https://bugzilla.novell.com/show_bug.cgi?id=356226 Matthias Koenig changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |mkoenig@novell.com AssignedTo|bnc-team-screening@forge.provo.novell.com |mkoenig@novell.com Status|NEW |ASSIGNED -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User jos.baudrez@telenet.be added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c2 --- Comment #2 from Jos Baudrez 2008-01-30 09:22:15 MST --- Created an attachment (id=192367) --> (https://bugzilla.novell.com/attachment.cgi?id=192367) NVIDIA RAID 0 metadata Matthias asked for this info. The complete output hitherto.... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User mkoenig@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c4 Matthias Koenig changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|mkoenig@novell.com |kernel-maintainers@forge.provo.novell.com Status|ASSIGNED |NEW --- Comment #4 from Matthias Koenig 2008-01-31 05:32:21 MST --- Yes, this looks bogus. dmraid currently does not do any writing of metadata. It just searches for metadata on disk, interprets it and sets up a device-mapper table. Can you please give some more information about the hardware (type of mainboard)? The offset and size values are different for 10.2 and 10.3. As the metadata offset for nv should be: #define NV_CONFIGOFFSET ((di->sectors - 2) << 9) If the size of the device is detected wrong dmraid looks at the wrong location for metadata. This still does not explain why there is a mirror raid detected on 10.3. But obviously the problem is the wrong device size detection. Reassigning to kernel team. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User jos.baudrez@telenet.be added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c6 --- Comment #6 from Jos Baudrez 2008-01-31 14:07:59 MST --- Mathias (or to whom it may concern), I had some spare moments, so the bug bite me again. And some investigation showed that it really IS a KERNEL bug! As i told you before, I'm running kernel 2.6.22.5-12, from a regular Susu-repo whose name I forgot (it was under /home:) only with the extension -mine to diff it from a possible stock kernel with the same number. I saw today that the SuSE 10.3 stock kernel on the install DVD is kernel 2.6.22.5-31 !!! only a few numbers apart! The temptation crept ........ So I did 'the thing' and compiled the SuSE 10.3 kernel 2.6.22.5-31 on my machine, with EXACT the same .config as my (earlier -mine) kernel, and guess what.... BINGO - the mirror appeared again ;-) !!! Exact replica of the symptoms under SuSE 10.3 install DVD I described earlier! (COLD) Rebooting under my old faithfull 2.6.22.5-12, and all went well again... So - what's next ? Yours - jos. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User jos.baudrez@telenet.be added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c7 --- Comment #7 from Jos Baudrez 2008-02-14 00:28:16 MST --- And now, for something completely else ;-) Started Sysrescue-cd this morning (based on Gentoo base system - release 1.12.9) for some maintenance. It worked like a charm (as it does most of the times). Only one detail: the kernel there is __2.6.23.08__ ?!?!? an inspection with dmraid -n showed that the offset is correct !?!?! I am lost in kernel-space ..... :-) Grtz - jos. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User oen@dsl.pipex.com added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c9 --- Comment #9 from Occo Eric Nolf 2008-02-18 19:01:56 MST --- Several comments to bug #333753 show there have been differences in RAID detection between 10.2 and 10.3 from the very start. No one ever bothered to take #333753 seriously. It is still critical and unsolved - this bug #356226 is just a clone. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User jos.baudrez@telenet.be added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c11 Jos Baudrez changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|nfbrown@novell.com | --- Comment #11 from Jos Baudrez 2008-02-20 08:06:38 MST --- Tejun & all others ! Alas the libata.ignore_hpa=0 doesn't work either! I stil face a mirror (hey nice one - isn't it ? ;-) ) and have to cold-boot the machine before a see a stripe of light..... I do include the complete findings (method Watson, method!) in the tar.gz file weirdo_mirror. All has been done starting with the rescue-DVD (except gentoo, with is on an USB stick ;-) ) Hope this helps! - and tnx for the effort ! -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User teheo@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c13 Tejun Heo changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |jos.baudrez@telenet.be --- Comment #13 from Tejun Heo 2008-02-20 18:58:19 MST --- Hmm... libata.ignore_hpa isn't being passed to libata properly. <6>ata3.00: HPA unlocked: 145223999 -> 145226112, native 145226112 <6>ata3.00: ATA-7: WDC WD740ADFD-00NLR4, 21.07QR4, max UDMA/133 <6>ata3.00: 145226112 sectors, multi 16: LBA48 NCQ (depth 0/32) <6>ata3.00: configured for UDMA/133 See the HPA unlocked lines? libata.ignore_hpa=0 should be honored if libata is being loaded from initrd. You're using mkinitrd from openSUSE 10.3, right? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User teheo@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c15 --- Comment #15 from Tejun Heo 2008-02-20 20:31:33 MST --- (In reply to comment #14 from Neil Brown)

Re: comment #10.

It would not be easy to get dmraid to try multiple device sizes, but probably it is not impossible. However dmraid is a user-space program that reads the metadata then sets up dm-tables in the kernel. So the two different possible sizes need to be available to userspace. Maybe an 'hpa-size' sysfs attribute???

Yeah, probably.

It would possibly be a lot easier if there was an ioctl to set the apparent device size to match the hpa size, or maybe even two different devices, one for each size (OK, that really is asking too much).

Hmmm... Making two different devices was suggested once by Jeff Garzik too but I worry how the userland would cope with it (ie. udev, hal and friends). It can easily confuse userland tools. I'll think about the ioctl but I think it would be better to keep the HPA region to be accessible whether dm is attached or not and if dm is going to revert the setting after scan, it just makes more sense to use the BIOS size directly instead of locking and unlocking HPA for metadata scanning. I suppose that it can be tedious to try to retrieve matching sysfs node. Does dmraid require sysfs for anything else?

Re: comments #11/#13. As libata is loaded as a module, a boot time kernel parameter won't work. You would need to get a line like:

options libata ignore_hpa=0

into the modprobe.conf in the initrd. Maybe try adding that line to /etc/modprobe.conf or /etc/modprobe.d/libata and rerunning mkinitrd ??

Our initrd is smart enough to pass residual xxxx.yyyy params to modules contained in the initrd although it does produce spurious invalid parameter warning during boot. That mechanism doesn't work for modules loaded by udev after root fs is mounted. I don't think it's the case here but nevertheless please try the suggested config files. Thanks. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User teheo@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c17 --- Comment #17 from Tejun Heo 2008-02-21 19:10:59 MST --- Ah.. if you're booting off the installation media, you need to use different parameter. Options="libata.ignore_hpa=0" The reason why unloading and re-loading doesn't work is because the first module loading unlocks HPA and it doesn't get re-locked on unload. And the reason why soft reboot doesn't fix the problem is you BIOS forgets to restore HPA setting on softreset. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User jos.baudrez@telenet.be added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c19 --- Comment #19 from Jos Baudrez 2008-03-11 02:29:08 MST --- Hello Tejun, Thanks for pointing out the hpa-locking issue. I am confident that this is the solution to the behavior I mentioned before. But I am still unable to pass the option to the kernel for booting. This makes it still impossible to install SuSE 10.3 on my machine. Is there a possible solution to this ? Tnx! -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User jos.baudrez@telenet.be added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c21 --- Comment #21 from Jos Baudrez 2008-03-11 03:22:33 MST --- I am not booting an installed system! I am trying to install SuSE 10.3 on my system, with a striped-RAID already configured, starting from the install-DVD. The installation doesn't see the striped-raid and insists it is an mirrored-raid, which has been pointed out to be the effect of the hpa_unlocking (btw, i suppose: tnx Micro$oft ?). When i boot from the installation dvd (SuSE 10.3 - SuSE 11-alpha2), I can't pass the option 'libata.ignore_hpa=0' to the kernel. I always get as a result:

...

...

...

...

grep -i hpa boot.msg <6>Command line: BOOT_IMAGE=linux initrd=initrd,12801024.spl splash=silent rescue=1 vga=0x31a libata.ignore_hpa=0 <5>Kernel command line: BOOT_IMAGE=linux initrd=initrd,12801024.spl splash=silent rescue=1 vga=0x31a libata.ignore_hpa=0 <3>Unknown boot option `libata.ignore_hpa=0': ignoring <6>ata3.00: HPA unlocked: 145223999 -> 145226112, native 145226112 <6>ata4.00: HPA unlocked: 145223999 -> 145226112, native 145226112 <6>ata5.00: HPA unlocked: 976771055 -> 976773168, native 976773168

So the striped-array is never recognized, and I can't install ..... I understand the /etc/sysconfig/kernel is not necessary now ? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User jos.baudrez@telenet.be added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c23 Jos Baudrez changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|jos.baudrez@telenet.be | --- Comment #23 from Jos Baudrez 2008-03-11 04:15:13 MST --- ... And that solved the problem completely ! Which proves life can be simple ;-) Everything works like a charm .... and although I know such a raid is a-poor-man-solution, it gives a very responsive machine ..... again ;-) A GREAT BIG THANK YOU ! Jos. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=356226 User nfbrown@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=356226#c25 Neil Brown changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED --- Comment #25 from Neil Brown 2008-05-26 22:15:38 MDT --- I'm closing this as 'FIXED'... even though we didn't actually fix anything in the SuSE distro... I hope that is OK. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.