https://bugzilla.novell.com/show_bug.cgi?id=371657 User jeffm@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c1 Jeff Mahoney changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jeffm@novell.com Status|NEW |NEEDINFO Info Provider| |deanjo@sasktel.net --- Comment #1 from Jeff Mahoney 2008-03-19 11:03:50 MST --- Can you provide the output of "dmsetup table" and also fdisk -l? -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c3 --- Comment #3 from Dean Hilkewich 2008-03-19 23:07:10 MST --- Created an attachment (id=203066) --> (https://bugzilla.novell.com/attachment.cgi?id=203066) output of "dmsetup table" and also fdisk -l Output of dmtable setup and fdisk -l -- 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=371657 Jeff Mahoney changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|kernel-maintainers@forge.provo.novell.com |jeffm@novell.com Status|NEEDINFO |ASSIGNED Info Provider|deanjo@sasktel.net | -- 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=371657 User jeffm@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c6 Jeff Mahoney changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |deanjo@sasktel.net --- Comment #6 from Jeff Mahoney 2008-03-31 09:55:27 MST --- Ok, so I've test up a test environment that consists of 4 sparse files as loop back devices that match the disks in your system: $ ls -la sd* -rw-r--r-- 1 jeffm users 251000194048 2008-03-31 10:37 sda -rw-r--r-- 1 jeffm users 251000194048 2008-03-31 10:37 sdb -rw-r--r-- 1 jeffm users 500107863040 2008-03-31 10:38 sde -rw-r--r-- 1 jeffm users 500107863040 2008-03-31 10:38 sdf Then I created the dm striping environment to match yours: $ dmsetup table | grep nvidia|sort nvidia_bdjbjgci: 0 980469248 striped 2 128 7:0 0 7:1 0 nvidia_bdjbjgci_part1: 0 40965687 linear 253:20 63 nvidia_bdjbjgci_part2: 0 204812685 linear 253:20 40965750 nvidia_bdjbjgci_part3: 0 208845 linear 253:20 245778435 nvidia_bdjbjgci_part4: 0 734475735 linear 253:20 245987280 nvidia_bdjbjgci_part5: 0 41945652 linear 253:24 63 nvidia_bdjbjgci_part6: 0 1028097 linear 253:24 41945778 nvidia_bdjbjgci_part7: 0 691501797 linear 253:24 42973938 nvidia_fjfafgaa: 0 1953546240 striped 2 128 7:2 0 7:3 0 nvidia_fjfafgaa_part1: 0 1953536067 linear 253:28 63 .. and I can't recreate the problem. I seek to the end -3k and write 4k, and get a short write as expected. Can you provide more information as to how you're triggering this? -- 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=371657 User cthiel@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c8 Christoph Thiel changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|deanjo@sasktel.net | --- Comment #8 from Christoph Thiel 2008-04-25 02:39:09 MST --- Info provided in comment #7? -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c9 Dean Hilkewich changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P2 - High Target Milestone|Beta 2 |--- Version|Factory |Beta 2 --- Comment #9 from Dean Hilkewich 2008-05-05 09:04:35 MST --- Still present in beta 2 -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c11 --- Comment #11 from Dean Hilkewich 2008-05-09 11:05:37 MST --- I would have to disagree with the assessment that software raid is a better solution. It has many downfalls when a system has multiple os's on the system. After all isn't "interoperability" one of the goals of Novell/openSUSE? Had this been an issue like grub not being able to load windows I'm pretty sure it would have been marked Critical. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c13 Dean Hilkewich changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |REMIND Version|Beta 2 |RC 1 --- Comment #13 from Dean Hilkewich 2008-06-08 10:26:05 MDT --- An update would be nice. Controllers that this has been found on are Nforce 3 250 Nforce 570 Nforce 780a Silicon Image 3114 -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c15 --- Comment #15 from Dean Hilkewich 2008-06-09 08:57:58 MDT --- OK I have some additional info on this. It seems that it is not isolated to the drives. The issue occurs when a additional raid set is added. If you create a raid 0 set using only two drives (any of the drives) hooked up everything is fine. If you create a dmraid partition afterwards the system gives the above mentioned errors. If you do a install on each of the raid setups individually and then plug them all in they all register successfully and no errors occur when the partitions are mounted. If you try to delete the partitions using disk utility however you cannot remove the extended partition or the /boot partition. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c17 --- Comment #17 from Dean Hilkewich 2008-06-11 08:56:02 MDT --- Here is the solution to this, the partitioner should have this process automated. http://le-gall.net/sylvain+violaine/blog/index.php?2007/12/04/32-solving-the... -- 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=371657 User jeffm@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c18 Jeff Mahoney changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |mkoenig@novell.com AssignedTo|jeffm@novell.com |kasievers@novell.com Status|REOPENED |NEW --- Comment #18 from Jeff Mahoney 2008-06-18 10:51:10 MDT --- Great, thanks for the research. I don't think the partitioner is the right place to do this, though. YaST should make things easier, but not be the only method of partitioning sanely. udev should probably just be smarter about scanning the partition tables on device-spanning fake raids like this. I'm reassigning this to the udev maintainer and adding the dmraid maintainer to the CC to coordinate methods of automatically identifying dmraid component devices. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c20 --- Comment #20 from Dean Hilkewich 2008-09-27 11:38:12 MDT --- Any update, has this been fixed in 11.1? I couldn't tell because of the 11B1 64 being a bad iso and uninstallable. -- 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=371657 User kasievers@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c22 Kay Sievers changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |jeffm@novell.com --- Comment #22 from Kay Sievers 2008-10-07 04:48:28 MDT --- The "attempt to access beyond end of device" is caused by the kernel partition table parser, allowing partitions to be created, which pretend to be larger than the entire physical disk. The result of BLKGETSIZE64 and "size" in sysfs returns invalid values. Userspace could find that out by doing the consistency check, the kernel didn't do, by reading the disk size, reading the start and size of the partition, and check if it's in the limits of the disk. But doing that sound really like the wrong solution. Shouldn't the kernel partition table scanner just prevent the entire and invalid partition from showing up, or at least limit the partition size to the physical end of the disk? Jeff, any idea? -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c24 --- Comment #24 from Dean Hilkewich 2008-10-09 10:41:39 MDT --- Yes, I can do that. You will just have to give me a day to clear off those harddrives and reset up the dmraid on those drives. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c29 --- Comment #29 from Dean Hilkewich 2008-10-13 09:47:47 MDT --- What I don't understand is why this doesn't occur if I have 2 separate raid 0 setups with 2 different OS's installed to each array. For Example: I install openSUSE 11 to one set leaving the other set disconnected. Disconnect the first set, reconnect the second set install opensuse 11 onto those. Reconnect all drives and mount them the errors disappear. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c31 --- Comment #31 from Dean Hilkewich 2008-10-26 00:52:26 MDT --- Thanks for the patch Kay, but do you have an idea why if you have two installations of the OS on separate raid sets that they can easily mount each others raid volumes without the messages appearing? What does the partitioner do different on install to a raid set volume that it does not do for a secondary raid set? -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c33 --- Comment #33 from Dean Hilkewich 2008-10-26 11:40:33 MDT --- (In reply to comment #32 from Kay Sievers)

(In reply to comment #31 from Dean Hilkewich)

...

Thanks for the patch Kay, but do you have an idea why if you have two installations of the OS on separate raid sets that they can easily mount each others raid volumes without the messages appearing? What does the partitioner do different on install to a raid set volume that it does not do for a secondary raid set?

No, I have no idea. The messages should happen when probing for a raid setup (signature at the end of the volume) on any partition that is larger than the disk itself (table on first raid 0 disk, pointing to sectors of second raid0 disk).

Udev will try to read the end of all the partitions the kernel announces. I have no idea why this does not to happen in your case.

Right, therefore the partitioner has to be doing something different when creating a secondary raid set then it does on the primary raid set which leads us back to the partitioner being the cause of the issues with it not writing the table on the first disk properly of the second raid set. -- 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=371657 Kay Sievers changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P1 - Urgent |P4 - Low -- 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=371657 User kasievers@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c36 --- Comment #36 from Kay Sievers 2008-11-17 17:39:13 MST --- (In reply to comment #35 from Dean Hilkewich)

So despite this being around forever as a bug @ a P1 and now being marked as a P4 that this bug will not be fixed for 11.1?

I do not think that the general problem causing this "bug" will be fixed. The problem is that the kernel parses partitions and creates block devices without knowing about the context or the metadata of the volume. All of these partitions should not exist in the first place, because the disk is part of a raid setup. But the kernel does not know that, and creates completely invalid partitions. There are two possible fixes, one would be to teach the kernel about all possible raid metadata, if the kernel wants to continue parsing partition tables. The other option would be to parse partitions only in userspace and not in the kernel. Both options are unlikely to happen _now_. In the upstream kernel we made sure (it's merged, but still tested and not in a released version) that partitions point only to a valid storage area, which will prevent the "access beyond end of device" warning. But still, it's only a cosmetic change to the underlying problem. We can port the upstream "cosmetic" fix to 11.1, if it survives the released upstream kernel (the former fix got removed because it broke some setups), but I do not see any "nice" fix to the general problem. The problem is known and exists since a while, without any good idea to fix it so far. Your specific behavior, that your second array behaves differently from the first one, nobody has an idea why this could be, so unfortunately, for now, I'm not sure what we can expect to be fixed. -- 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=371657 User kasievers@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c38 Kay Sievers changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |kasievers@novell.com AssignedTo|kasievers@novell.com |bnc-team-screening@forge.provo.novell.com Component|Kernel |YaST2 QAContact|qa@suse.de |jsrain@novell.com --- Comment #38 from Kay Sievers 2008-11-17 18:07:08 MST --- Yeah, but the problem is that none of the things which access the partitions read the partition type value, and should still continue to open the device, and cause the "access beyond end of device" errors. I still have no idea why the messages disappear. But if it helps to change the partition type, let's see if the partitioner can do something. Reassigning to Yast. Summary: dmraid setups create kernel partitions which are completely invalid, and cause "access beyond end of device" errors on probing for metadata. Dean, the bug reporter, sees that setting a different partition type in the partition table makes the bug disappear. This may be not a sufficient fix for the general problem with dmraid setups, but if it helps, please look if we can do something here. Details in comment#37. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c39 Dean Hilkewich changed: What |Removed |Added ---------------------------------------------------------------------------- Component|YaST2 |Kernel --- Comment #39 from Dean Hilkewich 2008-11-17 18:55:34 MST --- It should be noted that this already happens automatically in the Suse partitioner on the primary raid. It gets marked as a type f This only occurs on a secondary set where it does not create the same type f partition. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c41 --- Comment #41 from Dean Hilkewich 2008-11-18 09:58:09 MST --- I don't see how it is when using another partitioner resolves the issue. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c43 --- Comment #43 from Dean Hilkewich 2008-11-18 10:44:35 MST --- (In reply to comment #42 from Kay Sievers)

...

So is it a kernel issue or not?

...

I don't see how it is when using another partitioner resolves the issue.

Dean, you changed it to "kernel" yesterday:

Dean Hilkewich changed:

What |Removed |Added ---------------------------------------- Component|YaST2 |Kernel

Please do not change theses fields, they are used to coordinate who's looking at the bug internally.

I swear I did not do that. If I did it was purely unintentional. -- 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.

So if this really differs on several dmraid volumes the behavior must come from parted (bug or feature?). Maybe our parted maintainer knowns more. AFAIU the parted code, it should set the type of extended partition to 0xf if

https://bugzilla.novell.com/show_bug.cgi?id=371657 User puzel@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c45 Petr Uzel changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |puzel@novell.com Status|NEEDINFO |NEW Info Provider|puzel@novell.com | --- Comment #45 from Petr Uzel 2008-11-19 05:16:27 MST --- (In reply to comment #44 from Arvin Schnell) the storage device supports LBA and 0x5 otherwise.

YaST uses parted to create extended partitions and does not specify whether 0x05 or 0x0F should be used as type. The type can be forced with 'parted /dev/hdx set 1 type 0xf' in case its needed.

-- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c47 --- Comment #47 from Dean Hilkewich 2008-11-19 13:04:36 MST --- Created an attachment (id=253613) --> (https://bugzilla.novell.com/attachment.cgi?id=253613) contents of sys/firmware/edd Here is the contents of sys/firmware/edd you requested. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c49 --- Comment #49 from Dean Hilkewich 2008-11-19 15:56:10 MST --- Created an attachment (id=253669) --> (https://bugzilla.novell.com/attachment.cgi?id=253669) hwinfo output The two Maxtors are running Raid 0 on dmraid and the 4 Seagates are running as Raid 5 using the fakeraid controller. The Maxtors are the primary, the seagates the secondary. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c51 --- Comment #51 from Dean Hilkewich 2008-11-21 13:48:11 MST --- Arvin Schnell changed: What |Removed |Added ---------------------------------------------------------------------------- AssignedTo|yast2-maintainers@suse.de |aschnell@novell.com Status|NEEDINFO |ASSIGNED Info Provider|deanjo@sasktel.net | Target Milestone|--- |Future/Later So I'll have to dig this bug report AGAIN and pray that it actually gets looked at in a future release? This bug was reported back in 10.3. -- 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=371657 User deanjo@sasktel.net added comment https://bugzilla.novell.com/show_bug.cgi?id=371657#c53 --- Comment #53 from Dean Hilkewich 2008-11-22 14:33:36 MST --- It was marked as P1 forever until Kay decided for some reason to knock it down for some unexplained reason to P4 on the 17th. -- 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.