In our system, virtual hosts configuration is stored in redis. During connection setup, when the SNI is received, we would like to query redis for the correct certificate and key pair to use for the TLS connection and create a new Context instance with that attached.
The bulk of the code is similar to the accepted answer here: Twisted listenSSL virtualhosts
The issue we are facing is that, since accessing the certificates involves an additional network operation, we would like to make the set_tlsext_servername_callback function return a deferred.
Is there a way to tell Twisted/pyOpenSSL to wait until the deferred fires?
Edit: I found this link which seems promising, but falls short of providing a solution: https://mta.openssl.org/pipermail/openssl-dev/2015-January/000480.html
You can find an example of Twisted and SNI here: https://pypi.python.org/pypi/txsni. I would really, really like that callback to be able to take a Deferred. I think that the way to do this would be to pause the underlying transport from delivering any further bytes either in or out (stopReading/stopWriting) and then resume when the Deferred fires, after doing the rest of the SNI dance. However, I'm not even sure if this is possible with OpenSSL, because the SNI is received with the rest of ClientHello and you may need to be able to react immediately to serve the correct certificate. In this worst-possible-case scenario, you could feed the first chunk of bytes you receive into a dummy memory-BIO, wait for the TLS handshake, throw it away and never deliver any generated responses, and then don't initialize your "real" sub-transport until you've decided on which context object to use.
Hope this helps - and if you figure it out, please contribute a patch to TxSNI or Twisted!
Related
I tried posting this on ask.openstack but it has been stuck in the moderator for 5 days now. I thought I'd try here.
I was trying to debug a Nova issue and wanted to decode the SSL / TLS packets being exchanged using Wireshark. Part of the changes I was making was setting Nova up to use SSL / TLS and I wanted to be sure that part of it I had set correctly. I eventually figure out my issues from the various log files but I'm somewhat assuming that being able to watch the network traffic may help in some very difficult cases.
The exchange uses TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384 at one point. According to this security stackexchannge question, there is a "pre-master secret" or various other terms. I've wrestled with this before in a previous life doing IPSec. Usually you can set debug in the application and it will spew out the secret into the log file. I tried "debug = true" under Default in nova.conf and got lots of debug but no secret. There was two items that looked interesting that were reported as **** in the log: keystone_authtoken.memcache_secret_key and neutron.metadata_proxy_shared_secret. I wasn't sure if those were the secrets I was looking for or not. In this case, I'm looking at the nova-api traffic going to port 8774.
Also, since all of openstack is Python and uses the same "request" and "certifi" packages, it may be possible to generalize this to all of the openstack components.
nova --version report 9.1.1
The similar problem is described here: GWT IllegalArgumentException: encodedRequest cannot be empty
My GWT application is deployed in Tomcat6, which is linked with Apache by using Coyote/JK2 connectors. For SSO I use the mod_auth_sspi/1.0.4.
When I use IE8, pages is not displayed, but for Firefox everything OK. In Tomcat logs I see the following:
SEVERE: Exception while dispatching incoming RPC call
java.lang.IllegalArgumentException: encodedRequest cannot be empty
at com.google.gwt.user.server.rpc.RPC.decodeRequest(RPC.java:232)
at org.spring4gwt.server.SpringGwtRemoteServiceServlet.processCall(SpringGwtRemoteServiceServlet.java:32)
at com.google.gwt.user.server.rpc.RemoteServiceServlet.processPost(RemoteServiceServlet.java:248)
at com.google.gwt.user.server.rpc.AbstractRemoteServiceServlet.doPost(AbstractRemoteServiceServlet.java:62)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:643)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:723)
at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:290)
at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:206)
at gov.department.it.server.RequestInterceptorFilter.doFilter(RequestInterceptorFilter.java:90)
at org.apache.catalina.core.ApplicationFilterChain.internalDoFilter(ApplicationFilterChain.java:235)
at org.apache.catalina.core.ApplicationFilterChain.doFilter(ApplicationFilterChain.java:206)
at org.apache.catalina.core.StandardWrapperValve.invoke(StandardWrapperValve.java:233)
at org.apache.catalina.core.StandardContextValve.invoke(StandardContextValve.java:191)
at org.apache.catalina.core.StandardHostValve.invoke(StandardHostValve.java:127)
at org.apache.catalina.valves.ErrorReportValve.invoke(ErrorReportValve.java:103)
at org.apache.catalina.core.StandardEngineValve.invoke(StandardEngineValve.java:109)
at org.apache.catalina.connector.CoyoteAdapter.service(CoyoteAdapter.java:293)
at org.apache.jk.server.JkCoyoteHandler.invoke(JkCoyoteHandler.java:190)
at org.apache.jk.common.HandlerRequest.invoke(HandlerRequest.java:311)
at org.apache.jk.common.ChannelSocket.invoke(ChannelSocket.java:776)
at org.apache.jk.common.ChannelSocket.processConnection(ChannelSocket.java:705)
at org.apache.jk.common.ChannelSocket$SocketConnection.runIt(ChannelSocket.java:898)
at org.apache.tomcat.util.threads.ThreadPool$ControlRunnable.run(ThreadPool.java:690)
at java.lang.Thread.run(Thread.java:619)
What have I tried so far:
1) Can't find the registry key DisableNTLMPreAuth (IMHO it's not the solution, because in my case IE 8 is actively used).
