Annotation of libwww/Library/src/HTFilter.c, revision 2.14
2.1 frystyk 1: /*
2: ** BEFORE AND AFTER FILTERS
3: **
4: ** (c) COPYRIGHT MIT 1995.
5: ** Please first read the full copyright statement in the file COPYRIGH.
2.14 ! frystyk 6: ** @(#) $Id: HTFilter.c,v 2.13 1996/09/09 16:23:04 frystyk Exp $
2.1 frystyk 7: **
8: ** This module implrments a set of default filters that can be registerd
9: ** as BEFORE and AFTER filters to the Net manager
10: ** Authors
11: ** HFN Henrik Frystyk, frystyk@w.org
12: ** History
13: ** Jul 4, 96 Written
14: */
15:
16: /* Library include files */
17: #include "WWWLib.h"
18: #include "WWWCache.h"
19: #include "WWWHTTP.h"
20: #include "HTLog.h"
21: #include "HTAccess.h"
2.10 frystyk 22: #include "HTProxy.h"
23: #include "HTRules.h"
2.1 frystyk 24: #include "HTFilter.h" /* Implemented here */
25:
26: /* ------------------------------------------------------------------------- */
27:
28: /*
29: ** Proxy and Gateway BEFORE filter
30: ** -------------------------------
31: ** Checks for registerd proxy servers or gateways and sees whether this
32: ** request should be redirected to a proxy or a gateway. Proxies have
33: ** higher priority than gateways so we look for them first!
34: ** For HTTP/1.0 and HTTP/1.1 we may only send a full URL (including the
35: ** host portion) to proxy servers. Therefore, we tell the Library whether
36: ** to use the full URL or the traditional HTTP one without the host part.
37: */
38: PUBLIC int HTProxyFilter (HTRequest * request, void * param, int status)
39: {
40: HTParentAnchor * anchor = HTRequest_anchor(request);
2.2 frystyk 41: char * addr = HTAnchor_physical(anchor);
2.1 frystyk 42: char * physical = NULL;
43: if ((physical = HTProxy_find(addr))) {
2.6 frystyk 44: HTRequest_setFullURI(request, YES); /* For now */
2.5 frystyk 45: HTRequest_setProxy(request, physical);
2.8 frystyk 46: HT_FREE(physical);
2.6 frystyk 47: #if 0
48: /* Don't paste the URLs together anymore */
2.1 frystyk 49: StrAllocCat(physical, addr);
2.5 frystyk 50: HTAnchor_setPhysical(anchor, physical);
2.6 frystyk 51: #endif
2.1 frystyk 52: } else if ((physical = HTGateway_find(addr))) {
53: /*
54: ** A gateway URL is crated by chopping off any leading "/" to make the
55: ** host into part of path
56: */
57: char * path =
58: HTParse(addr, "", PARSE_HOST + PARSE_PATH + PARSE_PUNCTUATION);
59: char * gatewayed = HTParse(path+1, physical, PARSE_ALL);
60: HTAnchor_setPhysical(anchor, gatewayed);
61: HT_FREE(path);
62: HT_FREE(gatewayed);
63: HTRequest_setFullURI(request, NO);
2.6 frystyk 64: HTRequest_deleteProxy(request);
2.1 frystyk 65: } else {
2.6 frystyk 66: HTRequest_setFullURI(request, NO); /* For now */
67: HTRequest_deleteProxy(request);
2.1 frystyk 68: }
69: return HT_OK;
70: }
71:
72: /*
73: ** Rule Translation BEFORE Filter
74: ** ------------------------------
75: ** If we have a set of rules loaded (see the Rule manager) then check
76: ** before each request whether how that should be translated. The trick
77: ** is that a parent anchor has a "address" which is the part from the URL
78: ** we used when we created the anchor. However, it also have a "physical
79: ** address" which is the place we are actually going to look for the
2.2 frystyk 80: ** resource. Hence this filter translates the physical address
