Apache Shiro 学习记录1

　　最近几天在学习Apache Shiro......看了一些大神们的教程.....感觉收获不少.....但是毕竟教程也只是指引一下方向....即使是精品教程,仍然有很多东西都没有说明....所以自己也稍微研究了一下...记录了一下我的研究发现....教程点这里

　　这篇教程的最后提到了strategy.....然后给出了4个方法.....但是并没有怎么详细说明.....我想说说我的理解.....(我的理解可能会有很多错误)

---

　　我想先说说登陆验证的大致流程....大致......

Subject

从用户那里收集完用户名密码以后我们会调用subject.login(token)这个方法去登陆.....Subject是一个接口,没有定义login的具体实现.....Shiro里只有一个类实现了这个接口,是DelegatingSubject这个类.这个类里的方法login方法如下:

   public void login(AuthenticationToken token) throws AuthenticationException {
        clearRunAsIdentitiesInternal();
        Subject subject = securityManager.login(this, token);

        PrincipalCollection principals;

        String host = null;

        if (subject instanceof DelegatingSubject) {
            DelegatingSubject delegating = (DelegatingSubject) subject;
            //we have to do this in case there are assumed identities - we don't want to lose the 'real' principals:
            principals = delegating.principals;
            host = delegating.host;
        } else {
            principals = subject.getPrincipals();
        }

        if (principals == null || principals.isEmpty()) {
            String msg = "Principals returned from securityManager.login( token ) returned a null or " +
                    "empty value.  This value must be non null and populated with one or more elements.";
            throw new IllegalStateException(msg);
        }
        this.principals = principals;
        this.authenticated = true;
        if (token instanceof HostAuthenticationToken) {
            host = ((HostAuthenticationToken) token).getHost();
        }
        if (host != null) {
            this.host = host;
        }
        Session session = subject.getSession(false);
        if (session != null) {
            this.session = decorate(session);
        } else {
            this.session = null;
        }
    }

代码各种复杂=.= ..............我也看不懂....但是我看到第三行,明白了subject其实也是让securityManager来执行login操作的......那我们去看看securityManager吧......

SecurityManager

SecurityManager也是一个接口,各种继承,实现其他接口......还好实现类只有一个DefaultSecurityManager类...并且login方法直到这个类才被实现....

    public Subject login(Subject subject, AuthenticationToken token) throws AuthenticationException {
        AuthenticationInfo info;
        try {
            info = authenticate(token);
        } catch (AuthenticationException ae) {
            try {
                onFailedLogin(token, ae, subject);
            } catch (Exception e) {
                if (log.isInfoEnabled()) {
                    log.info("onFailedLogin method threw an " +
                            "exception.  Logging and propagating original AuthenticationException.", e);
                }
            }
            throw ae; //propagate
        }

        Subject loggedIn = createSubject(token, info, subject);

        onSuccessfulLogin(token, info, loggedIn);

        return loggedIn;
    }

我觉得这里最重要的就是第四行的authenticate方法.......

这个方法是在DefaultSecurityManager的N层父类AuthenticatingSecurityManager里实现的....

    public AuthenticationInfo authenticate(AuthenticationToken token) throws AuthenticationException {
        return this.authenticator.authenticate(token);
    }

这里又可以看出SecurityManager的login方法其实也是委托给认证器authenticator调用authenticate(token)方法来实现的.

public abstract class AuthenticatingSecurityManager extends RealmSecurityManager {

    /**
     * The internal <code>Authenticator</code> delegate instance that this SecurityManager instance will use
     * to perform all authentication operations.
     */
    private Authenticator authenticator;

    /**
     * Default no-arg constructor that initializes its internal
     * <code>authenticator</code> instance to a
     * {@link org.apache.shiro.authc.pam.ModularRealmAuthenticator ModularRealmAuthenticator}.
     */
    public AuthenticatingSecurityManager() {
        super();
        this.authenticator = new ModularRealmAuthenticator();
    }
    ..................
}

认证器authenticator的默认实现是ModularRealmAuthenticator....this.authenticator = new ModularRealmAuthenticator();...所以我们再来看看ModularRealmAuthenticator..........

