@realize

Quote:

I have UDMA working fine on my peg2 its NOT MOL. Its your Amiga One (Teron) board that is having problems with DMA.

I've had DMA working on my machine for years in both OS4 and Linux. I've found kernel 2.6.31-1 with Ubuntu 9.04 on my machine to be quite stable. I can both use DMA and USB2 transfers with Linux.

I'm also using a Sii0680 IDE controller exclusively to avoid DMA trouble so I should not be troubled by DMA issues.

So would MoL be programming the IDE registers directly from within OSX? That's the one way I can see it crashing. Otherwise OSX must be badly programmed if it crashes over a drive error. I haven't seen this on a real Mac, it just delays. The DMA issue should be restricted to Linux.

I tried to look Kurt up and found a dead website. I don't know any other details about him. Was his MoL package ever uploaded?

The 2.4 kernel you propose is very old and predates the 2.6.18 kernel I am having trouble with and I wouldn't rely on it. Old kernels miss out on later patches, exhibit bugs and are unstable on newer hardware. For example they break with larger memory and may not have 0680 IDE or SATA support.

Last edited by Hypex on 03-May-2014 at 03:20 AM.

@realize



I found the web site here: http://www.mrhardwarecomputers.com/

but cannot find the version of MOL that Kurt distributed.

Can anyone out there in AmigaLand help?


Scott

_________________
AmigaOne uA1-c 512M RAM - Only Amiga Makes It Possible!
Check my blog AmigaOne Computing

@Hypex

[quote] I've got this working before so I don't know why Mol on the AmigaOne is so buggy! /quote]

Yes, its quite frustrating! But I am using kernel 2.6.18 (well, 2.6.18-6 I guess).....so how do I get the original version of MOL that worked with 2.6.18?

Scott

_________________
AmigaOne uA1-c 512M RAM - Only Amiga Makes It Possible!
Check my blog AmigaOne Computing

@Hypex

Some interesting observations.....

When I boot into Wheezy, Squeeze or even Lenny (with kernel 2.6.31) I have a proc/device-tree/cpus file.
When I boot into Lenny using the MOL kernel 2.6.18, there is no /proc/device-tree directory!
Rather there is a /proc/cpuinfo file instead of /proc/device-tree/cpus file.

This makes me wonder if the MOL kernel I am using, which is looking for the file /proc/device-tree/cpus was built for a different kernel than 2.6.18.....and yet, when I go to install MOL it specifically only works with the 2.6.18 kernel. But if the 2.6.18 kernel does not have a /proc/device-tree/cpus file, why is this version of MOL looking for it?

I suspect the MOL that was built that I am using somehow got messed up with a combination of two different kernels or something.

So, do we have a version of MOL built for kernel 2.6.31 somewhere?

Scott

_________________
AmigaOne uA1-c 512M RAM - Only Amiga Makes It Possible!
Check my blog AmigaOne Computing

@scabit

Geri can confirm but I'd say the newer MoL source that was compiled was indeed made to work with a later kernel. But can compile fine on an older kernel.

AFAIK we don't have any versions built for 2.6.31 or newer. On that thread you can see how I tried and gave up trying to compile on 2.6.32.

However, it has been a tradition of A1 Linux installers to include MoL. Perhaps I should install all later kernels, the source headers and see if I can compile MoL again. Then the problem becomes incompatible distros. But having a MoL to download for every later kernel which works would be good.

@Hypex

[quote] AFAIK we don't have any versions built for 2.6.31 or newer. /quote]

I am so sad....the MOL kernel I compiled last year (or was it 2012?) was either for kernel 2.6.31 or 3.1...and man, it took a lot of "little tricks" to get it to compile and build successfully. Geri gave me a lot of advice getting it up and running....but at that time I did not have a MAC OSX DVD to fully test it. It did build fine and run in test mode just fine with the new kernel though.
Too bad my hard drive got trashed a few months ago...it had my new MOL and sources on it.

I will try your trick of creating a usb copy of the DVD and see if that helps the install.....

But I suspect you or I will need to create a new MOL built for a newer kernel to get back to a stable working point.....

But...didn't Geri say that the MOL and kernel he uploaded to sourceforge work for him on his A1? The one I am using shows MAC OSX Boot Loader 0.9.73-SVN
Geri, is that what you are using?

Hey, where has Geri been anyway? Haven't heard from him in a while....


Scott

_________________
AmigaOne uA1-c 512M RAM - Only Amiga Makes It Possible!
Check my blog AmigaOne Computing

@scabit

you have to compile the kernel mods for MOL manually to your specific linux kernel. Take my advice bro you are spending alot of fruitless time trying to to get newer 2.6 kernels working with mol for your a1 (is there Anyone who has mol working on 2.6+ kernel??)

Yes mr hardware may be defunct :(

@Hypex

So...I tried to copy my MACOSX dvd to my 16 Gig memory stick using the dd command.

dd does NOT work. I did this: dd if=/dev/sg1 of=/dev/sdb1

wait ... no dvd access, no usb access...nothing...wait for 20 minutes, still nothing.
CTRL-C, message shows 0 blocks copied.

Tried this about a dozen times. Tried it with both devices NOT mounted, and tried it with the usb stick mounted. Tried it with the usb stick formatted to EXT3. Nothing.

dd does NOTHING.

So much for that idea......

Scott

_________________
AmigaOne uA1-c 512M RAM - Only Amiga Makes It Possible!
Check my blog AmigaOne Computing

@scabit

Maybe try gparted instead of dd?

@scabit

Geri seems to have disappeared. Although he did do a kernel update to SF.

He has updated MoL there as well for use with the 2.6.18-6 kernel which should be the best we have. But to as to which MoL works for Geri on his A1. All MoL's work for Geri on his A1! We've had this argument before. Me and him.

I have problems he doesn't see.

Here is an older version that should work with older 2.6.18:
http://home.lizzy.com.au/dmaen/downloads/mol_0.9.73-2_powerpc.deb
http://home.lizzy.com.au/dmaen/downloads/mol-kmods_0.9.73-2_powerpc.deb

Now, the dd copy fail. Did you do this as root? And what kernel?

I just did a test. I tried to copy from DVD to my OSX partition on HD. I used the 2.6.31-1 kernel with my old Ubuntu install. Using the same format as for dd I was able to copy from DVD to a HD partition. DVD was mounted, HD partition was not. Took about 5 mins, 5.7MB/s it said.

I also was able to boot off the volume in MoL but as my only OSX volume it had nowhere to install! Disk Utility ran fine. But the Installer Log froze my machine. So that is what is doing it.

Perhaps I should try with 2.6.18-6 to see if dd fails with it. And of course copy it to a USB stick if I can make a free one.

As a last resort on USB format your USB stick as HFS+ and copy the DVD files over.

Failing that. You can try to copy it to a spare partition on HD like I've done. If you have a spare. This is the only way I have ever installed it IIRC. Sad.

BTW, I also tried to find the latest source but I can never get it! Don't know where I found other one I have. That SF site always has old source which annoys me. I wish they'd update it! Last updated 2013 but the source says 2007! Argh!!

Last edited by Hypex on 05-May-2014 at 03:24 PM.

@Hypex

Quote:

Hypex wrote: @scabit Geri seems to have disappeared. Although he did do a kernel update to SF.

Not disappeared, just busy clearing a backlog of other work.
Yes, I uploaded a new kernel package to sf.net, but it was only a side-product of the work trying to debug the lockups of 3.x kernels on my A1XE G4.

Quote:

He has updated MoL there as well for use with the 2.6.18-6 kernel which should be the best we have. But to as to which MoL works for Geri on his A1. All MoL's work for Geri on his A1! We've had this argument before. Me and him.

Quote:

I have problems he doesn't see.

Before I try to read the whole thread now, of what problems exactly are we talking about?

BTW: I never heard of this 8GB limit/problem. My OSX installation resides on a partition at the end of a 80GB harddisk. But naturally OSX only sees this partition as its harddisk.

Quote:

I just did a test. I tried to copy from DVD to my OSX partition on HD. I used the 2.6.31-1 kernel with my old Ubuntu install. Using the same format as for dd I was able to copy from DVD to a HD partition. DVD was mounted, HD partition was not. Took about 5 mins, 5.7MB/s it said.

Sounds slow. But IIRC dd usually uses a blocksize of 512 bytes, which wastes quite a lot of speed. Running it with a blocksize of 1MB or increasing the count parameter should speed it up.

Quote:

I also was able to boot off the volume in MoL but as my only OSX volume it had nowhere to install! Disk Utility ran fine. But the Installer Log froze my machine. So that is what is doing it.

Was this with a SATA controller? To be honest I think I never tested MoL together with a SATA controller. But OSX occasionally also locked up on my A1. IIRC some OSX update mitigated the problem.

Note that MoL just provides access to Linux' block devices, so any DMA problem will also affect OSX.

Quote:

BTW, I also tried to find the latest source but I can never get it! Don't know where I found other one I have. That SF site always has old source which annoys me. I wish they'd update it! Last updated 2013 but the source says 2007! Argh!!

I guess you have to do a SVN checkout to get the latest source code version.

BTW: one could give xeno74's MoL version a try. KVM support should be included in all 3.x kernels for the AmigaOne.

From a previous post:
Quote:

That's strange. a "/dev/sg1" should have worked just as well. In any case looks like I use "/dev/cdrom" and SCSI auto probe. Don't know why "/dev/cdrom" was breaking on your system.

/dev/cdrom resp. /dev/srx should be the right device. I would have to give MoL a try with a SATA controller first to see, if that makes a difference.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri

Quote:

Not disappeared, just busy clearing a backlog of other work.

That's why I said it seems. Yes I thought you were busy with work.

Any closer to tracking that G4 bug? I may be being simple here but what happens, if you can, disable AltiVec? That and caches I know are different.

Quote:

Before I try to read the whole thread now, of what problems exactly are we talking about?

This page might do it. But for me the main one is DMA corruption which I confirmed a few days ago. I can boot a 10.4 DVD (and also 10.3 CD) but when I open Disk Utility it crashes when reading drives. It also always fails to install off optical disc. It reports a problem with the disc. I have to image the OSX disc onto a spare partition and boot from there.

I also found a read error in the OSX install log but bringing up the window freezes my A1. Hard reset needed. This is a new one.

I found "ide=nodma" in the Linux boot args was acknowledged but drives were still in DMA mode when queried. Using hdparm to disable DMA improved the situation but it still eventually broke and I still couldn't install OSX off disc.

Quote:

BTW: I never heard of this 8GB limit/problem.

It's an old one but IIRC I had to deal with it in the early MoL days. My OSX partition was 40GB into the drive before I accidently starting installing Debian on it. Now it has an OSX disc image.

Quote:

Sounds slow.

Does but it didn't take a long time. Wasn't like half an hour. BTW don't know what that head banging was about. Since it worked.

Quote:

Was this with a SATA controller?

No as I only have my 0680 IDE. But Scott is having problems and has SATA.

Quote:

Note that MoL just provides access to Linux' block devices,

Then kernel 2.6.18-6 has a problem with DMA, on both 0680 and SATA. And by the looks of it something else by my freeze.

Quote:

I guess you have to do a SVN checkout to get the latest source code version.

I did that once but when I examimed the code lately using HTTP SVN checkout on SF it was still old files not updated in 7 years!

Perhaps I should just SVN it anyway to my system.

Quote:

BTW: one could give xeno74's MoL version a try.

Could be an idea. Be annoying if it worked for us and not him.

Quote:

/dev/cdrom resp. /dev/srx should be the right device.

IIRC /dev/cdrom existed on his system but when he added it in MoL could not find the block device as if it wasn't mounted or linked to the block device.

@Hypex

Quote:

Hypex wrote: @Geri Quote: Not disappeared, just busy clearing a backlog of other work. That's why I said it seems. Yes I thought you were busy with work.

Well, I hate it when a plan fails! You may know what I mean...

Quote:

Any closer to tracking that G4 bug? I may be being simple here but what happens, if you can, disable AltiVec? That and caches I know are different.

Not really, I have to do a full bisect run between kernel 2.6.32 and 3.2. Thankfully I managed to build a Debian Wheezy powerpc cross compiler so that I can do full-blown kernel builds in under 15 min. But a full bisect run will still need some time...

