- stunnel-users - stunnel.org

Re: [stunnel-users] stunnel-users Digest, Vol 183, Issue 4
by Brent Kimberley 26 Oct '19

26 Oct '19

Chris.Does this help? https://stackoverflow.com/questions/40454338/no-shared-cipher-at-ssl-accept… If you can rule out libraries like OpenSSL, then look at your config + initial setup exchange. Date: Fri, 25 Oct 2019 11:11:54 -0400 From: Christopher Schultz <chris(a)christopherschultz.net> Subject: Re: [stunnel-users] Strange connection failure in one environmentOn 10/25/19 10:09, Christopher Schultz wrote: > All, > > I've been using stunnel in two environments (dev, prod) for a long time > without any problems. Recently, my dev environment started acting funny > and I can't connect to it from outside the box. > > Can someone take a look and let me know if you have any suggestions for > where to look for a problem? > > Both environments have the following things in common: > > 1. Hosted in Amazon EC2, no load-balancer in the way > 2. Configuration requires client-certificate to connect > 3. All certificates are valid, self-signed, and properly-trusted by both > sides > 4. TLS configuration has been locked-down to TLSv1.2, selected cipher > suites, FIPS mode=off > 5. All versions are the same: stunnel 4.56 w/OpenSSL 1.0.2k-fips > > The production (working) environment happens to be i686 and the > development environment happens to be x86-86, but I don't believe that > is relevant. > > When I use e.g. "openssl s_client" to connect to the production > environment and I *do not* provide a client certificate, I am able to > perform the initial TLS handshake, get a cipher suite negotiated, etc. > and then the connection fails because I didn't provide the client > certificate, of course. I *can* see in the handshake the list of allowed > client certificates. > > When I do the same in development, I get a handshake failure. No allowed > client certificates are shown. No nothing. > > If I connect on localhost to the dev server, I get what I'm expecting: > allowed client certificates are listed, connection is closed because I'm > not using the client certificate. Connecting from another host gets my a > handshake failure. > > Again, there are no load-balancers or anything between the outside and > the EC2 instance. I'm connecting as directly as it's possible to > connect. The box definitely allows incoming connections on the port I'm > trying to use; the AWS security group is configured correctly. > > I have tried dropping ALL security configuration on the dev server's > stunnel.conf such as client-cert requirements, TLS protocols, cipher > suites, etc. and I get the same behavior every time. I'm starting to > think that it has nothing to do with my stunnel.conf configuration at > all, but I'm at a loss as to where to look, next. > > Any ideas? Some more information: 0. The error I get on the client is "handshake failure" and the stunnel server drops this log message: SSL_accept: 1408A0C1: error:1408A0C1:SSL routines:ssl3_get_client_hello:no shared cipher Note that I have disabled all but TLSv1.2 on the server. Removing this restriction does not change the behavior. 1. I have multiple stunnel configuration files on this server. Actually, I have 4 of them. Connections to ports defined in 2 of these files are not connecting successfully. Connections to ports defined in the OTHER two files *are* connecting successfully. The configurations seem to follow a pattern: those using RSA certificates as the server-certificate are working as expected, while those with EC server-certificate are failing. When I say "working" versus "failing", I mean that this command will give me a cipher suite and master key, but still drop the connection because I'm not providing a client-certificate for these tests: $ openssl s_client -connect host:port 2. I have a Java-based service that *is* able to connect through this stunnel instance just fine. It's running on a recent version of Java 8. My CLI client (not OpenSSL) is also running the same version. My CLI client cannot connect. *weird* In both cases, I am using EC client certificates, but the certificates are different from each other. Both of these certificates are trusted by the server. 3. When using OpenSSL 1.0.2t, I *can* connect, get the list of acceptable client certificates, etc. even without providing a client certificate. When using OpenSSL 1.1.1d, I can *not* connect. So perhaps the inside/outside networking thing I was thinking the problem might be is incorrect. AFAIK, both versions of OpenSSL should be able to use EC certificates and cipher suites. Thanks, -chris

1 0

Strange connection failure in one environment
by Christopher Schultz 25 Oct '19

25 Oct '19

All, I've been using stunnel in two environments (dev, prod) for a long time without any problems. Recently, my dev environment started acting funny and I can't connect to it from outside the box. Can someone take a look and let me know if you have any suggestions for where to look for a problem? Both environments have the following things in common: 1. Hosted in Amazon EC2, no load-balancer in the way 2. Configuration requires client-certificate to connect 3. All certificates are valid, self-signed, and properly-trusted by both sides 4. TLS configuration has been locked-down to TLSv1.2, selected cipher suites, FIPS mode=off 5. All versions are the same: stunnel 4.56 w/OpenSSL 1.0.2k-fips The production (working) environment happens to be i686 and the development environment happens to be x86-86, but I don't believe that is relevant. When I use e.g. "openssl s_client" to connect to the production environment and I *do not* provide a client certificate, I am able to perform the initial TLS handshake, get a cipher suite negotiated, etc. and then the connection fails because I didn't provide the client certificate, of course. I *can* see in the handshake the list of allowed client certificates. When I do the same in development, I get a handshake failure. No allowed client certificates are shown. No nothing. If I connect on localhost to the dev server, I get what I'm expecting: allowed client certificates are listed, connection is closed because I'm not using the client certificate. Connecting from another host gets my a handshake failure. Again, there are no load-balancers or anything between the outside and the EC2 instance. I'm connecting as directly as it's possible to connect. The box definitely allows incoming connections on the port I'm trying to use; the AWS security group is configured correctly. I have tried dropping ALL security configuration on the dev server's stunnel.conf such as client-cert requirements, TLS protocols, cipher suites, etc. and I get the same behavior every time. I'm starting to think that it has nothing to do with my stunnel.conf configuration at all, but I'm at a loss as to where to look, next. Any ideas? -chris

1 1

ssl offloading with stunnel
by simona vittori 23 Oct '19

23 Oct '19

Hello,Is it possible to have the following architecture with stunnel? stunnel client --> balancer --> db server Balancer doesn't have stunnel product installed on it. It receives encrypted packages from stunnel client and redirect clear traffico to DB server. Is it possible to realize this configuration? Thank a lot. Regards Simona

1 0

stunnel : user identification in mutual auth
by Denis Polushin 22 Oct '19

22 Oct '19

Hi All, Haven't found the answer for this issue. The scheme is : TLS-client <==tls==> stunnel-server <==open==> App-server In user session stunnel-server perform authorization for client with its certificate (verify=2) and send request further to App-server. How does App-server can identify user in this session? To grand permissions. Ideally it would be good to know CN or EKU of user certificate. Is it possible? Thanks a lot!! Denis

1 0

segfault with v5.55
by Mark Brookes 30 Sep '19

30 Sep '19

We are currently experiencing a segfault with version 5.55 openssl version 1.1.0k and are trying to narrow down what the issue may be. The first 2 segfaults are about 10 mins apart. The below is from the stunnel log just before and then after the event. Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77582]: TLS state (accept): SSLv3/TLS write finished Sep 23 11:47:31 xxxx stunnel: LOG7[77582]: TLS state (accept): SSLv3/TLS read change cipher spec Sep 23 11:47:31 xxxx stunnel: LOG7[77582]: TLS state (accept): SSLv3/TLS read finished Sep 23 11:47:31 xxxx stunnel: LOG7[77580]: TLS state (accept): SSLv3/TLS write finished Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Deallocating application specific data for session connect address Sep 23 11:47:31 xxxx stunnel: LOG7[77581]: Remove session callback Sep 23 11:52:58 xxxx stunnel: LOG7[ui]: Clients allowed=62937 Sep 23 11:52:58 xxxx stunnel: LOG5[ui]: stunnel 5.55 on x86_64-pc-linux-gnu platform Sep 23 11:52:58 xxxx stunnel: LOG5[ui]: Compiled/running with OpenSSL 1.1.0k 28 May 2019 Sep 23 11:52:58 xxxx stunnel: LOG5[ui]: Threading:PTHREAD Sockets:POLL,IPv6 TLS:ENGINE,OCSP,PSK,SNI Sep 23 11:52:58 xxxx stunnel: LOG7[ui]: errno: (*__errno_location ()) The following is from the kernel log. stunnel[31253]: segfault at 0 ip 00007f91214cacea sp 00007f91205ea9f8 error 6 in libpthread-2.12.so[7f91214c0000+17000] The kernel log after the second segfault- stunnel[1868]: segfault at 0 ip 00007f5b35398cea sp 00007f5b344c59f8 error 6 in libpthread-2.12.so[7f5b3538e000+17000] This is all the information we have at the moment. -- Mark Brookes Loadbalancer.org Ltd. www.loadbalancer.org +44 (0)330 380 1064 mark(a)loadbalancer.org

