It can be done.
--snip from web site you posted
map-drive = 0x80 / to = 0x81
map-drive = 0x81 / to = 0x80
snip--
Is the key. It switches the bios flag master to slave, slave to master, and fools Winbloze.
I made an error, I should have said master and slave of ide0, not ide1 and ide0. But I believe winnt/xp does not like booting from master ide1 to master ide0 either. Win98 has no trouble with this and you can edit the environment variables in dos before win98 complains about it.
One ide device can handle two drives. For some reason all the advice I have read on the net says that the master is faster than the slave but i have not experienced any difference, especially in newer motherboards. The only draw back to having 2 hard drives on one device is that they can't both talk to each other or to another device at the same time. So if you are burning a cd on a fast cd-rw while encoding the iso on the fly, having the source hard drive on ide0 slave and the destination cd-rw on ide1 slave should not be a problem. On the other hand, having your
source hard drive on ide0 master and the destination cd-rw on ide0 slave could possibly cause buffer underun while the two drives compete to communicate. To my understanding, each device, master and slave, uses almost all the wires on a 40 pin cable (80 pin cable uses some sort of inductive electrical property to be able to communicate at high mhz, but still goes into regular 40 pin connector). One of the cable's wires tells the drives whose turn it is to talk by switching on and off, off for master on for slave (don't quote me on that). This is similar to how scsi works, but is not even close to the performance level scsi runs it at. Here is one of my setups for cd burning and hard drive cloning:
/ master 60gig hard drive (/dev/hda)
ide0 -
\ slave 48x cdrom (/dev/hdb)
/ master 26x (read) 16x (write) 4x (erase and write) cd-r/rw (/dev/hdc)
ide1 -
\ slave misc hard drive (/dev/hdd)
The hard drives are not both masters because the computer mostly burns from hda (data) to hdc (cd-rw) and sometimes clones cd's from hdb (cdrom) to hdc (cd-rw) or clone partitions from hda (data) to hdd (clone). These three data tranfer paths are all from one ide device to the other and never from master to slave.The only drawback I see in this configuration is burning from hdd to hdc (slave to master), but the machine that has this config never does that. I setup this from all the advice I read on the net that the masters (hda,hdc) are faster and therefore a safer burn at high speed. Example of possible speed increase: Installing a big application suite of bloat-ware for winbloze and then installing the same suite from another cdrom drive on the same machine but from a different ide device makes installation goes 3-4 times faster.
If you are not going to transfer huge amounts of data around, especially from the Linux drive to the xp drive, and you are going to burn/encode cdroms or burn/play dvds, here is what i recommend:
/ master winxp hard drive (/dev/hda)
ide0 -
\ slave linux hard drive (/dev/hdb)
/ master cd-r/rw or dvd-r/rw (/dev/hdc)
ide1 -
\ slave unused, temp, or backup device (/dev/hdd)
lilo.conf (not tested, not finished, don't cut and paste)
boot = /dev/hdb
map=/boot/map (map to linux images, default)
install=/boot/boot.b (chain loader?, default)
LBA32 (default, but leave as reminder)
prompt
timeout = 50
delay = 300
vga = normal
verbose = 5 (gives lots of info, comment out or lower num. when successful)
disk=/dev/hda bios=0x80
disk=/dev/hdb bios=0x81
disk=/dev/hdc bios=0x82
disk=/dev/hdd bios=0x83
image = /boot/stable_kernel.img
root = /dev/hdb1
label = Stable
read-only
image = /boot/test-kern.img
root = /dev/hdb1
label = TestKernel
read-only
optional
image = /boot/original.bare.i
root = /dev/hdb1
label = bare.i
other = /dev.hda1
label = WinXP
table = /dev/hda
change
automatic
map-drive = 0x80 to = 0x81
map-drive = 0x81 to = 0x80
Here is a big guide --
http://home.san.rr.com/johninsd/pub/...ersion21_docs/ -- and the source is close by. I downloaded the latest source code and will skim through it soon, but I'm a noob in C.
I am not familiar with new Ami bios's, but it is fine. I was not aware your comp was fairly new, I am used to Pentium pro 300's and the like with fussy bios's. I have a fairly new award bios on a machine that auto detects the same values and is capable of booting win98, zip-slack, and slack9.1, in both auto and "matched" modes. There may be an option to have the bios read and match the drives' partition tables. This will probably not affect linux, but it may speed up windows. Any change to the bios can be reversed, but could prove fatal to a drive depending on the os. I don't recommend changing it unless the drive is a tad slow or you can afford the time to rebuild the os on it. Of course I tried it, but... thats what I do.
just for fun, man sfdisk, it has some interesting info, but don't use it
grub may be a better choice to use with winxp, but I haven't looked into grub much, but I read somewhere that it is more forgiving than lilo and can be altered at boot prompt.
OK, that being said, I can now die happy? -crap, not yet, must study...
i've just managed to get an install of gentoo finally running on my systed (after about 20 hours of emerging 
)
. lilo works fine, but only the gentoo bit.
