topic Re: Support Pack for Proliant Firmware Update over PXE in BladeSystem - General

Support Pack for Proliant Firmware Update over PXE

Gerardo_Arceri — Fri, 28 Oct 2011 15:29:18 GMT

Anyone knows if the SPP iso image can be booted over PXE just like the old Firmware DVD?

Re: Support Pack for Proliant Firmware Update over PXE

Gerardo_Arceri — Mon, 31 Oct 2011 15:40:56 GMT

The following PXE boot menu entry did the trick

label vsos
MENU LABEL Interactive Firmware Update Version 2011.09
kernel vmlinuz.FW11
append initrd=initrd.img.FW11 media=cdrom rw root=/dev/ram0 ramdisk_size=257144 init=/bin/init loglevel=3 ide=nodma ide=noraid pnpbios=off vga=791 splash=silent showopts TYPE=MANUAL iso1=nfs://kickstart.nfs/osbuild/extra/install_packages/hp_firmware/SPP2011090.2011_0930.13.iso iso1mnt=/mnt/bootdevice pci=bfsort
ipappend 2

We had to patch the initrd.img to allow the DVD to be booted from interfacees other than eth0.

Let me know if you are interested in the patch.

Re: Support Pack for Proliant Firmware Update over PXE

chuckk281 — Thu, 17 Nov 2011 16:51:24 GMT

Yes it can.

Re: Support Pack for Proliant Firmware Update over PXE

Voodoo66 — Thu, 29 Dec 2011 07:07:06 GMT

Hi, I am interested in this patch. I am trying to set up PXE boot for firmware deployments. It would be nice if you can post it here.

Thankx and happy new year.

Re: Support Pack for Proliant Firmware Update over PXE

Peter Waller_1 — Tue, 17 Jan 2012 15:25:22 GMT

Hello, I'm very interested in this patch.

I'm going to integrate the service pack into our boot menu, but it does not work. Can you give me this patch?

Thank You and regards from Switzerland

Re: Support Pack for Proliant Firmware Update over PXE

Michael Pye — Thu, 19 Jan 2012 02:53:26 GMT

The main Support Pack for Proliant page states "Simplified delivery of HP ProLiant systems software and firmware into PXE bootable ISO images to reduce customer Qualification cycles Resource usage Maintenance windows Downtime".

No where that I can find is it detailed about how you pxeboot SPP. The release notes state:

ftp://ftp.hp.com/pub/softlib2/software1/doc/p1822529277/v72723/SPP2011.09.0ReleaseNotes.pdf

there are only 2 methods to depoly, mount and boot iso image offline, or mout iso and run SUM online.

If it is pxe bootable it's not offcially documnted like the 9.30 fw dvd was.

Re: Support Pack for Proliant Firmware Update over PXE

Chris-Smith — Tue, 06 Mar 2012 10:27:11 GMT

Any chance I can get a copy of the modified initrd.img also :)

Re: Support Pack for Proliant Firmware Update over PXE

sreitemeyer — Fri, 23 Mar 2012 11:30:33 GMT

Hi,

we have the same problem. The service pack is only using eth0.

It would be nice if you can send us the patch.

Re: Support Pack for Propellant Firmware Update over PXE

shocko — Fri, 30 Mar 2012 13:39:41 GMT

I have the same issue. I can't find documentation for the PXE boot params. This is so infuriating.

Re: Support Pack for Propellant Firmware Update over PXE

odenbach — Fri, 27 Apr 2012 15:43:56 GMT

Hi,

we had the same problem with SPP just using eth0 and had a deep dive into the SPP ISO. Let me explain the details:

There is already a function that loops over every network adapter and tries to get an address via DHCP. The problem is that if the dhcpcd client does not get an address via DHCP it makes a fallback to APIPA, so it gets an address like 169.254.x.x and then returns OK. The looping function then thinks that DHCP worked and skips the other interfaces.

The simple trick is now to disable APIPA in dhcpcd. This is done with the '-L' switch.