2) I have installed and configured the Native Windows Authentication Framework WAFFLE
web.xml:
...
<filter>
<filter-name>NegotiateSecurityFilter</filter-name>
<filter-class>waffle.servlet.NegotiateSecurityFilter</filter-class>
<init-param>
<param-name>waffle.servlet.spi.NegotiateSecurityFilterProvider/protocols</param-name>
<param-value>NTLM</param-value>
</init-param>
</filter>
...
<filter-mapping>
<filter-name>NegotiateSecurityFilter</filter-name>
<url-pattern>/my-app/*</url-pattern>
</filter-mapping>
...
But it did not help.
3) In worker.properties I set socket_keepalive=0, but it did not help too -
worker.ajp13.type=ajp13
worker.ajp13.host=localhost
worker.ajp13.port=8009
worker.ajp13.lbfactor=50
worker.ajp13.cachesize=10
worker.ajp13.cache_timeout=600
worker.ajp13.socket_keepalive=0
worker.ajp13.socket_timeout=300
What else can I try to do?
You have rediscovered the 7 year old bug #1 in mod_auth_sspi which has affected numerous projects, frustrated numerous developers, and caused uncountable wasted man-hours over the years. Yet it still stands unresolved because the maintainer doesn't consider it a bug. Nor has it been addressed by Microsoft for older browsers, because indications are that IE9 doesn't have this problem.
Cause
It is caused by IE trying to be 'smart' and sending a zero content-length POST (I named it 0POST to try making it an indexable term to benefit those who rediscover it in the next 7 years.) with an NTLM auth header in anticipation of being challenged by the server. IE does this when it has been authenticated before in that protection space. So it knows that it will be challenged again. Sadly mod_auth_sspi is not as smart as IE, so bad things happen on the server side when a 0POST arrives and it is let through to the apps without being challenged. Except that sometimes this can happen even for unprotected areas, if they are under an area that requires authentication.
Other browsers don't pretend to be as smart as IE and don't try to save a few bytes on the first round trip for "performance", so they don't run into this problem. Here is Microsoft's explanation of this behavior.
Horrible Workaround
In Apache httpd.conf set
SSPIPerRequestAuth On
This is equivalent to the DisableNTLMPreAuth IE client-side fix you mentioned, which is impractical for a large user group. Plus it amounts to crippling all non-Apache apps also, which may be capable of handling a 0POST. There are literally NO examples of this setting being discussed or its side effects explained on the web, so I am including this only link I found that sheds some light on it. Anyway, making one server side change seems to be the lesser of the two evils. Although now, by changing the server config, you have crippled all other innocent browsers visiting this site as well.
The problem with this workaround is that it forces EVERY request to perform an SSPI handshake which results in a lot of extra 401 traffic and can affect performance. For performance, NTLM authentication is treated as 'session-based' not 'request-based' which means that the handshake occurs only at the start of the session. When using this setting, you should also set filters to prevent your log filling up with 401s. Also note that this requires KeepAlive to be turned on.
I am not sure your setup is the same as the one described in the WAFFLE fix; were they using Apache like you? I think WAFFLE applies to Tomcat, whereas you have Apache in front, so Apache is handling authentication. You might consider using that setup instead of Apache. If you can use that setup, it may be a better option than this workaround because WAFFLE has explicitly accounted for 0POST and can handle it. The author had also discovered this gem while working with GWT like you.
Interestingly, for jcifs, a fix for this very issue was posted 9 Years ago. The author also provided an excellent explanation later:
The code in the filter examines all HTTP POST requests and determines
if they contain an NTLM type 1 message. If the request contains an
NTLM type 1 message, the filter responds with a dummy type 2 message
to entertain IE's desire to re-negotiate NTLM prior to submitting any
POST data. The browser should then respond with an NTLM type 3
message along with the post data which the filter should then allow to
chain to the rest of the web application.
A simple patch was also created for mod_auth_sspi 5 years ago, if you are interested. See diff in the author's own repo. I am not sure if I agree with that approach though. It tries to detect IE/0POST, whereas I think the right fix should be to detect if the client is requesting auth with a NTLM Type 1 header, as in the jcifs filter. (Type 1 simply means that it is the first message of the handshake)
I wonder if anyone has used alternatives to mod_auth_sspi like mod_auth_ntlm_winbind and if they don't exhibit this behavior. If you have, please leave a comment. We already know WAFFLE works, but it is not a mod_auth_sspi replacement.
One alternative is to forget NTLM and use Kerberos, (mod_auth_kerb) but many people find that too complicated to setup. IE will behave this way on any challenge-response scheme, so odds are that kerb auth could run into the same problem, since a similar 401 sequence happens in both cases. But being a different module, its possible it is capable of handling this.
Lastly, I should mention that there is yet another issue that this per-request auth workaround doesn't seem to fix. I haven't seen it discussed anywhere, but I have found that sometimes after the 0POST, the server waits for a very long time before it responds with the final 200 response with the results of the (proper) POST. This long delay happens only in the end though, NOT immediately in response to the 0POST. That goes fine, and the handshake completes, but the server doesn't respond until after a long wait which I have noticed is suspiciously close to 90 seconds, like some sort of timeout. The practical result of this is that when users log in, IE8 will sometimes hang for 90sec waiting for server response. I thought the KeepAlive might be causing it, but it is not even explicitly defined in my config, so I assume it is at the 15sec Apache default. But I am sure this is related to the 0POST, because it happens only right after a successful 0POST auth handshake. Our server is in a separate (2-way) trusted domain across a firewall, so maybe that has something to do with it.
Diverse Examples of This Issue
https://confluence.atlassian.com/display/JIRAKB/NullPointerException+when+Authenticating+from+IE
http://trac.edgewall.org/ticket/2696
http://trac.edgewall.org/ticket/4560
https://drupal.org/node/82530
http://www.webmasterworld.com/apache/3087425.htm
Why "Content-Length: 0" in POST requests?
https://jira.springsource.org/browse/SEC-1087
The most hilarious example is how IE's smartness affected Microsoft's own products! They themselves couldn't understand how to deal with IE's behavior, causing a bug in ISA Server 2006.
http://support.microsoft.com/kb/942638
I am trying to monitor https traffic with Fiddler, using current newest version:2.4.4.5
I've successfully set up https, certificates and I can see the full https encrypted traffic for example browsing my bank's web site.
...however...
When I trying to monitor an other server I got this error message in the response window:
"Failed to secure existing connection for 77.87.178.160. A call to SSPI failed, see inner exception. InnerException: System.ComponentModel.Win32Exception: The client and server cannot communicate, because they do not possess a common algorithm"
For full Fiddler window see:
The client is not a in this case browser, but a custom client program, which communicates with its own server.
My question: Is this exception misleading and in reality some other error prevents the secure channel to set up?
...or...
We have still chance to monitor this https communication?
Thx in advance
What is the client program?
This error typically indicates that that client application is only offering certain HTTPS ciphers, and those ciphers are not supported by Fiddler.
However, in this case, the specific problem here is almost certainly this: http://blogs.msdn.com/b/ieinternals/archive/2009/12/08/aes-is-not-a-valid-cipher-for-sslv3.aspx
The client is trying to use AES with SSLv3, but that isn't one of the valid ciphers for SSL3. As a consequence, the connection fails.
You might be able to workaround this by clicking Rules > Customize Rules. Scroll down to the Main() function and add the following line within the function:
CONFIG.oAcceptedServerHTTPSProtocols =
System.Security.Authentication.SslProtocols.Ssl3;
Please let me know if this works.
NOTE Current versions of Fiddler offer a UI link for this: Look at the lis of enabled protocols on the HTTPS tab.
Unbelievably this issue is still present some 6 years later.
Just installed the latest version of Fiddle (v5.0.20194.41348), and sure enough on Win7 using Chrome or IE it keeps failing with the dreaded error:
"fiddler.network.https> HTTPS handshake to google.com (for #1) failed. System.ComponentModel.Win32Exception The client and server cannot communicate, because they do not possess a common algorithm"
After some hours of testing, I found a middle ground solution which seems to work with virtually all websites. The aim was to get the highest possible security with no errors in the log. Without needing to add any code, simply changing this line under Tools > Options > HTTPS > Protocols is what worked for me (just copy and paste it):
<client>;ssl3;tls1.1;tls1.2
Basically removed the ssl2 and tls1.0 protocols which leaves us with some pretty decent security and no errors so far. Having spent hours of frustration with this error, hope someone out there might find this useful, and a big thanks to EricLaw who discovered the root of the problem.
Yes I too have seen this error when working outside of fiddler and it was connected with AuthenticateAsServer but only went wrong when using IE10 and not Chrome as the browser.
