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 : 141666 : string_endswith(const string& haystack, const string& needle)
162 : : {
163 [ + + ]: 141666 : return (haystack.size() >= needle.size() &&
164 : 138574 : equal(haystack.end()-needle.size(), haystack.end(),
165 : 141666 : 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 : 3378 : tmp_inc_metric(const string& mname, const string& lname, const string& lvalue):
576 [ + - + - ]: 3378 : m(mname), n(lname), v(lvalue)
577 : : {
578 [ + - ]: 3378 : add_metric (m, n, v, 1);
579 : 3378 : }
580 : 3378 : ~tmp_inc_metric()
581 : : {
582 : 3378 : add_metric (m, n, v, -1);
583 : 3378 : }
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 : 360751 : tmp_ms_metric(const string& mname, const string& lname, const string& lvalue):
591 [ + - + - ]: 360751 : m(mname), n(lname), v(lvalue)
592 : : {
593 : 360767 : clock_gettime (CLOCK_MONOTONIC, & ts_start);
594 : 360952 : }
595 : 361067 : ~tmp_ms_metric()
596 : : {
597 : 361067 : struct timespec ts_end;
598 : 361067 : clock_gettime (CLOCK_MONOTONIC, & ts_end);
599 : 361110 : double deltas = (ts_end.tv_sec - ts_start.tv_sec)
600 : 361110 : + (ts_end.tv_nsec - ts_start.tv_nsec)/1.e9;
601 : :
602 : 361110 : add_metric (m + "_milliseconds_sum", n, v, (deltas*1000.0));
603 : 361164 : inc_metric (m + "_milliseconds_count", n, v);
604 : 361146 : }
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 [ - - - - : 605 : 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 : 633 : MHD_RESULT mhd_send_response(MHD_Connection* c) const {
809 : 1266 : MHD_Response* r = MHD_create_response_from_buffer (message.size(),
810 : 633 : (void*) message.c_str(),
811 : : MHD_RESPMEM_MUST_COPY);
812 : 633 : add_mhd_response_header (r, "Content-Type", "text/plain");
813 : 633 : MHD_RESULT rc = MHD_queue_response (c, code, r);
814 : 633 : MHD_destroy_response (r);
815 : 633 : 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 : : };
848 : :
849 : :
850 : : struct elfutils_exception: public reportable_exception
851 : : {
852 : 0 : elfutils_exception(int rc, const string& msg):
853 [ # # # # : 0 : reportable_exception(string("elfutils error: ") + msg + ": " + string(elf_errmsg(rc) ?: "?")) {
# # # # ]
854 [ # # # # : 0 : inc_metric("error_count","elfutils",elf_errmsg(rc));
# # # # #
# ]
855 : 0 : }
856 : : };
857 : :
858 : :
859 : : ////////////////////////////////////////////////////////////////////////
860 : :
861 : : template <typename Payload>
862 : : class workq
863 : : {
864 : : unordered_set<Payload> q; // eliminate duplicates
865 : : mutex mtx;
866 : : condition_variable cv;
867 : : bool dead;
868 : : unsigned idlers; // number of threads busy with wait_idle / done_idle
869 : : unsigned fronters; // number of threads busy with wait_front / done_front
870 : :
871 : : public:
872 : 84 : workq() { dead = false; idlers = 0; fronters = 0; }
873 : 84 : ~workq() {}
874 : :
875 : 1314 : void push_back(const Payload& p)
876 : : {
877 : 1314 : unique_lock<mutex> lock(mtx);
878 [ + - ]: 1314 : q.insert (p);
879 [ + - + - : 2628 : set_metric("thread_work_pending","role","scan", q.size());
+ - + - ]
880 : 1314 : cv.notify_all();
881 : 1314 : }
882 : :
883 : : // kill this workqueue, wake up all idlers / scanners
884 : 84 : void nuke() {
885 : 84 : unique_lock<mutex> lock(mtx);
886 : : // optional: q.clear();
887 : 84 : dead = true;
888 : 84 : cv.notify_all();
889 : 84 : }
890 : :
891 : : // clear the workqueue, when scanning is interrupted with USR2
892 : 0 : void clear() {
893 : 0 : unique_lock<mutex> lock(mtx);
894 : 0 : q.clear();
895 [ # # # # : 0 : set_metric("thread_work_pending","role","scan", q.size());
# # # # ]
896 : : // NB: there may still be some live fronters
897 : 0 : cv.notify_all(); // maybe wake up waiting idlers
898 : 0 : }
899 : :
900 : : // block this scanner thread until there is work to do and no active idler
901 : 1602 : bool wait_front (Payload& p)
902 : : {
903 : 1602 : unique_lock<mutex> lock(mtx);
904 [ + + + + : 5920 : while (!dead && (q.size() == 0 || idlers > 0))
+ + ]
905 [ + - ]: 4318 : cv.wait(lock);
906 [ + + ]: 1602 : if (dead)
907 : : return false;
908 : : else
909 : : {
910 [ + - ]: 1314 : p = * q.begin();
911 : 1314 : q.erase (q.begin());
912 : 1314 : fronters ++; // prevent idlers from starting awhile, even if empty q
913 [ + - + - : 2628 : set_metric("thread_work_pending","role","scan", q.size());
+ - + - -
- ]
914 : : // NB: don't wake up idlers yet! The consumer is busy
915 : : // processing this element until it calls done_front().
916 : 1314 : return true;
917 : : }
918 : 1602 : }
919 : :
920 : : // notify waitq that scanner thread is done with that last item
921 : 1314 : void done_front ()
922 : : {
923 : 1314 : unique_lock<mutex> lock(mtx);
924 : 1314 : fronters --;
925 [ + + + + ]: 1314 : if (q.size() == 0 && fronters == 0)
926 : 100 : cv.notify_all(); // maybe wake up waiting idlers
927 : 1314 : }
928 : :
929 : : // block this idler thread until there is no work to do
930 : 681 : void wait_idle ()
931 : : {
932 : 681 : unique_lock<mutex> lock(mtx);
933 : 681 : cv.notify_all(); // maybe wake up waiting scanners
934 [ + + + + : 761 : while (!dead && ((q.size() != 0) || fronters > 0))
+ + ]
935 [ + - ]: 80 : cv.wait(lock);
936 [ + - ]: 681 : idlers ++;
937 : 681 : }
938 : :
939 : 599 : void done_idle ()
940 : : {
941 : 599 : unique_lock<mutex> lock(mtx);
942 : 599 : idlers --;
943 : 599 : cv.notify_all(); // maybe wake up waiting scanners, but probably not (shutting down)
944 : 599 : }
945 : : };
946 : :
947 : : typedef struct stat stat_t;
948 : : typedef pair<string,stat_t> scan_payload;
949 : : inline bool operator< (const scan_payload& a, const scan_payload& b)
950 : : {
951 : : return a.first < b.first; // don't bother compare the stat fields
952 : : }
953 : :
954 : : namespace std { // https://gcc.gnu.org/bugzilla/show_bug.cgi?id=56480
955 : : template<> struct hash<::scan_payload>
956 : : {
957 : 5371 : std::size_t operator() (const ::scan_payload& p) const noexcept
958 : : {
959 [ + + + + ]: 5371 : return hash<string>()(p.first);
960 : : }
961 : : };
962 : : template<> struct equal_to<::scan_payload>
963 : : {
964 : 546 : std::size_t operator() (const ::scan_payload& a, const ::scan_payload& b) const noexcept
965 : : {
966 [ + - - - ]: 546 : return a.first == b.first;
967 : : }
968 : : };
969 : : }
970 : :
971 : : static workq<scan_payload> scanq; // just a single one
972 : : // producer & idler: thread_main_fts_source_paths()
973 : : // consumer: thread_main_scanner()
974 : : // idler: thread_main_groom()
975 : :
976 : :
977 : : ////////////////////////////////////////////////////////////////////////
978 : :
979 : : // Unique set is a thread-safe structure that lends 'ownership' of a value
980 : : // to a thread. Other threads requesting the same thing are made to wait.
981 : : // It's like a semaphore-on-demand.
982 : : template <typename T>
983 : : class unique_set
984 : : {
985 : : private:
986 : : set<T> values;
987 : : mutex mtx;
988 : : condition_variable cv;
989 : : public:
990 : 60 : unique_set() {}
991 : 60 : ~unique_set() {}
992 : :
993 : 2597 : void acquire(const T& value)
994 : : {
995 : 2597 : unique_lock<mutex> lock(mtx);
996 [ + + ]: 7119 : while (values.find(value) != values.end())
997 [ + - ]: 4522 : cv.wait(lock);
998 [ + - ]: 2597 : values.insert(value);
999 : 2597 : }
1000 : :
1001 : 2597 : void release(const T& value)
1002 : : {
1003 : 2597 : unique_lock<mutex> lock(mtx);
1004 : : // assert (values.find(value) != values.end());
1005 : 2597 : values.erase(value);
1006 : 2597 : cv.notify_all();
1007 : 2597 : }
1008 : : };
1009 : :
1010 : :
1011 : : // This is the object that's instantiate to uniquely hold a value in a
1012 : : // RAII-pattern way.
1013 : : template <typename T>
1014 : : class unique_set_reserver
1015 : : {
1016 : : private:
1017 : : unique_set<T>& please_hold;
1018 : : T mine;
1019 : : public:
1020 : 2597 : unique_set_reserver(unique_set<T>& t, const T& value):
1021 [ + - ]: 2597 : please_hold(t), mine(value) { please_hold.acquire(mine); }
1022 : 2597 : ~unique_set_reserver() { please_hold.release(mine); }
1023 : : };
1024 : :
1025 : :
1026 : : ////////////////////////////////////////////////////////////////////////
1027 : :
1028 : : // periodic_barrier is a concurrency control object that lets N threads
1029 : : // periodically (based on counter value) agree to wait at a barrier,
1030 : : // let one of them carry out some work, then be set free
1031 : :
1032 : : class periodic_barrier
1033 : : {
1034 : : private:
1035 : : unsigned period; // number of count() reports to trigger barrier activation
1036 : : unsigned threads; // number of threads participating
1037 : : mutex mtx; // protects all the following fields
1038 : : unsigned counter; // count of count() reports in the current generation
1039 : : unsigned generation; // barrier activation generation
1040 : : unsigned waiting; // number of threads waiting for barrier
1041 : : bool dead; // bring out your
1042 : : condition_variable cv;
1043 : : public:
1044 : 72 : periodic_barrier(unsigned t, unsigned p):
1045 : 72 : period(p), threads(t), counter(0), generation(0), waiting(0), dead(false) { }
1046 : : virtual ~periodic_barrier() {}
1047 : :
1048 : : virtual void periodic_barrier_work() noexcept = 0;
1049 : 72 : void nuke() {
1050 : 72 : unique_lock<mutex> lock(mtx);
1051 : 72 : dead = true;
1052 : 72 : cv.notify_all();
1053 : 72 : }
1054 : :
1055 : 1602 : void count()
1056 : : {
1057 : 1602 : unique_lock<mutex> lock(mtx);
1058 : 1602 : unsigned prev_generation = this->generation;
1059 [ + + ]: 1602 : if (counter < period-1) // normal case: counter just freely running
1060 : : {
1061 : 1458 : counter ++;
1062 : 1458 : return;
1063 : : }
1064 [ + + ]: 144 : else if (counter == period-1) // we're the doer
1065 : : {
1066 : 36 : counter = period; // entering barrier holding phase
1067 : 36 : cv.notify_all();
1068 [ + + + - ]: 173 : while (waiting < threads-1 && !dead)
1069 [ + - ]: 101 : cv.wait(lock);
1070 : : // all other threads are now stuck in the barrier
1071 : 36 : this->periodic_barrier_work(); // NB: we're holding the mutex the whole time
1072 : : // reset for next barrier, releasing other waiters
1073 : 36 : counter = 0;
1074 : 36 : generation ++;
1075 : 36 : cv.notify_all();
1076 : 36 : return;
1077 : : }
1078 [ + - ]: 108 : else if (counter == period) // we're a waiter, in holding phase
1079 : : {
1080 : 108 : waiting ++;
1081 : 108 : cv.notify_all();
1082 [ + + + - : 406 : while (counter == period && generation == prev_generation && !dead)
+ - ]
1083 [ + - ]: 190 : cv.wait(lock);
1084 : 108 : waiting --;
1085 : 108 : return;
1086 : : }
1087 : 1602 : }
1088 : : };
1089 : :
1090 : :
1091 : :
1092 : : ////////////////////////////////////////////////////////////////////////
1093 : :
1094 : :
1095 : : // Print a standard timestamp.
1096 : : static ostream&
1097 : 51295 : timestamp (ostream &o)
1098 : : {
1099 : 51295 : char datebuf[80];
1100 : 51295 : char *now2 = NULL;
1101 : 51295 : time_t now_t = time(NULL);
1102 : 51293 : struct tm now;
1103 : 51293 : struct tm *nowp = gmtime_r (&now_t, &now);
1104 [ + - ]: 51296 : if (nowp)
1105 : : {
1106 : 51296 : (void) strftime (datebuf, sizeof (datebuf), "%c", nowp);
1107 : 51296 : now2 = datebuf;
1108 : : }
1109 : :
1110 : 51296 : return o << "[" << (now2 ? now2 : "") << "] "
1111 [ - + ]: 51296 : << "(" << getpid () << "/" << tid() << "): ";
1112 : : }
1113 : :
1114 : :
1115 : : // A little class that impersonates an ostream to the extent that it can
1116 : : // take << streaming operations. It batches up the bits into an internal
1117 : : // stringstream until it is destroyed; then flushes to the original ostream.
1118 : : // It adds a timestamp
1119 : : class obatched
1120 : : {
1121 : : private:
1122 : : ostream& o;
1123 : : stringstream stro;
1124 : : static mutex lock;
1125 : : public:
1126 : 51297 : obatched(ostream& oo, bool timestamp_p = true): o(oo)
1127 : : {
1128 [ + - ]: 51296 : if (timestamp_p)
1129 [ + - ]: 51296 : timestamp(stro);
1130 : 51297 : }
1131 : 51291 : ~obatched()
1132 : : {
1133 : 51291 : unique_lock<mutex> do_not_cross_the_streams(obatched::lock);
1134 : 51299 : o << stro.str();
1135 : 51299 : o.flush();
1136 : 51299 : }
1137 : : operator ostream& () { return stro; }
1138 [ - - + - : 41142 : template <typename T> ostream& operator << (const T& t) { stro << t; return stro; }
+ - + - +
- + - + -
- - - - -
- + - + -
+ - + - +
- + - + -
+ - + - +
- + - + -
- - + - +
- + - + -
+ - + - +
- + - - -
+ - + - +
- + - + -
+ - + - -
- + - - -
+ - - - +
- + - + -
- - - - -
- - - - -
- - + - -
- + - + -
- - + - +
- - - + -
- - + - -
- - - + -
+ - + - +
- + - + -
+ - - - ]
1139 : : };
1140 : : mutex obatched::lock; // just the one, since cout/cerr iostreams are not thread-safe
1141 : :
1142 : :
1143 : 697 : void reportable_exception::report(ostream& o) const {
1144 [ + - + - ]: 697 : obatched(o) << message << endl;
1145 : 697 : }
1146 : :
1147 : :
1148 : : ////////////////////////////////////////////////////////////////////////
1149 : :
1150 : :
1151 : : // RAII style sqlite prepared-statement holder that matches { } block lifetime
1152 : :
1153 : : struct sqlite_ps
1154 : : {
1155 : : private:
1156 : : sqlite3* db;
1157 : : const string nickname;
1158 : : const string sql;
1159 : : sqlite3_stmt *pp;
1160 : : // for step_timeout()/callback
1161 : : struct timespec ts_start;
1162 : : double ts_timeout;
1163 : :
1164 : : sqlite_ps(const sqlite_ps&); // make uncopyable
1165 : : sqlite_ps& operator=(const sqlite_ps &); // make unassignable
1166 : :
1167 : : public:
1168 [ + - ]: 9862 : sqlite_ps (sqlite3* d, const string& n, const string& s): db(d), nickname(n), sql(s) {
1169 : : // tmp_ms_metric tick("sqlite3","prep",nickname);
1170 [ + + ]: 9861 : if (verbose > 4)
1171 [ + - + - : 174 : obatched(clog) << nickname << " prep " << sql << endl;
+ - + - +
- ]
1172 [ + - ]: 9861 : int rc = sqlite3_prepare_v2 (db, sql.c_str(), -1 /* to \0 */, & this->pp, NULL);
1173 [ - + ]: 9862 : if (rc != SQLITE_OK)
1174 [ # # # # ]: 0 : throw sqlite_exception(rc, "prepare " + sql);
1175 : 19724 : this->reset_timeout(0.0);
1176 : 0 : }
1177 : :
1178 : 198283 : sqlite_ps& reset()
1179 : : {
1180 [ + - + - ]: 396590 : tmp_ms_metric tick("sqlite3","reset",nickname);
1181 [ + - ]: 198308 : sqlite3_reset(this->pp);
1182 : 198322 : return *this;
1183 : 198282 : }
1184 : :
1185 : 226802 : sqlite_ps& bind(int parameter, const string& str)
1186 : : {
1187 [ + + ]: 226802 : if (verbose > 4)
1188 [ + - + - : 196 : obatched(clog) << nickname << " bind " << parameter << "=" << str << endl;
+ - + - +
- + - ]
1189 : 226802 : int rc = sqlite3_bind_text (this->pp, parameter, str.c_str(), -1, SQLITE_TRANSIENT);
1190 [ - + ]: 226762 : if (rc != SQLITE_OK)
1191 [ # # # # ]: 0 : throw sqlite_exception(rc, "sqlite3 bind");
1192 : 226762 : return *this;
1193 : : }
1194 : :
1195 : 61846 : sqlite_ps& bind(int parameter, int64_t value)
1196 : : {
1197 [ + + ]: 61846 : if (verbose > 4)
1198 [ + - + - : 64 : obatched(clog) << nickname << " bind " << parameter << "=" << value << endl;
+ - + - +
- + - ]
1199 : 61846 : int rc = sqlite3_bind_int64 (this->pp, parameter, value);
1200 [ - + ]: 61852 : if (rc != SQLITE_OK)
1201 [ # # # # ]: 0 : throw sqlite_exception(rc, "sqlite3 bind");
1202 : 61852 : return *this;
1203 : : }
1204 : :
1205 : : sqlite_ps& bind(int parameter)
1206 : : {
1207 : : if (verbose > 4)
1208 : : obatched(clog) << nickname << " bind " << parameter << "=" << "NULL" << endl;
1209 : : int rc = sqlite3_bind_null (this->pp, parameter);
1210 : : if (rc != SQLITE_OK)
1211 : : throw sqlite_exception(rc, "sqlite3 bind");
1212 : : return *this;
1213 : : }
1214 : :
1215 : :
1216 : 121842 : void step_ok_done() {
1217 [ + - + - ]: 243722 : tmp_ms_metric tick("sqlite3","step_done",nickname);
1218 [ + - ]: 121880 : int rc = sqlite3_step (this->pp);
1219 [ + + ]: 121930 : if (verbose > 4)
1220 [ + - + - : 128 : obatched(clog) << nickname << " step-ok-done(" << sqlite3_errstr(rc) << ") " << sql << endl;
+ - + - +
- + - + -
+ - ]
1221 [ + + - + ]: 121930 : if (rc != SQLITE_OK && rc != SQLITE_DONE && rc != SQLITE_ROW)
1222 [ # # # # ]: 0 : throw sqlite_exception(rc, "sqlite3 step");
1223 [ + - ]: 121930 : (void) sqlite3_reset (this->pp);
1224 : 121929 : }
1225 : :
1226 : :
1227 : 40899 : int step() {
1228 [ + - + - ]: 81793 : tmp_ms_metric tick("sqlite3","step",nickname);
1229 [ + - ]: 40895 : int rc = sqlite3_step (this->pp);
1230 [ + + ]: 40900 : if (verbose > 4)
1231 [ + - + - : 62 : obatched(clog) << nickname << " step(" << sqlite3_errstr(rc) << ") " << sql << endl;
+ - + - +
- + - + -
+ - ]
1232 : 40900 : return rc;
1233 : 40900 : }
1234 : :
1235 : :
1236 : 9892 : void reset_timeout(double s) // set starting point for maximum elapsed time in step_timeouts()
1237 : : {
1238 : 9862 : clock_gettime (CLOCK_MONOTONIC, &this->ts_start);
1239 : 9862 : this->ts_timeout = s;
1240 : : }
1241 : :
1242 : :
1243 : 0 : static int sqlite3_progress_handler_cb (void *param)
1244 : : {
1245 : 0 : sqlite_ps *pp = (sqlite_ps*) param;
1246 : 0 : struct timespec ts_end;
1247 : 0 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
1248 : 0 : double deltas = (ts_end.tv_sec - pp->ts_start.tv_sec) + (ts_end.tv_nsec - pp->ts_start.tv_nsec)/1.e9;
1249 [ # # # # ]: 0 : return (interrupted || (deltas > pp->ts_timeout)); // non-zero => interrupt sqlite operation in progress
1250 : : }
1251 : :
1252 : :
1253 : 48 : int step_timeout() {
1254 : : // Do the same thing as step(), except wrapping it into a timeout
1255 : : // relative to the last reset_timeout() invocation.
1256 : : //
1257 : : // Do this by attaching a progress_handler to the database
1258 : : // connection, for the duration of this operation. It should be a
1259 : : // private connection to the calling thread, so other operations
1260 : : // cannot begin concurrently.
1261 : :
1262 : 48 : sqlite3_progress_handler(this->db, 10000 /* bytecode insns */,
1263 : : & sqlite3_progress_handler_cb, (void*) this);
1264 : 48 : int rc = this->step();
1265 : 48 : sqlite3_progress_handler(this->db, 0, 0, 0); // disable
1266 : 48 : struct timespec ts_end;
1267 : 48 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
1268 : 48 : double deltas = (ts_end.tv_sec - this->ts_start.tv_sec) + (ts_end.tv_nsec - this->ts_start.tv_nsec)/1.e9;
1269 [ - + ]: 48 : if (verbose > 3)
1270 [ # # # # : 0 : obatched(clog) << this->nickname << " progress-delta-final " << deltas << endl;
# # # # ]
1271 : 48 : return rc;
1272 : : }
1273 : :
1274 : :
1275 : 9853 : ~sqlite_ps () { sqlite3_finalize (this->pp); }
1276 [ + - + - : 7088 : operator sqlite3_stmt* () { return this->pp; }
+ - + - +
- + - + -
+ - + - +
- ]
1277 : : };
1278 : :
1279 : :
1280 : : ////////////////////////////////////////////////////////////////////////
1281 : :
1282 : :
1283 : : struct sqlite_checkpoint_pb: public periodic_barrier
1284 : : {
1285 : : // NB: don't use sqlite_ps since it can throw exceptions during ctor etc.
1286 : 72 : sqlite_checkpoint_pb(unsigned t, unsigned p):
1287 : 144 : periodic_barrier(t, p) { }
1288 : :
1289 : 36 : void periodic_barrier_work() noexcept
1290 : : {
1291 : 36 : (void) sqlite3_exec (db, "pragma wal_checkpoint(truncate);", NULL, NULL, NULL);
1292 : 36 : }
1293 : : };
1294 : :
1295 : : static periodic_barrier* scan_barrier = 0; // initialized in main()
1296 : :
1297 : :
1298 : : ////////////////////////////////////////////////////////////////////////
1299 : :
1300 : : // RAII style templated autocloser
1301 : :
1302 : : template <class Payload, class Ignore>
1303 : : struct defer_dtor
1304 : : {
1305 : : public:
1306 : : typedef Ignore (*dtor_fn) (Payload);
1307 : :
1308 : : private:
1309 : : Payload p;
1310 : : dtor_fn fn;
1311 : :
1312 : : public:
1313 : 8022 : defer_dtor(Payload _p, dtor_fn _fn): p(_p), fn(_fn) {}
1314 : 414 : ~defer_dtor() { (void) (*fn)(p); }
1315 : :
1316 : : private:
1317 : : defer_dtor(const defer_dtor<Payload,Ignore>&); // make uncopyable
1318 : : defer_dtor& operator=(const defer_dtor<Payload,Ignore> &); // make unassignable
1319 : : };
1320 : :
1321 : :
1322 : :
1323 : : ////////////////////////////////////////////////////////////////////////
1324 : :
1325 : :
1326 : : static string
1327 : 6776 : header_censor(const string& str)
1328 : : {
1329 : 6776 : string y;
1330 [ + + ]: 83926 : for (auto&& x : str)
1331 : : {
1332 [ + + ]: 77150 : if (isalnum(x) || x == '/' || x == '.' || x == ',' || x == '_' || x == ':')
1333 [ + - ]: 154294 : y += x;
1334 : : }
1335 : 6776 : return y;
1336 : 0 : }
1337 : :
1338 : :
1339 : : static string
1340 : 3388 : conninfo (struct MHD_Connection * conn)
1341 : : {
1342 : 3388 : char hostname[256]; // RFC1035
1343 : 3388 : char servname[256];
1344 : 3388 : int sts = -1;
1345 : :
1346 [ - + ]: 3388 : if (conn == 0)
1347 : 0 : return "internal";
1348 : :
1349 : : /* Look up client address data. */
1350 : 3388 : const union MHD_ConnectionInfo *u = MHD_get_connection_info (conn,
1351 : : MHD_CONNECTION_INFO_CLIENT_ADDRESS);
1352 [ + - ]: 3388 : struct sockaddr *so = u ? u->client_addr : 0;
1353 : :
1354 [ + - - + ]: 3388 : if (so && so->sa_family == AF_INET) {
1355 : 0 : sts = getnameinfo (so, sizeof (struct sockaddr_in),
1356 : : hostname, sizeof (hostname),
1357 : : servname, sizeof (servname),
1358 : : NI_NUMERICHOST | NI_NUMERICSERV);
1359 [ + - ]: 3388 : } else if (so && so->sa_family == AF_INET6) {
1360 : 3388 : struct sockaddr_in6* addr6 = (struct sockaddr_in6*) so;
1361 [ + - + - : 3388 : if (IN6_IS_ADDR_V4MAPPED(&addr6->sin6_addr)) {
+ + ]
1362 : 1626 : struct sockaddr_in addr4;
1363 : 1626 : memset (&addr4, 0, sizeof(addr4));
1364 : 1626 : addr4.sin_family = AF_INET;
1365 : 1626 : addr4.sin_port = addr6->sin6_port;
1366 : 1626 : memcpy (&addr4.sin_addr.s_addr, addr6->sin6_addr.s6_addr+12, sizeof(addr4.sin_addr.s_addr));
1367 : 1626 : sts = getnameinfo ((struct sockaddr*) &addr4, sizeof (addr4),
1368 : : hostname, sizeof (hostname),
1369 : : servname, sizeof (servname),
1370 : : NI_NUMERICHOST | NI_NUMERICSERV);
1371 : : } else {
1372 : 1762 : sts = getnameinfo (so, sizeof (struct sockaddr_in6),
1373 : : hostname, sizeof (hostname),
1374 : : servname, sizeof (servname),
1375 : : NI_NUMERICHOST | NI_NUMERICSERV);
1376 : : }
1377 : : }
1378 : :
1379 [ - + ]: 3388 : if (sts != 0) {
1380 : 0 : hostname[0] = servname[0] = '\0';
1381 : : }
1382 : :
1383 : : // extract headers relevant to administration
1384 [ + + ]: 3388 : const char* user_agent = MHD_lookup_connection_value (conn, MHD_HEADER_KIND, "User-Agent") ?: "";
1385 [ + + ]: 3388 : const char* x_forwarded_for = MHD_lookup_connection_value (conn, MHD_HEADER_KIND, "X-Forwarded-For") ?: "";
1386 : : // NB: these are untrustworthy, beware if machine-processing log files
1387 : :
1388 [ + - + - : 10164 : return string(hostname) + string(":") + string(servname) +
+ - + - +
- ]
1389 [ + - + - : 13552 : string(" UA:") + header_censor(string(user_agent)) +
+ - + - +
- ]
1390 [ + - + - : 10164 : string(" XFF:") + header_censor(string(x_forwarded_for));
+ - + - ]
1391 : : }
1392 : :
1393 : :
1394 : :
1395 : : ////////////////////////////////////////////////////////////////////////
1396 : :
1397 : : /* Wrapper for MHD_add_response_header that logs an error if we
1398 : : couldn't add the specified header. */
1399 : : static void
1400 : 13314 : add_mhd_response_header (struct MHD_Response *r,
1401 : : const char *h, const char *v)
1402 : : {
1403 [ - + ]: 13314 : if (MHD_add_response_header (r, h, v) == MHD_NO)
1404 [ # # # # : 0 : obatched(clog) << "Error: couldn't add '" << h << "' header" << endl;
# # ]
1405 : 13314 : }
1406 : :
1407 : : static void
1408 : 2152 : add_mhd_last_modified (struct MHD_Response *resp, time_t mtime)
1409 : : {
1410 : 2152 : struct tm now;
1411 : 2152 : struct tm *nowp = gmtime_r (&mtime, &now);
1412 [ + - ]: 2152 : if (nowp != NULL)
1413 : : {
1414 : 2152 : char datebuf[80];
1415 : 2152 : size_t rc = strftime (datebuf, sizeof (datebuf), "%a, %d %b %Y %T GMT",
1416 : : nowp);
1417 [ + - ]: 2152 : if (rc > 0 && rc < sizeof (datebuf))
1418 : 2152 : add_mhd_response_header (resp, "Last-Modified", datebuf);
1419 : : }
1420 : :
1421 : 2152 : add_mhd_response_header (resp, "Cache-Control", "public");
1422 : 2152 : }
1423 : :
1424 : : // quote all questionable characters of str for safe passage through a sh -c expansion.
1425 : : static string
1426 : 568 : shell_escape(const string& str)
1427 : : {
1428 : 568 : string y;
1429 [ + + ]: 72928 : for (auto&& x : str)
1430 : : {
1431 [ + + + + ]: 72360 : if (! isalnum(x) && x != '/')
1432 [ + - ]: 8966 : y += "\\";
1433 [ + - ]: 144720 : y += x;
1434 : : }
1435 : 568 : return y;
1436 : 0 : }
1437 : :
1438 : :
1439 : : // PR25548: Perform POSIX / RFC3986 style path canonicalization on the input string.
1440 : : //
1441 : : // Namely:
1442 : : // // -> /
1443 : : // /foo/../ -> /
1444 : : // /./ -> /
1445 : : //
1446 : : // This mapping is done on dwarf-side source path names, which may
1447 : : // include these constructs, so we can deal with debuginfod clients
1448 : : // that accidentally canonicalize the paths.
1449 : : //
1450 : : // realpath(3) is close but not quite right, because it also resolves
1451 : : // symbolic links. Symlinks at the debuginfod server have nothing to
1452 : : // do with the build-time symlinks, thus they must not be considered.
1453 : : //
1454 : : // see also curl Curl_dedotdotify() aka RFC3986, which we mostly follow here
1455 : : // see also libc __realpath()
1456 : : // see also llvm llvm::sys::path::remove_dots()
1457 : : static string
1458 : 16942 : canon_pathname (const string& input)
1459 : : {
1460 : 16942 : string i = input; // 5.2.4 (1)
1461 : 16942 : string o;
1462 : :
1463 : 16942 : while (i.size() != 0)
1464 : : {
1465 : : // 5.2.4 (2) A
1466 [ + - - + ]: 145114 : if (i.substr(0,3) == "../")
1467 [ # # ]: 0 : i = i.substr(3);
1468 [ + - - + ]: 145114 : else if(i.substr(0,2) == "./")
1469 [ # # ]: 0 : i = i.substr(2);
1470 : :
1471 : : // 5.2.4 (2) B
1472 [ + - + + ]: 145114 : else if (i.substr(0,3) == "/./")
1473 [ + - ]: 1872 : i = i.substr(2);
1474 [ - + ]: 143242 : else if (i == "/.")
1475 [ # # ]: 0 : i = ""; // no need to handle "/." complete-path-segment case; we're dealing with file names
1476 : :
1477 : : // 5.2.4 (2) C
1478 [ + - + + ]: 143242 : else if (i.substr(0,4) == "/../") {
1479 [ + - ]: 2518 : i = i.substr(3);
1480 : 2518 : string::size_type sl = o.rfind("/");
1481 [ + - ]: 2518 : if (sl != string::npos)
1482 [ + - ]: 2518 : o = o.substr(0, sl);
1483 : : else
1484 [ # # ]: 0 : o = "";
1485 [ - + ]: 140724 : } else if (i == "/..")
1486 [ # # ]: 0 : i = ""; // no need to handle "/.." complete-path-segment case; we're dealing with file names
1487 : :
1488 : : // 5.2.4 (2) D
1489 : : // no need to handle these cases; we're dealing with file names
1490 [ - + ]: 140724 : else if (i == ".")
1491 [ # # ]: 0 : i = "";
1492 [ - + ]: 140724 : else if (i == "..")
1493 [ # # ]: 0 : i = "";
1494 : :
1495 : : // POSIX special: map // to /
1496 [ + - + + ]: 140724 : else if (i.substr(0,2) == "//")
1497 [ + - ]: 128 : i = i.substr(1);
1498 : :
1499 : : // 5.2.4 (2) E
1500 : : else {
1501 [ - + ]: 140596 : string::size_type next_slash = i.find("/", (i[0]=='/' ? 1 : 0)); // skip first slash
1502 [ + - ]: 281192 : o += i.substr(0, next_slash);
1503 [ + + ]: 140596 : if (next_slash == string::npos)
1504 [ + - + + ]: 178998 : i = "";
1505 : : else
1506 [ + - ]: 123654 : i = i.substr(next_slash);
1507 : : }
1508 : : }
1509 : :
1510 : 16942 : return o;
1511 : 16942 : }
1512 : :
1513 : :
1514 : : // Estimate available free space for a given filesystem via statfs(2).
1515 : : // Return true if the free fraction is known to be smaller than the
1516 : : // given minimum percentage. Also update a related metric.
1517 : 4212 : bool statfs_free_enough_p(const string& path, const string& label, long minfree = 0)
1518 : : {
1519 : 4212 : struct statfs sfs;
1520 : 4212 : int rc = statfs(path.c_str(), &sfs);
1521 [ + + ]: 4211 : if (rc == 0)
1522 : : {
1523 : 4137 : double s = (double) sfs.f_bavail / (double) sfs.f_blocks;
1524 [ + - + - ]: 8275 : set_metric("filesys_free_ratio","purpose",label, s);
1525 : 4138 : return ((s * 100.0) < minfree);
1526 : : }
1527 : : return false;
1528 : : }
1529 : :
1530 : :
1531 : :
1532 : : // A map-like class that owns a cache of file descriptors (indexed by
1533 : : // file / content names).
1534 : : //
1535 : : // If only it could use fd's instead of file names ... but we can't
1536 : : // dup(2) to create independent descriptors for the same unlinked
1537 : : // files, so would have to use some goofy linux /proc/self/fd/%d
1538 : : // hack such as the following
1539 : :
1540 : : #if 0
1541 : : int superdup(int fd)
1542 : : {
1543 : : #ifdef __linux__
1544 : : char *fdpath = NULL;
1545 : : int rc = asprintf(& fdpath, "/proc/self/fd/%d", fd);
1546 : : int newfd;
1547 : : if (rc >= 0)
1548 : : newfd = open(fdpath, O_RDONLY);
1549 : : else
1550 : : newfd = -1;
1551 : : free (fdpath);
1552 : : return newfd;
1553 : : #else
1554 : : return -1;
1555 : : #endif
1556 : : }
1557 : : #endif
1558 : :
1559 : : class libarchive_fdcache
1560 : : {
1561 : : private:
1562 : : mutex fdcache_lock;
1563 : :
1564 : : typedef pair<string,string> key; // archive, entry
1565 : 1504 : struct fdcache_entry
1566 : : {
1567 : : string fd; // file name (probably in $TMPDIR), not an actual open fd (EMFILE)
1568 : : double fd_size_mb; // slightly rounded up megabytes
1569 : : time_t freshness; // when was this entry created or requested last
1570 : : unsigned request_count; // how many requests were made; or 0=prefetch only
1571 : : double latency; // how many seconds it took to extract the file
1572 : : };
1573 : :
1574 : : map<key,fdcache_entry> entries; // optimized for lookup
1575 : : time_t last_cleaning;
1576 : : long max_mbs;
1577 : :
1578 : : public:
1579 : 3006 : void set_metrics()
1580 : : {
1581 : 3006 : double fdcache_mb = 0.0;
1582 : 3006 : double prefetch_mb = 0.0;
1583 : 3006 : unsigned fdcache_count = 0;
1584 : 3006 : unsigned prefetch_count = 0;
1585 [ + + ]: 6152 : for (auto &i : entries) {
1586 [ + + ]: 3146 : if (i.second.request_count) {
1587 : 2550 : fdcache_mb += i.second.fd_size_mb;
1588 : 2550 : fdcache_count ++;
1589 : : } else {
1590 : 596 : prefetch_mb += i.second.fd_size_mb;
1591 : 596 : prefetch_count ++;
1592 : : }
1593 : : }
1594 [ + - ]: 3006 : set_metric("fdcache_bytes", fdcache_mb*1024.0*1024.0);
1595 [ + - ]: 3006 : set_metric("fdcache_count", fdcache_count);
1596 [ + - ]: 3006 : set_metric("fdcache_prefetch_bytes", prefetch_mb*1024.0*1024.0);
1597 [ + - ]: 3006 : set_metric("fdcache_prefetch_count", prefetch_count);
1598 : 3006 : }
1599 : :
1600 : 1504 : void intern(const string& a, const string& b, string fd, off_t sz,
1601 : : bool requested_p, double lat)
1602 : : {
1603 : 1504 : {
1604 : 1504 : unique_lock<mutex> lock(fdcache_lock);
1605 : 1504 : time_t now = time(NULL);
1606 : : // there is a chance it's already in here, just wasn't found last time
1607 : : // if so, there's nothing to do but count our luck
1608 [ + - ]: 1504 : auto i = entries.find(make_pair(a,b));
1609 [ - + ]: 1504 : if (i != entries.end())
1610 : : {
1611 [ # # # # : 0 : inc_metric("fdcache_op_count","op","redundant_intern");
# # # # ]
1612 [ # # ]: 0 : if (requested_p) i->second.request_count ++; // repeat prefetch doesn't count
1613 : 0 : i->second.freshness = now;
1614 : : // We need to nuke the temp file, since interning passes
1615 : : // responsibility over the path to this structure. It is
1616 : : // possible that the caller still has an fd open, but that's
1617 : : // OK.
1618 : 0 : unlink (fd.c_str());
1619 : 0 : return;
1620 : : }
1621 : 1504 : double mb = (sz+65535)/1048576.0; // round up to 64K block
1622 : 1504 : fdcache_entry n = { .fd=fd, .fd_size_mb=mb,
1623 : 1504 : .freshness=now, .request_count = requested_p?1U:0U,
1624 [ + - + + ]: 1504 : .latency=lat};
1625 [ + - + - : 1504 : entries.insert(make_pair(make_pair(a,b),n));
+ - ]
1626 : :
1627 [ + + ]: 1504 : if (requested_p)
1628 [ + - + - : 1816 : inc_metric("fdcache_op_count","op","enqueue");
+ - + - ]
1629 : : else
1630 [ + - + - : 1192 : inc_metric("fdcache_op_count","op","prefetch_enqueue");
+ - + - ]
1631 : :
1632 [ + + ]: 1504 : if (verbose > 3)
1633 [ + - + - ]: 4164 : obatched(clog) << "fdcache interned a=" << a << " b=" << b
1634 [ + - + - : 1388 : << " fd=" << fd << " mb=" << mb << " front=" << requested_p
+ - + - +
- + - + -
+ - ]
1635 [ + - + - : 1388 : << " latency=" << lat << endl;
+ - ]
1636 : :
1637 [ + - ]: 1504 : set_metrics();
1638 : 1504 : }
1639 : :
1640 : : // NB: we age the cache at lookup time too
1641 [ + - + + ]: 1504 : if (statfs_free_enough_p(tmpdir, "tmpdir", fdcache_mintmp))
1642 : : {
1643 [ + - + - : 2836 : inc_metric("fdcache_op_count","op","emerg-flush");
+ - ]
1644 [ + - ]: 2836 : obatched(clog) << "fdcache emergency flush for filling tmpdir" << endl;
1645 : 1418 : this->limit(0); // emergency flush
1646 : : }
1647 : : else // age cache normally
1648 : 86 : this->limit(max_mbs);
1649 : : }
1650 : :
1651 : 1012 : int lookup(const string& a, const string& b)
1652 : : {
1653 : 1012 : int fd = -1;
1654 : 1012 : {
1655 : 1012 : unique_lock<mutex> lock(fdcache_lock);
1656 [ + - ]: 1012 : auto i = entries.find(make_pair(a,b));
1657 [ + + ]: 1012 : if (i != entries.end())
1658 : : {
1659 [ + + ]: 100 : if (i->second.request_count == 0) // was a prefetch!
1660 : : {
1661 [ + - + - ]: 2 : inc_metric("fdcache_prefetch_saved_milliseconds_count");
1662 [ + - + - ]: 4 : add_metric("fdcache_prefetch_saved_milliseconds_sum", i->second.latency*1000.);
1663 : : }
1664 : 100 : i->second.request_count ++;
1665 : 100 : i->second.freshness = time(NULL);
1666 : : // brag about our success
1667 [ + - + - : 200 : inc_metric("fdcache_op_count","op","prefetch_access"); // backward compat
+ - + - ]
1668 [ + - + - ]: 100 : inc_metric("fdcache_saved_milliseconds_count");
1669 [ + - + - ]: 100 : add_metric("fdcache_saved_milliseconds_sum", i->second.latency*1000.);
1670 [ + - ]: 1012 : fd = open(i->second.fd.c_str(), O_RDONLY);
1671 : : }
1672 : 0 : }
1673 : :
1674 [ + + ]: 1012 : if (fd >= 0)
1675 [ + - + - : 200 : inc_metric("fdcache_op_count","op","lookup_hit");
+ - ]
1676 : : else
1677 [ + - + - : 1824 : inc_metric("fdcache_op_count","op","lookup_miss");
+ - ]
1678 : :
1679 : : // NB: no need to age the cache after just a lookup
1680 : :
1681 : 1012 : return fd;
1682 : : }
1683 : :
1684 : 1296 : int probe(const string& a, const string& b) // just a cache residency check - don't modify state, don't open
1685 : : {
1686 : 1296 : unique_lock<mutex> lock(fdcache_lock);
1687 [ + - ]: 1296 : auto i = entries.find(make_pair(a,b));
1688 [ - + ]: 1296 : if (i != entries.end()) {
1689 [ # # # # : 0 : inc_metric("fdcache_op_count","op","probe_hit");
# # # # ]
1690 : 0 : return true;
1691 : : } else {
1692 [ + - + - : 2592 : inc_metric("fdcache_op_count","op","probe_miss");
+ - + - ]
1693 : 1296 : return false;
1694 : : }
1695 : 1296 : }
1696 : :
1697 : 0 : void clear(const string& a, const string& b)
1698 : : {
1699 : 0 : unique_lock<mutex> lock(fdcache_lock);
1700 [ # # ]: 0 : auto i = entries.find(make_pair(a,b));
1701 [ # # ]: 0 : if (i != entries.end()) {
1702 [ # # # # : 0 : inc_metric("fdcache_op_count","op",
# # # # #
# ]
1703 [ # # ]: 0 : i->second.request_count > 0 ? "clear" : "prefetch_clear");
1704 : 0 : unlink (i->second.fd.c_str());
1705 : 0 : entries.erase(i);
1706 [ # # ]: 0 : set_metrics();
1707 : 0 : return;
1708 : : }
1709 : 0 : }
1710 : :
1711 : 1672 : void limit(long maxmbs, bool metrics_p = true)
1712 : : {
1713 : 1672 : time_t now = time(NULL);
1714 : :
1715 : : // avoid overly frequent limit operations
1716 [ + + + + ]: 1672 : if (maxmbs > 0 && (now - this->last_cleaning) < 10) // probably not worth parametrizing
1717 : 86 : return;
1718 : 1586 : this->last_cleaning = now;
1719 : :
1720 [ + + + + ]: 1586 : if (verbose > 3 && (this->max_mbs != maxmbs))
1721 [ + - + - ]: 192 : obatched(clog) << "fdcache limited to maxmbs=" << maxmbs << endl;
1722 : :
1723 : 1586 : unique_lock<mutex> lock(fdcache_lock);
1724 : :
1725 : 1586 : this->max_mbs = maxmbs;
1726 : 1586 : double total_mb = 0.0;
1727 : :
1728 : 1586 : map<double, pair<string,string>> sorted_entries;
1729 [ + + ]: 3090 : for (auto &i: entries)
1730 : : {
1731 : 1504 : total_mb += i.second.fd_size_mb;
1732 : :
1733 : : // need a scalar quantity that combines these inputs in a sensible way:
1734 : : //
1735 : : // 1) freshness of this entry (last time it was accessed)
1736 : : // 2) size of this entry
1737 : : // 3) number of times it has been accessed (or if just prefetched with 0 accesses)
1738 : : // 4) latency it required to extract
1739 : : //
1740 : : // The lower the "score", the earlier garbage collection will
1741 : : // nuke it, so to prioritize entries for preservation, the
1742 : : // score should be higher, and vice versa.
1743 : 1504 : time_t factor_1_freshness = (now - i.second.freshness); // seconds
1744 : 1504 : double factor_2_size = i.second.fd_size_mb; // megabytes
1745 : 1504 : unsigned factor_3_accesscount = i.second.request_count; // units
1746 : 1504 : double factor_4_latency = i.second.latency; // seconds
1747 : :
1748 : : #if 0
1749 : : double score = - factor_1_freshness; // simple LRU
1750 : : #endif
1751 : :
1752 [ + + ]: 1504 : double score = 0.
1753 : 1504 : - log1p(factor_1_freshness) // penalize old file
1754 : 1504 : - log1p(factor_2_size) // penalize large file
1755 : 1504 : + factor_4_latency * factor_3_accesscount; // reward slow + repeatedly read files
1756 : :
1757 [ + + ]: 1504 : if (verbose > 4)
1758 [ + - ]: 8 : obatched(clog) << "fdcache scored score=" << score
1759 [ + - + - ]: 8 : << " a=" << i.first.first << " b=" << i.first.second
1760 [ + - + - : 12 : << " f1=" << factor_1_freshness << " f2=" << factor_2_size
+ - + - +
- + - +
- ]
1761 [ + - + - : 4 : << " f3=" << factor_3_accesscount << " f4=" << factor_4_latency
+ - + - +
- ]
1762 : 4 : << endl;
1763 : :
1764 [ + - + - ]: 4512 : sorted_entries.insert(make_pair(score, i.first));
1765 : : }
1766 : :
1767 : 1586 : unsigned cleaned = 0;
1768 : 1586 : unsigned entries_original = entries.size();
1769 : 1586 : double cleaned_score_min = DBL_MAX;
1770 : 1586 : double cleaned_score_max = DBL_MIN;
1771 : :
1772 : : // drop as many entries[] as needed to bring total mb down to the threshold
1773 [ + + ]: 3090 : for (auto &i: sorted_entries) // in increasing score order!
1774 : : {
1775 [ - + ]: 1504 : if (this->max_mbs > 0 // if this is not a "clear entire table"
1776 [ # # ]: 0 : && total_mb < this->max_mbs) // we've cleared enough to meet threshold
1777 : : break; // stop clearing
1778 : :
1779 [ - + ]: 1504 : auto j = entries.find(i.second);
1780 [ - + ]: 1504 : if (j == entries.end())
1781 : 0 : continue; // should not happen
1782 : :
1783 [ + + ]: 1504 : if (cleaned == 0)
1784 : 1422 : cleaned_score_min = i.first;
1785 : 1504 : cleaned++;
1786 : 1504 : cleaned_score_max = i.first;
1787 : :
1788 [ + + ]: 1504 : if (verbose > 3)
1789 [ + - + - ]: 4164 : obatched(clog) << "fdcache evicted score=" << i.first
1790 [ + - + - ]: 2776 : << " a=" << i.second.first << " b=" << i.second.second
1791 [ + - + - : 4164 : << " fd=" << j->second.fd << " mb=" << j->second.fd_size_mb
+ - + - +
- + - ]
1792 [ + - + - : 1388 : << " rq=" << j->second.request_count << " lat=" << j->second.latency
+ - + - ]
1793 [ + - + - : 1388 : << " fr=" << (now - j->second.freshness)
+ - ]
1794 : 1388 : << endl;
1795 [ + + ]: 1504 : if (metrics_p)
1796 [ + - + - : 2836 : inc_metric("fdcache_op_count","op","evict");
+ - + - ]
1797 : :
1798 : 1504 : total_mb -= j->second.fd_size_mb;
1799 : 1504 : unlink (j->second.fd.c_str());
1800 : 1504 : entries.erase(j);
1801 : : }
1802 : :
1803 [ + + ]: 1586 : if (metrics_p)
1804 [ + - + - : 3004 : inc_metric("fdcache_op_count","op","evict_cycle");
+ - + - ]
1805 : :
1806 [ + - + + ]: 1586 : if (verbose > 1 && cleaned > 0)
1807 : : {
1808 [ + - + - : 4266 : obatched(clog) << "fdcache evicted num=" << cleaned << " of=" << entries_original
+ - + - ]
1809 [ + - + - : 1422 : << " min=" << cleaned_score_min << " max=" << cleaned_score_max
+ - + - +
- ]
1810 : 1422 : << endl;
1811 : : }
1812 : :
1813 [ + + + - ]: 1586 : if (metrics_p) set_metrics();
1814 : 1586 : }
1815 : :
1816 : :
1817 : 84 : ~libarchive_fdcache()
1818 : : {
1819 : : // unlink any fdcache entries in $TMPDIR
1820 : : // don't update metrics; those globals may be already destroyed
1821 : 84 : limit(0, false);
1822 : 84 : }
1823 : : };
1824 : : static libarchive_fdcache fdcache;
1825 : :
1826 : : /* Search ELF_FD for an ELF/DWARF section with name SECTION.
1827 : : If found copy the section to a temporary file and return
1828 : : its file descriptor, otherwise return -1.
1829 : :
1830 : : The temporary file's mtime will be set to PARENT_MTIME.
1831 : : B_SOURCE should be a description of the parent file suitable
1832 : : for printing to the log. */
1833 : :
1834 : : static int
1835 : 12 : extract_section (int elf_fd, int64_t parent_mtime,
1836 : : const string& b_source, const string& section,
1837 : : const timespec& extract_begin)
1838 : : {
1839 : : /* Search the fdcache. */
1840 : 12 : struct stat fs;
1841 : 12 : int fd = fdcache.lookup (b_source, section);
1842 [ - + ]: 12 : if (fd >= 0)
1843 : : {
1844 [ # # ]: 0 : if (fstat (fd, &fs) != 0)
1845 : : {
1846 [ # # ]: 0 : if (verbose)
1847 [ # # ]: 0 : obatched (clog) << "cannot fstate fdcache "
1848 [ # # # # : 0 : << b_source << " " << section << endl;
# # ]
1849 : 0 : close (fd);
1850 : 0 : return -1;
1851 : : }
1852 [ # # ]: 0 : if ((int64_t) fs.st_mtime != parent_mtime)
1853 : : {
1854 [ # # ]: 0 : if (verbose)
1855 [ # # ]: 0 : obatched(clog) << "mtime mismatch for "
1856 [ # # # # : 0 : << b_source << " " << section << endl;
# # ]
1857 : 0 : close (fd);
1858 : 0 : return -1;
1859 : : }
1860 : : /* Success. */
1861 : : return fd;
1862 : : }
1863 : :
1864 : 12 : Elf *elf = elf_begin (elf_fd, ELF_C_READ_MMAP_PRIVATE, NULL);
1865 [ - + ]: 12 : if (elf == NULL)
1866 : : return -1;
1867 : :
1868 : : /* Try to find the section and copy the contents into a separate file. */
1869 : 12 : try
1870 : : {
1871 : 12 : size_t shstrndx;
1872 [ + - ]: 12 : int rc = elf_getshdrstrndx (elf, &shstrndx);
1873 [ - + ]: 12 : if (rc < 0)
1874 [ # # # # ]: 0 : throw elfutils_exception (rc, "getshdrstrndx");
1875 : :
1876 : : Elf_Scn *scn = NULL;
1877 : 424 : while (true)
1878 : : {
1879 [ + - ]: 218 : scn = elf_nextscn (elf, scn);
1880 [ + - ]: 218 : if (scn == NULL)
1881 : : break;
1882 : 218 : GElf_Shdr shdr_storage;
1883 [ + - ]: 218 : GElf_Shdr *shdr = gelf_getshdr (scn, &shdr_storage);
1884 [ + - ]: 218 : if (shdr == NULL)
1885 : : break;
1886 : :
1887 [ + - ]: 218 : const char *scn_name = elf_strptr (elf, shstrndx, shdr->sh_name);
1888 [ + - ]: 218 : if (scn_name == NULL)
1889 : : break;
1890 [ + + ]: 218 : if (scn_name == section)
1891 : : {
1892 : 12 : Elf_Data *data = NULL;
1893 : :
1894 : : /* We found the desired section. */
1895 [ + - ]: 12 : data = elf_rawdata (scn, NULL);
1896 [ - + ]: 12 : if (data == NULL)
1897 [ # # # # : 0 : throw elfutils_exception (elf_errno (), "elfraw_data");
# # ]
1898 [ + + ]: 12 : if (data->d_buf == NULL)
1899 : : {
1900 [ + - + - ]: 8 : obatched(clog) << "section " << section
1901 [ + - + - ]: 4 : << " is empty" << endl;
1902 : 4 : break;
1903 : : }
1904 : :
1905 : : /* Create temporary file containing the section. */
1906 : 8 : char *tmppath = NULL;
1907 : 8 : rc = asprintf (&tmppath, "%s/debuginfod-section.XXXXXX", tmpdir.c_str());
1908 [ - + ]: 8 : if (rc < 0)
1909 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate tmppath");
1910 : 8 : defer_dtor<void*,void> tmmpath_freer (tmppath, free);
1911 [ + - ]: 8 : fd = mkstemp (tmppath);
1912 [ - + ]: 8 : if (fd < 0)
1913 [ # # # # ]: 0 : throw libc_exception (errno, "cannot create temporary file");
1914 : :
1915 [ + - ]: 8 : ssize_t res = write_retry (fd, data->d_buf, data->d_size);
1916 [ + - - + ]: 8 : if (res < 0 || (size_t) res != data->d_size) {
1917 [ # # ]: 0 : close (fd);
1918 : 0 : unlink (tmppath);
1919 [ # # # # ]: 0 : throw libc_exception (errno, "cannot write to temporary file");
1920 : : }
1921 : :
1922 : : /* Set mtime to be the same as the parent file's mtime. */
1923 : 8 : struct timespec tvs[2];
1924 [ - + ]: 8 : if (fstat (elf_fd, &fs) != 0) {
1925 [ # # ]: 0 : close (fd);
1926 : 0 : unlink (tmppath);
1927 [ # # # # ]: 0 : throw libc_exception (errno, "cannot fstat file");
1928 : : }
1929 : :
1930 : 8 : tvs[0].tv_sec = 0;
1931 : 8 : tvs[0].tv_nsec = UTIME_OMIT;
1932 : 8 : tvs[1] = fs.st_mtim;
1933 : 8 : (void) futimens (fd, tvs);
1934 : :
1935 : 8 : struct timespec extract_end;
1936 : 8 : clock_gettime (CLOCK_MONOTONIC, &extract_end);
1937 : 8 : double extract_time = (extract_end.tv_sec - extract_begin.tv_sec)
1938 : 8 : + (extract_end.tv_nsec - extract_begin.tv_nsec)/1.e9;
1939 : :
1940 : : /* Add to fdcache. */
1941 [ + - + - ]: 8 : fdcache.intern (b_source, section, tmppath, data->d_size, true, extract_time);
1942 : 8 : break;
1943 : 8 : }
1944 : 206 : }
1945 : : }
1946 [ - - ]: 0 : catch (const reportable_exception &e)
1947 : : {
1948 [ - - ]: 0 : e.report (clog);
1949 [ - - ]: 0 : close (fd);
1950 : 0 : fd = -1;
1951 : 0 : }
1952 : :
1953 : 12 : elf_end (elf);
1954 : : return fd;
1955 : : }
1956 : :
1957 : : static struct MHD_Response*
1958 : 1152 : handle_buildid_f_match (bool internal_req_t,
1959 : : int64_t b_mtime,
1960 : : const string& b_source0,
1961 : : const string& section,
1962 : : int *result_fd)
1963 : : {
1964 : 1152 : (void) internal_req_t; // ignored
1965 : :
1966 : 1152 : struct timespec extract_begin;
1967 : 1152 : clock_gettime (CLOCK_MONOTONIC, &extract_begin);
1968 : :
1969 : 1152 : int fd = open(b_source0.c_str(), O_RDONLY);
1970 [ - + ]: 1152 : if (fd < 0)
1971 [ # # # # : 0 : throw libc_exception (errno, string("open ") + b_source0);
# # ]
1972 : :
1973 : : // NB: use manual close(2) in error case instead of defer_dtor, because
1974 : : // in the normal case, we want to hand the fd over to libmicrohttpd for
1975 : : // file transfer.
1976 : :
1977 : 1152 : struct stat s;
1978 : 1152 : int rc = fstat(fd, &s);
1979 [ - + ]: 1152 : if (rc < 0)
1980 : : {
1981 : 0 : close(fd);
1982 [ # # # # : 0 : throw libc_exception (errno, string("fstat ") + b_source0);
# # ]
1983 : : }
1984 : :
1985 [ - + ]: 1152 : if ((int64_t) s.st_mtime != b_mtime)
1986 : : {
1987 [ # # ]: 0 : if (verbose)
1988 [ # # # # ]: 0 : obatched(clog) << "mtime mismatch for " << b_source0 << endl;
1989 : 0 : close(fd);
1990 : 0 : return 0;
1991 : : }
1992 : :
1993 [ + + ]: 1152 : if (!section.empty ())
1994 : : {
1995 : 6 : int scn_fd = extract_section (fd, s.st_mtime, b_source0, section, extract_begin);
1996 : 6 : close (fd);
1997 : :
1998 [ + + ]: 6 : if (scn_fd >= 0)
1999 : 4 : fd = scn_fd;
2000 : : else
2001 : : {
2002 [ + - ]: 2 : if (verbose)
2003 [ + - ]: 6 : obatched (clog) << "cannot find section " << section
2004 [ + - + - : 2 : << " for " << b_source0 << endl;
+ - ]
2005 : 2 : return 0;
2006 : : }
2007 : :
2008 : 4 : rc = fstat(fd, &s);
2009 [ - + ]: 4 : if (rc < 0)
2010 : : {
2011 : 0 : close (fd);
2012 [ # # # # ]: 0 : throw libc_exception (errno, string ("fstat ") + b_source0
2013 [ # # # # : 0 : + string (" ") + section);
# # # # ]
2014 : : }
2015 : : }
2016 : :
2017 : 1150 : struct MHD_Response* r = MHD_create_response_from_fd ((uint64_t) s.st_size, fd);
2018 [ + - + - : 2300 : inc_metric ("http_responses_total","result","file");
+ - ]
2019 [ - + ]: 1150 : if (r == 0)
2020 : : {
2021 [ # # ]: 0 : if (verbose)
2022 [ # # ]: 0 : obatched(clog) << "cannot create fd-response for " << b_source0
2023 [ # # # # : 0 : << " section=" << section << endl;
# # ]
2024 : 0 : close(fd);
2025 : : }
2026 : : else
2027 : : {
2028 : 1150 : add_mhd_response_header (r, "Content-Type", "application/octet-stream");
2029 [ + - ]: 1150 : add_mhd_response_header (r, "X-DEBUGINFOD-SIZE",
2030 : 1150 : to_string(s.st_size).c_str());
2031 : 1150 : add_mhd_response_header (r, "X-DEBUGINFOD-FILE", b_source0.c_str());
2032 : 1150 : add_mhd_last_modified (r, s.st_mtime);
2033 [ + - ]: 1150 : if (verbose > 1)
2034 [ + - + - : 2300 : obatched(clog) << "serving file " << b_source0 << " section=" << section << endl;
+ - + - ]
2035 : : /* libmicrohttpd will close it. */
2036 [ - + ]: 1150 : if (result_fd)
2037 : 1150 : *result_fd = fd;
2038 : : }
2039 : :
2040 : : return r;
2041 : : }
2042 : :
2043 : :
2044 : : #ifdef USE_LZMA
2045 : : struct lzma_exception: public reportable_exception
2046 : : {
2047 : 0 : lzma_exception(int rc, const string& msg):
2048 : : // liblzma doesn't have a lzma_ret -> string conversion function, so just
2049 : : // report the value.
2050 [ # # # # : 0 : reportable_exception(string ("lzma error: ") + msg + ": error " + to_string(rc)) {
# # ]
2051 [ # # # # : 0 : inc_metric("error_count","lzma",to_string(rc));
# # ]
2052 : 0 : }
2053 : : };
2054 : :
2055 : : // Neither RPM nor deb files support seeking to a specific file in the package.
2056 : : // Instead, to extract a specific file, we normally need to read the archive
2057 : : // sequentially until we find the file. This is very slow for files at the end
2058 : : // of a large package with lots of files, like kernel debuginfo.
2059 : : //
2060 : : // However, if the compression format used in the archive supports seeking, we
2061 : : // can accelerate this. As of July 2024, xz is the only widely-used format that
2062 : : // supports seeking, and usually only in multi-threaded mode. Luckily, the
2063 : : // kernel-debuginfo package in Fedora and its downstreams, and the
2064 : : // linux-image-*-dbg package in Debian and its downstreams, all happen to use
2065 : : // this.
2066 : : //
2067 : : // The xz format [1] ends with an index of independently compressed blocks in
2068 : : // the stream. In RPM and deb files, the xz stream is the last thing in the
2069 : : // file, so we assume that the xz Stream Footer is at the end of the package
2070 : : // file and do everything relative to that. For each file in the archive, we
2071 : : // remember the size and offset of the file data in the uncompressed xz stream,
2072 : : // then we use the index to seek to that offset when we need that file.
2073 : : //
2074 : : // 1: https://xz.tukaani.org/format/xz-file-format.txt
2075 : :
2076 : : // Return whether an archive supports seeking.
2077 : : static bool
2078 : 1092 : is_seekable_archive (const string& rps, struct archive* a)
2079 : : {
2080 : : // Only xz supports seeking.
2081 [ + + ]: 1092 : if (archive_filter_code (a, 0) != ARCHIVE_FILTER_XZ)
2082 : : return false;
2083 : :
2084 : 390 : int fd = open (rps.c_str(), O_RDONLY);
2085 [ - + ]: 390 : if (fd < 0)
2086 : : return false;
2087 : 390 : defer_dtor<int,int> fd_closer (fd, close);
2088 : :
2089 : : // Seek to the xz Stream Footer. We assume that it's the last thing in the
2090 : : // file, which is true for RPM and deb files.
2091 : 390 : off_t footer_pos = -LZMA_STREAM_HEADER_SIZE;
2092 [ - + ]: 390 : if (lseek (fd, footer_pos, SEEK_END) == -1)
2093 : : return false;
2094 : :
2095 : : // Decode the Stream Footer.
2096 : : uint8_t footer[LZMA_STREAM_HEADER_SIZE];
2097 : : size_t footer_read = 0;
2098 [ + + ]: 780 : while (footer_read < sizeof (footer))
2099 : : {
2100 [ - + ]: 390 : ssize_t bytes_read = read (fd, footer + footer_read,
2101 : : sizeof (footer) - footer_read);
2102 [ - + ]: 390 : if (bytes_read < 0)
2103 : : {
2104 [ # # ]: 0 : if (errno == EINTR)
2105 : 0 : continue;
2106 : : return false;
2107 : : }
2108 [ - + ]: 390 : if (bytes_read == 0)
2109 : : return false;
2110 : 390 : footer_read += bytes_read;
2111 : : }
2112 : :
2113 : 390 : lzma_stream_flags stream_flags;
2114 : 390 : lzma_ret ret = lzma_stream_footer_decode (&stream_flags, footer);
2115 [ - + ]: 390 : if (ret != LZMA_OK)
2116 : : return false;
2117 : :
2118 : : // Seek to the xz Index.
2119 [ - + ]: 390 : if (lseek (fd, footer_pos - stream_flags.backward_size, SEEK_END) == -1)
2120 : : return false;
2121 : :
2122 : : // Decode the Number of Records in the Index. liblzma doesn't have an API for
2123 : : // this if you don't want to decode the whole Index, so we have to do it
2124 : : // ourselves.
2125 : : //
2126 : : // We need 1 byte for the Index Indicator plus 1-9 bytes for the
2127 : : // variable-length integer Number of Records.
2128 : : uint8_t index[10];
2129 : : size_t index_read = 0;
2130 [ + + ]: 780 : while (index_read == 0) {
2131 [ + - ]: 390 : ssize_t bytes_read = read (fd, index, sizeof (index));
2132 [ - + ]: 390 : if (bytes_read < 0)
2133 : : {
2134 [ # # ]: 0 : if (errno == EINTR)
2135 : 0 : continue;
2136 : : return false;
2137 : : }
2138 [ - + ]: 390 : if (bytes_read == 0)
2139 : : return false;
2140 : 390 : index_read += bytes_read;
2141 : : }
2142 : : // The Index Indicator must be 0.
2143 [ - + ]: 390 : if (index[0] != 0)
2144 : : return false;
2145 : :
2146 : 390 : lzma_vli num_records;
2147 : 390 : size_t pos = 0;
2148 : 390 : size_t in_pos = 1;
2149 : 390 : while (true)
2150 : : {
2151 [ - + ]: 390 : if (in_pos >= index_read)
2152 : : {
2153 [ # # ]: 0 : ssize_t bytes_read = read (fd, index, sizeof (index));
2154 [ # # ]: 0 : if (bytes_read < 0)
2155 : : {
2156 [ # # ]: 0 : if (errno == EINTR)
2157 : 0 : continue;
2158 : : return false;
2159 : : }
2160 [ # # ]: 0 : if (bytes_read == 0)
2161 : : return false;
2162 : 0 : index_read = bytes_read;
2163 : 0 : in_pos = 0;
2164 : : }
2165 : 390 : ret = lzma_vli_decode (&num_records, &pos, index, &in_pos, index_read);
2166 [ - + ]: 390 : if (ret == LZMA_STREAM_END)
2167 : : break;
2168 [ # # ]: 0 : else if (ret != LZMA_OK)
2169 : : return false;
2170 : : }
2171 : :
2172 [ + + ]: 390 : if (verbose > 3)
2173 [ + - + - : 676 : obatched(clog) << rps << " has " << num_records << " xz Blocks" << endl;
+ - + - +
- ]
2174 : :
2175 : : // The file is only seekable if it has more than one Block.
2176 : 390 : return num_records > 1;
2177 : 390 : }
2178 : :
2179 : : // Read the Index at the end of an xz file.