So just do the following:

- Extract the original initrd:

# mkdir /tmp/initrd

# cd /tmp/initrd

# gunzip -c /tftp/hp/initrd.img | cpio -i

- Extract the root filesystem:

# mkdir rootfs

# cd rootfs

# busybox unlzma -c ../fs.tar.lzma | tar xvf -

- Apply the small patch (the -L switch to dhcpcd) to root/compaq/netconfig.sh:

root@kitty[compaq]# diff -c netconfig.sh /tmp/initrd/root/compaq/netconfig.sh
*** netconfig.sh   Fri Mar 2 08:36:47 2012
--- /tmp/initrd/root/compaq/netconfig.sh   Fri Apr 27 17:06:29 2012
***************
*** 40,46 ****
          fi
          if [ $PROTOCOL == "dhcp" ]; then
                  echo Attempting to configure $1 using DHCP
!                 dhcpcd -t 45 $1
                  TEMP=$?
                  echo return from dhcpcd $TEMP
                  if [ $TEMP != 0 ]; then
--- 40,46 ----
          fi
          if [ $PROTOCOL == "dhcp" ]; then
                  echo Attempting to configure $1 using DHCP
!                 dhcpcd -L -t 45 $1
                  TEMP=$?
                  echo return from dhcpcd $TEMP
                  if [ $TEMP != 0 ]; then

(if you like you can decrease the timeout of 45 seconds to a lower value to make the boot process faster)

- repack the rootfs and remove the unpacked data:

# tar cf - . | lzma -c > ../fs.tar.lzma

# cd ..

# rm -rf rootfs

- repack the initrd:

# find . | cpio --create --'format=newc' | gzip > /tftp/hp/initrd.img

So this helps servers boot from _any_ network device that has DHCP configured.

Christopher

Re: Support Pack for Propellant Firmware Update over PXE

shocko — Tue, 01 May 2012 08:25:38 GMT

Excellent work oedenbach! very useful info.

Re: Support Pack for Proliant Firmware Update over PXE

Michael Leu — Thu, 07 Jun 2012 15:57:19 GMT

Excellent post odenbach, thanks!

It seems with the newest SPP 2012.06.0(B) they switched a lot around inside the initrd. So here is a updated version of patching netconfig.sh with the new location.

It includes another small change for booting with a be2net (BL460c G7) and bnx2 (DL385 G2) based network card.

ISO=<replace with full path to the SPP.iso>
INITRD=<replace with full path for the new initrd>

# unpack
mkdir /tmp/spp /tmp/initrd-spp
mount -o loop $ISO /tmp/spp
cd /tmp/initrd-spp && gzip -cd /tmp/spp/system/initrd.img | cpio -id
# disable APIPA in dhcpcd (by odenbach)
sed -i 's/dhcpcd -t 45 $1/dhcpcd -L -t 45 $1/' sbin/netconfig.sh
# add be2net and bnx2 drivers
sed -i 's/NET_MODULES="e100 e1000 igb qlcnic tg3 bnx2x"/NET_MODULES="e100 e1000 igb qlcnic tg3 bnx2x be2net bnx2"/' etc/initrd.functions
# fix lazy greps
sed -i 's/grep $mod/grep ^${mod}$/' etc/initrd.functions
# repack and cleanup
find . | cpio --quiet -o -H newc | gzip -9 > $INITRD
cd - && rm -Rf /tmp/initrd-spp
umount /tmp/spp && rmdir /tmp/spp

Additionally the PXE boot menu entry also needs to be changed from

media=cdrom

media=net

So a complete entry might look something like this:

label spp
  kernel vmlinuz-spp
  append initrd=initrd.img-spp media=net rw root=/dev/ram0 ramdisk_size=815744 init=/bin/init loglevel=3 ide=nodma ide=noraid nopat pnpbios=off vga=791 splash=silent hp_fibre showopts noexec32=off numa=off nox2apic TYPE=AUTOMATIC iso1=nfs://[IP]/[PATH]/SPP2012060B.2012_0525.1.iso iso1mnt=/mnt/bootdevice

By the way, adding the parameter "d3bug" in combination with "console=ttyS0,115200" and the iLO VSP lets you see some debugging infomation :)

Hope this helps...

Re: Support Pack for Proliant Firmware Update over PXE

Peter Capriccio — Thu, 14 Jun 2012 02:09:14 GMT

Just wondering if anyone is seeing this, or is it just me...

I've switched to using the "Madis" method (not mounting the ISO, but unpacking the contents and NFS serving out the directory -- this way, you can easily add updates) and for this release I've run into a problem where it doesn't create the link from "/mnt/bootdevice" to "/mnt/cdrom" -- from the "supermount_isos" function in "etc/initrd.functions":

.
.
+ mkdir -p /mnt/bootdevice
+ mount_url.sh --dir /mnt/bootdevice nfs://192.168.1.254/SPP-2012.06.0B/ 
+ [ /mnt/bootdevice == /mnt/bootdevice ]
+ rm -f /mnt/boot
+ rm -f /mnt/cdrom
+ ln -s
BusyBox v1.14.1 (2012-04-04 02:06:19 UTC) multi-call binary

Usage: ln [OPTION] TARGET... LINK_NAME|DIRECTORY

Create a link named LINK_NAME or DIRECTORY to the specified TARGET.
Use '--' to indicate that all following arguments are non-options.

Options:
        -s      Make symlinks instead of hardlinks
        -f      Remove existing destination files
        -n      Don't dereference symlinks - treat like normal file
        -b      Make a backup of the target (if exists) before link operation
        -S suf  Use suffix instead of ~ when making backup files

+ ln -s /mnt/cdrom
ln: cdrom/cdrom: File exists
+ eval iso=$iso2
+ iso=
+ eval isoopts=$iso2opts
+ isoopts=
+ eval isomnt=$iso2mnt
+ isomnt=
+ [ -n  ]
+ break
.
.

The side effect of this is that when the browser starts up, you can progress through the language selection and EULA, but when you click on "Launch HPSUM", it just dumps you back to the reboot screen in the browser. I can launch hpsum from a debug window (Ctrl+Alt+d+b+x), but this doesn't go over too well with the point-and-click crowd :)...

My workaround was to edit the "langsel.htm" file (after the squashfs is loaded) and replace the "/mnt/cdrom" references with "/mnt/bootdevice" in the "runLin" function instead of trying to fix the link after the squashfs replaces the initrd. This did the trick for me, but in the end I wound up replacing the entire "langsel.htm" to bypass the language selection and EULA. I added the following changes to the "init" script:

.
.
#Mount the squash filesystem
mount_fs

# "fix" xorg.conf (change "*Depth" from 24 back to 16 for older systems)
sed -i '/Depth/s/24/16/g' "${CHROOT_DIR}"/etc/X11/xorg.default-us

# "fix" references to "/mnt/cdrom" and remove EULA, etc, in the langsel.htm file in the squash
# sed -i 's|/mnt/cdrom|/mnt/bootdevice|g' "${CHROOT_DIR}"/opt/hp/hp-phoenix/srv/www/langsel.htm
cp /langsel.htm.wbu "${CHROOT_DIR}"/opt/hp/hp-phoenix/srv/www/langsel.htm
.
.

Re: Support Pack for Proliant Firmware Update over PXE

Christian Otto Stelter — Thu, 14 Jun 2012 09:50:20 GMT

@Peter Capriccio wrote:
Just wondering if anyone is seeing this, or is it just me...

/me too. At least the full automatic mode runs without problems after patching netconfig.sh - Thx to Michael Leu!

Re: Support Pack for Proliant Firmware Update over PXE

Gerardo_Arceri — Sat, 28 Jul 2012 12:35:56 GMT

I've made a zip with the patches on netconfig.sh and initrd.functions along with simple instructions.

