2 * Copyright 2001-2007 Internet2
\r
4 * Licensed under the Apache License, Version 2.0 (the "License");
\r
5 * you may not use this file except in compliance with the License.
\r
6 * You may obtain a copy of the License at
\r
8 * http://www.apache.org/licenses/LICENSE-2.0
\r
10 * Unless required by applicable law or agreed to in writing, software
\r
11 * distributed under the License is distributed on an "AS IS" BASIS,
\r
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
\r
13 * See the License for the specific language governing permissions and
\r
14 * limitations under the License.
\r
17 /** XMLRequestMapper.cpp
\r
19 * XML-based RequestMapper implementation
\r
22 #include "internal.h"
\r
23 #include "exceptions.h"
\r
24 #include "AccessControl.h"
\r
25 #include "RequestMapper.h"
\r
26 #include "SPRequest.h"
\r
27 #include "util/DOMPropertySet.h"
\r
28 #include "util/SPConstants.h"
\r
30 #include <xmltooling/util/NDC.h>
\r
31 #include <xmltooling/util/ReloadableXMLFile.h>
\r
32 #include <xmltooling/util/XMLHelper.h>
\r
33 #include <xercesc/util/XMLUniDefs.hpp>
\r
34 #include <xercesc/util/regx/RegularExpression.hpp>
\r
36 using namespace shibsp;
\r
37 using namespace xmltooling;
\r
38 using namespace std;
\r
42 // Blocks access when an ACL plugin fails to load.
\r
43 class AccessControlDummy : public AccessControl
\r
52 aclresult_t authorized(const SPRequest& request, const Session* session) const {
\r
53 return shib_acl_false;
\r
57 class Override : public DOMPropertySet, public DOMNodeFilter
\r
60 Override() : m_acl(NULL) {}
\r
61 Override(const DOMElement* e, Category& log, const Override* base=NULL);
\r
64 // Provides filter to exclude special config elements.
\r
65 short acceptNode(const DOMNode* node) const {
\r
66 return FILTER_REJECT;
\r
69 const Override* locate(const HTTPRequest& request) const;
\r
70 AccessControl* getAC() const { return (m_acl ? m_acl : (getParent() ? dynamic_cast<const Override*>(getParent())->getAC() : NULL)); }
\r
73 void loadACL(const DOMElement* e, Category& log);
\r
75 map<string,Override*> m_map;
\r
76 vector< pair<RegularExpression*,Override*> > m_regexps;
\r
77 vector< pair< pair<string,RegularExpression*>,Override*> > m_queries;
\r
80 AccessControl* m_acl;
\r
83 class XMLRequestMapperImpl : public Override
\r
86 XMLRequestMapperImpl(const DOMElement* e, Category& log);
\r
88 ~XMLRequestMapperImpl() {
\r
90 m_document->release();
\r
93 void setDocument(DOMDocument* doc) {
\r
97 const Override* findOverride(const char* vhost, const HTTPRequest& request) const;
\r
100 map<string,Override*> m_extras;
\r
101 DOMDocument* m_document;
\r
104 #if defined (_MSC_VER)
\r
105 #pragma warning( push )
\r
106 #pragma warning( disable : 4250 )
\r
109 class XMLRequestMapper : public RequestMapper, public ReloadableXMLFile
\r
112 XMLRequestMapper(const DOMElement* e) : ReloadableXMLFile(e,Category::getInstance(SHIBSP_LOGCAT".RequestMapper")), m_impl(NULL) {
\r
116 ~XMLRequestMapper() {
\r
120 Settings getSettings(const HTTPRequest& request) const;
\r
123 pair<bool,DOMElement*> load();
\r
