Re: Problems with netgroups and ldap.

[Mark Burgess]

I could not find the function on my new linux. It looks like a thread
rentrancy problem - very strange. Is this a multiprocessor machine? 

It might take me a little while to look at this as I am about to travel
to the US for a week.

M

On Fri, 2005-04-08 at 10:12 -0700, Mark Keller wrote:
> >> Mark, this sounds odd. Nothing has changed here in the cfengine code for
> > many years. I don't think the innetgr function is portable. It looks
> > solarissy. Have you tried doing an strace (truss) to see exactly what it
> > gets stuck on?
> >
> 
> The innetgr function seems to be on Solaris, Redhat EL3, Fedora, HPUX 11, 
> FreeBSD, and Mac OS X in netdb.h (just doing a quick google search anyhow). I 
> suppose it might not be on all OS's though. Am-utils uses it exclusively for 
> netgroup calls for all OS's it supports and so does pam_access, which are 
> working for me.
> 
> So I added a couple of debug statements to the original code:
>    case netgroup:  setnetgrent(ebuff);
>                    Debug1("set netgroup to %s\n",ebuff);
> 
>                    while (getnetgrent(&machine,&user,&domain))
>                       {
>                       Debug1("getnetgrent m = %s, u = %s, d = %s in netgroup 
> %s\n",machine,user,domain,ebuff); 
>                       if (strcmp(machine,VDEFAULTBINSERVER.name) == 0)
>                          {
>                          Debug1("Matched %s in netgroup %s\n",machine,ebuff);
>                          AddClassToHeap(GROUPBUFF);
>                          break;
>                          }
> ...truncated...
>                       }
>                    
>                    endnetgrent();
>                    break;
> 
> Running "cfagent -Bvq -d1", I get the following:
> 
> ....truncated.....
> ==============================BEGIN NEW ACTION Groups:=============
> 
> 
> Resetting CLASS to ANY
> 
> LVALUE betahosts
> HandleLVALUE(betahosts) in action Groups:
> EQUALS =
> LEFTBRACK
> RVAL-VAROBJ address@hidden
> 
> HandleGroupRvalue(address@hidden)
> Netgroup rval, lookup NIS group (ops-beta-sos5)
> ExpandVarstring(ops-beta-sos5)
> HandleGroupRVal(ops-beta-sos5) group (betahosts), type=1
> set netgroup to ops-beta-sos5
> cfengine:: Time out
> 
> It takes a couple of minutes before I get the "Time out" error and then 
> cfagent just sits there for an extremely long time.
> 
> 
> Now Running "truss -f cfagent -Bvq -d1", I get the following:
> 
> ....truncated.......
> 8328:   write(5, "160301\08610\0\082\0809F".., 182)     = 182
> 8328:   read(5, "140301\001", 5)                        = 5
> 8328:   read(5, "01", 1)                                = 1
> 8328:   read(5, "160301\0  ", 5)                        = 5
> 8328:   read(5, " O 09D12 G1EF981 kF59D b".., 32)       = 32
> 8328:   time()                                          = 1112978413
> 8328:   write(5, "170301\01ED8 ]BE 3 =B0 ]".., 35)      = 35
> 8328:   time()                                          = 1112978413
> 8328:   poll(0xFFBF9708, 1, 30000)                      = 1
> 8328:   read(5, "170301\01E", 5)                        = 5
> 8328:   read(5, "C8C5A8C1EB N $1A19 g H05".., 30)       = 30
> 8328:   time()                                          = 1112978413
> 8328:   setsockopt(5, SOL_SOCKET, SO_KEEPALIVE, 0xFFBFB8D0, 4, 1) = 0
> 8328:   fcntl(5, F_SETFD, 0x00000001)                   = 0
> 8328:   getsockname(5, 0xFEFF00C0, 0xFFBFB8CC, 1)       = 0
> 8328:   getpeername(5, 0xFEFF00D0, 0xFFBFB8C8, 1)       = 0
> 8328:   time()                                          = 1112978413
> 8328:   getpid()                                        = 8328 [8327]
> 8328:   getuid()                                        = 0 [0]
> 8328:   time()                                          = 1112978413
> 8328:   write(5, "170301\095D58BA9\r FAE S".., 154)     = 154
> 8328:   poll(0xFFBF96B0, 1, -1)                         = 1
> 8328:   read(5, "17030101 .", 5)                        = 5
> 8328:   read(5, "\tF20215D380\0B3A41BDD83".., 302)      = 302
> 8328:   poll(0xFFBF96B0, 1, -1)                         = 1
> 8328:   read(5, "170301\01E", 5)                        = 5
> 8328:   read(5, "9CBC\b G02 [0691B2\nFBF0".., 30)       = 30
> 8328:   time()                                          = 1112978413
> 8328:   time()                                          = 1112978413
> 8328:   sigaction(SIGPIPE, 0xFFBFC0C0, 0x00000000)      = 0
> set netgroup to ops-beta-sos5
> 8328:   write(1, " s e t   n e t g r o u p".., 30)      = 30
> 8328:   sigaction(SIGPIPE, 0xFFBFC118, 0xFEFF10E0)      = 0
> 8328:   mmap(0x00000000, 110592, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANON, 
> -1, 0) = 0xFEDA0000
> 8328:   lwp_park(0x00000000, 0)         (sleeping...)
> 8328:       Received signal #14, SIGALRM, in lwp_park() [caught]
> 8328:   lwp_park(0x00000000, 0)                         Err#4 EINTR
> 8328:   sigprocmask(SIG_SETMASK, 0xFFBFBC54, 0x00000000) = 0
> 8328:   alarm(0)                                        = 0
> cfengine:: Time out
> 8328:   write(1, " c f e n g i n e : :   T".., 20)      = 20
> 8328:   sigprocmask(SIG_SETMASK, 0xFF07A074, 0xFFBFBA08) = 0
> 8328:   lwp_unpark(1, 1)                                = 0
> 8328:   setcontext(0xFFBFBA18)
> 8328:   lwp_park(0x00000000, 0)                         = 0
> 8328:   lwp_park(0x00000000, 0)         (sleeping...)
> 
> Here is takes a couple of minutes between the first lwp_park and the "Time 
> out" error, then stays on the last lwp_park for an extremely long time. Might 
> be infinite, but it is at least hours. If you need more trussing please let 
> me know.
> 
> So to me seems like something is going wrong around the getnetgrent function. 
> I can't tell what might be the problem as my C coding/debugging is rusty. My 
> small test code using getnetgrent seems to work just fine.
> 
> Thanks,
> 
> Mark Keller
> 
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> Bug-cfengine mailing list
> address@hidden
> http://lists.gnu.org/mailman/listinfo/bug-cfengine