Solaris fcntl CPU/Lock update

[Pierre Belanger]

I first thought that the libraries were not threaded but it
appears they are. But you have a compile options to disable
threads.

alkonost{pbelang1}-openldap-2.1.4% ./configure --help|grep -i thread
   --with-threads          with threads [auto]

Don't know what the impacts are but since Samba is non-threaded
I guess there's no impact in compile the libraries with the
"--without-threads" option.

Pierre B.

Pierre Belanger wrote:
> Did you try to compile using OpenLDAP? I think that can
> be a "quick" fix.
> 
> Otherwise, open a ticket with Sun?
> 
> Pierre B.
> 
> Jeff Mandel wrote:
> 
>>>
>>>
>>>> trace from 12327:
>>>> (gdb) bt
>>>> #0  0xfecd9794 in __sigprocmask () from /usr/lib/libthread.so.1
>>>> #1  0xfecce1e8 in _deliversigs () from /usr/lib/libthread.so.1
>>>> #2  0xfecd05c4 in thr_sigsetmask () from /usr/lib/libthread.so.1
>>>> #3  <signal handler called>
>>>> #4  0xb in ?? ()
>>>> #5  0xfecdb1f0 in usleep () from /usr/lib/libthread.so.1
>>>> #6  0xf7d1c in do_lock_spin (fsp=0x26da48, conn=0x257c68, lock_pid=0,
>>>>    count=20, offset=2147483559, lock_type=WRITE_LOCK) at 
>>>> locking/locking.c:175
>>>> #7  0x58bb4 in reply_lockingX (conn=0x257c68, inbuf=0x2173c9 "",
>>>>    outbuf=0x237819 "", length=75, bufsize=16644) at smbd/reply.c:4714
>>>> #8  0x739bc in switch_message (type=36, inbuf=0x2173c9 "", 
>>>> outbuf=0x237819 "",
>>>>    size=75, bufsize=16644) at smbd/process.c:774
>>>> #9  0x73a48 in construct_reply (inbuf=0x2173c9 "", outbuf=0x237819 "",
>>>>    size=75, bufsize=16644) at smbd/process.c:803
>>>> #10 0x73cf4 in process_smb (inbuf=0x2173c9 "", outbuf=0x237819 "")
>>>>    at smbd/process.c:897
>>>> #11 0x74814 in smbd_process () at smbd/process.c:1294
>>>> #12 0x314b4 in main (argc=0, argv=0xffbefeac) at smbd/server.c:832
>>>> (gdb) The program is running.  Quit anyway (and detach it)? (y or n) y
>>>>   
>>>
>>>
>>>
>>> This is a much more interesting backtrace than the
>>> other. Why is smbd linking in pthread libraries ?
>>> smbd is *NOT* a threaded program.
>>>
>>> The backtrace you have here shows smbd trying to get
>>> a fcntl lock on behalf of a client and failing to get it
>>> instantaneously, so going into the lock spin code. It
>>> will try 3 times, sleeping for 10 usec between each attempt,
>>> and then return a fail to the client. This is not in itself
>>> a spinning bug or problem, smbd is meant to do this.
>>>
>>> I'm worried about the interaction between the Solaris lwp
>>> libc code and smbd.... smbd should be a simple program
>>> with only one thread of execution.
>>>
>>> Jeremy.
>>>
>>>  
>>>
>> Since it seems that libthread shows up because nss_ldap (padl's) is 
>> used, does anyone have suggestions for dealing with this at the 
>> nss_ldap level?
>>
>> Jeff
>>
> 
>