Milter-greylist

Hello list!

I've compiled and installed milter-greylist-1.6rc1 on Solaris 9 (sparc).
It works fine to compile if you install the following:
- gcc
- flex
- bison
(and add the option CFLAGS=-D__EXTENSIONS__ to the Makefile)

(This, because of that Solaris lex and yacc is missing to many important feature)

Everything works fine, except for "peer" option (replication of greylist/whitelist data between MX-peers). I've tried to truss and "debug" as much as i can with the help from Emmanuel Dreyfus, but i'm stuck...

The message i get in the syslog looks like this:

Dec 13 20:00:55 gemini milter-greylist: [ID 226885 mail.debug] iBDJ0rOa004257: addr = 151.156.192.51, from = <root@...>, rcpt = <kalle@...>
Dec 13 20:00:55 gemini milter-greylist: [ID 173742 mail.debug] created: 151.156.192.51 from <root@...> to <kalle@....se> delayed for 60s
Dec 13 20:00:55 gemini milter-greylist: [ID 928669 mail.info] iBDJ0rOa004257: addr 151.156.192.51 from <root@...> to <kalle@...> delayed for 00:01:00
Dec 13 20:00:55 gemini sendmail[4257]: [ID 801593 mail.info] iBDJ0rOa004257: milter=greylist, action=rcpt, reject=451 4.7.1 Greylisting in action, please come back in 00:01:00
Dec 13 20:00:55 gemini milter-greylist: [ID 795507 mail.debug] sync_sender running
Dec 13 20:00:55 gemini milter-greylist: [ID 634049 mail.error] local_addr: bind failed: Error 0
Dec 13 20:00:55 gemini milter-greylist: [ID 750011 mail.debug] sync_sender sleeping, done 0 entries in 0.010411s

Apparently "sync_sender" routine fails (problem binding a free port on the local machine?). I've also snoop:ed the network, and i can't see any milter-greylist "sync"-trafic. Has someone running Solaris noticed or fixed this problem for version 1.6rc1?
If so, please send a patch to Emmanuel Dreyfus...

I've noticed that most milter coding is done on BSD-systems.
But the company that i work for prefers Solaris :-(

TIA,
/P

Hi,

>>>>> On Tue, 14 Dec 2004 12:00:25 +0100
>>>>> <fredrik.pettai@...> said:

pettai> Dec 13 20:00:55 gemini milter-greylist: [ID 634049 mail.error] local_addr: bind failed: Error 0

It's strange.  It seems bind() returns an error without setting errno.
Please try following patch.  Solaris9 might not allow bind to port 0.

Index: sync.c
diff -u -p sync.c.orig sync.c
--- sync.c.orig	Fri Nov 12 23:22:22 2004
+++ sync.c	Tue Dec 14 23:21:44 2004
@@ -1176,12 +1176,12 @@ local_addr(sa, salen)
 	memcpy(&addr, sa, salen);
 	switch(sa->sa_family) {
 	case AF_INET:
-		SA4(&addr)->sin_port = 0;
+		SA4(&addr)->sin_port = 1;
 		break;
 
 #ifdef AF_INET6
 	case AF_INET6:
-		SA6(&addr)->sin6_port = 0;
+		SA6(&addr)->sin6_port = 1;
 		break;
 #endif
 

--
Hajimu UMEMOTO @ Internet Mutual Aid Society Yokohama, Japan
ume@...  ume@{,jp.}FreeBSD.org
http://www.imasy.org/~ume/

On Tue, Dec 14, 2004 at 12:00:25PM +0100, fredrik.pettai@... wrote:
> Dec 13 20:00:55 gemini milter-greylist: [ID 795507 mail.debug] sync_sender running
> Dec 13 20:00:55 gemini milter-greylist: [ID 634049 mail.error] local_addr: bind failed: Error 0
> Dec 13 20:00:55 gemini milter-greylist: [ID 750011 mail.debug] sync_sender sleeping, done 0 entries in 0.010411s

Could you please provide output from "truss -f" for this bug?

	Kind regards

-- 
Matthias Scheler                                  http://scheler.de/~matthias/

On Tue, Dec 14, 2004 at 11:29:36PM +0900, Hajimu UMEMOTO wrote:
> It's strange.  It seems bind() returns an error without setting errno.

It shouldn't. If it does it is a bug. But we need "truss -f" output to
know what's going wrong there.

> Please try following patch.

That patch is highly broken. The bind() call might be rejected because
port 1 is a privileged port and/or in use.

> Solaris9 might not allow bind to port 0.

It's part of the socket API and therefore supported.

[Yes, I'm 100% sure because I actually maintain that code in Solaris.]

	Kind regards

-- 
Matthias Scheler                                  http://scheler.de/~matthias/

On Thu, Dec 16, 2004 at 05:09:28AM +0000, Matthias Scheler wrote:
> 
> On Tue, Dec 14, 2004 at 12:00:25PM +0100, fredrik.pettai@... wrote:
> > Dec 13 20:00:55 gemini milter-greylist: [ID 795507 mail.debug] sync_sender running
> > Dec 13 20:00:55 gemini milter-greylist: [ID 634049 mail.error] local_addr: bind failed: Error 0
> > Dec 13 20:00:55 gemini milter-greylist: [ID 750011 mail.debug] sync_sender sleeping, done 0 entries in 0.010411s
> 
> Could you please provide output from "truss -f" for this bug?

Better make that "truss -v all -o trussfile -f ...".

	Kind regards

-- 
Matthias Scheler                                  http://scheler.de/~matthias/

On Thu, Dec 16, 2004 at 05:16:57AM +0000, Matthias Scheler wrote:
> > Could you please provide output from "truss -f" for this bug?
> Better make that "truss -v all -o trussfile -f ...".

Here is the output which Fredrik kindly provided:

7906/4:         so_socket(PF_INET, SOCK_DGRAM, IPPROTO_UDP, "", 1) = 8
7906/4:         bind(8, 0x000632E8, 16, 3)                      Err#126 EADDRNOTAVAIL
7906/4:                 AF_INET  name = 151.156.192.100  port = 0

So bind() worked as expected. The question is why errno got overwritten.
Looking at the code the code in "sync.c" ...

	if (bind(sfd, sa, salen) == -1) {
		if (errno != EADDRNOTAVAIL) {
			syslog(LOG_ERR, "local_addr: bind failed: %s\n",
			    strerror(errno));
			islocal = -1;
		} else {
			islocal = 0;
		}
	} else {
		islocal = 1;
	}

... I cannot see why it was overwritten unless this code has been
interrupted by a signal handler.

	Kind regards 

-- 
Matthias Scheler                                  http://scheler.de/~matthias/

--- In milter-greylist@yahoogroups.com, Matthias Scheler <tron@z...>
wrote:
> On Thu, Dec 16, 2004 at 05:16:57AM +0000, Matthias Scheler wrote:
> > > Could you please provide output from "truss -f" for this bug?
> > Better make that "truss -v all -o trussfile -f ...".
> 
> Here is the output which Fredrik kindly provided:
> 
> 7906/4:         so_socket(PF_INET, SOCK_DGRAM, IPPROTO_UDP, "", 1) = 8
> 7906/4:         bind(8, 0x000632E8, 16, 3)                     
Err#126 EADDRNOTAVAIL
> 7906/4:                 AF_INET  name = 151.156.192.100  port = 0
> 
> So bind() worked as expected. The question is why errno got overwritten.
> Looking at the code the code in "sync.c" ...
> 
> 	if (bind(sfd, sa, salen) == -1) {
> 		if (errno != EADDRNOTAVAIL) {
> 			syslog(LOG_ERR, "local_addr: bind failed: %s\n",
> 			    strerror(errno));
> 			islocal = -1;
> 		} else {
> 			islocal = 0;
> 		}
> 	} else {
> 		islocal = 1;
> 	}
> 
> ... I cannot see why it was overwritten unless this code has been
> interrupted by a signal handler.

Could it be due to lack of some #define that would make errno
thread-safe?  Just a wild guess...

BTW, 'syslog(LOG_ERR, "local_addr: bind failed: %m")' is a bit shorter :)

Eugene

On Thu, Dec 16, 2004 at 03:46:16PM -0000, egcrosser wrote:
> Could it be due to lack of some #define that would make errno
> thread-safe?  Just a wild guess...

Yes, indeed. The program should be compiled with "cc -mt ..." and linked
with "-pthread". I'm not sure what the correct options for GCC are.

	Kind regards

-- 
Matthias Scheler                                  http://scheler.de/~matthias/

Matthias Scheler <tron@...> wrote:

> ... I cannot see why it was overwritten unless this code has been
> interrupted by a signal handler.

You would see it in the kernel trace, woudln't you?

-- 
Emmanuel Dreyfus
Il y a 10 sortes de personnes dans le monde: ceux qui comprennent 
le binaire et ceux qui ne le comprennent pas.
manu@...

egcrosser <egcrosser@...> wrote:

> BTW, 'syslog(LOG_ERR, "local_addr: bind failed: %m")' is a bit shorter :)

I didn't know that one. That's nifty. Is it portable?

-- 
Emmanuel Dreyfus
Il y a 10 sortes de personnes dans le monde: ceux qui comprennent 
le binaire et ceux qui ne le comprennent pas.
manu@...

Matthias Scheler <tron@...> wrote:

> > Could it be due to lack of some #define that would make errno
> > thread-safe?  Just a wild guess... 
> Yes, indeed. The program should be compiled with "cc -mt ..." and linked
> with "-pthread". I'm not sure what the correct options for GCC are.

If someone wants to contribute patches to configure.ac so that
milter-greylist build automatically take care of that, don't hesitate.

-- 
Emmanuel Dreyfus
Il y a 10 sortes de personnes dans le monde: ceux qui comprennent 
le binaire et ceux qui ne le comprennent pas.
manu@...

On Thu, 16 Dec 2004 manu@... wrote:
> egcrosser <egcrosser@...> wrote:
> > BTW, 'syslog(LOG_ERR, "local_addr: bind failed: %m")' is a bit shorter :)
> I didn't know that one. That's nifty. Is it portable?

It appears to be a gcc-specific extension.

-Dan

--- In milter-greylist@yahoogroups.com, Dan Hollis <goemon@a...> wrote:
> On Thu, 16 Dec 2004 manu@n... wrote:
> > egcrosser <egcrosser@y...> wrote:
> > > BTW, 'syslog(LOG_ERR, "local_addr: bind failed: %m")' is a bit
shorter :)
> > I didn't know that one. That's nifty. Is it portable?
> 
> It appears to be a gcc-specific extension.

Nope.
I beleive it's standard.  I still remember Unices that did not have
syslog() at all, but no one that had syslog() without "%m".

Eugene

FreeBSD xxx.xxxx.xx 4.10-STABLE
"The message is identical to a printf(3) format string, except that `%m'
     is replaced by the current error message."

SunOS xxxxx 5.7
"occurrences  of  %m  in  the  format
     string  pointed  to  by the message argument are replaced by
     the error message string associated with the  current  value
     of errno."

etc.

On Fri, 17 Dec 2004, egcrosser wrote:
> --- In milter-greylist@yahoogroups.com, Dan Hollis <goemon@a...> wrote:
> > On Thu, 16 Dec 2004 manu@n... wrote:
> > > egcrosser <egcrosser@y...> wrote:
> > > > BTW, 'syslog(LOG_ERR, "local_addr: bind failed: %m")' is a bit shorter :)
> > > I didn't know that one. That's nifty. Is it portable?
> > It appears to be a gcc-specific extension.
> Nope.
> I beleive it's standard.  I still remember Unices that did not have
> syslog() at all, but no one that had syslog() without "%m".

I did some research, I guess you can safely use it for syslog _only_, but 
using %m in any other format strings (printf etc) cannot be guaranteed on 
anything other than gnu libc.

-Dan

On Thu, Dec 16, 2004 at 08:40:42PM +0100, Emmanuel Dreyfus wrote:
> > ... I cannot see why it was overwritten unless this code has been
> > interrupted by a signal handler.
> You would see it in the kernel trace, woudln't you?

Yes, I think so. That makes the "not compiled for threads" option most likely.

	Kind regards

-- 
Matthias Scheler                                  http://scheler.de/~matthias/

Hi, 
  I am new to this list and have been running milter-greylist for
about 2 weeks.  Now i am trying to sync between 2 MTAs and am getting
this same problem.  I have read all that I could find in this thread
but could not find a answer on how to fix this.  Could someone point
me in the right direction please.

Thanks
Keith



--- In milter-greylist@yahoogroups.com, <fredrik.pettai@v...> wrote:
> 
> Hello list!
> 
> I've compiled and installed milter-greylist-1.6rc1 on Solaris 9 (sparc).
> It works fine to compile if you install the following:
> - gcc
> - flex
> - bison
> (and add the option CFLAGS=-D__EXTENSIONS__ to the Makefile)
> 
> (This, because of that Solaris lex and yacc is missing to many
important feature)
> 
> Everything works fine, except for "peer" option (replication of
greylist/whitelist data between MX-peers). I've tried to truss and
"debug" as much as i can with the help from Emmanuel Dreyfus, but i'm
stuck...
> 
> The message i get in the syslog looks like this:
> 
> Dec 13 20:00:55 gemini milter-greylist: [ID 226885 mail.debug]
iBDJ0rOa004257: addr = 151.156.192.51, from = <root@b...>, rcpt =
<kalle@g...>
> Dec 13 20:00:55 gemini milter-greylist: [ID 173742 mail.debug]
created: 151.156.192.51 from <root@b...> to <kalle@g...> delayed for 60s
> Dec 13 20:00:55 gemini milter-greylist: [ID 928669 mail.info]
iBDJ0rOa004257: addr 151.156.192.51 from <root@b...> to <kalle@g...>
delayed for 00:01:00
> Dec 13 20:00:55 gemini sendmail[4257]: [ID 801593 mail.info]
iBDJ0rOa004257: milter=greylist, action=rcpt, reject=451 4.7.1
Greylisting in action, please come back in 00:01:00
> Dec 13 20:00:55 gemini milter-greylist: [ID 795507 mail.debug]
sync_sender running
> Dec 13 20:00:55 gemini milter-greylist: [ID 634049 mail.error]
local_addr: bind failed: Error 0
> Dec 13 20:00:55 gemini milter-greylist: [ID 750011 mail.debug]
sync_sender sleeping, done 0 entries in 0.010411s
> 
> Apparently "sync_sender" routine fails (problem binding a free port
on the local machine?). I've also snoop:ed the network, and i can't
see any milter-greylist "sync"-trafic. Has someone running Solaris
noticed or fixed this problem for version 1.6rc1?

> If so, please send a patch to Emmanuel Dreyfus...
> 
> I've noticed that most milter coding is done on BSD-systems.
> But the company that i work for prefers Solaris :-(
> 
> TIA,
> /P