On 04/08/10 16:07 +0900, Jorgen Lundman wrote:
>
> Since the message that we frequently see before we have trouble comes from
> this code:
>
> 6719 oo = rfs4_findopenowner(owner, &create, args->seqid);
> 6720 if (oo == NULL) {
> 6721 *cs->statusp = resp->status = NFS4ERR_STALE_CLIENTID;
>
>
>
> and the function rfs4_findopenowner() calls rfs4_dbsearch(), whose first
> line is:
>
> 522 NFS4_DEBUG(table->debug & SEARCH_DEBUG,
> 523 (CE_NOTE, "Searching for key %p in table %s by %s",
> 524 key, table->name, idx->keyname));
>
> I thought it would be interesting to enable this message.
>
> Ie, set SEARCH_DEBUG for NFS4_DEBUG. It would appear that it comes from:
>
> rfs4_table_create()
>
> rfs4_database_create(uint32_t flags)
>
> rfs4_server_state = rfs4_database_create(rfs4_database_debug);
>
> static uint32_t rfs4_database_debug = 0x00;
>
>
>
> And since SEARCH_DEBUG is 0x0001, I have attempted to set:
>
> # mdb -kw
> > rfs4_database_debug/D
> rfs4_database_debug:
> rfs4_database_debug: 0
> > rfs4_database_debug/W0t1
> > rfs4_database_debug/D
> rfs4_database_debug:
> rfs4_database_debug: 1
>
> but also I added /etc/system
>
> set nfssrv:rfs4_database_debug=1
>
> with a reboot.
>
>
>
> I can confirm that rfs4_database_debug is now set to 1.
>
> But I don't see a huge amount of debug info that I somewhat expected to
> see. It uses CNM_NOTE, which I assume will be picked up by kern.debug and
> placed in /var/adm/messages. Is this not correct?
>
I haven't debugged nfsv4 state database tables yet, but the steps
which you have outlined above look ok. Can you verify from mdb
that SEARCH_DEBUG is set in openowner table?
<snip>
rfs4_openowner_idx::print rfs4_index_t dbi_table->dbt_debug
</snip>
If it's not set then you might solve this problem by setting
dbt_debug for this particular table to 1.
cheers
-jan
_______________________________________________
nfs-discuss mailing list
[email protected]