Odd thing is that it did not break all the time for IE10 using SslProtocols.Tls for the protocol so I will add a bit of code to switch the protocol if one fails
The protocol that can be used also seems to change on if you are using a proxy server like Fiddler or using an invisible server by hijacking the DNS via the hosts file to divert traffic to the server
I'm using a Windows certification authority (AD CS) to issue certificates for the participants in a secure connection between a RabbitMQ Windows service and a client.
The Subject of my client certificate is my Distinguished Name (DN) in LDAP:
"CN=John Ruiz,CN=Users,DC=devexample,DC=com"
When I attempt to establish this connection, the server throws an exception and closes the connection and I see this erlang stack trace in the rabbit log:
=ERROR REPORT==== 30-Dec-2011::10:33:24 ===
exception on TCP connection <0.331.0> from 10.1.30.70:52269
{channel0_error,starting,
{error,{case_clause,[{printableString,"Users"},
{printableString,"John Ruiz"}]},
'connection.start_ok',
[{rabbit_ssl,find_by_type,2,[]},
{rabbit_auth_mechanism_ssl,init,1,[]},
{rabbit_reader,handle_method0,2,[]},
{rabbit_reader,handle_method0,3,[]},
{rabbit_reader,handle_input,3,[]},
{rabbit_reader,recvloop,2,[]},
{rabbit_reader,start_connection,7,[]},
{proc_lib,init_p_do_apply,3,
[{file,"proc_lib.erl"},{line,227}]}]}}
Looking through the last two lines in the stack trace, I found the two files involved:
rabbit_ssl.erl
rabbit_auth_mechanism_ssl.erl
The problem is that I've neither read nor written erlang before, so I don't know why find_by_type is throwing an exception. My best guess is that since there are two CN=* elements in the list of relative DNs (RDNs), that the result of the call to lists:flatten is an array whereas the expected result is a scalar.
Can someone familiar with erlang please confirm or correct my assumption? If you see a way in which this code could be improved to handle the case I've just described (instead of throwing an exception), I would really appreciate it so that I can suggest it on the RabbitMQ mailing list.
Your guess is correct. It crashes because there are two CN=* elements. Looking at the code it seems like a lot of it depends on there only being one CN. CN itself is used as the username for the ssl session I think so having multiple makes little sense.
I recently needed to configure CocoaHttpServer, which we're using in our application with success, to handle HTTPS connections coming from a client application (running on Android devices). This is fine - there is copious sample code which allows for this, and we were able to enable the secure server without issue.
In practice we were seeing incredibly long SSL negotiation phases while the client was doing its handshaking with our server - upwards of 70 seconds.
Through a long series of searches, I found that the delay was because of the calculation of Diffie-Hellman parameters used by default when SSL is enabled in CFSocket. This thread is where I first started to find the answer to my issue.
To match what our Windows server was doing (using a less-secure SSL cipher) I needed to set the cipher explicitly on the Mac, which isn't easy when using AsyncSocket as a wrapper for the socket communications.
Our Windows server was using:
TLS_RSA_WITH_RC4_128_MD5 )(0x04)
RC4 128 bits MD5 RSA
Our Macintosh server was using:
TLS_DHE_RSA_WITH_AES_256_CBC_SHA (0x039)
AES 256 bits SHA-1 Ephemeral Diffie-Hellman key exchange using RSA certificate
The difference in "security" is large, but likely not worth the effort/computation/delay that we were seeing. Security Theater?
Please note that there are different ciphers that can be chosen - I chose to use the same one as our Windows implementation for consistency.
With information from another question mentioned above, I figured out how to set the cipher for CFSocket to use the same as Windows, and the code appears to be now quite a bit better - like it really works! CFSocket isn't directly exposing the SecureTransport support, which makes this kind of hard, but defining a particular key makes it work nicely.
For posterity, here's the code I've added to -onSocketWillConnect: in our HTTPConnection class:
// define this value; it isn't exposed by CFSocketStream.h
const extern CFStringRef kCFStreamPropertySocketSSLContext;
...
CFReadStreamRef stream = [sock getCFReadStream];
CFDataRef data = (CFDataRef) CFReadStreamCopyProperty(stream, kCFStreamPropertySocketSSLContext);
// Extract the SSLContextRef from the CFData
SSLContextRef sslContext;
CFDataGetBytes(data, CFRangeMake(0, sizeof(SSLContextRef)), (UInt8*)&sslContext);
SSLCipherSuite *ciphers = (SSLCipherSuite *)malloc(1 * sizeof(SSLCipherSuite));
ciphers[0] = SSL_RSA_WITH_RC4_128_MD5; // Basic cipher - not Diffie-Hellman
SSLSetEnabledCiphers(sslContext, ciphers, 1);
I hope this helps anyone working through the same issue as I - I'd be happy to share some more code and advice if needed.
For what it's worth, I contributed a patch to CocoaAsyncSocket about a week before you had this issue. Sorry that I didn't notice your question back then. :-)