Installer crashing with 'SCSI 1:0:0:0: rejecting I/O to dead device'

[beranger5ca@…]

Well, the installer failed after being for a few seconds at "Copying files: 11%": it crashed, then the panels crashed, then the whole system was dead.

Rebooting and trying it over was useless.

Rebooting and trying the "Test CD for defects" was also useless: "No md5sums file found on live media." Oops! (Of course I have checked the MD5 sum of the downloaded ISO file, but it would have been nice to have the LiveCD self-check working too!)

So I assumed the CD is OK. After all, the live session works perfectly!

Rebooting one more time and selecting the text install menu option cleared up a little the mystery: what happens past the 11% of files copying is a huge amount of lines saying this:

SCSI 1:0:0:0: rejecting I/O to dead device

No further details, AFAICT.

My disks are IDE, but since newer kernels are treating everything as SCSI/SATA... it's not my fault, you know.

Debian Etch works with this system, but it's based on the kernel 2.6.18, still using the legacy IDE driver.

So I couldn't install Parsix on the desktop PC, to my deep regret.

What else could I try?

I attempted to install it with:
-- / on /dev/sdb3
-- swap on /dev/sdb1
Note: sda and sdb are HDDs; sdc and sdd are DVD and CD-RW units.

[alan@…]

Hi,

Thanks for reposting the bug, output of the following commands would be helpful for us:

# lspci
# fdisk -l
# dmesg | grep ATA

Also would you please check the downloaded ISO's MD5 checksum?

# md5sum /path/to/isofile

Compare it with MD5 files in the download section.

Alan

[beranger5ca@…]

would you please check the downloaded ISO's MD5 checksum?

I have checked it, it's correct, and this is stated in the initial report.

I'll come back in ~6-7 hours with lspci, etc. etc.

I will also burn a second CD -- the MD5 sum of the ISO file is correct, but maybe the burning had some errors.

[beranger5ca@…]

lspci

[beranger5ca@…]

lspci -vv

[beranger5ca@…]

dmesg

[beranger5ca@…]

dmesg | grep ATA

[beranger5ca@…]

fdisk -l

[beranger5ca@…]

Attachments show the required system info.

I have burned another CD, with the same image that I downloaded from another server (because I deleted the first one). md5sum was checked and correct.

This time, the GUI installer crashed at 22% (maybe there was more free RAM this time?!), and the whole Live system was left in a corrupt state (no path to anything, corrupted mounted livesystem).

The text installer went not up to 68% before raising a bunch of:
SCSI 1:0:0:0: rejecting I/O to dead device

I tried mount, just to obtain a lot of:
SQUASHFS: sb_bread failed reading block
SQUASHFS: Unable to read code block

I still don't think the burned CD is corrupt, as the Live system works perfectly once you set the correct screen resolution! OTOH, the LiveCD can't check for itself the md5sum...

Ideas?

[alan@…]

OK,

I checked the attachments and you system has nothing strange. So please do me a favor and test the following methods:

1) try to install in text mode
2) try to install in text mode + nodma ( press F6 when you chose text mode installation from the menu and type nodma and press enter)
3) I searched a little, this error message is somehow a strange thing, is it possible to test with another CD-ROM/DVD-ROM drive?

FYI: You can always set your desired screen resolution with F6 key on the live CD boot menu and typing xscreen=XxY eg, xscreen=1440x900...

Thanks,
Alan

[beranger5ca@…]

1) try to install in text mode

I did this last night and (as posted), this time the text installer went not up to 68%. Weird.

I'll check one more time with "nodma" and possibly some other options (noacpi, noapic).

is it possible to test with another CD-ROM/DVD-ROM drive?

I could swap the two DVD/CD-RW units. Notice however that I have used both of them (with no preference) to install *dozens* of different Linux distros in the last 2-3 years and I haven't had any such error. For instance, I just installed Vector 5.9 on the same PC, burned on the same CD burner, etc.

xscreen=1440x900

I wasn't aware that it would honor such options. But there is one more issue, not mentioned at Bug 82: "ati" defaults to "fglrx", but in my case the 100% free "radeon" driver performs better. Does it honor "xmodule=radeon"? (I hope so.)

I'll post back results in ~6-7 hours.

[beranger5ca@…]

ISSUE SOLVED. "nodma" made the installation possible!

The strange part is that a relatively similar kernel, Fedora's 2.6.23.1 works in similar conditions, i.e. installing from a LiveCD. There is something wrong with Parsix's kernel.

I wonder however how should I use the installed system! Is it safe to use the DMA now?!

Maybe I could force the legacy libata (but I guess sdX would revert to hdX) and use one of the following options:
libata.dma=0 Disable all PATA and SATA DMA
libata.dma=1 PATA and SATA Disk DMA only
libata.dma=2 ATAPI (CDROM) DMA only
libata.dma=3 enables DMA for disks and CDROMs, but not CFs.
libata.dma=4 Compact Flash DMA only

Or maybe I should keep things as is (on the installed system), and disable DMA on the optical units with something like:
sdc=nodma sdd=nodma

[alan@…]

OK

Just remove nodma from the GRUB's kernel line in the installed system. It's now safe to enable DMA. It was a weired problem!

[alan@…]

Would you please check if this problem has been solved with Viola TEST-1?

[beranger5ca@…]

Alan, the problem is *NOT* solved with the same PC hardware! (Not a new PC, on the contrary, older than a laptop on which the CD is installable.)

Unfortunately, this time nodma does *NOT* help :-( So I could not install it on the PC :-(

[mldoering@…]

as a follow-up some more libata woes:

I was unable to install/update Viola on a Via C3 Notebook locking up while installing. HW might have some quirks, but I was able to circumvent those by disabling dma for the HD with e.g. hdparm on Ramon.

Since the change to libata/pata_via in Kernel unfortunately setting dma to off with hdparm isn't supported anymore leaving me a bit stranded. Kernelparameters I already tried ist acpi=off noapic and nolapic but to no avail.

Any suggestion the get leagcy ide support working with Violas Live CD would be appreciated.

thanks in advance

Michael "mld" Doering
​http://zakalwe.fi/uade/

[alan@…]

As we shipped a brand new installer series, we will be happy if you are still interested to test if this issue has been resolved or no. Please wait for Parsix Kev TEST-2. Thanks.