#!/bin/sh -x
disk=ad8
disklabel -W ${disk}
dd if=/dev/zero of=/dev/${disk} bs=1k count=10
fdisk -BI ${disk}
disklabel -w -B ${disk}s1 auto
disklabel ${disk}s1 >/tmp/label.$$
disklabel ${disk}s1 |
egrep unused |
sed "s/c:/a:/" |
sed "s/unused/4.2BSD/" >>/tmp/label.$$
disklabel -R -B ${disk}s1 /tmp/label.$$
rm -f /tmp/label.$$
boot0cfg ${disk}
On Friday, Jan 17, 2003, at 12:09 US/Pacific, Dirk-Willem van Gulik wrote:
In short: A working procedure for making an ATA disk bootable does not
work reliably with CF cards/with a Soekris.
On FreeBSD with a normal IDE/ATA hard disk the following sequence:
fdisk -BI ad0
disklabel -w -B ad0s1 auto
disklabel -e ad0s1
... newfs what is needed
boot0cfg -v -B ad0
will reliably turn an IDE disk into something bootable (followed after a
make install-world DESTDIR=..)
Is this the correct sequence ?
And though I can boot from CF cards; things are bit hit&miss; and it is
not clear to me exactly what the right procedure is.
The issue is that I seemingly get random results. And simply re-running
the script or rebooting changes behavour.
Issues seen are:
-> Flashcards simply do not want to be 'seen' anymore with
/stand/sysinstall; it will bail out wit no disk detected.
-> The flashcard show *different results* on successive run
with disklabel after an init iwth
fdisk -BI ad0:
c: 249824 0 unused 0 0 # (Cyl. 0 - 247*)
or
c: 249824 0 unsued 1024 8192 16 # (Cyl. 0 - 247*)
quite random; with the same card and may change with a reboot.
Then after a disklabel edit once sees either one of the following
three:
a: 249824 0 4.2BSD 0 0 # (Cyl. 0 - 247*)
c: 249824 0 unused 0 0 # (Cyl. 0 - 247*)
or
a: 249824 0 4.2BSD 1024 8192 16 # (Cyl. 0 - 247*)
c: 249824 0 unsued 0 0 16 # (Cyl. 0 - 247*)
or
a: 249824 0 4.2BSD 1024 8192 16 # (Cyl. 0 - 247*)
c: 249824 0 unsued 1024 8192 16 # (Cyl. 0 - 247*)
Again; mixed across cycles and boots. Is there anything in the
new GEOM or BIOS which can be remembered across reboots, including
power cycles.
Right now the following commands :
boot soekris diskless
with GENERIC kernel using usual /var and /tmp on mfs.
setenv TERM vt100
fdisk -BI ad0
disklabel -w -B ad0s1 auto
disklabel -e ad0s1
-> add 'a: 4.2BSD line'
-> remove fsize/bsize
a: 249824 0 4.2BSD
a: 249824 0 untitled
disklabel ad0s1 will output 8 out of 11 times:
a: 249824 0 4.2BSD 1024 8192 16 # (Cyl. 0 - 247*)
c: 249824 0 unsued 0 0 16 # (Cyl. 0 - 247*)
boot0cfg -v -B ad0
9-10 cases will give
GEOM: new disk ad0
ar: FreeBSD check1 failed
ad0: <Hitachi CVM1.1.1/Rev 1.01> ATA-0 disk at ata0-master
ad0: 122MB (250368 sectors), 978 C, 8 H, 32 S, 512 B
ad0: 1 secs/int, 1 depth queue, BIOSPIO
ad0: piomode=1 dmamode=-1 udmamode=-1 cblid=0
GEOM: Configure ad0s1, start 16384 length 127909888 end 127926271
GEOM: Add ad0s1 hot[0] start 512 length 276 end 787
GEOM: Configure ad0s1a, start 0 length 127909888 end 127909887
GEOM: Configure ad0s1c, start 0 length 127909888 end 127909887
Manual root filesystem specification:
<fstype>:<device> Mount <device> using filesystem <fstype>
eg. ufs:da0s1a
? List valid disk boot devices
<empty line> Abort manual input
mountroot> ufs:ad0s1a
panic: Root mount failed, startup aborted.
Debugger("panic")
Stopped at Debugger+0x54: xchgl %ebx,in_Debugger.0
and sometimes:
mountroot> ufs:ad0s1a
Mounting root from ufs:ad0s1a
start_init: trying /sbin/init
..hang...
and then again, 1 in 50 times it will boot just fine. What am I doing
wrong. How can I make this more reliable ?
Dw
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-hackers" in the body of the message
