| SRDB ID | Synopsis | Date | ||
| 48145 | Sun Fire[TM] 15K: I/O adapter is NOT detected by HPOST | 29 Oct 2002 |
| Status | Issued |
| Description |
- Problem Statement:
setkeyswitch: 3V PCI I/O adapter is NOT detected by HPOST final configuration,
but is detected and probed by OBP.
- Symptoms:
HWAD daemon detects presence of the hpci I/O cassettes in slot IO4 and IO5
as reported by the SMS command showboards:
showboards -v
...
Location Pwr Type of Board ... Domain
------ --- ----------- ------
IO4/C3V0 Off Unknown - - B
IO4/C5V0 Off Unknown - - B
IO4/C3V1 Off Unknown - - B
IO4/C5V1 Off Unknown - - B
IO5/C3V0 Off Unknown - - B
IO5/C5V0 Off Unknown - - B
IO5/C3V1 Off Unknown - - B
IO5/C5V1 Off Unknown - - B
After installation of PCI I/O adapters into IO4 and IO5 hpci C3V0 I/O cassettes,
and after executing 'setkeyswitch -dB on', hpost may detect a card as NOT present:
I/O_Brds: IOC P1/Bus/Adapt IOC P0/Bus/Adapt
Slot Gen Type P1 B1/10 B0/10 P0 B1/eb10 B0/10 (e=ENet, b=BBC)
IO04: P hsPCI P p _m p _m P p PP_p p _p
IO05: P hsPCI P p _m p _m P p PP_p p _m <===
Configured in 333 with 4 procs, 8.000 GBytes, 3 IO adapters.
Where IO4-c3v0 , IO4-c5v0 , IO5-c5v0 are reported as present and IO5-c3v0 is missing.
showboards -v detects hpci I/O cassette presence and power on:
showboards -v :
Location Pwr Type of Board ... Domain
------ --- ----------- ------
IO4/C3V0 On C3V - B
IO4/C5V0 On C5V - B
IO4/C3V1 On C3V - B
IO4/C5V1 On C5V - B
IO5/C3V0 On C3V - B
IO5/C5V0 On C5V - B
IO5/C3V1 On C3V - B
IO5/C5V1 On C5V - B
After installation of OBP probe recognizes all 4 pci cards (from the
domain console log):
/pci@bc,600000/ ==> IO5-c3v0
/pci@bc,700000/ ==> IO5-c5v0
/pci@9c,600000/ ==> IO4-c3v0
/pci@9c,700000/ ==> IO4-c5v0
Mar 6 14:22:23 2002 Probing PCI B pci
Mar 6 14:22:24 2002 Probing /pci@9c,700000 Device 1
Mar 6 14:22:24 2002 Probing /pci@9c,700000/pci@1 Device 0 pci108e,1000 SUNW,hme
Mar 6 14:22:26 2002 Probing /pci@9c,700000/pci@1 Device 4 SUNW,isptwo sd st
Mar 6 14:22:29 2002 Probing /pci@9c,700000 Device 2 bootbus-controller iosram
Mar 6 14:22:31 2002 Probing /pci@9c,700000 Device 3 pci108e,1100 network firewire usb
Mar 6 14:22:31 2002 Probing PCI A pci
Mar 6 14:22:32 2002 Probing /pci@9c,600000 Device 1 scsi scsi
Mar 6 14:22:32 2002 Probing PCI B pci
Mar 6 14:22:33 2002 Probing /pci@9d,700000 Device 1 Nothing there
Mar 6 14:22:38 2002 Probing PCI A pci
Mar 6 14:22:38 2002 Probing /pci@9d,600000 Device 1 Nothing there
Mar 6 14:22:39 2002 Probing PCI B pci
Mar 6 14:22:39 2002 Probing /pci@bc,700000 Device 1 pci
Mar 6 14:22:40 2002 Probing /pci@bc,700000/pci@1 Device 0 pci108e,1000 SUNW,hme
Mar 6 14:22:41 2002 Probing /pci@bc,700000/pci@1 Device 4 SUNW,isptwo sd st
Mar 6 14:22:44 2002 Probing /pci@bc,700000 Device 2 bootbus-controller iosram ^
Mar 6 14:22:45 2002 Probing /pci@bc,700000 Device 3 pci108e,1100 network firewire usb
Mar 6 14:22:45 2002 Probing PCI A pci
Mar 6 14:22:46 2002 Probing /pci@bc,600000 Device 1 scsi scsi
Mar 6 14:22:46 2002 Probing PCI B pci
Mar 6 14:22:47 4 2002 Probing /pci@bd,600000 Device 1 Nothing there
Mar 6 14:22:48 4 2002 Probing /pci@bd,700000 Device 1 Nothing there
After the domain boots, prtdiag detects the PCI adapters:
IO Port Bus Freq Bus Dev,
Slot ID Type ID Side MHz Freq Func State Name Model
---------- ---- ---- ---- ---- ---- ---- ----- -------------------------------- ----------------------
/IO04/C5V0 PCI 156 B 33 33 0,1 ok SUNW,hme-pci108e,1001 SUNW,cheerio
/IO04/C5V0 PCI 156 B 33 33 4,0 ok SUNW,isptwo-pci1077,1020/sd (blo+ QLGC,ISP1040B
/IO04/C3V0 PCI 156 A 66 66 1,0 ok scsi-pci1077,2200.1077.2.5