But I'm not so sure anymore that we're talking only about a G4 problem. I wouldn't be surprised if these lockups have a similar root cause like a sporadic lockup of my e1000 1Gbit network card in my A1SE, even if the effect is worse on the A1XE. The lockups only happen with the onboard IDE controller, but not with the SATA controller...

Quote:

Quote: Before I try to read the whole thread now, of what problems exactly are we talking about? This page might do it. But for me the main one is DMA corruption which I confirmed a few days ago. I can boot a 10.4 DVD (and also 10.3 CD) but when I open Disk Utility it crashes when reading drives. It also always fails to install off optical disc. It reports a problem with the disc. I have to image the OSX disc onto a spare partition and boot from there.

Well, remember that the DMA fix used for the AmigaOne is not the best solution. So it would still allow DMA corruption to occur in some corner cases...even if this is more a theoretical problem on the 6xx CPUs according to some professional Linux developer.

Quote:

I also found a read error in the OSX install log but bringing up the window freezes my A1. Hard reset needed. This is a new one.

Hmm, hard to tell if the error is in the Linux kernel, MoL or OSX. As I mentioned before I had to install all the OSX updates to get a halfway stable installation.

Quote:

I found "ide=nodma" in the Linux boot args was acknowledged but drives were still in DMA mode when queried. Using hdparm to disable DMA improved the situation but it still eventually broke and I still couldn't install OSX off disc.

Could it be that hdparm was setup to reenable DMA during startup?

Quote:

It's an old one but IIRC I had to deal with it in the early MoL days. My OSX partition was 40GB into the drive before I accidently starting installing Debian on it. Now it has an OSX disc image.

You mean, OSX is installed in a image file now?

Quote:

Does but it didn't take a long time. Wasn't like half an hour. BTW don't know what that head banging was about. Since it worked.

Okay.

Quote:

No as I only have my 0680 IDE. But Scott is having problems and has SATA.

For SATA it could as well have various reasons. I'll have to give it a try...

Quote:

Then kernel 2.6.18-6 has a problem with DMA, on both 0680 and SATA. And by the looks of it something else by my freeze.

I can't say anything about the 0680 controller, but I remember that the sata_sil driver did see some data corruption fixes after the 2.6.18 kernel.

Quote:

I did that once but when I examimed the code lately using HTTP SVN checkout on SF it was still old files not updated in 7 years!

I think the SVN checkout worked for me, when I tried to compile MoL with KVM support. However I can't remember how I did it...

Quote:

Could be an idea. Be annoying if it worked for us and not him.

Quote:

IIRC /dev/cdrom existed on his system but when he added it in MoL could not find the block device as if it wasn't mounted or linked to the block device.

Strange, /dev/cdrom is working fine here!?

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Hypex

Quote:

Could be an idea. Be annoying if it worked for us and not him.

Don't worry. If you need my version, then I can upload it. Lubuntu 14.04 PowerPC works great as virtual machine on QEMU with KVM. But Mac OS X Jaguar, Panther, and Tiger don't work with KVM on MOL and QEMU.

I asked Alexander Graf the developer of MOL with KVM support why Mac OS X doesn't boot. His answer was:

Quote:

Not sure what's going wrong here. Maybe the MOL patchery to circumvent split real mode doesn't work? Maybe there's something not working with the dcbz32 patchery? You could try to trace the guest through /sys/kernel/debug/tracing and see where it's stuck. You can also tell XNU to be more verbose by passing in a debug= parameter - though I don't remember the exact syntax anymore. SMP should be fine. The reason XNU doesn't work with QEMU + KVM is split real mode. We don't emulate that one properly in KVM yet. Alex

_________________
www.amigalinux.org

@xeno74

Quote:

xeno74 wrote: I asked Alexander Graf the developer of MOL with KVM support why Mac OS X doesn't boot. His answer was: Quote: Not sure what's going wrong here. Maybe the MOL patchery to circumvent split real mode doesn't work? Maybe there's something not working with the dcbz32 patchery?

