Success making bootable and persistent USB key with SYSLINUX

[zoltanthegypsy]

Well Done!

I'm still stuck on "geometry" as one of the keys to bootability, though. I get maximum bootability by formatting (and creating the partition) with the HP utility even though I then install grub in the mbr - which overwrites the boot code but not the part table. Touching the part table (other than setting the active flag) with Linux fdisk limits bootability on some of my test boxen. (Pretty sure about this - I need to re-test, though with Linux fdisk in expert mode).

I've also tried putting grub in the partition's boot sector and using a box-stock XP mbr w/the HP part table. That seems to boot as well as the HP + grub mbr approach: machines that reject fob formatted w/Linux will still boot this combo.

Still, I'd be curious to know if any other failures happen, when trying to boot from a USB thumbdrive that has been correctly recognized by the BIOS. What BIOS versions are in the computers that worked to boot drives formatted with the HP formatter but didn't work to boot drives formatted with other tools?

That, to me, is by far the hardest problem: getting the BIOS to see the drive, in the first place! If the BIOS can see the USB thumbdrive well enough to set it up as the bootable "hard drive" (the 0x80 device), and run the MBR properly, things just work from there, I have found.

One older Intel MB/BIOS will try to boot my test fob if I use the HP formatter, but hangs solid at "GRUB".

I will try to find the time to experiment more w/geometry and see if my theory has any merit. AND note the BIOSes involved.

Z.

[Krellan2]

Nice, thanks for checking into it. I've looked into it about as much as I can, given that my computers either all boot perfectly or fail to recognize the USB key at all = no in-between behaviour at all, unlike what you have seen.

As for geometry, a thought: since USB keys have no hardware CHS geometry at all (a good thing!), maybe a BIOS of yours is synthesizing the geometry from the CHS ending values in the partition table. It's using the head and sector numbers given, to set the geometry. That may be why the HP formatter limits itself to only including full cylinders in the CHS values, to guarantee that the head and sector numbers will be maxed out. Even though they are technically wrong (the total CHS count is then undersized, and doesn't match the LBA count). If the BIOS is reading these somehow, to establish a geometry, this would at least get the BIOS to establish the proper */255/63 values.

Ironically, including a proper CHS count in the partition table, instead of undersized like this, would cause failures, as the BIOS would then be getting bogus counts from the final, partial, cylinder at the end of the disk.

The BIOS could even be doing this from the extra Microsoft geometry fields in the PBR, at the beginning of FAT and FAT32 partitions. I did not check these at all. You might want to check these numbers and see if they compare with what is in the PBR.

Another experiment would be to try and replicate the HP formatter's patch, to GRUB's boot sector code: just force it outright to use LBA, and completely disable the use of CHS entirely, for loading any more sectors, no matter what the size of the disk is, or what the BIOS claims it supports. Since GRUB is open source, it should be a lot easier to make this patch (although the MBR is so small that it can fairly easily be completely reverse-engineered, as it was on that webpage I mentioned earlier).

Good luck!

Josh

[hardawayd]

I am having trouble getting knoppix on usb to connect to the internet. Can someone tell me if knoppix has the ipw2200 firmware necessary for my dell laptop centrino wireless chip to work?

[Krellan2]

I am having trouble getting knoppix on usb to connect to the internet. Can someone tell me if knoppix has the ipw2200 firmware necessary for my dell laptop centrino wireless chip to work?

I don't know about that chip, but I doubt it. The problem with firmware is that it is copyrighted. So, Knoppix can't distribute it. For a CD-based live distro like Knoppix, this really sucks, because you will have to remaster the CD yourself or download the firmware each time you boot up the CD. That's just one of the many problems with binary-only firmware.

http://ipw2200.sourceforge.net/firmware.php?fid=5
(big ugly licence agreement here)

Since you have a USB key, this will be slightly easier, because you can just copy the firmware file to your USB key and load it from there each time.

Unfortunately, it is very hard to find "clean" 802.11g cards: chipsets that don't require firmware or ndiswrapper in order to run. There are some reverse-engineered drivers that don't require firmware, but they tend to be unstable, as the reverser doesn't have enough information to understand the device fully and make a truly reliable driver (ACX111, Broadcom).

The only card I have found so far, that has both a manufacturer-supported Linux driver and no firmware/ndiswrapper requirement, is RaLink. Are there any others?

[zoltanthegypsy]

I am having trouble getting knoppix on usb to connect to the internet. Can someone tell me if knoppix has the ipw2200 firmware necessary for my dell laptop centrino wireless chip to work?


I don't know about that chip, but I doubt it. The problem with firmware is that it is copyrighted. So, Knoppix can't distribute it.

Actually, the firmware _does_ seem to be included w/4.0.2. At least, my T43 Centrino w/intel 2915 wifi comes up and associates OK. IIRC, the 2915 uses the ipw2200 driver and firmware. So far, it doesn't do DHCP correctly though. I'm working on that.

Z.

[zoltanthegypsy]

For the record, w/Knoppix 4.0.2, the ipw-2.2-xxx firmware is in /usr/lib/hotplug/firmware.

This is getting quite a way off topic

Z.

[spalla]

hey there i followed all the instruction, but i keep getting the same error message:

could not find kernel image: linux

any idea what i did wrong?

thank you for your help