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 #ifdef SHIBSP_XERCESC_SHORT_ACCEPTNODE
70 acceptNode(const DOMNode* node) const {
74 const Override* locate(const HTTPRequest& request) const;
75 AccessControl* getAC() const { return (m_acl ? m_acl : (getParent() ? dynamic_cast<const Override*>(getParent())->getAC() : NULL)); }
78 void loadACL(const DOMElement* e, Category& log);
80 map<string,Override*> m_map;
81 vector< pair<RegularExpression*,Override*> > m_regexps;
82 vector< pair< pair<string,RegularExpression*>,Override*> > m_queries;
88 class XMLRequestMapperImpl : public Override
91 XMLRequestMapperImpl(const DOMElement* e, Category& log);
93 ~XMLRequestMapperImpl() {
95 m_document->release();
98 void setDocument(DOMDocument* doc) {
102 const Override* findOverride(const char* vhost, const HTTPRequest& request) const;
105 map<string,Override*> m_extras;
106 DOMDocument* m_document;
109 #if defined (_MSC_VER)
110 #pragma warning( push )
111 #pragma warning( disable : 4250 )
114 class XMLRequestMapper : public RequestMapper, public ReloadableXMLFile
117 XMLRequestMapper(const DOMElement* e) : ReloadableXMLFile(e,Category::getInstance(SHIBSP_LOGCAT".RequestMapper")), m_impl(NULL) {
121 ~XMLRequestMapper() {
125 Settings getSettings(const HTTPRequest& request) const;
128 pair<bool,DOMElement*> load();
131 XMLRequestMapperImpl* m_impl;
134 #if defined (_MSC_VER)
135 #pragma warning( pop )
138 RequestMapper* SHIBSP_DLLLOCAL XMLRequestMapperFactory(const DOMElement* const & e)
140 return new XMLRequestMapper(e);
143 static const XMLCh _AccessControl[] = UNICODE_LITERAL_13(A,c,c,e,s,s,C,o,n,t,r,o,l);
144 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);
145 static const XMLCh Host[] = UNICODE_LITERAL_4(H,o,s,t);
146 static const XMLCh HostRegex[] = UNICODE_LITERAL_9(H,o,s,t,R,e,g,e,x);
147 static const XMLCh htaccess[] = UNICODE_LITERAL_8(h,t,a,c,c,e,s,s);
148 static const XMLCh ignoreCase[] = UNICODE_LITERAL_10(i,g,n,o,r,e,C,a,s,e);
149 static const XMLCh ignoreOption[] = UNICODE_LITERAL_1(i);
150 static const XMLCh Path[] = UNICODE_LITERAL_4(P,a,t,h);
151 static const XMLCh PathRegex[] = UNICODE_LITERAL_9(P,a,t,h,R,e,g,e,x);
152 static const XMLCh Query[] = UNICODE_LITERAL_5(Q,u,e,r,y);
153 static const XMLCh name[] = UNICODE_LITERAL_4(n,a,m,e);
154 static const XMLCh regex[] = UNICODE_LITERAL_5(r,e,g,e,x);
155 static const XMLCh _type[] = UNICODE_LITERAL_4(t,y,p,e);
158 void SHIBSP_API shibsp::registerRequestMappers()
160 SPConfig& conf=SPConfig::getConfig();
161 conf.RequestMapperManager.registerFactory(XML_REQUEST_MAPPER, XMLRequestMapperFactory);
162 conf.RequestMapperManager.registerFactory(NATIVE_REQUEST_MAPPER, XMLRequestMapperFactory);
165 void Override::loadACL(const DOMElement* e, Category& log)
168 const DOMElement* acl=XMLHelper::getFirstChildElement(e,htaccess);
170 log.info("building Apache htaccess AccessControl provider...");
171 m_acl=SPConfig::getConfig().AccessControlManager.newPlugin(HT_ACCESS_CONTROL,acl);
174 acl=XMLHelper::getFirstChildElement(e,_AccessControl);
176 log.info("building XML-based AccessControl provider...");
177 m_acl=SPConfig::getConfig().AccessControlManager.newPlugin(XML_ACCESS_CONTROL,acl);
180 acl=XMLHelper::getFirstChildElement(e,AccessControlProvider);
182 auto_ptr_char type(acl->getAttributeNS(NULL,_type));
183 log.info("building AccessControl provider of type %s...",type.get());
184 m_acl=SPConfig::getConfig().AccessControlManager.newPlugin(type.get(),acl);
189 catch (exception& ex) {
190 log.crit("exception building AccessControl provider: %s", ex.what());
191 m_acl = new AccessControlDummy();
195 Override::Override(const DOMElement* e, Category& log, const Override* base) : m_acl(NULL)
198 // Load the property set.
202 // Load any AccessControl provider.
205 // Handle nested Paths.
206 DOMElement* path = XMLHelper::getFirstChildElement(e,Path);
207 for (int i=1; path; ++i, path=XMLHelper::getNextSiblingElement(path,Path)) {
208 const XMLCh* n=path->getAttributeNS(NULL,name);
210 // Skip any leading slashes.
211 while (n && *n==chForwardSlash)
214 // Check for empty name.
216 log.warn("skipping Path element (%d) with empty name attribute", i);
220 // Check for an embedded slash.
221 int slash=XMLString::indexOf(n,chForwardSlash);
223 // Copy the first path segment.
224 XMLCh* namebuf=new XMLCh[slash + 1];
225 for (int pos=0; pos < slash; pos++)
227 namebuf[slash]=chNull;
229 // Move past the slash in the original pathname.
232 // Skip any leading slashes again.
233 while (*n==chForwardSlash)
237 // Create a placeholder Path element for the first path segment and replant under it.
238 DOMElement* newpath=path->getOwnerDocument()->createElementNS(shibspconstants::SHIB2SPCONFIG_NS,Path);
239 newpath->setAttributeNS(NULL,name,namebuf);
240 path->setAttributeNS(NULL,name,n);
241 path->getParentNode()->replaceChild(newpath,path);
242 newpath->appendChild(path);
244 // Repoint our locals at the new parent.
246 n=path->getAttributeNS(NULL,name);
249 // All we had was a pathname with trailing slash(es), so just reset it without them.
250 path->setAttributeNS(NULL,name,namebuf);
251 n=path->getAttributeNS(NULL,name);
256 Override* o=new Override(path,log,this);
257 pair<bool,const char*> name=o->getString("name");
258 char* dup=strdup(name.second);
259 for (char* pch=dup; *pch; pch++)
261 if (m_map.count(dup)) {
262 log.warn("skipping duplicate Path element (%s)",dup);
268 log.debug("added Path mapping (%s)", dup);
272 if (!XMLString::equals(e->getLocalName(), PathRegex)) {
273 // Handle nested PathRegexs.
274 path = XMLHelper::getFirstChildElement(e,PathRegex);
275 for (int i=1; path; ++i, path=XMLHelper::getNextSiblingElement(path,PathRegex)) {
276 const XMLCh* n=path->getAttributeNS(NULL,regex);
278 log.warn("skipping PathRegex element (%d) with empty regex attribute",i);
282 auto_ptr<Override> o(new Override(path,log,this));
284 const XMLCh* flag=path->getAttributeNS(NULL,ignoreCase);
286 auto_ptr<RegularExpression> re(
287 new RegularExpression(n, (flag && (*flag==chLatin_f || *flag==chDigit_0)) ? &chNull : ignoreOption)
289 m_regexps.push_back(make_pair(re.release(), o.release()));
291 catch (XMLException& ex) {
292 auto_ptr_char tmp(ex.getMessage());
293 log.error("caught exception while parsing PathRegex regular expression (%d): %s", i, tmp.get());
294 throw ConfigurationException("Invalid regular expression in PathRegex element.");
297 if (log.isDebugEnabled())
298 log.debug("added <PathRegex> mapping (%s)", m_regexps.back().second->getString("regex").second);
302 // Handle nested Querys.
303 path = XMLHelper::getFirstChildElement(e,Query);
304 for (int i=1; path; ++i, path=XMLHelper::getNextSiblingElement(path,Query)) {
305 const XMLCh* n=path->getAttributeNS(NULL,name);
307 log.warn("skipping Query element (%d) with empty name attribute",i);
310 auto_ptr_char ntemp(n);
311 const XMLCh* v=path->getAttributeNS(NULL,regex);
313 auto_ptr<Override> o(new Override(path,log,this));
315 RegularExpression* re = NULL;
317 re = new RegularExpression(v);
318 m_queries.push_back(make_pair(make_pair(string(ntemp.get()),re), o.release()));
320 catch (XMLException& ex) {
321 auto_ptr_char tmp(ex.getMessage());
322 log.error("caught exception while parsing Query regular expression (%d): %s", i, tmp.get());
323 throw ConfigurationException("Invalid regular expression in Query element.");
326 log.debug("added <Query> mapping (%s)", ntemp.get());
331 for_each(m_map.begin(),m_map.end(),xmltooling::cleanup_pair<string,Override>());
332 for (vector< pair<RegularExpression*,Override*> >::iterator i = m_regexps.begin(); i != m_regexps.end(); ++i) {
336 for (vector< pair< pair<string,RegularExpression*>,Override*> >::iterator j = m_queries.begin(); j != m_queries.end(); ++j) {
337 delete j->first.second;
344 Override::~Override()
347 for_each(m_map.begin(),m_map.end(),xmltooling::cleanup_pair<string,Override>());
348 for (vector< pair<RegularExpression*,Override*> >::iterator i = m_regexps.begin(); i != m_regexps.end(); ++i) {
352 for (vector< pair< pair<string,RegularExpression*>,Override*> >::iterator j = m_queries.begin(); j != m_queries.end(); ++j) {
353 delete j->first.second;
358 const Override* Override::locate(const HTTPRequest& request) const
360 // This function is confusing because it's *not* recursive.
361 // The whole path is tokenized and mapped in a loop, so the
362 // path parameter starts with the entire request path and
363 // we can skip the leading slash as irrelevant.
364 const char* path = request.getRequestURI();
368 // Now we copy the path, chop the query string, and lower case it.
369 char* dup=strdup(path);
370 char* sep=strchr(dup,'?');
373 for (char* pch=dup; *pch; pch++)
376 // Default is for the current object to provide settings.
377 const Override* o=this;
379 // Tokenize the path by segment and try and map each segment.
382 const char* token=strtok_r(dup,"/",&pos);
384 const char* token=strtok(dup,"/");
387 map<string,Override*>::const_iterator i=o->m_map.find(token);
388 if (i==o->m_map.end())
389 break; // Once there's no match, we've consumed as much of the path as possible here.
390 // We found a match, so reset the settings pointer.
393 // We descended a step down the path, so we need to advance the original
394 // parameter for the regex step later.
395 path += strlen(token);
399 // Get the next segment, if any.
401 token=strtok_r(NULL,"/",&pos);
403 token=strtok(NULL,"/");
409 // If there's anything left, we try for a regex match on the rest of the path minus the query string.
412 path2 = path2.substr(0,path2.find('?'));
414 for (vector< pair<RegularExpression*,Override*> >::const_iterator re = o->m_regexps.begin(); re != o->m_regexps.end(); ++re) {
415 if (re->first->matches(path2.c_str())) {
422 // Finally, check for query string matches. This is another "unrolled" recursive descent in a loop.
426 for (vector< pair< pair<string,RegularExpression*>,Override*> >::const_iterator q = o->m_queries.begin(); !descended && q != o->m_queries.end(); ++q) {
427 vector<const char*> vals;
428 if (request.getParameters(q->first.first.c_str(), vals)) {
429 if (q->first.second) {
430 // We have to match one of the values.
431 for (vector<const char*>::const_iterator v = vals.begin(); v != vals.end(); ++v) {
432 if (q->first.second->matches(*v)) {
440 // The simple presence of the parameter is sufficient to match.
451 XMLRequestMapperImpl::XMLRequestMapperImpl(const DOMElement* e, Category& log) : m_document(NULL)
454 xmltooling::NDC ndc("XMLRequestMapperImpl");
457 // Load the property set.
460 // Load any AccessControl provider.
463 // Loop over the HostRegex elements.
464 const DOMElement* host = XMLHelper::getFirstChildElement(e,HostRegex);
465 for (int i=1; host; ++i, host=XMLHelper::getNextSiblingElement(host,HostRegex)) {
466 const XMLCh* n=host->getAttributeNS(NULL,regex);
468 log.warn("Skipping HostRegex element (%d) with empty regex attribute",i);
472 auto_ptr<Override> o(new Override(host,log,this));
474 const XMLCh* flag=host->getAttributeNS(NULL,ignoreCase);
476 auto_ptr<RegularExpression> re(
477 new RegularExpression(n, (flag && (*flag==chLatin_f || *flag==chDigit_0)) ? &chNull : ignoreOption)
479 m_regexps.push_back(make_pair(re.release(), o.release()));
481 catch (XMLException& ex) {
482 auto_ptr_char tmp(ex.getMessage());
483 log.error("caught exception while parsing HostRegex regular expression (%d): %s", i, tmp.get());
486 log.debug("Added <HostRegex> mapping for %s", m_regexps.back().second->getString("regex").second);
489 // Loop over the Host elements.
490 host = XMLHelper::getFirstChildElement(e,Host);
491 for (int i=1; host; ++i, host=XMLHelper::getNextSiblingElement(host,Host)) {
492 const XMLCh* n=host->getAttributeNS(NULL,name);
494 log.warn("Skipping Host element (%d) with empty name attribute",i);
498 Override* o=new Override(host,log,this);
499 pair<bool,const char*> name=o->getString("name");
500 pair<bool,const char*> scheme=o->getString("scheme");
501 pair<bool,const char*> port=o->getString("port");
503 char* dup=strdup(name.second);
504 for (char* pch=dup; *pch; pch++)
506 auto_ptr<char> dupwrap(dup);
508 if (!scheme.first && port.first) {
509 // No scheme, but a port, so assume http.
510 scheme = pair<bool,const char*>(true,"http");
512 else if (scheme.first && !port.first) {
513 // Scheme, no port, so default it.
514 // XXX Use getservbyname instead?
516 if (!strcmp(scheme.second,"http"))
518 else if (!strcmp(scheme.second,"https"))
520 else if (!strcmp(scheme.second,"ftp"))
522 else if (!strcmp(scheme.second,"ldap"))
524 else if (!strcmp(scheme.second,"ldaps"))
529 string url(scheme.second);
530 url=url + "://" + dup;
532 // Is this the default port?
533 if ((!strcmp(scheme.second,"http") && !strcmp(port.second,"80")) ||
534 (!strcmp(scheme.second,"https") && !strcmp(port.second,"443")) ||
535 (!strcmp(scheme.second,"ftp") && !strcmp(port.second,"21")) ||
536 (!strcmp(scheme.second,"ldap") && !strcmp(port.second,"389")) ||
537 (!strcmp(scheme.second,"ldaps") && !strcmp(port.second,"636"))) {
538 // First store a port-less version.
539 if (m_map.count(url) || m_extras.count(url)) {
540 log.warn("Skipping duplicate Host element (%s)",url.c_str());
545 log.debug("Added <Host> mapping for %s",url.c_str());
547 // Now append the port. We use the extras vector, to avoid double freeing the object later.
548 url=url + ':' + port.second;
550 log.debug("Added <Host> mapping for %s",url.c_str());
553 url=url + ':' + port.second;
554 if (m_map.count(url) || m_extras.count(url)) {
555 log.warn("Skipping duplicate Host element (%s)",url.c_str());
560 log.debug("Added <Host> mapping for %s",url.c_str());
564 // No scheme or port, so we enter dual hosts on http:80 and https:443
565 string url("http://");
567 if (m_map.count(url) || m_extras.count(url)) {
568 log.warn("Skipping duplicate Host element (%s)",url.c_str());
573 log.debug("Added <Host> mapping for %s",url.c_str());
576 if (m_map.count(url) || m_extras.count(url)) {
577 log.warn("Skipping duplicate Host element (%s)",url.c_str());
581 log.debug("Added <Host> mapping for %s",url.c_str());
585 if (m_map.count(url) || m_extras.count(url)) {
586 log.warn("Skipping duplicate Host element (%s)",url.c_str());
590 log.debug("Added <Host> mapping for %s",url.c_str());
593 if (m_map.count(url) || m_extras.count(url)) {
594 log.warn("Skipping duplicate Host element (%s)",url.c_str());
598 log.debug("Added <Host> mapping for %s",url.c_str());
603 const Override* XMLRequestMapperImpl::findOverride(const char* vhost, const HTTPRequest& request) const
605 const Override* o=NULL;
606 map<string,Override*>::const_iterator i=m_map.find(vhost);
610 i=m_extras.find(vhost);
611 if (i!=m_extras.end())
614 for (vector< pair<RegularExpression*,Override*> >::const_iterator re = m_regexps.begin(); !o && re != m_regexps.end(); ++re) {
615 if (re->first->matches(vhost))
621 return o ? o->locate(request) : this;
624 pair<bool,DOMElement*> XMLRequestMapper::load()
626 // Load from source using base class.
627 pair<bool,DOMElement*> raw = ReloadableXMLFile::load();
629 // If we own it, wrap it.
630 XercesJanitor<DOMDocument> docjanitor(raw.first ? raw.second->getOwnerDocument() : NULL);
632 XMLRequestMapperImpl* impl = new XMLRequestMapperImpl(raw.second,m_log);
634 // If we held the document, transfer it to the impl. If we didn't, it's a no-op.
635 impl->setDocument(docjanitor.release());
640 return make_pair(false,(DOMElement*)NULL);
643 RequestMapper::Settings XMLRequestMapper::getSettings(const HTTPRequest& request) const
646 vhost << request.getScheme() << "://" << request.getHostname() << ':' << request.getPort();
647 const Override* o=m_impl->findOverride(vhost.str().c_str(), request);
648 return Settings(o,o->getAC());