Re: [Sisuite-users] Smart Array 6i cciss not loading

Fri, 09 Sep 2005 07:10:41 -0700

Peter/David,
 
I have not had an opportunity to try your Kernel yet, but hope to do so shortly.  Currently this test system resides on a stand alone network, so I haven't had a chance to burn the CD yet.
 
In any event some more information about the problem.
 
When the cciss driver is loaded it creates a structure under /dev/cciss//host0/target0/disc.  This is all well and good except for the fact that the auto install script looks for /dev/cciss/disc0/disc.
 
So by modifying the install script before the partitioning has begun and adding some lines:
 
mkdir /dev/cciss/disc0
cp -p /dev/cciss/host0/target0/disc /dev/cciss/disc0
mknod -m 600 /dev/cciss/disc0/part1 b 104 1
mknod -m 600 /dev/cciss/disc0/partX b 104 X
 
(In my case my major block value was 104, others may differ) and then adding a mknod line for each partition that would get created.
 
This has gotten me to the point where all the partitions are created, the image is transferred, and now I'm back to bombing on the good old Grub install, referencing the ever dreaded Grub.pm file! Mark Seber was able to provide an updated version that worked in the past, but in looking at some of his posts in July it would seem that the new grub included with the 3.5.1 edition doesn't take too kindly to his changes, but we'll see.
 
So the question remains, is there any way to get a cciss.ko driver that creates the proper /dev structure to begin with so modifying the autoinstall script becomes not necessary? (At least I assume that the cciss.ko driver is responsible for the /dev structure creation) It would really be nice to see /dev/cciss/c0d0pX called, which would also fix the grub issue, but that's probably asking for too much as this would mean changing the autoinstall script creation tool that would probably throw off a bunch of other stuff in the process.  Maybe a specific compaq smart array edition of System Imager??  It appears the cciss driver gets called as a module by the kernel, so it probably resides in the initrd.img(??) or does it somehow get loaded through the boel_binaries.tar.gz (which I'm still not clear how at file gets accessed)?  When I tried to cpio the initrd.img to take a look at its contents it threw a bunch of end of file errors letting me know how unhappy it was, so I've got no concept of what exactly is contained within the initrd.
 
Time to go play with grub.
 
Thanks,
 
Seth


 
On 9/8/05, Peter Mueller <[EMAIL PROTECTED]> wrote:
Does my 2.4 'SI-3.4.1' kernel work with SI-3.5.1 and RHEL4?
I expect it to have a 40% chance of actually working.
 
For your development question, I can answer that.
leads to ->
 

Developer Access

The Subversion server is now up and running: [WWW] http://svn.sisuite.org

If you would like to request developer access, send your public ssh key (one with a passphrase, of course) in a gpg encrypted or signed email to dannf_at_dannf.org or [MAILTO] [EMAIL PROTECTED].

Once your key is in place, you can perform checkouts like so:

  • Development trunk: 

svn checkout svn+ssh://[EMAIL PROTECTED]/var/lib/svn/systemimager/trunk/
 systemimager/
Regards,
 
P

From: [EMAIL PROTECTED] on behalf of Seth Haines
Sent: Thu 9/8/2005 12:31 PM
To: [email protected]
Subject: Re: [Sisuite-users] Smart Array 6i cciss not loading

 

David,

 

I manually ran a "modprobe cciss" from the BOEL command line in order to insure than an lsmod would show the cciss module loaded, it seemed to work in my case, resulting in the /dev/cciss/host0/target0/disc directory structure that was I referring to in my pervious mail. 

 

In this case the cciss module is contained with /lib/modules/.../kernel/drivers/devices/block/cciss.ko or something along those lines. It's extremely small as previously mentioned. 

 

What I've been playing around with is swaping out different kernels and initrd.img files from the pxe boot.  Haven't had much luck as most of them halt on boot in a kernel panic because none of the normal file system structure is there.

 

What we need is the ability to recompile the BOEL kernel with an up to date cciss driver, or figure out some what to recompile a redhat kernel with support for systemimager.

 

Under the /usr/share/systemimager/boot/i386/standard you'll find the boel tar.gz.  I've just now exploded it on a working RedHat box and started playing around with its contents.  This tar.gz gets called during the pxe boot process and produces the file system that you see when the shell out occurs during an error.

 

Since I'll be pxe booting, or CD booting, the size of the kernel/initrd.img doesn't matter to me one bit, so going with another kernel that will interact with systemimager may be the better of the two solutions above.

 

To rebuild the boel kernel we need the source, and I currently do not know where to obtain it, and neither does google at this point.  Maybe it's in the systemimager cvs repository and I just haven't looked hard enough yet. I'll let you know what I find.

 

Thanks,

 

Seth

 

On 9/8/05, [EMAIL PROTECTED] < [EMAIL PROTECTED]> wrote:


Thanks for the reply Seth ... I was beginning to think there was no one on the list.

The floppy/cd boots correctly and bombs exactly as you show below.
  - from the shell at that point a lsmod does not show cciss loaded.
  - /dev/cciss... does not exist.
  - I would attempt to load ccisss.o if I could find it - I presume it's in the initrd ?

