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