126 XMLRequestMapperImpl* m_impl;
\r
129 #if defined (_MSC_VER)
\r
130 #pragma warning( pop )
\r
133 RequestMapper* SHIBSP_DLLLOCAL XMLRequestMapperFactory(const DOMElement* const & e)
\r
135 return new XMLRequestMapper(e);
\r
138 static const XMLCh _AccessControl[] = UNICODE_LITERAL_13(A,c,c,e,s,s,C,o,n,t,r,o,l);
\r
139 static const XMLCh AccessControlProvider[] = UNICODE_LITERAL_21(A,c,c,e,s,s,C,o,n,t,r,o,l,P,r,o,v,i,d,e,r);
\r
140 static const XMLCh Host[] = UNICODE_LITERAL_4(H,o,s,t);
\r
141 static const XMLCh HostRegex[] = UNICODE_LITERAL_9(H,o,s,t,R,e,g,e,x);
\r
142 static const XMLCh htaccess[] = UNICODE_LITERAL_8(h,t,a,c,c,e,s,s);
\r
143 static const XMLCh ignoreCase[] = UNICODE_LITERAL_10(i,g,n,o,r,e,C,a,s,e);
\r
144 static const XMLCh ignoreOption[] = UNICODE_LITERAL_1(i);
\r
145 static const XMLCh Path[] = UNICODE_LITERAL_4(P,a,t,h);
\r
146 static const XMLCh PathRegex[] = UNICODE_LITERAL_9(P,a,t,h,R,e,g,e,x);
\r
147 static const XMLCh Query[] = UNICODE_LITERAL_5(Q,u,e,r,y);
\r
148 static const XMLCh name[] = UNICODE_LITERAL_4(n,a,m,e);
\r
149 static const XMLCh regex[] = UNICODE_LITERAL_5(r,e,g,e,x);
\r
150 static const XMLCh _type[] = UNICODE_LITERAL_4(t,y,p,e);
\r
153 void SHIBSP_API shibsp::registerRequestMappers()
\r
155 SPConfig& conf=SPConfig::getConfig();
\r
156 conf.RequestMapperManager.registerFactory(XML_REQUEST_MAPPER, XMLRequestMapperFactory);
\r
157 conf.RequestMapperManager.registerFactory(NATIVE_REQUEST_MAPPER, XMLRequestMapperFactory);
\r
160 void Override::loadACL(const DOMElement* e, Category& log)
\r
163 const DOMElement* acl=XMLHelper::getFirstChildElement(e,htaccess);
\r
165 log.info("building Apache htaccess AccessControl provider...");
\r
166 m_acl=SPConfig::getConfig().AccessControlManager.newPlugin(HT_ACCESS_CONTROL,acl);
\r
169 acl=XMLHelper::getFirstChildElement(e,_AccessControl);
\r
171 log.info("building XML-based AccessControl provider...");
\r
172 m_acl=SPConfig::getConfig().AccessControlManager.newPlugin(XML_ACCESS_CONTROL,acl);
\r
175 acl=XMLHelper::getFirstChildElement(e,AccessControlProvider);
\r
177 auto_ptr_char type(acl->getAttributeNS(NULL,_type));
\r
178 log.info("building AccessControl provider of type %s...",type.get());
\r
179 m_acl=SPConfig::getConfig().AccessControlManager.newPlugin(type.get(),acl);
\r
184 catch (exception& ex) {
\r
185 log.crit("exception building AccessControl provider: %s", ex.what());
\r
186 m_acl = new AccessControlDummy();
\r
190 Override::Override(const DOMElement* e, Category& log, const Override* base) : m_acl(NULL)
\r
193 // Load the property set.
\r
197 // Load any AccessControl provider.
\r
200 // Handle nested Paths.
\r
201 DOMElement* path = XMLHelper::getFirstChildElement(e,Path);
\r
202 for (int i=1; path; ++i, path=XMLHelper::getNextSiblingElement(path,Path)) {
\r
203 const XMLCh* n=path->getAttributeNS(NULL,name);
\r
205 // Skip any leading slashes.
\r
206 while (n && *n==chForwardSlash)
\r
209 // Check for empty name.
\r
211 log.warn("skipping Path element (%d) with empty name attribute", i);
\r
215 // Check for an embedded slash.
\r
216 int slash=XMLString::indexOf(n,chForwardSlash);
\r
218 // Copy the first path segment.
\r
219 XMLCh* namebuf=new XMLCh[slash + 1];
\r
220 for (int pos=0; pos < slash; pos++)
\r
221 namebuf[pos]=n[pos];
\r
222 namebuf[slash]=chNull;
\r
224 // Move past the slash in the original pathname.
\r
227 // Skip any leading slashes again.
\r
228 while (*n==chForwardSlash)
\r
232 // Create a placeholder Path element for the first path segment and replant under it.
\r
233 DOMElement* newpath=path->getOwnerDocument()->createElementNS(shibspconstants::SHIB2SPCONFIG_NS,Path);
\r
234 newpath->setAttributeNS(NULL,name,namebuf);
\r
235 path->setAttributeNS(NULL,name,n);
\r
236 path->getParentNode()->replaceChild(newpath,path);
\r
237 newpath->appendChild(path);
\r
239 // Repoint our locals at the new parent.
\r
241 n=path->getAttributeNS(NULL,name);
\r
244 // All we had was a pathname with trailing slash(es), so just reset it without them.
\r
245 path->setAttributeNS(NULL,name,namebuf);
\r
246 n=path->getAttributeNS(NULL,name);
\r
251 Override* o=new Override(path,log,this);
\r
252 pair<bool,const char*> name=o->getString("name");
\r
253 char* dup=strdup(name.second);
\r
254 for (char* pch=dup; *pch; pch++)
\r
255 *pch=tolower(*pch);
\r
256 if (m_map.count(dup)) {
\r
257 log.warn("skipping duplicate Path element (%s)",dup);
\r
263 log.debug("added Path mapping (%s)", dup);
\r
267 if (!XMLString::equals(e->getLocalName(), PathRegex)) {
\r
268 // Handle nested PathRegexs.
\r
269 path = XMLHelper::getFirstChildElement(e,PathRegex);
\r
270 for (int i=1; path; ++i, path=XMLHelper::getNextSiblingElement(path,PathRegex)) {
\r
271 const XMLCh* n=path->getAttributeNS(NULL,regex);
\r
273 log.warn("skipping PathRegex element (%d) with empty regex attribute",i);
\r
277 auto_ptr<Override> o(new Override(path,log,this));
\r
279 const XMLCh* flag=path->getAttributeNS(NULL,ignoreCase);
\r
281 auto_ptr<RegularExpression> re(
\r
282 new RegularExpression(n, (flag && (*flag==chLatin_f || *flag==chDigit_0)) ? &chNull : ignoreOption)
\r
284 m_regexps.push_back(make_pair(re.release(), o.release()));
\r
286 catch (XMLException& ex) {
\r
287 auto_ptr_char tmp(ex.getMessage());
\r
288 log.error("caught exception while parsing PathRegex regular expression (%d): %s", i, tmp.get());
\r
289 throw ConfigurationException("Invalid regular expression in PathRegex element.");
\r
292 if (log.isDebugEnabled())
\r
293 log.debug("added <PathRegex> mapping (%s)", m_regexps.back().second->getString("regex").second);
\r
297 // Handle nested Querys.
\r
298 path = XMLHelper::getFirstChildElement(e,Query);
\r
299 for (int i=1; path; ++i, path=XMLHelper::getNextSiblingElement(path,Query)) {
\r
300 const XMLCh* n=path->getAttributeNS(NULL,name);
\r
302 log.warn("skipping Query element (%d) with empty name attribute",i);
\r
305 auto_ptr_char ntemp(n);
\r
306 const XMLCh* v=path->getAttributeNS(NULL,regex);
\r
308 auto_ptr<Override> o(new Override(path,log,this));
\r
310 RegularExpression* re = NULL;
\r
312 re = new RegularExpression(v);
\r
313 m_queries.push_back(make_pair(make_pair(ntemp.get(),re), o.release()));
\r
315 catch (XMLException& ex) {
\r
316 auto_ptr_char tmp(ex.getMessage());
\r
317 log.error("caught exception while parsing Query regular expression (%d): %s", i, tmp.get());
\r
318 throw ConfigurationException("Invalid regular expression in Query element.");
\r
321 log.debug("added <Query> mapping (%s)", ntemp.get());
\r
324 catch (exception&) {
\r
326 for_each(m_map.begin(),m_map.end(),xmltooling::cleanup_pair<string,Override>());
\r
327 for (vector< pair<RegularExpression*,Override*> >::iterator i = m_regexps.begin(); i != m_regexps.end(); ++i) {
\r
331 for (vector< pair< pair<string,RegularExpression*>,Override*> >::iterator j = m_queries.begin(); j != m_queries.end(); ++j) {
\r
332 delete j->first.second;
\r
339 Override::~Override()
\r
342 for_each(m_map.begin(),m_map.end(),xmltooling::cleanup_pair<string,Override>());
\r
343 for (vector< pair<RegularExpression*,Override*> >::iterator i = m_regexps.begin(); i != m_regexps.end(); ++i) {
\r
347 for (vector< pair< pair<string,RegularExpression*>,Override*> >::iterator j = m_queries.begin(); j != m_queries.end(); ++j) {
\r
348 delete j->first.second;
\r
353 const Override* Override::locate(const HTTPRequest& request) const
\r
355 // This function is confusing because it's *not* recursive.
\r
356 // The whole path is tokenized and mapped in a loop, so the
\r
357 // path parameter starts with the entire request path and
\r
358 // we can skip the leading slash as irrelevant.
\r
359 const char* path = request.getRequestURI();
\r
363 // Now we copy the path, chop the query string, and lower case it.
\r
364 char* dup=strdup(path);
\r
365 char* sep=strchr(dup,'?');
\r
368 for (char* pch=dup; *pch; pch++)
\r
369 *pch=tolower(*pch);
\r
371 // Default is for the current object to provide settings.
\r
372 const Override* o=this;
\r
374 // Tokenize the path by segment and try and map each segment.
\r
375 #ifdef HAVE_STRTOK_R
\r
377 const char* token=strtok_r(dup,"/",&pos);
\r
379 const char* token=strtok(dup,"/");
\r
382 map<string,Override*>::const_iterator i=o->m_map.find(token);
\r
383 if (i==o->m_map.end())
\r
384 break; // Once there's no match, we've consumed as much of the path as possible here.
\r
385 // We found a match, so reset the settings pointer.
\r
388 // We descended a step down the path, so we need to advance the original
\r
389 // parameter for the regex step later.
\r
390 path += strlen(token);
\r
394 // Get the next segment, if any.
\r
395 #ifdef HAVE_STRTOK_R
\r
396 token=strtok_r(NULL,"/",&pos);
\r
398 token=strtok(NULL,"/");
\r
404 // If there's anything left, we try for a regex match on the rest of the path minus the query string.
\r
406 string path2(path);
\r
407 path2 = path2.substr(0,path2.find('?'));
\r
409 for (vector< pair<RegularExpression*,Override*> >::const_iterator re = o->m_regexps.begin(); re != o->m_regexps.end(); ++re) {
\r
410 if (re->first->matches(path2.c_str())) {
\r
417 // Finally, check for query string matches. This is another "unrolled" recursive descent in a loop.
\r
421 for (vector< pair< pair<string,RegularExpression*>,Override*> >::const_iterator q = o->m_queries.begin(); !descended && q != o->m_queries.end(); ++q) {
\r
422 vector<const char*> vals;
\r
423 if (request.getParameters(q->first.first.c_str(), vals)) {
\r
424 if (q->first.second) {
\r
425 // We have to match one of the values.
\r
426 for (vector<const char*>::const_iterator v = vals.begin(); v != vals.end(); ++v) {
\r
427 if (q->first.second->matches(*v)) {
\r
435 // The simple presence of the parameter is sufficient to match.