Looks like he did a rework of this patchery recently:
http://www.spinics.net/lists/kvm/msg102126.html

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri

Thanks for the hint. I'll try out the patch with the new kernel 3.14.2.

_________________
www.amigalinux.org

@Geri

I tried this patch. Unfortunately Mac OS X Jaguar and Tiger didn't boot.

_________________
www.amigalinux.org

@Geri

Quote:

Well, I hate it when a plan fails! You may know what I mean...

Yes I think so.

Quote:

Thankfully I managed to build a Debian Wheezy powerpc cross compiler

Yes I saw that. Looks like a lot of work in all those files.

Quote:

The lockups only happen with the onboard IDE controller, but not with the SATA controller...

Knowing the problems I stopped using mine years ago.

Quote:

Well, remember that the DMA fix used for the AmigaOne is not the best solution

Depending what DMA fix you are talking about. My DMA fix is a Sii 0680 IDE controller and AFAIK it is rock solid. So my machine shouldn't have any DMA corruption. Never seen any with OS4.1 or Ubuntu.

Quote:

As I mentioned before I had to install all the OSX updates to get a halfway stable installation.

That's not an option I have when running the installer.

Quote:

Could it be that hdparm was setup to reenable DMA during startup?

Not that I was aware of. You mean as it boots, not when I run it? When the kernel boots I can see it examine the drive, what PIO/DMA mode is supports, and how it sets it to max DMA, for all drives.

Quote:

You mean, OSX is installed in a image file now?

Not exactly, the OSX DVD is installed in an image file now.

I can boot the installer DVD off my HD.

Quote:

I think the SVN checkout worked for me,

Okay just tested and looks like it worked for me as well. Now all we need are some 2.6.31-1 and 2.6.32-5 headers. I think the rest are covered.

Quote:

Strange, /dev/cdrom is working fine here!?

And here also.

Last edited by Hypex on 06-May-2014 at 04:45 PM.

@xeno74

Quote:

xeno74 wrote: @Geri I tried this patch. Unfortunately Mac OS X Jaguar and Tiger didn't boot.

Yeah, it rather seems to be a patch related to the new powerpc64le support, when taking a closer look...

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Hypex

Quote:

Hypex wrote: @Geri Quote: Well, I hate it when a plan fails! You may know what I mean... Yes I think so.

Quote:

Yes I saw that. Looks like a lot of work in all those files.

This are just the toolchain binary packages. For a full toolchain one needs to install also some library dependencies using xapt with the "-m" multiarch option. But the multiarch support in Debian Wheezy makes that pretty easy.

Quote:

Quote: The lockups only happen with the onboard IDE controller, but not with the SATA controller... Knowing the problems I stopped using mine years ago.

This behavior is definitely new, as the onboard IDE controller works with kernel 2.6.32 on my A1XE.

Quote:

Quote: Well, remember that the DMA fix used for the AmigaOne is not the best solution Depending what DMA fix you are talking about. My DMA fix is a Sii 0680 IDE controller and AFAIK it is rock solid. So my machine shouldn't have any DMA corruption. Never seen any with OS4.1 or Ubuntu.

You're right. In this case I'm talking about the non-coherent DMA workaround in the Linux kernel.

Quote:

Quote: As I mentioned before I had to install all the OSX updates to get a halfway stable installation. That's not an option I have when running the installer.

Unfortunately. I guess there is also no way to integrate the latest updates into the OSX installer, as it can be done for Windows install CDs with the service packs.

Quote:

Quote: Could it be that hdparm was setup to reenable DMA during startup? Not that I was aware of. You mean as it boots, not when I run it? When the kernel boots I can see it examine the drive, what PIO/DMA mode is supports, and how it sets it to max DMA, for all drives.

Yes, I mean when it boots. How does your kernel command line look like?

Quote:

Not exactly, the OSX DVD is installed in an image file now. I can boot the installer DVD off my HD.

That should be faster anyway.

Quote:

Okay just tested and looks like it worked for me as well. Now all we need are some 2.6.31-1 and 2.6.32-5 headers. I think the rest are covered.

Well, I don't have headers for the 2.6.x kernels, but you could give the 3.x kernels a try that should have better KVM support anyway.

Quote:

And here also.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -