Re: Netboot E250 -- tftp problem

2008-01-22 Thread Geert Stappers
Op 22-01-2008 om 11:01 schreef Hero_xbd!.RRR:
 Geert Stappers wrote:
 My advise is to check the version the tftpd server.
 Over here says `dpkg -l tftpd-hpa` that I'm using version 0.48-1.

 For reference you may wireshark 
 http://www.stappers.nl/gst/temporary/e250/rodebes.pcap
 Rodebes is the hostname of an Ultra5 at address 172.24.0.15

 Greet,

Great, is is actual Geert

By the way, I prefer 'stappers', it goes back to the mid 1990
when I on the same mailinglist as Geert Uyterhoeven.
With 'stappers' it is unambiguous which Geert is meant.
Stappers is also my IRC nickname and Debian account name.

 I have followed your instructions, and the memory seems to be alright. (it 
 fills 55 and aa as expected)
 I have removed -r blocksize and it gives no difference.
 (Maybe I should try out inetd? Almost all the documents around the net 
 netbooting sparc use inetd instead of standalone tftpd.)

 I have also checked the version:

 Openboot 3.12
 tftpd-hpa 0.48-1

 I compared the .pcap file with yours, and it seems there is no difference 
 except for the check sum error. (Wireshark said 'maybe caused by UDP 
 checksum offload? ' -- what does that mean? Can it give us some clue?)

Over here I can't reproduce the error when the TFTP server is standalone.
Still it is a good thing try how the TFTP server behaves under inetd.

Because netbooting SUNs is a working thing, I'm curious what blocks it
this time. Could it be firewall software on the TFTP server?


Cheers
Geert Stappers
-- 
Bonus points for respecting the discussion order
(in other words: Reply _below_ the text ... )


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Hero_xbd!.RRR

Geert Stappers wrote:

By the way, I prefer 'stappers', it goes back to the mid 1990
when I on the same mailinglist as Geert Uyterhoeven.
With 'stappers' it is unambiguous which Geert is meant.
Stappers is also my IRC nickname and Debian account name.

  
OK, stappers. I'm sorry that I don't quite familiar with Latin names. I 
will be careful.



Because netbooting SUNs is a working thing, I'm curious what blocks it
this time.

Me too.

 Could it be firewall software on the TFTP server?
  


I am using fail2ban[1] on the TFTP server, which functions on iptables 
and is supposed to modify only ftp ftps ftp-data ftps-data sshports. 
Except it, I did not do anything with firewall.


Oh, maybe I should use a shiny minimum debian mechine to act as the TFTP 
server. That will eliminate various potential blocks.


Is it possible the NICs and Net Switch produce the problem? Does the 
working rarp session eliminate all the possibility?


Stappers, I have noticed that your .pcap file also has one packet with 
checksum error. Does it infer that checksum error is in some degree 
irrelevant?


1. http://www.fail2ban.org

--
续本达
Xu, Benda

Fundamental Science of Mathematics and Physics,
Tsinghua University, Beijing, P.R.China

http://learn.tsinghua.edu.cn:8080/2005012177/index.html
http://thirsty.phys.tsinghua.edu.cn/MyWiki/heroxbd



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Geert Stappers
Op 22-01-2008 om 21:31 schreef Hero_xbd!.RRR:
 Geert Stappers wrote:
  Could it be firewall software on the TFTP server?

 I am using fail2ban[1] on the TFTP server, which functions on iptables and 
 is supposed to modify only ftp ftps ftp-data ftps-data sshports. Except 
 it, I did not do anything with firewall.

Right now, we don't known what is blocking us,
we have to put fail2ban on the list of suspects ...

 Oh, maybe I should use a shiny minimum debian mechine to act as the TFTP 
 server. That will eliminate various potential blocks.

I think that is a very good option.

 Is it possible the NICs and Net Switch produce the problem?

Unlikely, but put it the bottom of the list with suspects.

 Does the working rarp session eliminate all the possibility?

The RARP session was indeed succesfull.
But RARP is on a different network layer as TFTP is.
So it could indeed that a network component like a switch,
does screw TFTP, but doesn't mangle RARP

 Stappers, I have noticed that your .pcap file also has one packet with 
 checksum error. Does it infer that checksum error is in some degree 
 irrelevant?

Rechecking http://www.stappers.nl/gst/temporary/e250/rodebes.pcap
reveals indeed a checksum error. However it is the last TFTP block of the
file. The block size is 28, not the regular 512.
The download is succesfull, the program works fine and there
is surely not 'Data access error'.

I just did `ls  /var/lib/tftpboot/AC18000F` to get less then 512 bytes
in the file that my Sparc (at address 172.24.0.15) downloads with
the 'ok' prompt command `load net`.

The single TFTP packet had according wireshark also a checksum error,
but the transfer was fine. ( checked with `4000 200 55 fill`, `load net`
and `4000 200 dump` )



 1. http://www.fail2ban.org


Cheers
Geert Stappers

P.S.

About latin names:
It is indeed common to use the first name,
it is just me who choose an uncommon family name as nick name.


P.P.S.

I'm subscribed to the mailinglist


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Martin
On Tue, 2008-01-22 at 21:31 +0800, Hero_xbd!.RRR wrote:
snip

 I am using fail2ban[1] on the TFTP server, which functions on iptables 
 and is supposed to modify only ftp ftps ftp-data ftps-data sshports. 
 Except it, I did not do anything with firewall.
Apologies if I'm telling you something you already know but that would
mean your TFTP server isn't protected.

 
 Oh, maybe I should use a shiny minimum debian mechine to act as the TFTP 
 server. That will eliminate various potential blocks.
 
 Is it possible the NICs and Net Switch produce the problem? Does the 
 working rarp session eliminate all the possibility?
This seems vaugely familiar.  I have a vauge memory of having problems
getting a Sun machine to netboot and finding out it was a cabling
problem.  Have you tried swaping the Sun machine for another boxen and
pulling the netboot file from the server?  This should tell you if it's
a cabling / switch problem.

HTH

Cheers,
 - Martin



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Daniel Smolik

Hero_xbd!.RRR napsal(a):

Hi, everybody!

I am trying to install a debian-sparc system on an old Enterprise 
250(E250) in our lab.


The CD-ROM on E250 is broken, so I need to do netboot and installation.

The problem is when netboot, there are problem with tftp downloading.

What I have done:

1. Setup a netboot server(Debian Sid i386), with rarpd and tftpd-hpa.
1.1. Edit /etc/ether and start rarpd.
1.2. Start tftpd-hpa(standalone, without inetd) with -l -s /tftpboot -r 
blksize option(edited in /etc/default/tftpd-hpa).

1.3. Getting a file from tftp server at another workstation works.

2. Turn on E250(in the same LAN as netboot server), Press Stop-A to get 
the OPENBOOT prompt, issueboot net, it reads that:

Boot device:/[EMAIL PROTECTED],4000/[EMAIL PROTECTED],1 File and Args:
Time out waiting for ARP/RARP packet
4200 Data Access Error
and throw me to the ok prompt again.

3.1 To solve the problem, I did net-watch-all in E250's OPENBOOT 
prompt, it returned no error.
3.2 I used Wireshark(aka. ethereal) to sniffer network packets, It shows 
that E250 has issued a RARP query, and netboot server has respond one 
then E250 has got an IP address(But when I ping from netboot server to 
E250, it gives no response). Then E250 has queried an image file named 
according to its newly received IP address(hexadecimal, I can also see 
in tftp's log this request), but when netboot server respond with the 
tftp packet, wireshark shows the udp packet encapsuring tftp data has 
udp checksum error


4. The hardware of netboot server is:

CPU: Intel Cereron 1.7 Ghz
Chipset: Intel i845G
NIC: Realtek 8139



I don't know what to do now, I have tried everything I know and 
everything I can find on the net, but still stack with this problem. 
Please give me some suggestions.


Thanks in advance!

I don't know if my advice help you, but I use netboot on E250 and E450 
and Ultra 20 and Ultra 5 many times without problem.But first thing that 
I do is upgrading OPB to latest version and switch to DHCP :-)


I download OPB firmware from sunsolve put it on tftp server and boot it 
to upgrade OPB it works great and I use it.


Dan






--
Mydatex s r.o.
http://www.mydatex.cz
email: [EMAIL PROTECTED]
mob: 604200362
tel: 226210085


--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



netboot E250

2008-01-22 Thread Eric Rapilly
Sorry, but it'snt an I386 that you should install : I386 is an INTEL 
processor, meanwhile on a sun it's a SPARC processor, a RISC processor .



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Jachym Cepicky
Hi,
Hero_xbd!.RRR píše v Po 21. 01. 2008 v 15:43 +0800:


 3.2 I used Wireshark(aka. ethereal) to sniffer network packets, It shows 
 that E250 has issued a RARP query, and netboot server has respond one 
 then E250 has got an IP address(But when I ping from netboot server to 
 E250, it gives no response). Then E250 has queried an image file named 
 according to its newly received IP address(hexadecimal, I can also see 
 in tftp's log this request), but when netboot server respond with the 
 tftp packet, wireshark shows the udp packet encapsuring tftp data has 
 udp checksum error
 

you have similar problems, I had to face. I had to (re)start the tftp
server right after first tftp request from the sun workstation. Then
everything worked well.

It is truth, that this is not usable, if you want to boot from network
regularly, but for installation it was enough.

Good luck

Jachym
-- 
Jachym Cepicky
e-mail: [EMAIL PROTECTED]
URL: http://les-ejk.cz
GPG: http://www.les-ejk.cz/pgp/jachym_cepicky-gpg.pub



signature.asc
Description: Toto je digitálně	 podepsaná část	 zprávy


Re: netboot E250

2008-01-22 Thread Hero_xbd!.RRR

Eric Rapilly wrote:
Sorry, but it'snt an I386 that you should install : I386 is an INTEL 
processor, meanwhile on a sun it's a SPARC processor, a RISC 
processor .

Eric, I am using the sparc64 netboot image found here:

http://ftp.se.debian.org/debian/dists/testing/main/installer-sparc/current/images/sparc64/netboot/2.6/boot.img

(I guess this is just the same as stable netboot image.)

I guess it is the right file for Sun4u (UltraSPARC II 300Mhz) architecture.

But I have not gone that far. Right now no file can be transfered to the 
Sparc box through TFTP without error, whether an i386 or a sparc netboot 
image.


Cheers,

--
续本达
Xu, Benda

Fundamental Science of Mathematics and Physics,
Tsinghua University, Beijing, P.R.China

http://learn.tsinghua.edu.cn:8080/2005012177/index.html
http://thirsty.phys.tsinghua.edu.cn/MyWiki/heroxbd



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Hero_xbd!.RRR

Martin wrote:

Apologies if I'm telling you something you already know but that would
mean your TFTP server isn't protected.
  

But does that metter? We just want to get the SUN box to netboot, don't we?

This seems vaugely familiar.  I have a vauge memory of having problems
getting a Sun machine to netboot and finding out it was a cabling
problem.  Have you tried swaping the Sun machine for another boxen and
pulling the netboot file from the server?  This should tell you if it's
a cabling / switch problem.
  
I have successfully pulled the files through TFTP to my PC in dormitory 
(TFTP server is at lab). Does it infer there is no problem here?


--
续本达
Xu, Benda

Fundamental Science of Mathematics and Physics,
Tsinghua University, Beijing, P.R.China

http://learn.tsinghua.edu.cn:8080/2005012177/index.html
http://thirsty.phys.tsinghua.edu.cn/MyWiki/heroxbd



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread jacob
On Wed, Jan 23, 2008 at 09:38:30AM +0800, Hero_xbd!.RRR wrote:
 Martin wrote:
 But does that metter? We just want to get the SUN box to netboot, don't we?
 This seems vaugely familiar.  I have a vauge memory of having problems
 getting a Sun machine to netboot and finding out it was a cabling
 problem.  Have you tried swaping the Sun machine for another boxen and
 pulling the netboot file from the server?  This should tell you if it's
 a cabling / switch problem.
   
 I have successfully pulled the files through TFTP to my PC in dormitory  
 (TFTP server is at lab). Does it infer there is no problem here?


Are you using tftpd-hpa? I know that with MIPS, IIRC, other tftp servers
didn't work. I have encoutered that problem before. I seem to recall
some discussion (possibly on debian-mips) about explicitly recommending
tftpd-hpa in the installation manual, but it doesn't seem to.