1 0

Re: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections
by Lynch, Andrew 27 Sep '19

27 Sep '19

I’ve been digging around the code and I am no longer sure whether stunnel is at fault or it is actually an issue in OpenSSL! load_key_engine() in ctx.c shoves the key into the SSL context – SSL_CTX_use_PrivateKey(section->ctx, pkey). So the key should be freed when that context is disposed of. The SIGHUP ends up in case SIGNAL_RELOAD_CONFIG of signal_pipe_dispatch(), which calls options_free() [stunnel.c line 830]. Eventually there is an SSL_CTX_free() call, but it looks like this does not actually free the private key. Ephemeral keys are freed and loads of other stuff, but not the private key associated with the certificate. I may try to build a hacked OpenSSL on Monday to see what happens if SSL_CTX_free also frees the private key… Regards, Andrew. From: Brent Kimberley [mailto:brent_kimberley@rogers.com] Sent: Friday, September 27, 2019 4:29 PM To: stunnel-users(a)stunnel.org; Lynch, Andrew Subject: Re: stunnel-users Digest, Vol 182, Issue 14 >> Over time, after numerous reloads of stunnel (kill -HUP) the HSM reports that its connection table is full. >> Logging from the engine shows that stunnel is never freeing the keys and therefore the engine is not closing the associated sessions with the HSM. Each stunnel reload opens 70 new sessions until eventually the HSM's configured limit is exceeded. Where does the signal dispatcher free keys on SIG_HUP? https://github.com/mtrojnar/stunnel/blob/master/src/stunnel.c Date: Fri, 27 Sep 2019 07:23:29 +0000 From: "Lynch, Andrew" <andrew.lynch(a)atos.net<mailto:andrew.lynch@atos.net>> To: Eric Eberhard <flash(a)vicsmba.com<mailto:flash@vicsmba.com>> Cc: "stunnel-users(a)stunnel.org<mailto:stunnel-users@stunnel.org>" <stunnel-users(a)stunnel.org<mailto:stunnel-users@stunnel.org>> Subject: Re: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections Hi Eric, Thank you for your suggestion. It is certainly worth looking into, although I suspect it may be impractical in our environment. Now we have a single stunnel.conf with 70 services which would translate into 70 additional entries in inetd.conf. Each of those entries then needs to have its own reduced stunnel.conf so that it will only load the cert and key for its endpoint. The stunnel HOWTO recommends daemon mode over inetd mode due to the overhead for forking and SSL initialization. We would have additional overhead for engine init and HSM connection. Regards, Andrew. -----Original Message----- From: Eric Eberhard [mailto:flash@vicsmba.com<mailto:flash@vicsmba.com>] Sent: Thursday, September 26, 2019 7:20 PM To: Lynch, Andrew; stunnel-users(a)stunnel.org<mailto:stunnel-users@stunnel.org> Subject: RE: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections? Try running in inetd mode -- even if you don't like this you will learn something. Inetd will close connections as needed. E -----Original Message----- From: stunnel-users [mailto:stunnel-users-bounces@stunnel.org<mailto:stunnel-users-bounces@stunnel.org>] On Behalf Of Lynch, Andrew Sent: Thursday, September 26, 2019 5:56 AM To: stunnel-users(a)stunnel.org<mailto:stunnel-users@stunnel.org> Subject: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections? Hi, We are using stunnel as a server to terminate incoming TLS connections. The config has around 70 services with certificates whose EC private keys are stored in an HSM and accessed using an OpenSSL engine. Over time, after numerous reloads of stunnel (kill -HUP) the HSM reports that its connection table is full. Logging from the engine shows that stunnel is never freeing the keys and therefore the engine is not closing the associated sessions with the HSM. Each stunnel reload opens 70 new sessions until eventually the HSM's configured limit is exceeded. This behaviour has been observed on Suse Enterprise Linux 12.3 with the system-provided stunnel-5.00-4.3.4, but I can reproduce it with my own build of the current version 5.55. Is this a known issue? It appears that other (ephemeral) keys are being freed, just not those associated with the service certificates. Currently our workaround is to perform a full restart instead of a reload. This closes all HSM sessions when the process terminates, but of course it also kills any open client connections so it can only be done during the scheduled maintenance windows. Regards, Andrew. _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org<mailto:stunnel-users@stunnel.org> https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users ------------------------------ Subject: Digest Footer _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org<mailto:stunnel-users@stunnel.org> https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users ------------------------------ End of stunnel-users Digest, Vol 182, Issue 14 **********************************************