81: ** (if any translations are found)
2.1 frystyk 82: */
83: PUBLIC int HTRuleFilter (HTRequest * request, void * param, int status)
84: {
85: HTList * list = HTRule_global();
86: HTParentAnchor * anchor = HTRequest_anchor(request);
2.2 frystyk 87: char * addr = HTAnchor_physical(anchor);
2.1 frystyk 88: char * physical = HTRule_translate(list, addr, NO);
89: if (!physical) {
90: HTRequest_addError(request, ERR_FATAL, NO, HTERR_FORBIDDEN,
91: NULL, 0, "HTRuleFilter");
92: return HT_ERROR;
93: }
94: HTAnchor_setPhysical(anchor, physical);
95: HT_FREE(physical);
96: return HT_OK;
97: }
98:
99: /*
100: ** Cache Validation BEFORE Filter
101: ** ------------------------------
102: ** Check the cache mode to see if we can use an already loaded version
103: ** of this document. If so and our copy is valid then we don't have
104: ** to go out and get it unless we are forced to
2.3 frystyk 105: ** We only check the cache in caseof a GET request. Otherwise, we go
106: ** directly to the source.
2.1 frystyk 107: */
108: PUBLIC int HTCacheFilter (HTRequest * request, void * param, int status)
109: {
110: HTParentAnchor * anchor = HTRequest_anchor(request);
2.12 frystyk 111: HTCache * cache = NULL;
2.1 frystyk 112: HTReload mode = HTRequest_reloadMode(request);
2.3 frystyk 113: HTMethod method = HTRequest_method(request);
2.12 frystyk 114: HTDisconnectedMode disconnect = HTCacheMode_disconnected();
115: BOOL validate = NO;
2.3 frystyk 116:
117: /*
2.12 frystyk 118: ** If the cache is disabled all together then it won't help looking, huh?
2.3 frystyk 119: */
2.12 frystyk 120: if (!HTCacheMode_enabled()) return HT_OK;
121: if (CACHE_TRACE) HTTrace("Cachefilter. Checking persistent cache\n");
2.3 frystyk 122:
2.1 frystyk 123: /*
2.12 frystyk 124: ** Now check the cache...
2.1 frystyk 125: */
2.12 frystyk 126: if (method != METHOD_GET) {
127: if (CACHE_TRACE) HTTrace("Cachefilter. We only check GET methods\n");
128: } else if (mode == HT_CACHE_FLUSH) {
129: /*
130: ** If the mode if "Force Reload" then don't even bother to check the
131: ** cache - we flush everything we know abut this document anyway.
2.1 frystyk 132: ** Add the appropriate request headers. We use both the "pragma"
133: ** and the "cache-control" headers in order to be
2.12 frystyk 134: ** backwards compatible with HTTP/1.0
2.1 frystyk 135: */
2.12 frystyk 136: validate = YES;
2.1 frystyk 137: HTRequest_addGnHd(request, HT_G_PRAGMA_NO_CACHE);
2.11 frystyk 138: HTRequest_addCacheControl(request, "no-cache", "");
2.1 frystyk 139:
140: /*
141: ** We also flush the information in the anchor
142: */
143: HTAnchor_clearHeader(anchor);
144:
2.12 frystyk 145: } else {
146: /*
147: ** Check the persistent cache manager. If we have a cache hit then
148: ** continue to see if the reload mode requires us to do a validation
149: ** check. This filter assumes that we can get the cached version
150: ** through one of our protocol modules (for example the file module)
151: */
152: cache = HTCache_find(anchor);
2.1 frystyk 153: if (cache) {
2.12 frystyk 154: mode = HTMAX(mode, HTCache_isFresh(cache, request));
155:
156: /*
157: ** Now check the mode and add the right headers for the validation
158: ** If we are to validate a cache entry then we get a lock
159: ** on it so that not other requests can steal it.
