Sun Fire 15K Server
 
Sun Fire E20K Server
 
Sun Fire E25K Server
 
Sun Fire 12K Server
 

PLA-Support>Sun Systems>SPARC>Enterprise>SN-SPARC: SF-Exxk
 
_Old GCS Categories>Sun Microsystems>Boards>Sun Fire Link
 

Applies to:

Goal

showdevices gathers device information from one or more Sun Fire[TM] high-end system domains. The command uses dca(1M) as a proxy to gather the information from the domains.

This article discusses a specific condition where "showdevices -d <domain-tag>" command would fail as follows :

v4u-15ka-sc1:sms-svc:24> showdevices -d A
Unable to get device information from domain A

However, besides the above anomaly, all other interactions between the main system controller (SC) and the domain proceeds as expected:

SC -- rcfgadm -d <domain-tag> proceeds successfully
DR ops initiated from the SC -- addboard and/or deleteboard proceeds successfully
DR ops initiated from the SC -- rcfgadm -c disconnect and/or rcfgadm -c configure proceeds successfully

Solution

The following messages logs were observed from the "showdevices" failure:

Domain OS's /var/adm/messages

     Sep  8 12:27:27 v4u-15ka-a dcs: [ID 985267 daemon.error] <12073> resource info init error (1)

SMS platform logs

     Sep  8 12:27:05 2005 v4u-15ka-sc1 showdevices[18788]: [0 3638613703618192 ERR ri_init.cc 85] rcfgaRequestProxy->ri_init failed. status= 4318

SMS domain logs

     Sep  8 12:27:04 2005 v4u-15ka-sc1 dca[7635]-A(): [4318 3638612328642665 ERR DCSInterface.cc 378] message receive failed: DCSInterface :: receiveResponse errCode:500

The above condition persists in spite of the fact that the DCA service on the SC and the DCS service on the domain OS have been terminated and then restarted. In fact, the problem persisted in spite of the fact that domain was keyswitched off/on.

The following observations were made through the course of further investigations.

Applying truss to the DCS service on the domain OS in parallel with executing "showdevices" from the main SC revealed the following:

After resolving the symbolic link of a specific path-name involving "/dev/wci17d" using resolvepath():

     12105:  resolvepath("/dev/wci17d", "/devices/SUNW,wci@17d,0:SUNW,wci-rsm", 1024) = 36

It was not able to successfully refer to the database of /devices to /dev mappings (i.e. /dev/.devlink_db) to complete the resolution of the device to its /devices mapping.

Given the above observations, it is possible to conclude that the problem encountered stems from the fact that the /devices -- /dev mappings were not populated as expected:

     # strings .devlink_db | grep -i wci
/SUNW,wci@17d,0
SUNW,wci-rsm
wci17d
../devices/SUNW,wci@17d,0:SUNW,wci-rsm

Due to the fact that WCI Remote Shared Memory device driver (i.e. wrsm) had not been modload-ed to the kernel and resulting device node being placed into a "not attached" state:

     SUNW,wci (driver not attached)

Hence, the "showdevices" anomaly can be addressed by ensuring that the "wrsm" device driver is loaded on the domain OS bootup. To do this, run devfsadm with no options as root, or perform a reconfiguration boot (boot -r).

Internal Comments

see CR 4530555.

This solution has no attachment