[Hpipaq21x-devel] boot observations

[Giovino, Jeffrey D.]

Again, sorry for potential duplicate questions,
still working on getting on the mail list...


-----Original Message-----
From: Oliver Ford [mailto:ipaqlinux at oliford.co.uk] 
Sent: Wednesday, March 24, 2010 5:08 AM
To: Giovino, Jeffrey D.
Cc: hpipaq21x-devel at linuxtogo.org
Subject: Re: [Hpipaq21x-devel] boot observations

Giovino, Jeffrey D. wrote:
> Thank you for the courteous reply. 
> It was refreshing to see somebody actually consider the topic before outright dismissing it!
>   
I'm an experimental scientist by profession, I'm used to trusting 
observations over my existing beliefs.
... you obviously do not support any gov't contracts :)


> Can you try again and make a copy of the partition table (512 bytes from 
> the start of the overall device), e.g.
> "dd if=/dev/sdb of=mbr.bin bs=512 count=1" but replace /dev/sdb with the 
> device (like /dev/sdb without any number, not like /dev/sdb1)
> before and after writing the tarball to the ext2 partition.
>
> Hmm, very odd one that.
>
> ... I will put this on my todo list. Will need to ramp up on dd however first.
> ... is this an experiment or a fix?
>   
Experiment, I'm trying to figure out what could possibly be changing to 
upset the boot ROM.


> ... following up on your thoughts above:
> 	1) fdisk - Format one large SD for 1 cyl (255x63) fat16 and remainder ext2
> 	2) mkfs.ext2
> 	3) un-tarball rootfs
>  	4) fdisk for only partition one with dormant ext2 filesystem
> 	5) have u-boot or the kernel modify the partition table and add back the ext2 second partition.
> 	6) this won't survive a reboot, but might provide insight.
>   
Might work, but it really shouldn't be necessary.

> ... I also noticed that when formatting the SD card it 
> ... was necessary to run mkfs.ext2 -L <ANGSTROM? /dev/db2;
> ... BUT running mkfs.msdos [-F 16] -n BOOT /dev/sdb1, although 
> ... intuitively correct, caused it to fail at boot time.
>   
I don't follow, are you saying the volume labels matter, or am I missing 
something?
... I am saying it worked without mkfs.msdos and failed with it (counterintuitively.)
... I should not have polluted the point with the labels

> ... the wiki could use a step by step fdisk/mkfs.ext2 procedure as well; I would be happy to provide
>   
Yes, that'd help a lot.

> Yup, that's a good procedure a few other people have had to use before. 
> If you have time, could you (or anyone else who feels like volunteering) 
> write something about that on the wiki?
> ... I would be glad to. Please point me in the right direction; Or should I simply send you a html/text file?
>
>   
Either add it to this one:
http://www.handhelds.org/moin/moin.cgi/HpIpaq214HowTo

or make a new page, something like 'HpIpaq214FAQ' or 
'HPIpaq214Troubleshooting'.

You have to sign up to the wiki to edit, but anyone can do it.
... I will give it a try.
... I have been keeping running notes and will post when I get my head above water.

>
>   
> Not sure exactly what you mean they do. Some people often do stuff with 
> loading a small rootfs to start with that has all the kernel modules 
> which can then access the big partitions. They do it to keep the kernel 
> small enough to store in flash. We don't have that problem because the 
> kernel contains all the drivers (because I never got the hang of working 
> with modules) and can be as big as we want since U-boot takes it from 
> the FAT16 partition.
> ... this topic addresses the ability to boot from large SD cards.
> ... this device mentioned can boot at least 2G SD cards that I have tested.
> ... here is a link to what I was referring to:
>  http://groups.google.com/group/beagleboard/msg/18b3c91e659ea206
>   
That seems to be because U-boot requires their RAM to be setup so that 
have a separate piece of code to do that first. Actually, the ipaq is 
similar since there is already a piece of code that runs before U-boot 
called the OBM (OEM boot module). It is that code which tests for the 
0-1 buttons and boots from the MMC card accordingly.
> ... this might have something to do with topic one above
> ... Is the dd procedure you mentioned already the way to get large SD cards to boot?
>
>   
No, that was just to make a copy of the SD-cards MBR to see what is 
happening.
> ... I am working toward building my application from source using your Angstrom hpipaq21x and the gnu-eabi-gcc
> ... I have already ran a hello world compiled with this compiler from another arm device
>   
OK, I'll be interested to here what you're building and how far you get.
... porting an existing C-openGL based cockpit display.
... I will be opening up to get to the serial pins for that.
... or possibly using a serial CF card if its magically supported by Angtrom
... Do you know if the serial signal is -5v or -12V EIA232 or are they 3.3ttl that needs a MAX232 conversion.
... The next real hurdle is to get the pxa310 egl 1.0 driver and libraries going
... or running open GL through X11, which I am afraid will be very cumbersome)
... currently I have ported the app to PSP, iPhone, beagleboard, and Android on Nexus One, and cygwin

If you're thinking of doing any serious embedded development though, 
It's definitely worth the time installing and getting used to 
openEmbedded. It also lets you edit all the existing apps.
... will do. I will keep notes and post as well.
... is MACHINE=ipaq214 recognized by the open embedded tree or only your git tree?

... Ever consider building Android? 
... Lots of java based app out there that would run w/o any HW dependancies.


> The term 'xscale' refers to a huge range of processors. It's possible 
> that their advert refers to a much higher spec model, or a software 
> openGL that uses the main CPU optimisations. If you find anything out, 
> let us know!
> ... please let me know if this is the processor:
> ... http://www.marvell.com/products/processors/applications/pxa_3xx/PXA310_PB_R4.pdf
> ... see page 2 para 1: 'ecosystem'
>   
Yes, thats the one. It definitely  does have the graphics acceleration 
and both windows and that directFB driver use it. I'll keep pushing my 
stuff up to date with the mainline kernel every few months so maybe 
sometime later this year it should just magically turn up. Still needs 
an X11 server for it though, unless someone wants to try switching us to 
a distro based on directFB rather than X11?
... how much of the ~80M do you think the X11 package is responsible for?



... new Q: I noticed in your march 2010 build wifi-up calls udhcpd instead of udhcpc.
... are there any low level diagnostics to test with.
... will be keeping notes on this as well, cant get this to work for WEP:
	/etc/wpa_supplicant.conf
ctrl_interface=DIR=/var/run/wpa_supplicant GROUP=users

network={
	ssid="netName"
	scaen_ssid=1
	key_mgmt=NONE
	wep_key0=nnnnnnnnnn
	wep_tx_keyidx=0
}