160: */
161: if (mode == HT_CACHE_END_VALIDATE) {
162: /*
163: ** If we were asked to end-to-end validate the cached object
164: ** then use a max-age=0 cache control directive
165: */
166: validate = YES;
167: HTCache_getLock(cache, request);
168: HTRequest_addCacheControl(request, "max-age", "0");
169: } else if (mode == HT_CACHE_VALIDATE) {
2.11 frystyk 170: /*
2.12 frystyk 171: ** If we were asked to validate the cached object then
2.11 frystyk 172: ** use the etag or the last modified for cache validation
173: */
2.12 frystyk 174: validate = YES;
175: HTCache_getLock(cache, request);
2.11 frystyk 176: HTRequest_addRqHd(request, HT_C_IF_NONE_MATCH | HT_C_IMS);
2.12 frystyk 177: } else {
178: /*
179: ** The entity does not require any validation at all. We
180: ** can just go ahead and get it from the cache
181: */
182: char * name = HTCache_name(cache);
183: HTAnchor_setPhysical(anchor, name);
184: HTCache_addHit(cache);
185: HT_FREE(name);
2.11 frystyk 186: }
187: }
2.12 frystyk 188: }
189:
190: /*
191: ** If we are in disconnected mode and we are to validate an entry
192: ** then check whether what mode of disconnected mode we're in. If
193: ** we are to use our own cache then return a "504 Gateway Timeout"
194: */
195: if ((!cache || validate) && disconnect != HT_DISCONNECT_NONE) {
196: if (disconnect == HT_DISCONNECT_EXTERNAL)
197: HTRequest_addCacheControl(request, "only-if-cached", "");
198: else {
199: HTRequest_addError(request, ERR_FATAL, NO,
200: HTERR_GATE_TIMEOUT, "Disconnected Cache Mode",
201: 0, "HTCacheFilter");
202: return HT_ERROR;
203: }
2.11 frystyk 204: }
205: return HT_OK;
206: }
207:
208: /*
2.12 frystyk 209: ** Cache Update AFTER filter
210: ** -------------------------
211: ** On our way out we catch the metainformation and stores it in
212: ** our persistent store. If we have a cache validation (a 304
213: ** response then we use the new metainformation and merges it with
214: ** the existing information already captured in the cache.
215: */
216: PUBLIC int HTCacheUpdateFilter (HTRequest * request, void * param, int status)
217: {
218: HTParentAnchor * anchor = HTRequest_anchor(request);
219: HTCache * cache = HTCache_find(anchor);
220:
221: /*
222: ** If this request resulted in a "304 Not Modified" response then
223: ** we merge the new metainformation with the old.
224: */
225: if (CACHE_TRACE) HTTrace("Cache....... Merging metainformation\n");
226:
227: /*
228: ** It may in fact be that the information in the 304 response
229: ** told us that we can't cache the entity anymore. If this is the
230: ** case then flush it now. Otherwise prepare for a cache read
231: */
232: if (HTAnchor_cachable(anchor) == NO) {
233: HTCache_remove(cache);
234: } else {
2.13 frystyk 235: HTCache_updateMeta(cache, request);
2.12 frystyk 236: HTRequest_setReloadMode(request, HT_CACHE_OK);
237: }
238:
239: /*
240: ** Start request directly from the cache. As with the redirection filter
241: ** we reuse the same request object which means that we must
242: ** keep this around until the cache load request has terminated
243: ** In the case of a
244: */
245: {
246: HTLoad(request, NO);
247: return HT_ERROR;
248: }
249: }
250:
251: /*
2.11 frystyk 252: ** Check the Memory Cache (History list)
253: ** -------------------------------------
254: ** Check if document is already loaded. The user can define whether
255: ** the history list should follow normal expiration or work as a
256: ** traditional history list where expired documents are not updated.
