Htc Universal, ... running Debian EABI Options

[IIZIIm]

I got so far as to pass memsize=0x8000000 to haret before booting linux, but still 64M. Haven't found any ref of running linux on the Universal with 128M.

I used to have an entry mem=64M in startup.txt, maybe you just try mem=128M like in Blueangel"s startup.txt :http://www.linuxtogo.org/~htcpxa/htcblueangel/GPEphone/startup.txt

Can you enter cat /proc/meminfo?

[ShiroiKuma]

I think mem=128M should do it, actually...

There's something wrong with the upgrade, as the 128M I was seeing before is flash size not RAM, so something's not right... I talked with SRS and am sending the unit back in today for them to check what's wrong.

[ShiroiKuma]

SRS confirmed they flashed the wrong ROM onto the device and that's why the 128MB wasn't working. Should be fine now. I should be receiving it beginning next week, and will report how linux works with 128MB. Hopefully, using 128MB should be no prob under linux, however no one has confirmed it still on the Universal, so hopefully there will be no more unwanted surprises.

[ShiroiKuma]

So I got the Universal back, with the correct ROM reflashed. It has 128MB memory now, works much better. Linux starts fine, without the mem=128M argument added, no changes to startup.txt necessary, finds 128MB of memory. X works great without swap now, it's definitely much better.

One problem though with suspend. When you suspend, there is no way to wake the Universal up, it doesn't respond to the power key. I saw a mention somewhere, specifically focusing on this, i.e. that expanding to 128MB, you will need to modify something, so it doesn't have the suspend-resume problem. Am trying to find the reference now, maybe it was on xda-developers. If anyone has any hints, I'd appreciate the help.

[ShiroiKuma]

OK, just found the reference on xda-developers. It's a comment by cr2, reading:

CODE

The only problem will pop up at resume, because the
HTC IPL sets the PXA memory controller as if there were 64MB RAM.
But it can be easily fixed by resetting them in the CPU resume function.

Anyone know how to do this?

[ShiroiKuma]

So, I have run various tests, but can't get the device to resume from suspend. Allthough cr2 in the above post says it's easy to fix, I don't know how to do this, if it entails editing the IPL or what. So unless we figure this out, currently 128MB is unuseable on the Universal for Linux.

The device resumes properly, when I start haret with mem=64M so am using that for now... (

[Thorsten]

OK, just found the reference on xda-developers. It's a comment by cr2, reading:

CODE

The only problem will pop up at resume, because the
HTC IPL sets the PXA memory controller as if there were 64MB RAM.
But it can be easily fixed by resetting them in the CPU resume function.

Anyone know how to do this?

I lack a bit of the context, but I suspect he talks about a fix in the linux kernel. Can you post the link?

Can you write something about the positive effect of the upgrade? My primary problem wit my universal is that is is horrible slow on linux. For my use case (remote linux server administration, nagios) I need VPN, ssh and a webbrowser (grafical, not lynx). But already opening a terminal window is very slow, a webbrowser is only usable with lots of passion...

Can you tell me how much speed improvements you got?

Are there other positive effects?

Does any body know if the kernel is still in active development? I can only find a quite old 2.6.21-hh20.

Greetings

Thorsten

[ShiroiKuma]

I lack a bit of the context, but I suspect he talks about a fix in the linux kernel. Can you post the link?

The link is http://forum.xda-developers.com/showpost.p...mp;postcount=95

I've inspected the htc-universal specifics of the linux kernel, and couldn't identify anything. I think he meant editing the IPL, which is the phone bootloader, which however seems strange, since if that was the prob, the phone wouldn't come out of suspend under WM, or maybe WM doesn't really do a deep suspend like Linux does.

Can you write something about the positive effect of the upgrade? My primary problem wit my universal is that is is horrible slow on linux. For my use case (remote linux server administration, nagios) I need VPN, ssh and a webbrowser (grafical, not lynx). But already opening a terminal window is very slow, a webbrowser is only usable with lots of passion...

Not much I can say, since I need to use the phone and without suspend it's useless. So I'm not using the 128MB now, until I figure out how to make it resume...

[Thorsten]

Hi, I contacted cr2, and he confirmed the need of a kernel patch. I hope I find some time in the next days to have a look on it.

greetings

Thorsten

[ShiroiKuma]

In fact, I got a response from him at the same time. He requested feedback to haret's "dump pxa27x" command. I'll be posting soon. Hopefully this'll enable us to use these 128...

[ShiroiKuma]

I've just posted the output of the command cr2 requested at http://forum.xda-developers.com/showpost.p...;postcount=2085

[ShiroiKuma]

Hi, I contacted cr2, and he confirmed the need of a kernel patch. I hope I find some time in the next days to have a look on it.

greetings

Thorsten

Do you have an idea of what and where needs to be patched? Over at xda-developers, cr2 hinted that haret's dump pxa27x command output should be compared between a 128MB and a 64MB Universal, to give an idea of the patch presumably. I've posted the 128MB Universal output at http://forum.xda-developers.com/showpost.p...;postcount=2088

However, I don't have the 64MB output. Could you post it? Or could someone else post the haret's dump pxa27x command's output here? Instructions on how to capture the output easily on a Universal I've posted at: http://sumoudou.org/index.php?cmd=%E7%9B%B...PC%20connection

[Thorsten]

Do you have an idea of what and where needs to be patched? Over at xda-developers, cr2 hinted that haret's dump pxa27x command output should be compared between a 128MB and a 64MB Universal, to give an idea of the patch presumably. I've posted the 128MB Universal output at http://forum.xda-developers.com/showpost.p...;postcount=2088

However, I don't have the 64MB output. Could you post it? Or could someone else post the haret's dump pxa27x command's output here? Instructions on how to capture the output easily on a Universal I've posted at: http://sumoudou.org/index.php?cmd=%E7%9B%B...PC%20connection

Yes, I have a, at least rough, idea where to patch. I the pxa specific code are some parts for the power management. These are not so big, and it seems easy to find, where the resume is done.
You start your dump with "1st part:" , is there a 2nd missing? I did not see the memory registers on a quick view.

I'll make a 64MB dump today, and post it for comparsion.

greetings

Thorsten

[Thorsten]

You start your dump with "1st part:" , is there a 2nd missing? I did not see the memory registers on a quick view.

I'll make a 64MB dump today, and post it for comparsion.

Sorry, I now found your 2nd part, and also the post with the comparsion log....

[ShiroiKuma]

Yes, I have a, at least rough, idea where to patch. I the pxa specific code are some parts for the power management. These are not so big, and it seems easy to find, where the resume is done.

I'll check further, so far haven't found anything that would indicate to me the right place to patch. In http://forum.xda-developers.com/showpost.p...;postcount=2094 cr2 indicates patching the IPL, which is the bootloader, but for that I have no idea, so hopefully patching the kernel will prove to be the way...