[minix3] Tip: if MINIX 3 CD doesn't boot

[Evgeniy Ivanov]

Hello,

Some time ago there were few discussions like "installation CD doesn't
boot" and "MINIX 3 doesn't work on real hardware" (it was about
booting CD too).
As Thomas Veerman points, boot problems (CD) are caused by EL TORITO
issues (bios unsupports it). It can be either old BIOS or lugs like
with Dell hardware.

The easiest and fastest solution I found it to use Grub4Dos to boot
from CD (or even from just ISO).

Grub4Dos installation for any Windows NT:
http://grub4dos.sourceforge.net/wiki/index.php/Grub4dos_tutorial#Booting_GRUB_for_DOS_via_the_Windows_NT.2F2000.2FXP.2F2003_boot_manager

Chainloading to CDROM:
http://grub4dos.sourceforge.net/wiki/index.php/Grub4dos_tutorial#CDROM_related_subjects

hth

--
Evgeniy Ivanov

--
You received this message because you are subscribed to the Google Groups "minix3" group.
To post to this group, send email to min...@googlegroups.com.
To unsubscribe from this group, send email to minix3+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/minix3?hl=en.

[Colin Fowler]

Funnily enough, I'm having this problem at the moment. Today I have
spent over 8 hours trying to get this machine to boot the Minix CD.

This machine is a Dell Precision 490 workstation and refuses to boot
the CD. It runs Slackware and FreeBSD.

There's no Windows installed so I can't use Grub4dos.
Normal grub doesn't see CD drives.
This machine has no floppy drive
I then tried booting a memdisk floppy image from grub. Not enough low memory.
I then installed lilo
Using this I could boot an image of smart disk manager via memdisk. It
spins up the disk and then the machine hangs.
Grrrrr
Using the same method I managed to create a boot floppy from a dell
windows BIOS exe using a script on Linux to extract the bios etc. I
updated the machine bios using this and memdisk.
I tried again, no joy.

I've replaced the CD drive with a PATA one. No joy

The furthest I have got is by booting the bootflop.img from the minix
download site using the above memdisk method.
It spins up the disk, I get "unrecognized magic number" and then it
continues probing and eventually fails with FS panic (main.c): Cannot
open RAM image device.
I assume this is because the bootflop is from the 3.0.7 days.

I mounted the 3.1.6 ISO, but it only contains a readme.txt and a
version.txt. No bootflop.img stuck somewhere convenient.

Anyone got ANY ideas?

I do need Minix on this particular machine and I do need to be able to
install quickly from CDROM. I'm working on a device driver and the odd
time I do something dumb, manage to lock the machine and need to
reboot. MFS doesn't like sudden shutdowns and I have to reinstall
sometimes when fsck can't recover the fs.


5am am. Saturday night. Time to go home :)

Any ideas appreciated. My next step is to install X and virtualbox on
the slackware partition and using raw access to the main drive install
in a VM

Colin

[Evgeniy Ivanov]

On Sun, Apr 25, 2010 at 7:15 AM, Colin Fowler <eleth...@gmail.com> wrote:
> Funnily enough, I'm having this problem at the moment. Today I have
> spent over 8 hours trying to get this machine to boot the Minix CD.
>
> This machine is a Dell Precision 490 workstation and refuses to boot
> the CD. It runs Slackware and FreeBSD.
>
> There's no Windows installed so I can't use Grub4dos.

You can install it on Linux too. Also normal grub allow you to create
bootable grub CD, thus I assume grub4dos can do the same for you. See
grub4dos for linux installation instructions.
Also Ben Gras has newboot iso for MINIX 3:
http://www.few.vu.nl/~beng/minix3_1_7_newboot.iso.gz
I think you could try it before playing with Grub.

> I do need Minix on this particular machine and I do need to be able to
> install quickly from CDROM. I'm working on a device driver and the odd
> time I do something dumb, manage to lock the machine and need to
> reboot. MFS doesn't like sudden shutdowns and I have to reinstall
> sometimes when fsck can't recover the fs.

It's recommended to use virtual machines for such development. Then
you can use snapshots and easily recover from similar failures. For
some time I worked on real hardware, but now I tried VM and I don't
want return to bare metal now :)

> 5am am. Saturday night. Time to go home :)
>
> Any ideas appreciated. My next step is to install X and virtualbox on
> the slackware partition and using raw access to the main drive install
> in a VM

It's an option.

[bootbhai]

there is only one thing tat u can do mate!!!

throw d CD down into d deepest ocean!!

[Colin Fowler]

Thanks for the reply. VM development is pretty much useless for me as
the hardware I develop for is not emulated by Minix. This machine has
a variant of the hardware I wrote a driver for onboard the motherboard
so I can't rip it out like a PCI card and stick it in a more
cooperative machine. After I went home I actually installed X and
virtualbox remotely on Slackware so I still have that option if
Grub4Dos (A very deceptive name!) does not work out for me under
Linux. Changing to a 3.1.7 prerelease right now is not an option as
I'd be changing the development platform for my driver before it was
done. That could introduce lots of subtle little bugs. When the driver
is done, I will of course test and make any changes necessary to get
it working under 3.1.7

I'm off now to give Grub4dos a go :)

regards,
Colin

[Colin Fowler]

FINALLY!

I have grub4dos loading via lilo.
cdrom --init told me "NO CDROMS FOUND". After a lot of cursing, I
installed an old spare CDROM from a Mac G4, rebooted and it picked up
the drive. The install is terribly slow though and I keep
getting AT0-D0: timout on command 0xc8. Hopefully the install completes.

regards,
Colin

[tota]

> Some time ago there were few discussions like "installation CD doesn't
> boot" and "MINIX 3 doesn't work on real hardware" (it was about
> booting CD too).

May be problem boot MINIX 3 on real hardware in file usr/src/kernel/
arch/i386/system.c
line:

write_cr4(read_cr4() | CR4_OSFXSR | CR4_OSXMMEXCPT);

try change on:

write_cr4(read_cr4() | CR4_OSFXSR );

?

ToTa

[Ben Gras]

There's a 'current development' iso on the download page now:

http://www.minix3.org/download/

It includes a fix for the topic of this thread, i.e., the new booting
problem, hopefully fixing it for all machines and not breaking it for
any. Testers welcome. There are other improvements and bugfixes too of
course, see the svn log for details :)

[Tom Chandler]

3.1.7 - 6844 loops on native hardware, same as 3.1.6 iso. (at least on my platform)

 

3.1.5 is the last release that I can boot native.

 

Thank You

Tom C.

[tota]

What type of computer this is, information such as brand, model and
CPU?

> 3.1.7 - 6844 loops on native hardware, same as 3.1.6 iso. (at least on my
> platform)
>
> 3.1.5 is the last release that I can boot native.

[Larson, Timothy E.]

> 3.1.7 - 6844 loops on native hardware, same as 3.1.6 iso. (at least
> on my platform)

What is your platform? Not only would it be useful in general to know, but specifically I want to know if I should save myself the bother of testing this on my own box. :)

Tim
--
Tim Larson
App Admin, Intercall Production Services

[Thomas Veerman]

Tom Chandler wrote:
> 3.1.7 - 6844 loops on native hardware, same as 3.1.6 iso. (at least on
> my platform)
>
> 3.1.5 is the last release that I can boot native.
>

How far into the boot process do you get before your computer resets?
What output do you get? Does it show it's loading the kernel and system
servers?

--
Thomas Veerman

[Victor van der Veen]

On Mon, 2010-04-26 at 16:40 -0700, Ben Gras wrote:
> There's a 'current development' iso on the download page now:
>
> http://www.minix3.org/download/
>
> It includes a fix for the topic of this thread, i.e., the new booting
> problem, hopefully fixing it for all machines and not breaking it for
> any. Testers welcome. There are other improvements and bugfixes too of
> course, see the svn log for details :)
>

This one boots on my Dell Optiplex 740. However, I can't press "1" in
the boot monitor so I have to wait 10 seconds.

Output when pressing 1:


1

Loading Boot image 3.1.7.
kernel ds rs pm vfs memory log tty mfs vm pfs init (6342k)






*** kernel messages:
0x7ce0 0x7d95 0x7648 0x32f7 0x2d9e 0x77f1 0x7530 0x2787 0x25b9 0x100a
rs 2 0xa070 0x7e78 0x7662 0x7697 0x6b76 0x4f0c 0x303a 0x7ce0 0x7d95
0x7648 0x32f7 0x2d9e 0x77f1 0x7530 0x2787 0x25b9 0x100a
kernel pani: cause_sig: signal manager gets lethal signal 6 for itself
kernel: 0x510a 0x489d 0x6219 0x40c3 0x410a 0xe84
cs: RPL 0, ind 6 of GDT -> base 0x00001000 size 0x00012e14 exec DPL 0
ds: RPL 0, ind 3 of GDT -> base 0x00014000 size 0x18fec000 nonexec DPL 0

cd>




Victor van der Veen

[Antoine LECA]

Ben Gras wrote:
> There's a 'current development' iso on the download page now:
>
> http://www.minix3.org/download/
>
> It includes a fix for the topic of this thread, i.e., the new booting
> problem, hopefully fixing it for all machines and not breaking it for
> any. Testers welcome.

Great work Ben, congratulations.

Works here :-), with a Dell GX270 (2004, BIOS at last release A07),
while official 3.1.6 won't boot.


However, and while it is almost certainly not related, I see another bug
later: once the kernel is started, I got the "Initiating legacy i8253
timer" message (I guess this one is OK, "vintage" computer), immediately
followed by a cryptic (cannot know easily which process it is):
sys_privctl failed for proc 73132: -12
Then the system went a bit more further, looks for and finds the CD etc.
(I know this means all the servers started and the ramdisk etc/rc is at
hands), then hangs with the HD led on at
Loading ramdisk from /dev/c0d2p1
(Clearly the stop is in loadramdisk.)

I tried with no_apic=1 in the boot monitor, no joy.
I also tried fiddling with bios_wini=yes, but the CD does not show up
while enumerating using BIOS (I guess there is a call to terminate
ElTorito emulation.)

> There are other improvements and bugfixes too of course, see the
> svn log for details :)

Yes, I saw that you quitted the hack with the added 16-byte prefix and
I should thank you for doing it, I felt it was very ugly. :-)


Antoine

[Victor van der Veen]

Oh, and after installing, the system does not boot. Same error message
as below. After entering the boot monitor, any option results in this
behavior ("1", "2" or wait for auto boot).

Victor van der Veen

[Tomas Hruby]

> I tried with no_apic=1 in the boot monitor, no joy.

It is set by default. If apic were used, it would not initialize the
legacy timer but apic. apic timer is safe nowadays on real hardware,
it ticks too fast in VMs :-/ It does not really matter which timer you
use. Using apic is more for catching problems on different hardware.
Nevertheless it is good to avoid it in case of other troubles ;-)

Cheers, T.

[Antoine LECA]

Tomas Hruby wrote:
>> I tried with no_apic=1 in the boot monitor, no joy.
>
> It is set by default. If apic were used, it would not initialize the
> legacy timer but apic. apic timer is safe nowadays on real hardware,
> it ticks too fast in VMs :-/ It does not really matter which timer you
> use. Using apic is more for catching problems on different hardware.
> Nevertheless it is good to avoid it in case of other troubles ;-)

Thanks Tom for the explanations.



By the way, back to my earlier problem, I noticed the error message for
sys_privctl failed for proc 73132: -12 was linked with the ATA driver,
it does not occur with the BIOS driver. Hope it helps...


Antoine

[PaulD]

On Apr 27, 12:02 pm, Antoine LECA <Antoine.Lec...@gmail.com> wrote:
> Ben Gras wrote:
> > There's a 'current development' iso on the download page now:
>
> >http://www.minix3.org/download/
>
> > It includes a fix for the topic of this thread, i.e., the new booting
> > problem, hopefully fixing it for all machines and not breaking it for
> > any. Testers welcome.
>
> Great work Ben, congratulations.

As describe in https://gforge.cs.vu.nl/gf/project/minix/tracker/?action=TrackerItemEdit&tracker_item_id=431
3.1.5 was working fine for me, but not 3.1.6.
I am please to say that minix3_1_7_ide_r6844.iso works fine for me.
I can login as root, began setup, but stop before really messing my
current installation.
Anyway, I suppose this is not a full CD image, being only 15 MB
compressed, where 3.1.6 compressed image was about 462 MB.

[Tomas Hruby]

> Anyway, I suppose this is not a full CD image, being only 15 MB
> compressed, where 3.1.6 compressed image was about 462 MB.

It does not include packages. You must install from network. T.

[Ben Gras]

> Output when pressing 1:
>
> 1
>
> Loading Boot image 3.1.7.
> kernel ds rs pm vfs memory log tty mfs vm pfs init (6342k)
>
> *** kernel messages:
>  0x7ce0 0x7d95 0x7648 0x32f7 0x2d9e 0x77f1 0x7530 0x2787 0x25b9 0x100a
> rs 2 0xa070 0x7e78 0x7662 0x7697 0x6b76 0x4f0c 0x303a 0x7ce0 0x7d95
> 0x7648 0x32f7 0x2d9e 0x77f1 0x7530 0x2787 0x25b9 0x100a
> kernel pani: cause_sig: signal manager gets lethal signal 6 for itself
> kernel: 0x510a 0x489d 0x6219 0x40c3 0x410a 0xe84
> cs: RPL 0, ind 6 of GDT -> base 0x00001000 size 0x00012e14 exec DPL 0
> ds: RPL 0, ind 3 of GDT -> base 0x00014000 size 0x18fec000 nonexec DPL 0
>
> cd>

I'm pretty sure this is the samet hing as I've seen this on a test
machine recently, as it happens - workaround will be out soon that
will hopefully solve this too.

[Evgeniy Ivanov]

tota was right about flags in CR4. So patch on its way and it will
solve problems on computers without SSE1 instruction set (processors
prior PIII).


--
Evgeniy Ivanov

[Tom Chandler]

Sorry about lack of info.  my error.....

 

It is a PREMICO P2-233 with 512meg.  It has an Intel processor

 

I see later in this thread, there may be a patch for processor before P3's.

This may solve the problem.

 

thank you

Tom c.

 

 

Loading Boot image 3.1.7.
kernel ds rs pm vfs memory log tty mfs vm pfs init (6342k)

[Evgeniy Ivanov]

On Wed, Apr 28, 2010 at 3:11 AM, Tom Chandler <tchan...@gmail.com> wrote:
> Sorry about lack of info.  my error.....
>
> It is a PREMICO P2-233 with 512meg.  It has an Intel processor
>
> I see later in this thread, there may be a patch for processor before P3's.
> This may solve the problem.

Yes, it will fix the problem. On your P2 there is no CR4_OSXMMEXCPT in
CR4 (on P2 that bit is reserved and attempting to set a reserved bit
in register CR4 to a value other than its original value results in a
general-protection exception #GP.
There will be a more correct patch today for trunk. You can either use
that patch for your version or just remove CR4_OSXMMEXCPT like was
suggested.



--
Evgeniy Ivanov

[Ben Gras]

Dear all,

There's a new iso on the download page of r6875. The following
problems mentioned in this thread are hopefully solved:
- looping boot problem (CR4 bit set without SSE1 support), found and
debugged by Tota and Evgeniy Ivanov (reported here by Tom Chandler)
- RS panic with deep stacktrace reported by Victor van der Veen -
this is tty panicing on "kbc_read failed to complete". the 'solution'
is to try harder ;) (which works on test hardware I have access to)

Testers welcome.

[Victor van der Veen]

Forks fine on the Dell Optiplex 740. Nice job! :)

[Antoine LECA]

Ben Gras a écrit :

> There's a new iso on the download page of r6875. The following
> problems mentioned in this thread are hopefully solved:
> - looping boot problem (CR4 bit set without SSE1 support), found and
> debugged by Tota and Evgeniy Ivanov (reported here by Tom Chandler)
> - RS panic with deep stacktrace reported by Victor van der Veen -
> this is tty panicing on "kbc_read failed to complete". the 'solution'
> is to try harder ;) (which works on test hardware I have access to)
>
> Testers welcome.

Works here on the Dell GX270 (P4 2.3 GHz, CpuID F.24) which was stuck
with rev 6844, yet there is still a message about an error with the ATA
driver and privctl returning -12 (EAGAIN?), and I found it to be quite a
bit long to load the ramdisk although it finally succeeded, then I could
not get any farther since the keyboard is not operational.

OTOH the same disk boots fine on an Atom board (which was brocken with
3.1.6)


Antoine

[PaulD]

On 28 avr, 12:51, Victor van der Veen <vvdv...@gmail.com> wrote:
> Forks fine on the Dell Optiplex 740. Nice job! :)
>
> On Wed, 2010-04-28 at 07:48 -0700, Ben Gras wrote:
> > Dear all,
>
> > There's a new iso on the download page of r6875. The following

Well, I had no success with r6875. I did by the book:
-Download the iso in library with unreliable network
-Don't test md5sum
-Mount the USB key on /mnt
-Directly burn the cd image on /mnt by piping the output of gunzip to
burncd
-While seeing the image was bigger then the expected about 125 Mb,
used gzip -l to see that the
file had 93% compression, 14 Mb compressed, arount 240Mb uncompressed
-Ignore the Geometry warnings at the end saying that partitions on the
CD did not began on cylinders boundaries
-Reboot as fast as possible after burncd finished

Strangely enough, when I rebooted I got:
1 Regular Minix 3
Can't load boot/image: No such file or directory

I found out in boot monitor that:
'ls /boot' reports no file in that directory.
Nor in any directories, except for /bin.

I guess I will have to retry more slowly. :-)

[Ben Gras]

Great!

[Larson, Timothy E.]

Ooooooo, dare I try it on my GX60? :) :)

Tim
--
Tim Larson
App Admin, Intercall Production Services

[Evgeniy Ivanov]

On Wed, Apr 28, 2010 at 10:43 PM, Larson, Timothy E. <TELa...@west.com> wrote:
> Ooooooo, dare I try it on my GX60?  :)  :)

Like SAS says: "Who dares wins" ;)



--
Evgeniy Ivanov

[PaulD]

I checked my md5 sum of my r6875.iso.gz image and found it ok.
So I re-burned on a new media, and it works fine now.

[Tom Chandler]

A BIG THANK YOU to all on the list who put up with my comments/questions and

 

A REALLY BIG THANK YOU to all who worked on the problem.

 

I have installed 3.1.7 - 6875 on my P2-233 Intel base processor and it is working

 

          FLAWLESSLY.

 

Again, a sincere THANK YOU to all.

 

Tom C.

[Larson, Timothy E.]

> There's a new iso on the download page of r6875. The following
> problems mentioned in this thread are hopefully solved:
> - looping boot problem (CR4 bit set without SSE1 support), found
> and
> debugged by Tota and Evgeniy Ivanov (reported here by Tom Chandler)
> - RS panic with deep stacktrace reported by Victor van der Veen -
> this is tty panicing on "kbc_read failed to complete". the
> 'solution'
> is to try harder ;) (which works on test hardware I have access to)
>
> Testers welcome.

Why is this only 14 MB while previous releases are well over 400? I presume it is because this installs a "bare minimum" system only. If so, how to install the rest of the system?

Thanks,

Tim
--
Tim Larson
App Admin, Intercall Production Services

[Evgeniy Ivanov]

On Tue, May 4, 2010 at 1:46 AM, Larson, Timothy E. <TELa...@west.com> wrote:
>> There's a new iso on the download page of r6875. The following
>> problems mentioned in this thread are hopefully solved:
>>   - looping boot problem (CR4 bit set without SSE1 support), found
>> and
>> debugged by Tota and Evgeniy Ivanov (reported here by Tom Chandler)
>>   - RS panic with deep stacktrace reported by Victor van der Veen -
>> this is tty panicing on "kbc_read failed to complete". the
>> 'solution'
>> is to try harder ;) (which works on test hardware I have access to)
>>
>> Testers welcome.
>
> Why is this only 14 MB while previous releases are well over 400?  I presume it is because this installs a "bare minimum" system only.  If so, how to install the rest of the system?

It doesn't include packages. You can install packages from the network
using packman or to download them manually.


--
Evgeniy Ivanov

[Ben Gras]

Hi Antoine,

For you I'd made: http://www.few.vu.nl/~beng/minix3_1_7_ide_copy.iso.bz2
to see if increasing various priv limits really solves your problem.

[Larson, Timothy E.]

> It doesn't include packages. You can install packages from the
> network
> using packman or to download them manually.

Ah, great. I always install packages from the network anyway...

I tried r6875 today. It booted on my Dell GX60 (hooray!) but the file copy process hung at include/bzlib.h with 7868 files remaining. After a short delay I heard the CD and HD spin up and was hopeful it was going to resume, but then they spun down again.

Tim
--
Tim Larson
App Admin, Intercall Production Services

[Ben Gras]

> I tried r6875 today.  It booted on my Dell GX60 (hooray!) but the file copy process hung at include/bzlib.h with 7868 files remaining.  After a short delay I heard the CD and HD spin up and was hopeful it was going to resume, but then they spun down again.

Do you think this is a minix bug? If so, I'd like you to substantiate
that by confirming an older version does install, and another copy of
the current version still doesn't, if possible.

(Interesting experiment: set atapi_dma=1 in the boot monitor before
installing to enable atapi dma, i used this as a workaround for
(different) installation problems in virtualbox at one point.)

[Larson, Timothy E.]

> > I tried r6875 today. It booted on my Dell GX60 (hooray!) but the
> file copy process hung at include/bzlib.h with 7868 files remaining.
> After a short delay I heard the CD and HD spin up and was hopeful
> it was going to resume, but then they spun down again.
>
> Do you think this is a minix bug? If so, I'd like you to
> substantiate
> that by confirming an older version does install, and another copy
> of
> the current version still doesn't, if possible.

This is the same hardware that I ran 3.1.5 on, but couldn't install 3.1.6 on.

My CD verified after burning, but I can try another.

> (Interesting experiment: set atapi_dma=1 in the boot monitor before
> installing to enable atapi dma, i used this as a workaround for
> (different) installation problems in virtualbox at one point.)

I'll give this a try, too.

Tim
--
Tim Larson
App Admin, Intercall Production Services