[comp.os.minix] ST Minix -- Mounting the hard disk on /

karl@decwrl.dec.com (Karl Rowley) (11/10/88)

ST Minix is a pretty impressive piece of work, and it works well with the
Atari hard disk.

For those who have 1 megabyte of memory and a hard disk, there do appear to
be some drawbacks to the system as it is distributed.  In order to load the
root filesystem from the hard disk, you must have a fairly small partition
dedicated to it.  Another partition is required if you want to keep /usr on
the hard disk.  Since the Atari software only supports four hard disk 
partitions, this leaves a maximum of two partitions for other purposes.

The hard disk is probably fast enough to make the ramdisk unnecessary in 
some situations.  At boot time it would be nice to be able to simply 
mount the root filesystem from one partition on the hard disk, and forget
about the ramdisk completely.  

Of course, one could use a ramdisk that was only large enough to hold the
necessary files for getting the system started.  However, the ramdisk would
still be mounted on / and the hard disk would still have to be mounted on
some subdirectory to /.  In this situation /tmp and /usr could not both be
on one hard disk partition.

What would have to be changed to eliminate use of the ramdisk?  Does such
a change require recompiling the kernel?

				Karl Rowley
				Evans and Sutherland Computer Division
				Mountain View, California
				...!decwrl!escd!karl
				escd!karl@decwrl.dec.com

dtynan@sultra.UUCP (Der Tynan) (11/11/88)

In article <5269@louie.udel.EDU>, escd!es66!karl@decwrl.dec.com (Karl Rowley) writes:
> 
> The hard disk is probably fast enough to make the ramdisk unnecessary in 
> some situations.  At boot time it would be nice to be able to simply 
> mount the root filesystem from one partition on the hard disk, and forget
> about the ramdisk completely.  
> 
> What would have to be changed to eliminate use of the ramdisk?  Does such
> a change require recompiling the kernel?
> 
> 				Karl Rowley

I've done this for the IBM-PC version, for two reasons; i/  I want to run
tasks in the background, which you can't really do, when ~240K is missing,
and ii/  I was sick of running out of space in /tmp.  The changes I made,
were relatively simple.  There are two variations on this theme.  One uses
the /dev/ram purely for /tmp, which speeds up compiles, and alleviates the
'no space' issue somewhat, but it still takes up precious PC memory.  What
I did is remove all traces of the RAMDISK, from the kernel (yes, it has to
be recompiled).  I don't have the information with me, but I had to edit
files in fs, mm, and of course, kernel.  Be careful when doing this, because
I managed to reinitialize my hard disk in the process.  Personally, I think
that there is probably enough interest to warrant using #ifdefs in the code.
If there's enough interest, I'll post the diffs.
						- Der
-- 
	dtynan@sultra.UUCP  (Dermot Tynan @ Tynan Computers)
	{mips,pyramid}!sultra!dtynan

 ---  God invented alcohol to keep the Irish from taking over the planet  ---

V050KY8G@ubvmsc.cc.buffalo.edu (11/11/88)

The message written to the newsgroup concerned having the root of the file
system be a hard disk partition on MINIX-ST.  Although I work with PC AT 
MINIX, I assume it is sufficiently similar enough to make an intelligent 
comment.

For my system, I changed the #define's for BOOT_DEV and ROOT_DEV and 
recompiled.  Actually, what happens in my system, as set up now, is the 
RAMdisk is somewhat unnecessarily loaded from a hard disk partition, then 
wiped out with mkfs, then mounted on /tmp.  The root of the file system is 
another hard disk partition, and is so declared at load/boot time by
ROOT_DEV being "compiled in."

Seems to me that perhaps a more elegant solution would be to use chroot() 
to move the root file system to a hard disk partition, perhaps after it is 
mounted (could it be done something like chroot to /hd-mnt-dir/dev/hd0 ??).

Has anyone else implemented using the RAMdisk as mounted on /tmp ?????

beattie@visenix.UUCP (Brian Beattie) (11/12/88)

In article <2637@sultra.UUCP> dtynan@sultra.UUCP (Der Tynan) writes:
^In article <5269@louie.udel.EDU>, escd!es66!karl@decwrl.dec.com (Karl Rowley) writes:
^> 
^> The hard disk is probably fast enough to make the ramdisk unnecessary in 
	[text deleted]
^> 				Karl Rowley
^
^I've done this for the IBM-PC version, for two reasons; i/  I want to run
	[ text deleted ]
^I did is remove all traces of the RAMDISK, from the kernel (yes, it has to
^be recompiled).  I don't have the information with me, but I had to edit
^files in fs, mm, and of course, kernel. 
I did a similar thing.  If ROOT_DEV has a major device of 3 (wini) I do
not create a ramdisk and skip the copy of BOOT_DEV to ROOT_DEV.  The
changes were limited to fs/main.c.
^                                                        Personally, I think
^that there is probably enough interest to warrant using #ifdefs in the code.
^If there's enough interest, I'll post the diffs.
^						- Der
likewise when things settle down here.
^-- 
^	dtynan@sultra.UUCP  (Dermot Tynan @ Tynan Computers)
^	{mips,pyramid}!sultra!dtynan
^
-- 
    _ANYONE_     | Brian Beattie          (703)471-7552
can sell software| 11525 Hickory Cluster, Reston, VA. 22090 
that has already | beattie@visenix.UU.NET
been written     | ...uunet!visenix!beattie