nfs client hanging

[J. Bruce Fields]

On Wed, Jul 23, 2008 at 04:27:24PM +0200, Guillaume Rousse wrote:
> J. Bruce Fields a écrit :
>> Thanks; let us know what you're able to figure out.
> I finally managed to get one, and it seems the problem actually stand in  
> ldap libraries when resolving users:
>
> (gdb) bt
> #0  0x00007fcc3c7a9a55 in raise () from /lib64/libc.so.6
> #1  0x00007fcc3c7ab1e0 in abort () from /lib64/libc.so.6
> #2  0x00007fcc3c7a2d6f in __assert_fail () from /lib64/libc.so.6
> #3  0x00007fcc3d6f8ca9 in ber_flush2 () from /usr/lib64/liblber-2.4.so.2
> #4  0x00007fcc3d922aac in ldap_int_flush_request ()
>    from /usr/lib64/libldap-2.4.so.2
> #5  0x00007fcc3d922f43 in ldap_send_server_request ()
>    from /usr/lib64/libldap-2.4.so.2
> #6  0x00007fcc3d923099 in ldap_send_initial_request ()
>    from /usr/lib64/libldap-2.4.so.2
> #7  0x00007fcc3d911ccd in ldap_search () from /usr/lib64/libldap-2.4.so.2
> #8  0x00007fcc3d911e4a in ldap_search_st () from /usr/lib64/libldap-2.4.so.2
> #9  0x00007fcc3c13a093 in do_search_s (
>     base=0x7fcc3c347d94 "ou=users,dc=msr-inria,dc=inria,dc=fr", scope=2,
>     filter=0x7fff4659e9f0 "(&(objectClass=posixAccount)(uid=arbelaez))",
>     attrs=0x7fcc3c348840, sizelimit=1, res=0x7fff4659f278) at  
> ldap-nss.c:2705
> #10 0x00007fcc3c1391c9 in do_with_reconnect (
>     base=0x7fcc3c347d94 "ou=users,dc=msr-inria,dc=inria,dc=fr", scope=2,
>     filter=0x7fff4659e9f0 "(&(objectClass=posixAccount)(uid=arbelaez))",
>     attrs=0x7fcc3c348840, sizelimit=1, private=0x7fff4659f278,
>     search_func=0x7fcc3c139ff0 <do_search_s>) at ldap-nss.c:2596
> #11 0x00007fcc3c139e13 in _nss_ldap_search_s (args=0x7fff4659f2f0,
>     filterprot=0x7fcc3c34c320 "(&(objectClass=posixAccount)(uid=%s))",
>     sel=LM_PASSWD, user_attrs=0xffffffffffffffff, sizelimit=1,
>     res=0x7fff4659f278) at ldap-nss.c:3120
> #12 0x00007fcc3c13a649 in _nss_ldap_getbyname (args=0x7fff4659f2f0,
>     result=0xb5b760, buffer=0xb5b790 "svn", buflen=1024,
>     errnop=0x7fcc3e58f690,
>     filterprot=0x7fcc3c34c320 "(&(objectClass=posixAccount)(uid=%s))",
>     sel=LM_PASSWD, parser=0x7fcc3c13ae80 <_nss_ldap_parse_pw>)
>     at ldap-nss.c:3467
> #13 0x00007fcc3c13ae57 in _nss_ldap_getpwnam_r (name=<value optimized out>,
>     result=0x110f, buffer=0x6 <Address 0x6 out of bounds>,
>     buflen=18446744073709551615, errnop=0x0) at ldap-pwd.c:210
> #14 0x00007fcc3c812188 in __getpwnam_r (name=0xc3ff30 "arbelaez",
>     resbuf=0xb5b760, buffer=0xb5b790 "svn", buflen=1024,  
> result=0x7fff4659f3c0)
>     at ../nss/getXXbyYY_r.c:226
> #15 0x00007fcc3db47e32 in write () from /usr/lib64/libnfsidmap.so.0
> #16 0x00007fcc3db48001 in write () from /usr/lib64/libnfsidmap.so.0
> #17 0x0000000000404002 in ?? ()
> #18 0x0000000000404255 in ?? ()
> #19 0x00007fcc3e175748 in event_base_loop () from  
> /usr/lib64/libevent-1.4.so.2
> #20 0x00000000004048c5 in ?? ()
> #21 0x00007fcc3c797074 in __libc_start_main (main=0x404380, argc=1,
>     ubp_av=0x7fff4659fbb8, init=<value optimized out>,
>     fini=<value optimized out>, rtld_fini=<value optimized out>,
>     stack_end=0x7fff4659fba8) at libc-start.c:220
> #22 0x0000000000401ed9 in write ()
> #23 0x00007fff4659fba8 in ?? ()
> #24 0x0000000000000000 in ?? ()
>
> Unfortunatly, due to missing symbols in the debug package, I can't  
> identify which is the exact issue, apart a failed assertion in liblber  
> ber_flush2() function.
>
> Anyway, even if the problem seems to be not directly related to nfs4,  
> could rpc.idpmapd be made more robust for this kind of issues ?

We'd need to understand better exactly what the problem is and why it's
happening.  If you don't have the time to dig through the ldap sources
to figure out what makes that assertion fail, then I guess the next
thing to do would be to try to get help from whoever's responsible for
the ldap code, and/or file a bug against whichever package it comes
from.

--b.

--b.