Update to 2.6.20-16 kernel breaks SATA in Feisty/Cassandra!

Questions about other topics - please check if your question fits better in another category before posting here
Forum rules
Before you post please read this

Update to 2.6.20-16 kernel breaks SATA in Feisty/Cassandra!

Postby Lolo Uila on Sat Jun 02, 2007 2:13 am

I did a clean install of Cassandra (final) on my "test" system and it installs and restarts fine. Everything seems to function properly, and all drives are accessible. It will continue to run fine if I don't do the "recommended" updates that I get prompted to install.

There is a HAL update that seems to go in fine. Then a moment later I am asked to do 18 more updates, and those too seem to install fine. I get a message that all updates installed successfully and my system is up to date. Then the restart icon appears in the tray telling me a restart is needed.

So I click on the icon and the system shuts down normally, but it won't start up again. The system reboots and the Mint splash screen appears, and when the progress bar is about 1.5 ticks up from the beginning it hangs for a very long time. Eventually the splash screen goes away and I am greeted by part of the text boot process, with a red [fail] message that looks like it's after "Loading hardware drivers..."

There is also another error:
"modprobe: WARNING: Not loading blacklisted module ipv6"

but I see that message a lot, even on systems that boot and run fine.

Then the system does a check of the root file system and that passes, and the next message is "* Checking file systems..." and it doesn't get any further than that.

If I unplug my 2 SATA data drives the system will boot normally.

Something that is being updated is breaking SATA support in Cassandra!

This system is pretty standard Intel hardware. 865/ICH5 chipset with a P4 2.4GHz processor and a gig of RAM. It's got an IDE boot drive and optical drive, and 2 sata drives for extra storage.

Anyone else run into this?

And, is there anything I can do to fix it?
Last edited by Lolo Uila on Mon Jun 04, 2007 4:04 am, edited 1 time in total.
User avatar
Lolo Uila
Level 5
Level 5
 
Posts: 547
Joined: Thu Mar 15, 2007 3:40 am
Location: Kapolei, Hawaii

Linux Mint is funded by ads and donations.
 

Postby D1Wayne on Sat Jun 02, 2007 3:27 am

Hi Lolo Uila

Also have a pair of 37gb SATA raptors and a ide HD

2 boots ago this showed up in a incomplete boot (after latest updates>

perhaps timing issue, syslog and syslog,0 snippets before and after boots

Code: Select all
Jun  1 03:14:47 x-desktop kernel: [   33.903336] scsi2 : sata_nv
Jun  1 03:14:47 x-desktop kernel: [   34.212336] ata3: SATA link down (SStatus 0 SControl 300)
Jun  1 03:14:47 x-desktop kernel: [   34.222758] ATA: abnormal status 0x7F on port 0x000109e7
Jun  1 03:14:47 x-desktop kernel: [   34.222802] scsi3 : sata_nv
Jun  1 03:14:47 x-desktop kernel: [   34.531943] ata4: SATA link down (SStatus 0 SControl 300)


I re-booted and
now only 1 has the abnormal status
links spotted as bugs in kernel
might explain several freezes I've had

Perhaps there is a guru out there that knows enough about sata to be of assistance

if you check your logs and see if you have similar info it might point to something one of the guru could figure out
D1Wayne
Level 3
Level 3
 
Posts: 198
Joined: Tue Apr 24, 2007 6:11 am
Location: Pacific North West

Postby Lolo Uila on Sat Jun 02, 2007 7:09 am

Looks like it's a problem with the new kernel... :(

https://bugs.launchpad.net/ubuntu/+sour ... bug/116996

Hopefully a fix is coming soon.
User avatar
Lolo Uila
Level 5
Level 5
 
Posts: 547
Joined: Thu Mar 15, 2007 3:40 am
Location: Kapolei, Hawaii

Postby Lolo Uila on Sat Jun 02, 2007 7:45 am

Further reading indicates it may be limited to Intel ICH4, ICH5 & ICH6 controllers. So... if you have SATA drives on any of those controllers, you probably should not do the kernel updates until this has been fixed.