ModularRealmAuthenticator

ModularRealmAuthenticator的authenticate方法是继承自父类AbstractAuthenticator的..这个方法还是final的...

    public final AuthenticationInfo authenticate(AuthenticationToken token) throws AuthenticationException {

        if (token == null) {
            throw new IllegalArgumentException("Method argumet (authentication token) cannot be null.");
        }

        log.trace("Authentication attempt received for token [{}]", token);

        AuthenticationInfo info;
        try {
            info = doAuthenticate(token);
            if (info == null) {
                String msg = "No account information found for authentication token [" + token + "] by this " +
                        "Authenticator instance.  Please check that it is configured correctly.";
                throw new AuthenticationException(msg);
            }
        } catch (Throwable t) {
            AuthenticationException ae = null;
            if (t instanceof AuthenticationException) {
                ae = (AuthenticationException) t;
            }
            if (ae == null) {
                //Exception thrown was not an expected AuthenticationException.  Therefore it is probably a little more
                //severe or unexpected.  So, wrap in an AuthenticationException, log to warn, and propagate:
                String msg = "Authentication failed for token submission [" + token + "].  Possible unexpected " +
                        "error? (Typical or expected login exceptions should extend from AuthenticationException).";
                ae = new AuthenticationException(msg, t);
            }
            try {
                notifyFailure(token, ae);
            } catch (Throwable t2) {
                if (log.isWarnEnabled()) {
                    String msg = "Unable to send notification for failed authentication attempt - listener error?.  " +
                            "Please check your AuthenticationListener implementation(s).  Logging sending exception " +
                            "and propagating original AuthenticationException instead...";
                    log.warn(msg, t2);
                }
            }


            throw ae;
        }

        log.debug("Authentication successful for token [{}].  Returned account [{}]", token, info);

        notifySuccess(token, info);

        return info;
    }

从中我们可以看出比较重要的是第11行 info = doAuthenticate(token);

doAuthenticate(token)方法在ModularRealmAuthenticator之中被override过...

    protected AuthenticationInfo doAuthenticate(AuthenticationToken authenticationToken) throws AuthenticationException {
        assertRealmsConfigured();
        Collection<Realm> realms = getRealms();
        if (realms.size() == 1) {
            return doSingleRealmAuthentication(realms.iterator().next(), authenticationToken);
        } else {
            return doMultiRealmAuthentication(realms, authenticationToken);
        }
    }

这里还是比较明显的,根据定义的Realm数量来决定是调用doSingleRealmAuthentication方法还是调用doMultiRealmAuthentication方法...

我们来看看doMultiRealmAuthentication方法里到底做了些什么呢..这个方法是定义在ModularRealmAuthenticator里的...

    protected AuthenticationInfo doMultiRealmAuthentication(Collection<Realm> realms, AuthenticationToken token) {

        AuthenticationStrategy strategy = getAuthenticationStrategy();

        AuthenticationInfo aggregate = strategy.beforeAllAttempts(realms, token);

        if (log.isTraceEnabled()) {
            log.trace("Iterating through {} realms for PAM authentication", realms.size());
        }

        for (Realm realm : realms) {

            aggregate = strategy.beforeAttempt(realm, token, aggregate);

            if (realm.supports(token)) {

                log.trace("Attempting to authenticate token [{}] using realm [{}]", token, realm);

                AuthenticationInfo info = null;
                Throwable t = null;
                try {
                    info = realm.getAuthenticationInfo(token);
                } catch (Throwable throwable) {
                    t = throwable;
                    if (log.isDebugEnabled()) {
                        String msg = "Realm [" + realm + "] threw an exception during a multi-realm authentication attempt:";
                        log.debug(msg, t);
                    }
                }

                aggregate = strategy.afterAttempt(realm, token, info, aggregate, t);

            } else {
                log.debug("Realm [{}] does not support token {}.  Skipping realm.", realm, token);
            }
        }

        aggregate = strategy.afterAllAttempts(token, aggregate);

        return aggregate;
    }

虽然很多看不明白....但是我大致流程能看懂:

在调用每个Realm之前,先执行strategy.beforeAllAttempts(realms, token);

然后再进入 for (Realm realm : realms),就是准备去调用每个realm来认证..不过每个Realm认证之前还要调用aggregate = strategy.beforeAttempt(realm, token, aggregate);

然后得到单个Realm的认证结果 info = realm.getAuthenticationInfo(token);

再调用aggregate = strategy.afterAttempt(realm, token, info, aggregate, t);这里可能会把info的认证信息和aggregate合并,就是把principle合并....不过还是要看具体的strategy的.

