[syslinux] PXELinux and serial number

[Scott Mewett]

On Fri, 2005-03-18 at 22:30 -0600, Michael_E_Brown at Dell.com wrote:
> I built an (unreleased, non-oss) identical thing to what you describe
> below for internal use here a while back. 
> 

Too bad.

> It isn't that hard. I would say that you don't even need PXElinux
> support to get the serial number. What my system did is boot the system
> into a stripped-down Linux that grabbed the system serial number and
> applicable hardware config, then registered itself with a master server.
> At this point, the master server knows the MAC address and can set up
> PXElinux config files based on MAC address.

Yeah, That's way the current design is spec'd out.
Since it's really only ever done once, it's not a big deal. But were
trying to make some servers FRU's so if the hardware gets swapped and
has tons of memory, then the we want as few reboot's as possible to get
the box backup.

Thanks
Scott


> --
> Michael
> 
> > -----Original Message-----
> > From: syslinux-bounces at zytor.com 
> > [mailto:syslinux-bounces at zytor.com] On Behalf Of Scott Mewett
> > Sent: Friday, March 18, 2005 10:18 PM
> > To: H. Peter Anvin
> > Cc: syslinux at zytor.com
> > Subject: Re: [syslinux] PXELinux and serial number
> > 
> > 
> > Very cool.
> > 
> > Were in the design stage of our project right now. I should 
> > know in the next few weeks if it gets approved to build out a 
> > working concept. If that happens then I'll try and get some 
> > hardware or $$ donation for your effort.
> > 
> > The general idea of the project is to build a managed pxe 
> > system. External applications would register with a database 
> > backed system to setup boot choices for a particular serial 
> > number. The backend system would control the creation of the 
> > tftp config files for pxelinux. For example a sysadmin would 
> > go through a web page and create a profile for a server to be 
> > kickstarted. The backend would create an appropriate pxelinux 
> > config file for that serial number or mac address. The the 
> > server would pxe boot using the config for that server.
> > 
> > For that scenario you don't need a back end db, but we want 
> > to expand it way beyond that, so that you can boot systems 
> > for example without a config, go to a web page to show 
> > unprovisioned servers and then create your config from there. 
> > Once you finish your config on the web page the system would 
> > take care of rebooting the box automatically so that the 
> > install could start. We are also thinking of having systems 
> > always pxe boot, even though the system may tell it to boot 
> > of the hard disk. There may also be systems that will boot 
> > and run over the network without using an OS on a hard disk. 
> > We want all that controlled by the db backend. All of the 
> > config files in the tftpboot dir could be recreated at will 
> > using the data in the db.
> > 
> > There are commercial systems that do some of this. But none 
> > that really do everything we want or are too closed a system 
> > to expand and do additional things.
> > 
> > Is anyone aware of any open source systems that already do 
> > some of this? Sorry if this is a little of topic. But since I 
> > want to base it on pxelinux, I figured this would be a good 
> > audience to bring it up with.
> > 
> > It is my intention, if I get support from my employer, to 
> > open source this system as well. I've been successful with 
> > that on a couple other project I've done.
> > 
> > Needless to say, pxelinux rocks. :-)
> > 
> > Thanks
> > Scott
> > 
> > 
> > On Fri, 2005-03-18 at 18:03 -0800, H. Peter Anvin wrote:
> > > Scott Mewett wrote:
> > > > 
> > > > Since there is only the serial number that we are looking 
> > for, being 
> > > > able to read all the table structures would not be necessary.
> > > > 
> > > > What do you think?
> > > > 
> > > 
> > > Doesn't look too bad.  Find the _DMI_ header in the BIOS ROM area, 
> > > then
> > > it's just a linked list of tables, and the serial number is 
> > in table 1. 
> > >   Get the index, and scan past the requisite number of 
> > nulls, to find 
> > > the appropriate string; if there is no _DMI_ header, if 
> > table 1 doesn't 
> > > exist or is less than 8 bytes long, the string number is zero, or 
> > > finding double-null during the scan is an error, meaning no data is 
> > > available.
> > > 
> > > This should be doable.
> > > 
> > > 	-hpa
> > 
> > _______________________________________________
> > SYSLINUX mailing list
> > Submissions to SYSLINUX at zytor.com
> > Unsubscribe or set options at: 
> > http://www.zytor.com/mailman/listinfo/syslinux
> > Please do not send private replies to mailing list traffic.
> >