\r
441 } while (descended);
\r
446 XMLRequestMapperImpl::XMLRequestMapperImpl(const DOMElement* e, Category& log) : m_document(NULL)
\r
449 xmltooling::NDC ndc("XMLRequestMapperImpl");
\r
452 // Load the property set.
\r
455 // Load any AccessControl provider.
\r
458 // Loop over the HostRegex elements.
\r
459 const DOMElement* host = XMLHelper::getFirstChildElement(e,HostRegex);
\r
460 for (int i=1; host; ++i, host=XMLHelper::getNextSiblingElement(host,HostRegex)) {
\r
461 const XMLCh* n=host->getAttributeNS(NULL,regex);
\r
463 log.warn("Skipping HostRegex element (%d) with empty regex attribute",i);
\r
467 auto_ptr<Override> o(new Override(host,log,this));
\r
469 const XMLCh* flag=host->getAttributeNS(NULL,ignoreCase);
\r
471 auto_ptr<RegularExpression> re(
\r
472 new RegularExpression(n, (flag && (*flag==chLatin_f || *flag==chDigit_0)) ? &chNull : ignoreOption)
\r
474 m_regexps.push_back(make_pair(re.release(), o.release()));
\r
476 catch (XMLException& ex) {
\r
477 auto_ptr_char tmp(ex.getMessage());
\r
478 log.error("caught exception while parsing HostRegex regular expression (%d): %s", i, tmp.get());
\r
481 log.debug("Added <HostRegex> mapping for %s", m_regexps.back().second->getString("regex").second);
\r
484 // Loop over the Host elements.
\r
485 host = XMLHelper::getFirstChildElement(e,Host);
\r
486 for (int i=1; host; ++i, host=XMLHelper::getNextSiblingElement(host,Host)) {
\r
487 const XMLCh* n=host->getAttributeNS(NULL,name);
\r
489 log.warn("Skipping Host element (%d) with empty name attribute",i);
\r
493 Override* o=new Override(host,log,this);
\r
494 pair<bool,const char*> name=o->getString("name");
\r
495 pair<bool,const char*> scheme=o->getString("scheme");
\r
496 pair<bool,const char*> port=o->getString("port");
\r
498 char* dup=strdup(name.second);
\r
499 for (char* pch=dup; *pch; pch++)
\r
500 *pch=tolower(*pch);
\r
501 auto_ptr<char> dupwrap(dup);
\r
503 if (!scheme.first && port.first) {
\r
504 // No scheme, but a port, so assume http.
\r
505 scheme = pair<bool,const char*>(true,"http");
\r
507 else if (scheme.first && !port.first) {
\r
508 // Scheme, no port, so default it.
\r
509 // XXX Use getservbyname instead?
\r
511 if (!strcmp(scheme.second,"http"))
\r
512 port.second = "80";
\r
513 else if (!strcmp(scheme.second,"https"))
\r
514 port.second = "443";
\r
515 else if (!strcmp(scheme.second,"ftp"))
\r
516 port.second = "21";
\r
517 else if (!strcmp(scheme.second,"ldap"))
\r
518 port.second = "389";
\r
519 else if (!strcmp(scheme.second,"ldaps"))
\r
520 port.second = "636";
\r
523 if (scheme.first) {
\r
524 string url(scheme.second);
\r
525 url=url + "://" + dup;
\r
527 // Is this the default port?
\r
528 if ((!strcmp(scheme.second,"http") && !strcmp(port.second,"80")) ||
\r
529 (!strcmp(scheme.second,"https") && !strcmp(port.second,"443")) ||
\r
530 (!strcmp(scheme.second,"ftp") && !strcmp(port.second,"21")) ||
\r
531 (!strcmp(scheme.second,"ldap") && !strcmp(port.second,"389")) ||
\r
532 (!strcmp(scheme.second,"ldaps") && !strcmp(port.second,"636"))) {
\r
533 // First store a port-less version.
\r
534 if (m_map.count(url) || m_extras.count(url)) {
\r
535 log.warn("Skipping duplicate Host element (%s)",url.c_str());
\r
540 log.debug("Added <Host> mapping for %s",url.c_str());
\r
542 // Now append the port. We use the extras vector, to avoid double freeing the object later.
\r
543 url=url + ':' + port.second;
\r
545 log.debug("Added <Host> mapping for %s",url.c_str());
\r
548 url=url + ':' + port.second;
\r
549 if (m_map.count(url) || m_extras.count(url)) {
\r
550 log.warn("Skipping duplicate Host element (%s)",url.c_str());
\r
555 log.debug("Added <Host> mapping for %s",url.c_str());
\r
559 // No scheme or port, so we enter dual hosts on http:80 and https:443
\r
560 string url("http://");
\r
562 if (m_map.count(url) || m_extras.count(url)) {
\r
563 log.warn("Skipping duplicate Host element (%s)",url.c_str());
\r
568 log.debug("Added <Host> mapping for %s",url.c_str());
\r
571 if (m_map.count(url) || m_extras.count(url)) {
\r
572 log.warn("Skipping duplicate Host element (%s)",url.c_str());
\r
576 log.debug("Added <Host> mapping for %s",url.c_str());
\r
580 if (m_map.count(url) || m_extras.count(url)) {
\r
581 log.warn("Skipping duplicate Host element (%s)",url.c_str());
\r
585 log.debug("Added <Host> mapping for %s",url.c_str());
\r
587 url = url + ":443";
\r
588 if (m_map.count(url) || m_extras.count(url)) {
\r
589 log.warn("Skipping duplicate Host element (%s)",url.c_str());
\r
593 log.debug("Added <Host> mapping for %s",url.c_str());
\r
598 const Override* XMLRequestMapperImpl::findOverride(const char* vhost, const HTTPRequest& request) const
\r
600 const Override* o=NULL;
\r
601 map<string,Override*>::const_iterator i=m_map.find(vhost);
\r
602 if (i!=m_map.end())
\r
605 i=m_extras.find(vhost);
\r
606 if (i!=m_extras.end())
\r
609 for (vector< pair<RegularExpression*,Override*> >::const_iterator re = m_regexps.begin(); !o && re != m_regexps.end(); ++re) {
\r
610 if (re->first->matches(vhost))
\r
616 return o ? o->locate(request) : this;
\r
619 pair<bool,DOMElement*> XMLRequestMapper::load()
\r
621 // Load from source using base class.
\r
622 pair<bool,DOMElement*> raw = ReloadableXMLFile::load();
\r
624 // If we own it, wrap it.
\r
625 XercesJanitor<DOMDocument> docjanitor(raw.first ? raw.second->getOwnerDocument() : NULL);
\r
627 XMLRequestMapperImpl* impl = new XMLRequestMapperImpl(raw.second,m_log);
\r
629 // If we held the document, transfer it to the impl. If we didn't, it's a no-op.
\r
630 impl->setDocument(docjanitor.release());
\r
635 return make_pair(false,(DOMElement*)NULL);
\r
638 RequestMapper::Settings XMLRequestMapper::getSettings(const HTTPRequest& request) const
\r
640 ostringstream vhost;
\r
641 vhost << request.getScheme() << "://" << request.getHostname() << ':' << request.getPort();
\r
643 const Override* o=m_impl->findOverride(vhost.str().c_str(), request);
\r
645 if (m_log.isDebugEnabled()) {
\r
647 xmltooling::NDC ndc("getSettings");
\r
649 pair<bool,const char*> ret=o->getString("applicationId");
\r
650 m_log.debug("mapped %s%s to %s", vhost.str().c_str(), request.getRequestURI() ? request.getRequestURI() : "", ret.second);
\r
653 return Settings(o,o->getAC());
\r