257: ** We don't check for anything but existence proof of a document
258: ** associated with the anchor as the definition is left to the application
259: */
260: PUBLIC int HTMemoryCacheFilter (HTRequest * request, void * param, int status)
261: {
262: HTReload validation = HTRequest_reloadMode(request);
263: HTParentAnchor * anchor = HTRequest_anchor(request);
264: void * document = HTAnchor_document(anchor);
2.1 frystyk 265:
2.11 frystyk 266: /*
2.14 ! frystyk 267: ** We only check the memory cache if it's a GET method
! 268: */
! 269: if (HTRequest_method(request) != METHOD_GET) {
! 270: if (CACHE_TRACE) HTTrace("Mem Cache... We only check GET methods\n");
! 271: return HT_OK;
! 272: }
! 273:
! 274: /*
2.11 frystyk 275: ** If we are asked to flush the persistent cache then there is no reason
276: ** to do anything here - we're flushing it anyway. Also if no document
277: ** then just exit from this filter.
278: */
279: if (!document || validation > HT_CACHE_FLUSH_MEM) {
280: if (CACHE_TRACE) HTTrace("Mem Cache... No fresh document...\n");
281: return HT_OK;
282: }
2.1 frystyk 283:
2.11 frystyk 284: /*
285: ** If we have a document object associated with this anchor then we also
286: ** have the object in the history list. Depending on what the user asked,
287: ** we can add a cache validator
288: */
2.14 ! frystyk 289: #if 0
2.11 frystyk 290: if (document) {
291: if (validation != HT_CACHE_FLUSH_MEM) {
2.14 ! frystyk 292: HTExpiresMode expires = HTCacheMode_expires();
2.11 frystyk 293: if (CACHE_TRACE)
294: HTTrace("Mem Cache... Document already in memory\n");
295: if (expires != HT_EXPIRES_IGNORE) {
296:
297: /*
298: ** Ask the cache manager if this object has expired. Also
299: ** check if we should care about expiration or not.
300: */
301: if (!HTCache_isValid(anchor)) {
302: if (expires == HT_EXPIRES_NOTIFY) {
303:
304: /*
305: ** See if we have a function registered for outputting errors.
306: ** If so then call it and present the message to the user
307: */
308: HTAlertCallback * cbf = HTAlert_find(HT_A_MESSAGE);
309: if (cbf)
310: (*cbf)(request, HT_A_MESSAGE, HTERR_CACHE_EXPIRED,
311: NULL, HTRequest_error(request), NULL);
312: } else {
313: if (CACHE_TRACE) HTTrace("Mem Cache... Expired - autoreload\n");
314: HTRequest_addRqHd(request, HT_C_IF_NONE_MATCH | HT_C_IMS);
315: return HT_OK; /* Must go get it */
316: }
317: }
2.1 frystyk 318: }
319: }
2.14 ! frystyk 320: return HT_LOADED; /* Got it! */
! 321: }
! 322: return HT_OK;
! 323: #endif
! 324: if (document && validation != HT_CACHE_FLUSH_MEM) {
! 325: if (CACHE_TRACE) HTTrace("Mem Cache... Document already in memory\n");
! 326: return HT_LOADED;
2.1 frystyk 327: }
328: return HT_OK;
329: }
330:
331: /*
332: ** Error and Information AFTER filter
333: ** ----------------------------------
334: ** It checks the status code from a request and generates an
335: ** error/information message if required.
336: */
337: PUBLIC int HTInfoFilter (HTRequest * request, void * param, int status)
338: {
339: HTParentAnchor * anchor = HTRequest_anchor(request);
340: char * uri = HTAnchor_address((HTAnchor*) anchor);
341: switch (status) {
342: case HT_RETRY:
343: if (PROT_TRACE)
344: HTTrace("Load End.... NOT AVAILABLE, RETRY AT %ld\n",
345: HTRequest_retryTime(request));
346: break;
347:
348: case HT_ERROR:
349: {
350: /*
351: ** See if we have a function registered for outputting errors.
352: ** If so then call it and present the message to the user
353: */
354: HTAlertCallback *cbf = HTAlert_find(HT_A_MESSAGE);
355: if (cbf) (*cbf)(request, HT_A_MESSAGE, HT_MSG_NULL, NULL,
356: HTRequest_error(request), NULL);
357: if (PROT_TRACE)
358: HTTrace("Load End.... ERROR: Can't access `%s\'\n",
359: uri ? uri : "<UNKNOWN>");
2.3 frystyk 360: break;
2.7 frystyk 361: }
362:
363: case HT_NO_DATA:
364: {
365: /*
366: ** The document was empty
367: */
368: HTAlertCallback *cbf = HTAlert_find(HT_A_MESSAGE);
369: if (cbf) (*cbf)(request, HT_A_MESSAGE, HT_MSG_NULL, NULL,
370: HTRequest_error(request), NULL);
371: if (PROT_TRACE)
372: HTTrace("Load End.... EMPTY: No content `%s\'\n",
373: uri ? uri : "<UNKNOWN>");
374: break;
375: }
2.3 frystyk 376:
377: case HT_LOADED:
378: {
379: /*
380: ** Even though we have received a loaded status the thing we have
381: ** loaded successfully may in fact be an error message. We therefore
382: ** look at the error stack to see what to do.
383: */
384: HTAlertCallback *cbf = HTAlert_find(HT_A_MESSAGE);
385: if (cbf) (*cbf)(request, HT_A_MESSAGE, HT_MSG_NULL, NULL,
386: HTRequest_error(request), NULL);
387: if (PROT_TRACE) HTTrace("Load End.... OK: `%s\'\n", uri);
2.1 frystyk 388: break;
389: }
390:
391: default:
392: if (PROT_TRACE)
393: HTTrace("Load End.... Request ended with code %d\n", status);
394: break;
395: }
396:
397: HT_FREE(uri);
398: return HT_OK;
399: }
400:
401: /*
402: ** Redirection AFTER filter
403: ** ------------------------
404: ** The redirection handler only handles redirections
405: ** on the GET or HEAD method (or any other safe method)
406: */
407: PUBLIC int HTRedirectFilter (HTRequest * request, void * param, int status)
408: {
409: HTMethod method = HTRequest_method(request);
410: HTAnchor * new_anchor = HTRequest_redirection(request);
2.7 frystyk 411: if (!new_anchor) {
412: if (PROT_TRACE) HTTrace("Redirection. No destination\n");
413: return HT_OK;
414: }
415:
2.1 frystyk 416: /*
417: ** Only do redirect on GET and HEAD
418: */
2.7 frystyk 419: if (!HTMethod_isSafe(method)) {
2.4 frystyk 420: HTAlertCallback * prompt = HTAlert_find(HT_A_CONFIRM);
421: if (prompt) {
422: if ((*prompt)(request, HT_A_CONFIRM, HT_MSG_REDIRECTION,
423: NULL, NULL, NULL) != YES)
424: return HT_ERROR;
425: }
2.1 frystyk 426: }
427:
428: /*
429: ** Start new request with the redirect anchor found in the headers.
430: ** Note that we reuse the same request object which means that we must
431: ** keep this around until the redirected request has terminated. It also
432: ** allows us in an easy way to keep track of the number of redirections
433: ** so that we can detect endless loops.
434: */
2.4 frystyk 435: if (HTRequest_doRetry(request)) {
2.1 frystyk 436: HTLoadAnchor(new_anchor, request);
2.9 frystyk 437: } else {
438: HTRequest_addError(request, ERR_FATAL, NO, HTERR_MAX_REDIRECT,
439: NULL, 0, "HTRedirectFilter");
440: }
441:
442: /*
443: ** By returning HT_ERROR we make sure that this is the last handler to be
444: ** called. We do this as we don't want any other filter to delete the
445: ** request object now when we have just started a new one ourselves
446: */
447: return HT_ERROR;
448: }
449:
450: /*
451: ** Retry through Proxy Filter
452: ** --------------------------
453: ** This filter handles a 305 Use Proxy response and retries the request
454: ** through the proxy
455: */
456: PUBLIC int HTUseProxyFilter (HTRequest * request, void * param, int status)
457: {
458: HTAnchor * proxy_anchor = HTRequest_redirection(request);
459: if (!proxy_anchor) {
460: if (PROT_TRACE) HTTrace("Use Proxy... No proxy location\n");
461: return HT_OK;
462: }
463:
464: /*
465: ** Add the proxy to the list. Assume HTTP access method only!
466: */
467: {
468: char * addr = HTAnchor_address(proxy_anchor);
469: HTProxy_add("http", addr);
470: HT_FREE(addr);
471: }
472:
473: /*
474: ** Start new request through the proxy if we haven't reached the max
475: ** number of redirections for this request
476: */
477: if (HTRequest_doRetry(request)) {
478: HTLoadAnchor(proxy_anchor, request);
2.1 frystyk 479: } else {
480: HTRequest_addError(request, ERR_FATAL, NO, HTERR_MAX_REDIRECT,
481: NULL, 0, "HTRedirectFilter");
482: }
483:
484: /*
485: ** By returning HT_ERROR we make sure that this is the last handler to be
486: ** called. We do this as we don't want any other filter to delete the
487: ** request object now when we have just started a new one ourselves
488: */
489: return HT_ERROR;
490: }
491:
492: /*
493: ** Client side authentication BEFORE filter
494: ** ----------------------------------------
495: ** The filter generates the credentials required to access a document
496: ** Getting the credentials may involve asking the user
497: */
498: PUBLIC int HTCredentialsFilter (HTRequest * request, void * param, int status)
499: {
500: /*
501: ** Ask the authentication module to call the right credentials generator
502: ** that understands this scheme
503: */
504: if (HTAA_beforeFilter(request, param, status) == HT_OK) {
505: if (PROT_TRACE) HTTrace("Credentials. verified\n");
506: return HT_OK;
507: } else {
508: HTRequest_addError(request, ERR_FATAL, NO, HTERR_UNAUTHORIZED,
509: NULL, 0, "HTCredentialsFilter");
510: return HT_ERROR;
511: }
512: }
513:
514: /*
515: ** Client side authentication AFTER filter
516: ** ---------------------------------------
517: ** The client side authentication filter uses the
518: ** user dialog messages registered in the HTAlert module.
519: ** By default these are the ones used by the line mode browser but you can
520: ** just register something else.
521: */
522: PUBLIC int HTAuthFilter (HTRequest * request, void * param, int status)
523: {
524: /*
525: ** Ask the authentication module to call the right challenge parser
526: ** that understands this scheme
527: */
528: if (HTAA_afterFilter(request, param, status) == HT_OK) {
529:
530: /*
531: ** Start request with new credentials. As with the redirection filter
532: ** we reuse the same request object which means that we must
533: ** keep this around until the redirected request has terminated
534: */
535: HTLoad(request, NO);
536:
537: /*
538: ** We return HT_ERROR to make sure that this is the last handler to be
539: ** called. We do this as we don't want any other filter to delete the
540: ** request object now when we have just started a new one ourselves
541: */
542: return HT_ERROR;
543: }
544: return HT_OK;
545: }
546:
547: /*
548: ** Request Loggin AFTER filter
549: ** ---------------------------
550: ** Default Logging filter using the log manager provided by HTLog.c
551: */
552: PUBLIC int HTLogFilter (HTRequest * request, void * param, int status)
553: {
554: if (request) {
555: if (HTLog_isOpen()) HTLog_add(request, status);
556: return HT_OK;
557: }
558: return HT_ERROR;
559: }
Webmaster