[Samba] smbd+ldapsam infinite loop, cpu=100%

[Jason C. Waters]

Whats your log level set at?  I've had it set at 10 before and its 
pegged out the processor.  Maybe thats all it is.

The Ranger wrote:
> Hello,
> 
> We have been running samba v2 for quite some time without any problems. 
> But after the upgrade to samba-3.0.5 some anomalies have come out. I've 
> compiled it with LDAP authentication support; the LDAP server is 
> installed on the same machine  and the communication is done via sockets.
> 
> After working fine for some time, and as usual creating multiple threads 
> from the smbd proccess, it suddenly goes into an infinite loop;
> 
> Running strace on that thread shows:
> 
> open("/etc/passwd", O_RDONLY)           = 28
> fcntl64(28, F_GETFD)                    = 0
> fcntl64(28, F_SETFD, FD_CLOEXEC)        = 0
> fstat64(28, {st_mode=S_IFREG|0644, st_size=852, ...}) = 0
> old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, 
> -1, 0) = 0x4033d000
> read(28, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 852
> read(28, "", 4096)                      = 0
> close(28)                               = 0
> munmap(0x4033d000, 4096)                = 0
> rt_sigaction(SIGPIPE, {SIG_IGN}, {SIG_DFL}, 8) = 0
> rt_sigprocmask(SIG_UNBLOCK, [PIPE], NULL, 8) = 0
> getpid()                                = 19340
> geteuid32()                             = 65534
> getpid()                                = 19340
> geteuid32()                             = 65534
> time(NULL)                              = 1093339331
> write(26, "0\201\305\2\3%\302\233c\201\275\4\25ou=users,dc=21k,dc"..., 
> 200) = 200
> select(1024, [26], [], NULL, NULL)      = 1 (in [26])
> read(26, "0\16\2\3%\302\233e", 8)       = 8
> read(26, "\7\n\1\0\4\0\4\0", 8)         = 8
> time(NULL)                              = 1093339331
> time([1093339331])                      = 1093339331
> rt_sigaction(SIGPIPE, {SIG_DFL}, {SIG_IGN}, 8) = 0
> rt_sigprocmask(SIG_UNBLOCK, [PIPE], NULL, 8) = 0
> open("/etc/passwd", O_RDONLY)           = 28
> fcntl64(28, F_GETFD)                    = 0
> fcntl64(28, F_SETFD, FD_CLOEXEC)        = 0
> fstat64(28, {st_mode=S_IFREG|0644, st_size=852, ...}) = 0
> old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, 
> -1, 0) = 0x4033d000
> read(28, "root:x:0:0:root:/root:/bin/bash\n"..., 4096) = 852
> read(28, "", 4096)                      = 0
> close(28)                               = 0
> munmap(0x4033d000, 4096)                = 0
> rt_sigaction(SIGPIPE, {SIG_IGN}, {SIG_DFL}, 8) = 0
> 
> .....
> 
> 
> While normally working smbd threads are owen by the user who is logged 
> into the domain, the looping ones are always run by the nobody-user:
> 
> nobody   19339  4.3  1.2  7012 3976 ?        S    02:26  29:07 
> /usr/local/samba/sbin/smbd -D
> nobody   19341  4.2  1.2  7012 3976 ?        S    02:26  28:30 
> /usr/local/samba/sbin/smbd -D
> nobody   19342  4.0  1.2  7012 3980 ?        S    02:26  26:56 
> /usr/local/samba/sbin/smbd -D
> nobody   19343  3.9  1.2  7012 3976 ?        S    02:26  26:15 
> /usr/local/samba/sbin/smbd -D
> 
> 
> I've tested the exactly the same configuration on the other machine and 
> it works fine, no looping.
> 
> 
> Any ideas, how to solve this problem or what to look for?
> 
> 
> 
> -- 
> Best wishes,
> The Ranger
>