 There was an item on May 2, 2005 Subject [Sisuite-users] Re: cciss disk problem DL360
 where Steve Grandinetti responded :

The device support for the SmartArray device wasn't loaded, either via
the loadable module, or via static support in your kernel. By default,
typically, the SmartArray support is via loadable module.  There is a
'discover' task that runs as part of the BOEL env. startup, that
identifies which h/w is present and loads the appropriate modules.  If
indeed your SmartArray support is dynamically loaded, it's possible that
the 'discover' utility didn't identify the SmartArray devices.

If you continue to have trouble in this area, you can statically build
in the support by modifying the ./patches/linux.i386.config file,
changing the following two config variables from

CONFIG_BLK_CPQ_DA=m
CONFIG_BLK_CPQ_CISS_DA=m
to

CONFIG_BLK_CPQ_DA=y
CONFIG_BLK_CPQ_CISS_DA=y

And rebuild the kernel.


- The real issue is to have the new array detected correctly and the module loaded.
   In the meantime how does one "rebuild the kernel" ?

Thanks


David K Livingstone
CN Signals and Communications
10229 127 Avenue floor 2
Edmonton, AB, T5E 0B9
Ph  : 780 472-3959 Fax : 780 472-3050
Email: [EMAIL PROTECTED]


Seth Haines <[EMAIL PROTECTED]>
Sent by: [EMAIL PROTECTED]

2005/09/08 11:16
Please respond to sisuite-users

       
        To:         [email protected]
        cc:        
        Subject:        Re: [Sisuite-users] Smart Array 6i cciss not loading




David,

Where does the imaging process bomb?  Is it when parted starts trying
to create the partitions and format them?

I'm running into the same problem with some DL380 G4's (RHEL4 Install).

Partitioning /dev/cciss/disc0/disc...
Old partition tabe for /dev/cciss/disc0/disc
Error: Could not stat device /dev/cciss/disc0/disc.  No such file or directory.
parted -s -- /dev/cciss/disc0/disc........
Error: Could not stat device /dev/cciss/disc0/disc.  No such file or directory.

I remember seeing some traffic back in mid July from Mark Seger (HP
SystemImager Guru) but I never really saw the outcome of the
discussion.  I'm starting to go through the archives now to see if
there is anything I can find.

Incidentally you should consider installing the HP Proliant Support
Pack 7.3, which replaces the tg3 nic card driver (that has potential
to drop packets under heavy load) with the broadcom bcm5700.

Thanks,

Seth

On 9/6/05, [EMAIL PROTECTED] < [EMAIL PROTECTED]> wrote:
>
> I curently run Systemimager 3.0.0 on numerous HP Proliant DL380 G3's
> witthout a problem. They
> use e1000 nic driver and cciss(Smart Array 5i) diriver when booting from
> floppy. I have loaded
> RH EL 3 with the latest updated.
>
> I recently acquired newer DL380 G4 servers(Smart Array 6i and tg3 nics) and
> I ran into numerous problems :
>
> 1. The 3.0.0 standard boot floppy does not recognize the Smart Array 6i and
> therefore no disks
>      are available.
> 2. The nics are not recognized.
>
> I have tried the following :
>
> 1. Using Systemimager 3.4.1
>     - After much effort(another story) I finally generated a boot floppy
> from the std boot pkg however
>       once again the Smart Array 6i(cciss) was not recognized.
>
> 2. Downloaded Ole Holm Nielsen's 3.5.2 rpm's(generate for RH 4.0 and posted
> to this list)  and
>     generated a bootable cd(no chance of floppy on this one:). Still no luck
> with with the
>     cciss driver not loaded. The nic's however are recognized.
>
> So the immediate question is what do I have to do to get the Smart Array 6i
> recognized ?
> - I have subsequently loaded the machine via RH cd's(RH 3 upadte 5) and the
> array is
>   working fine. Looked at the cciss code on RH 3 and the kernel source both
> cciss.c  v 2.4.52.RH2 and the latest
>   v2.4.54.RH all have the 6i updates.
>
> Other Questions :
>  - I currently use local.cfg on my floppy boots : Can I use local.cfg on a
> bootable cd and if so how ?
>  - Making a bootable floppy with 3.4.1 was a headace :
>     - the si_mkautoinstalldiskette would exit with zero space on  the
> floppy. No mention is made in the manual on this.
>       I tried :
>         1. As per the suggestion in si_mkautoinstalldiskette(ie vi the code)
> I attemted to use /dev/fd0u1680
>             - This worked for "si_mkautoinstalldiskette --floppy
> /dev/fd0u1680"  however on subsequent
>                boot the rc script which attempts to load local.cfg does so
> using /dev/fd0 and it fails.
>             - si_mkautoinstalldiskette -out-file ddimage   will also fail as
> it hardcodes to 1.44 MB and fails.
>         2. Took all the comments out of lcoal.cfg and syslinux.cfg and
> removed message.txt altogether. This
>              then just fit on fd0.
>
> Thanks
>
>
>
>
> David K Livingstone
> CN Signals and Communications
> 10229 127 Avenue floor 2
> Edmonton, AB, T5E 0B9
> Ph  : 780 472-3959 Fax : 780 472-3050
> Email: [EMAIL PROTECTED]
>
>


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
Sisuite-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sisuite-users

Reply via email to