@Geri

Okay, here is a dumb question....in the modules file for building the kernel, do I need to have floppy and sata_sil? My SIL0680 only has PATA, not SATA and I have no floppy drive.

Quote:

Afterwards save the following text in "/etc/initramfs-tools/modules": # List of modules that you want to include in your initramfs. # They will be loaded at boot time in the order below. # # Syntax: module_name [args ...] # # You must run update-initramfs(8) to effect this change. # # Examples: # # raid1 # sd_mod floppy ide-pci-generic ide-cd_mod siimage via82cxxx sd_mod scsi_mod libata ata_generic pata_sil pata_via sata_sil sata_via dm-mod cdrom crc-t10dif mbcache ext2 ext3 jbd 3c59x pcspkr

Also, I discovered that in kernel 2.6.32-5 (and possibly the other kernels as well) that when I disable both front and rear internal USB ports in uboot, the front ports ARE REALLY disabled, but the rear USB ports continue to work!
This is on the microA1. Just something to be aware of I guess.

Scott

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

@Geri

Okay, I removed floppy and sata_sil from the modules file.
I changed the local.conf file to this:

# Local module settings
# Created by the Debian installer - modified by Scott Cabit 3-21-13

options ide_core nodma=0.0 nodma=0.1 nodma=1.0 nodma=1.1
# onboard network shouldn't use PCI MMIO!
options 3c59x global_use_mmio=0
# Turn off USB UHCI
#blacklist uhci_hcd
# Force SATA To use 1.5G speed, turn off NCQ, turn off reset
options libata.force=1.5G,noncq,norst

Line 10, "options libata.force=1.5G,noncq,norst gives me an error during boot that says:
"ignoring bad line starting with options"

I found this online in the Debian archives dated October 8, 2012:

>Adding "libata.force=1.5Gbps" to the boot parameters for the installer
> doesn't work. It leads to a message saying "unknown parameter 5Gbps".
> (Trying 3.0Gbps leads to "unknown parameter 0Gbps" so it looks like it
> could be a parsing issue.)
>It turns out this wasn't limited to just the installer. It's a problem
>with the libkmod2 package. Module parameters given in the kernel
>command line are parsed incorrectly if the parameter value has a '.' in
>it, like 1.5Gbps. I filed a bug report with a fix:
>http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=689872

Not sure if that is my problem, but I will try some more. So far this line doesn't work...perhaps I need to created 3 different options lines, one for each parameter?

SCott

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

@scabit

Quote:

scabit wrote: @Geri ... When i got the SATA to work in Lenny, it was because the Syba card was in the PCI slot, and I was booting (as I am now in Squeeze) from the internal micro IDE bus. The hardware configuration when Lenny identified the SATA drives as sda1,sda2 and sda4 was identical to what I am trying now to do with Squeeze.

Okay.

Quote:

I think (but may be wrong) that in my current setup - i.e. Syba card in PCI slot, no SIL, booting from internal IDE in microA1, the Syba PATA/RAID port devices should show up as hdeX and the SATA drives should show up as sdaX .... no? That is what I was expecting.

As far as I understand libata based drivers (like the sata_via driver) always use the sdx naming scheme, no matter if its a PATA or SATA device.

Quote:

scabit wrote: @Geri Okay, I just rebooted into Squeeze using kernel 2.6.31-1 and it looks like the RAID/PATA drives are being seen as sda1,sda2 and sda4! I KNOW that when I did this a couple weeks ago using the exact same hardware configuration and kernel image, they were hde1,hde2 amd hde4.

I would have expected the former. Not sure why they were named hdx back then.

Quote:

What exactly determines the name that is assigned to a hard drive?

As far as I understand it, the names are assigned by the kernel.

Quote:

Hurray! Perhaps now I can try the USB2 uboot modifications....

Note that I'm not sure, if one of the modifications will fix USB2 now, but at least we know that the controller is handled by the correct driver now.

Quote:

It makes sense to me to try option #2 first. I'll see if I can discover where this file exists and make the changes you suggest and see what happens.

Yes, that makes sense.

Quote:

Hmmmm...interesting. I'll test that out. So far I haven't seen any lockups due to large downloads - I just downloaded some huge stuff to install a vector graphics drawing program in Squeeze that was supposed to (but doesn't) recognize .eps files.

To be more precise: the lockups will happen only at nearly 100Mbit ethernet speed. I did some tests a long time ago, where my system locked up reproducibly after around 700MB of data transfered at a speed between 5 to 11MB/s. Forcing the driver to use PCI IO access allowed me to transfer several GBs of data at the same speed without any lockup.

Quote:

Pesky. I'll verify that that is true. I seem to recall that having the internal USB ports on somehow conflicted with the Syba USB ports.

Yes, it is pesky. The thing is that the USB hardware isn't really disabled. It is just not initialized completely and thus Linux wants to get it working.

Quote:

I'll let you know what I find.

Yes, please!

@scabit

Quote:

scabit wrote: @Geri Okay, here is a dumb question....in the modules file for building the kernel, do I need to have floppy and sata_sil? My SIL0680 only has PATA, not SATA and I have no floppy drive.

No, you can remove these drivers from the modules file. That's the good thing about using an initramfs based boot image, i.e. remove what you don't need.
Note however that some modules listed in the file are needed by other modules. The sata_/pata_* and the floppy modules however are save to remove. I just tried to create a small but universal modules file that covers most use cases.

Quote:

Also, I discovered that in kernel 2.6.32-5 (and possibly the other kernels as well) that when I disable both front and rear internal USB ports in uboot, the front ports ARE REALLY disabled, but the rear USB ports continue to work! This is on the microA1. Just something to be aware of I guess.

Yes, that's also the case on my A1XE. IIRC only the interrupt line of the front USB controller is not initialized, when the onboard USB is disabled.

@scabit

Quote:

scabit wrote: @Geri Okay, I removed floppy and sata_sil from the modules file. I changed the local.conf file to this: # Local module settings # Created by the Debian installer - modified by Scott Cabit 3-21-13 options ide_core nodma=0.0 nodma=0.1 nodma=1.0 nodma=1.1 # onboard network shouldn't use PCI MMIO! options 3c59x global_use_mmio=0 # Turn off USB UHCI #blacklist uhci_hcd # Force SATA To use 1.5G speed, turn off NCQ, turn off reset options libata.force=1.5G,noncq,norst Line 10, "options libata.force=1.5G,noncq,norst gives me an error during boot that says: "ignoring bad line starting with options"

The notation of "module.option" can only be used on the kernel command line. In a /etc/modprobe.d/ file you would have to write it this way:

Quote:

options libata force=1.5G,noncq,norst

Note the missing dot!

Quote:

I found this online in the Debian archives dated October 8, 2012: >Adding "libata.force=1.5Gbps" to the boot parameters for the installer > doesn't work. It leads to a message saying "unknown parameter 5Gbps". > (Trying 3.0Gbps leads to "unknown parameter 0Gbps" so it looks like it > could be a parsing issue.) >It turns out this wasn't limited to just the installer. It's a problem >with the libkmod2 package. Module parameters given in the kernel >command line are parsed incorrectly if the parameter value has a '.' in >it, like 1.5Gbps. I filed a bug report with a fix: >http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=689872 Not sure if that is my problem, but I will try some more. So far this line doesn't work...perhaps I need to created 3 different options lines, one for each parameter?

I don't think so (based on the kernel parameter documentation). Let's try out my suggestion above first.

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

- A1 Linux support -

@Geri

Quote:

options libata force=1.5G,noncq,norst

Of course, you are correct. Changed as you described, now no more erros during boot with the local.conf file.
Sadly, SATA still fails. Here is the dmesg "ata" output lines:

[ 0.022398] pci 0000:00:07.0: calling quirk_no_ata_d3+0x0/0x20
[ 0.022517] pci 0000:00:07.1: calling quirk_no_ata_d3+0x0/0x20
[ 0.022614] pci 0000:00:07.2: calling quirk_no_ata_d3+0x0/0x20
[ 0.022716] pci 0000:00:07.3: calling quirk_no_ata_d3+0x0/0x20
[ 0.022840] pci 0000:00:07.4: calling quirk_no_ata_d3+0x0/0x20
[ 0.022999] pci 0000:00:07.5: calling quirk_no_ata_d3+0x0/0x20
[ 0.023120] pci 0000:00:07.6: calling quirk_no_ata_d3+0x0/0x20
[ 0.023222] pci 0000:00:08.0: calling quirk_no_ata_d3+0x0/0x20
[ 0.023349] pci 0000:00:08.1: calling quirk_no_ata_d3+0x0/0x20
[ 0.023471] pci 0000:00:08.2: calling quirk_no_ata_d3+0x0/0x20
[ 0.023594] pci 0000:00:08.3: calling quirk_no_ata_d3+0x0/0x20
[ 0.023884] pci 0000:01:00.0: calling quirk_no_ata_d3+0x0/0x20
[ 2.533136] libata version 3.00 loaded.
[ 4.007784] sata_via 0000:00:08.3: version 2.4
[ 4.007855] sata_via 0000:00:08.3: calling quirk_via_vlink+0x0/0xd4
[ 4.007953] sata_via 0000:00:08.3: routed to hard irq line 9
[ 4.018316] scsi0 : sata_via
[ 4.025607] scsi1 : sata_via
[ 4.032544] scsi2 : sata_via
[ 4.039337] ata1: SATA max UDMA/133 port i16@0x8000 bmdma 0x8080 irq 9
[ 4.045950] ata2: SATA max UDMA/133 port i16@0x8020 bmdma 0x8088 irq 9
[ 4.052522] ata3: PATA max UDMA/133 port i16@0x8040 bmdma 0x8090 irq 9
[ 4.389909] ata1: SATA link down (SStatus 0 SControl 310)
[ 9.755560] ata2: link is slow to respond, please be patient (ready=0)
[ 14.407557] ata2: COMRESET failed (errno=-16)
[ 19.771554] ata2: link is slow to respond, please be patient (ready=0)
[ 24.423558] ata2: COMRESET failed (errno=-16)
[ 29.787554] ata2: link is slow to respond, please be patient (ready=0)
[ 59.471556] ata2: COMRESET failed (errno=-16)
[ 64.499552] ata2: COMRESET failed (errno=-16)
[ 64.506229] ata2: reset failed, giving up
[ 66.018341] EXT3-fs: mounted filesystem with ordered data mode.

So we still get a failure mounting the SATA.

On another subject, I noticed that when I used the "dd" command to copy my LEnny ext3 hard drive of 4.5 gigabytes (formatted in Etch) to a 24 gig partitioned as ext3 and formatted in Squeeze, the dd command completed...all the files were on the 24 gig drive... but gparted showed the drive as being nearly full...I think it showed up as 21Gig or so! What's up with that?
Does that mean that the ext3 block size on Lenny is, say 512 bytes while the block size on Squeeze is 2048 bytes and it copies block for block so ends up size * 4? Just wondering.

Scott

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

@scabit

Quote:

scabit wrote: @Geri Quote: options libata force=1.5G,noncq,norst Of course, you are correct. Changed as you described, now no more erros during boot with the local.conf file. Sadly, SATA still fails. Here is the dmesg "ata" output lines: [ 0.022398] pci 0000:00:07.0: calling quirk_no_ata_d3+0x0/0x20 [ 0.022517] pci 0000:00:07.1: calling quirk_no_ata_d3+0x0/0x20 [ 0.022614] pci 0000:00:07.2: calling quirk_no_ata_d3+0x0/0x20 [ 0.022716] pci 0000:00:07.3: calling quirk_no_ata_d3+0x0/0x20 [ 0.022840] pci 0000:00:07.4: calling quirk_no_ata_d3+0x0/0x20 [ 0.022999] pci 0000:00:07.5: calling quirk_no_ata_d3+0x0/0x20 [ 0.023120] pci 0000:00:07.6: calling quirk_no_ata_d3+0x0/0x20 [ 0.023222] pci 0000:00:08.0: calling quirk_no_ata_d3+0x0/0x20 [ 0.023349] pci 0000:00:08.1: calling quirk_no_ata_d3+0x0/0x20 [ 0.023471] pci 0000:00:08.2: calling quirk_no_ata_d3+0x0/0x20 [ 0.023594] pci 0000:00:08.3: calling quirk_no_ata_d3+0x0/0x20 [ 0.023884] pci 0000:01:00.0: calling quirk_no_ata_d3+0x0/0x20 [ 2.533136] libata version 3.00 loaded. [ 4.007784] sata_via 0000:00:08.3: version 2.4 [ 4.007855] sata_via 0000:00:08.3: calling quirk_via_vlink+0x0/0xd4 [ 4.007953] sata_via 0000:00:08.3: routed to hard irq line 9 [ 4.018316] scsi0 : sata_via [ 4.025607] scsi1 : sata_via [ 4.032544] scsi2 : sata_via [ 4.039337] ata1: SATA max UDMA/133 port i16@0x8000 bmdma 0x8080 irq 9 [ 4.045950] ata2: SATA max UDMA/133 port i16@0x8020 bmdma 0x8088 irq 9 [ 4.052522] ata3: PATA max UDMA/133 port i16@0x8040 bmdma 0x8090 irq 9 [ 4.389909] ata1: SATA link down (SStatus 0 SControl 310) [ 9.755560] ata2: link is slow to respond, please be patient (ready=0) [ 14.407557] ata2: COMRESET failed (errno=-16) [ 19.771554] ata2: link is slow to respond, please be patient (ready=0) [ 24.423558] ata2: COMRESET failed (errno=-16) [ 29.787554] ata2: link is slow to respond, please be patient (ready=0) [ 59.471556] ata2: COMRESET failed (errno=-16) [ 64.499552] ata2: COMRESET failed (errno=-16) [ 64.506229] ata2: reset failed, giving up [ 66.018341] EXT3-fs: mounted filesystem with ordered data mode. So we still get a failure mounting the SATA.

I would have expected some "FORCE: ..." messages, if the force option is recognized!? Maybe they are only shown after a link has been established.
Okay, I'll wade through the kernel development history. Hopefully there's a commit that tackles this problem. But if you find the time, then please also give the 3.2.0 kernel a try. Or do you have another SATA device that you could attach to the VIA controller temporarily for testing?

Quote:

On another subject, I noticed that when I used the "dd" command to copy my LEnny ext3 hard drive of 4.5 gigabytes (formatted in Etch) to a 24 gig partitioned as ext3 and formatted in Squeeze, the dd command completed...all the files were on the 24 gig drive... but gparted showed the drive as being nearly full...I think it showed up as 21Gig or so! What's up with that?

Just to avoid misunderstandings: can you please post the full dd copy command? Because one can easily destroy lots of data with a wrong dd command...

Quote:

Does that mean that the ext3 block size on Lenny is, say 512 bytes while the block size on Squeeze is 2048 bytes and it copies block for block so ends up size * 4? Just wondering.

If I understand you correctly, then you copied a 4.5GB harddisk (not a single partition) to a 24GB harddisk (i.e. dd if=/dev/sda of =/dev/sdb with sda = 4.5GB and sdb = 24GB harddisk). From my understanding and experience the result is a 24GB harddisk with a 4.5GB partition and a lot of free unpartitioned space (because dd overwrites the partition table and the existing 24GB partition with its filesystem). To use the full 24GB space you would have to extend the 4.5GB partition to the full 24GB of the new harddisk and resize the filesystem afterwards (the latter one my require a reboot so that Linux recognizes the modified partition table of the 24GB harddisk).

I think the default for dd is to do 512 byte copies. The block size of the ext3 file system doesn't matter for dd. The filesystem blocksize is chosen by mke2fs resp. mkfs.ext3 dependent on the size of the filesystem.

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

- A1 Linux support -

@Geri

Hmmm...one important thing I need to mention. The local.conf file is not getting run until AFTER gdm3 starts up. So when the sata buses are being tested, the force commands in the local.conf file have not yet been applied. I know this because when I had errors in line 10, the error messages appeared after the gdm3 startup, well after the sata buses had been tested and failed.

Quote:

If I understand you correctly, then you copied a 4.5GB harddisk (not a single partition) to a 24GB harddisk

No, it was a 10 Gig partition of which 5.30 gig was used (sorry, I wrote 4.5), to a 24 Gig partition that was formatted blank as ext3.
The exact dd command I used to copy my working LENNY partition to my new drive was:

dd if=/dev/sda2 of=/dev/hda12

And I must admit that at least the PATA portion of the Syba card (sda2) is working great. That is, I have never to my knowledge had any errors copying to or from the sda devices using the PATA 40 pin EIDE cabled drive. The dd command however from the sda2 partition to the hda12 partition brought over what gparted shows as a lot more data...instead of 5.30 Gig I ended up with something like 21 Gig.

As for kernel 3.2.0 - I downloaded it and did a dpkg --install. I have what looks like an initramfs image for it in boot. The is also a cuImage-amigaone-3.2.0 etc in /boot as well. Do I need to use a mkimage now to generate a bootable image file for the AmigaOne? If so, what are the start and entry parameters I need, (i.e. was 0x04000000 and 0x04000440 in 2.6.31-2). Is this where I need to start, with the mkimage command?

I tried something else as well since yesterday. I booted from the ETCH install CD and partitioned and formatted my hda12 drive using the Etch partitioner. I then rebooted, went into Squeeze, mounted sda2 and hda12 and used nautilus to copy (in superuser mode) all of the files from sda2 to hda12. hda12 now shows 5.57Gig, sda2 was 5.30Gig. Closer to what I expect.
But when I boot into sda12 LENNY, everything starts to boot up but stops after the root file system at hda12 has been mounted with an error about detecting the keyboard. It just sits there, CTRL-ALT-DEL and CTRL-ALT-F1 do nothing. I must hard reset.

Why do I want to install LENNY on my 80Gig hard drive you ask? So I can test the SATA interface from LENNY that worked without having to recable anything - the hardware setup will be exactly the same as what I am trying to do for Squeeze. i.e. 80Gig drive connected to internal IDE bus, SATA connected to the 160Gig drive from the Syba card. No SIL card installed. If LENNY works with SATA and SQUEEZE does not - without any hardware change, then maybe we can narrow down the issues in software. I just want to establish that the hardware configuration is correct and working.

Any ideas what is missing from my LENNY partition when I just use copy to move all the files over? It is doing essentially a cp -r since all subdirectories are populated.

Oh, why is it that the config files in /boot for kernels 2.6.31, 2.6.32 and 3.2.0 all have this line: CONFIG_ALTIVEC=y in the processor support sections? Just curious. My little microA1 certainly does not need altivec support, so I'm guessing these config files are not really used or are over-ridden somehow.


Thanks for your continuing patience!

SCott

Last edited by scabit on 22-Mar-2013 at 09:41 PM.
Last edited by scabit on 22-Mar-2013 at 09:23 PM.

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

@Geri

For fun, I tried generating a new kernel 3.2.0 image by modifying my makebootimage.sh from squeeze kernel 2.6.32-5 so that it uses the correct names for the 3.2.0 cuimage and initrd image, kept the same parameters, and executed the script. It generated a kernel image 8.2 megabytes in size. I added a line to my ever growing a1boot.conf to load this 3.2.0 image and mount my squeeze root at hda10 and rebooted.
I select the a1boot.conf line to start squeeze kernel 3.2.0, and the kernel image loads. THen I see a rapid flash of byte size hexidecimal data dumped on the screen, the the micro A1 reboots.

Guess I need some different parameters in my makebootimage.sh file?

Scott

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

@scabit

Quote:

scabit wrote: @Geri Hmmm...one important thing I need to mention. The local.conf file is not getting run until AFTER gdm3 starts up. So when the sata buses are being tested, the force commands in the local.conf file have not yet been applied. I know this because when I had errors in line 10, the error messages appeared after the gdm3 startup, well after the sata buses had been tested and failed.

The initramfs image and afterwards the boot image need to be regenerated whenever you change something in the /etc/modprobe.d/ files. Otherwise the modules included in the initramfs will not see the changes and the new options will not get applied at all.

Quote:

Quote: If I understand you correctly, then you copied a 4.5GB harddisk (not a single partition) to a 24GB harddisk No, it was a 10 Gig partition of which 5.30 gig was used (sorry, I wrote 4.5), to a 24 Gig partition that was formatted blank as ext3.

Now it's getting complicated.

Quote:

The exact dd command I used to copy my working LENNY partition to my new drive was: dd if=/dev/sda2 of=/dev/hda12

Okay. But did you resize the filesytem to 24G after the dd copy?

Quote:

And I must admit that at least the PATA portion of the Syba card (sda2) is working great. That is, I have never to my knowledge had any errors copying to or from the sda devices using the PATA 40 pin EIDE cabled drive. The dd command however from the sda2 partition to the hda12 partition brought over what gparted shows as a lot more data...instead of 5.30 Gig I ended up with something like 21 Gig.

I'm not sure how gparted comes to this number. What does "df --si" tell you about the filesystem size?

Quote:

As for kernel 3.2.0 - I downloaded it and did a dpkg --install. I have what looks like an initramfs image for it in boot. The is also a cuImage-amigaone-3.2.0 etc in /boot as well. Do I need to use a mkimage now to generate a bootable image file for the AmigaOne? If so, what are the start and entry parameters I need, (i.e. was 0x04000000 and 0x04000440 in 2.6.31-2). Is this where I need to start, with the mkimage command?

Basically you can copy the mkbootimg-* command for the 2.6.32 kernel and adapt it for the 3.2.0 kernel:

Quote:

mkimage -A ppc -O Linux -T multi -C none -a 0x04000000 -e 0x04000454 -n "Debian GNU/Linux 6.0 Squeeze" -d cuImage.amigaone.img-3.2.0-0.bpo.4-book3s-amigaone+:initrd.img-3.2.0-0.bpo.4-book3s-amigaone+ boot.img.amigaone-3.2.0-0.bpo.4-book3s-amigaone+

Sorry, I probably looked at an older build for the entry address of the kernel. The addresses above should be correct. Note that the cuImage.amigaone.img file is used for the boot image generation, as it does not yet contain a uImage header.

BTW mkimage can also tell you the load and entry address with the "-l" option:

Quote:

mkimage -l cuImage.amigaone-3.2.0-0.bpo.4-book3s-amigaone+

In this case the cuImage file without the .img extension is used, as it already comes with a uImage header!

Quote:

I tried something else as well since yesterday. I booted from the ETCH install CD and partitioned and formatted my hda12 drive using the Etch partitioner. I then rebooted, went into Squeeze, mounted sda2 and hda12 and used nautilus to copy (in superuser mode) all of the files from sda2 to hda12. hda12 now shows 5.57Gig, sda2 was 5.30Gig. Closer to what I expect.

I usually use "cp -a" for such a job. Nautilus should do the same (in superuser mode), but I trust cp more.

Quote:

But when I boot into sda12 LENNY, everything starts to boot up but stops after the root file system at hda12 has been mounted with an error about detecting the keyboard. It just sits there, CTRL-ALT-DEL and CTRL-ALT-F1 do nothing. I must hard reset.

Do you use a USB keyboard? (I can't remember, if I already asked this...)

Quote:

Why do I want to install LENNY on my 80Gig hard drive you ask? So I can test the SATA interface from LENNY that worked without having to recable anything - the hardware setup will be exactly the same as what I am trying to do for Squeeze. i.e. 80Gig drive connected to internal IDE bus, SATA connected to the 160Gig drive from the Syba card. No SIL card installed. If LENNY works with SATA and SQUEEZE does not - without any hardware change, then maybe we can narrow down the issues in software. I just want to establish that the hardware configuration is correct and working.

It definitely makes sense.

Quote:

Any ideas what is missing from my LENNY partition when I just use copy to move all the files over? It is doing essentially a cp -r since all subdirectories are populated.

Hmm, I could image some files in /dev haven't been copied correctly, but on the other side they are mostly generated automatically. Did you adapt /etc/fstab for the new root partition?

Quote:

Oh, why is it that the config files in /boot for kernels 2.6.31, 2.6.32 and 3.2.0 all have this line: CONFIG_ALTIVEC=y in the processor support sections? Just curious. My little microA1 certainly does not need altivec support, so I'm guessing these config files are not really used or are over-ridden somehow.

These config files define which parts of the kernel source are compiled into the kernel and as modules. Altivec support is included in the kernel, but naturally it will/can only be used, if the processor supports it. It wont hurt in your case.

Quote:

scabit wrote: @Geri For fun, I tried generating a new kernel 3.2.0 image by modifying my makebootimage.sh from squeeze kernel 2.6.32-5 so that it uses the correct names for the 3.2.0 cuimage and initrd image, kept the same parameters, and executed the script. It generated a kernel image 8.2 megabytes in size. I added a line to my ever growing a1boot.conf to load this 3.2.0 image and mount my squeeze root at hda10 and rebooted. I select the a1boot.conf line to start squeeze kernel 3.2.0, and the kernel image loads. THen I see a rapid flash of byte size hexidecimal data dumped on the screen, the the micro A1 reboots. Guess I need some different parameters in my makebootimage.sh file?

Please take a look at the script above.

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

- A1 Linux support -

@Geri

Quote:

The initramfs image and afterwards the boot image need to be regenerated whenever you change something in the /etc/modprobe.d/ files.

Ahh...okay. I'll need to try that.
Ignoring the dd copy problems I've had (for now), I used your mkimage script parameters to generate a working Squeeze 3.2.0-0 kernel which I have booted into and am using right now!

What changes exist between 2.6.32 and 3.2.0 that I should see given my particular hardware? That is, would you expect more success with the Syba USB2, or with the SATA I/O?

Okay...just modified this post. I forgot to add the VSATA commands back into my UBOOT bootcmd parameters. I am hooked up using the Syba EIDE port and it is working, but drives are listed as sdax and sdbx, instead of hda and sda as I am accusomted to with the previous kernels...so something has changed! I'll try the SATA port next....Nope... SATA fails just like with kernel 2.6.31 or 2.6.32. Retries during bootup then failed.

You also mentioned the /etc/fstab ... okay...here's a dumb question. Which one? When I go to boot my /boot directory with the a1boot.conf and the kernel image files, it uses these files to boot onto a different partition. So for example I may be booting from hda2 which has the /boot with the proper a1boot.conf and image files into root partition hde10, which has the squeeze install. Which /etc/fstab needs to have all my devices listed, and how can it adapt for when I boot from different partitions ...i.e. I sometimes boot from a partition on the SIL0680 160G drive, and sometimes from the internal IDE 80Gig drive. That's why I have been using either "LABEL=SQUEEZE" or "UUID=...." in my a1boot.conf files to ensure it finds the correct root drive.
No, I do not have a USB keyboard.

I'll get more to you later as I continue to test the newkernel 3.2.0.

THanks!

Last edited by scabit on 25-Mar-2013 at 08:52 PM.
Last edited by scabit on 25-Mar-2013 at 08:43 PM.
Last edited by scabit on 25-Mar-2013 at 07:03 PM.

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

@Geri
@Hypex

Okay...finally got my LENNY install on my 80Gig partition working.

1) After formatting the partition using an old ETCH install CD, and copying all the files from my working LENNY partition on my 160G drive partition to my 80G drive partition using cp -a,
2) I also had to compress everything under the /dev folder of my working LENNY partition and uncompress it into my new LENNY partition to get files such as tty1 etc
3) I added an entry to a1boot.conf on my booting 80G linux partition as follows:

Lenny_HDA12 /boot/uImage-2.6.18 root=/dev/hda12 l2cr=0x80000000
(where hda12 is the new LENNY partition into which I wish to boot)

4) I then booted into LENNY on hda12. Still lots of errors when I try to log on using GDM. So I went to a shell using CTRL-ALT-F1 and created a new user.
5) Restarted the login and logged on as the new user. All works!

Now to verify SATA works with this configuration.
Geri - I emailed you a pile of files and data showing what the working SATA configuration on my LENNY partition looks like...perhaps it will give you a clue as to why SATA is failing with the Squeeze 2.6.31, 2.6.32 and 3.2.0 kernels.

Scott


EDIT: Aha! Just tried using the SATA port in LENNY on my 80 Gig drive and it fails just like in Squeeze as shown here:

sata_via 0000:00:08.3: version 2.0
sata_via 0000:00:08.3: routed to hard irq line 9
ata1: SATA max UDMA/133 cmd 0x8000 ctl 0x800A bmdma 0x8080 irq 9
ata2: SATA max UDMA/133 cmd 0x8020 ctl 0x802A bmdma 0x8088 irq 9
scsi0 : sata_via
ata1: port is slow to respond, please be patient
ata1: port failed to respond (30 secs)
ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310)
ATA: abnormal status 0x80 on port 0x8007
scsi1 : sata_via
ata2: SATA link down (SStatus 0 SControl 310)
ATA: abnormal status 0x7F on port 0x8027

So, the SATA does NOT work on my 160G Seagate hard drive, but works okay on my 80 Gig drive. It may not be a problem with Squeeze at all, but with the SATA not working right with this Seagate drive!!!

Last edited by scabit on 27-Mar-2013 at 08:25 PM.

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

@all

Wohoo!

I pulled an old 8 Gig IDE hard drive, partitoned it in AmigaOS4.1 as two partitions, set up the machine with it attached as a regular EIDE drive and booted into Squeeze, then formatted the partitions with ext3 on this 8 Gig drive.
Powered down everything. Disconnected the 8G drive from the PATA port of the Syba card and plugged it into the SATA port using the EIDE/SATA converter I have.
Booted up into LENNY and verified that it saw the SATA 8G drive partitions. Yes! It did.
Booted into Squeeze kernel 2.6.32 (dated 3-20-13) and it too sees the SATA drive partitions. I am able to read and write to them just fine!

Hurray! So it looks like Squeeze with kernel 2.6.32 DOES work with both EIDE and SATA I/O on the VIA SATA controller of my Syba card.
Note: Kernel 2.6.31 does NOT work with SATA.

EDIT: The new kernel 3.2.0 also seems to fail and stop while booting when this SATA drive is plugged in. I'll test some more. Wait...it turns out, as with several Linux things, that restarting/rebooting causes the boot process to progress further each time until now, happily, kernel 3.2.0 DOES see the SATA drive partitions.

Now to experiment with USB2 some more!

Thanks!

Scott

Last edited by scabit on 28-Mar-2013 at 02:34 PM.
Last edited by scabit on 28-Mar-2013 at 12:56 PM.

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

@scabit

Quote:

scabit wrote: @Geri What changes exist between 2.6.32 and 3.2.0 that I should see given my particular hardware? That is, would you expect more success with the Syba USB2, or with the SATA I/O?

No, I just hoped it could fix the SATA link problem. But I don't expect it to fix the USB2.0 problem.

Quote:

Okay...just modified this post. I forgot to add the VSATA commands back into my UBOOT bootcmd parameters. I am hooked up using the Syba EIDE port and it is working, but drives are listed as sdax and sdbx, instead of hda and sda as I am accusomted to with the previous kernels...so something has changed!

Yes, the 3.2 kernel doesn't use the old IDE drivers anymore, but instead comes with a libata based driver for the onboard IDE controller (pata_via). Thus the drives are named "sdx" now.

Quote:

I'll try the SATA port next....Nope... SATA fails just like with kernel 2.6.31 or 2.6.32. Retries during bootup then failed.

Okay. I found a developer note in the meantime that some revisions of VIA SATA controllers don't like hard-resets of a SATA link. If I understand it correctly the sata_via controller was changed to issue a hard-reset by default (as indicated by the "COMRESET" message) now instead of a soft-reset. Thus I still hope the libata.force=nohrst option can help here.

Quote:

You also mentioned the /etc/fstab ... okay...here's a dumb question. Which one? When I go to boot my /boot directory with the a1boot.conf and the kernel image files, it uses these files to boot onto a different partition. So for example I may be booting from hda2 which has the /boot with the proper a1boot.conf and image files into root partition hde10, which has the squeeze install. Which /etc/fstab needs to have all my devices listed, and how can it adapt for when I boot from different partitions ...i.e. I sometimes boot from a partition on the SIL0680 160G drive, and sometimes from the internal IDE 80Gig drive. That's why I have been using either "LABEL=SQUEEZE" or "UUID=...." in my a1boot.conf files to ensure it finds the correct root drive.

Well, the root partition you specified in the a1boot.conf file (no matter whether it is a UUID, LABEL or plain device node specification) should also be listed as root partition in the /etc/fstab file on the partition the kernel should boot into.

Quote:

scabit wrote: @Geri @Hypex Okay...finally got my LENNY install on my 80Gig partition working. 1) After formatting the partition using an old ETCH install CD, and copying all the files from my working LENNY partition on my 160G drive partition to my 80G drive partition using cp -a,

I could debug the unrecognized boot partition problem in the meantime. After a whole day of debugging (and copying around whole partitions) it turned out that the inode size needs to be fixed to 128 bytes to make the partition readable by the slb_v2 bootloader. So, if you need to recreate the boot partition again, then you should try it with the "-I" option of the mkfs.ext3 command under Squeeze or Lenny:

Quote:

mkfs.ext3 -I 128 /dev/sdxy

Quote:

2) I also had to compress everything under the /dev folder of my working LENNY partition and uncompress it into my new LENNY partition to get files such as tty1 etc

Interesting. I don't think I ever had to do this, when copying a whole system installation.

Quote:

3) I added an entry to a1boot.conf on my booting 80G linux partition as follows: Lenny_HDA12 /boot/uImage-2.6.18 root=/dev/hda12 l2cr=0x80000000 (where hda12 is the new LENNY partition into which I wish to boot) 4) I then booted into LENNY on hda12. Still lots of errors when I try to log on using GDM. So I went to a shell using CTRL-ALT-F1 and created a new user. 5) Restarted the login and logged on as the new user. All works! Now to verify SATA works with this configuration. Geri - I emailed you a pile of files and data showing what the working SATA configuration on my LENNY partition looks like...perhaps it will give you a clue as to why SATA is failing with the Squeeze 2.6.31, 2.6.32 and 3.2.0 kernels.

Yes, I dug through them. Unfortunately I couldn't find any differences or abnormalities compared to the 2.6.32 logs.

Quote:

EDIT: Aha! Just tried using the SATA port in LENNY on my 80 Gig drive and it fails just like in Squeeze as shown here: sata_via 0000:00:08.3: version 2.0 sata_via 0000:00:08.3: routed to hard irq line 9 ata1: SATA max UDMA/133 cmd 0x8000 ctl 0x800A bmdma 0x8080 irq 9 ata2: SATA max UDMA/133 cmd 0x8020 ctl 0x802A bmdma 0x8088 irq 9 scsi0 : sata_via ata1: port is slow to respond, please be patient ata1: port failed to respond (30 secs) ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 310) ATA: abnormal status 0x80 on port 0x8007 scsi1 : sata_via ata2: SATA link down (SStatus 0 SControl 310) ATA: abnormal status 0x7F on port 0x8027 So, the SATA does NOT work on my 160G Seagate hard drive, but works okay on my 80 Gig drive. It may not be a problem with Squeeze at all, but with the SATA not working right with this Seagate drive!!!

I only know about problems with sil311x SATA controllers and Seagate SATA drives. So it looks more like Seagate SATA drives do not really follow the specification!?

Quote:

scabit wrote: @all Wohoo! I pulled an old 8 Gig IDE hard drive, partitoned it in AmigaOS4.1 as two partitions, set up the machine with it attached as a regular EIDE drive and booted into Squeeze, then formatted the partitions with ext3 on this 8 Gig drive. Powered down everything. Disconnected the 8G drive from the PATA port of the Syba card and plugged it into the SATA port using the EIDE/SATA converter I have. Booted up into LENNY and verified that it saw the SATA 8G drive partitions. Yes! It did. Booted into Squeeze kernel 2.6.32 (dated 3-20-13) and it too sees the SATA drive partitions. I am able to read and write to them just fine!

So it really seems to be a problem with specific drives!?

Quote:

Hurray! So it looks like Squeeze with kernel 2.6.32 DOES work with both EIDE and SATA I/O on the VIA SATA controller of my Syba card. Note: Kernel 2.6.31 does NOT work with SATA.

Strange! Does the 2.6.31 still show the "COMRESET failed" messages?

Quote:

EDIT: The new kernel 3.2.0 also seems to fail and stop while booting when this SATA drive is plugged in. I'll test some more. Wait...it turns out, as with several Linux things, that restarting/rebooting causes the boot process to progress further each time until now, happily, kernel 3.2.0 DOES see the SATA drive partitions.

Which drive are you referring to? The 160GB drive?

_________________
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 explaining the drive names and why they changed between kernel versions.
Wow! You figured out the formatting issue under Lenny and Squeeze! So simply using the command:
mkfs.ext3 -l 128 /dev/hdxx will format a partition readable by the SLB2!

Good work!

When I tried to copy files from one Lenny partition to another, I got errors with several files. I was in superuser mode, and neither source nor destination drive were my current root partition, but something about the /dev/ entries would not allow me to copy them. That's why I had to go back and gz them from the source then un-gz them to the destination. That somehow allowed them to be copied properly.

Quote:

So it really seems to be a problem with specific drives!?

Yes! My Western Digital 8 gig and WD 80 Gig work just fine with my IDE to SATA adapter connected to the SATA port of the Syba card, while the Seagate 160 Gig does not work at all and never once has.

Quote:

Does the 2.6.31 still show the "COMRESET failed" messages?

No, when I boot with a SATA Drive (that works in Lenny) connected using Squeeze with kernel 2.6.31, I get an error: Host system error- PCI problems? Unable to handle kernel paging address: 0x00000050

The boot process just locks up there.

With Squeeze kernel 2.6.32 I get 3 retries of the SATA bus, then the boot process continues and all is well. Similarly with Squeeze and kernel 3.2.0. Note - the test version of 2.6.32 I created does NOT work with SATA...can't recall the exact problem. I know I didn't get it working though, and I tried several times.

Quote:

Which drive are you referring to? The 160GB drive?

No, when I test the working SATA ports now I am using the 8 Gig Western Digital drive with two EXT3 partitions I created.

I have yet to try BOTH the Syba EIDE PATA and the SATA port being connected to drives at the exact same time. I am afraid it would fry my AmigaOne power supply, which is only 250 Watts. !!!

So, we are now doing really well! Of the four items that were to be examined:

1) Determine why SLB_V2 does NOT see partitions formatted by Lenny/Squeeze - SOLVED

2) Get VIA PCI card EIDE working - SOLVED in kernel 2.6.31, 2.6.32 and 3.2.0 Squeeze kernels

3) Get Syba SATA working - SOLVED in kernel 2.6.32 and 3.2.0 Squeeze

4) Get Syba VIA USB2 working - STILL TO BE SOLVED

3 out of 4 items now SOLVED! Great work Geri!

Should I try to use some of the UBOOT byte writes you suggested to try to get the Syba VIA USB2 working? Do you have any other ideas? I'd love to get this last item working.

Thanks!

Scott

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

@scabit

Quote:

scabit wrote: @Geri Thanks for explaining the drive names and why they changed between kernel versions. Wow! You figured out the formatting issue under Lenny and Squeeze! So simply using the command: mkfs.ext3 -l 128 /dev/hdxx will format a partition readable by the SLB2!

Yes, it works on my A1SE even after several reformats. Still an independent confirmation would be good!

Quote:

Good work!

It took long enough!

Quote:

When I tried to copy files from one Lenny partition to another, I got errors with several files. I was in superuser mode, and neither source nor destination drive were my current root partition, but something about the /dev/ entries would not allow me to copy them. That's why I had to go back and gz them from the source then un-gz them to the destination. That somehow allowed them to be copied properly.

The only problem I could think of, would be that the filesystem was mounted with the "nodev" option. This would forbid the filesystem to interpret character or block special devices. But I doubt that this would prevent a copy.

Quote:

Quote: So it really seems to be a problem with specific drives!? Yes! My Western Digital 8 gig and WD 80 Gig work just fine with my IDE to SATA adapter connected to the SATA port of the Syba card, while the Seagate 160 Gig does not work at all and never once has.

How old is this 160GB SATA drive?

Quote:

Quote: Does the 2.6.31 still show the "COMRESET failed" messages? No, when I boot with a SATA Drive (that works in Lenny) connected using Squeeze with kernel 2.6.31, I get an error: Host system error- PCI problems? Unable to handle kernel paging address: 0x00000050 The boot process just locks up there.

When does this message appear? When the sata_via driver is loaded?

Quote:

With Squeeze kernel 2.6.32 I get 3 retries of the SATA bus, then the boot process continues and all is well. Similarly with Squeeze and kernel 3.2.0. Note - the test version of 2.6.32 I created does NOT work with SATA...can't recall the exact problem. I know I didn't get it working though, and I tried several times.

So another Squeeze installation (the test version) with a 2.6.32 still doesn't work with SATA?

Quote:

I have yet to try BOTH the Syba EIDE PATA and the SATA port being connected to drives at the exact same time. I am afraid it would fry my AmigaOne power supply, which is only 250 Watts. !!!

I think the uA1 doesn't draw so much power and a harddisk usually takes around 10W. IMHO the power supply should be able to handle it for a short test.

Quote:

So, we are now doing really well! Of the four items that were to be examined: 1) Determine why SLB_V2 does NOT see partitions formatted by Lenny/Squeeze - SOLVED 2) Get VIA PCI card EIDE working - SOLVED in kernel 2.6.31, 2.6.32 and 3.2.0 Squeeze kernels 3) Get Syba SATA working - SOLVED in kernel 2.6.32 and 3.2.0 Squeeze

You solved the last case by yourself!

Quote:

4) Get Syba VIA USB2 working - STILL TO BE SOLVED 3 out of 4 items now SOLVED! Great work Geri!

I still have to dig through the kernel development history. Maybe there are some changes in newer kernels that address this issue. I'll try my best, but don't expect any miracles.

Quote:

Should I try to use some of the UBOOT byte writes you suggested to try to get the Syba VIA USB2 working? Do you have any other ideas? I'd love to get this last item working.

I'll search though the kernel history first, if that's okay. I think some patches were specific to the hardware/workarounds you tried out on your card.

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

- A1 Linux support -

@Geri

OT, but I just thought of an awesome improvement to the way Linux maps drives. What if we had a /dev/rdb/ which used the device names from the Amiga side? Like /dev/rdb/dh0 Obviously my patch to add device name (and SLB reading/writing) support to amiga-fdisk would be extra useful then (in case you've already got the same device name on another drive). But that patch has been ready for years, just need someone to file a bug report about the need for it

I for one think this would be a lot more logical than the current mapping, which is just a bit too simple for my taste.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@Hypex

Okay, while Geri is looking at the VIA USB2 patches for the kernels, I figured I would try going back to playing with Mac On Linux (MOL) and seeing if I can get it working on Squeeze kernel 3.2.0.
I downloaded the two .deb mol files from sourceforge and used dpkg --install to install. Ran molvconfig to set up the video modes. So far so good.
I run "startmol --test" and of course it tells me that I need to recompile the kernel modules for 3.2.0.
This is where I left off last time. I need to get the mol source files and then somehow recompile them for kernel 3.2.0 yes?
I've installed gcc on Squeeze. What else do I need to do to get this rebuilt and working?

Thanks,

Scott

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

@olegil

Quote:

olegil wrote: @Geri OT, but I just thought of an awesome improvement to the way Linux maps drives. What if we had a /dev/rdb/ which used the device names from the Amiga side? Like /dev/rdb/dh0

Good input! This crossed my mind too, when I was changing my Linux setup on the AmigaOne to use filesystem labels. I guess that would need support in the kernel and in udev. Then RDB partition names could be used similar to FS-labels or UUID identifiers:

Quote:

/dev/ directory structure: /dev/disks/by-rdb/HD0 -> /dev/sda1 mount entry in /etc/fstab: RDB=HD0 / ext3 defaults 0 0 or on the kernel command line: root=RDB=HD0

Quote:

Obviously my patch to add device name (and SLB reading/writing) support to amiga-fdisk would be extra useful then (in case you've already got the same device name on another drive). But that patch has been ready for years, just need someone to file a bug report about the need for it

True, that's on my TODO list as well! Unfortunately this list only gets longer, but not shorter.

BTW the debian-installer and IIRC the GNOME partitioning tool build upon libparted, which AFAIK has no knowledge about the RDB format. So there is still a lot of work left...

Quote:

I for one think this would be a lot more logical than the current mapping, which is just a bit too simple for my taste.

Well, I don't think the Linux guys are going to change this mapping, but at least we could make the mapping of RDB partition names to Linux device nodes directly usable for the user.

Last edited by Geri on 30-Mar-2013 at 09:29 PM.

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

- A1 Linux support -

@Hypex

Okay, found I needed to apt-get install autoconf and zlib1g-dev to be able to run the ./autogen.sh.
I run "make" but get errors in the very fist object file it goes to create, checklist.o.
The errors are "expected ')' before '*' token

What now?

Thanks

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

@scabit

Quote:

scabit wrote: @Hypex Okay, while Geri is looking at the VIA USB2 patches for the kernels, I figured I would try going back to playing with Mac On Linux (MOL) and seeing if I can get it working on Squeeze kernel 3.2.0. I downloaded the two .deb mol files from sourceforge and used dpkg --install to install. Ran molvconfig to set up the video modes. So far so good. I run "startmol --test" and of course it tells me that I need to recompile the kernel modules for 3.2.0. This is where I left off last time. I need to get the mol source files and then somehow recompile them for kernel 3.2.0 yes? I've installed gcc on Squeeze. What else do I need to do to get this rebuilt and working?

Uhh, MoL is a different story with newer Linux kernels (as usual).

The old MoL kernel module source code will not compile, because a core kernel function is not exported anymore for use by kernel modules. However MoL can now make use of the KVM (kernel virtual machine) feature of the Linux kernel. Unfortunately I didn't have the time yet to compile it and do a test run. Thus I don't even know, if it is working on the AmigaOne.

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

- A1 Linux support -

@Geri

Quote:

Unfortunately I didn't have the time yet to compile it and do a test run. Thus I don't even know, if it is working on the AmigaOne.

Hmmm...sounds painful. Okay, since I have the delightful ability to boot into Lenny using (I think) kernel 2.6.18, can't I run MOL from that partition? If so, will I need to rebuild anything at all, or just install it and see what it does?

Of course, I'll need to come up with a OS X 10.4 partition or DVD....still looking for something cheap to try out.

Scott

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