2 1

Re: [stunnel-users] stunnel-users Digest, Vol 182, Issue 14
by Eric Eberhard 27 Sep '19

27 Sep '19

I don’t suggest it as a final solution – although for me it works great and we have customers exchanging 2-4 million XML documents per day with no performance issues – BUT no matter what it is a good debugging tool. Inetd gets all the initial connections and closes them when done. It might or might or might not be involved with the HSM at all. You might also look at the process table (some form of ps) and grep for stunnel and relatives and see if they are hanging around. In theory, if a process dies, it “should” close it’s connections. If it is floating in space it will not. Inetd will kill the process – at least on Centos and AIX and SCO – for sure and for certain. E From: stunnel-users [mailto:stunnel-users-bounces@stunnel.org] On Behalf Of Brent Kimberley Sent: Friday, September 27, 2019 7:29 AM To: stunnel-users(a)stunnel.org; andrew.lynch(a)atos.net Subject: Re: [stunnel-users] stunnel-users Digest, Vol 182, Issue 14 >> Over time, after numerous reloads of stunnel (kill -HUP) the HSM reports that its connection table is full. >> Logging from the engine shows that stunnel is never freeing the keys and therefore the engine is not closing the associated sessions with the HSM. Each stunnel reload opens 70 new sessions until eventually the HSM's configured limit is exceeded. Where does the signal dispatcher free keys on SIG_HUP? https://github.com/mtrojnar/stunnel/blob/master/src/stunnel.c Date: Fri, 27 Sep 2019 07:23:29 +0000 From: "Lynch, Andrew" <andrew.lynch(a)atos.net <mailto:andrew.lynch@atos.net> > To: Eric Eberhard <flash(a)vicsmba.com <mailto:flash@vicsmba.com> > Cc: "stunnel-users(a)stunnel.org <mailto:stunnel-users@stunnel.org> " <stunnel-users(a)stunnel.org <mailto:stunnel-users@stunnel.org> > Subject: Re: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections Hi Eric, Thank you for your suggestion. It is certainly worth looking into, although I suspect it may be impractical in our environment. Now we have a single stunnel.conf with 70 services which would translate into 70 additional entries in inetd.conf. Each of those entries then needs to have its own reduced stunnel.conf so that it will only load the cert and key for its endpoint. The stunnel HOWTO recommends daemon mode over inetd mode due to the overhead for forking and SSL initialization. We would have additional overhead for engine init and HSM connection. Regards, Andrew. -----Original Message----- From: Eric Eberhard [mailto:flash@vicsmba.com <mailto:flash@vicsmba.com> ] Sent: Thursday, September 26, 2019 7:20 PM To: Lynch, Andrew; stunnel-users(a)stunnel.org <mailto:stunnel-users@stunnel.org> Subject: RE: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections? Try running in inetd mode -- even if you don't like this you will learn something. Inetd will close connections as needed. E -----Original Message----- From: stunnel-users [mailto:stunnel-users-bounces@stunnel.org <mailto:stunnel-users-bounces@stunnel.org> ] On Behalf Of Lynch, Andrew Sent: Thursday, September 26, 2019 5:56 AM To: stunnel-users(a)stunnel.org <mailto:stunnel-users@stunnel.org> Subject: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections? Hi, We are using stunnel as a server to terminate incoming TLS connections. The config has around 70 services with certificates whose EC private keys are stored in an HSM and accessed using an OpenSSL engine. Over time, after numerous reloads of stunnel (kill -HUP) the HSM reports that its connection table is full. Logging from the engine shows that stunnel is never freeing the keys and therefore the engine is not closing the associated sessions with the HSM. Each stunnel reload opens 70 new sessions until eventually the HSM's configured limit is exceeded. This behaviour has been observed on Suse Enterprise Linux 12.3 with the system-provided stunnel-5.00-4.3.4, but I can reproduce it with my own build of the current version 5.55. Is this a known issue? It appears that other (ephemeral) keys are being freed, just not those associated with the service certificates. Currently our workaround is to perform a full restart instead of a reload. This closes all HSM sessions when the process terminates, but of course it also kills any open client connections so it can only be done during the scheduled maintenance windows. Regards, Andrew. _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org <mailto:stunnel-users@stunnel.org> https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users ------------------------------ Subject: Digest Footer _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org <mailto:stunnel-users@stunnel.org> https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users ------------------------------ End of stunnel-users Digest, Vol 182, Issue 14 **********************************************

1 0

Re: [stunnel-users] stunnel-users Digest, Vol 182, Issue 14
by Brent Kimberley 27 Sep '19

27 Sep '19

>> Over time, after numerous reloads of stunnel (kill -HUP) the HSM reports that its connection table is full. >> Logging from the engine shows that stunnel is never freeing the keys and therefore the engine is not closing the associated sessions with the HSM. Each stunnel reload opens 70 new sessions until eventually the HSM's configured limit is exceeded. Where does the signal dispatcher free keys on SIG_HUP?https://github.com/mtrojnar/stunnel/blob/master/src/stunnel.c Date: Fri, 27 Sep 2019 07:23:29 +0000 From: "Lynch, Andrew" <andrew.lynch(a)atos.net> To: Eric Eberhard <flash(a)vicsmba.com> Cc: "stunnel-users(a)stunnel.org" <stunnel-users(a)stunnel.org> Subject: Re: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections Hi Eric, Thank you for your suggestion. It is certainly worth looking into, although I suspect it may be impractical in our environment. Now we have a single stunnel.conf with 70 services which would translate into 70 additional entries in inetd.conf. Each of those entries then needs to have its own reduced stunnel.conf so that it will only load the cert and key for its endpoint. The stunnel HOWTO recommends daemon mode over inetd mode due to the overhead for forking and SSL initialization. We would have additional overhead for engine init and HSM connection. Regards, Andrew. -----Original Message----- From: Eric Eberhard [mailto:flash@vicsmba.com] Sent: Thursday, September 26, 2019 7:20 PM To: Lynch, Andrew; stunnel-users(a)stunnel.org Subject: RE: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections? Try running in inetd mode -- even if you don't like this you will learn something. Inetd will close connections as needed. E -----Original Message----- From: stunnel-users [mailto:stunnel-users-bounces@stunnel.org] On Behalf Of Lynch, Andrew Sent: Thursday, September 26, 2019 5:56 AM To: stunnel-users(a)stunnel.org Subject: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections? Hi, We are using stunnel as a server to terminate incoming TLS connections. The config has around 70 services with certificates whose EC private keys are stored in an HSM and accessed using an OpenSSL engine. Over time, after numerous reloads of stunnel (kill -HUP) the HSM reports that its connection table is full. Logging from the engine shows that stunnel is never freeing the keys and therefore the engine is not closing the associated sessions with the HSM. Each stunnel reload opens 70 new sessions until eventually the HSM's configured limit is exceeded. This behaviour has been observed on Suse Enterprise Linux 12.3 with the system-provided stunnel-5.00-4.3.4, but I can reproduce it with my own build of the current version 5.55. Is this a known issue? It appears that other (ephemeral) keys are being freed, just not those associated with the service certificates. Currently our workaround is to perform a full restart instead of a reload. This closes all HSM sessions when the process terminates, but of course it also kills any open client connections so it can only be done during the scheduled maintenance windows. Regards, Andrew. _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users ------------------------------ Subject: Digest Footer _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users ------------------------------ End of stunnel-users Digest, Vol 182, Issue 14 **********************************************

1 0

Re: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections?
by Eric Eberhard 27 Sep '19

27 Sep '19

Try running in inetd mode -- even if you don't like this you will learn something. Inetd will close connections as needed. E -----Original Message----- From: stunnel-users [mailto:stunnel-users-bounces@stunnel.org] On Behalf Of Lynch, Andrew Sent: Thursday, September 26, 2019 5:56 AM To: stunnel-users(a)stunnel.org Subject: [stunnel-users] stunnel with OpenSSL engine: reload leaks HSM connections? Hi, We are using stunnel as a server to terminate incoming TLS connections. The config has around 70 services with certificates whose EC private keys are stored in an HSM and accessed using an OpenSSL engine. Over time, after numerous reloads of stunnel (kill -HUP) the HSM reports that its connection table is full. Logging from the engine shows that stunnel is never freeing the keys and therefore the engine is not closing the associated sessions with the HSM. Each stunnel reload opens 70 new sessions until eventually the HSM's configured limit is exceeded. This behaviour has been observed on Suse Enterprise Linux 12.3 with the system-provided stunnel-5.00-4.3.4, but I can reproduce it with my own build of the current version 5.55. Is this a known issue? It appears that other (ephemeral) keys are being freed, just not those associated with the service certificates. Currently our workaround is to perform a full restart instead of a reload. This closes all HSM sessions when the process terminates, but of course it also kills any open client connections so it can only be done during the scheduled maintenance windows. Regards, Andrew. _______________________________________________ stunnel-users mailing list stunnel-users(a)stunnel.org https://www.stunnel.org/cgi-bin/mailman/listinfo/stunnel-users

2 1

stunnel with OpenSSL engine: reload leaks HSM connections?
by Lynch, Andrew 26 Sep '19

26 Sep '19

Hi, We are using stunnel as a server to terminate incoming TLS connections. The config has around 70 services with certificates whose EC private keys are stored in an HSM and accessed using an OpenSSL engine. Over time, after numerous reloads of stunnel (kill -HUP) the HSM reports that its connection table is full. Logging from the engine shows that stunnel is never freeing the keys and therefore the engine is not closing the associated sessions with the HSM. Each stunnel reload opens 70 new sessions until eventually the HSM's configured limit is exceeded. This behaviour has been observed on Suse Enterprise Linux 12.3 with the system-provided stunnel-5.00-4.3.4, but I can reproduce it with my own build of the current version 5.55. Is this a known issue? It appears that other (ephemeral) keys are being freed, just not those associated with the service certificates. Currently our workaround is to perform a full restart instead of a reload. This closes all HSM sessions when the process terminates, but of course it also kills any open client connections so it can only be done during the scheduled maintenance windows. Regards, Andrew.

1 0