最后再调用 aggregate = strategy.afterAllAttempts(token, aggregate);

也就是说strategy.beforeAllAttempts(realms, token);调用1次,然后根据realm的数量调用N次strategy.beforeAttempt(realm, token, aggregate);N次realm.getAuthenticationInfo(token);N次 strategy.afterAttempt(realm, token, info, aggregate, t);再调用1次strategy.afterAllAttempts(token, aggregate);

另外,不是说你单个Realm里认证失败,比如抛出了UnknownAccountException..所有认证就终止了....仔细看第23,24行会发现Realm里抛出的异常是会被拦截下来的...然后传给了 strategy.afterAttempt(realm, token, info, aggregate, t);....发现了这个异常以后到底是算认证失败了还是继续看其他的Realm是根据strategy来的...比如AllSuccessfulStrategy的afterAttempt方法...发现了任意1个realm抛出的异常,就直接抛出异常,终止认证....但是AtLeastOneSuccessfulStrategy就不一样...它就会无视这个异常...继续下面realm来认证...然后根据afterAllAttempts里的AuthenticationInfo的汇总信息来判断到底是认证失败了要抛出异常还是认证是成功的....

Shiro默认的策略是AtLeastOneSuccessfulStrategy

---

各种Strategy

我觉得看到这里我们可以再来看看Shiro已经定义的三种Strategy了...教程里提到过:

FirstSuccessfulStrategy：只要有一个Realm验证成功即可，只返回第一个Realm身份验证成功的认证信息，其他的忽略；

AtLeastOneSuccessfulStrategy：只要有一个Realm验证成功即可，和FirstSuccessfulStrategy不同，返回所有Realm身份验证成功的认证信息；

AllSuccessfulStrategy：所有Realm验证成功才算成功，且返回所有Realm身份验证成功的认证信息，如果有一个失败就失败了。

这3种策略都继承自AbstractAuthenticationStrategy类..这个抽象类里定义了前面提到的2个after方法和2个before方法还有1个merge方法...

    protected AuthenticationInfo merge(AuthenticationInfo info, AuthenticationInfo aggregate) {
        if( aggregate instanceof MergableAuthenticationInfo ) {
            ((MergableAuthenticationInfo)aggregate).merge(info);
            return aggregate;
        } else {
            throw new IllegalArgumentException( "Attempt to merge authentication info from multiple realms, but aggregate " +
                      "AuthenticationInfo is not of type MergableAuthenticationInfo." );
        }
    }

merge方法会调用AuthenticationInfo的merge方法....大致过程是:

  public void merge(AuthenticationInfo info) {
        if (info == null || info.getPrincipals() == null || info.getPrincipals().isEmpty()) {
            return;
        }

        if (this.principals == null) {
            this.principals = info.getPrincipals();
        } else {
            if (!(this.principals instanceof MutablePrincipalCollection)) {
                this.principals = new SimplePrincipalCollection(this.principals);
            }
            ((MutablePrincipalCollection) this.principals).addAll(info.getPrincipals());
        }
    ......................
}

大致意思就是把info的principals合并到aggregate的principals上去....

AbstractAuthenticationStrategy的merge方法主要是在afterAttempt(Realm realm, AuthenticationToken token, AuthenticationInfo singleRealmInfo, AuthenticationInfo aggregateInfo, Throwable t)里被调用,因为经过一个Realm的认证以后会得到info对象...这个时候不同的认证策略就需要考虑是不是要把当前得到的info合并到已经有的前面几个Realm积累下来的aggregateInfo里去了...

