[9fans] Plan9 Installation

[9fans] Plan9 Installation

[mark_otto]

I am trying to install the initial 4 PC disk that I downloaded from 
plan9.bell-labs.com/plan9/pcdist.  I have done the first boot from a: but I 
get the following error when I do the file system installation from 
c:\plan9\b:

hd0: LBA 63463668 sectors, -1045626880 bytes 
.bad magic 0xb7600 not a plan9 executable 
Boot devices: hd!0 h!0
boot from:

When I type in hd!0, the boot from message returns.  When I type h!0 or 
h!0!/plan9/9dos, I get the ".bad magic 0xb7600 not a plan9 executable" 
message again.

I get the same error message with the basic plan9.ini in the "Read this 
section if you have trouble installing on the PC" section and with 
vgasize=1024x768x8 and ata1=irq=15.  The following is my basic plan9.ini:

mouseport=ps2
monitor=vga
vgasize=640x480x1
bootfile=hd!0!/plan9/9dos
rootdir=/plan9
bootdisk=#H/hd0disk:63

I know I have an unsupported PC, a Micon PC with a 199MHz Pentium using a 
Stealth 3D 2000 Pro, Diamond Multimedia Systems, Inc. video card.  I also 
know from the 9fans listserve digest and the /plan9/update directory at 
bell-labs that there are boddles for the S3 Virge Chipset if I can get the 
system running on another PC.  I am willing to find one of our older 486 
PCs with the [E]ISA bus, but I thought since I got this far in the 
installation that the bus and the video card were not the problem.  I want 
to see if I can use the basic system, connect to the network and our Sun 
unix boxes, and use our HPIIsi printer before I buy the CD rom and face our 
head of computer support who rather we only use Windows.

Should I give up and try this on supported PC or is there a way around this 
problem.  I am enclosing information on my PC below.  Thank you in advance 
for your help.

Mark Otto, Biological Statistician         work:    301-497-5872     
Population and Habitat Assessment Section  fax:     301-497-5871     
Office of Migratory Bird Management        e-mail:  Mark_Otto@FWS.Gov
U.S. Fish and Wildlife Service
Department of the Interior
11500 American Holly Drive
Laurel, Maryland  20708-4016

---------------------------------------
Here is the set up I have:

Motherboard:   Micron
CPU:           Intel 199mhz Pentium
Memory:        32MB
Disks:         2GB IDE Hard Drive partitioned into c: and d: drives, 1GB
               each (d: is empty in order to hold plan9) If this works, 
               I will get a SCSI drive for the fileserver.  The primary 
               IDE is IRQ=14 Port=01f0h and the secondary is IRQ=15 
               Port=01f0h  
Zip Drive:     IOmega Zip 100, ATAPI
CDROM:         Hitachi CRR-8330 Standard CD-ROM Device 
Sound Board:   Yamaha OPL3-SAx, ATAPI
Network:       3c905 Fast EtherLink card, IRQ=10 Port=0180h 
Mouse:         PS2 IRQ=10
Keyboard:      MS Natural Keyboard, IRQ=01 Port=0060h
Monitor:       Micron 700 FGx, Horizotal refresh=30-69KHz and vertical=50-160KHz
Video Card:    Stealth 3D 2000 Pro, Diamond Multimedia Systems, Inc, S3 Virge, 
               4MB video memory, IRQ=09 Port=03B0h

Re: [9fans] Plan9 installation

[vdharani]

Russ,

thanks for the reply.

>> making
>> bootsetup do its job? what is more, it doesnt do boot floppy creation
>> also.
>
> maybe a bug has been introduced.  people have been working
> on the installer recently.  installing plan 9 onto a disk already
> containing plan 9 is a tricky case and not often tested.

can we track and try to solve the issue? when something fails, people tend
to reinstall and the problem with bootsetup will panic new or even
experienced people. is there any reliable way of getting rid of the
problem? like zero-filling some area?

> venti is slow for reasons other than bad sector offsets.
> i doubt very much that making the sector offsets multiples of 16 matters,
> since cylinder boundaries are so often odd anyway.  the suggestion (which
> i think is my fault) was likely naive paranoia on my part.
in that case, can i remove the lines that talk about this alignment in
Wiki? i think it will mislead, otherwise.

> there should be a new venti by the end of the summer.  i have it running
> a linux system at mit.  i'm going to convert my home venti server over the
> summer and hopefully we'll convert the bell labs venti server too.
> the new venti code is at least 10x faster.
thats great!

>> - after i started plan9, i just did "fsys main snap" in fossil console.
>> it
>> went on for many hours (btw, i havent used fossil before). then i gave
>> the
>> same command again (thinking this time it is going to just come back
>> saying "done"). but it is going on and on. what is strange, i even
>> rebooted the machine. still 'stats' says system is busy. i am wondering
>> what it is doing. any idea?
>
> it's probably saving the file system to venti.  this can take a while, as
> we discussed on 9fans last month (look for a post by vester thacker).
i will check it up.

thanks
dharani

[9fans] Plan9 installation

[vdharani]

Hi,

I was trying to install plan9 this weekend and ran into lot of poblems. I
got many little questions in this exercise. I have listed the
questions/issues below. I will be very happy if you can give your
comments/answers.

- first, where can i get the latest cd image? can i get the latest one in
http://cm.bell-labs.com/magic/9down4e/ureg?haveread=yes&exportok=yes
or
"download the distribution archive" link of download section? or is that
both have the latest nightly build cd image?

- bootsetup seems to get confused many times. i tried with CDROM image and
bootsetup reported errors like "null concatenation" or didnt install 9fat
contents properly or install plan9.ini config lines, etc. everytime i had
one problem or the other. somehow if 9fat already exists or the disk has
some previous data, bootsetup gets confused. it is difficult to predict
what exactly is happening. what is the reliable, predictable way of making
bootsetup do its job? what is more, it doesnt do boot floppy creation
also.

- i tried with another hard disk i had and then i was able to proceed with
fossil+venti installation. it takes a long time but atleast i could
install plan9. i read that venti can work faster if arenas partition
sector offsets are multiples of 16. given that venti attempts to compress
8k blocks, how good is this facility? will it still be beneficial? and
what about fossil? will it be good if it can be on a similar boundary?

- a small routine could put arenas parition (and fossil, if needed) at the
facourable boundary in case of automatic paritioning. can we add the code
as needed to avoid manual work?

- the floppy disk based installation seems to support only fossil (no
venti). is it because the code needed doesnt fit into the floppy disk?

- after i started plan9, i just did "fsys main snap" in fossil console. it
went on for many hours (btw, i havent used fossil before). then i gave the
same command again (thinking this time it is going to just come back
saying "done"). but it is going on and on. what is strange, i even
rebooted the machine. still 'stats' says system is busy. i am wondering
what it is doing. any idea?

any help appreciated.

regards
dharani

Re: [9fans] Plan9 installation

[Russ Cox]

> - first, where can i get the latest cd image? can i get the latest one in
> http://cm.bell-labs.com/magic/9down4e/ureg?haveread=yes&exportok=yes
> or
> "download the distribution archive" link of download section? or is that
> both have the latest nightly build cd image?

they're the same file and both the latest nightly build cd.

> - bootsetup seems to get confused many times. i tried with CDROM image and
> bootsetup reported errors like "null concatenation" or didnt install 9fat
> contents properly or install plan9.ini config lines, etc. everytime i had
> one problem or the other. somehow if 9fat already exists or the disk has
> some previous data, bootsetup gets confused. it is difficult to predict
> what exactly is happening. what is the reliable, predictable way of making
> bootsetup do its job? what is more, it doesnt do boot floppy creation
> also.

maybe a bug has been introduced.  people have been working
on the installer recently.  installing plan 9 onto a disk already
containing plan 9 is a tricky case and not often tested.

> - i tried with another hard disk i had and then i was able to proceed with
> fossil+venti installation. it takes a long time but atleast i could
> install plan9. i read that venti can work faster if arenas partition
> sector offsets are multiples of 16. given that venti attempts to compress
> 8k blocks, how good is this facility? will it still be beneficial? and
> what about fossil? will it be good if it can be on a similar boundary?

venti is slow for reasons other than bad sector offsets.
i doubt very much that making the sector offsets multiples of 16 matters,
since cylinder boundaries are so often odd anyway.  the suggestion (which
i think is my fault) was likely naive paranoia on my part.

there should be a new venti by the end of the summer.  i have it running
a linux system at mit.  i'm going to convert my home venti server over the
summer and hopefully we'll convert the bell labs venti server too.
the new venti code is at least 10x faster.

> - a small routine could put arenas parition (and fossil, if needed) at the
> facourable boundary in case of automatic paritioning. can we add the code
> as needed to avoid manual work?

i don't think the boundary matters.

> - the floppy disk based installation seems to support only fossil (no
> venti). is it because the code needed doesnt fit into the floppy disk?

yes.

> - after i started plan9, i just did "fsys main snap" in fossil console. it
> went on for many hours (btw, i havent used fossil before). then i gave the
> same command again (thinking this time it is going to just come back
> saying "done"). but it is going on and on. what is strange, i even
> rebooted the machine. still 'stats' says system is busy. i am wondering
> what it is doing. any idea?

it's probably saving the file system to venti.  this can take a while, as
we discussed on 9fans last month (look for a post by vester thacker).

russ