- stunnel-users - stunnel.org

Re: [stunnel-users] ftp to sftp
by Justin Greene 16 Jan '09

16 Jan '09

I do not think that this can work. SFTP is not just encrypted FTP traffic, but an entirely different protocol. Justin ________________________________ From: aaa aaa [mailto:[email protected]] Sent: Friday, January 16, 2009 7:58 AM To: stunnel-users(a)mirt.net Subject: ftp to sftp Hello, I am using last edition of stunnel. I would like to use my favourite ftp client and connect to sftp. But my ftp clinet don't allow this so I would like to use stunnel if possible. How should I configure stunnel to do this (ports etc.)? I am a newbie with stunnel but I am using it for sending email for several weeks and everything works well. So I hope it is possible to use it for sftp too. Any help much appreciated. thanks in advance. ----------

1 0

0x09 bytes in stunnel-4.26/src/Makefile.am
by Andrew Schulman 14 Jan '09

14 Jan '09

For some strange reason, there are five hex 09 bytes in stunnel-4.26/src/Makefile.am. This broke my Cygwin build: I remove the Windows-only stuff from this file for Cygwin, and diff captured the 0x09 characters that got removed, but then when I later tried to apply the patch, patch choked on them. It seems that each of these bytes should really be 0x0A. Strange. Andrew.

2 2

Re: [stunnel-users] How do I do a one-time client connection?
by D. Richard Hipp 13 Jan '09

13 Jan '09

On Jan 13, 2009, at 11:07 AM, Karl J. Runge wrote: > On Tue, 13 Jan 2009, "D. Richard Hipp" <drh(a)hwaci.com> wrote: >> >> I'm sure I could get something like that to work on Unix. >> Unfortunately, I need a cross-platform solution:; unix & windows. >> And >> (as usual) the windows side is harder to work around. >> >> What I really need to do is send a known amount of binary data in a >> file over an SSL connection to a remote machine, then get back the >> reply of unknown length into a separate file: >> >> stunnel test.conf <test.in >test.out >> >> I'm open to any suggestions on how to do this, (such that it will >> work >> on unix and windows). > > So it is not an https URL like in your example? If it were, perhaps > wget or curl for windows would be a better path than stunnel. I'm using a simple web-page fetch for debugging. In actual use, the situation is more complicated and wget and curl won't work. Thx for the suggestion, though. > > > socat might be ported to windows, I'm not sure. If writing a small C > wrapper program for Windows is not acceptable, I'm not sure what you > can do. socat appears to be linux-mostly. (They have a MacOS-X port, which suggests that it is not even "unix" neutral.) So I'm guessing that windows is out of the question there. Looks like I'm just going to have to bite the bullet and link against OpenSSL directly..... > > > > BTW, I have made a one-time connection patch to stunnel that I use > with > my VNC tool SSVNC: > > http://libvncserver.cvs.sourceforge.net/viewvc/libvncserver/libvncserver/x1… > > It is a hack, but useful. My VNC tool works OK with stock stunnel > (everything else equal I prefer using), but I also distribute a > patched > stunnel for more reliable behavior in this one-time client connect > mode. > > Karl > D. Richard Hipp drh(a)hwaci.com

2 1

Re: [stunnel-users] How do I do a one-time client connection?
by stunnel-hm＠karlrunge.com 13 Jan '09

13 Jan '09

On Tue, 13 Jan 2009, "D. Richard Hipp" <drh(a)hwaci.com> wrote: > > Apparently stunnel will work as a pipe if I create a config file like > this: > > client=yes > connect=remosthost:port > > Then run stunnel and talk over standard input and standard output. One problem I have found in doing this is that if socketpair(2) is used to create the file descriptor that will be connected to stunnel's stdio, then stunnel's libwrap checking is activated because it detects it to be a socket. Using socketpair is a nice way to turn a TCP client into one that connects via a two-way pipe with stunnel (because the client app doesn't have to special case the two file descriptors for stunnel's stdin and stdout when reading and writing.) Since a user (i.e. not a sysadmin) has no control over the libwrap settings in /etc, stunnel's libwrap check will always fail when used this way. In my opinion, stunnel should allow disabling of libwrap (and syslog too) via the config file (i.e. not just compile time options). This makes the "one connection" client usage you describe work better. Karl

1 0

Re: [stunnel-users] How do I do a one-time client connection?
by D. Richard Hipp 13 Jan '09

13 Jan '09

Perhaps I spoke too soon about having solved my own problem. Suppose I have a configuration file (test.conf) like this: client=yes connect=www.fossil-scm.org:443 Then if I run: stunnel test.conf And I type "GET / HTTP/1.0" on standard input, the webpage comes back on standard output. So far so good. But if I do: stunnel test.conf <test.in >test.out Where file test.in contains the "GET / HTTP/1.0" test, I get nothing back! I also tried this: cat test.in | stunnel test.conf And various combinations of this: openssl s_client -connect www.fossil-scm.org:443 <test.in In every case, when I type the GET line on standard input it works. When I redirect from a file it does not. Any hints on the cause of the very unexpected and very mysterious behavior? D. Richard Hipp drh(a)hwaci.com

1 0

How do I do a one-time client connection?
by D. Richard Hipp 13 Jan '09

13 Jan '09

I'd like to invoke stunnel to handle a single SSL client-side connection (talking to an SSL server) with the assurance that the stunnel quit will automatically quit after that one connection closes. I did not see any options to support this. Did I overlook something, or is stunnel just not able to do this? D. Richard Hipp drh(a)hwaci.com

1 1

public domain [PATCH] to stunnel 4.27b1 for Windows CE / Windows Mobile compilation and unicode bug fixes
by Pierre DELAAGE 10 Jan '09

