[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Installer problem report (dual boot loadlin)



On Wed, Mar 15, 2006 at 12:24:55PM +1100, Lachlan Patrick wrote:
> Well, the question I wanted it to ask was "which partition
> should /boot go onto?" AFAI recall that question wasn't
> asked. I had two other ext2 partitions, so I was just
> surprised that the command to format /dev/hda6 was also
> taken to mean that /dev/hda6 was where /boot should go.
> I wonder how the installer decides where to place /boot
> if I had formatted multiple partitions? I also wonder what
> it would have done if I'd chosen not to reformat /dev/hda6
> as ext3 (it was an empty ext2 before)? My point is
> formatting/partitioning != choosing where to install.
> Maybe there was a screen asking that, but I don't recall
> one. I just remember being surprised I wasn't asked.

/boot goes to whatever is mounted as /boot in the partition setup.  If
you have a seperate partition for /boot you tell the partition setup you
want it used as /boot and it does.  If you don't have a seperate one, it
obviously has to go on whatever is used as / in the partition setup.

> As a user of Debian, but not a frequent installer of Debian,
> I had no way to know whether loadlin was 'good', all I knew
> was it worked. As far as I knew, loadlin was an officially
> supported and sanctioned and 'good' boot method... I was
> looking for it in the installer menu options. I just had no
> way of knowing that it was deprecated or obsolete... where
> is that written? Maybe I just missed it.

I don't think anyone ever considered loadlin a good sactioned boot
loader, just a hack for dos users to use as a way to get something
loaded, often as a way to start an installer before having a proper boot
loader installed.  And of course a decade ago when some distributions
supported installing with umsdos filesystem, loadlin was the way to boot
those.  Fortunately that mess has disappeared.  loadlin is the only
remnant left and certainly hasn't had any maintenance done on it in a
while.  I would not be surprised if the next time some change is
required in boot loaders to support a kernel boot change, loadlin will
not be updated.

> OK, but trust is a subjective thing. I had a working loadlin
> system for five years; I trusted it, it worked through
> several revisions of kernel 2.2. By contrast, I didn't
> trust LILO, having tried and failed to get it working
> (bear in mind I was a total Linux newbie five years ago,
> and I knew loadlin worked, so I switched back from LILO
> to loadlin when I couldn't get LILO working.) For me,
> "if it works, why change it?" I mean, I don't even know
> why GRUB is preferred to LILO now. I just wanted a method
> which boots, and I had one, so I thought, why change it?

For a dos user, a dos program makes sense.  And lilo can be a real pain.
Grub is much nicer.

> Oh, I realise that now, and I wasn't suggesting otherwise.
> But I only discovered that by trying it.
> 
> My perspective was odd... I had a previously installed
> Debian system, with a blank 2GB ext2 partition ready and
> waiting for a new installation, so I didn't _need_ to
> partition anything. The installer expected me to, though.
> So, only having a "write changes" button, when potentially
> nothing needed to change, seemed odd. But only from my
> perspective. For a new user, with an empty PC hard drive,
> partitioning would be necessary, and "write changes"
> would therefore be necessary.

Even if you don't want to create partitions you still have to tell it
which existing partition to use for what (even if you don't want to
format them).  The partition setup does all that.

> OK, I didn't know loadlin/FIPS wouldn't work with recent
> versions of Windows. Thanks for that info!

They are dos tools.  Windows 98 was the last version they work on.  That
is quite a long time ago.

> I had used that combination five years ago, and it worked
> fine, so until today I assumed it would still work, or
> was still supported, or that recent upgrades to those
> packages would work with recent versions of Windows.
> I'm a bit surprised that they've been allowed to lapse.
> Surely being able to install Linux and boot into it from
> an icon on the Windows desktop is a desirable way to
> convert the unwashed masses? Forcing people to start
> with a clean PC, or lose their existing Windows partition,
> seems IMHO to raise the barrier for entry. Or is it that
> Windows won't play fair and is putting files all over the
> disk to stop FIPS-like disk repartitioning?

You can't do anything like that from inside any 32bit windows.  You
could only do that on the win98 and older because they were a 32bit
shell running on top of DOS, so they could drop down to dos and run
loadlin.

> As it turns out, the initrd parameter solved the problem,
> and I now have loadlin booting kernel 2.4.27.
> 
> No, not much, apart from a vague unease about formatting,
> partitioning, or changing MBRs if I didn't need to (the
> disk had existing data and operating systems on it).
> But see below for a comment on incremental vs big changes.
> 
> I'm hearing you on FM, buddy.
> 
> I just expected (I don't know why) that the installer
> also had an upgrade-detection capability. I don't know
> why, I just assumed it would. (My apt database got
> corrupted a few years ago, long story omitted.)

No, redhat used to do that.  Debian's upgrade system is designed to handle
incremental updates during development as well as well release to
release updates.  It is all done while the system is running an usable.
No need to reboot other than to change kernels, not even for upgrades.
Debian is not windows, and not even redhat. :)

> I just had a working system so I was trying to keep it
> working the same way. The working system used a choice menu
> in the Win98 autoexec.bat file, firing off loadlin to boot
> into Debian after 3 seconds, unless the user hits an arrow
> key to select Win98 instead, in which case the Windows GUI
> starts. I had a vague preference for keeping this two-stage
> bootup so that if something went wrong with the installation
> I would still have a booting Win98 system, and I could
> still go back to the old Linux 2.2 system. GRUB was a
> new thing to learn. I'm sure it's all very shiny. It just
> wasn't incremental for me, it was a break from the old,
> with the potential to screw up my system if I did something
> wrong, whereas loadlin seemed lower risk (it just involved
> copying some files to /dev/hda1). One of the reasons why
> I like Debian is that it is incremental... old software
> generally doesn't just break because someone decides I must
> do a forced upgrade (unlike other operating systems).
> So, it's not that I didn't trust GRUB, or didn't trust it
> to modify the MBR correctly, it was more that I didn't
> trust *me* to get the configuration files right (since
> I'd earlier stuffed up a LILO config file). Whereas I
> did trust myself not to break anything with loadlin (it
> just involved copying a few files and editing a script).
> So, once bitten, twice shy.

Well I would say 99.9% of the time the debian installer gets the grub
setup correct and working for all the OSs on your machine.  It didn't
used to be like that and some tweaking was sometimes required.
Microsoft's solution is much simpler: Don't support other OSs and don't
support dual boot systems (except recognizing older windows versions).

> OK, well, I didn't realise that using loadlin wasn't very
> realistic, since it's been working for me for five years,
> and it now works with Debian 3.1.r1 for me. So, I just
> didn't know it was deprecated, and I'm unsure how I was
> supposed to know, short of asking here.

Sure it can work, but the installer doesn't have a reason to support it,
since loadlin is unusual to use, and to actually set it up fully
requires messing with autoexec.bat and config.sys on the DOS system, or
creating a batch file with the right options, and then adding some
support for updating this when new kernels installed (since that is done
for lilo and grub users already).  It just isn't realistic to support,
when there are much better solutions already supported.

> But it looks like GRUB is the way to go in future, and
> certainly for new machines (which don't have Win98!).
> So thanks for the info!

GRUB in the MBR is the right solution for almost all machines.

Len Sorensen



Reply to: