Bug 113749

Summary: CCISS Driver for SCSI-raid not found
Product: [openSUSE] SUSE LINUX 10.0 Reporter: Hans Witvliet <hwit>
Component: KernelAssignee: Hannes Reinecke <hare>
Status: RESOLVED FIXED QA Contact: Klaus Kämpf <kkaempf>
Severity: Blocker    
Priority: P5 - None CC: kgw
Version: Beta 2   
Target Milestone: ---   
Hardware: 64bit   
OS: SUSE Other   
Whiteboard:
Found By: Beta-NTS Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: requested: /var/log/boot.msg
as requested: output from dmesg 10.2
output from yast
output of lcpci
detailed output from lspci

Description Hans Witvliet 2005-08-29 07:29:07 UTC
I presume hardware related are normally beyond the scope of the beta tests. 
However, in this case the hardware was recognized by elder releases and not 
recognized anymore (10.0 beta-1, beta-2, both 32/64) 
 
HARDWARE DESCRIPTION: 
System: HP-proliant (blade) DL360, 3Ghz Xeon-64, 2048MB mem 
CD/DVD on first primary IDE interface, hda 
Two 146GB ultra 320 SCSI disks connected via HP Smart Array 6i controller 
 
SYMPTOMS: 
After booting installationprograms (V1.9.8) starts up 
Routine questions are asked (language, agreement...) 
 
Finally you get an error message: 
++++++++++++++++++++++++++++++++++++++ 
+ No hardware disks were found for the installation + 
+ Please check your hardware                              + 
++++++++++++++++++++++++++++++++++++++ 
 
 
FURTHER ANALYSIS: 
9.3: As said, the system normally runs Prof-9.3 
9.3: (no special handling during installation required) 
9.3:  
9.3: The installer/yast had appended a kernel option for startup: 
9.3: "resume=/dev/cciss/c0d0p1 
9.3: lsmod showed the use of the module "cciss" 
9.3: Root partition was mounted on /dev/cciss/c0d0p2 
9.3: From messages i learned: 
9.3: HPCISS driver v2.6.4 
9.3: Device 0x46 found 
9.3: CCISS: Using DAC cycles 
 
With this information, i tried again (10.0/b2-64) 
10.0/b2: After booting from CD1, i provided same option: 
10.0/b2: "resume=/dev/cciss/c0d0p1" 
10.0/b2: 
10.0/b2 This yielded the same results 
10.0/b2 Manually trying: fdisk /dev/cciss/c0d0p2  
10.0/b2 returned: unable to open 
10.0/b2 Manually tried: "insmod ccis" without result 
 
Needles to say, before burning, md5's were checked. 
different sets of CDROMs were used, and test performed on three different 
blades. 
 
(Will re-test with 10.0-b3 asasp)
Comment 1 Olaf Kirch 2005-08-30 07:25:13 UTC
The resume option is added by the installation, and is needed for suspend/ 
resume to work properly. 
 
Please try to boot without this resume option, and append the boot log 
and dmesg to this report. 
Comment 2 Olaf Kirch 2005-08-30 07:26:07 UTC
*** Bug 113948 has been marked as a duplicate of this bug. ***
Comment 4 Klaus Wagner 2005-08-30 09:55:28 UTC
Suggestions for original reporter: your system might have problems
with PCI bus/HW ressources administration. The ACPI code which handles
this isn't sufficiently reliable in many BIOSes. To narrow this down:
  -  try booting the "Installation -- Safe settings" entry and see
     whether the SmartArray 6i now gets detected/the cciss module
     now loads successfully.

  -  It also might be (depending on your particular hardware) that
     a kernel parameter like "acpi=off" is needed.

  -  try booting SuSE Linux 10.0 beta3


There's a verified case here that *SL 10.0b3* properly loads cciss and
detects a SmartArray 6i, so it can't be just the driver being unable to
handle the hardware.
Comment 5 Hans Witvliet 2005-08-31 21:24:12 UTC
Created attachment 48378 [details]
requested: /var/log/boot.msg

after booting 10.2, no cmdline options
Comment 6 Hans Witvliet 2005-08-31 21:25:04 UTC
Created attachment 48379 [details]
as requested: output from dmesg 10.2
Comment 7 Hans Witvliet 2005-08-31 21:33:45 UTC
Created attachment 48381 [details]
output from yast

as suggested tried:
a) with no option during boot
b) using "safe mode
c) specifying acpi off

No show, cciss driver not loaded
(perhaps 64-bit specific?)
Comment 8 Olaf Kirch 2005-09-01 08:58:57 UTC
It appears as if your device has 
 
PCI Vendor/Device ID 0e11 0046 
Sub-Vendor/Device ID 0e11 3201 
 
0046 is PCI_DEVICE_ID_COMPAQ_CISSC. Looking at the device table in 
drivers/block/cciss.c, there are several entries for that device ID, but 
none of them has a matching sub-device ID. 
 
And indeed, the SA 6300 isn't listed in the supported hardware: 
 
MODULE_SUPPORTED_DEVICE("HP SA5i SA5i+ SA532 SA5300 SA5312 SA641 SA642 SA6400" 
                        " SA6i P600 P800 E400 E300"); 
 
Maybe we should ask HP about this 
Comment 9 Olaf Kirch 2005-09-01 09:02:48 UTC
BTW I extracted these PCI IDs from the yast logs - do they agree with 
what lspci reports? 
Comment 10 Klaus Wagner 2005-09-01 09:11:20 UTC
Side note for narrowing down purposes: the working SmartArray 6i
mentioned in comment #4 turns out to be a different device. It gets
reported by lspci as follows:

0000:04:03.0 RAID bus controller: Compaq Computer Corporation Smart Array 64xx
(rev 01)
        Subsystem: Compaq Computer Corporation: Unknown device 4091

0000:04:03.0 Class 0104: 0e11:0046 (rev 01)
        Subsystem: 0e11:4091

So, once more, one needs to be truly detailed w.r.t device specification:
"SmartArray 6i" isn't enough...
Comment 11 Klaus Kämpf 2005-09-01 11:09:14 UTC
comment #9: YaST gets its data from hwinfo which in turn uses the same source 
as lspci, the /proc filesystem 
Comment 12 Olaf Kirch 2005-09-01 11:19:15 UTC
I assumed as much. But I wanted to double check these values with the 
customer, because the numbers in the yast log are a) decimal, and b) seem to 
be offset by 65536, and c) buried in a 800-character line. So there's plenty 
of opportunity for me to get it wrong :-) 
Comment 13 Andreas Jaeger 2005-09-03 15:22:55 UTC
Does this still occur with Beta4?
Comment 14 Hans Witvliet 2005-09-05 06:58:23 UTC
Created attachment 48740 [details]
output of lcpci

As reuested, output of lspci.
As you see, does not give away much more information

Will try B4 today
(previous attempt failed, because off bad media)
Comment 15 Olaf Kirch 2005-09-05 07:35:19 UTC
Sorry, I was a bit unspecific in my request. To obtain the PCI IDs of  
your device, call "lspci -s 0000:02:01.0 -nv". 0000:02:01.0 is the bus 
ID of your RAID array, as shown by the lspci output above. 
Comment 16 Hans Witvliet 2005-09-06 19:34:01 UTC
Created attachment 48992 [details]
detailed output from lspci

Firstly, here the requested output.

Secondly, raid controller is now recognized by 10.0/B4
Any changes?
Investigate it further, or close it?
Comment 17 Klaus Wagner 2005-09-07 07:38:30 UTC
Just checked the "provides the needed information" box (assuming Hans
did simply forget) to state-change the bug. Hans, hare, okir: do you agree? 
Comment 18 Olaf Kirch 2005-09-07 08:03:23 UTC
I simply can't tell what may have caused the array to be recognized now. 
 
But I'm happy it's detected. Closing as fixed. 
 
Thanks, all! 
Comment 19 Hans Witvliet 2005-09-07 19:05:43 UTC
Yes, i'm happy that it's solved and closed.
Thanks for all the effort you all have put into it.

Kind regard, back to hunting..