10 Jan '09

Dear Sir, Please find enclosed a patch in "diff -cr orig work" format, applying to stunnel4.27b1 as found here ftp://stunnel.mirt.net/stunnel/. This patch maily addresses compilation and unicode issues for Windows CE targets. I use MS EVC 4sp2 compiler with WCE 420 SDK, on a vista host platform. Once debugged the code works fine on WM5 and WM6 HTC smartphones. It needs a windows CE openssl lib (I logged a patch to openssl snap 20090102 because it has some compilation issues on windows CE platform). * For the sake of completeness openssl 098i and 098j (latest) must be ALSO patched to compile for wince, with this patch I submitted to openssl team : http://marc.info/?l=openssl-dev&m=123099015013538&w=2 The present stunnel patch addresses the following issues : 1/ COMPILATION FAILURE : on common.h "EINVAL redefinition" : in fact the compile options WX (see evc.mak) does not tolerate any warning : to avoid a warning on EINVAL redef, we must protect this undef by ifdef. Note that there is still a problem : if the symbol is NOT defined at all, it will NEVER be defined by that revised code : this is because the compiler always generate an error if it sees something like : #ifdef EINVAL #undef EINVAL #define EINVAL WSAEINVAL #else #define EINVAL WSAEINVAL #endif because it seems to not like two defines even separated by a else. Just putting the #define outside of a block dealing with undef, does not solve the compilation issue. 2/ COMPILATION FAILURE : related to wcecompat/setjmp.h SOLUTION : in evc.mak, needed /D ARM to compile with wcecompat/setjmp.h. This symbol is not documented in MSDN but clearly present is header files. Moreover It is consistent with other processor symbols described in the sdk. 3/ COMPILATION FAILURE : related to gui.c pb with undeclared hmainmenu identifier, because "endif use_win32" misplaced. 3/ RC COMPILATION FAILURE : related to resources.rc CAUSE : "VERSION" unknown SOLUTION : /dversion in rflags in evc.mak 4/ LINK ERROR : fatal error LNK1181: cannot open input file 'libeay32.lib' cause : bad path name in evc.mak for openssl lib. since openssl >> 098a (eg 098h) "out32dll" folder is named "out32dll_targetCPU". 5/ LINK ERROR on gui.obj : error LNK2001: unresolved external symbol _beginthread _$(TARGETCPU) Cause : in sthreads.c, misplaced "#endif os2" Note : corrected various erroneous comments on #if endif in sthreads.c ************* OPERATIONAL ERRORS (at run-time) using stunnel to enable certificate based authentication for POPS/SMPTPS between a smartphone and a linux server, I faced this problems : When first trying to authenticate a dlgbox is displayed on the smartphone to input the password of the private key protecting the client certificate: 6/ RUN-TIME ERROR : "squares" displayed instead of chars in "password input dialog box" titlebar Well, this is only a cosmetic issue, but this is still an issue. Cause : this is a typical unicode issue in gui.c pass_proc/wm_initdialog : the sendmessage to display the title uses a concatenation of two strings : a constant extracted from the resource file, and converted to wchar by TEXT macro in unicode context, and a key file name given by ui_data->section->key. The problem is that the file name is always of type 8bit char, leading to concatenation "scrambling" when concatenated with a unicode string. SOLUTION : it seems not possible to change the type of ui_data->section->key to tchar/wchar because it is used in various ansi code. So the simplest way to correct this pb is just to make a local conversion from ansi to tchar in gui.c/pass_proc/initdialog 7/ RUN-TIME ERROR : It is impossible to validate a password on a private key protecting a certificate: The password is ALWAYS rejected (after many tests with 3 retries). Cause : the password is input as unicode (because the message EM_GETLINE on the control IDE_PASSEDIT is unicode ready) but stored and managed as an ansi string so the password can never be validated. NOTE : contrary to the documentation of windows CE sdk, the EM_GETLINE msg IS REALLY unicode ready. Indeed this is also the case in classic win32 sdk for PC platform (described in the msdn library). SOLUTION : the input password must be converted to ansi before being validated vs openssl 2009.01.05 13:00:01 LOG3[4849170:5712826]: Wrong pass phrase: retrying 2009.01.05 13:00:05 LOG3[4849170:5712826]: error stack: 140B3009 : error:140B3009:SSL routines:SSL_CTX_use_RSAPrivateKey_file:PEM lib 2009.01.05 13:00:05 LOG3[4849170:5712826]: error stack: 906A065 : error:0906A065:PEM routines:PEM_do_header:bad decrypt 2009.01.05 13:00:05 LOG3[4849170:5712826]: SSL_CTX_use_RSAPrivateKey_file: 6065064: error:06065064:digital envelope routines:EVP_DecryptFinal_ex:bad decrypt 2009.01.05 13:00:05 LOG3[4849170:5712826]: Server is down I hope you will find this patch useful. Thank you for your excellent work, Yours sincerely and happy new year, Pierre Delaage Note : I use stunnel to establish a simple "vpn" between smartphones and a corporate linux server mainly for HTTPS/POPS/SMTPS support. Stunnel is very relevant in that matter, over solutions based on SSH (although we use also ssh), from a communication cost point of view : ssh establishes permanent socket between client and server, so that the communication is charged by the mobile network provider : and these charges are very expensive. On the contrary stunnel only establishes ssl sockets on demand so that financial charges are limited to strict necessary. Please note that stunnel brings "client based certificate authentication" to POP/SMTP mobile mail user agents which only supports SSL with server authentication, but NO client authentication, such as mobile outlook.Here again stunnel is very relevant. Note 2 : the unicode bug fixes should benefit also to the win32 stunnel version for PC. Note 3 : this mail has also been sent to webmaster(a)stunnel.org, but with no response at all. Only in work/src: client.obj diff -cr orig/src/common.h work/src/common.h *** orig/src/common.h 2008-09-01 08:06:24.000000000 +0200 --- work/src/common.h 2009-01-05 23:16:03.052680000 +0100 *************** *** 187,198 **** #define EINPROGRESS WSAEINPROGRESS #define EWOULDBLOCK WSAEWOULDBLOCK #define EISCONN WSAEISCONN ! #undef EINVAL #define EINVAL WSAEINVAL #include <process.h> /* _beginthread */ #include <tchar.h> #define NO_IDEA #define OPENSSL_NO_IDEA --- 187,204 ---- #define EINPROGRESS WSAEINPROGRESS #define EWOULDBLOCK WSAEWOULDBLOCK #define EISCONN WSAEISCONN ! ! #ifdef EINVAL // pdelaage, required to avoid compilation "error", see WX option in evc.mak and ms-evc420 compiler ! #undef EINVAL // TODO : if einval is not defined, this code does not define it; but putting a #define in an #else leads again to compilation error! #define EINVAL WSAEINVAL + #endif + #include <process.h> /* _beginthread */ #include <tchar.h> + + #define NO_IDEA #define OPENSSL_NO_IDEA Only in work/src: ctx.obj diff -cr orig/src/evc.mak work/src/evc.mak *** orig/src/evc.mak 2008-09-20 22:23:21.000000000 +0200 --- work/src/evc.mak 2009-01-04 10:09:32.439894500 +0100 *************** *** 1,23 **** # wce.mak for stunnel.exe by Michal Trojnara 2006 # ! WCEVER=300 # Modify this to point to your actual openssl compile directory # (You did already compile openssl, didn't you???) ! SSLDIR=..\..\build\openssl-0.9.8i ! COMPATDIR=..\..\build\wcecompat CEUTILSDIR=..\..