Jacob


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Hero_xbd!.RRR

Martin wrote:

On Wed, 2008-01-23 at 09:38 +0800, Hero_xbd!.RRR wrote:

  

This seems vaugely familiar.  I have a vauge memory of having problems
getting a Sun machine to netboot and finding out it was a cabling
problem.  Have you tried swaping the Sun machine for another boxen and
pulling the netboot file from the server?  This should tell you if it's
a cabling / switch problem.
  
  
I have successfully pulled the files through TFTP to my PC in dormitory 
(TFTP server is at lab). Does it infer there is no problem here?



That depends on the architecture of the network.  If it's something like
this:


PC in dorm -- ... some network ... -- switch -- server
 ^
 |
 |
 V
E250

(Where arrows are ethernet cables)

then you will have tested all but the cable that goes from the switch to
the E250.  If your topology is different then what you have tested will
be different.
  
Hmm, I think I can interchange the cable of E250 and server to test(The 
Switch is not accessible to me now)

PS If you get really stuck netbooting, you could pull the disks, dump a
minimal filesystem on them and then boot from disk.
  
E250 is using an old scsi interface. Here in my university you can 
rarely see scsi, and if there is one, it is of new interface...


I myself is not fimilar with scsi, I do not know the specification of 
the bus(I know they're different, one is old one is new)


Oh, perhaps I can have an IDE PCI card on E250 use an IDE HDD to boot. 
While I am still curious about what is blocking me from netbooting...


--
续本达
Xu, Benda

Fundamental Science of Mathematics and Physics,
Tsinghua University, Beijing, P.R.China

http://learn.tsinghua.edu.cn:8080/2005012177/index.html
http://thirsty.phys.tsinghua.edu.cn/MyWiki/heroxbd



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Hero_xbd!.RRR

Geert Stappers wrote:

But the '4200 Data Access Error' is a scary one ...

4000 is the default download start address.
4200 is start plus 200. 0x200 is 512 decimal, the size of a default TFTP packet.

How to classify 'Data Access Error' is the challenge.

If there are problems with memory, the 'ok' wouldn't probably not been
show. Just do a simple check by these five separate command at the 'ok'
prompt:
 4 400 dump
 4 400 55 fill
 4 400 dump
 4 400 AA fill
 4 400 dump
to make sure it is really no memory problem.
  

Hi, I have got a great break through! It seems to be a memory problem!

I have replaced NIC Realtck 8139 with Intel e100, and there is no more 
checksum error now. But it still hang with the second TFTP packet, 
leaving me:


ok 4200 Data Access Error

But, when I do

ok 4000 400 aa fill

and then

ok boot net

The TFTP packets hang with the third packet now! I can't believe it!

So I went on and got perfectly to packet 146 or so. I am going tired so 
I just did


ok 4000 a0 aa fill

Then the Openboot hangs!

I power it on again, and did

ok test-all

No error.

Our E250 have 1024MB memory, far more than enough. What seems to be the 
problem now?


I am copying the Openboot info (after power on) below:

Sun (TM) Enterprise 250 (2 X UltraSPARC-II 296MHz), Keyboard Present

OpenBoot 3.12, 1024 MB memory installed, Serial #11548348
Ethernet address 8:0:20:b0:36:bc, Host ID: 80b036bc.

--
续本达
Xu, Benda

Fundamental Science of Mathematics and Physics,
Tsinghua University, Beijing, P.R.China

http://learn.tsinghua.edu.cn:8080/2005012177/index.html
http://thirsty.phys.tsinghua.edu.cn/MyWiki/heroxbd



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: Netboot E250 -- tftp problem

2008-01-22 Thread Hero_xbd!.RRR
Hero_xbd!.RRR wrote: 

Hi, I have got a great break through! It seems to be a memory problem!

I have replaced NIC Realtck 8139 with Intel e100, and there is no more 
checksum error now. But it still hang with the second TFTP packet, 
leaving me:


ok 4200 Data Access Error

...

I have tried again as follows:

ok boot net

4400 Data Access Error

ok dump 4000 600

0  Fast Data Access MMU Miss

I think this MMU Miss message is an important clue.

--
续本达
Xu, Benda

Fundamental Science of Mathematics and Physics,
Tsinghua University, Beijing, P.R.China

http://learn.tsinghua.edu.cn:8080/2005012177/index.html
http://thirsty.phys.tsinghua.edu.cn/MyWiki/heroxbd



--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]