举例:

AllSuccessfulStrategy类

    public AuthenticationInfo beforeAttempt(Realm realm, AuthenticationToken token, AuthenticationInfo info) throws AuthenticationException {
        if (!realm.supports(token)) {
            String msg = "Realm [" + realm + "] of type [" + realm.getClass().getName() + "] does not support " +
                    " the submitted AuthenticationToken [" + token + "].  The [" + getClass().getName() +
                    "] implementation requires all configured realm(s) to support and be able to process the submitted " +
                    "AuthenticationToken.";
            throw new UnsupportedTokenException(msg);
        }

        return info;
    }

第2行,如果realm不支持token,显然这个realm是认证失败的...而AllSuccessfulStrategy策略需要所有Realm都认证成功才行...果断抛出异常...

    public AuthenticationInfo afterAttempt(Realm realm, AuthenticationToken token, AuthenticationInfo info, AuthenticationInfo aggregate, Throwable t)
            throws AuthenticationException {
        if (t != null) {
            if (t instanceof AuthenticationException) {
                //propagate:
                throw ((AuthenticationException) t);
            } else {
                String msg = "Unable to acquire account data from realm [" + realm + "].  The [" +
                        getClass().getName() + " implementation requires all configured realm(s) to operate successfully " +
                        "for a successful authentication.";
                throw new AuthenticationException(msg, t);
            }
        }
        if (info == null) {
            String msg = "Realm [" + realm + "] could not find any associated account data for the submitted " +
                    "AuthenticationToken [" + token + "].  The [" + getClass().getName() + "] implementation requires " +
                    "all configured realm(s) to acquire valid account data for a submitted token during the " +
                    "log-in process.";
            throw new UnknownAccountException(msg);
        }

        log.debug("Account successfully authenticated using realm [{}]", realm);

        // If non-null account is returned, then the realm was able to authenticate the
        // user - so merge the account with any accumulated before:
        merge(info, aggregate);

        return aggregate;
    }

Throwable t, t是从Realm里抛出来的,如果Realm抛出了异常,不管什么原因,认证肯定是失败了...所以Strategy也要抛出异常,即认证失败了......info = null的道理也是一样的....

AtLeastOneSuccessfulStrategy类:

    public AuthenticationInfo afterAllAttempts(AuthenticationToken token, AuthenticationInfo aggregate) throws AuthenticationException {
        //we know if one or more were able to succesfully authenticate if the aggregated account object does not
        //contain null or empty data:
        if (aggregate == null || CollectionUtils.isEmpty(aggregate.getPrincipals())) {
            throw new AuthenticationException("Authentication token of type [" + token.getClass() + "] " +
                    "could not be authenticated by any configured realms.  Please ensure that at least one realm can " +
                    "authenticate these tokens.");
        }

        return aggregate;
    }

所有Realm的认证都执行完毕之后,如果AuthenticationInfo 的合并结果aggregate还是null或者空的话那么说明所有Realm的认证都失败的...那么就应该抛出异常,说明认证失败了....

FirstSuccessfulStrategy类:

public class FirstSuccessfulStrategy extends AbstractAuthenticationStrategy {

    /**
     * Returns {@code null} immediately, relying on this class's {@link #merge merge} implementation to return
     * only the first {@code info} object it encounters, ignoring all subsequent ones.
     */
    public AuthenticationInfo beforeAllAttempts(Collection<? extends Realm> realms, AuthenticationToken token) throws AuthenticationException {
        return null;
    }

    /**
     * Returns the specified {@code aggregate} instance if is non null and valid (that is, has principals and they are
     * not empty) immediately, or, if it is null or not valid, the {@code info} argument is returned instead.
     * <p/>
     * This logic ensures that the first valid info encountered is the one retained and all subsequent ones are ignored,
     * since this strategy mandates that only the info from the first successfully authenticated realm be used.
     */
    protected AuthenticationInfo merge(AuthenticationInfo info, AuthenticationInfo aggregate) {
        if (aggregate != null && !CollectionUtils.isEmpty(aggregate.getPrincipals())) {
            return aggregate;
        }
        return info != null ? info : aggregate;
    }
}

为什么要override beforeAllAttempts方法返回null我也不懂....父类是返回new SimpleAuthenticationInfo();.....

override merge方法是因为FirstSuccessfulStrategy策略只返回第一个验证成功的Realm的AuthenticationInfo ...

如果aggregate不是空的或者null,说明前面的Realm有成功过,那么根据策略,这个时候不应该把本次realm得到的info合并进去.....而是直接返回前面验证成功的AuthenticationInfo ......