2180 : : static lzma_index*
2181 : 202 : read_xz_index (int fd)
2182 : : {
2183 : 202 : off_t footer_pos = -LZMA_STREAM_HEADER_SIZE;
2184 [ - + ]: 202 : if (lseek (fd, footer_pos, SEEK_END) == -1)
2185 [ # # # # ]: 0 : throw libc_exception (errno, "lseek");
2186 : :
2187 : : uint8_t footer[LZMA_STREAM_HEADER_SIZE];
2188 : : size_t footer_read = 0;
2189 [ + + ]: 404 : while (footer_read < sizeof (footer))
2190 : : {
2191 [ - + ]: 202 : ssize_t bytes_read = read (fd, footer + footer_read,
2192 : : sizeof (footer) - footer_read);
2193 [ - + ]: 202 : if (bytes_read < 0)
2194 : : {
2195 [ # # ]: 0 : if (errno == EINTR)
2196 : 0 : continue;
2197 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2198 : : }
2199 [ - + ]: 202 : if (bytes_read == 0)
2200 [ # # ]: 0 : throw reportable_exception ("truncated file");
2201 : 202 : footer_read += bytes_read;
2202 : : }
2203 : :
2204 : 202 : lzma_stream_flags stream_flags;
2205 : 202 : lzma_ret ret = lzma_stream_footer_decode (&stream_flags, footer);
2206 [ - + ]: 202 : if (ret != LZMA_OK)
2207 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_stream_footer_decode");
2208 : :
2209 [ - + ]: 202 : if (lseek (fd, footer_pos - stream_flags.backward_size, SEEK_END) == -1)
2210 [ # # # # ]: 0 : throw libc_exception (errno, "lseek");
2211 : :
2212 : 202 : lzma_stream strm = LZMA_STREAM_INIT;
2213 : 202 : lzma_index* index = NULL;
2214 : 202 : ret = lzma_index_decoder (&strm, &index, UINT64_MAX);
2215 [ - + ]: 202 : if (ret != LZMA_OK)
2216 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_index_decoder");
2217 : 202 : defer_dtor<lzma_stream*,void> strm_ender (&strm, lzma_end);
2218 : :
2219 : 202 : uint8_t in_buf[4096];
2220 : 202 : while (true)
2221 : : {
2222 [ + - ]: 202 : if (strm.avail_in == 0)
2223 : : {
2224 [ + - ]: 202 : ssize_t bytes_read = read (fd, in_buf, sizeof (in_buf));
2225 [ - + ]: 202 : if (bytes_read < 0)
2226 : : {
2227 [ # # ]: 0 : if (errno == EINTR)
2228 : 0 : continue;
2229 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2230 : : }
2231 [ - + ]: 202 : if (bytes_read == 0)
2232 [ # # ]: 0 : throw reportable_exception ("truncated file");
2233 : 202 : strm.avail_in = bytes_read;
2234 : 202 : strm.next_in = in_buf;
2235 : : }
2236 : :
2237 : 202 : ret = lzma_code (&strm, LZMA_RUN);
2238 [ - + ]: 202 : if (ret == LZMA_STREAM_END)
2239 : : break;
2240 [ # # ]: 0 : else if (ret != LZMA_OK)
2241 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_code index");
2242 : : }
2243 : :
2244 : 202 : ret = lzma_index_stream_flags (index, &stream_flags);
2245 [ - + ]: 202 : if (ret != LZMA_OK)
2246 : : {
2247 : 0 : lzma_index_end (index, NULL);
2248 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_index_stream_flags");
2249 : : }
2250 : 202 : return index;
2251 : 202 : }
2252 : :
2253 : : static void
2254 : 202 : my_lzma_index_end (lzma_index* index)
2255 : : {
2256 : 202 : lzma_index_end (index, NULL);
2257 : 202 : }
2258 : :
2259 : : static void
2260 : 210 : free_lzma_block_filter_options (lzma_block* block)
2261 : : {
2262 [ + + ]: 1050 : for (int i = 0; i < LZMA_FILTERS_MAX; i++)
2263 : : {
2264 : 840 : free (block->filters[i].options);
2265 : 840 : block->filters[i].options = NULL;
2266 : : }
2267 : 210 : }
2268 : :
2269 : : static void
2270 : 202 : free_lzma_block_filters (lzma_block* block)
2271 : : {
2272 [ + - ]: 202 : if (block->filters != NULL)
2273 : : {
2274 : 202 : free_lzma_block_filter_options (block);
2275 : 202 : free (block->filters);
2276 : : }
2277 : 202 : }
2278 : :
2279 : : static void
2280 : 202 : extract_xz_blocks_into_fd (const string& srcpath,
2281 : : int src,
2282 : : int dst,
2283 : : lzma_index_iter* iter,
2284 : : uint64_t offset,
2285 : : uint64_t size)
2286 : : {
2287 : : // Seek to the Block. Seeking from the end using the compressed size from the
2288 : : // footer means we don't need to know where the xz stream starts in the
2289 : : // archive.
2290 [ - + ]: 202 : if (lseek (src,
2291 : 202 : (off_t) iter->block.compressed_stream_offset
2292 : 202 : - (off_t) iter->stream.compressed_size,
2293 : : SEEK_END) == -1)
2294 [ # # # # ]: 0 : throw libc_exception (errno, "lseek");
2295 : :
2296 : 202 : offset -= iter->block.uncompressed_file_offset;
2297 : :
2298 : 202 : lzma_block block{};
2299 : 202 : block.filters = (lzma_filter*) calloc (LZMA_FILTERS_MAX + 1,
2300 : : sizeof (lzma_filter));
2301 [ - + ]: 202 : if (block.filters == NULL)
2302 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate lzma_block filters");
2303 : 202 : defer_dtor<lzma_block*,void> filters_freer (&block, free_lzma_block_filters);
2304 : :
2305 : 202 : uint8_t in_buf[4096];
2306 : 202 : uint8_t out_buf[4096];
2307 : 202 : size_t header_read = 0;
2308 : 202 : bool need_log_extracting = verbose > 3;
2309 : 8 : while (true)
2310 : : {
2311 : : // The first byte of the Block is the encoded Block Header Size. Read the
2312 : : // first byte and whatever extra fits in the buffer.
2313 [ + + ]: 412 : while (header_read == 0)
2314 : : {
2315 [ + - ]: 202 : ssize_t bytes_read = read (src, in_buf, sizeof (in_buf));
2316 [ - + ]: 202 : if (bytes_read < 0)
2317 : : {
2318 [ # # ]: 0 : if (errno == EINTR)
2319 : 0 : continue;
2320 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2321 : : }
2322 [ - + ]: 202 : if (bytes_read == 0)
2323 [ # # ]: 0 : throw reportable_exception ("truncated file");
2324 : 202 : header_read += bytes_read;
2325 : : }
2326 : :
2327 : 210 : block.header_size = lzma_block_header_size_decode (in_buf[0]);
2328 : :
2329 : : // If we didn't buffer the whole Block Header earlier, get the rest.
2330 : 210 : eu_static_assert (sizeof (in_buf)
2331 : : >= lzma_block_header_size_decode (UINT8_MAX));
2332 [ - + ]: 210 : while (header_read < block.header_size)
2333 : : {
2334 [ # # ]: 0 : ssize_t bytes_read = read (src, in_buf + header_read,
2335 : : sizeof (in_buf) - header_read);
2336 [ # # ]: 0 : if (bytes_read < 0)
2337 : : {
2338 [ # # ]: 0 : if (errno == EINTR)
2339 : 0 : continue;
2340 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2341 : : }
2342 [ # # ]: 0 : if (bytes_read == 0)
2343 [ # # ]: 0 : throw reportable_exception ("truncated file");
2344 : 0 : header_read += bytes_read;
2345 : : }
2346 : :
2347 : : // Decode the Block Header.
2348 : 210 : block.check = iter->stream.flags->check;
2349 : 210 : lzma_ret ret = lzma_block_header_decode (&block, NULL, in_buf);
2350 [ - + ]: 210 : if (ret != LZMA_OK)
2351 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_block_header_decode");
2352 : 210 : ret = lzma_block_compressed_size (&block, iter->block.unpadded_size);
2353 [ - + ]: 210 : if (ret != LZMA_OK)
2354 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_block_compressed_size");
2355 : :
2356 : : // Start decoding the Block data.
2357 : 210 : lzma_stream strm = LZMA_STREAM_INIT;
2358 : 210 : ret = lzma_block_decoder (&strm, &block);
2359 [ - + ]: 210 : if (ret != LZMA_OK)
2360 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_block_decoder");
2361 : 210 : defer_dtor<lzma_stream*,void> strm_ender (&strm, lzma_end);
2362 : :
2363 : : // We might still have some input buffered from when we read the header.
2364 : 210 : strm.avail_in = header_read - block.header_size;
2365 : 210 : strm.next_in = in_buf + block.header_size;
2366 : 210 : strm.avail_out = sizeof (out_buf);
2367 : 210 : strm.next_out = out_buf;
2368 : 11734 : while (true)
2369 : : {
2370 [ + + ]: 11734 : if (strm.avail_in == 0)
2371 : : {
2372 [ + - ]: 14 : ssize_t bytes_read = read (src, in_buf, sizeof (in_buf));
2373 [ - + ]: 14 : if (bytes_read < 0)
2374 : : {
2375 [ # # ]: 0 : if (errno == EINTR)
2376 : 0 : continue;
2377 [ # # # # ]: 0 : throw libc_exception (errno, "read");
2378 : : }
2379 [ - + ]: 14 : if (bytes_read == 0)
2380 [ # # ]: 0 : throw reportable_exception ("truncated file");
2381 : 14 : strm.avail_in = bytes_read;
2382 : 14 : strm.next_in = in_buf;
2383 : : }
2384 : :
2385 : 11734 : ret = lzma_code (&strm, LZMA_RUN);
2386 [ - + ]: 11734 : if (ret != LZMA_OK && ret != LZMA_STREAM_END)
2387 [ # # # # ]: 0 : throw lzma_exception (ret, "lzma_code block");
2388 : :
2389 : : // Throw away anything we decode until we reach the offset, then
2390 : : // start writing to the destination.
2391 [ + + ]: 11734 : if (strm.total_out > offset)
2392 : : {
2393 : 3006 : size_t bytes_to_write = strm.next_out - out_buf;
2394 : 3006 : uint8_t* buf_to_write = out_buf;
2395 : :
2396 : : // Ignore anything in the buffer before the offset.
2397 [ + + ]: 3006 : if (bytes_to_write > strm.total_out - offset)
2398 : : {
2399 : 196 : buf_to_write += bytes_to_write - (strm.total_out - offset);
2400 : 196 : bytes_to_write = strm.total_out - offset;
2401 : : }
2402 : :
2403 : : // Ignore anything after the size.
2404 [ + + ]: 3006 : if (strm.total_out - offset >= size)
2405 : 202 : bytes_to_write -= strm.total_out - offset - size;
2406 : :
2407 [ + + ]: 3006 : if (need_log_extracting)
2408 : : {
2409 [ + - + - ]: 240 : obatched(clog) << "extracting from xz archive " << srcpath
2410 [ + - + - : 120 : << " size=" << size << endl;
+ - ]
2411 : 120 : need_log_extracting = false;
2412 : : }
2413 : :
2414 [ + + ]: 6012 : while (bytes_to_write > 0)
2415 : : {
2416 [ + - ]: 3006 : ssize_t written = write (dst, buf_to_write, bytes_to_write);
2417 [ - + ]: 3006 : if (written < 0)
2418 : : {
2419 [ # # ]: 0 : if (errno == EAGAIN)
2420 : 0 : continue;
2421 [ # # # # ]: 0 : throw libc_exception (errno, "write");
2422 : : }
2423 : 3006 : bytes_to_write -= written;
2424 : 3006 : buf_to_write += written;
2425 : : }
2426 : :
2427 : : // If we reached the size, we're done.
2428 [ + + ]: 3006 : if (strm.total_out - offset >= size)
2429 : 202 : return;
2430 : : }
2431 : :
2432 : 11532 : strm.avail_out = sizeof (out_buf);
2433 : 11532 : strm.next_out = out_buf;
2434 : :
2435 [ + + ]: 11532 : if (ret == LZMA_STREAM_END)
2436 : : break;
2437 : : }
2438 : :
2439 : : // This Block didn't have enough data. Go to the next one.
2440 [ - + ]: 8 : if (lzma_index_iter_next (iter, LZMA_INDEX_ITER_BLOCK))
2441 [ # # ]: 0 : throw reportable_exception ("no more blocks");
2442 [ + - ]: 8 : if (strm.total_out > offset)
2443 : 8 : size -= strm.total_out - offset;
2444 : 8 : offset = 0;
2445 : : // If we had any buffered input left, move it to the beginning of the
2446 : : // buffer to decode the next Block Header.
2447 [ + - ]: 8 : if (strm.avail_in > 0)
2448 : : {
2449 : 8 : memmove (in_buf, strm.next_in, strm.avail_in);
2450 : 8 : header_read = strm.avail_in;
2451 : : }
2452 : : else
2453 : : header_read = 0;
2454 : 8 : free_lzma_block_filter_options (&block);
2455 : 210 : }
2456 : 202 : }
2457 : :
2458 : : static int
2459 : 202 : extract_from_seekable_archive (const string& srcpath,
2460 : : char* tmppath,
2461 : : uint64_t offset,
2462 : : uint64_t size)
2463 : : {
2464 [ + - + - : 404 : inc_metric ("seekable_archive_extraction_attempts","type","xz");
+ - ]
2465 : 202 : try
2466 : : {
2467 [ + - ]: 202 : int src = open (srcpath.c_str(), O_RDONLY);
2468 [ - + ]: 202 : if (src < 0)
2469 [ # # # # : 0 : throw libc_exception (errno, string("open ") + srcpath);
# # ]
2470 : 202 : defer_dtor<int,int> src_closer (src, close);
2471 : :
2472 [ + - ]: 202 : lzma_index* index = read_xz_index (src);
2473 : 202 : defer_dtor<lzma_index*,void> index_ender (index, my_lzma_index_end);
2474 : :
2475 : : // Find the Block containing the offset.
2476 : 202 : lzma_index_iter iter;
2477 : 202 : lzma_index_iter_init (&iter, index);
2478 [ - + ]: 202 : if (lzma_index_iter_locate (&iter, offset))
2479 [ # # ]: 0 : throw reportable_exception ("offset not found");
2480 : :
2481 [ + + ]: 202 : if (verbose > 3)
2482 [ + - + - ]: 360 : obatched(clog) << "seeking in xz archive " << srcpath
2483 [ + - + - : 120 : << " offset=" << offset << " block_offset="
+ - ]
2484 [ + - + - ]: 120 : << iter.block.uncompressed_file_offset << endl;
2485 : :
2486 [ + - ]: 202 : int dst = mkstemp (tmppath);
2487 [ - + ]: 202 : if (dst < 0)
2488 [ # # # # ]: 0 : throw libc_exception (errno, "cannot create temporary file");
2489 : :
2490 : 202 : try
2491 : : {
2492 [ + - ]: 202 : extract_xz_blocks_into_fd (srcpath, src, dst, &iter, offset, size);
2493 : : }
2494 : 0 : catch (...)
2495 : : {
2496 : 0 : unlink (tmppath);
2497 [ - - ]: 0 : close (dst);
2498 : 0 : throw;
2499 : 0 : }
2500 : :
2501 [ + - + - : 404 : inc_metric ("seekable_archive_extraction_successes","type","xz");
+ - + - ]
2502 : 202 : return dst;
2503 : 202 : }
2504 [ - - ]: 0 : catch (const reportable_exception &e)
2505 : : {
2506 [ - - - - : 0 : inc_metric ("seekable_archive_extraction_failures","type","xz");
- - - - ]
2507 [ - - ]: 0 : if (verbose)
2508 [ - - - - ]: 0 : obatched(clog) << "failed to extract from seekable xz archive "
2509 [ - - - - : 0 : << srcpath << ": " << e.message << endl;
- - ]
2510 : 0 : return -1;
2511 : 0 : }
2512 : : }
2513 : : #else
2514 : : static bool
2515 : : is_seekable_archive (const string& rps __attribute__ ((unused)),
2516 : : struct archive* a __attribute__ ((unused)))
2517 : : {
2518 : : return false;
2519 : : }
2520 : : static int
2521 : : extract_from_seekable_archive (const string& srcpath __attribute__ ((unused)),
2522 : : char* tmppath __attribute__ ((unused)),
2523 : : uint64_t offset __attribute__ ((unused)),
2524 : : uint64_t size __attribute__ ((unused)))
2525 : : {
2526 : : return -1;
2527 : : }
2528 : : #endif
2529 : :
2530 : :
2531 : : // For security/portability reasons, many distro-package archives have
2532 : : // a "./" in front of path names; others have nothing, others have
2533 : : // "/". Canonicalize them all to a single leading "/", with the
2534 : : // assumption that this matches the dwarf-derived file names too.
2535 : 3966 : string canonicalized_archive_entry_pathname(struct archive_entry *e)
2536 : : {
2537 : 3966 : string fn = archive_entry_pathname(e);
2538 [ - + ]: 3966 : if (fn.size() == 0)
2539 : 0 : return fn;
2540 [ - + ]: 3966 : if (fn[0] == '/')
2541 : 0 : return fn;
2542 [ + + ]: 3966 : if (fn[0] == '.')
2543 [ + - ]: 2726 : return fn.substr(1);
2544 : : else
2545 [ + - + - ]: 2480 : return string("/")+fn;
2546 : 3966 : }
2547 : :
2548 : :
2549 : : // NB: takes ownership of, and may reassign, fd.
2550 : : static struct MHD_Response*
2551 : 1000 : create_buildid_r_response (int64_t b_mtime0,
2552 : : const string& b_source0,
2553 : : const string& b_source1,
2554 : : const string& section,
2555 : : const string& ima_sig,
2556 : : const char* tmppath,
2557 : : int& fd,
2558 : : off_t size,
2559 : : time_t mtime,
2560 : : const string& metric,
2561 : : const struct timespec& extract_begin)
2562 : : {
2563 [ + + ]: 1000 : if (tmppath != NULL)
2564 : : {
2565 : 900 : struct timespec extract_end;
2566 : 900 : clock_gettime (CLOCK_MONOTONIC, &extract_end);
2567 : 900 : double extract_time = (extract_end.tv_sec - extract_begin.tv_sec)
2568 : 900 : + (extract_end.tv_nsec - extract_begin.tv_nsec)/1.e9;
2569 [ + - ]: 1800 : fdcache.intern(b_source0, b_source1, tmppath, size, true, extract_time);
2570 : : }
2571 : :
2572 [ + + ]: 1000 : if (!section.empty ())
2573 : : {
2574 [ + - ]: 6 : int scn_fd = extract_section (fd, b_mtime0,
2575 [ + - ]: 12 : b_source0 + ":" + b_source1,
2576 : : section, extract_begin);
2577 : 6 : close (fd);
2578 [ + + ]: 6 : if (scn_fd >= 0)
2579 : 4 : fd = scn_fd;
2580 : : else
2581 : : {
2582 [ + - ]: 2 : if (verbose)
2583 [ + - ]: 6 : obatched (clog) << "cannot find section " << section
2584 : : << " for archive " << b_source0
2585 [ + - + - : 2 : << " file " << b_source1 << endl;
+ - + - +
- ]
2586 : 2 : return 0;
2587 : : }
2588 : :
2589 : 4 : struct stat fs;
2590 [ - + ]: 4 : if (fstat (fd, &fs) < 0)
2591 : : {
2592 : 0 : close (fd);
2593 : 0 : throw libc_exception (errno,
2594 [ # # # # : 0 : string ("fstat ") + b_source0 + string (" ") + section);
# # # # #
# # # ]
2595 : : }
2596 : 4 : size = fs.st_size;
2597 : : }
2598 : :
2599 : 998 : struct MHD_Response* r = MHD_create_response_from_fd (size, fd);
2600 [ - + ]: 998 : if (r == 0)
2601 : : {
2602 [ # # ]: 0 : if (verbose)
2603 [ # # # # ]: 0 : obatched(clog) << "cannot create fd-response for " << b_source0 << endl;
2604 : 0 : close(fd);
2605 : : }
2606 : : else
2607 : : {
2608 [ + - + - ]: 1996 : inc_metric ("http_responses_total","result",metric);
2609 : 998 : add_mhd_response_header (r, "Content-Type", "application/octet-stream");
2610 [ + - ]: 998 : add_mhd_response_header (r, "X-DEBUGINFOD-SIZE", to_string(size).c_str());
2611 : 998 : add_mhd_response_header (r, "X-DEBUGINFOD-ARCHIVE", b_source0.c_str());
2612 : 998 : add_mhd_response_header (r, "X-DEBUGINFOD-FILE", b_source1.c_str());
2613 [ - + ]: 998 : if(!ima_sig.empty()) add_mhd_response_header(r, "X-DEBUGINFOD-IMASIGNATURE", ima_sig.c_str());
2614 : 998 : add_mhd_last_modified (r, mtime);
2615 [ - + ]: 998 : if (verbose > 1)
2616 [ + - ]: 2994 : obatched(clog) << "serving " << metric << " " << b_source0
2617 : : << " file " << b_source1
2618 : : << " section=" << section
2619 [ + - + - : 998 : << " IMA signature=" << ima_sig << endl;
+ - + - +
- + - + -
+ - + - ]
2620 : : /* libmicrohttpd will close fd. */
2621 : : }
2622 : : return r;
2623 : : }
2624 : :
2625 : : static struct MHD_Response*
2626 : 1058 : handle_buildid_r_match (bool internal_req_p,
2627 : : int64_t b_mtime,
2628 : : const string& b_source0,
2629 : : const string& b_source1,
2630 : : int64_t b_id0,
2631 : : int64_t b_id1,
2632 : : const string& section,
2633 : : int *result_fd)
2634 : : {
2635 : 1058 : struct timespec extract_begin;
2636 : 1058 : clock_gettime (CLOCK_MONOTONIC, &extract_begin);
2637 : :
2638 : 1058 : struct stat fs;
2639 : 1058 : int rc = stat (b_source0.c_str(), &fs);
2640 [ + + ]: 1058 : if (rc != 0)
2641 [ + - + - : 116 : throw libc_exception (errno, string("stat ") + b_source0);
+ - ]
2642 : :
2643 [ - + ]: 1000 : if ((int64_t) fs.st_mtime != b_mtime)
2644 : : {
2645 [ # # ]: 0 : if (verbose)
2646 [ # # # # ]: 0 : obatched(clog) << "mtime mismatch for " << b_source0 << endl;
2647 : 0 : return 0;
2648 : : }
2649 : :
2650 : : // Extract the IMA per-file signature (if it exists)
2651 : 1000 : string ima_sig = "";
2652 : : #ifdef ENABLE_IMA_VERIFICATION
2653 : : do
2654 : : {
2655 : : FD_t rpm_fd;
2656 : : if(!(rpm_fd = Fopen(b_source0.c_str(), "r.ufdio"))) // read, uncompressed, rpm/rpmio.h
2657 : : {
2658 : : if (verbose) obatched(clog) << "There was an error while opening " << b_source0 << endl;
2659 : : break; // Exit IMA extraction
2660 : : }
2661 : :
2662 : : Header rpm_hdr;
2663 : : if(RPMRC_FAIL == rpmReadPackageFile(NULL, rpm_fd, b_source0.c_str(), &rpm_hdr))
2664 : : {
2665 : : if (verbose) obatched(clog) << "There was an error while reading the header of " << b_source0 << endl;
2666 : : Fclose(rpm_fd);
2667 : : break; // Exit IMA extraction
2668 : : }
2669 : :
2670 : : // Fill sig_tag_data with an alloc'd copy of the array of IMA signatures (if they exist)
2671 : : struct rpmtd_s sig_tag_data;
2672 : : rpmtdReset(&sig_tag_data);
2673 : : do{ /* A do-while so we can break out of the koji sigcache checking on failure */
2674 : : if(requires_koji_sigcache_mapping)
2675 : : {
2676 : : /* NB: Koji builds result in a directory structure like the following
2677 : : - PACKAGE/VERSION/RELEASE
2678 : : - ARCH1
2679 : : - foo.rpm // The rpm known by debuginfod
2680 : : - ...
2681 : : - ARCHN
2682 : : - data
2683 : : - signed // Periodically purged (and not scanned by debuginfod)
2684 : : - sigcache
2685 : : - ARCH1
2686 : : - foo.rpm.sig // An empty rpm header
2687 : : - ...
2688 : : - ARCHN
2689 : : - PACKAGE_KEYID1
2690 : : - ARCH1
2691 : : - foo.rpm.sig // The header of the signed rpm. This is the file we need to extract the IMA signatures
2692 : : - ...
2693 : : - ARCHN
2694 : : - ...
2695 : : - PACKAGE_KEYIDn
2696 : :
2697 : : We therefore need to do a mapping:
2698 : :
2699 : : P/V/R/A/N-V-R.A.rpm ->
2700 : : P/V/R/data/sigcache/KEYID/A/N-V-R.A.rpm.sig
2701 : :
2702 : : There are 2 key insights here
2703 : :
2704 : : 1. We need to go 2 directories down from sigcache to get to the
2705 : : rpm header. So to distinguish ARCH1/foo.rpm.sig and
2706 : : PACKAGE_KEYID1/ARCH1/foo.rpm.sig we can look 2 directories down
2707 : :
2708 : : 2. It's safe to assume that the user will have all of the
2709 : : required verification certs. So we can pick from any of the
2710 : : PACKAGE_KEYID* directories. For simplicity we choose first we
2711 : : match against
2712 : :
2713 : : See: https://pagure.io/koji/issue/3670
2714 : : */
2715 : :
2716 : : // Do the mapping from b_source0 to the koji path for the signed rpm header
2717 : : string signed_rpm_path = b_source0;
2718 : : size_t insert_pos = string::npos;
2719 : : for(int i = 0; i < 2; i++) insert_pos = signed_rpm_path.rfind("/", insert_pos) - 1;
2720 : : string globbed_path = signed_rpm_path.insert(insert_pos + 1, "/data/sigcache/*").append(".sig"); // The globbed path we're seeking
2721 : : glob_t pglob;
2722 : : int grc;
2723 : : if(0 != (grc = glob(globbed_path.c_str(), GLOB_NOSORT, NULL, &pglob)))
2724 : : {
2725 : : // Break out, but only report real errors
2726 : : if (verbose && grc != GLOB_NOMATCH) obatched(clog) << "There was an error (" << strerror(errno) << ") globbing " << globbed_path << endl;
2727 : : break; // Exit koji sigcache check
2728 : : }
2729 : : signed_rpm_path = pglob.gl_pathv[0]; // See insight 2 above
2730 : : globfree(&pglob);
2731 : :
2732 : : if (verbose > 2) obatched(clog) << "attempting IMA signature extraction from koji header " << signed_rpm_path << endl;
2733 : :
2734 : : FD_t sig_rpm_fd;
2735 : : if(NULL == (sig_rpm_fd = Fopen(signed_rpm_path.c_str(), "r")))
2736 : : {
2737 : : if (verbose) obatched(clog) << "There was an error while opening " << signed_rpm_path << endl;
2738 : : break; // Exit koji sigcache check
2739 : : }
2740 : :
2741 : : Header sig_hdr = headerRead(sig_rpm_fd, HEADER_MAGIC_YES /* Validate magic too */ );
2742 : : if (!sig_hdr || 1 != headerGet(sig_hdr, RPMSIGTAG_FILESIGNATURES, &sig_tag_data, HEADERGET_ALLOC))
2743 : : {
2744 : : if (verbose) obatched(clog) << "Unable to extract RPMSIGTAG_FILESIGNATURES from " << signed_rpm_path << endl;
2745 : : }
2746 : : headerFree(sig_hdr); // We can free here since sig_tag_data has an alloc'd copy of the data
2747 : : Fclose(sig_rpm_fd);
2748 : : }
2749 : : }while(false);
2750 : :
2751 : : if(0 == sig_tag_data.count)
2752 : : {
2753 : : // In the general case (or a fallback from the koji sigcache mapping not finding signatures)
2754 : : // we can just (try) extract the signatures from the rpm header
2755 : : if (1 != headerGet(rpm_hdr, RPMTAG_FILESIGNATURES, &sig_tag_data, HEADERGET_ALLOC))
2756 : : {
2757 : : if (verbose) obatched(clog) << "Unable to extract RPMTAG_FILESIGNATURES from " << b_source0 << endl;
2758 : : }
2759 : : }
2760 : : // Search the array for the signature coresponding to b_source1
2761 : : int idx = -1;
2762 : : char *sig = NULL;
2763 : : rpmfi hdr_fi = rpmfiNew(NULL, rpm_hdr, RPMTAG_BASENAMES, RPMFI_FLAGS_QUERY);
2764 : : do
2765 : : {
2766 : : sig = (char*)rpmtdNextString(&sig_tag_data);
2767 : : idx = rpmfiNext(hdr_fi);
2768 : : }
2769 : : while (idx != -1 && 0 != strcmp(b_source1.c_str(), rpmfiFN(hdr_fi)));
2770 : : rpmfiFree(hdr_fi);
2771 : :
2772 : : if(sig && 0 != strlen(sig) && idx != -1)
2773 : : {
2774 : : if (verbose > 2) obatched(clog) << "Found IMA signature for " << b_source1 << ":\n" << sig << endl;
2775 : : ima_sig = sig;
2776 : : inc_metric("http_responses_total","extra","ima-sigs-extracted");
2777 : : }
2778 : : else
2779 : : {
2780 : : if (verbose > 2) obatched(clog) << "Could not find IMA signature for " << b_source1 << endl;
2781 : : }
2782 : :
2783 : : rpmtdFreeData (&sig_tag_data);
2784 : : headerFree(rpm_hdr);
2785 : : Fclose(rpm_fd);
2786 : : } while(false);
2787 : : #endif
2788 : :
2789 : : // check for a match in the fdcache first
2790 [ + - ]: 1000 : int fd = fdcache.lookup(b_source0, b_source1);
2791 [ + + ]: 1000 : while (fd >= 0) // got one!; NB: this is really an if() with a possible branch out to the end
2792 : : {
2793 : 100 : rc = fstat(fd, &fs);
2794 [ - + ]: 100 : if (rc < 0) // disappeared?
2795 : : {
2796 [ # # ]: 0 : if (verbose)
2797 [ # # # # : 0 : obatched(clog) << "cannot fstat fdcache " << b_source0 << endl;
# # ]
2798 [ # # ]: 0 : close(fd);
2799 [ # # ]: 0 : fdcache.clear(b_source0, b_source1);
2800 : : break; // branch out of if "loop", to try new libarchive fetch attempt
2801 : : }
2802 : :
2803 [ + - + - ]: 100 : struct MHD_Response* r = create_buildid_r_response (b_mtime, b_source0,
2804 : : b_source1, section,
2805 : : ima_sig, NULL, fd,
2806 : : fs.st_size,
2807 : : fs.st_mtime,
2808 : : "archive fdcache",
2809 : : extract_begin);
2810 [ + - ]: 100 : if (r == 0)
2811 : : break; // branch out of if "loop", to try new libarchive fetch attempt
2812 [ + - ]: 100 : if (result_fd)
2813 : 100 : *result_fd = fd;
2814 : : return r;
2815 : : // NB: see, we never go around the 'loop' more than once
2816 : : }
2817 : :
2818 : : // no match ... look for a seekable entry
2819 : 900 : bool populate_seekable = ! passive_p;
2820 : 900 : unique_ptr<sqlite_ps> pp (new sqlite_ps (internal_req_p ? db : dbq,
2821 : : "rpm-seekable-query",
2822 : : "select type, size, offset, mtime from " BUILDIDS "_r_seekable "
2823 [ + - + - : 900 : "where file = ? and content = ?"));
+ - + + +
- ]
2824 [ + - + - : 900 : rc = pp->reset().bind(1, b_id0).bind(2, b_id1).step();
+ - + - ]
2825 [ + + ]: 900 : if (rc != SQLITE_DONE)
2826 : : {
2827 [ - + ]: 202 : if (rc != SQLITE_ROW)
2828 [ # # # # ]: 0 : throw sqlite_exception(rc, "step");
2829 : : // if we found a match in _r_seekable but we fail to extract it, don't
2830 : : // bother populating it again
2831 : 202 : populate_seekable = false;
2832 [ + - ]: 202 : const char* seekable_type = (const char*) sqlite3_column_text (*pp, 0);
2833 [ + - - + ]: 202 : if (seekable_type != NULL && strcmp (seekable_type, "xz") == 0)
2834 : : {
2835 [ + - ]: 202 : int64_t seekable_size = sqlite3_column_int64 (*pp, 1);
2836 [ + - ]: 202 : int64_t seekable_offset = sqlite3_column_int64 (*pp, 2);
2837 [ + - ]: 202 : int64_t seekable_mtime = sqlite3_column_int64 (*pp, 3);
2838 : :
2839 : 202 : char* tmppath = NULL;
2840 [ - + ]: 202 : if (asprintf (&tmppath, "%s/debuginfod-fdcache.XXXXXX", tmpdir.c_str()) < 0)
2841 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate tmppath");
2842 : 202 : defer_dtor<void*,void> tmmpath_freer (tmppath, free);
2843 : :
2844 [ + - ]: 202 : fd = extract_from_seekable_archive (b_source0, tmppath,
2845 : : seekable_offset, seekable_size);
2846 [ + - ]: 202 : if (fd >= 0)
2847 : : {
2848 : : // Set the mtime so the fdcache file mtimes propagate to future webapi
2849 : : // clients.
2850 : 202 : struct timespec tvs[2];
2851 : 202 : tvs[0].tv_sec = 0;
2852 : 202 : tvs[0].tv_nsec = UTIME_OMIT;
2853 : 202 : tvs[1].tv_sec = seekable_mtime;
2854 : 202 : tvs[1].tv_nsec = 0;
2855 : 202 : (void) futimens (fd, tvs); /* best effort */
2856 [ + - + - ]: 202 : struct MHD_Response* r = create_buildid_r_response (b_mtime,
2857 : : b_source0,
2858 : : b_source1,
2859 : : section,
2860 : : ima_sig,
2861 : : tmppath, fd,
2862 : : seekable_size,
2863 : : seekable_mtime,
2864 : : "seekable xz archive",
2865 : : extract_begin);
2866 [ + - ]: 202 : if (r != 0 && result_fd)
2867 : 202 : *result_fd = fd;
2868 : 202 : return r;
2869 : : }
2870 : 202 : }
2871 : : }
2872 : 698 : pp.reset();
2873 : :
2874 : : // still no match ... grumble, must process the archive
2875 [ + - ]: 698 : string archive_decoder = "/dev/null";
2876 [ + - ]: 698 : string archive_extension = "";
2877 [ + + ]: 1942 : for (auto&& arch : scan_archives)
2878 [ + + ]: 1244 : if (string_endswith(b_source0, arch.first))
2879 : : {
2880 [ + - ]: 698 : archive_extension = arch.first;
2881 [ + - ]: 1942 : archive_decoder = arch.second;
2882 : : }
2883 : 698 : FILE* fp;
2884 : :
2885 : 698 : defer_dtor<FILE*,int>::dtor_fn dfn;
2886 [ + + ]: 698 : if (archive_decoder != "cat")
2887 : : {
2888 [ + - + - : 1056 : string popen_cmd = archive_decoder + " " + shell_escape(b_source0);
+ - ]
2889 [ + - ]: 528 : fp = popen (popen_cmd.c_str(), "r"); // "e" O_CLOEXEC?
2890 : 528 : dfn = pclose;
2891 [ - + ]: 528 : if (fp == NULL)
2892 [ # # # # : 0 : throw libc_exception (errno, string("popen ") + popen_cmd);
# # ]
2893 : 528 : }
2894 : : else
2895 : : {
2896 [ + - ]: 170 : fp = fopen (b_source0.c_str(), "r");
2897 : 170 : dfn = fclose;
2898 [ - + ]: 170 : if (fp == NULL)
2899 [ # # # # : 0 : throw libc_exception (errno, string("fopen ") + b_source0);
# # ]
2900 : : }
2901 : 698 : defer_dtor<FILE*,int> fp_closer (fp, dfn);
2902 : :
2903 : 698 : struct archive *a;
2904 [ + - ]: 698 : a = archive_read_new();
2905 [ - + ]: 698 : if (a == NULL)
2906 [ # # # # ]: 0 : throw archive_exception("cannot create archive reader");
2907 : 698 : defer_dtor<struct archive*,int> archive_closer (a, archive_read_free);
2908 : :
2909 [ + - ]: 698 : rc = archive_read_support_format_all(a);
2910 [ - + ]: 698 : if (rc != ARCHIVE_OK)
2911 [ # # # # ]: 0 : throw archive_exception(a, "cannot select all format");
2912 [ + - ]: 698 : rc = archive_read_support_filter_all(a);
2913 [ - + ]: 698 : if (rc != ARCHIVE_OK)
2914 [ # # # # ]: 0 : throw archive_exception(a, "cannot select all filters");
2915 : :
2916 [ + - ]: 698 : rc = archive_read_open_FILE (a, fp);
2917 [ - + ]: 698 : if (rc != ARCHIVE_OK)
2918 : : {
2919 [ # # # # : 0 : obatched(clog) << "cannot open archive from pipe " << b_source0 << endl;
# # ]
2920 [ # # # # ]: 0 : throw archive_exception(a, "cannot open archive from pipe");
2921 : : }
2922 : :
2923 : : // If the archive was scanned in a version without _r_seekable, then we may
2924 : : // need to populate _r_seekable now. This can be removed the next time
2925 : : // BUILDIDS is updated.
2926 [ + + ]: 698 : if (populate_seekable)
2927 : : {
2928 [ + - ]: 696 : populate_seekable = is_seekable_archive (b_source0, a);
2929 [ + - ]: 696 : if (populate_seekable)
2930 : : {
2931 : : // NB: the names are already interned
2932 : 0 : pp.reset(new sqlite_ps (db, "rpm-seekable-insert2",
2933 : : "insert or ignore into " BUILDIDS "_r_seekable (file, content, type, size, offset, mtime) "
2934 : : "values (?, "
2935 : : "(select id from " BUILDIDS "_files "
2936 : : "where dirname = (select id from " BUILDIDS "_fileparts where name = ?) "
2937 : : "and basename = (select id from " BUILDIDS "_fileparts where name = ?) "
2938 [ # # # # : 0 : "), 'xz', ?, ?, ?)"));
# # # # ]
2939 : : }
2940 : : }
2941 : :
2942 : : // archive traversal is in five stages:
2943 : : // 1) before we find a matching entry, insert it into _r_seekable if needed or
2944 : : // skip it otherwise
2945 : : // 2) extract the matching entry (set r = result). Also insert it into
2946 : : // _r_seekable if needed
2947 : : // 3) extract some number of prefetched entries (just into fdcache). Also
2948 : : // insert them into _r_seekable if needed
2949 : : // 4) if needed, insert all of the remaining entries into _r_seekable
2950 : : // 5) abort any further processing
2951 : 698 : struct MHD_Response* r = 0; // will set in stage 2
2952 [ + + ]: 698 : unsigned prefetch_count =
2953 : : internal_req_p ? 0 : fdcache_prefetch; // will decrement in stage 3
2954 : :
2955 [ + + - + ]: 10822 : while(r == 0 || prefetch_count > 0 || populate_seekable) // stage 1-4
2956 : : {
2957 [ + - ]: 10782 : if (interrupted)
2958 : : break;
2959 : :
2960 : 10782 : struct archive_entry *e;
2961 [ + - ]: 10782 : rc = archive_read_next_header (a, &e);
2962 [ + + ]: 10782 : if (rc != ARCHIVE_OK)
2963 : : break;
2964 : :
2965 [ + - + + ]: 10126 : if (! S_ISREG(archive_entry_mode (e))) // skip non-files completely
2966 : 9428 : continue;
2967 : :
2968 [ + - ]: 2926 : string fn = canonicalized_archive_entry_pathname (e);
2969 : :
2970 [ - + ]: 2926 : if (populate_seekable)
2971 : : {
2972 : 0 : string dn, bn;
2973 : 0 : size_t slash = fn.rfind('/');
2974 [ # # ]: 0 : if (slash == std::string::npos) {
2975 [ # # ]: 0 : dn = "";
2976 [ # # ]: 0 : bn = fn;
2977 : : } else {
2978 [ # # ]: 0 : dn = fn.substr(0, slash);
2979 [ # # ]: 0 : bn = fn.substr(slash + 1);
2980 : : }
2981 : :
2982 [ # # ]: 0 : int64_t seekable_size = archive_entry_size (e);
2983 [ # # ]: 0 : int64_t seekable_offset = archive_filter_bytes (a, 0);
2984 [ # # ]: 0 : time_t seekable_mtime = archive_entry_mtime (e);
2985 : :
2986 [ # # ]: 0 : pp->reset();
2987 [ # # ]: 0 : pp->bind(1, b_id0);
2988 [ # # ]: 0 : pp->bind(2, dn);
2989 [ # # ]: 0 : pp->bind(3, bn);
2990 [ # # ]: 0 : pp->bind(4, seekable_size);
2991 [ # # ]: 0 : pp->bind(5, seekable_offset);
2992 [ # # ]: 0 : pp->bind(6, seekable_mtime);
2993 [ # # ]: 0 : rc = pp->step();
2994 [ # # ]: 0 : if (rc != SQLITE_DONE)
2995 [ # # # # ]: 0 : obatched(clog) << "recording seekable file=" << fn
2996 [ # # # # : 0 : << " sqlite3 error: " << (sqlite3_errstr(rc) ?: "?") << endl;
# # # # #
# ]
2997 [ # # ]: 0 : else if (verbose > 2)
2998 [ # # # # ]: 0 : obatched(clog) << "recorded seekable file=" << fn
2999 [ # # # # ]: 0 : << " size=" << seekable_size
3000 [ # # # # ]: 0 : << " offset=" << seekable_offset
3001 [ # # # # : 0 : << " mtime=" << seekable_mtime << endl;
# # ]
3002 [ # # ]: 0 : if (r != 0 && prefetch_count == 0) // stage 4
3003 : 0 : continue;
3004 : 0 : }
3005 : :
3006 [ + + + + ]: 2926 : if ((r == 0) && (fn != b_source1)) // stage 1
3007 : 1632 : continue;
3008 : :
3009 [ + - - + ]: 1294 : if (fdcache.probe (b_source0, fn) && // skip if already interned
3010 [ # # ]: 0 : fn != b_source1) // but only if we'd just be prefetching, PR29474
3011 : 0 : continue;
3012 : :
3013 : : // extract this file to a temporary file
3014 : 1294 : char* tmppath = NULL;
3015 : 1294 : rc = asprintf (&tmppath, "%s/debuginfod-fdcache.XXXXXX", tmpdir.c_str());
3016 [ - + ]: 1294 : if (rc < 0)
3017 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate tmppath");
3018 : 1294 : defer_dtor<void*,void> tmmpath_freer (tmppath, free);
3019 [ + - ]: 1294 : fd = mkstemp (tmppath);
3020 [ - + ]: 1294 : if (fd < 0)
3021 [ # # # # ]: 0 : throw libc_exception (errno, "cannot create temporary file");
3022 : : // NB: don't unlink (tmppath), as fdcache will take charge of it.
3023 : :
3024 : : // NB: this can take many uninterruptible seconds for a huge file
3025 [ + - ]: 1294 : rc = archive_read_data_into_fd (a, fd);
3026 [ - + ]: 1294 : if (rc != ARCHIVE_OK) // e.g. ENOSPC!
3027 : : {
3028 [ # # ]: 0 : close (fd);
3029 : 0 : unlink (tmppath);
3030 [ # # # # ]: 0 : throw archive_exception(a, "cannot extract file");
3031 : : }
3032 : :
3033 : : // Set the mtime so the fdcache file mtimes, even prefetched ones,
3034 : : // propagate to future webapi clients.
3035 : 1294 : struct timespec tvs[2];
3036 : 1294 : tvs[0].tv_sec = 0;
3037 : 1294 : tvs[0].tv_nsec = UTIME_OMIT;
3038 [ + - ]: 1294 : tvs[1].tv_sec = archive_entry_mtime(e);
3039 [ + - ]: 1294 : tvs[1].tv_nsec = archive_entry_mtime_nsec(e);
3040 : 1294 : (void) futimens (fd, tvs); /* best effort */
3041 : :
3042 [ + + ]: 1294 : if (r != 0) // stage 3
3043 : : {
3044 : 596 : struct timespec extract_end;
3045 : 596 : clock_gettime (CLOCK_MONOTONIC, &extract_end);
3046 : 596 : double extract_time = (extract_end.tv_sec - extract_begin.tv_sec)
3047 : 596 : + (extract_end.tv_nsec - extract_begin.tv_nsec)/1.e9;
3048 : : // NB: now we know we have a complete reusable file; make fdcache
3049 : : // responsible for unlinking it later.
3050 [ + - + - : 596 : fdcache.intern(b_source0, fn,
+ - ]
3051 : : tmppath, archive_entry_size(e),
3052 : : false, extract_time); // prefetched ones go to the prefetch cache
3053 : 596 : prefetch_count --;
3054 [ + - ]: 596 : close (fd); // we're not saving this fd to make a mhd-response from!
3055 : 596 : continue;
3056 : 596 : }
3057 : :
3058 [ + - + - : 698 : r = create_buildid_r_response (b_mtime, b_source0, b_source1, section,
+ - ]
3059 : : ima_sig, tmppath, fd,
3060 : : archive_entry_size(e),
3061 : : archive_entry_mtime(e),
3062 [ + - ]: 698 : archive_extension + " archive",
3063 : : extract_begin);
3064 [ + + ]: 698 : if (r == 0)
3065 : : break; // assume no chance of better luck around another iteration; no other copies of same file
3066 [ + - ]: 696 : if (result_fd)
3067 : 696 : *result_fd = fd;
3068 : 3522 : }
3069 : :
3070 : : // XXX: rpm/file not found: delete this R entry?
3071 : 698 : return r;
3072 : 1900 : }
3073 : :
3074 : : void
3075 : 650 : add_client_federation_headers(debuginfod_client *client, MHD_Connection* conn){
3076 : : // Transcribe incoming User-Agent:
3077 [ - + ]: 650 : string ua = MHD_lookup_connection_value (conn, MHD_HEADER_KIND, "User-Agent") ?: "";
3078 [ + - + - ]: 654 : string ua_complete = string("User-Agent: ") + ua;
3079 [ + - ]: 650 : debuginfod_add_http_header (client, ua_complete.c_str());
3080 : :
3081 : : // Compute larger XFF:, for avoiding info loss during
3082 : : // federation, and for future cyclicity detection.
3083 [ + - + + : 1278 : string xff = MHD_lookup_connection_value (conn, MHD_HEADER_KIND, "X-Forwarded-For") ?: "";
+ - ]
3084 [ + + ]: 650 : if (xff != "")
3085 [ + - ]: 52 : xff += string(", "); // comma separated list
3086 : :
3087 : 650 : unsigned int xff_count = 0;
3088 [ + + ]: 978 : for (auto&& i : xff){
3089 [ + + ]: 328 : if (i == ',') xff_count++;
3090 : : }
3091 : :
3092 : : // if X-Forwarded-For: exceeds N hops,
3093 : : // do not delegate a local lookup miss to upstream debuginfods.
3094 [ + + ]: 650 : if (xff_count >= forwarded_ttl_limit)
3095 : 4 : throw reportable_exception(MHD_HTTP_NOT_FOUND, "not found, --forwared-ttl-limit reached \
3096 [ + - ]: 8 : and will not query the upstream servers");
3097 : :
3098 : : // Compute the client's numeric IP address only - so can't merge with conninfo()
3099 [ + - ]: 646 : const union MHD_ConnectionInfo *u = MHD_get_connection_info (conn,
3100 : : MHD_CONNECTION_INFO_CLIENT_ADDRESS);
3101 [ + - ]: 646 : struct sockaddr *so = u ? u->client_addr : 0;
3102 : 646 : char hostname[256] = ""; // RFC1035
3103 [ + - - + ]: 646 : if (so && so->sa_family == AF_INET) {
3104 [ # # ]: 0 : (void) getnameinfo (so, sizeof (struct sockaddr_in), hostname, sizeof (hostname), NULL, 0,
3105 : : NI_NUMERICHOST);
3106 [ + - ]: 646 : } else if (so && so->sa_family == AF_INET6) {
3107 : 646 : struct sockaddr_in6* addr6 = (struct sockaddr_in6*) so;
3108 [ + - + - : 646 : if (IN6_IS_ADDR_V4MAPPED(&addr6->sin6_addr)) {
- + ]
3109 : 646 : struct sockaddr_in addr4;
3110 [ + - ]: 646 : memset (&addr4, 0, sizeof(addr4));
3111 : 646 : addr4.sin_family = AF_INET;
3112 : 646 : addr4.sin_port = addr6->sin6_port;
3113 [ + - ]: 646 : memcpy (&addr4.sin_addr.s_addr, addr6->sin6_addr.s6_addr+12, sizeof(addr4.sin_addr.s_addr));
3114 [ + - ]: 646 : (void) getnameinfo ((struct sockaddr*) &addr4, sizeof (addr4),
3115 : : hostname, sizeof (hostname), NULL, 0,
3116 : : NI_NUMERICHOST);
3117 : : } else {
3118 [ # # ]: 0 : (void) getnameinfo (so, sizeof (struct sockaddr_in6), hostname, sizeof (hostname), NULL, 0,
3119 : : NI_NUMERICHOST);
3120 : : }
3121 : : }
3122 : :
3123 [ + - + - : 1296 : string xff_complete = string("X-Forwarded-For: ")+xff+string(hostname);
+ - + - ]
3124 [ + - ]: 646 : debuginfod_add_http_header (client, xff_complete.c_str());
3125 : 654 : }
3126 : :
3127 : : static struct MHD_Response*
3128 : 2210 : handle_buildid_match (bool internal_req_p,
3129 : : int64_t b_mtime,
3130 : : const string& b_stype,
3131 : : const string& b_source0,
3132 : : const string& b_source1,
3133 : : int64_t b_id0,
3134 : : int64_t b_id1,
3135 : : const string& section,
3136 : : int *result_fd)
3137 : : {
3138 : 2210 : try
3139 : : {
3140 [ + + ]: 2210 : if (b_stype == "F")
3141 [ + - ]: 1152 : return handle_buildid_f_match(internal_req_p, b_mtime, b_source0,
3142 : : section, result_fd);
3143 [ + - ]: 1058 : else if (b_stype == "R")
3144 [ + + ]: 1058 : return handle_buildid_r_match(internal_req_p, b_mtime, b_source0,
3145 : : b_source1, b_id0, b_id1, section,
3146 : : result_fd);
3147 : : }
3148 [ - + ]: 58 : catch (const reportable_exception &e)
3149 : : {
3150 [ + - ]: 58 : e.report(clog);
3151 : : // Report but swallow libc etc. errors here; let the caller
3152 : : // iterate to other matches of the content.
3153 : 58 : }
3154 : :
3155 : : return 0;
3156 : : }
3157 : :
3158 : :
3159 : : static int
3160 : 4 : debuginfod_find_progress (debuginfod_client *, long a, long b)
3161 : : {
3162 [ - + ]: 4 : if (verbose > 4)
3163 [ # # # # : 0 : obatched(clog) << "federated debuginfod progress=" << a << "/" << b << endl;
# # # # ]
3164 : :
3165 : 4 : return interrupted;
3166 : : }
3167 : :
3168 : :
3169 : : // a little lru pool of debuginfod_client*s for reuse between query threads
3170 : :
3171 : : mutex dc_pool_lock;
3172 : : deque<debuginfod_client*> dc_pool;
3173 : :
3174 : 670 : debuginfod_client* debuginfod_pool_begin()
3175 : : {
3176 : 670 : unique_lock<mutex> lock(dc_pool_lock);
3177 [ + + ]: 670 : if (dc_pool.size() > 0)
3178 : : {
3179 [ + - + - : 1276 : inc_metric("dc_pool_op_count","op","begin-reuse");
+ - + - ]
3180 : 638 : debuginfod_client *c = dc_pool.front();
3181 : 638 : dc_pool.pop_front();
3182 : 638 : return c;
3183 : : }
3184 [ + - + - : 64 : inc_metric("dc_pool_op_count","op","begin-new");
+ - + - ]
3185 [ + - ]: 32 : return debuginfod_begin();
3186 : 670 : }
3187 : :
3188 : :
3189 : 164 : void debuginfod_pool_groom()
3190 : : {
3191 : 164 : unique_lock<mutex> lock(dc_pool_lock);
3192 [ + + ]: 196 : while (dc_pool.size() > 0)
3193 : : {
3194 [ + - + - : 64 : inc_metric("dc_pool_op_count","op","end");
+ - + - ]
3195 [ + - ]: 32 : debuginfod_end(dc_pool.front());
3196 : 32 : dc_pool.pop_front();
3197 : : }
3198 : 164 : }
3199 : :
3200 : :
3201 : 670 : void debuginfod_pool_end(debuginfod_client* c)
3202 : : {
3203 : 670 : unique_lock<mutex> lock(dc_pool_lock);
3204 [ + - + - : 1340 : inc_metric("dc_pool_op_count","op","end-save");
+ - + - ]
3205 [ + - ]: 670 : dc_pool.push_front(c); // accelerate reuse, vs. push_back
3206 : 670 : }
3207 : :
3208 : :
3209 : : static struct MHD_Response*
3210 : 2797 : handle_buildid (MHD_Connection* conn,
3211 : : const string& buildid /* unsafe */,
3212 : : string& artifacttype /* unsafe, cleanse on exception/return */,
3213 : : const string& suffix /* unsafe */,
3214 : : int *result_fd)
3215 : : {
3216 : : // validate artifacttype
3217 [ + + ]: 2797 : string atype_code;
3218 [ + + + - ]: 2797 : if (artifacttype == "debuginfo") atype_code = "D";
3219 [ + + + - ]: 1829 : else if (artifacttype == "executable") atype_code = "E";
3220 [ + + + - ]: 1138 : else if (artifacttype == "source") atype_code = "S";
3221 [ + + + - ]: 12 : else if (artifacttype == "section") atype_code = "I";
3222 : : else {
3223 [ + - ]: 4 : artifacttype = "invalid"; // PR28242 ensure http_resposes metrics don't propagate unclean user data
3224 [ + - ]: 8 : throw reportable_exception("invalid artifacttype");
3225 : : }
3226 : :
3227 [ + + ]: 2793 : if (conn != 0)
3228 [ + - + - : 5831 : inc_metric("http_requests_total", "type", artifacttype);
+ - ]
3229 : :
3230 [ + + ]: 2793 : string section;
3231 [ + + ]: 2793 : if (atype_code == "I")
3232 : : {
3233 [ - + ]: 8 : if (suffix.size () < 2)
3234 [ # # ]: 0 : throw reportable_exception ("invalid section suffix");
3235 : :
3236 : : // Remove leading '/'
3237 [ + - ]: 8 : section = suffix.substr(1);
3238 : : }
3239 : :
3240 [ + + - + ]: 2793 : if (atype_code == "S" && suffix == "")
3241 [ # # ]: 0 : throw reportable_exception("invalid source suffix");
3242 : :
3243 : : // validate buildid
3244 [ + + ]: 2793 : if ((buildid.size() < 2) || // not empty
3245 [ + + + - : 5581 : (buildid.size() % 2) || // even number
+ - ]
3246 : 2788 : (buildid.find_first_not_of("0123456789abcdef") != string::npos)) // pure tasty lowercase hex
3247 [ + - ]: 10 : throw reportable_exception("invalid buildid");
3248 : :
3249 [ + - ]: 2788 : if (verbose > 1)
3250 [ + - + - ]: 8364 : obatched(clog) << "searching for buildid=" << buildid << " artifacttype=" << artifacttype
3251 [ + - + - : 2788 : << " suffix=" << suffix << endl;
+ - + - +
- ]
3252 : :
3253 : : // If invoked from the scanner threads, use the scanners' read-write
3254 : : // connection. Otherwise use the web query threads' read-only connection.
3255 [ + + ]: 2788 : sqlite3 *thisdb = (conn == 0) ? db : dbq;
3256 : :
3257 : 2788 : sqlite_ps *pp = 0;
3258 : :
3259 [ + + ]: 2788 : if (atype_code == "D")
3260 : : {
3261 : 968 : pp = new sqlite_ps (thisdb, "mhd-query-d",
3262 : : "select mtime, sourcetype, source0, source1, id0, id1 from " BUILDIDS "_query_d2 where buildid = ? "
3263 [ + - + - : 1936 : "order by mtime desc");
+ - + - ]
3264 [ + - ]: 968 : pp->reset();
3265 [ + - ]: 968 : pp->bind(1, buildid);
3266 : : }
3267 [ + + ]: 1820 : else if (atype_code == "E")
3268 : : {
3269 : 686 : pp = new sqlite_ps (thisdb, "mhd-query-e",
3270 : : "select mtime, sourcetype, source0, source1, id0, id1 from " BUILDIDS "_query_e2 where buildid = ? "
3271 [ + - + - : 1372 : "order by mtime desc");
+ - + - ]
3272 [ + - ]: 686 : pp->reset();
3273 [ + - ]: 686 : pp->bind(1, buildid);
3274 : : }
3275 [ + + ]: 1134 : else if (atype_code == "S")
3276 : : {
3277 : : // PR25548
3278 : : // Incoming source queries may come in with either dwarf-level OR canonicalized paths.
3279 : : // We let the query pass with either one.
3280 : :
3281 : 1126 : pp = new sqlite_ps (thisdb, "mhd-query-s",
3282 : : "select mtime, sourcetype, source0, source1 from " BUILDIDS "_query_s where buildid = ? and artifactsrc in (?,?) "
3283 [ + - + - : 2252 : "order by sharedprefix(source0,source0ref) desc, mtime desc");
+ - + - ]
3284 [ + - ]: 1126 : pp->reset();
3285 [ + - ]: 1126 : pp->bind(1, buildid);
3286 : : // NB: we don't store the non-canonicalized path names any more, but old databases
3287 : : // might have them (and no canon ones), so we keep searching for both.
3288 [ + - ]: 1126 : pp->bind(2, suffix);
3289 [ + - + - ]: 2893 : pp->bind(3, canon_pathname(suffix));
3290 : : }
3291 [ + - ]: 8 : else if (atype_code == "I")
3292 : : {
3293 : 8 : pp = new sqlite_ps (thisdb, "mhd-query-i",
3294 : : "select mtime, sourcetype, source0, source1, 1 as debug_p from " BUILDIDS "_query_d2 where buildid = ? "
3295 : : "union all "
3296 : : "select mtime, sourcetype, source0, source1, 0 as debug_p from " BUILDIDS "_query_e2 where buildid = ? "
3297 [ + - + - : 16 : "order by debug_p desc, mtime desc");
+ - + - ]
3298 [ + - ]: 8 : pp->reset();
3299 [ + - ]: 8 : pp->bind(1, buildid);
3300 [ + - ]: 8 : pp->bind(2, buildid);
3301 : : }
3302 : 2788 : unique_ptr<sqlite_ps> ps_closer(pp); // release pp if exception or return
3303 : :
3304 : 2788 : bool do_upstream_section_query = true;
3305 : :
3306 : : // consume all the rows
3307 : 2912 : while (1)
3308 : : {
3309 [ + - ]: 2850 : int rc = pp->step();
3310 [ + + ]: 2850 : if (rc == SQLITE_DONE) break;
3311 [ - + ]: 2210 : if (rc != SQLITE_ROW)
3312 [ # # # # ]: 0 : throw sqlite_exception(rc, "step");
3313 : :
3314 [ + - ]: 2210 : int64_t b_mtime = sqlite3_column_int64 (*pp, 0);
3315 [ + - - + : 2210 : string b_stype = string((const char*) sqlite3_column_text (*pp, 1) ?: ""); /* by DDL may not be NULL */
+ - ]
3316 [ + - - + : 2210 : string b_source0 = string((const char*) sqlite3_column_text (*pp, 2) ?: ""); /* may be NULL */
+ - ]
3317 [ + - + + : 3362 : string b_source1 = string((const char*) sqlite3_column_text (*pp, 3) ?: ""); /* may be NULL */
+ - ]
3318 : 2210 : int64_t b_id0 = 0, b_id1 = 0;
3319 [ + + + + ]: 2210 : if (atype_code == "D" || atype_code == "E")
3320 : : {
3321 [ + - ]: 1070 : b_id0 = sqlite3_column_int64 (*pp, 4);
3322 [ + - ]: 1070 : b_id1 = sqlite3_column_int64 (*pp, 5);
3323 : : }
3324 : :
3325 [ + - ]: 2210 : if (verbose > 1)
3326 [ + - + - ]: 6630 : obatched(clog) << "found mtime=" << b_mtime << " stype=" << b_stype
3327 [ + - + - : 2210 : << " source0=" << b_source0 << " source1=" << b_source1 << endl;
+ - + - +
- + - +
- ]
3328 : :
3329 : : // Try accessing the located match.
3330 : : // XXX: in case of multiple matches, attempt them in parallel?
3331 [ + - ]: 2210 : auto r = handle_buildid_match (conn ? false : true,
3332 : : b_mtime, b_stype, b_source0, b_source1,
3333 : : b_id0, b_id1, section, result_fd);
3334 [ + + ]: 2210 : if (r)
3335 : 2148 : return r;
3336 : :
3337 : : // If a debuginfo file matching BUILDID was found but didn't contain
3338 : : // the desired section, then the section should not exist. Don't
3339 : : // bother querying upstream servers.
3340 [ + + + - : 62 : if (!section.empty () && (sqlite3_column_int (*pp, 4) == 1))
- + ]
3341 : : {
3342 : 4 : struct stat st;
3343 : :
3344 : : // For "F" sourcetype, check if the debuginfo exists. For "R"
3345 : : // sourcetype, check if the debuginfo was interned into the fdcache.
3346 [ - + ]: 2 : if ((b_stype == "F" && (stat (b_source0.c_str (), &st) == 0))
3347 [ + + + - : 4 : || (b_stype == "R" && fdcache.probe (b_source0, b_source1)))
+ - + - ]
3348 : : do_upstream_section_query = false;
3349 : : }
3350 : 2210 : }
3351 [ + - ]: 640 : pp->reset();
3352 : :
3353 [ - + ]: 640 : if (!do_upstream_section_query)
3354 [ # # ]: 0 : throw reportable_exception(MHD_HTTP_NOT_FOUND, "not found");
3355 : :
3356 : : // We couldn't find it in the database. Last ditch effort
3357 : : // is to defer to other debuginfo servers.
3358 : :
3359 : 640 : int fd = -1;
3360 [ + - ]: 640 : debuginfod_client *client = debuginfod_pool_begin ();
3361 [ - + ]: 640 : if (client == NULL)
3362 [ # # # # ]: 0 : throw libc_exception(errno, "debuginfod client pool alloc");
3363 : 640 : defer_dtor<debuginfod_client*,void> client_closer (client, debuginfod_pool_end);
3364 : :
3365 [ + - ]: 640 : debuginfod_set_progressfn (client, & debuginfod_find_progress);
3366 : :
3367 [ + + ]: 640 : if (conn)
3368 [ + + ]: 620 : add_client_federation_headers(client, conn);
3369 : :
3370 [ + + ]: 636 : if (artifacttype == "debuginfo")
3371 [ + - ]: 88 : fd = debuginfod_find_debuginfo (client,
3372 [ + - ]: 88 : (const unsigned char*) buildid.c_str(),
3373 : : 0, NULL);
3374 [ + + ]: 548 : else if (artifacttype == "executable")
3375 [ + - ]: 546 : fd = debuginfod_find_executable (client,
3376 [ + - ]: 546 : (const unsigned char*) buildid.c_str(),
3377 : : 0, NULL);
3378 [ + - ]: 2 : else if (artifacttype == "source")
3379 [ + - ]: 2 : fd = debuginfod_find_source (client,
3380 [ + - ]: 2 : (const unsigned char*) buildid.c_str(),
3381 : : 0, suffix.c_str(), NULL);
3382 [ # # ]: 0 : else if (artifacttype == "section")
3383 [ # # ]: 0 : fd = debuginfod_find_section (client,
3384 [ # # ]: 0 : (const unsigned char*) buildid.c_str(),
3385 : : 0, section.c_str(), NULL);
3386 : :
3387 [ + + ]: 636 : if (fd >= 0)
3388 : : {
3389 [ + - ]: 4 : if (conn != 0)
3390 [ + - + - : 644 : inc_metric ("http_responses_total","result","upstream");
+ - + - ]
3391 : 4 : struct stat s;
3392 : 4 : int rc = fstat (fd, &s);
3393 [ + - ]: 4 : if (rc == 0)
3394 : : {
3395 [ + - ]: 4 : auto r = MHD_create_response_from_fd ((uint64_t) s.st_size, fd);
3396 [ + - ]: 4 : if (r)
3397 : : {
3398 [ + - ]: 4 : add_mhd_response_header (r, "Content-Type",
3399 : : "application/octet-stream");
3400 : : // Copy the incoming headers
3401 [ + - ]: 4 : const char * hdrs = debuginfod_get_headers(client);
3402 [ + - ]: 4 : string header_dup;
3403 [ + - ]: 4 : if (hdrs)
3404 [ + - ]: 4 : header_dup = string(hdrs);
3405 : : // Parse the "header: value\n" lines into (h,v) tuples and pass on
3406 : 20 : while(1)
3407 : : {
3408 : 12 : size_t newline = header_dup.find('\n');
3409 [ + + ]: 12 : if (newline == string::npos) break;
3410 : 8 : size_t colon = header_dup.find(':');
3411 [ + - ]: 8 : if (colon == string::npos) break;
3412 [ + - ]: 8 : string header = header_dup.substr(0,colon);
3413 [ + - ]: 8 : string value = header_dup.substr(colon+1,newline-colon-1);
3414 : : // strip leading spaces from value
3415 : 8 : size_t nonspace = value.find_first_not_of(" ");
3416 [ + - ]: 8 : if (nonspace != string::npos)
3417 [ + - ]: 8 : value = value.substr(nonspace);
3418 [ + - ]: 8 : add_mhd_response_header(r, header.c_str(), value.c_str());
3419 [ + - ]: 8 : header_dup = header_dup.substr(newline+1);
3420 : 8 : }
3421 : :
3422 [ + - ]: 4 : add_mhd_last_modified (r, s.st_mtime);
3423 [ + - ]: 4 : if (verbose > 1)
3424 [ + - + - ]: 8 : obatched(clog) << "serving file from upstream debuginfod/cache" << endl;
3425 [ + - ]: 4 : if (result_fd)
3426 : 4 : *result_fd = fd;
3427 : 4 : return r; // NB: don't close fd; libmicrohttpd will
3428 : 0 : }
3429 : : }
3430 [ # # ]: 0 : close (fd);
3431 : : }
3432 : : else
3433 [ + + ]: 632 : switch(fd)
3434 : : {
3435 : : case -ENOSYS:
3436 : : break;
3437 : : case -ENOENT:
3438 : : break;
3439 : 532 : default: // some more tricky error
3440 [ + - + - ]: 1064 : throw libc_exception(-fd, "upstream debuginfod query failed");
3441 : : }
3442 : :
3443 [ + - ]: 200 : throw reportable_exception(MHD_HTTP_NOT_FOUND, "not found");
3444 : 3429 : }
3445 : :
3446 : :
3447 : : ////////////////////////////////////////////////////////////////////////
3448 : :
3449 : : static map<string,double> metrics; // arbitrary data for /metrics query
3450 : : // NB: store int64_t since all our metrics are integers; prometheus accepts double
3451 : : static mutex metrics_lock;
3452 : : // NB: these objects get released during the process exit via global dtors
3453 : : // do not call them from within other global dtors
3454 : :
3455 : : // utility function for assembling prometheus-compatible
3456 : : // name="escaped-value" strings
3457 : : // https://prometheus.io/docs/instrumenting/exposition_formats/
3458 : : static string
3459 : 826389 : metric_label(const string& name, const string& value)
3460 : : {
3461 : 826389 : string x = name + "=\"";
3462 [ + + ]: 14221211 : for (auto&& c : value)
3463 [ - - - + ]: 13395441 : switch(c)
3464 : : {
3465 [ # # ]: 0 : case '\\': x += "\\\\"; break;
3466 [ # # ]: 0 : case '\"': x += "\\\""; break;
3467 [ # # ]: 0 : case '\n': x += "\\n"; break;
3468 [ + - ]: 26790202 : default: x += c; break;
3469 : : }
3470 [ + - ]: 825770 : x += "\"";
3471 : 825957 : return x;
3472 : 0 : }
3473 : :
3474 : :
3475 : : // add prometheus-format metric name + label tuple (if any) + value
3476 : :
3477 : : static void
3478 : 12192 : set_metric(const string& metric, double value)
3479 : : {
3480 : 12192 : unique_lock<mutex> lock(metrics_lock);
3481 [ + - ]: 12192 : metrics[metric] = value;
3482 : 12192 : }
3483 : : static void
3484 : 102 : inc_metric(const string& metric)
3485 : : {
3486 : 102 : unique_lock<mutex> lock(metrics_lock);
3487 [ + - ]: 102 : metrics[metric] ++;
3488 : 102 : }
3489 : : static void
3490 : 8337 : set_metric(const string& metric,
3491 : : const string& lname, const string& lvalue,
3492 : : double value)
3493 : : {
3494 [ + - + - ]: 16677 : string key = (metric + "{" + metric_label(lname, lvalue) + "}");
3495 [ + - ]: 8340 : unique_lock<mutex> lock(metrics_lock);
3496 [ + - ]: 8340 : metrics[key] = value;
3497 : 8340 : }
3498 : :
3499 : : static void
3500 : 398550 : inc_metric(const string& metric,
3501 : : const string& lname, const string& lvalue)
3502 : : {
3503 [ + - + - ]: 797090 : string key = (metric + "{" + metric_label(lname, lvalue) + "}");
3504 [ + - ]: 398534 : unique_lock<mutex> lock(metrics_lock);
3505 [ + - ]: 398573 : metrics[key] ++;
3506 : 398542 : }
3507 : : static void
3508 : 379435 : add_metric(const string& metric,
3509 : : const string& lname, const string& lvalue,
3510 : : double value)
3511 : : {
3512 [ + - + - ]: 758909 : string key = (metric + "{" + metric_label(lname, lvalue) + "}");
3513 [ + - ]: 379472 : unique_lock<mutex> lock(metrics_lock);
3514 [ + - ]: 379592 : metrics[key] += value;
3515 : 379585 : }
3516 : : static void
3517 : 102 : add_metric(const string& metric,
3518 : : double value)
3519 : : {
3520 : 102 : unique_lock<mutex> lock(metrics_lock);
3521 [ + - ]: 102 : metrics[metric] += value;
3522 : 102 : }
3523 : :
3524 : :
3525 : : // and more for higher arity labels if needed
3526 : :
3527 : : static void
3528 : 10164 : inc_metric(const string& metric,
3529 : : const string& lname, const string& lvalue,
3530 : : const string& rname, const string& rvalue)
3531 : : {
3532 [ + - ]: 20328 : string key = (metric + "{"
3533 [ + - + - ]: 40656 : + metric_label(lname, lvalue) + ","
3534 [ + - ]: 30492 : + metric_label(rname, rvalue) + "}");
3535 [ + - ]: 10164 : unique_lock<mutex> lock(metrics_lock);
3536 [ + - ]: 10164 : metrics[key] ++;
3537 : 10164 : }
3538 : : static void
3539 : 10164 : add_metric(const string& metric,
3540 : : const string& lname, const string& lvalue,
3541 : : const string& rname, const string& rvalue,
3542 : : double value)
3543 : : {
3544 [ + - ]: 20328 : string key = (metric + "{"
3545 [ + - + - ]: 40656 : + metric_label(lname, lvalue) + ","
3546 [ + - ]: 30492 : + metric_label(rname, rvalue) + "}");
3547 [ + - ]: 10164 : unique_lock<mutex> lock(metrics_lock);
3548 [ + - ]: 10164 : metrics[key] += value;
3549 : 10164 : }
3550 : :
3551 : : static struct MHD_Response*
3552 : 751 : handle_metrics (off_t* size)
3553 : : {
3554 : 751 : stringstream o;
3555 : 751 : {
3556 [ + - ]: 751 : unique_lock<mutex> lock(metrics_lock);
3557 [ + + ]: 79024 : for (auto&& i : metrics)
3558 [ + - ]: 78273 : o << i.first
3559 : : << " "
3560 [ + - + - ]: 78273 : << std::setprecision(std::numeric_limits<double>::digits10 + 1)
3561 [ + - + - ]: 78273 : << i.second
3562 : 78273 : << endl;
3563 : 0 : }
3564 [ + - ]: 751 : const string& os = o.str();
3565 [ + - ]: 751 : MHD_Response* r = MHD_create_response_from_buffer (os.size(),
3566 [ + - ]: 751 : (void*) os.c_str(),
3567 : : MHD_RESPMEM_MUST_COPY);
3568 [ + - ]: 751 : if (r != NULL)
3569 : : {
3570 [ + - ]: 751 : *size = os.size();
3571 [ + - ]: 751 : add_mhd_response_header (r, "Content-Type", "text/plain");
3572 : : }
3573 : 1502 : return r;
3574 : 751 : }
3575 : :
3576 : :
3577 : : static struct MHD_Response*
3578 : 30 : handle_metadata (MHD_Connection* conn,
3579 : : string key, string value, off_t* size)
3580 : : {
3581 : 30 : MHD_Response* r;
3582 : : // Because this query can take on the order of many seconds, we need
3583 : : // to prevent DoS against the other normal quick queries, so we use
3584 : : // a dedicated database connection.
3585 : 30 : sqlite3 *thisdb = 0;
3586 : 30 : int rc = sqlite3_open_v2 (db_path.c_str(), &thisdb, (SQLITE_OPEN_READONLY
3587 : : |SQLITE_OPEN_URI
3588 : : |SQLITE_OPEN_PRIVATECACHE
3589 : : |SQLITE_OPEN_NOMUTEX), /* private to us */
3590 : : NULL);
3591 [ - + ]: 30 : if (rc)
3592 [ # # # # ]: 0 : throw sqlite_exception(rc, "cannot open database for metadata query");
3593 : 30 : defer_dtor<sqlite3*,int> sqlite_db_closer (thisdb, sqlite3_close_v2);
3594 : :
3595 : : // Query locally for matching e, d files
3596 [ + + ]: 30 : string op;
3597 [ + + ]: 30 : if (key == "glob")
3598 [ + - ]: 26 : op = "glob";
3599 [ + - ]: 4 : else if (key == "file")
3600 [ + - ]: 4 : op = "=";
3601 : : else
3602 [ # # ]: 0 : throw reportable_exception("/metadata webapi error, unsupported key");
3603 : :
3604 : : // Since PR30378, the file names are segmented into two tables. We
3605 : : // could do a glob/= search over the _files_v view that combines
3606 : : // them, but that means that the entire _files_v thing has to be
3607 : : // materialized & scanned to do the query. Slow! Instead, we can
3608 : : // segment the incoming file/glob pattern into dirname / basename
3609 : : // parts, and apply them to the corresponding table. This is done
3610 : : // by splitting the value at the last "/". If absent, the same
3611 : : // convention as is used in register_file_name().
3612 : :
3613 : 30 : string dirname, bname; // basename is a "poisoned" identifier on some distros
3614 : 30 : size_t slash = value.rfind('/');
3615 [ - + ]: 30 : if (slash == std::string::npos) {
3616 [ # # ]: 0 : dirname = "";
3617 [ # # ]: 0 : bname = value;
3618 : : } else {
3619 [ + - ]: 30 : dirname = value.substr(0, slash);
3620 [ + - ]: 30 : bname = value.substr(slash+1);
3621 : : }
3622 : :
3623 : : // NB: further optimization is possible: replacing the 'glob' op
3624 : : // with simple equality, if the corresponding value segment lacks
3625 : : // metacharacters. sqlite may or may not be smart enough to do so,
3626 : : // so we help out.
3627 [ + - ]: 30 : string metacharacters = "[]*?";
3628 [ + + + + : 56 : string dop = (op == "glob" && dirname.find_first_of(metacharacters) == string::npos) ? "=" : op;
+ - + - ]
3629 [ + + - + : 56 : string bop = (op == "glob" && bname.find_first_of(metacharacters) == string::npos) ? "=" : op;
- - + - ]
3630 : :
3631 : 30 : string sql = string(
3632 : : // explicit query r_de and f_de once here, rather than the query_d and query_e
3633 : : // separately, because they scan the same tables, so we'd double the work
3634 : : "select d1.executable_p, d1.debuginfo_p, 0 as source_p, "
3635 : : " b1.hex, f1d.name || '/' || f1b.name as file, a1.name as archive "
3636 : : "from " BUILDIDS "_r_de d1, " BUILDIDS "_files f1, " BUILDIDS "_fileparts f1b, " BUILDIDS "_fileparts f1d, "
3637 : : BUILDIDS "_buildids b1, " BUILDIDS "_files_v a1 "
3638 : : "where f1.id = d1.content and a1.id = d1.file and d1.buildid = b1.id "
3639 [ + - + - ]: 90 : " and f1d.name " + dop + " ? and f1b.name " + bop + " ? and f1.dirname = f1d.id and f1.basename = f1b.id "
3640 : : "union all \n"
3641 : : "select d2.executable_p, d2.debuginfo_p, 0, "
3642 : : " b2.hex, f2d.name || '/' || f2b.name, NULL "
3643 : : "from " BUILDIDS "_f_de d2, " BUILDIDS "_files f2, " BUILDIDS "_fileparts f2b, " BUILDIDS "_fileparts f2d, "
3644 : : BUILDIDS "_buildids b2 "
3645 : : "where f2.id = d2.file and d2.buildid = b2.id "
3646 [ + - + - ]: 90 : " and f2d.name " + dop + " ? and f2b.name " + bop + " ? "
3647 : 30 : " and f2.dirname = f2d.id and f2.basename = f2b.id");
3648 : :
3649 : : // NB: we could query source file names too, thusly:
3650 : : //
3651 : : // select * from " BUILDIDS "_buildids b, " BUILDIDS "_files_v f1, " BUILDIDS "_r_sref sr
3652 : : // where b.id = sr.buildid and f1.id = sr.artifactsrc and f1.name " + op + "?"
3653 : : // UNION ALL something with BUILDIDS "_f_s"
3654 : : //
3655 : : // But the first part of this query cannot run fast without the same index temp-created
3656 : : // during "maxigroom":
3657 : : // create index " BUILDIDS "_r_sref_arc on " BUILDIDS "_r_sref(artifactsrc);
3658 : : // and unfortunately this index is HUGE. It's similar to the size of the _r_sref
3659 : : // table, which is already the largest part of a debuginfod index. Adding that index
3660 : : // would nearly double the .sqlite db size.
3661 : :
3662 [ + - + - : 30 : sqlite_ps *pp = new sqlite_ps (thisdb, "mhd-query-meta-glob", sql);
+ - ]
3663 [ + - ]: 30 : pp->reset();
3664 [ + - ]: 30 : pp->bind(1, dirname);
3665 [ + - ]: 30 : pp->bind(2, bname);
3666 [ + - ]: 30 : pp->bind(3, dirname);
3667 [ + - ]: 30 : pp->bind(4, bname);
3668 : 30 : unique_ptr<sqlite_ps> ps_closer(pp); // release pp if exception or return
3669 : 30 : pp->reset_timeout(metadata_maxtime_s);
3670 : :
3671 [ + - ]: 30 : json_object *metadata = json_object_new_object();
3672 [ - + - - : 30 : if (!metadata) throw libc_exception(ENOMEM, "json allocation");
- - ]
3673 : 30 : defer_dtor<json_object*,int> metadata_d(metadata, json_object_put);
3674 [ + - ]: 30 : json_object *metadata_arr = json_object_new_array();
3675 [ - + - - : 30 : if (!metadata_arr) throw libc_exception(ENOMEM, "json allocation");
- - ]
3676 [ + - ]: 30 : json_object_object_add(metadata, "results", metadata_arr);
3677 : : // consume all the rows
3678 : :
3679 : 48 : bool metadata_complete = true;
3680 : 48 : while (1)
3681 : : {
3682 [ + - ]: 48 : rc = pp->step_timeout();
3683 [ + + ]: 48 : if (rc == SQLITE_DONE) // success
3684 : : break;
3685 [ + - ]: 18 : if (rc == SQLITE_ABORT || rc == SQLITE_INTERRUPT) // interrupted such as by timeout
3686 : : {
3687 : : metadata_complete = false;
3688 : : break;
3689 : : }
3690 [ - + ]: 18 : if (rc != SQLITE_ROW) // error
3691 [ # # # # ]: 0 : throw sqlite_exception(rc, "step");
3692 : :
3693 [ + - ]: 18 : int m_executable_p = sqlite3_column_int (*pp, 0);
3694 [ + - ]: 18 : int m_debuginfo_p = sqlite3_column_int (*pp, 1);
3695 [ + - ]: 18 : int m_source_p = sqlite3_column_int (*pp, 2);
3696 [ + - - + : 18 : string m_buildid = (const char*) sqlite3_column_text (*pp, 3) ?: ""; // should always be non-null
+ - ]
3697 [ + - - + : 18 : string m_file = (const char*) sqlite3_column_text (*pp, 4) ?: "";
+ - ]
3698 [ + - - + : 18 : string m_archive = (const char*) sqlite3_column_text (*pp, 5) ?: "";
+ - ]
3699 : :
3700 : : // Confirm that m_file matches in the fnmatch(FNM_PATHNAME)
3701 : : // sense, since sqlite's GLOB operator is a looser filter.
3702 [ + - + - : 18 : if (key == "glob" && fnmatch(value.c_str(), m_file.c_str(), FNM_PATHNAME) != 0)
- + ]
3703 : 0 : continue;
3704 : :
3705 : 36 : auto add_metadata = [metadata_arr, m_buildid, m_file, m_archive](const string& type) {
3706 : 18 : json_object* entry = json_object_new_object();
3707 [ - + - - : 18 : if (NULL == entry) throw libc_exception (ENOMEM, "cannot allocate json");
- - ]
3708 : 18 : defer_dtor<json_object*,int> entry_d(entry, json_object_put);
3709 : :
3710 : 90 : auto add_entry_metadata = [entry](const char* k, string v) {
3711 : 72 : json_object* s;
3712 [ + - ]: 72 : if(v != "") {
3713 : 72 : s = json_object_new_string(v.c_str());
3714 [ - + - - : 72 : if (NULL == s) throw libc_exception (ENOMEM, "cannot allocate json");
- - ]
3715 : 72 : json_object_object_add(entry, k, s);
3716 : : }
3717 : 72 : };
3718 : :
3719 [ + - + - ]: 18 : add_entry_metadata("type", type.c_str());
3720 [ + - + - ]: 18 : add_entry_metadata("buildid", m_buildid);
3721 [ + - + - ]: 18 : add_entry_metadata("file", m_file);
3722 [ + - + - : 36 : if (m_archive != "") add_entry_metadata("archive", m_archive);
+ - ]
3723 [ - + ]: 18 : if (verbose > 3)
3724 [ # # ]: 0 : obatched(clog) << "metadata found local "
3725 : : << json_object_to_json_string_ext(entry,
3726 [ # # # # : 0 : JSON_C_TO_STRING_PRETTY)
# # ]
3727 : 0 : << endl;
3728 : :
3729 : : // Increase ref count to switch its ownership
3730 [ + - + - ]: 18 : json_object_array_add(metadata_arr, json_object_get(entry));
3731 [ + - + - : 36 : };
+ - ]
3732 : :
3733 [ + - + - : 36 : if (m_executable_p) add_metadata("executable");
+ - ]
3734 [ - + - - : 18 : if (m_debuginfo_p) add_metadata("debuginfo");
- - ]
3735 [ - + - - : 18 : if (m_source_p) add_metadata("source");
- - ]
3736 : 18 : }
3737 [ + - ]: 30 : pp->reset();
3738 : :
3739 [ + - ]: 30 : unsigned num_local_results = json_object_array_length(metadata_arr);
3740 : :
3741 : : // Query upstream as well
3742 [ + - ]: 30 : debuginfod_client *client = debuginfod_pool_begin();
3743 [ + - ]: 30 : if (client != NULL)
3744 : : {
3745 [ + - ]: 30 : add_client_federation_headers(client, conn);
3746 : :
3747 : 30 : int upstream_metadata_fd;
3748 : 30 : char *upstream_metadata_file = NULL;
3749 [ + - ]: 30 : upstream_metadata_fd = debuginfod_find_metadata(client, key.c_str(), (char*)value.c_str(),
3750 : : &upstream_metadata_file);
3751 [ + + ]: 30 : if (upstream_metadata_fd >= 0) {
3752 : : /* json-c >= 0.13 has json_object_from_fd(). */
3753 [ + - ]: 18 : json_object *upstream_metadata_json = json_object_from_file(upstream_metadata_file);
3754 : 18 : free (upstream_metadata_file);
3755 : 18 : json_object *upstream_metadata_json_arr;
3756 : 18 : json_object *upstream_complete;
3757 [ - + ]: 18 : if (NULL != upstream_metadata_json &&
3758 [ + - + - : 36 : json_object_object_get_ex(upstream_metadata_json, "results", &upstream_metadata_json_arr) &&
- + ]
3759 [ + - ]: 18 : json_object_object_get_ex(upstream_metadata_json, "complete", &upstream_complete))
3760 : : {
3761 [ + - ]: 18 : metadata_complete &= json_object_get_boolean(upstream_complete);
3762 [ + - + + ]: 22 : for (int i = 0, n = json_object_array_length(upstream_metadata_json_arr); i < n; i++)
3763 : : {
3764 [ + - ]: 4 : json_object *entry = json_object_array_get_idx(upstream_metadata_json_arr, i);
3765 [ - + ]: 4 : if (verbose > 3)
3766 [ # # ]: 0 : obatched(clog) << "metadata found remote "
3767 : : << json_object_to_json_string_ext(entry,
3768 [ # # # # : 0 : JSON_C_TO_STRING_PRETTY)
# # ]
3769 : 0 : << endl;
3770 : :
3771 [ + - ]: 4 : json_object_get(entry); // increment reference count
3772 [ + - ]: 4 : json_object_array_add(metadata_arr, entry);
3773 : : }
3774 [ + - ]: 18 : json_object_put(upstream_metadata_json);
3775 : : }
3776 [ + - ]: 18 : close(upstream_metadata_fd);
3777 : : }
3778 [ + - ]: 30 : debuginfod_pool_end (client);
3779 : : }
3780 : :
3781 [ + - ]: 30 : unsigned num_total_results = json_object_array_length(metadata_arr);
3782 : :
3783 [ + - ]: 30 : if (verbose > 2)
3784 [ + - + - ]: 90 : obatched(clog) << "metadata found local=" << num_local_results
3785 [ + - + - ]: 30 : << " remote=" << (num_total_results-num_local_results)
3786 [ + - + - : 30 : << " total=" << num_total_results
+ - ]
3787 : 30 : << endl;
3788 : :
3789 [ + - + - ]: 30 : json_object_object_add(metadata, "complete", json_object_new_boolean(metadata_complete));
3790 [ + - ]: 30 : const char* metadata_str = json_object_to_json_string(metadata);
3791 [ - + ]: 30 : if (!metadata_str)
3792 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate json");
3793 [ + - ]: 30 : r = MHD_create_response_from_buffer (strlen(metadata_str),
3794 : : (void*) metadata_str,
3795 : : MHD_RESPMEM_MUST_COPY);
3796 : 30 : *size = strlen(metadata_str);
3797 [ + - ]: 30 : if (r)
3798 [ + - ]: 30 : add_mhd_response_header(r, "Content-Type", "application/json");
3799 : 30 : return r;
3800 : 30 : }
3801 : :
3802 : :
3803 : : static struct MHD_Response*
3804 : 4 : handle_root (off_t* size)
3805 : : {
3806 : 4 : MHD_Response* r;
3807 [ + - ]: 4 : if (cust_homepage_file != "")
3808 : 4 : try
3809 : : {
3810 [ + - ]: 4 : int fd = open (cust_homepage_file.c_str(), O_RDONLY);
3811 [ + + ]: 4 : if (fd != -1) {
3812 : 2 : struct stat buf;
3813 : 2 : stat (cust_homepage_file.c_str(), &buf);
3814 [ + - ]: 2 : r = MHD_create_response_from_fd(buf.st_size, fd);
3815 : : // NB: MHD owns and handles the fd from now. Must not close()!
3816 [ + - ]: 2 : if (r != NULL)
3817 : : {
3818 : 2 : *size = buf.st_size;
3819 [ + - ]: 2 : add_mhd_response_header (r, "Content-Type", "text/html");
3820 : : }
3821 : : } else {
3822 [ + - + - ]: 4 : throw libc_exception (errno, "cannot open file " + cust_homepage_file);
3823 : : }
3824 : 2 : return r;
3825 : : }
3826 [ - + ]: 2 : catch (const reportable_exception& e)
3827 : : {
3828 [ + - ]: 2 : e.report(clog);
3829 : 2 : }
3830 : :
3831 [ + - + - ]: 4 : static string version = "debuginfod (" + string (PACKAGE_NAME) + ") "
3832 [ + - + - : 66 : + string (PACKAGE_VERSION);
+ - + - ]
3833 : 2 : r = MHD_create_response_from_buffer (version.size (),
3834 : 2 : (void *) version.c_str (),
3835 : : MHD_RESPMEM_PERSISTENT);
3836 [ + - ]: 2 : if (r != NULL)
3837 : : {
3838 : 2 : *size = version.size ();
3839 : 2 : add_mhd_response_header (r, "Content-Type", "text/plain");
3840 : : }
3841 : : return r;
3842 : : }
3843 : :
3844 : :
3845 : : static struct MHD_Response*
3846 : 2 : handle_options (off_t* size)
3847 : : {
3848 : 2 : static char empty_body[] = " ";
3849 : 2 : MHD_Response* r = MHD_create_response_from_buffer (1, empty_body,
3850 : : MHD_RESPMEM_PERSISTENT);
3851 [ + - ]: 2 : if (r != NULL)
3852 : : {
3853 : 2 : *size = 1;
3854 : 2 : add_mhd_response_header (r, "Access-Control-Allow-Origin", "*");
3855 : 2 : add_mhd_response_header (r, "Access-Control-Allow-Methods", "GET, OPTIONS");
3856 : 2 : add_mhd_response_header (r, "Access-Control-Allow-Headers", "cache-control");
3857 : : }
3858 : 2 : return r;
3859 : : }
3860 : :
3861 : :
3862 : : ////////////////////////////////////////////////////////////////////////
3863 : :
3864 : :
3865 : : /* libmicrohttpd callback */
3866 : : static MHD_RESULT
3867 : 6780 : handler_cb (void * /*cls*/,
3868 : : struct MHD_Connection *connection,
3869 : : const char *url,
3870 : : const char *method,
3871 : : const char * /*version*/,
3872 : : const char * /*upload_data*/,
3873 : : size_t * /*upload_data_size*/,
3874 : : void ** ptr)
3875 : : {
3876 : 6780 : struct MHD_Response *r = NULL;
3877 : 6780 : string url_copy = url;
3878 : :
3879 : : /* libmicrohttpd always makes (at least) two callbacks: once just
3880 : : past the headers, and one after the request body is finished
3881 : : being received. If we process things early (first callback) and
3882 : : queue a response, libmicrohttpd would suppress http keep-alive
3883 : : (via connection->read_closed = true). */
3884 : 6780 : static int aptr; /* just some random object to use as a flag */
3885 [ + + ]: 6780 : if (&aptr != *ptr)
3886 : : {
3887 : : /* do never respond on first call */
3888 : 3390 : *ptr = &aptr;
3889 : 3390 : return MHD_YES;
3890 : : }
3891 : 3390 : *ptr = NULL; /* reset when done */
3892 : :
3893 [ + - ]: 3390 : const char *maxsize_string = MHD_lookup_connection_value(connection, MHD_HEADER_KIND, "X-DEBUGINFOD-MAXSIZE");
3894 : 3390 : long maxsize = 0;
3895 [ + + + - ]: 3390 : if (maxsize_string != NULL && maxsize_string[0] != '\0')
3896 : 2 : maxsize = atol(maxsize_string);
3897 : : else
3898 : : maxsize = 0;
3899 : :
3900 : : #if MHD_VERSION >= 0x00097002
3901 : 3390 : enum MHD_Result rc;
3902 : : #else
3903 : : int rc = MHD_NO; // mhd
3904 : : #endif
3905 : 3390 : int http_code = 500;
3906 : 3390 : off_t http_size = -1;
3907 : 3390 : struct timespec ts_start, ts_end;
3908 : 3390 : clock_gettime (CLOCK_MONOTONIC, &ts_start);
3909 : 3390 : double afteryou = 0.0;
3910 [ + + ]: 3390 : string artifacttype, suffix;
3911 : 3390 : string urlargs; // for logging
3912 : :
3913 : 3390 : try
3914 : : {
3915 [ + + + - : 3390 : if (webapi_cors && method == string("OPTIONS"))
+ + ]
3916 : : {
3917 [ + - + - : 4 : inc_metric("http_requests_total", "type", method);
+ - + - ]
3918 [ + - ]: 2 : r = handle_options(& http_size);
3919 [ + - ]: 2 : rc = MHD_queue_response (connection, MHD_HTTP_OK, r);
3920 : 2 : http_code = MHD_HTTP_OK;
3921 [ + - ]: 2 : MHD_destroy_response (r);
3922 : 2 : return rc;
3923 : : }
3924 [ + - - + ]: 4021 : else if (string(method) != "GET")
3925 [ # # ]: 0 : throw reportable_exception(400, "we support OPTIONS+GET only");
3926 : :
3927 : : /* Start decoding the URL. */
3928 : 3388 : size_t slash1 = url_copy.find('/', 1);
3929 [ + - ]: 3388 : string url1 = url_copy.substr(0, slash1); // ok even if slash1 not found
3930 : :
3931 [ + + - + ]: 3388 : if (slash1 != string::npos && url1 == "/buildid")
3932 : : {
3933 : : // PR27863: block this thread awhile if another thread is already busy
3934 : : // fetching the exact same thing. This is better for Everyone.
3935 : : // The latecomer says "... after you!" and waits.
3936 [ + - + - : 5819 : add_metric ("thread_busy", "role", "http-buildid-after-you", 1);
+ - + - ]
3937 : : #ifdef HAVE_PTHREAD_SETNAME_NP
3938 : 2597 : (void) pthread_setname_np (pthread_self(), "mhd-buildid-after-you");
3939 : : #endif
3940 : 2597 : struct timespec tsay_start, tsay_end;
3941 : 2597 : clock_gettime (CLOCK_MONOTONIC, &tsay_start);
3942 [ + + + - ]: 2657 : static unique_set<string> busy_urls;
3943 [ + - ]: 2597 : unique_set_reserver<string> after_you(busy_urls, url_copy);
3944 : 2597 : clock_gettime (CLOCK_MONOTONIC, &tsay_end);
3945 : 2597 : afteryou = (tsay_end.tv_sec - tsay_start.tv_sec) + (tsay_end.tv_nsec - tsay_start.tv_nsec)/1.e9;
3946 [ + - + - : 5194 : add_metric ("thread_busy", "role", "http-buildid-after-you", -1);
+ - + - ]
3947 : :
3948 [ + - + - : 5194 : tmp_inc_metric m ("thread_busy", "role", "http-buildid");
+ - + - ]
3949 : : #ifdef HAVE_PTHREAD_SETNAME_NP
3950 : 2597 : (void) pthread_setname_np (pthread_self(), "mhd-buildid");
3951 : : #endif
3952 : 2597 : size_t slash2 = url_copy.find('/', slash1+1);
3953 [ - + ]: 2597 : if (slash2 == string::npos)
3954 [ # # ]: 0 : throw reportable_exception("/buildid/ webapi error, need buildid");
3955 : :
3956 [ + - ]: 2597 : string buildid = url_copy.substr(slash1+1, slash2-slash1-1);
3957 : :
3958 : 2597 : size_t slash3 = url_copy.find('/', slash2+1);
3959 : :
3960 [ + + ]: 2597 : if (slash3 == string::npos)
3961 : : {
3962 [ + - ]: 1463 : artifacttype = url_copy.substr(slash2+1);
3963 [ + - ]: 1463 : suffix = "";
3964 : : }
3965 : : else
3966 : : {
3967 [ + - ]: 1134 : artifacttype = url_copy.substr(slash2+1, slash3-slash2-1);
3968 [ + - ]: 1759 : suffix = url_copy.substr(slash3); // include the slash in the suffix
3969 : : }
3970 : :
3971 : : // get the resulting fd so we can report its size
3972 : 2597 : int fd;
3973 [ + + ]: 2597 : r = handle_buildid (connection, buildid, artifacttype, suffix, &fd);
3974 : 1972 : if (r)
3975 : : {
3976 : 1972 : struct stat fs;
3977 [ + - ]: 1972 : if (fstat(fd, &fs) == 0)
3978 : 1972 : http_size = fs.st_size;
3979 : : // libmicrohttpd will close (fd);
3980 : : }
3981 : 3222 : }
3982 [ + + ]: 791 : else if (url1 == "/metrics")
3983 : : {
3984 [ + - + - : 1502 : tmp_inc_metric m ("thread_busy", "role", "http-metrics");
+ - + - ]
3985 [ + - ]: 751 : artifacttype = "metrics";
3986 [ + - + - : 1502 : inc_metric("http_requests_total", "type", artifacttype);
+ - ]
3987 [ + - ]: 751 : r = handle_metrics(& http_size);
3988 : 751 : }
3989 [ + + ]: 40 : else if (url1 == "/metadata")
3990 : : {
3991 [ + - + - : 60 : tmp_inc_metric m ("thread_busy", "role", "http-metadata");
+ - + - ]
3992 [ + - ]: 30 : const char* key = MHD_lookup_connection_value(connection, MHD_GET_ARGUMENT_KIND, "key");
3993 [ + - ]: 30 : const char* value = MHD_lookup_connection_value(connection, MHD_GET_ARGUMENT_KIND, "value");
3994 [ - + ]: 30 : if (NULL == value || NULL == key)
3995 [ # # ]: 0 : throw reportable_exception("/metadata webapi error, need key and value");
3996 : :
3997 [ + - + - : 30 : urlargs = string("?key=") + string(key) + string("&value=") + string(value); // apprx., for logging
+ - + - +
- + - +
- ]
3998 [ + - ]: 30 : artifacttype = "metadata";
3999 [ + - + - : 60 : inc_metric("http_requests_total", "type", artifacttype);
+ - ]
4000 [ + - + - : 30 : r = handle_metadata(connection, key, value, &http_size);
+ - ]
4001 : 30 : }
4002 [ + + ]: 10 : else if (url1 == "/")
4003 : : {
4004 [ + - ]: 4 : artifacttype = "/";
4005 [ + - + - : 641 : inc_metric("http_requests_total", "type", artifacttype);
+ - ]
4006 [ + - ]: 4 : r = handle_root(& http_size);
4007 : : }
4008 : : else
4009 [ + - + - ]: 12 : throw reportable_exception("webapi error, unrecognized '" + url1 + "'");
4010 : :
4011 [ - + ]: 2757 : if (r == 0)
4012 [ # # ]: 0 : throw reportable_exception("internal error, missing response");
4013 : :
4014 [ + + + - ]: 2757 : if (maxsize > 0 && http_size > maxsize)
4015 : : {
4016 [ + - ]: 2 : MHD_destroy_response(r);
4017 [ + - + - ]: 4 : throw reportable_exception(406, "File too large, max size=" + std::to_string(maxsize));
4018 : : }
4019 : :
4020 [ + + ]: 2755 : if (webapi_cors)
4021 : : // add ACAO header for all successful requests
4022 [ + - ]: 132 : add_mhd_response_header (r, "Access-Control-Allow-Origin", "*");
4023 [ + + + + ]: 2755 : if ((cust_homepage_redirect) != "" && (url1 == "/"))
4024 : : {
4025 : : // redirect to given custom --homepage
4026 [ + - ]: 4 : MHD_add_response_header(r, "Location", cust_homepage_redirect.c_str());
4027 [ + - ]: 4 : rc = MHD_queue_response (connection, MHD_HTTP_FOUND, r);
4028 : : http_code = MHD_HTTP_FOUND;
4029 : : }
4030 : : else
4031 : : {
4032 [ + - ]: 2751 : rc = MHD_queue_response (connection, MHD_HTTP_OK, r);
4033 : : http_code = MHD_HTTP_OK;
4034 : : }
4035 [ + - ]: 2755 : MHD_destroy_response (r);
4036 : 3388 : }
4037 [ - + ]: 633 : catch (const reportable_exception& e)
4038 : : {
4039 [ + - + - : 1266 : inc_metric("http_responses_total","result","error");
+ - + - ]
4040 [ + - ]: 633 : e.report(clog);
4041 : 633 : http_code = e.code;
4042 [ + - ]: 633 : http_size = e.message.size();
4043 [ + - ]: 633 : rc = e.mhd_send_response (connection);
4044 : 633 : }
4045 : :
4046 : 3388 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
4047 : 3388 : double deltas = (ts_end.tv_sec - ts_start.tv_sec) + (ts_end.tv_nsec - ts_start.tv_nsec)/1.e9;
4048 : : // afteryou: delay waiting for other client's identical query to complete
4049 : : // deltas: total latency, including afteryou waiting
4050 [ + - + - ]: 6776 : obatched(clog) << conninfo(connection)
4051 : : << ' ' << method << ' ' << url << urlargs
4052 [ + - + - : 3388 : << ' ' << http_code << ' ' << http_size
+ - + - +
- + - + -
+ - ]
4053 [ + - + - : 3388 : << ' ' << (int)(afteryou*1000) << '+' << (int)((deltas-afteryou)*1000) << "ms"
+ - + - +
- + - +
- ]
4054 : 3388 : << endl;
4055 : :
4056 : : // related prometheus metrics
4057 : 3388 : string http_code_str = to_string(http_code);
4058 [ + - + - : 6776 : add_metric("http_responses_transfer_bytes_sum",
+ - + - ]
4059 : : "code", http_code_str, "type", artifacttype, http_size);
4060 [ + - + - : 6776 : inc_metric("http_responses_transfer_bytes_count",
+ - + - ]
4061 : : "code", http_code_str, "type", artifacttype);
4062 : :
4063 [ + - + - : 6776 : add_metric("http_responses_duration_milliseconds_sum",
+ - + - ]
4064 : : "code", http_code_str, "type", artifacttype, deltas*1000); // prometheus prefers _seconds and floating point
4065 [ + - + - : 6776 : inc_metric("http_responses_duration_milliseconds_count",
+ - + - ]
4066 : : "code", http_code_str, "type", artifacttype);
4067 : :
4068 [ + - + - : 6776 : add_metric("http_responses_after_you_milliseconds_sum",
+ - + - ]
4069 : : "code", http_code_str, "type", artifacttype, afteryou*1000);
4070 [ + - + - : 6776 : inc_metric("http_responses_after_you_milliseconds_count",
+ - + - ]
4071 : : "code", http_code_str, "type", artifacttype);
4072 : :
4073 : 3388 : return rc;
4074 : 13558 : }
4075 : :
4076 : :
4077 : : ////////////////////////////////////////////////////////////////////////
4078 : : // borrowed originally from src/nm.c get_local_names()
4079 : :
4080 : : static void
4081 : 412 : dwarf_extract_source_paths (Elf *elf, set<string>& debug_sourcefiles)
4082 : : noexcept // no exceptions - so we can simplify the altdbg resource release at end
4083 : : {
4084 : 412 : Dwarf* dbg = dwarf_begin_elf (elf, DWARF_C_READ, NULL);
4085 [ - + ]: 412 : if (dbg == NULL)
4086 : 0 : return;
4087 : :
4088 : 412 : Dwarf* altdbg = NULL;
4089 : 412 : int altdbg_fd = -1;
4090 : :
4091 : : // DWZ handling: if we have an unsatisfied debug-alt-link, add an
4092 : : // empty string into the outgoing sourcefiles set, so the caller
4093 : : // should know that our data is incomplete.
4094 : 412 : const char *alt_name_p;
4095 : 412 : const void *alt_build_id; // elfutils-owned memory
4096 : 412 : ssize_t sz = dwelf_dwarf_gnu_debugaltlink (dbg, &alt_name_p, &alt_build_id);
4097 [ + + ]: 412 : if (sz > 0) // got one!
4098 : : {
4099 : 200 : string buildid;
4100 : 200 : unsigned char* build_id_bytes = (unsigned char*) alt_build_id;
4101 [ + + ]: 4200 : for (ssize_t idx=0; idx<sz; idx++)
4102 : : {
4103 : 4000 : buildid += "0123456789abcdef"[build_id_bytes[idx] >> 4];
4104 : 4000 : buildid += "0123456789abcdef"[build_id_bytes[idx] & 0xf];
4105 : : }
4106 : :
4107 [ + + ]: 200 : if (verbose > 3)
4108 : 156 : obatched(clog) << "Need altdebug buildid=" << buildid << endl;
4109 : :
4110 : : // but is it unsatisfied the normal elfutils ways?
4111 : 200 : Dwarf* alt = dwarf_getalt (dbg);
4112 [ + - ]: 200 : if (alt == NULL)
4113 : : {
4114 : : // Yup, unsatisfied the normal way. Maybe we can satisfy it
4115 : : // from our own debuginfod database.
4116 : 200 : int alt_fd;
4117 : 200 : struct MHD_Response *r = 0;
4118 : 200 : try
4119 : : {
4120 [ + - ]: 200 : string artifacttype = "debuginfo";
4121 [ + - + + ]: 220 : r = handle_buildid (0, buildid, artifacttype, "", &alt_fd);
4122 : : // NB: no need for ACAO etc. headers; this is not getting sent to a client
4123 : 20 : }
4124 [ - + ]: 20 : catch (const reportable_exception& e)
4125 : : {
4126 : : // swallow exceptions
4127 : 20 : }
4128 : :
4129 : : // NB: this is not actually recursive! This invokes the web-query
4130 : : // path, which cannot get back into the scan code paths.
4131 : 200 : if (r)
4132 : : {
4133 : : // Found it!
4134 : 180 : altdbg_fd = dup(alt_fd); // ok if this fails, downstream failures ok
4135 : 180 : alt = altdbg = dwarf_begin (altdbg_fd, DWARF_C_READ);
4136 : : // NB: must close this dwarf and this fd at the bottom of the function!
4137 : 180 : MHD_destroy_response (r); // will close alt_fd
4138 [ - + ]: 180 : if (alt)
4139 : 180 : dwarf_setalt (dbg, alt);
4140 : : }
4141 : : }
4142 : : else
4143 : : {
4144 : : // NB: dwarf_setalt(alt) inappropriate - already done!
4145 : : // NB: altdbg will stay 0 so nothing tries to redundantly dealloc.
4146 : : }
4147 : :
4148 [ + + ]: 200 : if (alt)
4149 : : {
4150 [ + + ]: 180 : if (verbose > 3)
4151 : 156 : obatched(clog) << "Resolved altdebug buildid=" << buildid << endl;
4152 : : }
4153 : : else // (alt == NULL) - signal possible presence of poor debuginfo
4154 : : {
4155 : 20 : debug_sourcefiles.insert("");
4156 [ + - ]: 20 : if (verbose > 3)
4157 : 0 : obatched(clog) << "Unresolved altdebug buildid=" << buildid << endl;
4158 : : }
4159 : 200 : }
4160 : :
4161 : 412 : Dwarf_Off offset = 0;
4162 : 412 : Dwarf_Off old_offset;
4163 : 412 : size_t hsize;
4164 : :
4165 [ + + ]: 9435 : while (dwarf_nextcu (dbg, old_offset = offset, &offset, &hsize, NULL, NULL, NULL) == 0)
4166 : : {
4167 : 9024 : Dwarf_Die cudie_mem;
4168 : 9024 : Dwarf_Die *cudie = dwarf_offdie (dbg, old_offset + hsize, &cudie_mem);
4169 : :
4170 [ - + ]: 9024 : if (cudie == NULL)
4171 : 36 : continue;
4172 [ + + ]: 9024 : if (dwarf_tag (cudie) != DW_TAG_compile_unit)
4173 : 36 : continue;
4174 : :
4175 [ - + ]: 8988 : const char *cuname = dwarf_diename(cudie) ?: "unknown";
4176 : :
4177 : 8988 : Dwarf_Files *files;
4178 : 8988 : size_t nfiles;
4179 [ - + ]: 8988 : if (dwarf_getsrcfiles (cudie, &files, &nfiles) != 0)
4180 : 0 : continue;
4181 : :
4182 : : // extract DW_AT_comp_dir to resolve relative file names
4183 : 8988 : const char *comp_dir = "";
4184 : 8988 : const char *const *dirs;
4185 : 8988 : size_t ndirs;
4186 [ - + ]: 8988 : if (dwarf_getsrcdirs (files, &dirs, &ndirs) == 0 &&
4187 [ - + ]: 8988 : dirs[0] != NULL)
4188 : : comp_dir = dirs[0];
4189 : : if (comp_dir == NULL)
4190 : : comp_dir = "";
4191 : :
4192 [ + + ]: 8988 : if (verbose > 3)
4193 : 13812 : obatched(clog) << "searching for sources for cu=" << cuname << " comp_dir=" << comp_dir
4194 : 6906 : << " #files=" << nfiles << " #dirs=" << ndirs << endl;
4195 : :
4196 [ - + - - ]: 8988 : if (comp_dir[0] == '\0' && cuname[0] != '/')
4197 : : {
4198 [ # # ]: 0 : if (verbose > 3)
4199 : 0 : obatched(clog) << "skipping cu=" << cuname << " due to empty comp_dir" << endl;
4200 : 0 : continue;
4201 : : }
4202 : :
4203 [ + + ]: 147228 : for (size_t f = 1; f < nfiles; f++)
4204 : : {
4205 : 138241 : const char *hat = dwarf_filesrc (files, f, NULL, NULL);
4206 [ - + ]: 138278 : if (hat == NULL)
4207 : 1690 : continue;
4208 : :
4209 [ + + ]: 138261 : if (string(hat) == "<built-in>"
4210 [ + + + + ]: 276530 : || string_endswith(hat, "<built-in>")) // gcc intrinsics, don't bother record
4211 : 1690 : continue;
4212 : :
4213 [ + + ]: 136571 : string waldo;
4214 [ + + ]: 136571 : if (hat[0] == '/') // absolute
4215 : 85814 : waldo = (string (hat));
4216 [ + - ]: 50757 : else if (comp_dir[0] != '\0') // comp_dir relative
4217 : 50757 : waldo = (string (comp_dir) + string("/") + string (hat));
4218 : : else
4219 : : {
4220 [ # # ]: 0 : if (verbose > 3)
4221 : 0 : obatched(clog) << "skipping hat=" << hat << " due to empty comp_dir" << endl;
4222 : 0 : continue;
4223 : : }
4224 : :
4225 : : // NB: this is the 'waldo' that a dbginfo client will have
4226 : : // to supply for us to give them the file The comp_dir
4227 : : // prefixing is a definite complication. Otherwise we'd
4228 : : // have to return a setof comp_dirs (one per CU!) with
4229 : : // corresponding filesrc[] names, instead of one absolute
4230 : : // resoved set. Maybe we'll have to do that anyway. XXX
4231 : :
4232 [ + + ]: 136585 : if (verbose > 4)
4233 [ - + ]: 32 : obatched(clog) << waldo
4234 [ - + ]: 16 : << (debug_sourcefiles.find(waldo)==debug_sourcefiles.end() ? " new" : " dup") << endl;
4235 : :
4236 : 136585 : debug_sourcefiles.insert (waldo);
4237 : 136550 : }
4238 : : }
4239 : :
4240 : 412 : dwarf_end(dbg);
4241 [ + + ]: 412 : if (altdbg)
4242 : 180 : dwarf_end(altdbg);
4243 [ + + ]: 412 : if (altdbg_fd >= 0)
4244 : 180 : close(altdbg_fd);
4245 : : }
4246 : :
4247 : :
4248 : :
4249 : : static void
4250 : 1708 : elf_classify (int fd, bool &executable_p, bool &debuginfo_p, string &buildid, set<string>& debug_sourcefiles)
4251 : : {
4252 : 1708 : Elf *elf = elf_begin (fd, ELF_C_READ_MMAP_PRIVATE, NULL);
4253 [ + - ]: 1708 : if (elf == NULL)
4254 : : return;
4255 : :
4256 : 1708 : try // catch our types of errors and clean up the Elf* object
4257 : : {
4258 [ + - + + ]: 1708 : if (elf_kind (elf) != ELF_K_ELF)
4259 : : {
4260 [ + - ]: 906 : elf_end (elf);
4261 : 946 : return;
4262 : : }
4263 : :
4264 : 802 : GElf_Ehdr ehdr_storage;
4265 [ + - ]: 802 : GElf_Ehdr *ehdr = gelf_getehdr (elf, &ehdr_storage);
4266 [ - + ]: 802 : if (ehdr == NULL)
4267 : : {
4268 [ # # ]: 0 : elf_end (elf);
4269 : : return;
4270 : : }
4271 : 802 : auto elf_type = ehdr->e_type;
4272 : :
4273 : 802 : const void *build_id; // elfutils-owned memory
4274 [ + - ]: 802 : ssize_t sz = dwelf_elf_gnu_build_id (elf, & build_id);
4275 [ + + ]: 802 : if (sz <= 0)
4276 : : {
4277 : : // It's not a diagnostic-worthy error for an elf file to lack build-id.
4278 : : // It might just be very old.
4279 [ + - ]: 40 : elf_end (elf);
4280 : : return;
4281 : : }
4282 : :
4283 : : // build_id is a raw byte array; convert to hexadecimal *lowercase*
4284 : 762 : unsigned char* build_id_bytes = (unsigned char*) build_id;
4285 [ + + ]: 15997 : for (ssize_t idx=0; idx<sz; idx++)
4286 : : {
4287 [ + - ]: 15235 : buildid += "0123456789abcdef"[build_id_bytes[idx] >> 4];
4288 [ + - ]: 30472 : buildid += "0123456789abcdef"[build_id_bytes[idx] & 0xf];
4289 : : }
4290 : :
4291 : : // now decide whether it's an executable - namely, any allocatable section has
4292 : : // PROGBITS;
4293 [ + + ]: 762 : if (elf_type == ET_EXEC || elf_type == ET_DYN)
4294 : : {
4295 : 700 : size_t shnum;
4296 [ + - ]: 700 : int rc = elf_getshdrnum (elf, &shnum);
4297 [ - + ]: 700 : if (rc < 0)
4298 [ # # # # ]: 0 : throw elfutils_exception(rc, "getshdrnum");
4299 : :
4300 : 700 : executable_p = false;
4301 [ + + ]: 13340 : for (size_t sc = 0; sc < shnum; sc++)
4302 : : {
4303 [ + - ]: 13014 : Elf_Scn *scn = elf_getscn (elf, sc);
4304 [ - + ]: 13014 : if (scn == NULL)
4305 : 0 : continue;
4306 : :
4307 : 13014 : GElf_Shdr shdr_mem;
4308 [ + - ]: 13014 : GElf_Shdr *shdr = gelf_getshdr (scn, &shdr_mem);
4309 [ - + ]: 13014 : if (shdr == NULL)
4310 : 0 : continue;
4311 : :
4312 : : // allocated (loadable / vm-addr-assigned) section with available content?
4313 [ + + + + ]: 13014 : if ((shdr->sh_type == SHT_PROGBITS) && (shdr->sh_flags & SHF_ALLOC))
4314 : : {
4315 [ - + ]: 374 : if (verbose > 4)
4316 [ # # # # : 0 : obatched(clog) << "executable due to SHF_ALLOC SHT_PROGBITS sc=" << sc << endl;
# # ]
4317 : 374 : executable_p = true;
4318 : 374 : break; // no need to keep looking for others
4319 : : }
4320 : : } // iterate over sections
4321 : : } // executable_p classification
4322 : :
4323 : : // now decide whether it's a debuginfo - namely, if it has any .debug* or .zdebug* sections
4324 : : // logic mostly stolen from fweimer@redhat.com's elfclassify drafts
4325 : 762 : size_t shstrndx;
4326 [ + - ]: 762 : int rc = elf_getshdrstrndx (elf, &shstrndx);
4327 [ - + ]: 762 : if (rc < 0)
4328 [ # # # # ]: 0 : throw elfutils_exception(rc, "getshdrstrndx");
4329 : :
4330 : : Elf_Scn *scn = NULL;
4331 : : bool symtab_p = false;
4332 : : bool bits_alloc_p = false;
4333 : 40886 : while (true)
4334 : : {
4335 [ + - ]: 20824 : scn = elf_nextscn (elf, scn);
4336 [ + + ]: 20763 : if (scn == NULL)
4337 : : break;
4338 : 20413 : GElf_Shdr shdr_storage;
4339 [ + - ]: 20413 : GElf_Shdr *shdr = gelf_getshdr (scn, &shdr_storage);
4340 [ + - ]: 20410 : if (shdr == NULL)
4341 : : break;
4342 [ + - ]: 20410 : const char *section_name = elf_strptr (elf, shstrndx, shdr->sh_name);
4343 [ + - ]: 20474 : if (section_name == NULL)
4344 : : break;
4345 [ + + ]: 20474 : if (startswith (section_name, ".debug_line") ||
4346 [ - + ]: 20062 : startswith (section_name, ".zdebug_line"))
4347 : : {
4348 : 412 : debuginfo_p = true;
4349 [ + - ]: 412 : if (scan_source_info)
4350 : 412 : dwarf_extract_source_paths (elf, debug_sourcefiles);
4351 : : break; // expecting only one .*debug_line, so no need to look for others
4352 : : }
4353 [ + + ]: 20062 : else if (startswith (section_name, ".debug_") ||
4354 [ + - ]: 18765 : startswith (section_name, ".zdebug_"))
4355 : : {
4356 : 1250 : debuginfo_p = true;
4357 : : // NB: don't break; need to parse .debug_line for sources
4358 : : }
4359 [ + + ]: 18812 : else if (shdr->sh_type == SHT_SYMTAB)
4360 : : {
4361 : : symtab_p = true;
4362 : : }
4363 : 18788 : else if (shdr->sh_type != SHT_NOBITS
4364 [ + + ]: 18788 : && shdr->sh_type != SHT_NOTE
4365 [ + + ]: 9185 : && (shdr->sh_flags & SHF_ALLOC) != 0)
4366 : : {
4367 : 20062 : bits_alloc_p = true;
4368 : : }
4369 : 20062 : }
4370 : :
4371 : : // For more expansive elf/split-debuginfo classification, we
4372 : : // want to identify as debuginfo "strip -s"-produced files
4373 : : // without .debug_info* (like libicudata), but we don't want to
4374 : : // identify "strip -g" executables (with .symtab left there).
4375 [ - + ]: 762 : if (symtab_p && !bits_alloc_p)
4376 : 0 : debuginfo_p = true;
4377 : : }
4378 [ # # ]: 0 : catch (const reportable_exception& e)
4379 : : {
4380 [ # # ]: 0 : e.report(clog);
4381 : 0 : }
4382 : 762 : elf_end (elf);
4383 : : }
4384 : :
4385 : :
4386 : : // Intern the given file name in two parts (dirname & basename) and
4387 : : // return the resulting file's id.
4388 : : static int64_t
4389 : 34058 : register_file_name(sqlite_ps& ps_upsert_fileparts,
4390 : : sqlite_ps& ps_upsert_file,
4391 : : sqlite_ps& ps_lookup_file,
4392 : : const string& name)
4393 : : {
4394 : 34058 : std::size_t slash = name.rfind('/');
4395 [ + + ]: 34060 : string dirname, filename;
4396 [ + + ]: 34060 : if (slash == std::string::npos)
4397 : : {
4398 [ + - ]: 90 : dirname = "";
4399 [ + - ]: 90 : filename = name;
4400 : : }
4401 : : else
4402 : : {
4403 [ + - ]: 33970 : dirname = name.substr(0, slash);
4404 [ + - ]: 33971 : filename = name.substr(slash+1);
4405 : : }
4406 : : // NB: see also handle_metadata()
4407 : :
4408 : : // intern the two substrings
4409 : 34061 : ps_upsert_fileparts
4410 [ + - ]: 34061 : .reset()
4411 [ + - ]: 34057 : .bind(1, dirname)
4412 [ + - ]: 34049 : .step_ok_done();
4413 : 34062 : ps_upsert_fileparts
4414 [ + - ]: 34062 : .reset()
4415 [ + - ]: 34062 : .bind(1, filename)
4416 [ + - ]: 34060 : .step_ok_done();
4417 : :
4418 : : // intern the tuple
4419 : 34062 : ps_upsert_file
4420 [ + - ]: 34062 : .reset()
4421 [ + - ]: 34061 : .bind(1, dirname)
4422 [ + - ]: 34060 : .bind(2, filename)
4423 [ + - ]: 34058 : .step_ok_done();
4424 : :
4425 : : // look up the tuple's id
4426 : 34062 : ps_lookup_file
4427 [ + - ]: 34062 : .reset()
4428 [ + - ]: 34062 : .bind(1, dirname)
4429 [ + - ]: 34059 : .bind(2, filename);
4430 [ + - ]: 34061 : int rc = ps_lookup_file.step();
4431 [ - + - - : 34062 : if (rc != SQLITE_ROW) throw sqlite_exception(rc, "step");
- - ]
4432 : :
4433 [ + - ]: 34062 : int64_t id = sqlite3_column_int64 (ps_lookup_file, 0);
4434 [ + - ]: 34060 : ps_lookup_file.reset();
4435 : 34062 : return id;
4436 : 34062 : }
4437 : :
4438 : :
4439 : :
4440 : : static void
4441 : 1106 : scan_source_file (const string& rps, const stat_t& st,
4442 : : sqlite_ps& ps_upsert_buildids,
4443 : : sqlite_ps& ps_upsert_fileparts,
4444 : : sqlite_ps& ps_upsert_file,
4445 : : sqlite_ps& ps_lookup_file,
4446 : : sqlite_ps& ps_upsert_de,
4447 : : sqlite_ps& ps_upsert_s,
4448 : : sqlite_ps& ps_query,
4449 : : sqlite_ps& ps_scan_done,
4450 : : unsigned& fts_cached,
4451 : : unsigned& fts_executable,
4452 : : unsigned& fts_debuginfo,
4453 : : unsigned& fts_sourcefiles)
4454 : : {
4455 : 1106 : int64_t fileid = register_file_name(ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, rps);
4456 : :
4457 : : /* See if we know of it already. */
4458 : 1106 : int rc = ps_query
4459 : 1106 : .reset()
4460 : 1106 : .bind(1, fileid)
4461 : 1106 : .bind(2, st.st_mtime)
4462 : 1106 : .step();
4463 : 1106 : ps_query.reset();
4464 [ + + ]: 1106 : if (rc == SQLITE_ROW) // i.e., a result, as opposed to DONE (no results)
4465 : : // no need to recheck a file/version we already know
4466 : : // specifically, no need to elf-begin a file we already determined is non-elf
4467 : : // (so is stored with buildid=NULL)
4468 : : {
4469 : 438 : fts_cached++;
4470 : 438 : return;
4471 : : }
4472 : :
4473 : 668 : bool executable_p = false, debuginfo_p = false; // E and/or D
4474 [ + - ]: 668 : string buildid;
4475 [ + - ]: 668 : set<string> sourcefiles;
4476 : :
4477 [ + - ]: 668 : int fd = open (rps.c_str(), O_RDONLY);
4478 : 668 : try
4479 : : {
4480 [ + - ]: 668 : if (fd >= 0)
4481 [ + - ]: 668 : elf_classify (fd, executable_p, debuginfo_p, buildid, sourcefiles);
4482 : : else
4483 [ # # # # : 0 : throw libc_exception(errno, string("open ") + rps);
# # ]
4484 [ + - + - : 1336 : add_metric ("scanned_bytes_total","source","file",
+ - ]
4485 [ + - ]: 668 : st.st_size);
4486 [ + - + - : 1336 : inc_metric ("scanned_files_total","source","file");
+ - + - ]
4487 : : }
4488 : : // NB: we catch exceptions here too, so that we can
4489 : : // cache the corrupt-elf case (!executable_p &&
4490 : : // !debuginfo_p) just below, just as if we had an
4491 : : // EPERM error from open(2).
4492 [ - - ]: 0 : catch (const reportable_exception& e)
4493 : : {
4494 [ - - ]: 0 : e.report(clog);
4495 : 0 : }
4496 : :
4497 [ + - ]: 668 : if (fd >= 0)
4498 [ + - ]: 668 : close (fd);
4499 : :
4500 [ + + ]: 668 : if (buildid == "")
4501 : : {
4502 : : // no point storing an elf file without buildid
4503 : 572 : executable_p = false;
4504 : 572 : debuginfo_p = false;
4505 : : }
4506 : : else
4507 : : {
4508 : : // register this build-id in the interning table
4509 : 96 : ps_upsert_buildids
4510 [ + - ]: 96 : .reset()
4511 [ + - ]: 96 : .bind(1, buildid)
4512 [ + - ]: 96 : .step_ok_done();
4513 : : }
4514 : :
4515 [ + + ]: 668 : if (executable_p)
4516 : 72 : fts_executable ++;
4517 [ + + ]: 668 : if (debuginfo_p)
4518 : 72 : fts_debuginfo ++;
4519 [ + + + + ]: 668 : if (executable_p || debuginfo_p)
4520 : : {
4521 : 96 : ps_upsert_de
4522 [ + - ]: 96 : .reset()
4523 [ + - ]: 96 : .bind(1, buildid)
4524 [ + + + - ]: 120 : .bind(2, debuginfo_p ? 1 : 0)
4525 [ + + + - ]: 120 : .bind(3, executable_p ? 1 : 0)
4526 [ + - ]: 96 : .bind(4, fileid)
4527 [ + - ]: 96 : .bind(5, st.st_mtime)
4528 [ + - ]: 96 : .step_ok_done();
4529 : : }
4530 [ + + ]: 668 : if (executable_p)
4531 [ + - + - : 144 : inc_metric("found_executable_total","source","files");
+ - + - ]
4532 [ + + ]: 668 : if (debuginfo_p)
4533 [ + - + - : 144 : inc_metric("found_debuginfo_total","source","files");
+ - + - ]
4534 : :
4535 [ + + + - ]: 668 : if (sourcefiles.size() && buildid != "")
4536 : : {
4537 : 72 : fts_sourcefiles += sourcefiles.size();
4538 : :
4539 [ + + ]: 15410 : for (auto&& dwarfsrc : sourcefiles)
4540 : : {
4541 [ + - ]: 15338 : char *srp = realpath(dwarfsrc.c_str(), NULL);
4542 [ - + ]: 15338 : if (srp == NULL) // also if DWZ unresolved dwarfsrc=""
4543 : 0 : continue; // unresolvable files are not a serious problem
4544 : : // throw libc_exception(errno, "fts/file realpath " + srcpath);
4545 [ + - ]: 15338 : string srps = string(srp);
4546 : 15338 : free (srp);
4547 : :
4548 : 15338 : struct stat sfs;
4549 : 15338 : rc = stat(srps.c_str(), &sfs);
4550 [ - + ]: 15338 : if (rc != 0)
4551 : 0 : continue;
4552 : :
4553 [ + - ]: 15338 : if (verbose > 2)
4554 [ + - + - ]: 46014 : obatched(clog) << "recorded buildid=" << buildid << " file=" << srps
4555 [ + - + - : 15338 : << " mtime=" << sfs.st_mtime
+ - + - ]
4556 [ + - + - : 15338 : << " as source " << dwarfsrc << endl;
+ - ]
4557 : :
4558 : : // PR25548: store canonicalized dwarfsrc path
4559 [ + - ]: 15338 : string dwarfsrc_canon = canon_pathname (dwarfsrc);
4560 [ + + ]: 15338 : if (dwarfsrc_canon != dwarfsrc)
4561 : : {
4562 [ + + ]: 2938 : if (verbose > 3)
4563 [ + - + - : 4568 : obatched(clog) << "canonicalized src=" << dwarfsrc << " alias=" << dwarfsrc_canon << endl;
+ - + - +
- ]
4564 : : }
4565 : :
4566 [ + - ]: 15338 : int64_t fileid1 = register_file_name (ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, dwarfsrc_canon);
4567 [ + - ]: 15338 : int64_t fileid2 = register_file_name (ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, srps);
4568 : :
4569 : 15338 : ps_upsert_s
4570 [ + - ]: 15338 : .reset()
4571 [ + - ]: 15337 : .bind(1, buildid)
4572 [ + - ]: 15337 : .bind(2, fileid1)
4573 [ + - ]: 15334 : .bind(3, fileid2)
4574 [ + - ]: 15338 : .bind(4, sfs.st_mtime)
4575 [ + - ]: 15336 : .step_ok_done();
4576 : :
4577 [ + - + - : 30676 : inc_metric("found_sourcerefs_total","source","files");
+ - + - ]
4578 : 15338 : }
4579 : : }
4580 : :
4581 : 668 : ps_scan_done
4582 [ + - ]: 668 : .reset()
4583 [ + - ]: 668 : .bind(1, fileid)
4584 [ + - ]: 668 : .bind(2, st.st_mtime)
4585 [ + - ]: 668 : .bind(3, st.st_size)
4586 [ + - ]: 668 : .step_ok_done();
4587 : :
4588 [ + - ]: 668 : if (verbose > 2)
4589 [ + - + - ]: 2004 : obatched(clog) << "recorded buildid=" << buildid << " file=" << rps
4590 [ + - + - : 668 : << " mtime=" << st.st_mtime << " atype="
+ - + - ]
4591 : : << (executable_p ? "E" : "")
4592 [ + - + + : 1860 : << (debuginfo_p ? "D" : "") << endl;
+ - + + +
- + - ]
4593 : 668 : }
4594 : :
4595 : :
4596 : :
4597 : :
4598 : :
4599 : : // Analyze given archive file of given age; record buildids / exec/debuginfo-ness of its
4600 : : // constituent files with given upsert statements.
4601 : : static void
4602 : 396 : archive_classify (const string& rps, string& archive_extension, int64_t archiveid,
4603 : : sqlite_ps& ps_upsert_buildids, sqlite_ps& ps_upsert_fileparts, sqlite_ps& ps_upsert_file,
4604 : : sqlite_ps& ps_lookup_file,
4605 : : sqlite_ps& ps_upsert_de, sqlite_ps& ps_upsert_sref, sqlite_ps& ps_upsert_sdef,
4606 : : sqlite_ps& ps_upsert_seekable,
4607 : : time_t mtime,
4608 : : unsigned& fts_executable, unsigned& fts_debuginfo, unsigned& fts_sref, unsigned& fts_sdef,
4609 : : bool& fts_sref_complete_p)
4610 : : {
4611 : 396 : string archive_decoder = "/dev/null";
4612 [ + + ]: 1024 : for (auto&& arch : scan_archives)
4613 [ + + ]: 628 : if (string_endswith(rps, arch.first))
4614 : : {
4615 [ + - ]: 396 : archive_extension = arch.first;
4616 [ + - ]: 1024 : archive_decoder = arch.second;
4617 : : }
4618 : :
4619 : 396 : FILE* fp;
4620 : 396 : defer_dtor<FILE*,int>::dtor_fn dfn;
4621 [ + + ]: 396 : if (archive_decoder != "cat")
4622 : : {
4623 [ + - + - : 80 : string popen_cmd = archive_decoder + " " + shell_escape(rps);
+ - ]
4624 [ + - ]: 40 : fp = popen (popen_cmd.c_str(), "r"); // "e" O_CLOEXEC?
4625 : 40 : dfn = pclose;
4626 [ - + ]: 40 : if (fp == NULL)
4627 [ # # # # : 0 : throw libc_exception (errno, string("popen ") + popen_cmd);
# # ]
4628 : 40 : }
4629 : : else
4630 : : {
4631 [ + - ]: 356 : fp = fopen (rps.c_str(), "r");
4632 : 356 : dfn = fclose;
4633 [ - + ]: 356 : if (fp == NULL)
4634 [ # # # # : 0 : throw libc_exception (errno, string("fopen ") + rps);
# # ]
4635 : : }
4636 : 396 : defer_dtor<FILE*,int> fp_closer (fp, dfn);
4637 : :
4638 : 396 : struct archive *a;
4639 [ + - ]: 396 : a = archive_read_new();
4640 [ - + ]: 396 : if (a == NULL)
4641 [ # # # # ]: 0 : throw archive_exception("cannot create archive reader");
4642 : 396 : defer_dtor<struct archive*,int> archive_closer (a, archive_read_free);
4643 : :
4644 [ + - ]: 396 : int rc = archive_read_support_format_all(a);
4645 [ - + ]: 396 : if (rc != ARCHIVE_OK)
4646 [ # # # # ]: 0 : throw archive_exception(a, "cannot select all formats");
4647 [ + - ]: 396 : rc = archive_read_support_filter_all(a);
4648 [ - + ]: 396 : if (rc != ARCHIVE_OK)
4649 [ # # # # ]: 0 : throw archive_exception(a, "cannot select all filters");
4650 : :
4651 [ + - ]: 396 : rc = archive_read_open_FILE (a, fp);
4652 [ - + ]: 396 : if (rc != ARCHIVE_OK)
4653 : : {
4654 [ # # # # : 0 : obatched(clog) << "cannot open archive from pipe " << rps << endl;
# # ]
4655 [ # # # # ]: 0 : throw archive_exception(a, "cannot open archive from pipe");
4656 : : }
4657 : :
4658 [ + + ]: 396 : if (verbose > 3)
4659 [ + - + - : 712 : obatched(clog) << "libarchive scanning " << rps << " id " << archiveid << endl;
+ - + - +
- ]
4660 : :
4661 [ + - ]: 396 : bool seekable = is_seekable_archive (rps, a);
4662 [ + - + + ]: 396 : if (verbose> 2 && seekable)
4663 [ + - + - : 64 : obatched(clog) << rps << " is seekable" << endl;
+ - ]
4664 : :
4665 : : bool any_exceptions = false;
4666 : 3774 : while(1) // parse archive entries
4667 : : {
4668 [ + - ]: 3774 : if (interrupted)
4669 : : break;
4670 : :
4671 : 3774 : try
4672 : : {
4673 : 3774 : struct archive_entry *e;
4674 [ + - ]: 3774 : rc = archive_read_next_header (a, &e);
4675 [ + + ]: 3774 : if (rc != ARCHIVE_OK)
4676 : : break;
4677 : :
4678 [ + - + + ]: 3378 : if (! S_ISREG(archive_entry_mode (e))) // skip non-files completely
4679 : 2338 : continue;
4680 : :
4681 [ + - ]: 1040 : string fn = canonicalized_archive_entry_pathname (e);
4682 : :
4683 [ + + ]: 1040 : if (verbose > 3)
4684 [ + - + - : 1756 : obatched(clog) << "libarchive checking " << fn << endl;
+ - ]
4685 : :
4686 [ + - ]: 1040 : int64_t seekable_size = archive_entry_size (e);
4687 [ + - ]: 1040 : int64_t seekable_offset = archive_filter_bytes (a, 0);
4688 [ + - ]: 1040 : time_t seekable_mtime = archive_entry_mtime (e);
4689 : :
4690 : : // extract this file to a temporary file
4691 : 1040 : char* tmppath = NULL;
4692 : 1040 : rc = asprintf (&tmppath, "%s/debuginfod-classify.XXXXXX", tmpdir.c_str());
4693 [ - + ]: 1040 : if (rc < 0)
4694 [ # # # # ]: 0 : throw libc_exception (ENOMEM, "cannot allocate tmppath");
4695 : 1040 : defer_dtor<void*,void> tmmpath_freer (tmppath, free);
4696 [ + - ]: 1040 : int fd = mkstemp (tmppath);
4697 [ - + ]: 1040 : if (fd < 0)
4698 [ # # # # ]: 0 : throw libc_exception (errno, "cannot create temporary file");
4699 : 1040 : unlink (tmppath); // unlink now so OS will release the file as soon as we close the fd
4700 : 1040 : defer_dtor<int,int> minifd_closer (fd, close);
4701 : :
4702 [ + - ]: 1040 : rc = archive_read_data_into_fd (a, fd);
4703 [ - + ]: 1040 : if (rc != ARCHIVE_OK) {
4704 [ # # ]: 0 : close (fd);
4705 [ # # # # ]: 0 : throw archive_exception(a, "cannot extract file");
4706 : : }
4707 : :
4708 : : // finally ... time to run elf_classify on this bad boy and update the database
4709 : 1040 : bool executable_p = false, debuginfo_p = false;
4710 [ + - ]: 1040 : string buildid;
4711 [ + - ]: 1040 : set<string> sourcefiles;
4712 [ + - ]: 1040 : elf_classify (fd, executable_p, debuginfo_p, buildid, sourcefiles);
4713 : : // NB: might throw
4714 : :
4715 [ + + ]: 1040 : if (buildid != "") // intern buildid
4716 : : {
4717 : 666 : ps_upsert_buildids
4718 [ + - ]: 666 : .reset()
4719 [ + - ]: 666 : .bind(1, buildid)
4720 [ + - ]: 666 : .step_ok_done();
4721 : : }
4722 : :
4723 [ + - ]: 1040 : int64_t fileid = register_file_name (ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, fn);
4724 : :
4725 [ + + ]: 1040 : if (sourcefiles.size() > 0) // sref records needed
4726 : : {
4727 : : // NB: we intern each source file once. Once raw, as it
4728 : : // appears in the DWARF file list coming back from
4729 : : // elf_classify() - because it'll end up in the
4730 : : // _norm.artifactsrc column. We don't also put another
4731 : : // version with a '.' at the front, even though that's
4732 : : // how rpm/cpio packs names, because we hide that from
4733 : : // the database for storage efficiency.
4734 : :
4735 [ + + ]: 802 : for (auto&& s : sourcefiles)
4736 : : {
4737 [ + + ]: 498 : if (s == "")
4738 : : {
4739 : 20 : fts_sref_complete_p = false;
4740 : 20 : continue;
4741 : : }
4742 : :
4743 : : // PR25548: store canonicalized source path
4744 : 478 : const string& dwarfsrc = s;
4745 [ + - ]: 478 : string dwarfsrc_canon = canon_pathname (dwarfsrc);
4746 [ + + ]: 478 : if (dwarfsrc_canon != dwarfsrc)
4747 : : {
4748 [ + - ]: 28 : if (verbose > 3)
4749 [ + - + - : 56 : obatched(clog) << "canonicalized src=" << dwarfsrc << " alias=" << dwarfsrc_canon << endl;
+ - + - +
- ]
4750 : : }
4751 : :
4752 [ + - ]: 478 : int64_t srcfileid = register_file_name(ps_upsert_fileparts, ps_upsert_file, ps_lookup_file,
4753 : : dwarfsrc_canon);
4754 : :
4755 : 478 : ps_upsert_sref
4756 [ + - ]: 478 : .reset()
4757 [ + - ]: 478 : .bind(1, buildid)
4758 [ + - ]: 478 : .bind(2, srcfileid)
4759 [ + - ]: 478 : .step_ok_done();
4760 : :
4761 : 478 : fts_sref ++;
4762 : 478 : }
4763 : : }
4764 : :
4765 [ + + ]: 1040 : if (executable_p)
4766 : 302 : fts_executable ++;
4767 [ + + ]: 1040 : if (debuginfo_p)
4768 : 366 : fts_debuginfo ++;
4769 : :
4770 [ + + + + ]: 1040 : if (executable_p || debuginfo_p)
4771 : : {
4772 : 666 : ps_upsert_de
4773 [ + - ]: 666 : .reset()
4774 [ + - ]: 666 : .bind(1, buildid)
4775 [ + + + - ]: 966 : .bind(2, debuginfo_p ? 1 : 0)
4776 [ + + + - ]: 1030 : .bind(3, executable_p ? 1 : 0)
4777 [ + - ]: 666 : .bind(4, archiveid)
4778 [ + - ]: 666 : .bind(5, mtime)
4779 [ + - ]: 666 : .bind(6, fileid)
4780 [ + - ]: 666 : .step_ok_done();
4781 [ + + ]: 666 : if (seekable)
4782 : 336 : ps_upsert_seekable
4783 [ + - ]: 336 : .reset()
4784 [ + - ]: 336 : .bind(1, archiveid)
4785 [ + - ]: 336 : .bind(2, fileid)
4786 [ + - ]: 336 : .bind(3, seekable_size)
4787 [ + - ]: 336 : .bind(4, seekable_offset)
4788 [ + - ]: 336 : .bind(5, seekable_mtime)
4789 [ + - ]: 336 : .step_ok_done();
4790 : : }
4791 : : else // potential source - sdef record
4792 : : {
4793 : 374 : fts_sdef ++;
4794 : 374 : ps_upsert_sdef
4795 [ + - ]: 374 : .reset()
4796 [ + - ]: 374 : .bind(1, archiveid)
4797 [ + - ]: 374 : .bind(2, mtime)
4798 [ + - ]: 374 : .bind(3, fileid)
4799 [ + - ]: 374 : .step_ok_done();
4800 : : }
4801 : :
4802 [ + - + + : 1040 : if ((verbose > 2) && (executable_p || debuginfo_p))
+ + ]
4803 : : {
4804 [ + - ]: 666 : obatched ob(clog);
4805 [ + - + - ]: 666 : auto& o = ob << "recorded buildid=" << buildid << " rpm=" << rps << " file=" << fn
4806 [ + - + - : 666 : << " mtime=" << mtime << " atype="
+ - + - +
- + - ]
4807 : : << (executable_p ? "E" : "")
4808 : : << (debuginfo_p ? "D" : "")
4809 [ + - + + : 1330 : << " sourcefiles=" << sourcefiles.size();
+ - + + +
- + - +
- ]
4810 [ + + ]: 666 : if (seekable)
4811 [ + - + - ]: 336 : o << " seekable size=" << seekable_size
4812 [ + - + - ]: 336 : << " offset=" << seekable_offset
4813 [ + - + - ]: 336 : << " mtime=" << seekable_mtime;
4814 [ + - ]: 666 : o << endl;
4815 : 666 : }
4816 : :
4817 : 1040 : }
4818 [ - - ]: 0 : catch (const reportable_exception& e)
4819 : : {
4820 [ - - ]: 0 : e.report(clog);
4821 : 0 : any_exceptions = true;
4822 : : // NB: but we allow the libarchive iteration to continue, in
4823 : : // case we can still gather some useful information. That
4824 : : // would allow some webapi queries to work, until later when
4825 : : // this archive is rescanned. (Its vitals won't go into the
4826 : : // _file_mtime_scanned table until after a successful scan.)
4827 : 0 : }
4828 : : }
4829 : :
4830 [ - + ]: 396 : if (any_exceptions)
4831 [ # # ]: 0 : throw reportable_exception("exceptions encountered during archive scan");
4832 : 396 : }
4833 : :
4834 : :
4835 : :
4836 : : // scan for archive files such as .rpm
4837 : : static void
4838 : 762 : scan_archive_file (const string& rps, const stat_t& st,
4839 : : sqlite_ps& ps_upsert_buildids,
4840 : : sqlite_ps& ps_upsert_fileparts,
4841 : : sqlite_ps& ps_upsert_file,
4842 : : sqlite_ps& ps_lookup_file,
4843 : : sqlite_ps& ps_upsert_de,
4844 : : sqlite_ps& ps_upsert_sref,
4845 : : sqlite_ps& ps_upsert_sdef,
4846 : : sqlite_ps& ps_upsert_seekable,
4847 : : sqlite_ps& ps_query,
4848 : : sqlite_ps& ps_scan_done,
4849 : : unsigned& fts_cached,
4850 : : unsigned& fts_executable,
4851 : : unsigned& fts_debuginfo,
4852 : : unsigned& fts_sref,
4853 : : unsigned& fts_sdef)
4854 : : {
4855 : : // intern the archive file name
4856 : 762 : int64_t archiveid = register_file_name (ps_upsert_fileparts, ps_upsert_file, ps_lookup_file, rps);
4857 : :
4858 : : /* See if we know of it already. */
4859 : 762 : int rc = ps_query
4860 : 762 : .reset()
4861 : 762 : .bind(1, archiveid)
4862 : 762 : .bind(2, st.st_mtime)
4863 : 762 : .step();
4864 : 762 : ps_query.reset();
4865 [ + + ]: 762 : if (rc == SQLITE_ROW) // i.e., a result, as opposed to DONE (no results)
4866 : : // no need to recheck a file/version we already know
4867 : : // specifically, no need to parse this archive again, since we already have
4868 : : // it as a D or E or S record,
4869 : : // (so is stored with buildid=NULL)
4870 : : {
4871 : 366 : fts_cached ++;
4872 : 366 : return;
4873 : : }
4874 : :
4875 : : // extract the archive contents
4876 : 396 : unsigned my_fts_executable = 0, my_fts_debuginfo = 0, my_fts_sref = 0, my_fts_sdef = 0;
4877 : 396 : bool my_fts_sref_complete_p = true;
4878 : 396 : bool any_exceptions = false;
4879 : 396 : try
4880 : : {
4881 [ + - ]: 396 : string archive_extension;
4882 : 396 : archive_classify (rps, archive_extension, archiveid,
4883 : : ps_upsert_buildids, ps_upsert_fileparts, ps_upsert_file, ps_lookup_file,
4884 : : ps_upsert_de, ps_upsert_sref, ps_upsert_sdef, ps_upsert_seekable, // dalt
4885 [ + - ]: 396 : st.st_mtime,
4886 : : my_fts_executable, my_fts_debuginfo, my_fts_sref, my_fts_sdef,
4887 : : my_fts_sref_complete_p);
4888 [ + - + - : 792 : add_metric ("scanned_bytes_total","source",archive_extension + " archive",
+ - ]
4889 [ + - ]: 396 : st.st_size);
4890 [ + - + - : 792 : inc_metric ("scanned_files_total","source",archive_extension + " archive");
+ - + - ]
4891 [ + - + - : 792 : add_metric("found_debuginfo_total","source",archive_extension + " archive",
+ - + - ]
4892 : : my_fts_debuginfo);
4893 [ + - + - : 792 : add_metric("found_executable_total","source",archive_extension + " archive",
+ - + - ]
4894 : : my_fts_executable);
4895 [ + - + - : 792 : add_metric("found_sourcerefs_total","source",archive_extension + " archive",
+ - + - ]
4896 : : my_fts_sref);
4897 : 396 : }
4898 [ - - ]: 0 : catch (const reportable_exception& e)
4899 : : {
4900 [ - - ]: 0 : e.report(clog);
4901 : 0 : any_exceptions = true;
4902 : 0 : }
4903 : :
4904 [ + - ]: 396 : if (verbose > 2)
4905 [ + - ]: 1188 : obatched(clog) << "scanned archive=" << rps
4906 [ + - + - ]: 396 : << " mtime=" << st.st_mtime
4907 [ + - ]: 396 : << " executables=" << my_fts_executable
4908 [ + - + - ]: 396 : << " debuginfos=" << my_fts_debuginfo
4909 [ + - + - ]: 396 : << " srefs=" << my_fts_sref
4910 [ + - + - ]: 396 : << " sdefs=" << my_fts_sdef
4911 [ + - + - : 396 : << " exceptions=" << any_exceptions
+ - + - ]
4912 : 396 : << endl;
4913 : :
4914 : 396 : fts_executable += my_fts_executable;
4915 : 396 : fts_debuginfo += my_fts_debuginfo;
4916 : 396 : fts_sref += my_fts_sref;
4917 : 396 : fts_sdef += my_fts_sdef;
4918 : :
4919 [ - + ]: 396 : if (any_exceptions)
4920 [ # # ]: 0 : throw reportable_exception("exceptions encountered during archive scan");
4921 : :
4922 [ + + ]: 396 : if (my_fts_sref_complete_p) // leave incomplete?
4923 : 394 : ps_scan_done
4924 : 394 : .reset()
4925 : 394 : .bind(1, archiveid)
4926 : 394 : .bind(2, st.st_mtime)
4927 : 394 : .bind(3, st.st_size)
4928 : 394 : .step_ok_done();
4929 : : }
4930 : :
4931 : :
4932 : :
4933 : : ////////////////////////////////////////////////////////////////////////
4934 : :
4935 : :
4936 : :
4937 : : // The thread that consumes file names off of the scanq. We hold
4938 : : // the persistent sqlite_ps's at this level and delegate file/archive
4939 : : // scanning to other functions.
4940 : : static void
4941 : 288 : scan ()
4942 : : {
4943 : : // all the prepared statements fit to use, the _f_ set:
4944 [ + - + - ]: 576 : sqlite_ps ps_f_upsert_buildids (db, "file-buildids-intern", "insert or ignore into " BUILDIDS "_buildids VALUES (NULL, ?);");
4945 [ + - + - : 576 : sqlite_ps ps_f_upsert_fileparts (db, "file-fileparts-intern", "insert or ignore into " BUILDIDS "_fileparts VALUES (NULL, ?);");
+ - ]
4946 [ + - ]: 288 : sqlite_ps ps_f_upsert_file (db, "file-file-intern", "insert or ignore into " BUILDIDS "_files VALUES (NULL, \n"
4947 : : "(select id from " BUILDIDS "_fileparts where name = ?),\n"
4948 [ + - + - : 576 : "(select id from " BUILDIDS "_fileparts where name = ?));");
+ - ]
4949 [ + - ]: 288 : sqlite_ps ps_f_lookup_file (db, "file-file-lookup",
4950 : : "select f.id\n"
4951 : : " from " BUILDIDS "_files f, " BUILDIDS "_fileparts p1, " BUILDIDS "_fileparts p2 \n"
4952 [ + - + - : 576 : " where f.dirname = p1.id and f.basename = p2.id and p1.name = ? and p2.name = ?;\n");
+ - ]
4953 [ + - ]: 288 : sqlite_ps ps_f_upsert_de (db, "file-de-upsert",
4954 : : "insert or ignore into " BUILDIDS "_f_de "
4955 : : "(buildid, debuginfo_p, executable_p, file, mtime) "
4956 : : "values ((select id from " BUILDIDS "_buildids where hex = ?),"
4957 [ + - + - : 576 : " ?,?,?,?);");
+ - ]
4958 [ + - ]: 288 : sqlite_ps ps_f_upsert_s (db, "file-s-upsert",
4959 : : "insert or ignore into " BUILDIDS "_f_s "
4960 : : "(buildid, artifactsrc, file, mtime) "
4961 : : "values ((select id from " BUILDIDS "_buildids where hex = ?),"
4962 [ + - + - : 576 : " ?,?,?);");
+ - ]
4963 [ + - ]: 288 : sqlite_ps ps_f_query (db, "file-negativehit-find",
4964 : : "select 1 from " BUILDIDS "_file_mtime_scanned where sourcetype = 'F' "
4965 [ + - + - : 576 : "and file = ? and mtime = ?;");
+ - ]
4966 [ + - ]: 288 : sqlite_ps ps_f_scan_done (db, "file-scanned",
4967 : : "insert or ignore into " BUILDIDS "_file_mtime_scanned (sourcetype, file, mtime, size)"
4968 [ + - + - : 576 : "values ('F', ?,?,?);");
+ - ]
4969 : :
4970 : : // and now for the _r_ set
4971 [ + - + - : 576 : sqlite_ps ps_r_upsert_buildids (db, "rpm-buildid-intern", "insert or ignore into " BUILDIDS "_buildids VALUES (NULL, ?);");
+ - ]
4972 [ + - + - : 576 : sqlite_ps ps_r_upsert_fileparts (db, "rpm-fileparts-intern", "insert or ignore into " BUILDIDS "_fileparts VALUES (NULL, ?);");
+ - ]
4973 [ + - ]: 288 : sqlite_ps ps_r_upsert_file (db, "rpm-file-intern", "insert or ignore into " BUILDIDS "_files VALUES (NULL, \n"
4974 : : "(select id from " BUILDIDS "_fileparts where name = ?),\n"
4975 [ + - + - : 576 : "(select id from " BUILDIDS "_fileparts where name = ?));");
+ - ]
4976 [ + - ]: 288 : sqlite_ps ps_r_lookup_file (db, "rpm-file-lookup",
4977 : : "select f.id\n"
4978 : : " from " BUILDIDS "_files f, " BUILDIDS "_fileparts p1, " BUILDIDS "_fileparts p2 \n"
4979 [ + - + - : 576 : " where f.dirname = p1.id and f.basename = p2.id and p1.name = ? and p2.name = ?;\n");
+ - ]
4980 [ + - ]: 288 : sqlite_ps ps_r_upsert_de (db, "rpm-de-insert",
4981 : : "insert or ignore into " BUILDIDS "_r_de (buildid, debuginfo_p, executable_p, file, mtime, content) values ("
4982 [ + - + - : 576 : "(select id from " BUILDIDS "_buildids where hex = ?), ?, ?, ?, ?, ?);");
+ - ]
4983 [ + - ]: 288 : sqlite_ps ps_r_upsert_sref (db, "rpm-sref-insert",
4984 : : "insert or ignore into " BUILDIDS "_r_sref (buildid, artifactsrc) values ("
4985 : : "(select id from " BUILDIDS "_buildids where hex = ?), "
4986 [ + - + - : 576 : "?);");
+ - ]
4987 [ + - ]: 288 : sqlite_ps ps_r_upsert_sdef (db, "rpm-sdef-insert",
4988 : : "insert or ignore into " BUILDIDS "_r_sdef (file, mtime, content) values ("
4989 [ + - + - : 576 : "?, ?, ?);");
+ - ]
4990 [ + - ]: 288 : sqlite_ps ps_r_upsert_seekable (db, "rpm-seekable-insert",
4991 : : "insert or ignore into " BUILDIDS "_r_seekable (file, content, type, size, offset, mtime) "
4992 [ + - + - : 576 : "values (?, ?, 'xz', ?, ?, ?);");
+ - ]
4993 [ + - ]: 288 : sqlite_ps ps_r_query (db, "rpm-negativehit-query",
4994 : : "select 1 from " BUILDIDS "_file_mtime_scanned where "
4995 [ + - + - : 576 : "sourcetype = 'R' and file = ? and mtime = ?;");
+ - ]
4996 [ + - ]: 288 : sqlite_ps ps_r_scan_done (db, "rpm-scanned",
4997 : : "insert or ignore into " BUILDIDS "_file_mtime_scanned (sourcetype, file, mtime, size)"
4998 [ + - + - : 576 : "values ('R', ?, ?, ?);");
+ - ]
4999 : :
5000 : :
5001 : 288 : unsigned fts_cached = 0, fts_executable = 0, fts_debuginfo = 0, fts_sourcefiles = 0;
5002 : 288 : unsigned fts_sref = 0, fts_sdef = 0;
5003 : :
5004 [ + - + - : 576 : add_metric("thread_count", "role", "scan", 1);
+ - + - ]
5005 [ + - + - : 576 : add_metric("thread_busy", "role", "scan", 1);
+ - + - ]
5006 [ + + ]: 1890 : while (! interrupted)
5007 : : {
5008 [ + - ]: 1602 : scan_payload p;
5009 : :
5010 [ + - + - : 3204 : add_metric("thread_busy", "role", "scan", -1);
+ - + - ]
5011 : : // NB: threads may be blocked within either of these two waiting
5012 : : // states, if the work queue happens to run dry. That's OK.
5013 [ + - + - ]: 1602 : if (scan_barrier) scan_barrier->count();
5014 [ + - ]: 1602 : bool gotone = scanq.wait_front(p);
5015 [ + - + - : 3203 : add_metric("thread_busy", "role", "scan", 1);
+ - + - ]
5016 : :
5017 [ + + ]: 1602 : if (! gotone) continue; // go back to waiting
5018 : :
5019 : 1314 : try
5020 : : {
5021 : 1314 : bool scan_archive = false;
5022 [ + + ]: 2886 : for (auto&& arch : scan_archives)
5023 [ + + ]: 1572 : if (string_endswith(p.first, arch.first))
5024 : 762 : scan_archive = true;
5025 : :
5026 [ + + ]: 1314 : if (scan_archive)
5027 [ + - ]: 762 : scan_archive_file (p.first, p.second,
5028 : : ps_r_upsert_buildids,
5029 : : ps_r_upsert_fileparts,
5030 : : ps_r_upsert_file,
5031 : : ps_r_lookup_file,
5032 : : ps_r_upsert_de,
5033 : : ps_r_upsert_sref,
5034 : : ps_r_upsert_sdef,
5035 : : ps_r_upsert_seekable,
5036 : : ps_r_query,
5037 : : ps_r_scan_done,
5038 : : fts_cached,
5039 : : fts_executable,
5040 : : fts_debuginfo,
5041 : : fts_sref,
5042 : : fts_sdef);
5043 : :
5044 [ + + ]: 1314 : if (scan_files) // NB: maybe "else if" ?
5045 [ + - ]: 1106 : scan_source_file (p.first, p.second,
5046 : : ps_f_upsert_buildids,
5047 : : ps_f_upsert_fileparts,
5048 : : ps_f_upsert_file,
5049 : : ps_f_lookup_file,
5050 : : ps_f_upsert_de,
5051 : : ps_f_upsert_s,
5052 : : ps_f_query,
5053 : : ps_f_scan_done,
5054 : : fts_cached, fts_executable, fts_debuginfo, fts_sourcefiles);
5055 : : }
5056 [ - - ]: 0 : catch (const reportable_exception& e)
5057 : : {
5058 [ - - ]: 0 : e.report(cerr);
5059 : 0 : }
5060 : :
5061 [ + - ]: 1314 : scanq.done_front(); // let idlers run
5062 : :
5063 : 1314 : if (fts_cached || fts_executable || fts_debuginfo || fts_sourcefiles || fts_sref || fts_sdef)
5064 : : {} // NB: not just if a successful scan - we might have encountered -ENOSPC & failed
5065 [ + - + - ]: 1314 : (void) statfs_free_enough_p(db_path, "database"); // report sqlite filesystem size
5066 [ + - + - ]: 1314 : (void) statfs_free_enough_p(tmpdir, "tmpdir"); // this too, in case of fdcache/tmpfile usage
5067 : :
5068 : : // finished a scanning step -- not a "loop", because we just
5069 : : // consume the traversal loop's work, whenever
5070 [ + - + - : 2628 : inc_metric("thread_work_total","role","scan");
+ - + - ]
5071 : 1602 : }
5072 : :
5073 [ + - + - : 576 : add_metric("thread_busy", "role", "scan", -1);
+ - + - ]
5074 : 288 : }
5075 : :
5076 : :
5077 : : // Use this function as the thread entry point, so it can catch our
5078 : : // fleet of exceptions (incl. the sqlite_ps ctors) and report.
5079 : : static void*
5080 : 288 : thread_main_scanner (void* arg)
5081 : : {
5082 : 288 : (void) arg;
5083 [ + + ]: 864 : while (! interrupted)
5084 : 288 : try
5085 : : {
5086 [ + - ]: 288 : scan();
5087 : : }
5088 [ - - ]: 0 : catch (const reportable_exception& e)
5089 : : {
5090 [ - - ]: 0 : e.report(cerr);
5091 : 0 : }
5092 : 288 : return 0;
5093 : : }
5094 : :
5095 : :
5096 : :
5097 : : // The thread that traverses all the source_paths and enqueues all the
5098 : : // matching files into the file/archive scan queue.
5099 : : static void
5100 : 126 : scan_source_paths()
5101 : : {
5102 : : // NB: fedora 31 glibc/fts(3) crashes inside fts_read() on empty
5103 : : // path list.
5104 [ + + ]: 126 : if (source_paths.empty())
5105 : 2 : return;
5106 : :
5107 : : // Turn the source_paths into an fts(3)-compatible char**. Since
5108 : : // source_paths[] does not change after argv processing, the
5109 : : // c_str()'s are safe to keep around awile.
5110 : 124 : vector<const char *> sps;
5111 [ + + ]: 332 : for (auto&& sp: source_paths)
5112 [ + - ]: 208 : sps.push_back(sp.c_str());
5113 [ + - - - ]: 124 : sps.push_back(NULL);
5114 : :
5115 [ + + + - ]: 234 : FTS *fts = fts_open ((char * const *)sps.data(),
5116 : : (traverse_logical ? FTS_LOGICAL : FTS_PHYSICAL|FTS_XDEV)
5117 : : | FTS_NOCHDIR /* multithreaded */,
5118 : : NULL);
5119 [ - + ]: 124 : if (fts == NULL)
5120 [ # # # # ]: 0 : throw libc_exception(errno, "cannot fts_open");
5121 : 124 : defer_dtor<FTS*,int> fts_cleanup (fts, fts_close);
5122 : :
5123 : 124 : struct timespec ts_start, ts_end;
5124 : 124 : clock_gettime (CLOCK_MONOTONIC, &ts_start);
5125 : 124 : unsigned fts_scanned = 0, fts_regex = 0;
5126 : :
5127 : 124 : FTSENT *f;
5128 [ + - + + ]: 2644 : while ((f = fts_read (fts)) != NULL)
5129 : : {
5130 [ + - ]: 2396 : if (interrupted) break;
5131 : :
5132 [ - + ]: 2396 : if (sigusr2 != forced_groom_count) // stop early if groom triggered
5133 : : {
5134 [ # # ]: 0 : scanq.clear(); // clear previously issued work for scanner threads
5135 : : break;
5136 : : }
5137 : :
5138 : 2396 : fts_scanned ++;
5139 : :
5140 [ + - ]: 2396 : if (verbose > 2)
5141 [ + - + - : 4792 : obatched(clog) << "fts traversing " << f->fts_path << endl;
+ - ]
5142 : :
5143 [ + + + + : 2396 : switch (f->fts_info)
+ ]
5144 : : {
5145 : 1424 : case FTS_F:
5146 : 1424 : {
5147 : : /* Found a file. Convert it to an absolute path, so
5148 : : the buildid database does not have relative path
5149 : : names that are unresolvable from a subsequent run
5150 : : in a different cwd. */
5151 [ + - ]: 1424 : char *rp = realpath(f->fts_path, NULL);
5152 [ - + ]: 1424 : if (rp == NULL)
5153 : 0 : continue; // ignore dangling symlink or such
5154 [ + - ]: 1424 : string rps = string(rp);
5155 : 1424 : free (rp);
5156 : :
5157 [ + - ]: 1424 : bool ri = !regexec (&file_include_regex, rps.c_str(), 0, 0, 0);
5158 [ + - ]: 1424 : bool rx = !regexec (&file_exclude_regex, rps.c_str(), 0, 0, 0);
5159 [ + + ]: 1424 : if (!ri || rx)
5160 : : {
5161 [ + - ]: 110 : if (verbose > 3)
5162 [ + - ]: 220 : obatched(clog) << "fts skipped by regex "
5163 [ + + + - : 134 : << (!ri ? "I" : "") << (rx ? "X" : "") << endl;
+ + + - +
- ]
5164 : 110 : fts_regex ++;
5165 [ + + ]: 110 : if (!ri)
5166 [ + - + - : 24 : inc_metric("traversed_total","type","file-skipped-I");
+ - + - ]
5167 [ + + ]: 110 : if (rx)
5168 [ + - + - : 196 : inc_metric("traversed_total","type","file-skipped-X");
+ - + - ]
5169 : : }
5170 : : else
5171 : : {
5172 [ + - + - ]: 1314 : scanq.push_back (make_pair(rps, *f->fts_statp));
5173 [ + - + - : 2628 : inc_metric("traversed_total","type","file");
+ - + - ]
5174 : : }
5175 : 0 : }
5176 : 1424 : break;
5177 : :
5178 : 4 : case FTS_ERR:
5179 : 4 : case FTS_NS:
5180 : : // report on some types of errors because they may reflect fixable misconfiguration
5181 : 4 : {
5182 [ + - + - : 8 : auto x = libc_exception(f->fts_errno, string("fts traversal ") + string(f->fts_path));
+ - + - ]
5183 [ + - ]: 4 : x.report(cerr);
5184 : 0 : }
5185 [ + - + - : 8 : inc_metric("traversed_total","type","error");
+ - + - ]
5186 : 4 : break;
5187 : :
5188 : 32 : case FTS_SL: // ignore, but count because debuginfod -L would traverse these
5189 [ + - + - : 64 : inc_metric("traversed_total","type","symlink");
+ - + - ]
5190 : 32 : break;
5191 : :
5192 : 468 : case FTS_D: // ignore
5193 [ + - + - : 936 : inc_metric("traversed_total","type","directory");
+ - + - ]
5194 : 468 : break;
5195 : :
5196 : 468 : default: // ignore
5197 [ + - + - : 936 : inc_metric("traversed_total","type","other");
+ - + - ]
5198 : 468 : break;
5199 : : }
5200 : : }
5201 : 124 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
5202 : 124 : double deltas = (ts_end.tv_sec - ts_start.tv_sec) + (ts_end.tv_nsec - ts_start.tv_nsec)/1.e9;
5203 : :
5204 [ + - + - : 372 : obatched(clog) << "fts traversed source paths in " << deltas << "s, scanned=" << fts_scanned
+ - + - ]
5205 [ + - + - : 124 : << ", regex-skipped=" << fts_regex << endl;
+ - ]
5206 [ + - ]: 248 : }
5207 : :
5208 : :
5209 : : static void*
5210 : 72 : thread_main_fts_source_paths (void* arg)
5211 : : {
5212 : 72 : (void) arg; // ignore; we operate on global data
5213 : :
5214 [ + - + - : 144 : set_metric("thread_tid", "role","traverse", tid());
+ - ]
5215 [ + - + - : 144 : add_metric("thread_count", "role", "traverse", 1);
+ - ]
5216 : :
5217 : 72 : time_t last_rescan = 0;
5218 : :
5219 [ + - ]: 334 : while (! interrupted)
5220 : : {
5221 : 334 : sleep (1);
5222 : 334 : scanq.wait_idle(); // don't start a new traversal while scanners haven't finished the job
5223 : 334 : scanq.done_idle(); // release the hounds
5224 [ + + ]: 334 : if (interrupted) break;
5225 : :
5226 : 262 : time_t now = time(NULL);
5227 : 262 : bool rescan_now = false;
5228 [ + + ]: 262 : if (last_rescan == 0) // at least one initial rescan is documented even for -t0
5229 : 70 : rescan_now = true;
5230 [ + + + + ]: 262 : if (rescan_s > 0 && (long)now > (long)(last_rescan + rescan_s))
5231 : 262 : rescan_now = true;
5232 [ + + ]: 262 : if (sigusr1 != forced_rescan_count)
5233 : : {
5234 : 58 : forced_rescan_count = sigusr1;
5235 : 58 : rescan_now = true;
5236 : : }
5237 [ + + ]: 262 : if (rescan_now)
5238 : : {
5239 [ + - + - : 252 : set_metric("thread_busy", "role","traverse", 1);
+ - ]
5240 : 126 : try
5241 : : {
5242 [ + - ]: 126 : scan_source_paths();
5243 : : }
5244 [ - - ]: 0 : catch (const reportable_exception& e)
5245 : : {
5246 [ - - ]: 0 : e.report(cerr);
5247 : 0 : }
5248 : 126 : last_rescan = time(NULL); // NB: now was before scanning
5249 : : // finished a traversal loop
5250 [ + - + - : 252 : inc_metric("thread_work_total", "role","traverse");
+ - ]
5251 [ + - + - : 252 : set_metric("thread_busy", "role","traverse", 0);
+ - ]
5252 : : }
5253 : : }
5254 : :
5255 : 72 : return 0;
5256 : : }
5257 : :
5258 : :
5259 : :
5260 : : ////////////////////////////////////////////////////////////////////////
5261 : :
5262 : : static void
5263 : 80 : database_stats_report()
5264 : : {
5265 : 80 : sqlite_ps ps_query (db, "database-overview",
5266 [ + - + - ]: 160 : "select label,quantity from " BUILDIDS "_stats");
5267 : :
5268 [ + - + - ]: 160 : obatched(clog) << "database record counts:" << endl;
5269 : 1840 : while (1)
5270 : : {
5271 [ + - ]: 960 : if (interrupted) break;
5272 [ + - ]: 960 : if (sigusr1 != forced_rescan_count) // stop early if scan triggered
5273 : : break;
5274 : :
5275 [ + - ]: 960 : int rc = ps_query.step();
5276 [ + + ]: 960 : if (rc == SQLITE_DONE) break;
5277 [ - + ]: 880 : if (rc != SQLITE_ROW)
5278 [ # # # # ]: 0 : throw sqlite_exception(rc, "step");
5279 : :
5280 [ + - ]: 880 : obatched(clog)
5281 [ + - - + : 880 : << ((const char*) sqlite3_column_text(ps_query, 0) ?: (const char*) "NULL")
+ - ]
5282 : : << " "
5283 [ + - + - : 1760 : << (sqlite3_column_text(ps_query, 1) ?: (const unsigned char*) "NULL")
- + + - ]
5284 : 880 : << endl;
5285 : :
5286 [ + - + - : 1760 : set_metric("groom", "statistic",
- + + - +
- + - +
- ]
5287 [ + - ]: 880 : ((const char*) sqlite3_column_text(ps_query, 0) ?: (const char*) "NULL"),
5288 : : (sqlite3_column_double(ps_query, 1)));
5289 : 880 : }
5290 : 80 : }
5291 : :
5292 : :
5293 : : // Do a round of database grooming that might take many minutes to run.
5294 : 80 : void groom()
5295 : : {
5296 [ + - ]: 160 : obatched(clog) << "grooming database" << endl;
5297 : :
5298 : 80 : struct timespec ts_start, ts_end;
5299 : 80 : clock_gettime (CLOCK_MONOTONIC, &ts_start);
5300 : :
5301 : : // scan for files that have disappeared
5302 : 80 : sqlite_ps files (db, "check old files",
5303 : : "select distinct s.mtime, s.file, f.name from "
5304 : : BUILDIDS "_file_mtime_scanned s, " BUILDIDS "_files_v f "
5305 [ + - + - ]: 160 : "where f.id = s.file");
5306 : : // NB: Because _ftime_mtime_scanned can contain both F and
5307 : : // R records for the same file, this query would return duplicates if the
5308 : : // DISTINCT qualifier were not there.
5309 [ + - ]: 80 : files.reset();
5310 : :
5311 : : // DECISION TIME - we enumerate stale fileids/mtimes
5312 [ + - ]: 80 : deque<pair<int64_t,int64_t> > stale_fileid_mtime;
5313 : :
5314 : 80 : time_t time_start = time(NULL);
5315 : 344 : while(1)
5316 : : {
5317 : : // PR28514: limit grooming iteration to O(rescan time), to avoid
5318 : : // slow filesystem tests over many files locking out rescans for
5319 : : // too long.
5320 [ + + - + ]: 212 : if (rescan_s > 0 && (long)time(NULL) > (long)(time_start + rescan_s))
5321 : : {
5322 [ # # # # : 0 : inc_metric("groomed_total", "decision", "aborted");
# # # # ]
5323 : 0 : break;
5324 : : }
5325 : :
5326 [ + - ]: 212 : if (interrupted) break;
5327 : :
5328 [ + - ]: 212 : int rc = files.step();
5329 [ + + ]: 212 : if (rc != SQLITE_ROW)
5330 : : break;
5331 : :
5332 [ + - ]: 132 : int64_t mtime = sqlite3_column_int64 (files, 0);
5333 [ + - ]: 132 : int64_t fileid = sqlite3_column_int64 (files, 1);
5334 [ + - - + ]: 132 : const char* filename = ((const char*) sqlite3_column_text (files, 2) ?: "");
5335 : 132 : struct stat s;
5336 : 132 : bool regex_file_drop = 0;
5337 : :
5338 [ + + ]: 132 : if (regex_groom)
5339 : : {
5340 [ + - ]: 16 : bool reg_include = !regexec (&file_include_regex, filename, 0, 0, 0);
5341 [ + - ]: 16 : bool reg_exclude = !regexec (&file_exclude_regex, filename, 0, 0, 0);
5342 : 16 : regex_file_drop = !reg_include || reg_exclude; // match logic of scan_source_paths
5343 : : }
5344 : :
5345 : 132 : rc = stat(filename, &s);
5346 [ + + - + ]: 132 : if ( regex_file_drop || rc < 0 || (mtime != (int64_t) s.st_mtime) )
5347 : : {
5348 [ + - ]: 24 : if (verbose > 2)
5349 [ + - + - : 48 : obatched(clog) << "groom: stale file=" << filename << " mtime=" << mtime << endl;
+ - + - +
- ]
5350 [ + - ]: 24 : stale_fileid_mtime.push_back(make_pair(fileid,mtime));
5351 [ + - + - : 48 : inc_metric("groomed_total", "decision", "stale");
+ - + - ]
5352 [ + - + - : 48 : set_metric("thread_work_pending","role","groom", stale_fileid_mtime.size());
+ - + - ]
5353 : : }
5354 : : else
5355 [ + - + - : 216 : inc_metric("groomed_total", "decision", "fresh");
+ - + - ]
5356 : :
5357 [ + - ]: 132 : if (sigusr1 != forced_rescan_count) // stop early if scan triggered
5358 : : break;
5359 : 132 : }
5360 [ + - ]: 80 : files.reset();
5361 : :
5362 : : // ACTION TIME
5363 : :
5364 : : // Now that we know which file/mtime tuples are stale, actually do
5365 : : // the deletion from the database. Doing this during the SELECT
5366 : : // iteration above results in undefined behaviour in sqlite, as per
5367 : : // https://www.sqlite.org/isolation.html
5368 : :
5369 : : // We could shuffle stale_fileid_mtime[] here. It'd let aborted
5370 : : // sequences of nuke operations resume at random locations, instead
5371 : : // of just starting over. But it doesn't matter much either way,
5372 : : // as long as we make progress.
5373 : :
5374 [ + - + - : 160 : sqlite_ps files_del_f_de (db, "nuke f_de", "delete from " BUILDIDS "_f_de where file = ? and mtime = ?");
+ - ]
5375 [ + - + - : 160 : sqlite_ps files_del_r_de (db, "nuke r_de", "delete from " BUILDIDS "_r_de where file = ? and mtime = ?");
+ - ]
5376 [ + - ]: 80 : sqlite_ps files_del_scan (db, "nuke f_m_s", "delete from " BUILDIDS "_file_mtime_scanned "
5377 [ + - + - : 160 : "where file = ? and mtime = ?");
+ - ]
5378 : :
5379 [ + + ]: 104 : while (! stale_fileid_mtime.empty())
5380 : : {
5381 : 24 : auto stale = stale_fileid_mtime.front();
5382 : 24 : stale_fileid_mtime.pop_front();
5383 [ + - + - : 48 : set_metric("thread_work_pending","role","groom", stale_fileid_mtime.size());
+ - + - ]
5384 : :
5385 : : // PR28514: limit grooming iteration to O(rescan time), to avoid
5386 : : // slow nuke_* queries over many files locking out rescans for too
5387 : : // long. We iterate over the files in random() sequence to avoid
5388 : : // partial checks going over the same set.
5389 [ - + - - ]: 24 : if (rescan_s > 0 && (long)time(NULL) > (long)(time_start + rescan_s))
5390 : : {
5391 [ # # # # : 0 : inc_metric("groomed_total", "action", "aborted");
# # # # ]
5392 : 0 : break;
5393 : : }
5394 : :
5395 [ + - ]: 24 : if (interrupted) break;
5396 : :
5397 : 24 : int64_t fileid = stale.first;
5398 : 24 : int64_t mtime = stale.second;
5399 [ + - + - : 24 : files_del_f_de.reset().bind(1,fileid).bind(2,mtime).step_ok_done();
+ - + - ]
5400 [ + - + - : 24 : files_del_r_de.reset().bind(1,fileid).bind(2,mtime).step_ok_done();
+ - + - ]
5401 [ + - + - : 24 : files_del_scan.reset().bind(1,fileid).bind(2,mtime).step_ok_done();
+ - + - ]
5402 [ + - + - : 48 : inc_metric("groomed_total", "action", "cleaned");
+ - + - ]
5403 : :
5404 [ + - ]: 24 : if (sigusr1 != forced_rescan_count) // stop early if scan triggered
5405 : : break;
5406 : : }
5407 : 80 : stale_fileid_mtime.clear(); // no need for this any longer
5408 [ + - + - : 160 : set_metric("thread_work_pending","role","groom", stale_fileid_mtime.size());
+ - + - ]
5409 : :
5410 : : // delete buildids with no references in _r_de or _f_de tables;
5411 : : // cascades to _r_sref & _f_s records
5412 [ + - ]: 80 : sqlite_ps buildids_del (db, "nuke orphan buildids",
5413 : : "delete from " BUILDIDS "_buildids "
5414 : : "where not exists (select 1 from " BUILDIDS "_f_de d where " BUILDIDS "_buildids.id = d.buildid) "
5415 [ + - + - : 160 : "and not exists (select 1 from " BUILDIDS "_r_de d where " BUILDIDS "_buildids.id = d.buildid)");
+ - ]
5416 [ + - + - ]: 80 : buildids_del.reset().step_ok_done();
5417 : :
5418 [ - + ]: 80 : if (interrupted) return;
5419 : :
5420 : : // NB: "vacuum" is too heavy for even daily runs: it rewrites the entire db, so is done as maxigroom -G
5421 [ + - + - : 160 : { sqlite_ps g (db, "incremental vacuum", "pragma incremental_vacuum"); g.reset().step_ok_done(); }
+ - + - +
- ]
5422 : : // https://www.sqlite.org/lang_analyze.html#approx
5423 [ + - + - : 160 : { sqlite_ps g (db, "analyze setup", "pragma analysis_limit = 1000;\n"); g.reset().step_ok_done(); }
+ - + - +
- ]
5424 [ + - + - : 160 : { sqlite_ps g (db, "analyze", "analyze"); g.reset().step_ok_done(); }
+ - + - +
- ]
5425 [ + - + - : 160 : { sqlite_ps g (db, "analyze reload", "analyze sqlite_schema"); g.reset().step_ok_done(); }
+ - + - +
- ]
5426 [ + - + - : 160 : { sqlite_ps g (db, "optimize", "pragma optimize"); g.reset().step_ok_done(); }
+ - + - +
- ]
5427 [ + - + - : 160 : { sqlite_ps g (db, "wal checkpoint", "pragma wal_checkpoint=truncate"); g.reset().step_ok_done(); }
+ - + - +
- ]
5428 : :
5429 [ + - ]: 80 : database_stats_report();
5430 : :
5431 [ + - + - ]: 80 : (void) statfs_free_enough_p(db_path, "database"); // report sqlite filesystem size
5432 : :
5433 [ + - ]: 80 : sqlite3_db_release_memory(db); // shrink the process if possible
5434 [ + - ]: 80 : sqlite3_db_release_memory(dbq); // ... for both connections
5435 [ + - ]: 80 : debuginfod_pool_groom(); // and release any debuginfod_client objects we've been holding onto
5436 : : #if HAVE_MALLOC_TRIM
5437 : 80 : malloc_trim(0); // PR31103: release memory allocated for temporary purposes
5438 : : #endif
5439 : :
5440 : : #if 0 /* PR31265: don't jettison cache unnecessarily */
5441 : : fdcache.limit(0); // release the fdcache contents
5442 : : fdcache.limit(fdcache_mbs); // restore status quo parameters
5443 : : #endif
5444 : :
5445 : 80 : clock_gettime (CLOCK_MONOTONIC, &ts_end);
5446 : 80 : double deltas = (ts_end.tv_sec - ts_start.tv_sec) + (ts_end.tv_nsec - ts_start.tv_nsec)/1.e9;
5447 : :
5448 [ + - + - : 160 : obatched(clog) << "groomed database in " << deltas << "s" << endl;
+ - + - ]
5449 : 80 : }
5450 : :
5451 : :
5452 : : static void*
5453 : 82 : thread_main_groom (void* /*arg*/)
5454 : : {
5455 [ + - + - : 164 : set_metric("thread_tid", "role", "groom", tid());
+ - ]
5456 [ + - + - : 164 : add_metric("thread_count", "role", "groom", 1);
+ - ]
5457 : :
5458 : 82 : time_t last_groom = 0;
5459 : :
5460 : 612 : while (1)
5461 : : {
5462 : 347 : sleep (1);
5463 : 347 : scanq.wait_idle(); // PR25394: block scanners during grooming!
5464 [ + + ]: 347 : if (interrupted) break;
5465 : :
5466 : 265 : time_t now = time(NULL);
5467 : 265 : bool groom_now = false;
5468 [ + + ]: 265 : if (last_groom == 0) // at least one initial groom is documented even for -g0
5469 : 74 : groom_now = true;
5470 [ + + + + ]: 265 : if (groom_s > 0 && (long)now > (long)(last_groom + groom_s))
5471 : 265 : groom_now = true;
5472 [ + + ]: 265 : if (sigusr2 != forced_groom_count)
5473 : : {
5474 : 6 : forced_groom_count = sigusr2;
5475 : 6 : groom_now = true;
5476 : : }
5477 [ + + ]: 265 : if (groom_now)
5478 : : {
5479 [ + - + - : 160 : set_metric("thread_busy", "role", "groom", 1);
+ - ]
5480 : 80 : try
5481 : : {
5482 [ + - ]: 80 : groom ();
5483 : : }
5484 [ - - ]: 0 : catch (const sqlite_exception& e)
5485 : : {
5486 [ - - - - : 0 : obatched(cerr) << e.message << endl;
- - ]
5487 : 0 : }
5488 : 80 : last_groom = time(NULL); // NB: now was before grooming
5489 : : // finished a grooming loop
5490 [ + - + - : 160 : inc_metric("thread_work_total", "role", "groom");
+ - ]
5491 [ + - + - : 160 : set_metric("thread_busy", "role", "groom", 0);
+ - ]
5492 : : }
5493 : :
5494 : 265 : scanq.done_idle();
5495 : 265 : }
5496 : :
5497 : 82 : return 0;
5498 : : }
5499 : :
5500 : :
5501 : : ////////////////////////////////////////////////////////////////////////
5502 : :
5503 : :
5504 : : static void
5505 : 84 : signal_handler (int /* sig */)
5506 : : {
5507 : 84 : interrupted ++;
5508 : :
5509 [ + + ]: 84 : if (db)
5510 : 82 : sqlite3_interrupt (db);
5511 [ + - ]: 84 : if (dbq)
5512 : 84 : sqlite3_interrupt (dbq);
5513 : :
5514 : : // NB: don't do anything else in here
5515 : 84 : }
5516 : :
5517 : : static void
5518 : 58 : sigusr1_handler (int /* sig */)
5519 : : {
5520 : 58 : sigusr1 ++;
5521 : : // NB: don't do anything else in here
5522 : 58 : }
5523 : :
5524 : : static void
5525 : 6 : sigusr2_handler (int /* sig */)
5526 : : {
5527 : 6 : sigusr2 ++;
5528 : : // NB: don't do anything else in here
5529 : 6 : }
5530 : :
5531 : :
5532 : : static void // error logging callback from libmicrohttpd internals
5533 : 0 : error_cb (void *arg, const char *fmt, va_list ap)
5534 : : {
5535 : 0 : (void) arg;
5536 [ # # # # : 0 : inc_metric("error_count","libmicrohttpd",fmt);
# # ]
5537 : 0 : char errmsg[512];
5538 : 0 : (void) vsnprintf (errmsg, sizeof(errmsg), fmt, ap); // ok if slightly truncated
5539 [ # # ]: 0 : obatched(cerr) << "libmicrohttpd error: " << errmsg; // MHD_DLOG calls already include \n
5540 : 0 : }
5541 : :
5542 : :
5543 : : // A user-defined sqlite function, to score the sharedness of the
5544 : : // prefix of two strings. This is used to compare candidate debuginfo
5545 : : // / source-rpm names, so that the closest match
5546 : : // (directory-topology-wise closest) is found. This is important in
5547 : : // case the same sref (source file name) is in many -debuginfo or
5548 : : // -debugsource RPMs, such as when multiple versions/releases of the
5549 : : // same package are in the database.
5550 : :
5551 : 1344 : static void sqlite3_sharedprefix_fn (sqlite3_context* c, int argc, sqlite3_value** argv)
5552 : : {
5553 [ - + ]: 1344 : if (argc != 2)
5554 : 0 : sqlite3_result_error(c, "expect 2 string arguments", -1);
5555 [ + - + + ]: 2688 : else if ((sqlite3_value_type(argv[0]) != SQLITE_TEXT) ||
5556 : 1344 : (sqlite3_value_type(argv[1]) != SQLITE_TEXT))
5557 : 1082 : sqlite3_result_null(c);
5558 : : else
5559 : : {
5560 : 262 : const unsigned char* a = sqlite3_value_text (argv[0]);
5561 : 262 : const unsigned char* b = sqlite3_value_text (argv[1]);
5562 : 262 : int i = 0;
5563 [ + + + - : 27722 : while (*a != '\0' && *b != '\0' && *a++ == *b++)
+ + ]
5564 : 27198 : i++;
5565 : 262 : sqlite3_result_int (c, i);
5566 : : }
5567 : 1344 : }
5568 : :
5569 : :
5570 : : static unsigned
5571 : 164 : default_concurrency() // guaranteed >= 1
5572 : : {
5573 : : // Prior to PR29975 & PR29976, we'd just use this:
5574 : 164 : unsigned sth = std::thread::hardware_concurrency();
5575 : : // ... but on many-CPU boxes, admins or distros may throttle
5576 : : // resources in such a way that debuginfod would mysteriously fail.
5577 : : // So we reduce the defaults:
5578 : :
5579 : 164 : unsigned aff = 0;
5580 : : #ifdef HAVE_SCHED_GETAFFINITY
5581 : 164 : {
5582 : 164 : int ret;
5583 : 164 : cpu_set_t mask;
5584 : 164 : CPU_ZERO(&mask);
5585 : 164 : ret = sched_getaffinity(0, sizeof(mask), &mask);
5586 [ + - ]: 164 : if (ret == 0)
5587 : 164 : aff = CPU_COUNT(&mask);
5588 : : }
5589 : : #endif
5590 : :
5591 : 164 : unsigned fn = 0;
5592 : : #ifdef HAVE_GETRLIMIT
5593 : 164 : {
5594 : 164 : struct rlimit rlim;
5595 : 164 : int rc = getrlimit(RLIMIT_NOFILE, &rlim);
5596 [ + - ]: 164 : if (rc == 0)
5597 [ - + ]: 164 : fn = max((rlim_t)1, (rlim.rlim_cur - 100) / 4);
5598 : : // at least 2 fds are used by each listener thread etc.
5599 : : // plus a bunch to account for shared libraries and such
5600 : : }
5601 : : #endif
5602 : :
5603 [ - + + - : 328 : unsigned d = min(max(sth, 1U),
- + ]
5604 [ + - ]: 164 : min(max(aff, 1U),
5605 [ - + ]: 164 : max(fn, 1U)));
5606 : 164 : return d;
5607 : : }
5608 : :
5609 : :
5610 : : // 30879: Something to help out in case of an uncaught exception.
5611 : 0 : void my_terminate_handler()
5612 : : {
5613 : : #if defined(__GLIBC__)
5614 : 0 : void *array[40];
5615 : 0 : int size = backtrace (array, 40);
5616 : 0 : backtrace_symbols_fd (array, size, STDERR_FILENO);
5617 : : #endif
5618 : : #if defined(__GLIBCXX__) || defined(__GLIBCPP__)
5619 : 0 : __gnu_cxx::__verbose_terminate_handler();
5620 : : #endif
5621 : 0 : abort();
5622 : : }
5623 : :
5624 : :
5625 : : int
5626 : 84 : main (int argc, char *argv[])
5627 : : {
5628 : 84 : (void) setlocale (LC_ALL, "");
5629 : 84 : (void) bindtextdomain (PACKAGE_TARNAME, LOCALEDIR);
5630 : 84 : (void) textdomain (PACKAGE_TARNAME);
5631 : :
5632 : 84 : std::set_terminate(& my_terminate_handler);
5633 : :
5634 : : /* Tell the library which version we are expecting. */
5635 : 84 : elf_version (EV_CURRENT);
5636 : :
5637 [ + - ]: 168 : tmpdir = string(getenv("TMPDIR") ?: "/tmp");
5638 : :
5639 : : /* Set computed default values. */
5640 [ - + + - : 84 : db_path = string(getenv("HOME") ?: "/") + string("/.debuginfod.sqlite"); /* XDG? */
+ - ]
5641 : 84 : int rc = regcomp (& file_include_regex, ".*", REG_EXTENDED|REG_NOSUB); // match everything
5642 [ - + ]: 84 : if (rc != 0)
5643 : 0 : error (EXIT_FAILURE, 0, "regcomp failure: %d", rc);
5644 : 84 : rc = regcomp (& file_exclude_regex, "^$", REG_EXTENDED|REG_NOSUB); // match nothing
5645 [ - + ]: 84 : if (rc != 0)
5646 : 0 : error (EXIT_FAILURE, 0, "regcomp failure: %d", rc);
5647 : :
5648 : : // default parameters for fdcache are computed from system stats
5649 : 84 : struct statfs sfs;
5650 : 84 : rc = statfs(tmpdir.c_str(), &sfs);
5651 [ - + ]: 84 : if (rc < 0)
5652 : 0 : fdcache_mbs = 1024; // 1 gigabyte
5653 : : else
5654 : 84 : fdcache_mbs = sfs.f_bavail * sfs.f_bsize / 1024 / 1024 / 4; // 25% of free space
5655 : 84 : fdcache_mintmp = 25; // emergency flush at 25% remaining (75% full)
5656 : 84 : fdcache_prefetch = 64; // guesstimate storage is this much less costly than re-decompression
5657 : :
5658 : : /* Parse and process arguments. */
5659 : 84 : memset(&http_sockaddr, 0, sizeof(http_sockaddr));
5660 : 84 : http_sockaddr.sin6_family = AF_UNSPEC;
5661 : 84 : int remaining;
5662 : 84 : (void) argp_parse (&argp, argc, argv, ARGP_IN_ORDER, &remaining, NULL);
5663 [ - + ]: 84 : if (remaining != argc)
5664 : 0 : error (EXIT_FAILURE, 0,
5665 : 0 : "unexpected argument: %s", argv[remaining]);
5666 : :
5667 [ + + + + : 84 : if (scan_archives.size()==0 && !scan_files && source_paths.size()>0)
- + ]
5668 [ # # ]: 0 : obatched(clog) << "warning: without -F -R -U -Z, ignoring PATHs" << endl;
5669 : :
5670 : 84 : fdcache.limit(fdcache_mbs);
5671 : :
5672 : 84 : (void) signal (SIGPIPE, SIG_IGN); // microhttpd can generate it incidentally, ignore
5673 : 84 : (void) signal (SIGINT, signal_handler); // ^C
5674 : 84 : (void) signal (SIGHUP, signal_handler); // EOF
5675 : 84 : (void) signal (SIGTERM, signal_handler); // systemd
5676 : 84 : (void) signal (SIGUSR1, sigusr1_handler); // end-user
5677 : 84 : (void) signal (SIGUSR2, sigusr2_handler); // end-user
5678 : :
5679 : : /* Get database ready. */
5680 [ + + ]: 84 : if (! passive_p)
5681 : : {
5682 : 82 : rc = sqlite3_open_v2 (db_path.c_str(), &db, (SQLITE_OPEN_READWRITE
5683 : : |SQLITE_OPEN_URI
5684 : : |SQLITE_OPEN_PRIVATECACHE
5685 : : |SQLITE_OPEN_CREATE
5686 : : |SQLITE_OPEN_FULLMUTEX), /* thread-safe */
5687 : : NULL);
5688 [ - + ]: 82 : if (rc == SQLITE_CORRUPT)
5689 : : {
5690 : 0 : (void) unlink (db_path.c_str());
5691 : 0 : error (EXIT_FAILURE, 0,
5692 : : "cannot open %s, deleted database: %s", db_path.c_str(), sqlite3_errmsg(db));
5693 : : }
5694 [ - + ]: 82 : else if (rc)
5695 : : {
5696 : 0 : error (EXIT_FAILURE, 0,
5697 : : "cannot open %s, consider deleting database: %s", db_path.c_str(), sqlite3_errmsg(db));
5698 : : }
5699 : : }
5700 : :
5701 : : // open the readonly query variant
5702 : : // NB: PRIVATECACHE allows web queries to operate in parallel with
5703 : : // much other grooming/scanning operation.
5704 : 84 : rc = sqlite3_open_v2 (db_path.c_str(), &dbq, (SQLITE_OPEN_READONLY
5705 : : |SQLITE_OPEN_URI
5706 : : |SQLITE_OPEN_PRIVATECACHE
5707 : : |SQLITE_OPEN_FULLMUTEX), /* thread-safe */
5708 : : NULL);
5709 [ - + ]: 84 : if (rc)
5710 : : {
5711 : 0 : error (EXIT_FAILURE, 0,
5712 : : "cannot open %s, consider deleting database: %s", db_path.c_str(), sqlite3_errmsg(dbq));
5713 : : }
5714 : :
5715 : :
5716 [ + - ]: 168 : obatched(clog) << "opened database " << db_path
5717 [ + + + - : 86 : << (db?" rw":"") << (dbq?" ro":"") << endl;
- + + - +
- ]
5718 [ + - + - ]: 168 : obatched(clog) << "sqlite version " << sqlite3_version << endl;
5719 [ + + + - : 250 : obatched(clog) << "service mode " << (passive_p ? "passive":"active") << endl;
+ - ]
5720 : :
5721 : : // add special string-prefix-similarity function used in rpm sref/sdef resolution
5722 : 84 : rc = sqlite3_create_function(dbq, "sharedprefix", 2, SQLITE_UTF8, NULL,
5723 : : & sqlite3_sharedprefix_fn, NULL, NULL);
5724 [ - + ]: 84 : if (rc != SQLITE_OK)
5725 : 0 : error (EXIT_FAILURE, 0,
5726 : : "cannot create sharedprefix function: %s", sqlite3_errmsg(dbq));
5727 : :
5728 [ + + ]: 84 : if (! passive_p)
5729 : : {
5730 [ + + ]: 82 : if (verbose > 3)
5731 [ + - + - ]: 92 : obatched(clog) << "ddl: " << DEBUGINFOD_SQLITE_DDL << endl;
5732 : 82 : rc = sqlite3_exec (db, DEBUGINFOD_SQLITE_DDL, NULL, NULL, NULL);
5733 [ - + ]: 82 : if (rc != SQLITE_OK)
5734 : : {
5735 : 0 : error (EXIT_FAILURE, 0,
5736 : : "cannot run database schema ddl: %s", sqlite3_errmsg(db));
5737 : : }
5738 : : }
5739 : :
5740 [ + - + - : 168 : obatched(clog) << "libmicrohttpd version " << MHD_get_version() << endl;
+ - ]
5741 : :
5742 : : /* If '-C' wasn't given or was given with no arg, pick a reasonable default
5743 : : for the number of worker threads. */
5744 [ + + ]: 84 : if (connection_pool == 0)
5745 : 80 : connection_pool = default_concurrency();
5746 : :
5747 : : /* Note that MHD_USE_EPOLL and MHD_USE_THREAD_PER_CONNECTION don't
5748 : : work together. */
5749 : 84 : unsigned int use_epoll = 0;
5750 : : #if MHD_VERSION >= 0x00095100
5751 : 84 : use_epoll = MHD_USE_EPOLL;
5752 : : #endif
5753 : :
5754 : 84 : unsigned int mhd_flags = (
5755 : : #if MHD_VERSION >= 0x00095300
5756 : : MHD_USE_INTERNAL_POLLING_THREAD
5757 : : #else
5758 : : MHD_USE_SELECT_INTERNALLY
5759 : : #endif
5760 : : | MHD_USE_DUAL_STACK
5761 : : | use_epoll
5762 : : #if MHD_VERSION >= 0x00095200
5763 : : | MHD_USE_ITC
5764 : : #endif
5765 : : | MHD_USE_DEBUG); /* report errors to stderr */
5766 : :
5767 : 84 : MHD_Daemon *dsa = NULL,
5768 : 84 : *d4 = NULL,
5769 : 84 : *d46 = NULL;
5770 : :
5771 [ - + ]: 84 : if (http_sockaddr.sin6_family != AF_UNSPEC)
5772 : : {
5773 [ # # ]: 0 : if (http_sockaddr.sin6_family == AF_INET)
5774 : 0 : ((sockaddr_in*)&http_sockaddr)->sin_port = htons(http_port);
5775 [ # # ]: 0 : if (http_sockaddr.sin6_family == AF_INET6)
5776 : 0 : http_sockaddr.sin6_port = htons(http_port);
5777 : : // Start httpd server threads on socket addr:port.
5778 : 0 : dsa = MHD_start_daemon (mhd_flags & ~MHD_USE_DUAL_STACK, http_port,
5779 : : NULL, NULL, /* default accept policy */
5780 : : handler_cb, NULL, /* handler callback */
5781 : : MHD_OPTION_EXTERNAL_LOGGER,
5782 : : error_cb, NULL,
5783 : : MHD_OPTION_SOCK_ADDR,
5784 : : (struct sockaddr *) &http_sockaddr,
5785 : : MHD_OPTION_THREAD_POOL_SIZE,
5786 : : (int)connection_pool,
5787 : : MHD_OPTION_END);
5788 : : }
5789 : : else
5790 : : {
5791 : : // Start httpd server threads. Use a single dual-homed pool.
5792 : 84 : d46 = MHD_start_daemon (mhd_flags, http_port,
5793 : : NULL, NULL, /* default accept policy */
5794 : : handler_cb, NULL, /* handler callback */
5795 : : MHD_OPTION_EXTERNAL_LOGGER,
5796 : : error_cb, NULL,
5797 : : MHD_OPTION_THREAD_POOL_SIZE,
5798 : : (int)connection_pool,
5799 : : MHD_OPTION_END);
5800 : 84 : addr_info = "IPv4 IPv6";
5801 [ - + ]: 84 : if (d46 == NULL)
5802 : : {
5803 : : // Cannot use dual_stack, use ipv4 only
5804 : 0 : mhd_flags &= ~(MHD_USE_DUAL_STACK);
5805 [ # # ]: 0 : d4 = MHD_start_daemon (mhd_flags, http_port,
5806 : : NULL, NULL, /* default accept policy */
5807 : : handler_cb, NULL, /* handler callback */
5808 : : MHD_OPTION_EXTERNAL_LOGGER,
5809 : : error_cb, NULL,
5810 : : (connection_pool
5811 : : ? MHD_OPTION_THREAD_POOL_SIZE
5812 : : : MHD_OPTION_END),
5813 : : (connection_pool
5814 : : ? (int)connection_pool
5815 : : : MHD_OPTION_END),
5816 : : MHD_OPTION_END);
5817 : 0 : addr_info = "IPv4";
5818 : : }
5819 : : }
5820 [ - + - - ]: 84 : if (d4 == NULL && d46 == NULL && dsa == NULL)
5821 : : {
5822 : 0 : sqlite3 *database = db;
5823 : 0 : sqlite3 *databaseq = dbq;
5824 : 0 : db = dbq = 0; // for signal_handler not to freak
5825 : 0 : sqlite3_close (databaseq);
5826 : 0 : sqlite3_close (database);
5827 : 0 : error (EXIT_FAILURE, 0, "cannot start http server on %s port %d",
5828 : : addr_info.c_str(), http_port);
5829 : : }
5830 : :
5831 [ + - ]: 168 : obatched(clog) << "started http server on "
5832 : : << addr_info
5833 [ + - ]: 84 : << " port=" << http_port
5834 [ + - + + : 162 : << (webapi_cors ? " with cors" : "")
+ - + - ]
5835 : 84 : << endl;
5836 : :
5837 : : // add maxigroom sql if -G given
5838 [ - + ]: 84 : if (maxigroom)
5839 : : {
5840 [ # # ]: 0 : obatched(clog) << "maxigrooming database, please wait." << endl;
5841 : : // NB: this index alone can nearly double the database size!
5842 : : // NB: this index would be necessary to run source-file metadata searches fast
5843 : 0 : extra_ddl.push_back("create index if not exists " BUILDIDS "_r_sref_arc on " BUILDIDS "_r_sref(artifactsrc);");
5844 : 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);");
5845 : 0 : extra_ddl.push_back("drop index if exists " BUILDIDS "_r_sref_arc;");
5846 : :
5847 : : // NB: we don't maxigroom the _files interning table. It'd require a temp index on all the
5848 : : // tables that have file foreign-keys, which is a lot.
5849 : :
5850 : : // NB: with =delete, may take up 3x disk space total during vacuum process
5851 : : // vs. =off (only 2x but may corrupt database if program dies mid-vacuum)
5852 : : // vs. =wal (>3x observed, but safe)
5853 : 0 : extra_ddl.push_back("pragma journal_mode=delete;");
5854 : 0 : extra_ddl.push_back("vacuum;");
5855 : 0 : extra_ddl.push_back("pragma journal_mode=wal;");
5856 : : }
5857 : :
5858 : : // run extra -D sql if given
5859 [ + + ]: 84 : if (! passive_p)
5860 [ - + ]: 82 : for (auto&& i: extra_ddl)
5861 : : {
5862 [ # # ]: 0 : if (verbose > 1)
5863 [ # # # # ]: 0 : obatched(clog) << "extra ddl:\n" << i << endl;
5864 : 0 : rc = sqlite3_exec (db, i.c_str(), NULL, NULL, NULL);
5865 [ # # # # ]: 0 : if (rc != SQLITE_OK && rc != SQLITE_DONE && rc != SQLITE_ROW)
5866 : 0 : error (0, 0,
5867 : : "warning: cannot run database extra ddl %s: %s", i.c_str(), sqlite3_errmsg(db));
5868 : :
5869 [ # # ]: 0 : if (maxigroom)
5870 [ # # ]: 0 : obatched(clog) << "maxigroomed database" << endl;
5871 : : }
5872 : :
5873 [ + + ]: 84 : if (! passive_p)
5874 [ + - + - ]: 164 : obatched(clog) << "search concurrency " << concurrency << endl;
5875 : 84 : obatched(clog) << "webapi connection pool " << connection_pool
5876 [ + - - + : 84 : << (connection_pool ? "" : " (unlimited)") << endl;
+ - + - ]
5877 [ + + ]: 84 : if (! passive_p) {
5878 [ + - + - ]: 164 : obatched(clog) << "rescan time " << rescan_s << endl;
5879 [ + - + - ]: 164 : obatched(clog) << "scan checkpoint " << scan_checkpoint << endl;
5880 : : }
5881 [ + - + - ]: 168 : obatched(clog) << "fdcache mbs " << fdcache_mbs << endl;
5882 [ + - + - ]: 168 : obatched(clog) << "fdcache prefetch " << fdcache_prefetch << endl;
5883 [ + - + - ]: 168 : obatched(clog) << "fdcache tmpdir " << tmpdir << endl;
5884 [ + - + - ]: 168 : obatched(clog) << "fdcache tmpdir min% " << fdcache_mintmp << endl;
5885 [ + + ]: 84 : if (! passive_p)
5886 [ + - + - ]: 164 : obatched(clog) << "groom time " << groom_s << endl;
5887 [ + - + - ]: 168 : obatched(clog) << "forwarded ttl limit " << forwarded_ttl_limit << endl;
5888 : :
5889 [ + + ]: 84 : if (scan_archives.size()>0)
5890 : : {
5891 : 56 : obatched ob(clog);
5892 [ + - ]: 56 : auto& o = ob << "accepting archive types ";
5893 [ + + ]: 174 : for (auto&& arch : scan_archives)
5894 [ + - + - : 118 : o << arch.first << "(" << arch.second << ") ";
+ - + - ]
5895 [ + - ]: 56 : o << endl;
5896 : 56 : }
5897 : 84 : const char* du = getenv(DEBUGINFOD_URLS_ENV_VAR);
5898 [ + + + + ]: 84 : if (du && du[0] != '\0') // set to non-empty string?
5899 [ + - + - ]: 32 : obatched(clog) << "upstream debuginfod servers: " << du << endl;
5900 : :
5901 [ + + ]: 84 : vector<pthread_t> all_threads;
5902 : :
5903 [ + + ]: 84 : if (! passive_p)
5904 : : {
5905 : 82 : pthread_t pt;
5906 : 82 : rc = pthread_create (& pt, NULL, thread_main_groom, NULL);
5907 [ - + ]: 82 : if (rc)
5908 : 0 : error (EXIT_FAILURE, rc, "cannot spawn thread to groom database\n");
5909 : : else
5910 : : {
5911 : : #ifdef HAVE_PTHREAD_SETNAME_NP
5912 : 82 : (void) pthread_setname_np (pt, "groom");
5913 : : #endif
5914 [ + - ]: 82 : all_threads.push_back(pt);
5915 : : }
5916 : :
5917 [ + + + + ]: 82 : if (scan_files || scan_archives.size() > 0)
5918 : : {
5919 [ + - ]: 72 : if (scan_checkpoint > 0)
5920 [ + - ]: 72 : scan_barrier = new sqlite_checkpoint_pb(concurrency, (unsigned) scan_checkpoint);
5921 : :
5922 : 72 : rc = pthread_create (& pt, NULL, thread_main_fts_source_paths, NULL);
5923 [ - + ]: 72 : if (rc)
5924 : 0 : error (EXIT_FAILURE, rc, "cannot spawn thread to traverse source paths\n");
5925 : : #ifdef HAVE_PTHREAD_SETNAME_NP
5926 : 72 : (void) pthread_setname_np (pt, "traverse");
5927 : : #endif
5928 [ + - ]: 72 : all_threads.push_back(pt);
5929 : :
5930 [ + + ]: 360 : for (unsigned i=0; i<concurrency; i++)
5931 : : {
5932 : 288 : rc = pthread_create (& pt, NULL, thread_main_scanner, NULL);
5933 [ - + ]: 288 : if (rc)
5934 : 0 : error (EXIT_FAILURE, rc, "cannot spawn thread to scan source files / archives\n");
5935 : : #ifdef HAVE_PTHREAD_SETNAME_NP
5936 : 288 : (void) pthread_setname_np (pt, "scan");
5937 : : #endif
5938 [ + - ]: 288 : all_threads.push_back(pt);
5939 : : }
5940 : : }
5941 : : }
5942 : :
5943 : : /* Trivial main loop! */
5944 [ + - + - ]: 84 : set_metric("ready", 1);
5945 [ + + ]: 232 : while (! interrupted)
5946 [ + - ]: 148 : pause ();
5947 [ + - ]: 84 : scanq.nuke(); // wake up any remaining scanq-related threads, let them die
5948 [ + + + - ]: 84 : if (scan_barrier) scan_barrier->nuke(); // ... in case they're stuck in a barrier
5949 [ + - + - ]: 84 : set_metric("ready", 0);
5950 : :
5951 [ + - ]: 84 : if (verbose)
5952 [ + - + - : 168 : obatched(clog) << "stopping" << endl;
- - ]
5953 : :
5954 : : /* Join all our threads. */
5955 [ + + ]: 526 : for (auto&& it : all_threads)
5956 [ + - ]: 442 : pthread_join (it, NULL);
5957 : :
5958 : : /* Stop all the web service threads. */
5959 [ - + - - ]: 84 : if (dsa) MHD_stop_daemon (dsa);
5960 [ + - + - ]: 84 : if (d46) MHD_stop_daemon (d46);
5961 [ - + - - ]: 84 : if (d4) MHD_stop_daemon (d4);
5962 : :
5963 [ + + ]: 84 : if (! passive_p)
5964 : : {
5965 : : /* With all threads known dead, we can clean up the global resources. */
5966 [ + - ]: 82 : rc = sqlite3_exec (db, DEBUGINFOD_SQLITE_CLEANUP_DDL, NULL, NULL, NULL);
5967 [ - + ]: 82 : if (rc != SQLITE_OK)
5968 : : {
5969 [ # # ]: 0 : error (0, 0,
5970 : : "warning: cannot run database cleanup ddl: %s", sqlite3_errmsg(db));
5971 : : }
5972 : : }
5973 : :
5974 [ + - ]: 84 : debuginfod_pool_groom ();
5975 [ + + ]: 84 : delete scan_barrier;
5976 : :
5977 : : // NB: no problem with unconditional free here - an earlier failed regcomp would exit program
5978 [ + - ]: 84 : (void) regfree (& file_include_regex);
5979 [ + - ]: 84 : (void) regfree (& file_exclude_regex);
5980 : :
5981 : 84 : sqlite3 *database = db;
5982 : 84 : sqlite3 *databaseq = dbq;
5983 : 84 : db = dbq = 0; // for signal_handler not to freak
5984 [ + - ]: 84 : (void) sqlite3_close (databaseq);
5985 [ + + ]: 84 : if (! passive_p)
5986 [ + - ]: 82 : (void) sqlite3_close (database);
5987 : :
5988 [ + + ]: 84 : return 0;
5989 : 84 : }
|
