1/* Licensed to the Apache Software Foundation (ASF) under one or more
2 * contributor license agreements.  See the NOTICE file distributed with
3 * this work for additional information regarding copyright ownership.
4 * The ASF licenses this file to You under the Apache License, Version 2.0
5 * (the "License"); you may not use this file except in compliance with
6 * the License.  You may obtain a copy of the License at
7 *
8 *     http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17/*
18 * mod_expires.c
19 * version 0.0.11
20 * status beta
21 *
22 * Andrew Wilson <Andrew.Wilson@cm.cf.ac.uk> 26.Jan.96
23 *
24 * This module allows you to control the form of the Expires: header
25 * that Apache issues for each access.  Directives can appear in
26 * configuration files or in .htaccess files so expiry semantics can
27 * be defined on a per-directory basis.
28 *
29 * DIRECTIVE SYNTAX
30 *
31 * Valid directives are:
32 *
33 *     ExpiresActive on | off
34 *     ExpiresDefault <code><seconds>
35 *     ExpiresByType type/encoding <code><seconds>
36 *
37 * Valid values for <code> are:
38 *
39 *     'M'      expires header shows file modification date + <seconds>
40 *     'A'      expires header shows access time + <seconds>
41 *
42 *              [I'm not sure which of these is best under different
43 *              circumstances, I guess it's for other people to explore.
44 *              The effects may be indistinguishable for a number of cases]
45 *
46 * <seconds> should be an integer value [acceptable to atoi()]
47 *
48 * There is NO space between the <code> and <seconds>.
49 *
50 * For example, a directory which contains information which changes
51 * frequently might contain:
52 *
53 *     # reports generated by cron every hour.  don't let caches
54 *     # hold onto stale information
55 *     ExpiresDefault M3600
56 *
57 * Another example, our html pages can change all the time, the gifs
58 * tend not to change often:
59 *
60 *     # pages are hot (1 week), images are cold (1 month)
61 *     ExpiresByType text/html A604800
62 *     ExpiresByType image/gif A2592000
63 *
64 * Expires can be turned on for all URLs on the server by placing the
65 * following directive in a conf file:
66 *
67 *     ExpiresActive on
68 *
69 * ExpiresActive can also appear in .htaccess files, enabling the
70 * behaviour to be turned on or off for each chosen directory.
71 *
72 *     # turn off Expires behaviour in this directory
73 *     # and subdirectories
74 *     ExpiresActive off
75 *
76 * Directives defined for a directory are valid in subdirectories
77 * unless explicitly overridden by new directives in the subdirectory
78 * .htaccess files.
79 *
80 * ALTERNATIVE DIRECTIVE SYNTAX
81 *
82 * Directives can also be defined in a more readable syntax of the form:
83 *
84 *     ExpiresDefault "<base> [plus] {<num> <type>}*"
85 *     ExpiresByType type/encoding "<base> [plus] {<num> <type>}*"
86 *
87 * where <base> is one of:
88 *      access
89 *      now             equivalent to 'access'
90 *      modification
91 *
92 * where the 'plus' keyword is optional
93 *
94 * where <num> should be an integer value [acceptable to atoi()]
95 *
96 * where <type> is one of:
97 *      years
98 *      months
99 *      weeks
100 *      days
101 *      hours
102 *      minutes
103 *      seconds
104 *
105 * For example, any of the following directives can be used to make
106 * documents expire 1 month after being accessed, by default:
107 *
108 *      ExpiresDefault "access plus 1 month"
109 *      ExpiresDefault "access plus 4 weeks"
110 *      ExpiresDefault "access plus 30 days"
111 *
112 * The expiry time can be fine-tuned by adding several '<num> <type>'
113 * clauses:
114 *
115 *      ExpiresByType text/html "access plus 1 month 15 days 2 hours"
116 *      ExpiresByType image/gif "modification plus 5 hours 3 minutes"
117 *
118 * ---
119 *
120 * Change-log:
121 * 29.Jan.96    Hardened the add_* functions.  Server will now bail out
122 *              if bad directives are given in the conf files.
123 * 02.Feb.96    Returns DECLINED if not 'ExpiresActive on', giving other
124 *              expires-aware modules a chance to play with the same
125 *              directives. [Michael Rutman]
126 * 03.Feb.96    Call tzset() before localtime().  Trying to get the module
127 *              to work properly in non GMT timezones.
128 * 12.Feb.96    Modified directive syntax to allow more readable commands:
129 *                ExpiresDefault "now plus 10 days 20 seconds"
130 *                ExpiresDefault "access plus 30 days"
131 *                ExpiresDefault "modification plus 1 year 10 months 30 days"
132 * 13.Feb.96    Fix call to table_get() with NULL 2nd parameter [Rob Hartill]
133 * 19.Feb.96    Call gm_timestr_822() to get time formatted correctly, can't
134 *              rely on presence of HTTP_TIME_FORMAT in Apache 1.1+.
135 * 21.Feb.96    This version (0.0.9) reverses assumptions made in 0.0.8
136 *              about star/star handlers.  Reverting to 0.0.7 behaviour.
137 * 08.Jun.96    allows ExpiresDefault to be used with responses that use
138 *              the DefaultType by not DECLINING, but instead skipping
139 *              the table_get check and then looking for an ExpiresDefault.
140 *              [Rob Hartill]
141 * 04.Nov.96    'const' definitions added.
142 *
143 * TODO
144 * add support for Cache-Control: max-age=20 from the HTTP/1.1
145 * proposal (in this case, a ttl of 20 seconds) [ask roy]
146 * add per-file expiry and explicit expiry times - duplicates some
147 * of the mod_cern_meta.c functionality.  eg:
148 *              ExpiresExplicit index.html "modification plus 30 days"
149 *
150 * BUGS
151 * Hi, welcome to the internet.
152 */
153
154#include "apr.h"
155#include "apr_strings.h"
156#include "apr_lib.h"
157
158#define APR_WANT_STRFUNC
159#include "apr_want.h"
160
161#include "ap_config.h"
162#include "httpd.h"
163#include "http_config.h"
164#include "http_log.h"
165#include "http_request.h"
166#include "http_protocol.h"
167
168typedef struct {
169    int active;
170    int wildcards;
171    char *expiresdefault;
172    apr_table_t *expiresbytype;
173} expires_dir_config;
174
175/* from mod_dir, why is this alias used?
176 */
177#define DIR_CMD_PERMS OR_INDEXES
178
179#define ACTIVE_ON       1
180#define ACTIVE_OFF      0
181#define ACTIVE_DONTCARE 2
182
183module AP_MODULE_DECLARE_DATA expires_module;
184
185static void *create_dir_expires_config(apr_pool_t *p, char *dummy)
186{
187    expires_dir_config *new =
188    (expires_dir_config *) apr_pcalloc(p, sizeof(expires_dir_config));
189    new->active = ACTIVE_DONTCARE;
190    new->wildcards = 0;
191    new->expiresdefault = NULL;
192    new->expiresbytype = apr_table_make(p, 4);
193    return (void *) new;
194}
195
196static const char *set_expiresactive(cmd_parms *cmd, void *in_dir_config, int arg)
197{
198    expires_dir_config *dir_config = in_dir_config;
199
200    /* if we're here at all it's because someone explicitly
201     * set the active flag
202     */
203    dir_config->active = ACTIVE_ON;
204    if (arg == 0) {
205        dir_config->active = ACTIVE_OFF;
206    }
207    return NULL;
208}
209
210/* check_code() parse 'code' and return NULL or an error response
211 * string.  If we return NULL then real_code contains code converted
212 * to the cnnnn format.
213 */
214static char *check_code(apr_pool_t *p, const char *code, char **real_code)
215{
216    char *word;
217    char base = 'X';
218    int modifier = 0;
219    int num = 0;
220    int factor;
221
222    /* 0.0.4 compatibility?