\ceutils DSTDIR=ce:\stunnel - SDKDIR=C:\Windows CE Tools\wce$(WCEVER)\Pocket PC 2002 ! INCLUDES=-I$(SSLDIR)/inc32 -I$(COMPATDIR)\include -I"$(SDKDIR)\include" LIBS=libeay32.lib ssleay32.lib wcecompatex.lib winsock.lib ! CC=CL$(TARGETCPU) ! CFLAGS=/nologo /MC /O1i /W3 /WX /GF /Gy /DHOST=\"$(TARGETCPU)-WCE-eVC-$(WCEVER)\" /D$(TARGETCPU) /DUNDER_CE=$(WCEVER) /D_WIN32_WCE=$(WCEVER) /DUNICODE -D_UNICODE $(INCLUDES) ! RFLAGS=$(INCLUDES) ! LDFLAGS=/nologo /subsystem:windowsce,3.00 /machine:ARM /libpath:"$(SDKDIR)\lib\$(TARGETCPU)" /libpath:"$(COMPATDIR)\lib" /libpath:"$(SSLDIR)\out32dll" OBJS=stunnel.obj ssl.obj ctx.obj verify.obj file.obj client.obj protocol.obj sthreads.obj log.obj options.obj network.obj resolver.obj GUIOBJS=gui.obj resources.res --- 1,35 ---- # wce.mak for stunnel.exe by Michal Trojnara 2006 # ! WCEVER=420 ! # pdelaage : following flag required by (eg) winnt.h, and is different from targetcpu (armV4) ! WCETARGETCPU=ARM # Modify this to point to your actual openssl compile directory # (You did already compile openssl, didn't you???) ! SSLDIR=C:\Users\standard\Documents\Dvts\Contrib\openssl\openssl-0.9.8-stable-SNAP-20090102\openssl-0.9.8-stable-SNAP-20090102 ! COMPATDIR=C:\Users\standard\Documents\Dvts\openssl\build\wcecompat ! # pdelaage : ceutilsdir probably useless CEUTILSDIR=..\..\ceutils + # pdelaage "ce:" is not a correct location , but we never "make install" DSTDIR=ce:\stunnel ! SDKDIR="C:\Progra~1\Micros~2\wce$(WCEVER)\standardsdk" ! ! # pdelaage eroneous path for sdkdir INCLUDES=-I$(SSLDIR)/inc32 -I$(COMPATDIR)\include -I"$(SDKDIR)\include" ! INCLUDES=-I$(SSLDIR)\inc32 -I$(COMPATDIR)\include -I"$(SDKDIR)\include\$(TARGETCPU)" LIBS=libeay32.lib ssleay32.lib wcecompatex.lib winsock.lib ! # not correct because for armv4 cc is just clarm.exe. Moreover cc is already set in the ms wce$TARGETCPU.bat script, so it is not necessary to set it up here ! # CC=CL$(TARGETCPU) ! ! VERSION=4.27 ! DEFINES=/DVERSION=\"$(VERSION)\" ! # pdelaage added wcetargetcpu to cflags ! CFLAGS=/nologo /MC /O1i /W3 /WX /GF /Gy $(DEFINES) /DHOST=\"$(TARGETCPU)-WCE-eVC-$(WCEVER)\" /D$(WCETARGETCPU) /D$(TARGETCPU) /DUNDER_CE=$(WCEVER) /D_WIN32_WCE=$(WCEVER) /DUNICODE -D_UNICODE $(INCLUDES) ! RFLAGS=/DVERSION=\"$(VERSION)\" $(INCLUDES) ! # pdelaage : LDFLAGS: since openssl >> 098a (eg 098h) out32dll is out32dll_targetCPU ! LDFLAGS=/nologo /subsystem:windowsce,3.00 /machine:ARM /libpath:"$(SDKDIR)\lib\$(TARGETCPU)" /libpath:"$(COMPATDIR)\lib" /libpath:"$(SSLDIR)\out32dll_$(TARGETCPU)" OBJS=stunnel.obj ssl.obj ctx.obj verify.obj file.obj client.obj protocol.obj sthreads.obj log.obj options.obj network.obj resolver.obj GUIOBJS=gui.obj resources.res Only in work/src: file.obj diff -cr orig/src/gui.c work/src/gui.c *** orig/src/gui.c 2008-05-02 08:23:03.000000000 +0200 --- work/src/gui.c 2009-01-04 23:45:07.451820500 +0100 *************** *** 320,329 **** hwnd=CreateWindow(classname, win32_name, WS_TILEDWINDOW, CW_USEDEFAULT, CW_USEDEFAULT, CW_USEDEFAULT, CW_USEDEFAULT, NULL, hmainmenu, hInstance, NULL); ! #endif if(cmdline.service) /* do not allow to save file in the service mode */ EnableMenuItem(hmainmenu, IDM_SAVEAS, MF_GRAYED); if(error_mode) /* log window is hidden by default */ set_visible(1); --- 320,330 ---- hwnd=CreateWindow(classname, win32_name, WS_TILEDWINDOW, CW_USEDEFAULT, CW_USEDEFAULT, CW_USEDEFAULT, CW_USEDEFAULT, NULL, hmainmenu, hInstance, NULL); ! // commented by pdelaage on 20090104 #endif if(cmdline.service) /* do not allow to save file in the service mode */ EnableMenuItem(hmainmenu, IDM_SAVEAS, MF_GRAYED); + #endif // added by pdelaage on 20090104 if(error_mode) /* log window is hidden by default */ set_visible(1); *************** *** 523,536 **** WPARAM wParam, LPARAM lParam) { TCHAR titlebar[STRLEN]; WORD cchPassword; switch (message) { case WM_INITDIALOG: /* Set the default push button to "Cancel." */ SendMessage(hDlg, DM_SETDEFID, (WPARAM) IDCANCEL, (LPARAM) 0); _sntprintf(titlebar, STRLEN, TEXT("Private key: %s"), ! ui_data->section->key); SetWindowText(hDlg, titlebar); return TRUE; --- 524,548 ---- WPARAM wParam, LPARAM lParam) { TCHAR titlebar[STRLEN]; WORD cchPassword; + LPTSTR keyFileName;// added by pdelaage on 20090104 + TCHAR sPassword[PEM_BUFSIZE];// added by pdelaage on 20090104 + char* cPassword;// added by pdelaage on 20090104 switch (message) { case WM_INITDIALOG: /* Set the default push button to "Cancel." */ SendMessage(hDlg, DM_SETDEFID, (WPARAM) IDCANCEL, (LPARAM) 0); + // pdelaage on 20090104: here was a bug : _sntprintf is unicode ready but ui_data->section->key is always char. + // commented by pdelaage on 20090104 _sntprintf(titlebar, STRLEN, TEXT("Private key: %s"), + // ui_data->section->key); + + keyFileName = str2tstr(ui_data->section->key); _sntprintf(titlebar, STRLEN, TEXT("Private key: %s"), ! keyFileName); ! free(keyFileName); ! ! SetWindowText(hDlg, titlebar); return TRUE; *************** *** 549,561 **** } /* Put the number of characters into first word of buffer. */ ! *((LPWORD)ui_data->pass) = cchPassword; /* Get the characters. */ SendDlgItemMessage(hDlg, IDE_PASSEDIT, EM_GETLINE, ! (WPARAM) 0, /* line 0 */ (LPARAM) ui_data->pass); ! ui_data->pass[cchPassword] = 0; /* Null-terminate the string. */ EndDialog(hDlg, TRUE); return TRUE; --- 561,580 ---- } /* Put the number of characters into first word of buffer. */ ! *((LPWORD) sPassword) = cchPassword; /* Get the characters. */ SendDlgItemMessage(hDlg, IDE_PASSEDIT, EM_GETLINE, ! (WPARAM) 0, /* line 0 */ (LPARAM) sPassword);// pdelaage ui_data->pass); ! ! sPassword[cchPassword] = 0; /* Null-terminate the string. */ ! ! // convert input password to ANSI string (as ui_data->pass) ! cPassword = tstr2str(sPassword); ! strcpy(ui_data->pass, cPassword); ! free(cPassword); ! EndDialog(hDlg, TRUE); return TRUE; Only in work/src: gui.obj Only in work/src: log.obj diff -cr orig/src/makece.bat work/src/makece.bat *** orig/src/makece.bat 2005-12-23 20:57:24.000000000 +0100 --- work/src/makece.bat 2009-01-03 19:31:34.507379500 +0100 *************** *** 1,3 **** @echo off ! call "C:\Program Files\Microsoft eMbedded Tools\EVC\WCE300\BIN\WCEARM.BAT" nmake /NOLOGO -f evc.mak %1 %2 %3 %4 %5 %6 %7 %8 %9 --- 1,4 ---- @echo off ! ::call "C:\Program Files\Microsoft eMbedded Tools\EVC\WCE300\BIN\WCEARM.BAT" ! call "C:\Progra~1\MSEVC4\EVC\WCE420\BIN\WCEARMV4.BAT" nmake /NOLOGO -f evc.mak %1 %2 %3 %4 %5 %6 %7 %8 %9 Only in work/src: network.obj Only in work/src: nogui.obj Only in work/src: options.obj Only in work/src: protocol.obj Only in work/src: resolver.obj Only in work/src: resources.RES Only in work/src: ssl.obj diff -cr orig/src/sthreads.c work/src/sthreads.c *** orig/src/sthreads.c 2008-09-21 21:44:06.000000000 +0200 --- work/src/sthreads.c 2009-01-04 10:46:46.285200500 +0100 *************** *** 400,406 **** return 0; } ! #endif #ifdef USE_OS2 --- 400,406 ---- return 0; } ! #endif /* cmt added by pdelaage on 20090104 : USE_WIN32 */ #ifdef USE_OS2 *************** *** 437,442 **** --- 437,444 ---- return 0; } + #endif /* USE_OS2 */ // added by pdelaage on 20090104 + #ifdef _WIN32_WCE int _beginthread(void (*start_address)(void *), *************** *** 456,464 **** ExitThread(0); } ! #endif /* !defined(_WIN32_WCE) */ ! #endif /* USE_WIN32 */ #ifdef DEBUG_STACK_SIZE --- 458,467 ---- ExitThread(0); } ! #endif /* _WIN32_WCE */ /* pdelaage 20090104, following cmt is improper: !defined(_WIN32_WCE) */ ! // pdelaage 20090104 : following endif is misplaced, the endif usewin32 is already placed upper, but this endif relates ! // to os2, and is misplaced because it "squeezes" winCE code #endif /* USE_WIN32 */ #ifdef DEBUG_STACK_SIZE Only in work/src: sthreads.obj Files orig/src/stunnel.exe and work/src/stunnel.exe differ Only in work/src: stunnel.obj Only in work/src: tstunnel.exe Only in work/src: verify.obj

1 0

Re: [stunnel-users] Distinguished Name (DN) is a cleartext network communication?
by Christophe Nanteuil 09 Jan '09

09 Jan '09

2009/1/8 Michael Renner <michael.renner(a)gmx.de>: > A server should appear like a a 'normal' https webserver to others. More or > less interesting, or just a 404 error message. However: it should appear > harmless to others, just like a webserver. > Behind this should work a proxy server (squid). Only authorized users should > be able to use it. > If I understand well what you want to do, I think my patch "Identification Propagation patch using stunnel client certificates" can be useful : - Use my stunnel patch Using squid as a proxy : - In stunnel config file : select pertinent fields of the certiificate : DN -> identFields=1023 - Get a squid version compiled with --enable-ident-lookups option - In squid : use a home-made external authentification program which returns ok if the DN provided is good. Squid documentation will tell you what is the need for the home-made authentification program. I think it's easier with DansGuardian proxy : - In stunnel config file : select pertinent fields of the certiificate : only Issuer_CN -> identFields=32 - in /etc/dansguardian/dansguardian.conf, create 2 groups : filtergroups = 2 filtergroupslist = '/etc/dansguardian/lists/filtergroupslist' - create filter groups in /etc/dansguardian/lists/filtergroupslist. (All not matching users will go in group 1) : put a line like "your issuer CN equals 2" - in /etc/dansguardian/dansguardianf1.conf, put "groupmode=0" , ie banned - in /etc/dansguardian/dansguardianf2.conf, put "groupmode=2" , ie non filtered Note that the authentification of the users is done by stunnel. Stunnel forwards the identification to the proxy which does not need to authenticate since it was done by stunnel. If the ident provided by my patch is not convenient, you can modify it to match your needs. Hope this helps -- Christophe Nanteuil

1 0

Re: [stunnel-users] Distinguished Name (DN) is a cleartext network communication?
by Tobias Dussa 08 Jan '09

08 Jan '09

Hi, [...] > But this setup is senseless, since the DN is readable with a network sniffer. > It does not appear harmless any longer after a closer look into the network > traffic. But it have to. > Now, I need an other idea to implement such s service. One easy solution that comes to mind is to deploy an outer SSL tunnel for privacy and then set up an inner SSL tunnel for traffic separation. However, I think this would require stunnel on the client side as well, at least for legitimate users. Cheers, Toby. -- begin LOVE-LETTER-FOR-YOU.txt.vbs I am a signature virus. Distribute me until the bitter end http://piology.org/ILOVEYOU-Signature-FAQ.html

1 0

Stunnel (v4.26) Running, Working, but Sometimes Aborting
by Jeff Salisbury 08 Jan '09

08 Jan '09

Greetings, We are new to stunnel. We've installed Stunnel v4.26 on a handful of Windows XP machines, and we can send and receive emails through the Stunnel configuration. These XP machines reboot automatically every night. Many mornings, when the user sits down and tries to initially send email, the email hangs and Stunnel aborts (or has already aborted). If the user runs stunnel again, our sending and receiving of email is restored and stunnel seems rock-solid for the rest of the day. Has anybody else had stability problems with Stunnel v4.26? Any suggestions? FYI, relevant configuration information is shown below: Best regards, Jeff ================================== 1) Instead of installing stunnel on each machine, we have copied the following files into a network folder available to everyone in our office: Q:\bin\stunnel\stunnel.exe Q:\bin\stunnel\Q:\ibeay32.dll Q:\bin\stunnel\libssl32.dll Q:\bin\stunnel\stunnel.conf 2) Everyone has a shortcut installed in their "Startup" folder to run Q:\bin\stunnel\stunnel.exe. 3) Here is our stunnel.conf file: # GLOBAL OPTIONS client = yes output = C:\Documents and Settings\All Users\Application Data\stunnel-log.txt debug = 0 taskbar = yes # SERVICE-LEVEL OPTIONS [pop3] accept=995 connect=mail.ourpop.com:995 [smtp] accept=465 connect=mail.oursmtp.com:465

1 0

Distinguished Name (DN) is a cleartext network communication?
by Michael Renner 07 Jan '09

07 Jan '09

Moin, I am confused. Trying to use the DN as a kind of password replacement I saw that the DN goes unencrypted through the network, while the traffic itself is encrypted of cause. Analysing the network packages with wireshark I get this: 100104083828Z0..1.0...U....DE1.0...U....Germany1.0 ..U....Munich1.0...U. ..vbox4php1.0...U....stunnel1.0...U....mars.mtr.mynet1$0"..*.H.. .....michael.renner(a)gmx.de0.."0 This is, more or less, the content of the DN. Is there a chance to encrypt this? My setup should work, no matter if a client certificate exists or not (also independent of the DN's content) . But the appliation behind stunnel (started by stunnel using the 'exec' statement) may behave dependent on the DN! Any chance? Thanks -- |Michael Renner E-mail: michael.renner(a)gmx.de | |D-81541 Munich Germany ICQ: #112280325 | |Germany Don't drink as root! ESC:wq

1 0

Stunnel New User
by David Aitchison 07 Jan '09

07 Jan '09

Hello, I have a potential use for Stunnel on a project I am working on. Excuse the perhaps very silly questions, this is all new to me. I have read through the manual, and understand that Stunnel can either be run from the command line ad-hoc or as an application using the stunnel.conf configuration file. If I have Stunnel running on the server, does this mean that all mails that leave that server will be SSL encrypted? I have a requirement to send mails out over SSL, but my application doesn't support that, it does support sending out over standard SMTP port 25, but I need it to use port 465? Thanks in advance, David Aitchison

3 2

Re: [stunnel-users] Problem with the client certificate
by Michael Renner 03 Jan '09

03 Jan '09

On Thursday 01 January 2009, Christophe Nanteuil wrote: > Well, > I think the problem comes from the CAfile, ie the cert used to verify the > other certs. This is right. I thought that the CA file is not essential with verify depths less than 2. It seems this was a mistake. This time (yes, I use tinyCA), I made and used client- and server certificates as well as the CA. Now it works. Thanks for your help! > To test, I think it is better to have only one section per file : > - your key (section ----BEGIN RSA PRIVATE KEY----- to ---END RSA...) in the > key file corresponding to key=file in config file. > - your certificate (section ----BEGIN CERTIFICATE ----- to ---END > CERTIFICATE...) in the cert file corresponding to cert=file in the config > file. > These files must exist for server and client since you want mutual > authentication. > For verifying the certificate, you need to put the client certificate > (autosigned) in the CAfile option of the server config file. > The same for the client : put the server certificate in the CAfile option > of the client config file. > > on the server, the command 'openssl verify -CAfile client_cert.pem > client_cert.pem' should return ok > on the client, the command 'openssl verify -CAfile server_cert.pem > server_cert.pem' should return ok. > It can seem stupid but, as you have self-signed certificate, the CAfile is > the cert file. > > If you use tinyCA, you can easily build your own CA with one CA cert and > several certs for clients or servers. Then, you would be able to have the > same CAfile on the client and on the server. > > I hope this is clear enough. If not, write me back. > > > > > 2008/12/31 Michael Renner <michael.renner(a)gmx.de> > > > On Tuesday 30 December 2008, you wrote: > > > Hello, > > > > Happy new year! > > > > > * are the permissions correct on your files : > > > - key must belong to the user and have 0600 status (read only by the > > > > user) > > > > > - cert must belong to the user. > > > > I think the permissions are OK. The file is owed by root and loaded at > > the start > > Wrote 1024 new random bytes to /root/.rnd > > RAND_status claims sufficient entropy for the PRNG > > PRNG seeded successfully > > Certificate: /etc/stunnel/stunnelclient.pem > > Certificate loaded > > Key file: /etc/stunnel/stunnelclient.pem > > Private key loaded > > SSL context initialized for service BreakOut > > > > > * Is the content of the cert file of this form > > > > > > -----BEGIN CERTIFICATE----- > > > certificate data here > > > -----END CERTIFICATE----- > > > ? > > > and the content of the key file this form > > > -----BEGIN RSA PRIVATE KEY----- > > > key datat here > > > -----END RSA PRIVATE KEY----- > > > > I made several files. According to > > http://www.stunnel.org/faq/certs.html#ToC5 > > I got a file with a certificate, a RSA Key and a DH section (I removed > > the password for the certificate). > > > > According to http://www.stunnel.org/examples/client_cert.html I got a > > different file: it has a certificate and a RSA section and between them > > an other section: > > > > rcnyy/AbS1YPkdggJSnw+fqzg/L/QvQB6GTT5KWJzd0= > > -----END RSA PRIVATE KEY----- > > Certificate: > > Data: > > Version: 3 (0x2) > > Serial Number: 5 (0x5) > > Signature Algorithm: sha1WithRSAEncryption > > Issuer: C=DE, ST=Germany, L=Munich, O=vbox4php, OU=Rektorat, > > CN=DE/emailAddress=michael.renner(a)gmx.de > > Validity > > Not Before: Dec 28 20:37:19 2008 GMT > > Not After : Dec 28 20:37:19 2009 GMT > > Subject: C=DE, ST=Germany, O=vbox4php, OU=stunnel, > > CN=boulder.vbox4php.org/[email protected] > > Subject Public Key Info: > > Public Key Algorithm: rsaEncryption > > RSA Public Key: (1024 bit) > > Modulus (1024 bit): > > 00:b1:05:47:7a:27:4f:19:2b:18:72:e3:3c:f6:a6: > > . > > . > > 2b:55:2d:c9:dc:96:55:14:bb > > Exponent: 65537 (0x10001) > > X509v3 extensions: > > X509v3 Basic Constraints: > > CA:FALSE > > Netscape Cert Type: > > SSL Server > > Netscape Comment: > > TinyCA Generated Certificate > > X509v3 Subject Key Identifier: > > > > 86:F6:1F:71:29:AA:A5:61:DF:B2:81:F2:34:3A:A6:9E:58:C8:6A:5E X509v3 > > Authority Key Identifier: > > > > keyid:72:68:1A:0C:9D:E9:93:81:07:E9:36:71:75:33:05:C6:70:35:01:BF > > > > > > DirName:/C=DE/ST=Germany/L=Munich/O=vbox4php/OU=Rektorat/CN=DE/emailAddre > >ss= michael.renner(a)gmx.de > > serial:BC:97:82:4E:E3:9F:FE:5A > > > > X509v3 Issuer Alternative Name: > > email:[email protected] > > <email%3Amichael.renner(a)gmx.de> X509v3 Subject Alternative Name: > > email:[email protected] > > <email%3Amichael.renner(a)gmx.de> Signature Algorithm: > > sha1WithRSAEncryption > > 49:ef:06:aa:e5:71:b1:6e:23:87:02:9d:ce:56:e1:3b:77:5a: > > . > > . > > 41:93:92:ee:57:23:95:f3:99:62:27:6a:a4:b7:85:b4:92:86: > > 22:50:79:a0 > > -----BEGIN CERTIFICATE----- > > > > Anyhow: it fails: > > 2008.12.31 17:51:07 LOG4[13056:1073809760]: VERIFY ERROR: depth=0, > > error=unable to get local issuer > > certificate: /C=DE/ST=Germany/O=vbox4php/OU=stunnel/CN= > > boulder.vbox4php.org/[email protected] > > 2008.12.31 17:51:07 LOG3[13056:1073809760]: SSL_accept: 140890B2: > > error:140890B2:SSL routines:SSL3_GET_CLIENT_CERTIFICATE:no certificate > > returned > > > > With strace I can see the the key and the cert is OK on the client side: > > (I assume that it is only read once): > > > > [pid 11829] open("/etc/stunnel/stunnelserver.pem", O_RDONLY) = 4 > > [pid 11829] fstat(4, {st_mode=S_IFREG|0600, st_size=5521, ...}) = 0 > > [pid 11829] mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE| > > MAP_ANONYMOUS, -1, 0) = 0x7f8f5f13b000 > > [pid 11829] read(4, "-----BEGIN RSA PRIVATE KEY-----\n"..., 4096) = 4096 > > [pid 11829] read(4, "VQQDEwJERTEkMCIGCSqGSIb3\nDQEJARY"..., 4096) = 1425 > > [pid 11829] read(4, "", 4096) = 0 > > [pid 11829] close(4) = 0 > > [pid 11829] munmap(0x7f8f5f13b000, 4096) = 0 > > [pid 11829] write(2, "2008.12.31 17:52:56 LOG7[11829:1"..., 682008.12.31 > > 17:52:56 LOG7[11829:140253752059616]: Certificate loaded > > ) = 68 > > [pid 11829] write(2, "2008.12.31 17:52:56 LOG7[11829:1"..., 902008.12.31 > > 17:52:56 LOG7[11829:140253752059616]: Key > > file: /etc/stunnel/stunnelserver.pem > > ) = 90 > > [pid 11829] open("/etc/stunnel/stunnelserver.pem", O_RDONLY) = 4 > > [pid 11829] fstat(4, {st_mode=S_IFREG|0600, st_size=5521, ...}) = 0 > > [pid 11829] mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE| > > MAP_ANONYMOUS, -1, 0) = 0x7f8f5f13b000 > > [pid 11829] read(4, "-----BEGIN RSA PRIVATE KEY-----\n"..., 4096) = 4096 > > [pid 11829] close(4) = 0 > > [pid 11829] munmap(0x7f8f5f13b000, 4096) = 0 > > [pid 11829] write(2, "2008.12.31 17:52:56 LOG7[11829:1"..., 682008.12.31 > > 17:52:56 LOG7[11829:140253752059616]: Private key loaded > > > > > > While I see in the clients logfile: > > SSL state (connect): SSLv3 flush data > > SSL alert (read): fatal: bad certificate > > SSL_connect: 14094412: error:14094412:SSL routines:SSL3_READ_BYTES:sslv3 > > alert > > bad certificate > > Connection reset: 0 bytes sent to SSL, 0 bytes sent to socket > > > > > > Strange! > > > > > > One more hint? > > -- > > > > |Michael Renner E-mail: michael.renner(a)gmx.de | > > |D-81541 Munich Germany ICQ: #112280325 | > > |Germany Don't drink as root! ESC:wq -- |Michael Renner E-mail: michael.renner(a)gmx.de | |D-81541 Munich Germany ICQ: #112280325 | |Germany Don't drink as root! ESC:wq

1 0

hiding a proxy that appear like a web server?
by Michael Renner 03 Jan '09

03 Jan '09

Moin, I want to use stunnel (in conjunction with prtunnel) to reach my own proxy server out of a restricted network. The normal stunnel connection works so far without a problem. But I am looking for a possibility that this connection (reaching the endpoint of stunnel with my proxy server behind) appears like a normal webserver (maybe with just one simple index.html) to others. Any ideas? Thanks -- |Michael Renner E-mail: michael.renner(a)gmx.de | |D-81541 Munich Germany ICQ: #112280325 | |Germany Don't drink as root! ESC:wq

1 1

Re: [stunnel-users] Problem with the client certificate
by Michael Renner 31 Dec '08