IDE drives seem to be fine. If I unplug my SATA drives, or turn off the SATA controller in the BIOS my system boots fine.
User avatar
Lolo Uila
Level 5
Level 5
 
Posts: 547
Joined: Thu Mar 15, 2007 3:40 am
Location: Kapolei, Hawaii

Updates break Feisty/Cassandra for SATA!

Postby Lolo Uila on Sat Jun 02, 2007 2:59 pm

I unchecked the 2.6.20-16 kernel updates, thinking I could just install the other updates, and the kernel updates installed anyway! :x Fortunately it's just my test system, but still.... What's the point of the check boxes in the update manager if it's just going to ignore what I select and install them anyway?

And on top of that, it's still telling me those updates need to be installed, even though they were installed when I unchecked them!

Anyway... if anyone else runs into this problem and can't boot, you can still select the previous 2.6.20-15 kernel from the grub boot menu and that should let you boot. Kind of a pain to have to do that every time you want to boot. Fortunately Linux doesn't need to be restarted very often.

So, to recap:
The current 2.6.20-16 kernel breaks SATA support in Ubuntu Feisty and Mint Cassandra for most Intel ICH4, ICH5 & ICH6 controllers (although apparently not all of them). If you install the auto updates and are running any SATA drives on these controllers it is likely your system will fail to boot (or will boot with lots of problems).

If you run into this problem, hit escape at grub boot and select the previous 2.6.20-15 kernel from the menu and you should be able to get back into your system.

It looks like Debian had similar problems and it has been fixed. Hopefully Ubuntu will be fixed soon.
User avatar
Lolo Uila
Level 5
Level 5
 
Posts: 547
Joined: Thu Mar 15, 2007 3:40 am
Location: Kapolei, Hawaii

Postby Husse on Sat Jun 02, 2007 5:06 pm

The ICH controllers (chipset) do have problems, I'm not at all surprised - they are a pain in Windows as well...
There are some oddities in them and in motherboards that may explain why not all are hit....
Image
Don't fix it if it ain't broken, don't break it if you can't fix it
Husse
Level 21
Level 21
 
Posts: 19710
Joined: Sun Feb 11, 2007 7:22 am
Location: Near Borås Sweden

Postby clem on Sat Jun 02, 2007 6:19 pm

That reminds me of 2.6.17-11... if you don't need to upgrade kernels, hal, dbus, xorg and sensitive things like that.. don't. The hal upgrade was unstable enough not to be included in Cassandra stable. Ubuntu is based on sid, not on testing, and as packages are concerned we're currently based on a non-long-term-support version of it: Feisty, so we can expect them to break a few things now and then... not to mention that all repositories are activated (including backports). Mint releases very frequently... you shouldn't have to upgrade.

In /boot/grub/menu.lst you can make 2.6.20-15 default.

Clem
Image
User avatar
clem
Level 15
Level 15
 
Posts: 5545
Joined: Wed Nov 15, 2006 8:34 am

Postby Lolo Uila on Sat Jun 02, 2007 7:30 pm

I have never had any problems with the ICH controllers in Windows. I have several systems running on ICH5 and ICH5R controllers and they all work fine. My main system runs Win2K on a RAID-0 stripe and it is very fast (and, yes, I am aware of the risks of RAID-0, which should be called IAD since there is no redundancy).

It's surprizing this got out to the update channel. Those controllers are used on a LOT of motherboards, and SATA has become pretty popular in recent years. You'd think someone on the Ubuntu dev team would have that hardware. Oh well...

Clem,

Yes I shouldn't have to, but when I unchecked the kernel updates they installed anyway. The other updates said they were for security so I figured they were important. Either auto-update ignores the check boxes, or it was some kind of dependancy thing that forced the kernel update in spite of my de-selection.

For me this was a minor annoyance since it was only my test system. Fortunately my main box isn't running any SATA drives that Linux can see so the updates didn't mess anything up.

Aloha, Tim
User avatar
Lolo Uila
Level 5
Level 5
 
Posts: 547
Joined: Thu Mar 15, 2007 3:40 am
Location: Kapolei, Hawaii

Postby Husse on Sun Jun 03, 2007 6:26 am

@Lolo Uila
Long time since I had the unpleasant task of working with ICH 5 or 6, but it was unstable and completely impossible to make a decent RAID1 on it. Lost some money as I had to take back a bunch of mobos for that...
That said, it worked well on some motherboards, not on others...
It's not at all fun not to know if things will work or not when you sell the stuff....
Image
Don't fix it if it ain't broken, don't break it if you can't fix it
Husse
Level 21
Level 21
 
Posts: 19710
Joined: Sun Feb 11, 2007 7:22 am
Location: Near Borås Sweden

Postby Kladiator on Sun Jun 03, 2007 7:38 am

I experienced a similar problem with Cassandra: after 24 hours it was broken, I mean nothing was working anymore so I decided to do a fresh install of ... Bianca!
Now I have a question for you guys: is the problem only Feisty related or is it possible that the same disaster will happen also on Bianca?
Should I completely avoid updates?
If and when this bug is fixed, where can I find the announcement?
Thanks.
Kladiator
Level 1
Level 1
 
Posts: 22
Joined: Fri Dec 22, 2006 10:46 am
Location: near Venice

Postby exploder on Sun Jun 03, 2007 8:16 am

I was also a victim of the new kernel, Intel Chipset here too. I only have one sata hard drive and the system would boot. The new kernel gave me an extra CD-ROM drive that did not exist though! Oh well these things happen and I didn't lose any data because of it. Someone might want to create a sticky about this issue and maybe some notes in the Wiki to save other's from having this problem.

Another upgrade subject, if you have upgraded Gimp and would like the new splash screen back here is how to fix it.

From the Live CD go to terminal, sudo nautilus

Go to /usr/share/gimp/2.0/images and copy and save gimp-splash.png

Overwrite the file in the same directory (as root) in your installed version.

It is a minor thing but since someone took the time to make a nice looking splash I thought it was worth the time to put it back.
exploder
Level 15
Level 15
 
Posts: 5872
Joined: Tue Feb 13, 2007 10:50 am
Location: HartfordCity, Indiana USA

Postby D1Wayne on Sun Jun 03, 2007 9:11 am

would that be the extra kernel cd :D
D1Wayne
Level 3
Level 3
 
Posts: 198
Joined: Tue Apr 24, 2007 6:11 am
Location: Pacific North West

Postby exploder on Sun Jun 03, 2007 9:34 am

Nice one! Heck, the kernel Clem mentioned gave me an extra floppy drive on my old system!
exploder
Level 15
Level 15
 
Posts: 5872
Joined: Tue Feb 13, 2007 10:50 am
Location: HartfordCity, Indiana USA

Postby Lolo Uila on Mon Jun 04, 2007 4:21 am

Okay, I spoke too soon. The new kernel messes up my main system as well if the SATA controller is enabled. Even though there are no SATA drives accessible to Linux!

This is related to the second problem with the new kernel. It changes the drive mount ID. If your fstab file has not been edited you may be okay. If you are using /dev/sd... instead of UUID then you will get hit by this bug too. It seems the new kernel is changing the way the drives are mounted, again. Some drives are going back to /dev/hd... while others remain /dev/sd... but are given new letters or numbers. Even without a modded fstab file some people on the Ubuntu forums have reported their optical drives aren't working (optical drives do not have a UUID and can be hit by this bug).

Man this new kernel is really a major screw up.
User avatar
Lolo Uila
Level 5
Level 5
 
Posts: 547
Joined: Thu Mar 15, 2007 3:40 am
Location: Kapolei, Hawaii

Postby Lolo Uila on Mon Jun 04, 2007 7:12 am

Just checked, and I have FOUR optical drives now. Sheesh!

Two of them don't point to anything in particular, of course.

So it looks like my 4 IDE hard drives are still using the "sd" naming convention, but they have been changed from sda/b/c/d, to sdc/d/e/f. Note that these drives are not on the Intel controller; they are on an ITE controller chip. From what I have been reading, drives on the Intel controller get changed back to the "hd" naming convention used prior to Feisty (including optical drives).