223     */
224    if ((code[0] == 'A') || (code[0] == 'M')) {
225        *real_code = (char *)code;
226        return NULL;
227    }
228
229    /* <base> [plus] {<num> <type>}*
230     */
231
232    /* <base>
233     */
234    word = ap_getword_conf(p, &code);
235    if (!strncasecmp(word, "now", 1) ||
236        !strncasecmp(word, "access", 1)) {
237        base = 'A';
238    }
239    else if (!strncasecmp(word, "modification", 1)) {
240        base = 'M';
241    }
242    else {
243        return apr_pstrcat(p, "bad expires code, unrecognised <base> '",
244                       word, "'", NULL);
245    }
246
247    /* [plus]
248     */
249    word = ap_getword_conf(p, &code);
250    if (!strncasecmp(word, "plus", 1)) {
251        word = ap_getword_conf(p, &code);
252    }
253
254    /* {<num> <type>}*
255     */
256    while (word[0]) {
257        /* <num>
258         */
259        if (apr_isdigit(word[0])) {
260            num = atoi(word);
261        }
262        else {
263            return apr_pstrcat(p, "bad expires code, numeric value expected <num> '",
264                           word, "'", NULL);
265        }
266
267        /* <type>
268         */
269        word = ap_getword_conf(p, &code);
270        if (word[0] == '\0') {
271            return apr_pstrcat(p, "bad expires code, missing <type>", NULL);
272        }
273
274        if (!strncasecmp(word, "years", 1)) {
275            factor = 60 * 60 * 24 * 365;
276        }
277        else if (!strncasecmp(word, "months", 2)) {
278            factor = 60 * 60 * 24 * 30;
279        }
280        else if (!strncasecmp(word, "weeks", 1)) {
281            factor = 60 * 60 * 24 * 7;
282        }
283        else if (!strncasecmp(word, "days", 1)) {
284            factor = 60 * 60 * 24;
285        }
286        else if (!strncasecmp(word, "hours", 1)) {
287            factor = 60 * 60;
288        }
289        else if (!strncasecmp(word, "minutes", 2)) {
290            factor = 60;
291        }
292        else if (!strncasecmp(word, "seconds", 1)) {
293            factor = 1;
294        }
295        else {
296            return apr_pstrcat(p, "bad expires code, unrecognised <type>",
297                           "'", word, "'", NULL);
298        }
299
300        modifier = modifier + factor * num;
301
302        /* next <num>
303         */
304        word = ap_getword_conf(p, &code);
305    }
306
307    *real_code = apr_psprintf(p, "%c%d", base, modifier);
308
309    return NULL;
310}
311
312static const char *set_expiresbytype(cmd_parms *cmd, void *in_dir_config,
313                                     const char *mime, const char *code)
314{
315    expires_dir_config *dir_config = in_dir_config;
316    char *response, *real_code;
317    const char *check;
318
319    check = ap_strrchr_c(mime, '/');
320    if (check == NULL) {
321        return "Invalid mimetype: should contain a slash";
322    }
323    if ((strlen(++check) == 1) && (*check == '*')) {
324        dir_config->wildcards = 1;
325    }
326
327    if ((response = check_code(cmd->pool, code, &real_code)) == NULL) {
328        apr_table_setn(dir_config->expiresbytype, mime, real_code);
329        return NULL;
330    }
331    return apr_pstrcat(cmd->pool,
332                 "'ExpiresByType ", mime, " ", code, "': ", response, NULL);
333}
334
335static const char *set_expiresdefault(cmd_parms *cmd, void *in_dir_config,
336                                      const char *code)
337{
338    expires_dir_config * dir_config = in_dir_config;
339    char *response, *real_code;
340
341    if ((response = check_code(cmd->pool, code, &real_code)) == NULL) {
342        dir_config->expiresdefault = real_code;
343        return NULL;
344    }
345    return apr_pstrcat(cmd->pool,
346                   "'ExpiresDefault ", code, "': ", response, NULL);
347}
348
349static const command_rec expires_cmds[] =
350{
351    AP_INIT_FLAG("ExpiresActive", set_expiresactive, NULL, DIR_CMD_PERMS,
352                 "Limited to 'on' or 'off'"),
353    AP_INIT_TAKE2("ExpiresByType", set_expiresbytype, NULL, DIR_CMD_PERMS,
354                  "a MIME type followed by an expiry date code"),
355    AP_INIT_TAKE1("ExpiresDefault", set_expiresdefault, NULL, DIR_CMD_PERMS,
356                  "an expiry date code"),
357    {NULL}
358};
359
360static void *merge_expires_dir_configs(apr_pool_t *p, void *basev, void *addv)
361{
362    expires_dir_config *new = (expires_dir_config *) apr_pcalloc(p, sizeof(expires_dir_config));
363    expires_dir_config *base = (expires_dir_config *) basev;
364    expires_dir_config *add = (expires_dir_config *) addv;
365
366    if (add->active == ACTIVE_DONTCARE) {
367        new->active = base->active;
368    }
369    else {
370        new->active = add->active;
371    }
372
373    if (add->expiresdefault != NULL) {
374        new->expiresdefault = add->expiresdefault;
375    }
376    else {
377        new->expiresdefault = base->expiresdefault;
378    }
379    new->wildcards = add->wildcards;
380    new->expiresbytype = apr_table_overlay(p, add->expiresbytype,
381                                        base->expiresbytype);
382    return new;
383}
384
385/*
386 * Handle the setting of the expiration response header fields according
387 * to our criteria.
388 */
389
390static int set_expiration_fields(request_rec *r, const char *code,
391                                 apr_table_t *t)
392{
393    apr_time_t base;
394    apr_time_t additional;
395    apr_time_t expires;
396    int additional_sec;
397    char *timestr;
398
399    switch (code[0]) {
400    case 'M':
401        if (r->finfo.filetype == APR_NOFILE) {
402            /* file doesn't exist on disk, so we can't do anything based on
403             * modification time.  Note that this does _not_ log an error.
404             */
405            return DECLINED;
406        }
407        base = r->finfo.mtime;
408        additional_sec = atoi(&code[1]);
409        additional = apr_time_from_sec(additional_sec);
410        break;
411    case 'A':
412        /* there's been some discussion and it's possible that
413         * 'access time' will be stored in request structure
414         */
415        base = r->request_time;
416        additional_sec = atoi(&code[1]);
417        additional = apr_time_from_sec(additional_sec);
418        break;
419    default:
420        /* expecting the add_* routines to be case-hardened this
421         * is just a reminder that module is beta
422         */
423        ap_log_rerror(APLOG_MARK, APLOG_ERR, 0, r, APLOGNO(01500)
424                    "internal error: bad expires code: %s", r->filename);
425        return HTTP_INTERNAL_SERVER_ERROR;
426    }
427
428    expires = base + additional;
429    if (expires < r->request_time) {
430        expires = r->request_time;
431    }
432    apr_table_mergen(t, "Cache-Control",
433                     apr_psprintf(r->pool, "max-age=%" APR_TIME_T_FMT,
434                                  apr_time_sec(expires - r->request_time)));
435    timestr = apr_palloc(r->pool, APR_RFC822_DATE_LEN);
436    apr_rfc822_date(timestr, expires);
437    apr_table_setn(t, "Expires", timestr);
438    return OK;
439}
440
441/*
442 * Output filter to set the Expires response header field
443 * according to the content-type of the response -- if it hasn't
444 * already been set.
445 */
446static apr_status_t expires_filter(ap_filter_t *f,
447                                   apr_bucket_brigade *b)
448{
449    request_rec *r;
450    expires_dir_config *conf;
451    const char *expiry;
452    apr_table_t *t;
453
454    r = f->r;
455    conf = (expires_dir_config *) ap_get_module_config(r->per_dir_config,
456                                                       &expires_module);
457
458    /*
459     * Check to see which output header table we should use;
460     * mod_cgi loads script fields into r->err_headers_out,
461     * for instance.
462     */
463    expiry = apr_table_get(r->err_headers_out, "Expires");
464    if (expiry != NULL) {
465        t = r->err_headers_out;
466    }
467    else {
468        expiry = apr_table_get(r->headers_out, "Expires");
469        t = r->headers_out;
470    }
471    if (expiry == NULL) {
472        /*
473         * No expiration has been set, so we can apply any managed by
474         * this module.  First, check to see if there is an applicable
475         * ExpiresByType directive.
476         */
477        expiry = apr_table_get(conf->expiresbytype,
478                               ap_field_noparam(r->pool, r->content_type));
479        if (expiry == NULL) {
480            int usedefault = 1;
481            /*
482             * See if we have a wildcard entry for the major type.
483             */
484            if (conf->wildcards) {
485                char *checkmime;
486                char *spos;
487                checkmime = apr_pstrdup(r->pool, r->content_type);
488                spos = checkmime ? ap_strchr(checkmime, '/') : NULL;
489                if (spos != NULL) {
490                    /*
491                     * Without a '/' character, nothing we have will match.
492                     * However, we have one.
493                     */
494                    if (strlen(++spos) > 0) {
495                        *spos++ = '*';
496                        *spos = '\0';
497                    }
498                    else {
499                        checkmime = apr_pstrcat(r->pool, checkmime, "*", NULL);
500                    }
501                    expiry = apr_table_get(conf->expiresbytype, checkmime);
502                    usedefault = (expiry == NULL);
503                }
504            }
505            if (usedefault) {
506                /*
507                 * Use the ExpiresDefault directive
508                 */
509                expiry = conf->expiresdefault;
510            }
511        }
512        if (expiry != NULL) {
513            set_expiration_fields(r, expiry, t);
514        }
515    }
516    ap_remove_output_filter(f);
517    return ap_pass_brigade(f->next, b);
518}
519
520static void expires_insert_filter(request_rec *r)
521{
522    expires_dir_config *conf;
523
524    /* Don't add Expires headers to errors */
525    if (ap_is_HTTP_ERROR(r->status)) {
526        return;
527    }
528    /* Say no to subrequests */
529    if (r->main != NULL) {
530        return;
531    }
532    conf = (expires_dir_config *) ap_get_module_config(r->per_dir_config,
533                                                       &expires_module);
534
535    /* Check to see if the filter is enabled and if there are any applicable
536     * config directives for this directory scope
537     */
538    if (conf->active != ACTIVE_ON ||
539        (apr_is_empty_table(conf->expiresbytype) && !conf->expiresdefault)) {
540        return;
541    }
542    ap_add_output_filter("MOD_EXPIRES", NULL, r, r->connection);
543    return;
544}
545static void register_hooks(apr_pool_t *p)
546{
547    /* mod_expires needs to run *before* the cache save filter which is
548     * AP_FTYPE_CONTENT_SET-1.  Otherwise, our expires won't be honored.
549     */
550    ap_register_output_filter("MOD_EXPIRES", expires_filter, NULL,
551                              AP_FTYPE_CONTENT_SET-2);
552    ap_hook_insert_error_filter(expires_insert_filter, NULL, NULL, APR_HOOK_MIDDLE);
553    ap_hook_insert_filter(expires_insert_filter, NULL, NULL, APR_HOOK_MIDDLE);
554}
555
556AP_DECLARE_MODULE(expires) =
557{
558    STANDARD20_MODULE_STUFF,
559    create_dir_expires_config,  /* dir config creater */
560    merge_expires_dir_configs,  /* dir merger --- default is to override */
561    NULL,                       /* server config */
562    NULL,                       /* merge server configs */
563    expires_cmds,               /* command apr_table_t */
564    register_hooks              /* register hooks */
565};
566