31 Dec '08

On Tuesday 30 December 2008, you wrote: > Hello, Happy new year! > * are the permissions correct on your files : > - key must belong to the user and have 0600 status (read only by the user) > - cert must belong to the user. I think the permissions are OK. The file is owed by root and loaded at the start Wrote 1024 new random bytes to /root/.rnd RAND_status claims sufficient entropy for the PRNG PRNG seeded successfully Certificate: /etc/stunnel/stunnelclient.pem Certificate loaded Key file: /etc/stunnel/stunnelclient.pem Private key loaded SSL context initialized for service BreakOut > * Is the content of the cert file of this form > > -----BEGIN CERTIFICATE----- > certificate data here > -----END CERTIFICATE----- > ? > and the content of the key file this form > -----BEGIN RSA PRIVATE KEY----- > key datat here > -----END RSA PRIVATE KEY----- I made several files. According to http://www.stunnel.org/faq/certs.html#ToC5 I got a file with a certificate, a RSA Key and a DH section (I removed the password for the certificate). According to http://www.stunnel.org/examples/client_cert.html I got a different file: it has a certificate and a RSA section and between them an other section: rcnyy/AbS1YPkdggJSnw+fqzg/L/QvQB6GTT5KWJzd0= -----END RSA PRIVATE KEY----- Certificate: Data: Version: 3 (0x2) Serial Number: 5 (0x5) Signature Algorithm: sha1WithRSAEncryption Issuer: C=DE, ST=Germany, L=Munich, O=vbox4php, OU=Rektorat, CN=DE/emailAddress=michael.renner(a)gmx.de Validity Not Before: Dec 28 20:37:19 2008 GMT Not After : Dec 28 20:37:19 2009 GMT Subject: C=DE, ST=Germany, O=vbox4php, OU=stunnel, CN=boulder.vbox4php.org/[email protected] Subject Public Key Info: Public Key Algorithm: rsaEncryption RSA Public Key: (1024 bit) Modulus (1024 bit): 00:b1:05:47:7a:27:4f:19:2b:18:72:e3:3c:f6:a6: . . 2b:55:2d:c9:dc:96:55:14:bb Exponent: 65537 (0x10001) X509v3 extensions: X509v3 Basic Constraints: CA:FALSE Netscape Cert Type: SSL Server Netscape Comment: TinyCA Generated Certificate X509v3 Subject Key Identifier: 86:F6:1F:71:29:AA:A5:61:DF:B2:81:F2:34:3A:A6:9E:58:C8:6A:5E X509v3 Authority Key Identifier: keyid:72:68:1A:0C:9D:E9:93:81:07:E9:36:71:75:33:05:C6:70:35:01:BF DirName:/C=DE/ST=Germany/L=Munich/O=vbox4php/OU=Rektorat/CN=DE/[email protected] serial:BC:97:82:4E:E3:9F:FE:5A X509v3 Issuer Alternative Name: email:[email protected] X509v3 Subject Alternative Name: email:[email protected] Signature Algorithm: sha1WithRSAEncryption 49:ef:06:aa:e5:71:b1:6e:23:87:02:9d:ce:56:e1:3b:77:5a: . . 41:93:92:ee:57:23:95:f3:99:62:27:6a:a4:b7:85:b4:92:86: 22:50:79:a0 -----BEGIN CERTIFICATE----- Anyhow: it fails: 2008.12.31 17:51:07 LOG4[13056:1073809760]: VERIFY ERROR: depth=0, error=unable to get local issuer certificate: /C=DE/ST=Germany/O=vbox4php/OU=stunnel/CN=boulder.vbox4php.org/emailAddress=michael.renner(a)gmx.de 2008.12.31 17:51:07 LOG3[13056:1073809760]: SSL_accept: 140890B2: error:140890B2:SSL routines:SSL3_GET_CLIENT_CERTIFICATE:no certificate returned With strace I can see the the key and the cert is OK on the client side: (I assume that it is only read once): [pid 11829] open("/etc/stunnel/stunnelserver.pem", O_RDONLY) = 4 [pid 11829] fstat(4, {st_mode=S_IFREG|0600, st_size=5521, ...}) = 0 [pid 11829] mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE| MAP_ANONYMOUS, -1, 0) = 0x7f8f5f13b000 [pid 11829] read(4, "-----BEGIN RSA PRIVATE KEY-----\n"..., 4096) = 4096 [pid 11829] read(4, "VQQDEwJERTEkMCIGCSqGSIb3\nDQEJARY"..., 4096) = 1425 [pid 11829] read(4, "", 4096) = 0 [pid 11829] close(4) = 0 [pid 11829] munmap(0x7f8f5f13b000, 4096) = 0 [pid 11829] write(2, "2008.12.31 17:52:56 LOG7[11829:1"..., 682008.12.31 17:52:56 LOG7[11829:140253752059616]: Certificate loaded ) = 68 [pid 11829] write(2, "2008.12.31 17:52:56 LOG7[11829:1"..., 902008.12.31 17:52:56 LOG7[11829:140253752059616]: Key file: /etc/stunnel/stunnelserver.pem ) = 90 [pid 11829] open("/etc/stunnel/stunnelserver.pem", O_RDONLY) = 4 [pid 11829] fstat(4, {st_mode=S_IFREG|0600, st_size=5521, ...}) = 0 [pid 11829] mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE| MAP_ANONYMOUS, -1, 0) = 0x7f8f5f13b000 [pid 11829] read(4, "-----BEGIN RSA PRIVATE KEY-----\n"..., 4096) = 4096 [pid 11829] close(4) = 0 [pid 11829] munmap(0x7f8f5f13b000, 4096) = 0 [pid 11829] write(2, "2008.12.31 17:52:56 LOG7[11829:1"..., 682008.12.31 17:52:56 LOG7[11829:140253752059616]: Private key loaded While I see in the clients logfile: SSL state (connect): SSLv3 flush data SSL alert (read): fatal: bad certificate SSL_connect: 14094412: error:14094412:SSL routines:SSL3_READ_BYTES:sslv3 alert bad certificate Connection reset: 0 bytes sent to SSL, 0 bytes sent to socket Strange! One more hint? -- |Michael Renner E-mail: michael.renner(a)gmx.de | |D-81541 Munich Germany ICQ: #112280325 | |Germany Don't drink as root! ESC:wq

1 1

Problem with the client certificate
by Michael Renner 27 Dec '08

27 Dec '08

Moin, it seems that I understood something wrong with stunnel client certificates. First of all: until now I run stunnel 4.18 and stunnel 4.21 with debian and kubuntu without client certificates (with success). I use it to reach a proprietary application through an untrusted network. Now, I want the clients to authenticate against the server. Therefor, as a quick and dirty test, I used the same stunnel.pem file at the server and at the client side. The configuration at the server side (where the application is waiting for connections, in this example it is just a script) looks like this: cert = /etc/stunnel/stunnel.pem #chroot = /var/run/stunnel/ pid = /tmp/stunnel.pid setuid = nobody setgid = nogroup output = /var/log/stunnel4.log exec = /root/stest.sh verify = 1 [BreakOut] accept = 192.168.0.1:443 Here is the configuration at the client side, where a browser establish a connection cert = /etc/stunnel/stunnel.pem key = /etc/stunnel/stunnel.pem sslVersion = SSLv3 chroot = /var/lib/stunnel4/ setuid = stunnel4 setgid = stunnel4 pid = /stunnel4.pid socket = l:TCP_NODELAY=1 socket = r:TCP_NODELAY=1 verify = 0 debug = 7 output = /var/log/stunnel4/stunnel.log client = yes [BreakOut] accept = 127.0.0.1:8080 connect = 192.168.0.1:443 First of all: is it right to set 'verify = 1' at the server side, and 'verify = 0' at the client side? It seems right to me, because the client should authenticate, but not the other way. Anyhow: As soon as a connection is opened using this configuration it will terminate at the same second. renner@mars:~$ telnet localhost 8080 Trying 127.0.0.1... Connected to localhost. Escape character is '^]'. Connection closed by foreign host. In the clients logfile I see SSL state (connect): before/connect initialization SSL state (connect): SSLv3 write client hello A SSL state (connect): SSLv3 read server hello A VERIFY IGNORE: depth=0, /C=DE/ST=Germany/L=Munich/O=vbox4php/OU=stunnel/CN=boulder.vbox4php.org/emailAddress=michael.renner(a)gmx.de VERIFY OK: depth=0, /C=DE/ST=Germany/L=Munich/O=vbox4php/OU=stunnel/CN=boulder.vbox4php.org/emailAddress=michael.renner(a)gmx.de VERIFY IGNORE: depth=0, /C=DE/ST=Germany/L=Munich/O=vbox4php/OU=stunnel/CN=boulder.vbox4php.org/emailAddress=michael.renner(a)gmx.de VERIFY OK: depth=0, /C=DE/ST=Germany/L=Munich/O=vbox4php/OU=stunnel/CN=boulder.vbox4php.org/emailAddress=michael.renner(a)gmx.de SSL state (connect): SSLv3 read server certificate A SSL state (connect): SSLv3 read server certificate request A SSL state (connect): SSLv3 read server done A SSL state (connect): SSLv3 write client certificate A SSL state (connect): SSLv3 write client key exchange A SSL state (connect): SSLv3 write certificate verify A SSL state (connect): SSLv3 write change cipher spec A SSL state (connect): SSLv3 write finished A SSL state (connect): SSLv3 flush data SSL alert (read): fatal: bad certificate SSL_connect: 14094412: error:14094412:SSL routines:SSL3_READ_BYTES:sslv3 alert bad certificate Connection reset: 0 bytes sent to SSL, 0 bytes sent to socket BreakOut finished (0 left) And, more interesting, at the server side: BreakOut connected from 84.57.55.105:57987 VERIFY ERROR: depth=0, error=self signed certificate: /C=DE/ST=Germany/L=Munich/O=vbox4php/OU=stunnel/CN=boulder.vbox4php.org/emailAddress=michael.renner(a)gmx.de SSL_accept: 140890B2: error:140890B2:SSL routines:SSL3_GET_CLIENT_CERTIFICATE:no certificate returned Connection reset: 0 bytes sent to SSL, 0 bytes sent to socket Is something wrong with my self signed certificate? In my first tests I used one that I made with tinyCA, an other was made according http://www.stunnel.org/faq/certs.html: * openssl req -new -x509 -days 365 -nodes -out stunnel.pem -keyout stunnel.pem * openssl gendh 512 >> stunnel.pem * openssl rsa -in stunnel.pem -out new.pem * edit new.pem, merge some parts with stunnel.pem and rename it Any hint? Thanks -- |Michael Renner E-mail: michael.renner(a)gmx.de | |D-81541 Munich Germany ICQ: #112280325 | |Germany Don't drink as root! ESC:wq

1 0

Windows binary kits on 64 bit 2003 server or 2008 server
by Rich Jordan 24 Dec '08

24 Dec '08

Sorry to sully the list with wintel questions, but I haven't been able to locate an answer, and without systems available to test the obvious method of finding out is not available. We're running Stunnel V4.26 on a windows 2003 32-bit server, communicating with several better platforms running Stunnel V4.20 and newer, no issues. One interested customer is running windows 2003 server, 64-bit, and plans on moving to win2008 server 64 bit in the not too distant future. I do not have access to either OS/system and they're reluctant to install without knowing ahead of time what might happen. I'm hoping someone has tried the current STunnel binary kit on either or both of those systems. Will the kit run on 64-bit windows 2003, and is it usable on windows 2008? If not is this a situation where a build from source on either/both of those windows systems will produce working code? I've also found a number of hits for reputed OpenSSL ports for 64 bit windows. Any particular ones known to be good? Thanks for any info. Rich Illuc ivi, illud feci

1 0

Re: [stunnel-users] Wrapping a CherryPy server?
by Lucas Burdick 19 Dec '08

19 Dec '08

Hmm, I guess I didn't send my previous followup to the list. After setting the log to output properly, I was able to troubleshoot my problem. The chroot directory was non-existent and therefore didn't allow my .pid to be created. I have created the directory and everything now works as it should. Thanks for the help. Sorry for the trouble. Lucas -----Original Message----- From: Michael Renner [mailto:[email protected]] Sent: December-19-08 11:07 AM To: stunnel-users(a)mirt.net; Lucas Burdick Subject: Re: [stunnel-users] Wrapping a CherryPy server? On Friday 19 December 2008, Lucas Burdick wrote: > Hi There, Moin, > I?m trying to encrypt a cherrypy web page that I?m serving up. The page > itself works just fine on port 80, but I can?t seem to get stunnel to wrap > it. I?m using 4.26 on CentOS 5.2. I?ve put in my stunnel.conf: [...] > I?ve also tried commenting out the above mentioned .conf entry. I get the > error that inetd mode needs a remote host or executable. I?m trying to run > in daemon mode but I don?t think I?m doing it right. > > Please help, or ask me questions, or tell me to elaborate or something. can you call stunnel using the forground option and check the logfile? start stunnel like this ... $ stunnel -D debug -f -o /var/log/stunnel.log ........ CU -- |Michael Renner E-mail: michael.renner(a)gmx.de | |D-81541 Munich Germany ICQ: #112280325 | |Germany Don't drink as root! ESC:wq

1 0

Log rotation patch
by Christophe Nanteuil 17 Dec '08

17 Dec '08

Hello, NB : I propose this patch in alpha version. It has been tested only on a Linux Mandriva. (in the alpha version, it's recommended to use it on logs you might loose). Why to use this patch ? One reason I have to restart periodically stunnel is to do log rotation. This patch allows log rotation while running stunnel by implementing it directly inside. Therefore, the established connections are not closed by log rotation. Patch to version : 4.26 (should work on earlier versions). Type : new feature Description : Log rotation inside stunnel How to use this patch ? 3 new options are added in config file (see manual) maxLogSize : maximum size of log file (output option in config file). If logs have to be added, a log rotation is initiated. maxLogFiles : how many log files you want to keep. Default 7. useLogCompression : specify if rotated log files kept must be compressed or not. (requires zlib to be set). Just recompile stunnel after applying patch in stunnel directory: patch -p1 < /path/to/stunnel-logrot-0.10a.patch Test results are welcomed. Regards -- Christophe Nanteuil

1 0

Re: [stunnel-users] Stunnel 4 failing to connect to gmail
by Alex Limberis 15 Dec '08

15 Dec '08

HI Asif, Unfortunately that is the way gmail works. It changes the from to the gmail address rather than the computer domain. Unfortunately, this mostly makes it useless if you are trying to use it as a smarthost. -----Original Message----- From: stunnel-users-bounces(a)mirt.net [mailto:[email protected]] On Behalf Of Asif Iqbal Sent: Monday, December 15, 2008 2:06 PM To: Scott Gifford Cc: stunnel-users(a)mirt.net; nullmailer(a)lists.untroubled.org Subject: Re: [stunnel-users] Stunnel 4 failing to connect to gmail On Mon, Dec 15, 2008 at 4:49 PM, Scott Gifford <sgifford(a)suspectclass.com> wrote: > "Asif Iqbal" <vadud3(a)gmail.com> writes: > > [...] > >> I said it too soon. It passed that bind error but it still failing in ssl. >> I am seeing `SSL alert (write): fatal: handshake failure' > > Hello again Asif, glad you decided to try out stunnel! Yep, as you suggested :-) > > Try using port 465 instead of 587. That worked! I were able to send an email and I did receive it. It says the email came from `vadud3(a)gmail.com' instead of the sender, which is me(a)myhomecomputer.domain I am using this in /etc/nullmailer/remotes file localhost smtp --port=25 --user=vadud3 --pass=s3cr3t > > It looks like Google's STARTTLS command requires an EHLO to be sent > first, which is a someqhat unusual requirement. I don't think stunnel > does that. Port 465 will start talking SSL directly, without > requiring SMTP commands first. > > Good luck! > > ----Scott. > -- Asif Iqbal PGP Key: 0xE62693C5 KeyServer: pgp.mit.edu _______________________________________________ stunnel-users mailing list stunnel-users(a)mirt.net http://stunnel.mirt.net/mailman/listinfo/stunnel-users

1 0

Re: [stunnel-users] Cannot connect to SBC/yahoo to send (or telnet)
by Leandro Avila 11 Dec '08

11 Dec '08

Not sure if ranting will get you the kind of answer that you look for but it was certainly entertaining for me. Anyway... I agree, some times things can be a little frustraiting but I guess that depends on how much time on how much work you want to put into it. Sorry to hear you feel stupid. Everyone has to start from 0 at shome point and it takes time to get a good handle to how things work. The question is... do you really want to know how things work or just want them to work without having to do much? A google search of "SSL Explained" http://wildbill.nulldevice.net/presentations/sslpreso/ I liked slide 11/35 nice little diagram of SSL handshake wich helps to visualize in very general terms how an SSL connection works. The rest of the slides are good too check them out! In your particular case that forced you to change from plain-text to SSL connections to your mail server. You had a couple of options. - First and most obvious is to use SSL support build into your mail client (Thunderbird, Outlook,etc) most if not all modern mail clients support SSL natively. No need for stunnel there. - If your mail client does not support SSL then stunnel is the way to go. But your issue is on how stunnel is being used. As client. When you are a client... you always want to verify your server. You want to make sure you are connected to the right server. You want to use verify=2 in stunnel.conf How do we verify that the server I just connected is in fact the right server? That is the Certificate Authority's job. We ultimately have to trust the CA they issued the certificate for the server owner. They signed the certificate with their key. and we use the CA's certificate to verify that signature. Here is the catch... operating systems, and some programs are shipped with a prebuild collection of CA certificates. Unfortunatelly stunnel does not. However, most CAs let you download their certificates for your own use. In your particular case, looks like the server you are connecting to uses a certificate issued by Equifax Secure Certificate Authority Here they have the download page for the CA Certificates. http://www.geotrust.com/resources/root-certificates/ You pointed out that they refuse connections if the client presents a certificate. Yes, that is somehow a common practice. Mutual authentication (both server and client send certificates to each other and verify them) is impractical in some cases. Thanks ----------------- Leandro Avila

1 0

Cannot connect to SBC/yahoo to send (or telnet)
by James Moe 05 Dec '08

05 Dec '08

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello, (I sent this yesterday but that one seems to have gotten lost....) Stunnel v4.20. When connecting to SBC/Yahoo, the session is terminated with a "bad certificate" message. See the log below. The tech folks claim all is well at their end. Is there something I am missing here? Here is the conf file: ....[ conf ].... socket = l:TCP_NODELAY=1 socket = r:TCP_NODELAY=1 client = yes output = G:/c/voice/pmmdev/testcase/bin/stunnel.log verify = 0 debug = 7 cert = g:/c/voice/pmmdev/testcase/bin/sma-test.pem [sbc] accept = localhost:6325 connect = smtp.att.yahoo.com:465 ....[ end conf ].... ....[ connection log ].... 2008.11.11 00:14:17 LOG7[223:1737]: sbc accepted FD=15 from 127.0.0.1:61053 2008.11.11 00:14:17 LOG7[223:1737]: Creating a new thread 2008.11.11 00:14:17 LOG7[223:1737]: New thread created 2008.11.11 00:14:17 LOG7[251:1737]: sbc started 2008.11.11 00:14:17 LOG7[251:1737]: FD 15 in non-blocking mode 2008.11.11 00:14:17 LOG7[251:1737]: TCP_NODELAY option set on local socket 2008.11.11 00:14:17 LOG5[251:1737]: sbc accepted connection from 127.0.0.1:61053 2008.11.11 00:14:17 LOG7[251:1737]: FD 16 in non-blocking mode 2008.11.11 00:14:17 LOG7[251:1737]: sbc connecting 69.147.64.31:465 2008.11.11 00:14:17 LOG7[251:1737]: connect_wait: waiting 10 seconds 2008.11.11 00:14:17 LOG7[251:1737]: connect_wait: connected 2008.11.11 00:14:17 LOG5[251:1737]: sbc connected remote server from 192.168.69.14:61054 2008.11.11 00:14:17 LOG7[251:1737]: Remote FD=16 initialized 2008.11.11 00:14:17 LOG7[251:1737]: TCP_NODELAY option set on remote socket 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): before/connect initialization 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 write client hello A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 read server hello A 2008.11.11 00:14:17 LOG5[251:1737]: VERIFY IGNORE: depth=0, /C=US/ST=California/L=Santa Clara/O=Yahoo! Inc./OU=Yahoo/CN=smtp.att.yahoo.com 2008.11.11 00:14:17 LOG5[251:1737]: VERIFY OK: depth=0, /C=US/ST=California/L=Santa Clara/O=Yahoo! Inc./OU=Yahoo/CN=smtp.att.yahoo.com 2008.11.11 00:14:17 LOG5[251:1737]: VERIFY IGNORE: depth=0, /C=US/ST=California/L=Santa Clara/O=Yahoo! Inc./OU=Yahoo/CN=smtp.att.yahoo.com 2008.11.11 00:14:17 LOG5[251:1737]: VERIFY OK: depth=0, /C=US/ST=California/L=Santa Clara/O=Yahoo! Inc./OU=Yahoo/CN=smtp.att.yahoo.com 2008.11.11 00:14:17 LOG5[251:1737]: VERIFY IGNORE: depth=0, /C=US/ST=California/L=Santa Clara/O=Yahoo! Inc./OU=Yahoo/CN=smtp.att.yahoo.com 2008.11.11 00:14:17 LOG5[251:1737]: VERIFY OK: depth=0, /C=US/ST=California/L=Santa Clara/O=Yahoo! Inc./OU=Yahoo/CN=smtp.att.yahoo.com 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 read server certificate A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 read server certificate request A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 read server done A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 write client certificate A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 write client key exchange A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 write certificate verify A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 write change cipher spec A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 write finished A 2008.11.11 00:14:17 LOG7[251:1737]: SSL state (connect): SSLv3 flush data 2008.11.11 00:14:18 LOG7[251:1737]: SSL alert (read): fatal: bad certificate 2008.11.11 00:14:18 LOG3[251:1737]: SSL_connect: 14094412: error:14094412:SSL routines:SSL3_READ_BYTES:sslv3 alert bad certificate 2008.11.11 00:14:18 LOG5[251:1737]: Connection reset: 0 bytes sent to SSL, 0 bytes sent to socket 2008.11.11 00:14:18 LOG7[251:1737]: sbc finished (0 left) ....[ end log ].... - -- jimoe (at) sohnen-moe (dot) com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.7 (OS/2) iD8DBQFJGe4zzTcr8Prq0ZMRAhSPAJ4h6YHyR+/W5brb7FK1tbbW1zYZ+wCglxpC 9k2qqpP2hN99BL0TnsNhlnw= =P74g -----END PGP SIGNATURE-----

6 7

Test, test. Hello?
by James Moe 05 Dec '08

05 Dec '08

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hello, I have not received a posting for several days. Has the list really gone that quiet? - -- jimoe (at) sohnen-moe (dot) com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.7 (OS/2) iD4DBQFJOXHezTcr8Prq0ZMRAqqQAJ0aUC0ZyZC+85WF4WZN3nntQjfrxgCY/zjt 3Jlthcur9UTtXOnNkVcPFg== =scXZ -----END PGP SIGNATURE-----

1 0

Extending stunnels connect functionality
by bernhard.czech＠nobiscum.de 24 Nov '08

24 Nov '08

Hi all, we are looking for somebody to extend stunnel's multiple connect functionality a bit as of: - adding the ability to send a configurable string to each "connect" and compare the answer against an configured expectation string - the expectation should also be able to be interpreted as an regular expression. - the testing interval should be configurable - an alternative port should be configurable - it should be possible to call a script instead of doing a connection with the connect string provided as a parameter. - a way of monitoring might all paths might be good (i.e. writing a status file on regular basis) These things are pretty much the same things that LVS ( http://www.linuxvirtualserver.org/) provides. But we would like to reduce the complexity in our environment a bit. If somebody is interested to implement this feel free to contact me. If possible with an amount of expected reward. Kind regards, Bernhard Czech -- Bernhard Czech nobisCum Individual GmbH [T] +49 (0) 931 / 79 66 7-90 44 [F] +49 (0) 931 / 79 66 7-29 [M] +49 (0)175 / 90 51 45 8 [@] Bernhard.Czech(a)nobiscum.de [URL] www.nobiscum.de [Anschrift] Sedanstraße 23, 97082 Würzburg, Germany [Registergericht] Würzburg, HRB 5159 [Geschäftsführer] Markus Frisch, Hans-Peter Holl

1 0

Tunnel traffic from a network interface
by Kjetil Holien 23 Nov '08

23 Nov '08

Hi! Is it possible to capture traffic from one network interface on one computer, tunnel it to a second computer by stunnel and then send it through a network interface on that computer? This without changing any data on the network traffic (MAC adresses, IP etc). Thanks for any answers! greetings, Kjetil

1 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

stunnel-users