_t mask;
5589 : 164 : CPU_ZERO(&mask);
5590 : 164 : ret = sched_getaffinity(0, sizeof(mask), &mask);
5591 [ + - ]: 164 : if (ret == 0)
5592 : 164 : aff = CPU_COUNT(&mask);
5593 : : }
5594 : : #endif
5595 : :
5596 : 164 : unsigned fn = 0;
5597 : : #ifdef HAVE_GETRLIMIT
5598 : 164 : {
5599 : 164 : struct rlimit rlim;
5600 : 164 : int rc = getrlimit(RLIMIT_NOFILE, &rlim);
5601 [ + - ]: 164 : if (rc == 0)
5602 [ - + ]: 164 : fn = max((rlim_t)1, (rlim.rlim_cur - 100) / 4);
5603 : : // at least 2 fds are used by each listener thread etc.
5604 : : // plus a bunch to account for shared libraries and such
5605 : : }
5606 : : #endif
5607 : :
5608 [ - + + - : 328 : unsigned d = min(max(sth, 1U),
- + ]
5609 [ + - ]: 164 : min(max(aff, 1U),
5610 [ - + ]: 164 : max(fn, 1U)));
5611 : 164 : return d;
5612 : : }
5613 : :
5614 : :
5615 : : // 30879: Something to help out in case of an uncaught exception.
5616 : 0 : void my_terminate_handler()
5617 : : {
5618 : : #if defined(__GLIBC__)
5619 : 0 : void *array[40];
5620 : 0 : int size = backtrace (array, 40);
5621 : 0 : backtrace_symbols_fd (array, size, STDERR_FILENO);
5622 : : #endif
5623 : : #if defined(__GLIBCXX__) || defined(__GLIBCPP__)
5624 : 0 : __gnu_cxx::__verbose_terminate_handler();
5625 : : #endif
5626 : 0 : abort();
5627 : : }
5628 : :
5629 : :
5630 : : int
5631 : 84 : main (int argc, char *argv[])
5632 : : {
5633 : 84 : (void) setlocale (LC_ALL, "");
5634 : 84 : (void) bindtextdomain (PACKAGE_TARNAME, LOCALEDIR);
5635 : 84 : (void) textdomain (PACKAGE_TARNAME);
5636 : :
5637 : 84 : std::set_terminate(& my_terminate_handler);
5638 : :
5639 : : /* Tell the library which version we are expecting. */
5640 : 84 : elf_version (EV_CURRENT);
5641 : :
5642 [ + - ]: 168 : tmpdir = string(getenv("TMPDIR") ?: "/tmp");
5643 : :
5644 : : /* Set computed default values. */
5645 [ - + + - : 84 : db_path = string(getenv("HOME") ?: "/") + string("/.debuginfod.sqlite"); /* XDG? */
+ - ]
5646 : 84 : int rc = regcomp (& file_include_regex, ".*", REG_EXTENDED|REG_NOSUB); // match everything
5647 [ - + ]: 84 : if (rc != 0)
5648 : 0 : error (EXIT_FAILURE, 0, "regcomp failure: %d", rc);
5649 : 84 : rc = regcomp (& file_exclude_regex, "^$", REG_EXTENDED|REG_NOSUB); // match nothing
5650 [ - + ]: 84 : if (rc != 0)
5651 : 0 : error (EXIT_FAILURE, 0, "regcomp failure: %d", rc);
5652 : :
5653 : : // default parameters for fdcache are computed from system stats
5654 : 84 : struct statfs sfs;
5655 : 84 : rc = statfs(tmpdir.c_str(), &sfs);
5656 [ - + ]: 84 : if (rc < 0)
5657 : 0 : fdcache_mbs = 1024; // 1 gigabyte
5658 : : else
5659 : 84 : fdcache_mbs = sfs.f_bavail * sfs.f_bsize / 1024 / 1024 / 4; // 25% of free space
5660 : 84 : fdcache_mintmp = 25; // emergency flush at 25% remaining (75% full)
5661 : 84 : fdcache_prefetch = 64; // guesstimate storage is this much less costly than re-decompression
5662 : :
5663 : : /* Parse and process arguments. */
5664 : 84 : memset(&http_sockaddr, 0, sizeof(http_sockaddr));
5665 : 84 : http_sockaddr.sin6_family = AF_UNSPEC;
5666 : 84 : int remaining;
5667 : 84 : (void) argp_parse (&argp, argc, argv, ARGP_IN_ORDER, &remaining, NULL);
5668 [ - + ]: 84 : if (remaining != argc)
5669 : 0 : error (EXIT_FAILURE, 0,
5670 : 0 : "unexpected argument: %s", argv[remaining]);
5671 : :
5672 [ + + + + : 84 : if (scan_archives.size()==0 && !scan_files && source_paths.size()>0)
- + ]
5673 [ # # ]: 0 : obatched(clog) << "warning: without -F -R -U -Z, ignoring PATHs" << endl;
5674 : :
5675 : 84 : fdcache.limit(fdcache_mbs);
5676 : :
5677 : 84 : (void) signal (SIGPIPE, SIG_IGN); // microhttpd can generate it incidentally, ignore
5678 : 84 : (void) signal (SIGINT, signal_handler); // ^C
5679 : 84 : (void) signal (SIGHUP, signal_handler); // EOF
5680 : 84 : (void) signal (SIGTERM, signal_handler); // systemd
5681 : 84 : (void) signal (SIGUSR1, sigusr1_handler); // end-user
5682 : 84 : (void) signal (SIGUSR2, sigusr2_handler); // end-user
5683 : :
5684 : : /* Get database ready. */
5685 [ + + ]: 84 : if (! passive_p)
5686 : : {
5687 : 82 : rc = sqlite3_open_v2 (db_path.c_str(), &db, (SQLITE_OPEN_READWRITE
5688 : : |SQLITE_OPEN_URI
5689 : : |SQLITE_OPEN_PRIVATECACHE
5690 : : |SQLITE_OPEN_CREATE
5691 : : |SQLITE_OPEN_FULLMUTEX), /* thread-safe */
5692 : : NULL);
5693 [ - + ]: 82 : if (rc == SQLITE_CORRUPT)
5694 : : {
5695 : 0 : (void) unlink (db_path.c_str());
5696 : 0 : error (EXIT_FAILURE, 0,
5697 : : "cannot open %s, deleted database: %s", db_path.c_str(), sqlite3_errmsg(db));
5698 : : }
5699 [ - + ]: 82 : else if (rc)
5700 : : {
5701 : 0 : error (EXIT_FAILURE, 0,
5702 : : "cannot open %s, consider deleting database: %s", db_path.c_str(), sqlite3_errmsg(db));
5703 : : }
5704 : : }
5705 : :
5706 : : // open the readonly query variant
5707 : : // NB: PRIVATECACHE allows web queries to operate in parallel with
5708 : : // much other grooming/scanning operation.
5709 : 84 : rc = sqlite3_open_v2 (db_path.c_str(), &dbq, (SQLITE_OPEN_READONLY
5710 : : |SQLITE_OPEN_URI
5711 : : |SQLITE_OPEN_PRIVATECACHE
5712 : : |SQLITE_OPEN_FULLMUTEX), /* thread-safe */
5713 : : NULL);
5714 [ - + ]: 84 : if (rc)
5715 : : {
5716 : 0 : error (EXIT_FAILURE, 0,
5717 : : "cannot open %s, consider deleting database: %s", db_path.c_str(), sqlite3_errmsg(dbq));
5718 : : }
5719 : :
5720 : :
5721 [ + - ]: 168 : obatched(clog) << "opened database " << db_path
5722 [ + + + - : 86 : << (db?" rw":"") << (dbq?" ro":"") << endl;
- + + - +
- ]
5723 [ + - + - ]: 168 : obatched(clog) << "sqlite version " << sqlite3_version << endl;
5724 [ + + + - : 250 : obatched(clog) << "service mode " << (passive_p ? "passive":"active") << endl;
+ - ]
5725 : :
5726 : : // add special string-prefix-similarity function used in rpm sref/sdef resolution
5727 : 84 : rc = sqlite3_create_function(dbq, "sharedprefix", 2, SQLITE_UTF8, NULL,
5728 : : & sqlite3_sharedprefix_fn, NULL, NULL);
5729 [ - + ]: 84 : if (rc != SQLITE_OK)
5730 : 0 : error (EXIT_FAILURE, 0,
5731 : : "cannot create sharedprefix function: %s", sqlite3_errmsg(dbq));
5732 : :
5733 [ + + ]: 84 : if (! passive_p)
5734 : : {
5735 [ + + ]: 82 : if (verbose > 3)
5736 [ + - + - ]: 92 : obatched(clog) << "ddl: " << DEBUGINFOD_SQLITE_DDL << endl;
5737 : 82 : rc = sqlite3_exec (db, DEBUGINFOD_SQLITE_DDL, NULL, NULL, NULL);
5738 [ - + ]: 82 : if (rc != SQLITE_OK)
5739 : : {
5740 : 0 : error (EXIT_FAILURE, 0,
5741 : : "cannot run database schema ddl: %s", sqlite3_errmsg(db));
5742 : : }
5743 : : }
5744 : :
5745 [ + - + - : 168 : obatched(clog) << "libmicrohttpd version " << MHD_get_version() << endl;
+ - ]
5746 : :
5747 : : /* If '-C' wasn't given or was given with no arg, pick a reasonable default
5748 : : for the number of worker threads. */
5749 [ + + ]: 84 : if (connection_pool == 0)
5750 : 80 : connection_pool = default_concurrency();
5751 : :
5752 : : /* Note that MHD_USE_EPOLL and MHD_USE_THREAD_PER_CONNECTION don't
5753 : : work together. */
5754 : 84 : unsigned int use_epoll = 0;
5755 : : #if MHD_VERSION >= 0x00095100
5756 : 84 : use_epoll = MHD_USE_EPOLL;
5757 : : #endif
5758 : :
5759 : 84 : unsigned int mhd_flags = (
5760 : : #if MHD_VERSION >= 0x00095300
5761 : : MHD_USE_INTERNAL_POLLING_THREAD
5762 : : #else
5763 : : MHD_USE_SELECT_INTERNALLY
5764 : : #endif
5765 : : | MHD_USE_DUAL_STACK
5766 : : | use_epoll
5767 : : #if MHD_VERSION >= 0x00095200
5768 : : | MHD_USE_ITC
5769 : : #endif
5770 : : | MHD_USE_DEBUG); /* report errors to stderr */
5771 : :
5772 : 84 : MHD_Daemon *dsa = NULL,
5773 : 84 : *d4 = NULL,
5774 : 84 : *d46 = NULL;
5775 : :
5776 [ - + ]: 84 : if (http_sockaddr.sin6_family != AF_UNSPEC)
5777 : : {
5778 [ # # ]: 0 : if (http_sockaddr.sin6_family == AF_INET)
5779 : 0 : ((sockaddr_in*)&http_sockaddr)->sin_port = htons(http_port);
5780 [ # # ]: 0 : if (http_sockaddr.sin6_family == AF_INET6)
5781 : 0 : http_sockaddr.sin6_port = htons(http_port);
5782 : : // Start httpd server threads on socket addr:port.
5783 : 0 : dsa = MHD_start_daemon (mhd_flags & ~MHD_USE_DUAL_STACK, http_port,
5784 : : NULL, NULL, /* default accept policy */
5785 : : handler_cb, NULL, /* handler callback */
5786 : : MHD_OPTION_EXTERNAL_LOGGER,
5787 : : error_cb, NULL,
5788 : : MHD_OPTION_SOCK_ADDR,
5789 : : (struct sockaddr *) &http_sockaddr,
5790 : : MHD_OPTION_THREAD_POOL_SIZE,
5791 : : (int)connection_pool,
5792 : : MHD_OPTION_END);
5793 : : }
5794 : : else
5795 : : {
5796 : : // Start httpd server threads. Use a single dual-homed pool.
5797 : 84 : d46 = MHD_start_daemon (mhd_flags, http_port,
5798 : : NULL, NULL, /* default accept policy */
5799 : : handler_cb, NULL, /* handler callback */
5800 : : MHD_OPTION_EXTERNAL_LOGGER,
5801 : : error_cb, NULL,
5802 : : MHD_OPTION_THREAD_POOL_SIZE,
5803 : : (int)connection_pool,
5804 : : MHD_OPTION_END);
5805 : 84 : addr_info = "IPv4 IPv6";
5806 [ - + ]: 84 : if (d46 == NULL)
5807 : : {
5808 : : // Cannot use dual_stack, use ipv4 only
5809 : 0 : mhd_flags &= ~(MHD_USE_DUAL_STACK);
5810 [ # # ]: 0 : d4 = MHD_start_daemon (mhd_flags, http_port,
5811 : : NULL, NULL, /* default accept policy */
5812 : : handler_cb, NULL, /* handler callback */
5813 : : MHD_OPTION_EXTERNAL_LOGGER,
5814 : : error_cb, NULL,
5815 : : (connection_pool
5816 : : ? MHD_OPTION_THREAD_POOL_SIZE
5817 : : : MHD_OPTION_END),
5818 : : (connection_pool
5819 : : ? (int)connection_pool
5820 : : : MHD_OPTION_END),
5821 : : MHD_OPTION_END);
5822 : 0 : addr_info = "IPv4";
5823 : : }
5824 : : }
5825 [ - + - - ]: 84 : if (d4 == NULL && d46 == NULL && dsa == NULL)
5826 : : {
5827 : 0 : sqlite3 *database = db;
5828 : 0 : sqlite3 *databaseq = dbq;
5829 : 0 : db = dbq = 0; // for signal_handler not to freak
5830 : 0 : sqlite3_close (databaseq);
5831 : 0 : sqlite3_close (database);
5832 : 0 : error (EXIT_FAILURE, 0, "cannot start http server on %s port %d",
5833 : : addr_info.c_str(), http_port);
5834 : : }
5835 : :
5836 [ + - ]: 168 : obatched(clog) << "started http server on "
5837 : : << addr_info
5838 [ + - ]: 84 : << " port=" << http_port
5839 [ + - + + : 162 : << (webapi_cors ? " with cors" : "")
+ - + - ]
5840 : 84 : << endl;
5841 : :
5842 : : // add maxigroom sql if -G given
5843 [ - + ]: 84 : if (maxigroom)
5844 : : {
5845 [ # # ]: 0 : obatched(clog) << "maxigrooming database, please wait." << endl;
5846 : : // NB: this index alone can nearly double the database size!
5847 : : // NB: this index would be necessary to run source-file metadata searches fast
5848 : 0 : extra_ddl.push_back("create index if not exists " BUILDIDS "_r_sref_arc on " BUILDIDS "_r_sref(artifactsrc);");
5849 : 0 : extra_ddl.push_back("delete from " BUILDIDS "_r_sdef where not exists (select 1 from " BUILDIDS "_r_sref b where " BUILDIDS "_r_sdef.content = b.artifactsrc);");
5850 : 0 : extra_ddl.push_back("drop index if exists " BUILDIDS "_r_sref_arc;");
5851 : :
5852 : : // NB: we don't maxigroom the _files interning table. It'd require a temp index on all the
5853 : : // tables that have file foreign-keys, which is a lot.
5854 : :
5855 : : // NB: with =delete, may take up 3x disk space total during vacuum process
5856 : : // vs. =off (only 2x but may corrupt database if program dies mid-vacuum)
5857 : : // vs. =wal (>3x observed, but safe)
5858 : 0 : extra_ddl.push_back("pragma journal_mode=delete;");
5859 : 0 : extra_ddl.push_back("vacuum;");
5860 : 0 : extra_ddl.push_back("pragma journal_mode=wal;");
5861 : : }
5862 : :
5863 : : // run extra -D sql if given
5864 [ + + ]: 84 : if (! passive_p)
5865 [ - + ]: 82 : for (auto&& i: extra_ddl)
5866 : : {
5867 [ # # ]: 0 : if (verbose > 1)
5868 [ # # # # ]: 0 : obatched(clog) << "extra ddl:\n" << i << endl;
5869 : 0 : rc = sqlite3_exec (db, i.c_str(), NULL, NULL, NULL);
5870 [ # # # # ]: 0 : if (rc != SQLITE_OK && rc != SQLITE_DONE && rc != SQLITE_ROW)
5871 : 0 : error (0, 0,
5872 : : "warning: cannot run database extra ddl %s: %s", i.c_str(), sqlite3_errmsg(db));
5873 : :
5874 [ # # ]: 0 : if (maxigroom)
5875 [ # # ]: 0 : obatched(clog) << "maxigroomed database" << endl;
5876 : : }
5877 : :
5878 [ + + ]: 84 : if (! passive_p)
5879 [ + - + - ]: 164 : obatched(clog) << "search concurrency " << concurrency << endl;
5880 : 84 : obatched(clog) << "webapi connection pool " << connection_pool
5881 [ + - - + : 84 : << (connection_pool ? "" : " (unlimited)") << endl;
+ - + - ]
5882 [ + + ]: 84 : if (! passive_p) {
5883 [ + - + - ]: 164 : obatched(clog) << "rescan time " << rescan_s << endl;
5884 [ + - + - ]: 164 : obatched(clog) << "scan checkpoint " << scan_checkpoint << endl;
5885 : : }
5886 [ + - + - ]: 168 : obatched(clog) << "fdcache mbs " << fdcache_mbs << endl;
5887 [ + - + - ]: 168 : obatched(clog) << "fdcache prefetch " << fdcache_prefetch << endl;
5888 [ + - + - ]: 168 : obatched(clog) << "fdcache tmpdir " << tmpdir << endl;
5889 [ + - + - ]: 168 : obatched(clog) << "fdcache tmpdir min% " << fdcache_mintmp << endl;
5890 [ + + ]: 84 : if (! passive_p)
5891 [ + - + - ]: 164 : obatched(clog) << "groom time " << groom_s << endl;
5892 [ + - + - ]: 168 : obatched(clog) << "forwarded ttl limit " << forwarded_ttl_limit << endl;
5893 : :
5894 [ + + ]: 84 : if (scan_archives.size()>0)
5895 : : {
5896 : 56 : obatched ob(clog);
5897 [ + - ]: 56 : auto& o = ob << "accepting archive types ";
5898 [ + + ]: 174 : for (auto&& arch : scan_archives)
5899 [ + - + - : 118 : o << arch.first << "(" << arch.second << ") ";
+ - + - ]
5900 [ + - ]: 56 : o << endl;
5901 : 56 : }
5902 : 84 : const char* du = getenv(DEBUGINFOD_URLS_ENV_VAR);
5903 [ + + + + ]: 84 : if (du && du[0] != '\0') // set to non-empty string?
5904 [ + - + - ]: 32 : obatched(clog) << "upstream debuginfod servers: " << du << endl;
5905 : :
5906 [ + + ]: 84 : vector<pthread_t> all_threads;
5907 : :
5908 [ + + ]: 84 : if (! passive_p)
5909 : : {
5910 : 82 : pthread_t pt;
5911 : 82 : rc = pthread_create (& pt, NULL, thread_main_groom, NULL);
5912 [ - + ]: 82 : if (rc)
5913 : 0 : error (EXIT_FAILURE, rc, "cannot spawn thread to groom database\n");
5914 : : else
5915 : : {
5916 : : #ifdef HAVE_PTHREAD_SETNAME_NP
5917 : 82 : (void) pthread_setname_np (pt, "groom");
5918 : : #endif
5919 [ + - ]: 82 : all_threads.push_back(pt);
5920 : : }
5921 : :
5922 [ + + + + ]: 82 : if (scan_files || scan_archives.size() > 0)
5923 : : {
5924 [ + - ]: 72 : if (scan_checkpoint > 0)
5925 [ + - ]: 72 : scan_barrier = new sqlite_checkpoint_pb(concurrency, (unsigned) scan_checkpoint);
5926 : :
5927 : 72 : rc = pthread_create (& pt, NULL, thread_main_fts_source_paths, NULL);
5928 [ - + ]: 72 : if (rc)
5929 : 0 : error (EXIT_FAILURE, rc, "cannot spawn thread to traverse source paths\n");
5930 : : #ifdef HAVE_PTHREAD_SETNAME_NP
5931 : 72 : (void) pthread_setname_np (pt, "traverse");
5932 : : #endif
5933 [ + - ]: 72 : all_threads.push_back(pt);
5934 : :
5935 [ + + ]: 360 : for (unsigned i=0; i<concurrency; i++)
5936 : : {
5937 : 288 : rc = pthread_create (& pt, NULL, thread_main_scanner, NULL);
5938 [ - + ]: 288 : if (rc)
5939 : 0 : error (EXIT_FAILURE, rc, "cannot spawn thread to scan source files / archives\n");
5940 : : #ifdef HAVE_PTHREAD_SETNAME_NP
5941 : 288 : (void) pthread_setname_np (pt, "scan");
5942 : : #endif
5943 [ + - ]: 288 : all_threads.push_back(pt);
5944 : : }
5945 : : }
5946 : : }
5947 : :
5948 : : /* Trivial main loop! */
5949 [ + - + - ]: 84 : set_metric("ready", 1);
5950 [ + + ]: 232 : while (! interrupted)
5951 [ + - ]: 148 : pause ();
5952 [ + - ]: 84 : scanq.nuke(); // wake up any remaining scanq-related threads, let them die
5953 [ + + + - ]: 84 : if (scan_barrier) scan_barrier->nuke(); // ... in case they're stuck in a barrier
5954 [ + - + - ]: 84 : set_metric("ready", 0);
5955 : :
5956 [ + - ]: 84 : if (verbose)
5957 [ + - + - : 168 : obatched(clog) << "stopping" << endl;
- - ]
5958 : :
5959 : : /* Join all our threads. */
5960 [ + + ]: 526 : for (auto&& it : all_threads)
5961 [ + - ]: 442 : pthread_join (it, NULL);
5962 : :
5963 : : /* Stop all the web service threads. */
5964 [ - + - - ]: 84 : if (dsa) MHD_stop_daemon (dsa);
5965 [ + - + - ]: 84 : if (d46) MHD_stop_daemon (d46);
5966 [ - + - - ]: 84 : if (d4) MHD_stop_daemon (d4);
5967 : :
5968 [ + + ]: 84 : if (! passive_p)
5969 : : {
5970 : : /* With all threads known dead, we can clean up the global resources. */
5971 [ + - ]: 82 : rc = sqlite3_exec (db, DEBUGINFOD_SQLITE_CLEANUP_DDL, NULL, NULL, NULL);
5972 [ - + ]: 82 : if (rc != SQLITE_OK)
5973 : : {
5974 [ # # ]: 0 : error (0, 0,
5975 : : "warning: cannot run database cleanup ddl: %s", sqlite3_errmsg(db));
5976 : : }
5977 : : }
5978 : :
5979 [ + - ]: 84 : debuginfod_pool_groom ();
5980 [ + + ]: 84 : delete scan_barrier;
5981 : :
5982 : : // NB: no problem with unconditional free here - an earlier failed regcomp would exit program
5983 [ + - ]: 84 : (void) regfree (& file_include_regex);
5984 [ + - ]: 84 : (void) regfree (& file_exclude_regex);
5985 : :
5986 : 84 : sqlite3 *database = db;
5987 : 84 : sqlite3 *databaseq = dbq;
5988 : 84 : db = dbq = 0; // for signal_handler not to freak
5989 [ + - ]: 84 : (void) sqlite3_close (databaseq);
5990 [ + + ]: 84 : if (! passive_p)
5991 [ + - ]: 82 : (void) sqlite3_close (database);
5992 : :
5993 [ + + ]: 84 : return 0;
5994 : 84 : }
|