Hmmm... my optical drives used to be (from fstab):
/dev/scd0 /media/cdrom0 udf,iso9660 user,noauto 0 0
/dev/scd1 /media/cdrom1 udf,iso9660 user,noauto 0 0

And now they are "/dev/hda" & "/dev/hdc" according to the output of lshw.
description: IDE Channel 0
product: BENQ DVD DD DW1640
logical name: /dev/hda
description: IDE Channel 1
product: YAMAHA CRW-F1E
logical name: /dev/hdc

dmesg shows the same thing...
hda: ATAPI 48X DVD-ROM DVD-R CD-R/RW drive.
hdc: ATAPI 44X CD-ROM CD-R/RW drive

So...
/dev/scd0 is now /dev/hda (primary master drive)
/dev/scd1 is now /dev/hdc (secondary master drive)

Which means my mobile rack drives will be "hdb" and "hdd" when I plug them in. This is really messed up. This update was called a "security" update, but it obviously has some major, untested changes in it and should NOT have been released yet.
User avatar
Lolo Uila
Level 5
Level 5
 
Posts: 547
Joined: Thu Mar 15, 2007 3:40 am
Location: Kapolei, Hawaii

Postby Husse on Mon Jun 04, 2007 7:24 am

Agreed - this is no good.
But I have a nforce4 chipset and no problems what so ever with this update....
Not tested extensively enough
Image
Don't fix it if it ain't broken, don't break it if you can't fix it
Husse
Level 21
Level 21
 
Posts: 19710
Joined: Sun Feb 11, 2007 7:22 am
Location: Near Borås Sweden

Postby Lolo Uila on Mon Jun 04, 2007 7:45 am

Husse, :P

:D

Okay, I changed my fstab to reflect the new optical drive logical IDs and I'm back to just 2 drives with the correct mount points. Of course now they are going to "fix" it by changing things back and screw me up again. :roll:

Anyway... bed time.

Good night all, Tim
User avatar
Lolo Uila
Level 5
Level 5
 
Posts: 547
Joined: Thu Mar 15, 2007 3:40 am
Location: Kapolei, Hawaii

Postby Husse on Mon Jun 04, 2007 8:04 am

That's good.
We're lucky to have a distro that tells you which device is which UUID.
I mean you see something like
# /dev/sda1
#UUID=A42C6E042C6DD23A

I've seen some fstab with just the UUID....
Image
Don't fix it if it ain't broken, don't break it if you can't fix it
Husse
Level 21
Level 21
 
Posts: 19710
Joined: Sun Feb 11, 2007 7:22 am
Location: Near Borås Sweden

Postby screamin_jesus on Wed Jun 06, 2007 5:20 pm

Same thing happened to me, boot just froze, had to boot into the old kernel and uninstall the new one. I have an ICH5 Chipset. I have XP on my main 160 GB SATA drive, and Mint on an 80 GB IDE drive. My specs are: Dell 4600, SOcket 478 Pentium 4, 1 GB RAM, 7600GS, 160 GB WD SATA, 80 GB WD IDE. Ubuntu is great, but sometimes I seriously wonder if they bother to even test updates.
screamin_jesus
Level 1
Level 1
 
Posts: 15
Joined: Tue Jun 05, 2007 5:38 pm

Postby Husse on Wed Jun 06, 2007 6:16 pm

Welcome to Mint screamin_jesus
(well now, do I dare welcome this person :))
Sometimes one wonders that, but I don't know how large resources they have (to test different hardware). Canonical should have some but yet again...
Image
Don't fix it if it ain't broken, don't break it if you can't fix it
Husse
Level 21
Level 21
 
Posts: 19710
Joined: Sun Feb 11, 2007 7:22 am
Location: Near Borås Sweden

Linux Mint is funded by ads and donations.
 
Next

Return to Other topics

Who is online

Users browsing this forum: Bing [Bot] and 7 guests