/IO05/C5V0 PCI 188 B 33 33 0,1 ok SUNW,hme-pci108e,1001 SUNW,cheerio
/IO05/C5V0 PCI 188 B 33 33 4,0 ok SUNW,isptwo-pci1077,1020/sd (blo+ QLGC,ISP1040B
/IO05/C3V0 PCI 188 A 66 66 1,0 ok scsi-pci1077,2200.1077.2.5
SOLUTION SUMMARY:
- Troubleshooting:
A few questions to help understand the failure:
. Were the PCI adapters that resided in the cassettes used in other locations?
. In the case where POST failed to see adapters present, were the hpci I/O cassettes
that the PCI adapters resided in used before?
. Was there any attempt to re-seat the adapters into the hpci I/O cassettes and
the cassettes into the hsPCI ?
Check the connector of the PCI I/O adapter card for damage. If damaged, then replace
PCI adapter card and verify the hpci I/O cassette connector.
Check the hpci I/O cassette connector for damage or foreign debris within the connector.
If damaged, replace the hpci I/O cassette and verify the PCI adapter card connector.
Check by swapping the offending hpci cassette:
. If the PCI card is detected by HPOST, then service is complete.
. If the PCI card is still undetected, then replace the PCI I/O adapter card.
- Resolution:
The plastic connector, J2, which is the connector the PCI card plugs into, may have
been damaged when the NIC card was installed. The PCI card scrapes along the plastic
guide slot in J2 and "mashes" plastic onto the adjacent pin, B11, which is signal name
"EPCI_PRSNT2_L". The plastic was cleaned out and the problem was fixed, the same pin and
problem on all three cassettes. The cassettes have such a tight fit when trying to
install PCI cards that this guide slot gets a lot of pressure on it, and the plastic takes
a damage hit. The guide slot will have a groove in it from the pressure of installing the
PCI card.
HPOST did not detect the 3V PCI I/O adapter card due to a hardware failure on the hpci
I/O cassette or at the PCI adapter connector. Replace the hpci I/O cassette.
The latest 3V & 5V hpci I/O cassette assemblies (-07 rev50) greatly improved the PCI card
installation and accessibility.
- Summary of part number and patch ID's
FRUs: 3V hpci I/O cassette 501-5599-06
33/66mhz PCI I/O adapters
- References and bug IDs
4656169 : Inconsistency in results from hpost, obp and Solaris regarding IO
- Additional background information:
Problems have been observed inserting PCI I/O adapters into the hpci I/O cassettes
which cause damage to occur on the internal hpci I/O cassette connector.
By design, OBP probes PCI slots which are marked "present" but are missing the IO cards.
This is to allow for the possibility of hotplugging a card later since OBP must allocate
the PCI resource at this stage. OBP uses the PCI interface as the physical path to probe
for devices . However, POST adapter presence is determined by reading status registers within
the hot plug controller registers via I2C communication . This explains an inconsistency
between the PCI devices seen by hpost and OBP.
The PCI IO adapter installed in an hPCI cassette is supposed to ground either one or both
of the PRSNT[1:0]# interface signal pins. These 2 PCI signals interface to the HPC3130 hot
swap controller where their state can be read. If either pin is read as a zero then POST
would detect that an adapter is present. If both pins are a one, then no PCI adapter
is present. If an adapter card does not have one or both of the PRSNT[1:0]# signals
connected to ground, it is not PCI-compliant.
Meta-Data/Problem categorization:
Product/Platform: SF15K
Category:
- Keywords
3.3v hPCI cassette hpost OBP
INTERNAL SUMMARY:
SUBMITTER: Gino Valencia BUG REPORT ID: 4656169 APPLIES TO: Hardware/Sun Fire /15000 ATTACHMENTS: