irclog2html for ##pxe on 20070320

03:30.58*** join/##pxe treshome (i=hello@bas2-toronto12-1167861711.dsl.bell.ca)
05:16.20*** join/##pxe treshome (n=t@CABLE-72-53-79-186.cia.com)
09:17.51*** join/##pxe Norky (n=dan@milton.ocsl.co.uk)
10:04.15Norkypxe booting ia64 machines works
10:05.55Norkynothing particularly difficult about, except that I wanted the dhcp server to support ia32 and x86_64 machines without making a separate host entry for each machine
11:10.39Norkylet me know what more this needs, ScytheBlade1: http://dev.brantleyonline.com/wiki/index.php/Itanium
11:18.41Norkyhmm, I can't create a link from the Index http://dev.brantleyonline.com/wiki/index.php/PXE_Booting_Index, I gess I have to wait for you to do that
11:53.30NirkusNorky: does IA32 and AMD64 machines have a different client machine architecture identifier?
11:54.39Nirkusi thought only IA32 EFI and IA32 had different architecture identifiers
11:54.40NorkyI don't *think* so, but it's just a function of the firmware, so there's nothing stopping the PXE stack identifying itself with some helpful architecture data
11:54.59Norkyahh, ia32 EFI machines I have no clue about
11:55.38Nirkus11:05 < Norky> nothing particularly difficult about, except that I wanted the dhcp server to support ia32 and x86_64 machines without making a separate [..]
11:56.03Nirkusso they should use the same host entry, dont they?
11:56.25Norkyordinary ia32/x86_64 machines with traditional PC BIOSes
11:57.03Nirkusyes, they should bot set the IA32 architecture identifier.
11:57.17Norkymy dhcpd.conf makes /pexlinux.0 the default for all machines
11:57.37NirkusAFAIK there is no special x86_64 identifier assigned yet
11:57.39Norkyit's only those clietns that ID themselves as ia64 that shoudl get any thing different
11:58.00NorkyI'll try it in a minute with an ia32 machine
11:58.10Nirkushehe, ok.
11:58.13Nirkusill take a shower
11:58.46Norkyit'd be useful if they do, I can then stop the various ethernet switches and other embedded deices bitching about being given invalid config/firmware updates ;)
11:59.57NirkusNorky: you ever captured the IA64 EFI client PXE packets with wireshark or something similiar?
12:00.15Nirkusim interested in what DHCP options that machines set
12:01.05Norkyyeah, I've got a capture here
12:01.16Norkytcpdump format but wireshark will read it
12:01.21Norkywant a copy?
12:02.15Norkythere's actually another option just before the one I'm matching on that looks like it's a simpler way of identifying all ia64 machines
12:02.47NorkyI need to find some non-HP ia64 hardware for testing so I can work out waht, if any of this is peculiar to HP
12:05.46NorkyI have an idea that my config suggestion will only work for certain HP machines, never mind non-HP kit
12:22.33Nirkusre
12:22.41NirkusNorky: yeah, a copy would be nice
12:22.51Nirkusvangeuns@atis.uka.de
12:23.48Nirkusif you want to filter out some of the PXE machines, you almost certainly will have to use the MAC adrress or the UUID/GUID
12:24.04Norkywhat do you mean?
12:24.27Nirkuswhich of my statements do you mean? :)
12:24.43Norkyif you want to filter out some of the PXE machines, you almost certainly will have to use the MAC adrress or the UUID/GUID -- this
12:25.24Norkymy aim is to have *all* x86 machines boot pxelinux and go to a menu (which allows choice of OS installation, various diagnostics and rescue tools)
12:26.20Nirkusthen you just have to filter out the client machine architecture identifier
12:26.25Norkyand *all* ia64 machines should boot elilo, giving a text-based prompt to install various OSes
12:26.30Norkyexactly :)
12:26.46Nirkusit should be reported in the class identifier and in the client machine architecture DHCP option
12:27.02Norkyany time I ahve to specify a per-host entry for a machine I consider that I have failed :)
12:27.24Nirkusthat's possible with ISC DHCP AFAIR
12:27.33Nirkusbut you'll have to look up how
12:27.58Norkyahh, cool, my x86_64 machine (again HP) is sending: Vendor-Class Option 60, length 32: "PXEClient:Arch:00000:UNDI:002001"
12:29.18Nirkusyeah, 0 is the identifier of IA x86 PC
12:29.55Nirkusthis number should be sent in the DHCP option tag number 93, too
12:30.34Nirkusoption tag 93 is the client system architecture and it should always have a laength of 2 bytes
12:30.47Norkyaye, I had never bothered lookign for it before as the vast majority of nodes on this particular lans are x86 so I'd just assumed that as the default case
12:31.10Nirkus[93(1)|len(1)|arch identifier(2)]
12:31.21Nirkuslen should always have the value 2
12:32.00Norkyfor the purposes of initial pxe booting, x86_64 (both intel and amd variants) is functionally equivalent to 32bit x86, correct?
12:32.13Nirkusyap
12:32.41Norkythey're different once we get to teh stage of selecting a Linux kernel, but before that they can (must) be treated the same?
12:33.22Nirkusx86_64 is the same architecture as x86 but the processor has some additional features (64Bit, more commands, more registers) that you have to activate at boot time
12:33.40Nirkusexactly
12:34.29Nirkusand, of course you'll need 64Bit compiled libraries and binaries if you want to use the hardware as x64
12:34.57Nirkusbut you can always boot an IA32 system on any x86_64 hardware
12:35.23Norkythought thinking about it, I think I had some problems with the first gen of amd64 machines running pxelinux (fixed with the pxelinux.0 file shipped in x86_64 versions of RHEL) , though that's as likely to be jsut some 'new-bios' bug as the difference between ia32 and x86_64
12:36.21Nirkusthat may be
12:36.45Norkysorry, have only just sent that capture file
12:36.49Nirkustheoretically, IA32 systems and boot loaders should work w/o modification an any x86_64 hardware
12:36.59Nirkusnp
12:37.26Nirkusthanks, just received it :)
12:40.27NorkyI'll try more machines as I get my hands on them
12:43.23Nirkuscool :)
12:43.23Nirkuswith wireshark you can see the DHCP option tag 93 (client system identifier)
12:43.24Nirkuswhich is set to 2 for IA64 and 0 for IA32
12:43.24Nirkusso even IA64 EFI PXE implementations report that tag. nice :)
12:43.24NorkyI'm using IPCop as my machine router/infrastructure machine which ships ISC DHCP, however the web UI can't really cgive you the full range of control over the config file, and has a nasty habit of overwriting any changes made manually.
12:43.30Norkyyeah, I saw that once I'd opened the capture file in wireshark
12:50.22Norkyhmm, now if only vmware or qemu could emulate an EFI-based x86_64 machine...
12:58.09NirkusNorky: ftp://ftp.rfc-editor.org/in-notes/rfc4578.txt
12:58.27Nirkus0    Intel x86PC
12:58.36Nirkus2    EFI Itanium
12:58.51Nirkus6    EFI IA32
12:58.57Norkyahah
12:58.58Nirkus9    EFI x86-64
12:59.01Norkybrialliant, thank you :)
12:59.06Norkybrilliant*
12:59.07Nirkusnp
12:59.35NorkyI was looking in man 5 dhcp-options
13:00.12Nirkusyou should read the first 2 chapters of the Intel PXE spec v2.1 and that RFC i've posted
13:00.34Nirkusthan you'll have all the infoprmation about PXE, you need
13:00.45Nirkus=> pizza :(
13:00.47Nirkus:)
13:01.15Norkyyeah, but I wont have all the time I need ;)
13:36.14*** join/##pxe SirFunk (n=Sir@admin-147-222.potsdam.edu)
14:14.49Nirkushehe
14:30.20Norkygood pizza?
14:32.19Nirkusyap, very good one :)
14:33.20Nirkus<PROTECTED>
14:33.46Norkynormally, yes
14:34.11Norky172.16/16 represents a 'build' network, that we use to install/setup/test customer machines
14:34.47Nirkus<PROTECTED>
14:35.27Norkyit's that big because we coudl potentially have several racks full of 1u servers or blades each with 3 or 4 NICs each needing an address
14:35.32Nirkus<PROTECTED>
14:35.34NorkyI'm not worried if this gets logged
14:35.44Nirkusah ok
14:36.29Nirkusnormally it should be no security issue, but i usually dont let any configuration information from work be logged anywhere public ;)
14:36.57Nirkus~logs
14:37.03jbotlogs is probably apt/ibot/infobot/jbot/purl all log daily to http://ibot.rikers.org/<channelname>/ where channelname is html encoded ie: %23debian | lines that start with a space are not shown | some channels have stats at http://ibot.rikers.org/stats/<channelname>.html.gz
14:39.42Norkya 42U rack can support 6 blade enclosures (if you include the PSUs) each containing up to 16 blades with potentially 4 or 5 nics (including mgmt interface), 5*16*6=480 ip addresses per rack
14:40.13Norkywe could have a fair number such racks running at any one time
14:40.24Norkythough it's unlikely in practice :)
14:40.34Nirkus:)
15:30.17ScytheBlade1Norky, I'm behind a retarded router right now.. and can't access my own wiki as a result. I'll take a look later on today when I can :)
15:30.34Norkyno worries
15:31.20NorkyI don't consider it done myself, I will likely need to change it when and if  I get the chance to test with other EFI hardware
15:32.50ScytheBlade1That's the point of a wiki
15:32.54ScytheBlade1It's never done :)
15:58.44*** join/##pxe Norky (n=dan@milton.ocsl.co.uk) [NETSPLIT VICTIM]
17:03.39NirkusNorky: why do you expect other EFI hardware to behave differently?
17:04.05Norkycynicism/experience ;)
17:04.21Nirkusthe vendor class identifier of PXE clients has to include the client system architecture and UNDI version
17:05.10Nirkuss/include/contain/
17:05.28Nirkusuh, thx jbot :)
17:05.41NorkyI *think* that older versions of the EFI spec did not include PXE, so there could be olders machines out there that have some crappy vendor-implemented network booting stack, if any at all
17:06.04NorkyI'm just thinking of the worst possible thinks that coudl go wrong :)
17:06.25Norkys/oudl/ould/
17:06.33Norkylovely
17:06.40Nirkusthe only other network booting stack would be BOOTP which does not have any PXE APIs/ABIs, so you wouldnt be able to boot fromt hat ;)
17:07.24Norkynow if only the bot could correct my clumsy-fingered typing *before* it hits the channel ;)
17:07.34Nirkushihi
17:07.43Nirkusjbot: you're too slow
17:07.47Nirkus:>
17:08.39Nirkusand since PXE was specified before EFI (AFAIK), I guess it was in there from the start
17:09.21NirkusNorky: all new intel macs have EFI on x86-64
17:09.58Nirkusi wonder whether they report EFI x86 or EFI x86-64 when booting via PXE
17:10.11Nirkusif they are able to boot PXE :)
17:38.47NorkyI will try to steal my brother's MacMini :)
17:51.10Nirkusdoes that thing have a core duo?
17:51.29Nirkusmaybe it just has a x86 CPU inside :)
18:03.21*** join/##pxe jbot (i=ibot@pdpc/supporter/active/TimRiker/bot/apt)
18:03.21*** topic/##pxe is PXE HOWTO at: http://dev.brantleyonline.com/wiki/index.php/PXE_Booting_Index | Also see http://en.wikipedia.org/wiki/Preboot_Execution_Environment (also at de.wiki) | Need help? Stick around (for about 8 hours, we're probably sleeping or at school)! | Want logs? Try typing ~logs
20:50.53*** join/##pxe jbot (i=ibot@pdpc/supporter/active/TimRiker/bot/apt)
20:50.53*** topic/##pxe is PXE HOWTO at: http://dev.brantleyonline.com/wiki/index.php/PXE_Booting_Index | Also see http://en.wikipedia.org/wiki/Preboot_Execution_Environment (also at de.wiki) | Need help? Stick around (for about 8 hours, we're probably sleeping or at school)! | Want logs? Try typing ~logs
21:10.00*** join/##pxe jbot (i=ibot@pdpc/supporter/active/TimRiker/bot/apt)
21:10.00*** topic/##pxe is PXE HOWTO at: http://dev.brantleyonline.com/wiki/index.php/PXE_Booting_Index | Also see http://en.wikipedia.org/wiki/Preboot_Execution_Environment (also at de.wiki) | Need help? Stick around (for about 8 hours, we're probably sleeping or at school)! | Want logs? Try typing ~logs
22:35.12*** join/##pxe jbot (i=ibot@pdpc/supporter/active/TimRiker/bot/apt)
22:35.12*** topic/##pxe is PXE HOWTO at: http://dev.brantleyonline.com/wiki/index.php/PXE_Booting_Index | Also see http://en.wikipedia.org/wiki/Preboot_Execution_Environment (also at de.wiki) | Need help? Stick around (for about 8 hours, we're probably sleeping or at school)! | Want logs? Try typing ~logs
22:54.42*** join/##pxe SirFunk (n=Sir@cpe-74-71-205-222.twcny.res.rr.com)

Generated by irclog2html.pl by Jeff Waugh - find it at freshmeat.net! Modified by Tim Riker to work with blootbot logs, split per channel, etc.