What is the invalid command? Please, provide more information.

Reassigned to the kernel team.

Please provide hwinfo output from the system as well as kernel logfiles regarding the failure.

No further feedback. Please re-open if more information is available.

Created attachment 60437 [details]
Log from failure

Sorry it took so long to get. The command that fails says in the console log it is "unknown". Note that this is a dual boot machine and the disks are all NTFS partitions and run fine under XP SP2 plus all latest fixes as of today added. If you want more info can to tell me how to get it at it took ages to find out how to boot the rescue system and then mount the real root and a windows FAT partition to move the dafa across. 
Cheers, Andy

Re-open after data supplied

Thanks for the feedback, but something went wrong with the attachment, sorry. It appears to be empty and just consists of ascii zeroes... Could you please try attaching it again after making sure it's the right file indeed and displays OK locally?

Thanks!

Created attachment 60577 [details]
Log of failure

Log file provided. Strange - this time it got futher and loaded up CD's 2 -> 5 and then failed on reboot. The end of the console log is as follows...

hdf  dma_timer_expiry dma_status 0x60
hdf  dma timer retry
hdf  timeout waiting for dma
hde  status_error   status 0x58 {driveready seekcomplete datarequest}
ide  failed opcode was : unknown
cpu 0 machine check exception exception 4 bank 4 b200000000070f0f
tsc 26272ad37d
kernel panic - not syncing: machine check


Note I have run memory test on the machine (from the install disk) for 24 hours with no errors - just in case.....

Thanks, Andy

Sorry - I had the level of Linux incorrect but it did also fail on 10.0 the same way but as I am now on 10.1 Alpha (in the hope it would fix the problem) that is the one I will continue to use unless a newer beta comes out. Also added the chipset type IT8212

Created attachment 61131 [details]
dummy file

Have tried the Alpha 4 release but get the same problem. Have also tried this on another Athlon 64 processer machine with the same problem. It has its own 8212 card. 

Andy Norrie

Something's fishy here indeed. The log says:

<6>hde: 398297088 sectors (203928 MB) w/8192KiB Cache, CHS=24792/255/63, BUG

Note the BUG at the end of the line.

Jens, any idea why it would do that?

The BUG shows up because the drive has some weird values set in its IDENTIFY page. This is usually where you see the UDMA/DMA mode listed.

Does the failsafe boot change anything? It looks like problems with either dma or irq delivery.

Checking the ATA spec, it looks really strange indeed. word53 is the one that ide-dma checks to see what modes are supported, and the drive doesn't list either ultra or pio modes as valid. Only bit 1, which is the basic cyl/head/sec values.

Oh, it's the it821x driver doing these 'fixups' for non-raid volumes that aren't in pass-through mode. It actually looks like a very inferior mode looking at the driver, the problem is most likely that dma gets enabled which it wont support in this configuration.

Any chance you can change this setup?

Created attachment 61950 [details]
Disable DMA if not listed as supported

This might help the driver at least talk to the drives in this configuration.

Created attachment 61959 [details]
ITE Raid driver that USED to work in case it helps

Details on possible working driver attached.

It's not of much help I"m afraid, as it's the older SCSI variant. It really cannot help fix the current issue with the IDE layer and the new driver.

Good news, I think. I have applied the patch and all seems well however I cant be 100% sure
as this is the first time I patched the Kernel.

I applied the patch and it created an IDE_DMA.C.ORIG file.

I then did a make bzImage

Then added an extra section in grub to boot that kernel.
I made the grub stuf identical to normal boot except for the kernel name

I IPL'ed the new Kernel and I could get read/write access to the drives there
Read access to the NTFS ones and read/write to the FAT one.

If there are any other things I can test etc give me a shout via Email
at  GM1MQE@AOL.COM

Very many thanks, Andy Norrie

Thanks for testing, the patch will be in 10.1 (I have committed it).

Created attachment 63609 [details]
Possibly better patch

Any chance you can test this patch _instead_ of the previous one?

Yes I will but do I remove the origional patch first?

Yes, it's very important that you remove the old one first. You should see the BUG thing at boot again when you do that, so the newly built kernel with the replacement patch should display that again.

Been unable to get the patch supplied to apply to alpha-4 level. It looks as if its to do with the directories. Can you please have a look as I dont know much about the patch system. It seems to obgect to the --dif stuff as well.

Hmm, it applies fine and without offset for me:

axboe@nelson:[.]mnt/kscratch/linux-2.6.15 $ patch -p1 --dry-run < ~/it821x-dma-bug 
patching file drivers/ide/pci/it821x.c
axboe@nelson:[.]mnt/kscratch/linux-2.6.15 $ 

What happens for you?

Still cant get patch to fit.... Its erroring on the directories - they dont exist in my system. Sorry. I have downloaded and installed the BETA 1 level. Would it be at all possible to send my via Email a copy of the patched IT821x.c file with the patch installed as I can cope with that?

Next job... Learn how to patch kernels properly!!

Thanks Andy   GM1MQE@AOL.COM

Created attachment 65070 [details]
Patched driver

Here it is, the file with the patch applied.

There you go, please give that a test spin.

That was a fast response! OK I have tried it out and it "works for me". I did see the "BUG" msg at startup time in the log as you warned. I intend to reload the whole Beta 1 version tomorrow, redo the patch and set it all up properly and then give it some serious disk testing on that card.

Many thanks, Andy

Excellent, thanks a lot for testing. I've committed this change now, I wont be there before beta3 though.

Thanks for all your help. I will do a re-test when Beta-3 is available.

Regards, Andy

Just to confirm it all works beautifully on Beta 6. Sorry I was not able to test earlier but I have been unwell. Thanks for all your help.