Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

IDE drive works on RT/AOS!

1 view
Skip to first unread message

Christopher D Dukes

unread,
Oct 14, 1994, 11:32:39 PM10/14/94
to
In article <37mceh...@cs.utk.edu>, Keith Moore <mo...@cs.utk.edu> wrote:
>Summary: I just booted RT/AOS from a 545 Mb Seagate IDE drive.
>
>8. "newfs" didn't know anything about the new drive, so I used "mkfs"
>to create file systems on the new partitions. (I got partition sizes
>from "minidisk" under AOS).
Sounds fascinating... I'm going to have to go source code hopping
to find out where newfs figures things out. Perhaps it can be rewritten
to make guestimates based on information from minidisk.

Roger, if you see this. per chance did you already do this with
BSD44(almost)? Because I have absolutely no problems with the H310s that
I reformatted to 1225 cylinders and 35 tracks.
>
>9. Looked at the code in /usr/mdec/installboot to see how to
>write a partition to the disk. It turns out that you can use
>'minidisk' to get the sector address where the "boot" partition
>starts, and then use 'dd' to copy the boothd program onto the
>disk. I used a small 'c' program to build a boot block for
>this drive.
Would you mind releasing that 'c' program? I'm sure I can scrounge
something from the BSD44 source... but two bits of source code for
a given problem is always nice.
--
If you ask the wrong questions, you get answers like '42' or 'God'
cdd...@eos.ncsu.edu cdu...@nyx.cs.du.edu pak...@vnet.ibm.com
html://www2.ncsu.edu/eos/users/c/cddukes/mosaic

Roger Florkowski

unread,
Oct 15, 1994, 7:07:52 AM10/15/94
to

This is fantastic news !!! Thanks Keith !!!


In article <37nign$p...@taco.cc.ncsu.edu> cdd...@n00044-118dan.unity.ncsu.edu (Christopher D Dukes) writes:
>In article <37mceh...@cs.utk.edu>, Keith Moore <mo...@cs.utk.edu> wrote:
>>Summary: I just booted RT/AOS from a 545 Mb Seagate IDE drive.
>>
>>8. "newfs" didn't know anything about the new drive, so I used "mkfs"
>>to create file systems on the new partitions. (I got partition sizes
>>from "minidisk" under AOS).
>
>Sounds fascinating... I'm going to have to go source code hopping
>to find out where newfs figures things out. Perhaps it can be rewritten
>to make guestimates based on information from minidisk.
>
>Roger, if you see this. per chance did you already do this with
>BSD44(almost)? Because I have absolutely no problems with the H310s that
>I reformatted to 1225 cylinders and 35 tracks.

Of course I'll see it... I always read this newsgroup ;-)

AOS's newfs (ie: BSD-tahoe and earlier) depended solely on /etc/disktab
for how to partition the drive. (Else you can always do it yourself
with mkfs).

Starting with BSD-RENO, mkfs doesn't exist, and newfs is smarter.......
it queries the hard drive for the partition table and uses that
information. The only "gotcha" with the RT implementation is
if you use minidisk on a running system (ie: after you've booted,
and not from the sautil disk) *then* expect newfs to "do the right
think"..... it wont. You must reboot first. This is b/c the hd
driver reads the minidisk block only ONCE at boot time.
Starting with BSD44, protection is built in so that you can't
do a newfs on a mounted filesystem (and shoot yourself in the foot!)

>>9. Looked at the code in /usr/mdec/installboot to see how to
>>write a partition to the disk. It turns out that you can use
>>'minidisk' to get the sector address where the "boot" partition
>>starts, and then use 'dd' to copy the boothd program onto the
>>disk. I used a small 'c' program to build a boot block for
>>this drive.
>
>Would you mind releasing that 'c' program? I'm sure I can scrounge
>something from the BSD44 source... but two bits of source code for
>a given problem is always nice.

Its in the RT-FAQ. There are two useful programs, brec_dump and mkbrec.
They should be in /usr/ibm, but I forgot to install them. Sooo...
in the source tree, look in /sys/rt/stand/.


On a slightly related subject, the RT-NetBSD port is progressing
(very slowly!) A number of people have offered to help, thanks,
thanks, thanks! I was just donated a number of RT,s (5? 7? I
didnt count) so I have crash-n-burn heaven over here. I'll post
when there is something exciting to write about.

ROg.
--
Roger Florkowski - Advanced Network & Services, Inc. - ro...@ans.net

Keith Moore

unread,
Oct 14, 1994, 12:42:57 PM10/14/94
to
Summary: I just booted RT/AOS from a 545 Mb Seagate IDE drive.

Details:

1. I bought a dirt cheap ($15) IDE controller and a Seagate ST5660A
545 Mb IDE drive ($300 mail order, about $360 retail)

2. I configured the controller to be the second hard disk card, and
disabled its floppy controller, printer ports, serial ports, and game
port. (and that's a *minimal* IDE card!) Jumpered the IDE drive for
"master", with no "slave" attached.

3. Attached a 3 1/2 inch to 5 1/4 inch adapter to the drive, attached
drive mounting rails to the adapter, and slid the drive in slot C of
an RT tower. (The cabling was a bit of a stretch using the cables
that came with the controller; installation in other slots or in a
desktop unit might require custom cables.)

4. Booted the AOS sautil disk and told it to format the disk. I had
to answer several configuration questions about the drive geometry
(1057 cyl, 16 heads, 63 sectors/track), interleave (1 to 1), drive
type (I used ESDI), disk name (I used "ide545", but I think it's
better to pick a name that begins with "hd"), write precompensation
(none), sector size (512), etc.

5. Ate dinner while the disk was formatting.

6. Used the sautil minidisk utility to create a minidisk directory
and several partitions.

7. Booted AOS (from an EESDI drive)

8. "newfs" didn't know anything about the new drive, so I used "mkfs"
to create file systems on the new partitions. (I got partition sizes
from "minidisk" under AOS).

9. Looked at the code in /usr/mdec/installboot to see how to


write a partition to the disk. It turns out that you can use
'minidisk' to get the sector address where the "boot" partition
starts, and then use 'dd' to copy the boothd program onto the
disk. I used a small 'c' program to build a boot block for
this drive.

10. Copied a root file system (using dump/restore) from an EESDI
drive to the root partition on the IDE drive.

11. Halted and booted AOS from the IDE drive (using "hd(3,0)vmunix")

----
Here's the relevant lines from autoconf:

...
hdc1 adapter f0000170 IRQ 14 CPU level 4
hd2 at hdc1 slave 0
hd2: ide545 type (1057 cyl, 16 tracks, 63 sectors); interleave factor is 1 to 1
...

And here's the minidisk directory listing:

number=5 level=1 unused=5 first=0 last=4 bad_block=1063448 bad_size=1000
index iodn name date start size type

0 32746 boot Fri Oct 14 03:29:40 1994 252 756 01 ipl
1 32747 hd2a Fri Oct 14 08:21:41 1994 1008 32256 00
2 32748 hd2b Fri Oct 14 08:22:03 1994 33264 67536 20 swap
3 32749 hd2h Fri Oct 14 08:22:20 1994 100800 292320 00
4 32750 hd2g Fri Oct 14 08:22:35 1994 393120 670328 00

----
random notes:

a) to really test booting from the new drive, I'll need to remove
the EESDI controller and re-jumper the IDE controller to be hdc0.

b) I had various problems with using minidisk under AOS (except to
display partitions; that worked fine). I recommend always using
minidisk on the sautil disk when diddling partitions.

c) I first tried this on an old (c. 1989) Western Digital IDE drive.
That drive also worked, but AOS complained about a few errors during
autoconfig. I suspect that the newer drives work better in general
than the older ones, but I've only tried the two drives.

d) after power up, make *very* sure that the date is still correct
before scribbling on any files. (this won't affect the ide drive
itself too much, but it *will* affect things like NFS and dump/restore
that you might use to set things up on the new drive...)

----

I'm looking forward to installing this drive in a model 115. I used
to do RT development on a tower with 3 70Mb EESDI's (which wasn't
enough, just the best I could get), now I'll have 545 Mb in a desktop
unit which is *much* quieter. And at a very reasonable price!

--
Keith Moore NETLIB development group
Computer Science Department / University of Tennessee at Knoxville
107 Ayres Hall / Knoxville TN 37996-1301
Let's stamp out license managers in our lifetime.

0 new messages