3 * tailing reliable realtime streaming feeder for inn
4 * infile.c - monitoring and handling of input files
6 * Copyright (C) 2010 Ian Jackson <ijackson@chiark.greenend.org.uk>
8 * This program is free software: you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation, either version 3 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program. If not, see <http://www.gnu.org/licenses/>.
21 * (I believe that when you compile and link this as part of the inn2
22 * build, with the Makefile runes I have provided, all the libraries
23 * and files which end up included in innduct are licence-compatible
24 * with GPLv3. If not then please let me know. -Ian Jackson.)
29 /*========== monitoring of input files ==========*/
31 static void feedfile_eof(InputFile *ipf) {
32 assert(ipf != main_input_file); /* promised by tailing_try_read */
33 inputfile_reading_stop(ipf);
35 if (ipf == flushing_input_file) {
36 assert(sms==sm_SEPARATED || sms==sm_DROPPING);
37 if (main_input_file) inputfile_reading_start(main_input_file);
38 statemc_check_flushing_done();
39 } else if (ipf == backlog_input_file) {
40 statemc_check_backlog_done();
42 abort(); /* supposed to wait rather than get EOF on main input file */
46 InputFile *open_input_file(const char *path) {
47 int fd= open(path, O_RDWR);
49 if (errno==ENOENT) return 0;
50 sysdie("unable to open input file %s", path);
54 InputFile *ipf= xmalloc(sizeof(*ipf) + strlen(path) + 1);
55 memset(ipf,0,sizeof(*ipf));
59 LIST_INIT(ipf->queue);
60 strcpy(ipf->path, path);
65 void close_input_file(InputFile *ipf) { /* does not free */
66 assert(!ipf->readable_callback); /* must have had ->on_cancel */
67 assert(!ipf->filemon); /* must have had inputfile_reading_stop */
68 assert(!ipf->rd); /* must have had inputfile_reading_stop */
69 assert(!ipf->inprogress); /* no dangling pointers pointing here */
70 xclose_perhaps(&ipf->fd, "input file ", ipf->path);
74 /*---------- dealing with articles read in the input file ----------*/
76 static void *feedfile_got_bad_data(InputFile *ipf, off_t offset,
77 const char *data, const char *how) {
78 warn("corrupted file: %s, offset %lu: %s: in %s",
79 ipf->path, (unsigned long)offset, how, sanitise(data,-1));
80 ipf->counts.events[read_err]++;
81 if (ipf->counts.events[read_err] > max_bad_data_initial +
82 (ipf->counts.events[read_ok] + ipf->counts.events[read_blank])
84 crash("too much garbage in input file! (%d errs, %d ok, %d blank)",
85 ipf->counts.events[read_err], ipf->counts.events[read_ok],
86 ipf->counts.events[read_blank]);
90 static void *feedfile_read_err(oop_source *lp, oop_read *rd,
91 oop_rd_event ev, const char *errmsg,
92 int errnoval, const char *data, size_t recsz,
94 InputFile *ipf= ipf_v;
95 assert(ev == OOP_RD_SYSTEM);
97 syscrash("error reading input file: %s, offset %lu",
98 ipf->path, (unsigned long)ipf->offset);
101 static void *feedfile_got_article(oop_source *lp, oop_read *rd,
102 oop_rd_event ev, const char *errmsg,
103 int errnoval, const char *data, size_t recsz,
105 InputFile *ipf= ipf_v;
107 char tokentextbuf[sizeof(TOKEN)*2+3];
109 if (!data) { feedfile_eof(ipf); return OOP_CONTINUE; }
111 off_t old_offset= ipf->offset;
112 ipf->offset += recsz + !!(ev == OOP_RD_OK);
114 #define X_BAD_DATA(m) return feedfile_got_bad_data(ipf,old_offset,data,m);
116 if (ev==OOP_RD_PARTREC)
117 feedfile_got_bad_data(ipf,old_offset,data,"missing final newline");
118 /* but process it anyway */
120 if (ipf->skippinglong) {
121 if (ev==OOP_RD_OK) ipf->skippinglong= 0; /* fine now */
124 if (ev==OOP_RD_LONG) {
125 ipf->skippinglong= 1;
126 X_BAD_DATA("overly long line");
129 if (memchr(data,'\0',recsz)) X_BAD_DATA("nul byte");
130 if (!recsz) X_BAD_DATA("empty line");
133 if (strspn(data," ") != recsz) X_BAD_DATA("line partially blanked");
134 ipf->counts.events[read_blank]++;
138 char *space= strchr(data,' ');
139 int tokenlen= space-data;
140 int midlen= (int)recsz-tokenlen-1;
141 if (midlen <= 2) X_BAD_DATA("no room for messageid");
142 if (space[1]!='<' || space[midlen]!='>') X_BAD_DATA("invalid messageid");
144 if (tokenlen != sizeof(TOKEN)*2+2) X_BAD_DATA("token wrong length");
145 memcpy(tokentextbuf, data, tokenlen);
146 tokentextbuf[tokenlen]= 0;
147 if (!IsToken(tokentextbuf)) X_BAD_DATA("token wrong syntax");
149 ipf->counts.events[read_ok]++;
151 art= xmalloc(sizeof(*art) - 1 + midlen + 1);
152 memset(art,0,sizeof(*art));
153 art->state= art_Unchecked;
155 art->ipf= ipf; ipf->inprogress++;
156 art->token= TextToToken(tokentextbuf);
157 art->offset= old_offset;
158 art->blanklen= recsz;
159 strcpy(art->messageid, space+1);
161 if (ipf->autodefer >= 0) {
162 article_autodefer(ipf, art);
164 LIST_ADDTAIL(ipf->queue, art);
166 if (ipf==backlog_input_file)
167 article_check_expired(art);
170 if (sms==sm_NORMAL && ipf==main_input_file &&
171 ipf->offset >= target_max_feedfile_size)
172 statemc_start_flush("feed file size");
174 check_assign_articles(); /* may destroy conn but that's OK */
175 check_reading_pause_resume(ipf);
179 /*========== tailing input file ==========*/
181 static void *tailing_rable_call_time(oop_source *lp, struct timeval tv,
183 /* lifetime of ipf here is OK because destruction will cause
184 * on_cancel which will cancel this callback */
185 InputFile *ipf= user;
187 dbg("**TRACT** ipf=%p called",ipf);
188 if (!ipf->fake_readable) return OOP_CONTINUE;
190 /* we just keep calling readable until our caller (oop_rd)
191 * has called try_read, and try_read has found EOF so given EAGAIN */
192 dbg("**TRACT** ipf=%p reschedule",ipf);
193 loop->on_time(loop, OOP_TIME_NOW, tailing_rable_call_time, ipf);
195 return ipf->readable_callback(loop, &ipf->readable,
196 ipf->readable_callback_user);
199 static void tailing_on_cancel(struct oop_readable *rable) {
200 InputFile *ipf= (void*)rable;
201 dbg("**TOR** ipf=%p on_cancel",ipf);
203 if (ipf->filemon) filemon_stop(ipf);
204 dbg("**TRACT** ipf=%p cancel",ipf);
205 loop->cancel_time(loop, OOP_TIME_NOW, tailing_rable_call_time, ipf);
206 ipf->readable_callback= 0;
209 void tailing_make_readable(InputFile *ipf) {
210 dbg("**TRACT** ipf=%p makereadable rcb=%p",ipf,
211 (void*)ipf?ipf->readable_callback:0);
212 if (!ipf || !ipf->readable_callback) /* so callers can be naive */
214 ipf->fake_readable= 1;
215 loop->on_time(loop, OOP_TIME_NOW, tailing_rable_call_time, ipf);
218 static int tailing_on_readable(struct oop_readable *rable,
219 oop_readable_call *cb, void *user) {
220 InputFile *ipf= (void*)rable;
221 dbg("**TOR** ipf=%p on_readable",ipf);
223 tailing_on_cancel(rable);
224 ipf->readable_callback= cb;
225 ipf->readable_callback_user= user;
227 tailing_make_readable(ipf);
231 static ssize_t tailing_try_read(struct oop_readable *rable, void *buffer,
233 InputFile *ipf= (void*)rable;
235 ssize_t r= read(ipf->fd, buffer, length);
237 if (errno==EINTR) continue;
238 ipf->fake_readable= 0;
242 if (ipf==main_input_file) {
244 ipf->fake_readable= 0;
246 } else if (ipf==flushing_input_file) {
248 assert(sms==sm_SEPARATED || sms==sm_DROPPING);
249 } else if (ipf==backlog_input_file) {
255 dbg("**TOR** ipf=%p try_read r=%d",ipf,r);
260 /*---------- interface to start and stop an input file ----------*/
262 static const oop_rd_style feedfile_rdstyle= {
263 OOP_RD_DELIM_STRIP, '\n',
265 OOP_RD_SHORTREC_LONG,
268 void inputfile_reading_resume(InputFile *ipf) {
269 if (!ipf->rd) return;
270 if (!ipf->paused) return;
272 int r= oop_rd_read(ipf->rd, &feedfile_rdstyle, MAX_LINE_FEEDFILE,
273 feedfile_got_article,ipf, feedfile_read_err, ipf);
274 if (r) syscrash("unable start reading feedfile %s",ipf->path);
279 void inputfile_reading_pause(InputFile *ipf) {
280 if (!ipf->rd) return;
281 if (ipf->paused) return;
282 oop_rd_cancel(ipf->rd);
286 void inputfile_reading_start(InputFile *ipf) {
288 ipf->readable.on_readable= tailing_on_readable;
289 ipf->readable.on_cancel= tailing_on_cancel;
290 ipf->readable.try_read= tailing_try_read;
291 ipf->readable.delete_tidy= 0; /* we never call oop_rd_delete_{tidy,kill} */
292 ipf->readable.delete_kill= 0;
294 ipf->readable_callback= 0;
295 ipf->readable_callback_user= 0;
297 ipf->rd= oop_rd_new(loop, &ipf->readable, 0,0);
301 inputfile_reading_resume(ipf);
304 void inputfile_reading_stop(InputFile *ipf) {
306 inputfile_reading_pause(ipf);
307 oop_rd_delete(ipf->rd);
309 assert(!ipf->filemon); /* we shouldn't be monitoring it now */
312 void filepoll(void) {
313 tailing_make_readable(main_input_file);
314 tailing_make_readable(flushing_input_file);
317 char *dbg_report_ipf(InputFile *ipf) {
318 if (!ipf) return xasprintf("none");
320 const char *slash= strrchr(ipf->path,'/');
321 const char *path= slash ? slash+1 : ipf->path;
323 return xasprintf("%p/%s:queue=%d,ip=%ld,autodef=%ld,off=%ld,fd=%d%s%s%s",
325 ipf->queue.count, ipf->inprogress, ipf->autodefer,
326 (long)ipf->offset, ipf->fd,
327 ipf->rd ? "" : ",!rd",
328 ipf->skippinglong ? "*skiplong" : "",
329 ipf->rd && ipf->paused ? "*paused" : "");