On initrd.functions, the missing /mnt/cdrom is caused by a typo on line 386:

rm -f /mnt/cdrom >/dev/null 2>&1

You can't remove directories with rm (Hello QA TEAM ?????????), the solution is to:

rm -rf /mnt/cdrom >/dev/null 2>&1

I've also added the extra be2net and bnx modules, otherwise this will not work on G5s or G7+ blades.

on netconfig.sh the patch revolves around using the "ipappend 2" feature of pxelinux, which will append a BOOTIF=XX:XX:XX:XX:XX:XX parameter with the boot interface mac address, that can be later looked up with ifconfig before configuring the interface, take a look at netconfig.sh.patch for more info

This patch has been made by a coworker and adapted to work on the new SPP, we have been patching all FW/SPP DVD releases for quite a long time..

next and last, this is how my pxemenu entry looks like

label vsos1206
  MENU LABEL Interactive Firmware Update Version 2012.06
    kernel vmlinuz.SPP2012.06
    append initrd=initrd.img.SPP2012.06 media=net rw root=/dev/ram0 ramdisk_size=900000 init=/bin/init loglevel=3 ide=nodma ide=noraid nopat pnpbios=off vga=
791 splash=silent hp_fibre showopts noexec32=off numa=off nox2apic TYPE=MANUAL iso1=nfs://[your nfs ip goes here]/SPP2012.06.iso iso1mnt=/mnt/bootdevice
    ipappend 2

Re: Support Pack for Proliant Firmware Update over PXE

dcolpitts — Sat, 23 Jun 2012 14:07:10 GMT

So I've utilized Gerado's patch he posted, but I keep getting permission denied when mounting my iso image via SMB. This is happening on a ProLiant BL460c G7.

Any thoughts? The username and password are correct. It appears to be some sort of permissions issue in /mnt folder.

Does anyone know the exact mount command that is used with BusyBox? My command line includes:

iso1mnt=/mnt/bootdevice iso1=smb://myuserid:mypassword@10.131.200.13/hpssfw/SPP2012060B.2012_0525.1.iso

So it might help in my troubleshooting if I could manually mount the iso using mount and mount_url.sh.

dcc

Re: Support Pack for Proliant Firmware Update over PXE

dcolpitts — Sat, 23 Jun 2012 20:58:55 GMT

The issue I'm experiencing appears to be related to using the smb or cifs protocols. If I setup a NFS server, the iso image gets mount successfully. When I check the event log on my Windows server that the smb share is pointing at, I have no workgroup/domain listed, and username associated with the failed login event always has a : (colon) behind it.

dcc

Re: Support Pack for Proliant Firmware Update over PXE

Gerardo_Arceri — Mon, 25 Jun 2012 17:48:35 GMT

I extracted mount_url.sh from /sbin inside the initrd.

Tried to manually run with sh -x passing a fake url smb://user:pass@127.0.0.1/someshare/some.iso or cifs://user:password@127.0.0.1/someshare/file.iso and the script seems to be correctly parsing the URL, the mount command as executed by the script is:
+ mount -t cifs -o username=user,password=password //127.0.0.1/share /mnt/127.0.0.1/share

Re: Support Pack for Proliant Firmware Update over PXE

dcolpitts — Mon, 25 Jun 2012 18:27:31 GMT

I haven't had time to try this yet, but I'm wondering if perhaps I used the iso1options setting if that would help... So something like this:

iso1mnt=/mnt/bootdevice iso1=smb://10.131.200.13/hpssfw/SPP2012060B.2012_0525.1.iso iso1options="username=myuserid,workgroup=domain,password=mypassword"

dcc

Re: Support Pack for Proliant Firmware Update over PXE

Gerardo_Arceri — Tue, 26 Jun 2012 20:26:05 GMT

I updated the patches attached to the previous post.

Changelog:

-on initrd.functions the bnx2 driver was not being loaded

-Also on initrd.functions there was a bug on a grep line that prevented the same driver from loading (lsmod |cut -d' ' -f1 | grep $mod), this line choked if both bnx2 and bnx2x drivers were on the module load list , line replaced with (lsmod |cut -d' ' -f1 | grep ${mod}\$) to fix the problem.

As a consequence of these two bugs it was impossible to PXE boot if your servers used bnx2 interfaces to boot (like on bl680c g5)