https://bugzilla.novell.com/show_bug.cgi?id=335505 Jeff Mahoney changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jeffm@novell.com AssignedTo|kernel-maintainers@forge.provo.novell.com |teheo@novell.com -- 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=335505#c2 Thomas Scholz changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|th.sch@gmx.net | --- Comment #2 from Thomas Scholz 2007-10-28 07:26:00 MST --- Created an attachment (id=180948) --> (https://bugzilla.novell.com/attachment.cgi?id=180948) the requested log files I included also a view message - log lines around the crash happen this morning. It occures during a cron-backup-job while mounting my backup drive (/dev/sde1) using autofs. -- 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=335505#c4 Tejun Heo changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |NEEDINFO Info Provider| |th.sch@gmx.net --- Comment #4 from Tejun Heo 2007-10-29 03:08:12 MST --- Okay, let's forget about the vanilla kernel then. If you specify "brokenmodules=pata_pdc202xx_old" as kernel parameter, IDE driver will be used instead (harddrives will appear as /dev/hdX). Do the errors go away? -- 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=335505#c6 Robert Davies changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |rob.opensuse.linux@googlemail.com --- Comment #6 from Robert Davies 2007-10-31 19:49:02 MST --- I've probably seen this issue on an MSI K7D Master mainboard using AMD 760 MPX with chipset, AMD-768 (Opus) southbridge, Dual Athlon MP 2000+. I have Promise PDC20267 (FastTrak100/Ultra100) (rev 02). The errors were The system worked stable enough in past to compile the whole of Gentoo in past (gcc, glibc and whole kaboodle), and initial OS10.3 install onto an old spare disk attached to generic ATA seemed fine. Then I partitioned a 60GB Seagate on the Promise ATA-100 controller (which were the recommended sort at time of purchase), built file systems and moved /usr and /var onto the drive, erasing Gentoo. Had intermittent sluggish performance, and machine lock ups, after it had run a few hours. I saw similar error messages, with I/O errors so swapped out the IDE cable, and put the 60GB Seagate on host controller (hd0,0). Still got problems... possibly because / & /var were now on the Promise. But after swapping them over, checked the Seagate with badblocks and came up with some errors, possibly caused by bum writes previously. Been running the system all day, without trouble on a diskless Kubuntu-7.10, Linux ubuntu 2.6.22-14-generic configuration, and not seen repeat of troubles, even when dowloading to hard disk mounted. Will be able to run diagnostics on drive later (1st November), and be in good shape to test with a stock kernel if it helps track this down. It's basically cost me an install and all the time and effort post, trying to track problem down (this is much worse than the HPT366 driver problem which refused to boot), as I've got corrupted data. If you like, I could very quickly install Kubuntu onto the drive, and see if their kernel has trouble to when I switch to the PDC controller. Oct 29 02:35:57 oak SuSEfirewall2: Firewall rules successfully set Oct 29 03:00:21 oak kernel: ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 frozen Oct 29 03:00:21 oak kernel: ata3.00: cmd ca/00:00:40:e8:01/00:00:00:00:00/e0 tag 0 cdb 0x0 data 131072 out Oct 29 03:00:21 oak kernel: res 40/00:00:01:4f:c2/00:00:00:00:00/00 Emask 0x4 (timeout) Oct 29 03:00:21 oak kernel: ata3: soft resetting link Oct 29 03:00:21 oak kernel: ata3.00: configured for UDMA/100 Oct 29 03:00:21 oak kernel: ata3: EH complete Oct 29 03:00:21 oak kernel: ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 Oct 29 03:00:21 oak kernel: ata3.00: BMDMA stat 0x24 Oct 29 03:00:21 oak kernel: ata3.00: cmd ca/00:00:40:e8:01/00:00:00:00:00/e0 tag 0 cdb 0x0 data 131072 out Oct 29 03:00:21 oak kernel: res 51/84:00:40:e8:01/00:00:00:00:00/e0 Emask 0x10 (ATA bus error) Oct 29 03:00:21 oak kernel: ata3: soft resetting link Oct 29 03:00:22 oak kernel: ata3.00: configured for UDMA/100 Oct 29 03:00:22 oak kernel: ata3: EH complete Oct 29 03:00:22 oak kernel: ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 Oct 29 03:00:22 oak kernel: ata3.00: BMDMA stat 0x24 Oct 29 03:00:22 oak kernel: ata3.00: cmd ca/00:00:40:e8:01/00:00:00:00:00/e0 tag 0 cdb 0x0 data 131072 out Oct 29 03:00:22 oak kernel: res 51/84:00:40:e8:01/00:00:00:00:00/e0 Emask 0x10 (ATA bus error) Oct 29 03:00:22 oak kernel: ata3: soft resetting link Oct 29 03:00:22 oak kernel: ata3.00: configured for UDMA/100 Oct 29 03:00:22 oak kernel: ata3: EH complete Oct 29 03:00:22 oak kernel: ata3.00: limiting speed to UDMA/66:PIO4 Oct 29 03:00:22 oak kernel: ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 Oct 29 03:00:22 oak kernel: ata3.00: BMDMA stat 0x24 Oct 29 03:00:22 oak kernel: ata3.00: cmd ca/00:00:40:e8:01/00:00:00:00:00/e0 tag 0 cdb 0x0 data 131072 out Oct 29 03:00:22 oak kernel: res 51/84:00:40:e8:01/00:00:00:00:00/e0 Emask 0x10 (ATA bus error) Oct 29 03:00:22 oak kernel: ata3: soft resetting link Oct 29 03:00:22 oak kernel: ata3.00: configured for UDMA/66 Oct 29 03:00:22 oak kernel: ata3: EH complete Oct 29 03:00:22 oak kernel: ata3: EH complete Oct 29 03:00:23 oak kernel: ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 Oct 29 03:00:23 oak kernel: ata3.00: BMDMA stat 0x24 Oct 29 03:00:23 oak kernel: ata3.00: cmd ca/00:00:40:e8:01/00:00:00:00:00/e0 tag 0 cdb 0x0 data 131072 out Oct 29 03:00:23 oak kernel: res 51/84:00:40:e8:01/00:00:00:00:00/e0 Emask 0x10 (ATA bus error) Oct 29 03:00:23 oak kernel: ata3: soft resetting link Oct 29 03:00:23 oak kernel: ata3.00: configured for UDMA/66 Oct 29 03:00:23 oak kernel: ata3: EH complete Oct 29 03:00:23 oak kernel: ata3.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 Oct 29 03:00:23 oak kernel: ata3.00: BMDMA stat 0x24 Oct 29 03:00:23 oak kernel: ata3.00: cmd ca/00:00:40:e8:01/00:00:00:00:00/e0 tag 0 cdb 0x0 data 131072 out Oct 29 03:00:23 oak kernel: res 51/84:00:40:e8:01/00:00:00:00:00/e0 Emask 0x10 (ATA bus error) Oct 29 03:00:23 oak kernel: ata3: soft resetting link Oct 29 03:00:23 oak kernel: ata3.00: configured for UDMA/66 Oct 29 03:00:23 oak kernel: sd 2:0:0:0: [sdb] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE,SUGGEST_OK Oct 29 03:00:23 oak kernel: sd 2:0:0:0: [sdb] Sense Key : Aborted Command [current] [descriptor] Oct 29 03:00:23 oak kernel: Descriptor sense data with sense descriptors (in hex): Oct 29 03:00:23 oak kernel: 72 0b 47 00 00 00 00 0c 00 0a 80 00 00 00 00 00 Oct 29 03:00:23 oak kernel: 00 01 e8 40 Oct 29 03:00:23 oak kernel: sd 2:0:0:0: [sdb] Add. Sense: Scsi parity error Oct 29 03:00:23 oak kernel: end_request: I/O error, dev sdb, sector 124992 Oct 29 03:00:23 oak kernel: Buffer I/O error on device sdb2, logical block 0 Oct 29 03:00:23 oak kernel: lost page write due to I/O error on sdb2 Oct 29 03:00:23 oak kernel: Buffer I/O error on device sdb2, logical block 1 Oct 29 03:00:23 oak kernel: lost page write due to I/O error on sdb2 Oct 29 03:00:23 oak kernel: Buffer I/O error on device sdb2, logical block 2 Oct 29 03:00:23 oak kernel: lost page write due to I/O error on sdb2 Oct 29 03:00:23 oak kernel: Buffer I/O error on device sdb2, logical block 3 Oct 29 03:00:23 oak kernel: lost page write due to I/O error on sdb2 Oct 29 03:00:23 oak kernel: Buffer I/O error on device sdb2, logical block 4 Oct 29 03:00:23 oak kernel: lost page write due to I/O error on sdb2 Oct 29 03:00:23 oak kernel: Buffer I/O error on device sdb2, logical block 5 Oct 29 03:00:23 oak kernel: lost page write due to I/O error on sdb2 Oct 29 03:00:23 oak kernel: Buffer I/O error on device sdb2, logical block 6 Oct 29 03:00:23 oak kernel: lost page write due to I/O error on sdb2 Oct 29 03:00:23 oak kernel: Buffer I/O error on device sdb2, logical block 7 Oct 29 03:00:23 oak kernel: lost page write due to I/O error on sdb2 Ouch, ouch, ouch! -- 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=335505#c8 --- Comment #8 from Thomas Scholz 2007-11-08 12:05:34 MST --- (In reply to comment #5 from Tejun Heo)

Also, are the errors localized to one drive?

Ok. I could reproduce the freeze (in a few sec.) by doing a yes > test.out on my backup hard disk. So I removed it and got no further system crash up to now. I did some log file analyses in the past days and discovered that the last 3-4 system crashes was caused by the same (the removed) hard drive. Maybe there are defect blocks, I can't check the device using the ibm/hitachi drive fitness test because it is not able to start the tests on it. For other devices it works without problems and no errors was found, also for one of the disks that was connected to the Promise controller before I switched it to the onboard VIA controller. I used it for recording tv shows using mythtv. After a few hours of recording no write operations can be done but the file system is still mounted rw (and not full ;) ). I'm not sure if this is a hardware defect or a kernel issue, I assume that this kind of error will end up in a crash if the device is connected to the Promise controller. Because it is not a driver- or kernel task to work around potential broken hardware I'm not sure how we should proceed. Maybe we can close this bug, on the other hand I got this problems since I installed suse 10.2, using 10.3 it is much better. ATM the system is working fine without problems. At the weekend I will do some tests again with the Promise Controller. -- 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=335505#c10 --- Comment #10 from Robert Davies 2007-11-09 06:35:39 MST --- (In reply to comment #1 from Tejun Heo)

Promise controllers are known to have this type of problems under certain configurations and it's yet unknown what's the cause. Can you please reproduce the problem on kernel.org 2.6.23 kernel? Just build drivers and filesystems you need for the root fs into the kernel and don't bother with initrd.

After reproducing the problem on vanilla 2.6.23 kernel, we can bring this problem to upstream linux-ide@vger.kernel.org for further debugging.

Also, please post /var/log/boot.msg and full dmesg here (with the current kernel).

Tejun, I've kept a dual Athlon MP box with Promise controller on standby since 1st Nov, to aid with any testing effort should it help. I very much want these IDE ATA controllers to be supported well by 2.6 kernel. As the machine has been idle since 1st, apart from test of Live CD's, I can do whatever it takes on that box, and will make time to do it. So does this help? Should I update to OS 10.3 updates, and run bonnie, dbench etc? Then report back. I haven't opened up a new Bug report, because it looked like my problem was duplicate of Thomas's. -- 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=335505#c12 --- Comment #12 from Tejun Heo 2007-11-09 07:04:16 MST --- I still don't know what's wrong with your configuration yet. Please do the following. 1. install SL103 to a working drive on a working controller (the via one). 2. after boot into working system, reproduce timeouts by executing "dd if=/dev/sdX of=/dev/null bs=1M" where /dev/sdX is a drive connected to pata_pdc202xx_old. Store /var/log/boot.msg, the result of dmesg and "lspci -nnvvvxxx" somewhere. 3. boot with "brokenmodules=pata_pdc202xx_old". the system will load pdc202xx_old driver (the IDE one), if it doesn't, load it by executing "modprobe pdc202xx_old". 4. run "dd if=/dev/hdX of=/dev/null bs=1M" where /dev/hdX is a drive connected to pdc202xx_old. Store /var/log/boot.msg, the result of dmesg and "lspci -nnvvvxxx". 5. Post six files from #2 and #4 here. 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=335505#c14 --- Comment #14 from Robert Davies 2007-11-15 19:13:52 MST --- Kernel updated to SuSE kernel-default 2.6.22.12-0.1; 2x60GB Seagates which have passed DFT's. I suspected my "intermittent sluggish performance, and machine lock ups, after it had run a few hours" were caused by SuSE kernel 2.6.22.9-0.4 brown paper bag release. So I ran torture test on disks bonnie++ & dbench, including simulataneous on both disks on Host controller, tests found nothing system remained responsive despite heavy CPU and saturated I/O usage; just 1 SMART Prefailure warning about Raw_Read_Error_Rate (actually 62 to 63, then later 54 to 55), and some Hardware ECC recovered messages. I save smartctl -a output, and will attach. So swapped the newer disk sdb onto Promise controller - PDC20267 (FastTrak100/Ultra100) (rev 02) When swap partition activated on the disk placed on Promise controller, immediately ATA3 aka sdb caused Exceptions and the link was reset, without any need to run benchmark, much logging caused, though a partition was mountable from that disk. Error log attached previous comment. On starting test bonnie++ mounting partion from disk on Promise controller, the machine immediately seemed to freeze up, and there was a logged message, saying "Clocksource tsc unstable (delta = 1405645341 ns)". So this problem *IS* easily reproducible with latest SuSE kernel 2.6.22.12, you just need the controller, and a disk on it. The problem has moved with the controller not any particular hard drive. The machine is practically unusable when Disk is accessed attached to the Promise. Please can I provide any info, to progress this Bug investigation, it seems that it is impractical for th.sch@gmx.net to do so! -- 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=335505#c16 --- Comment #16 from Robert Davies 2007-11-15 19:53:00 MST --- Noticed this issue seems to stop clean shutdown, causing replay of transactions on reboot. Booted machine with suggest boot option : rob@oak:~> cat /proc/cmdline root=/dev/disk/by-label/OS-Root vga=0x317 resume=/dev/disk/by-label/OS-SWAP splash=silent brokenmodules=pata_pdc202xx_old Unfortunately rob@oak:~> lsmod | grep pdc pata_pdc202xx_old 12160 2 libata 137032 2 pata_pdc202xx_old,pata_amd Perhaps they're in my initrd, so I altered the initrd and rebooted. This time I got the non pata_ pdc module. So sdb is now hda. No warnings about 40 track cables etc. iostat shows throughput for bonnie++ 10x previous. The system is nice to use despite running the benchmark, it is like having the disk on the host controller. I can even run dbench at same time on the other disk and still have usable system. Getting up to 20,700 Blk_wrtn/s on each disk! My problem is caused by the pata_pdc202xx_old module! -- 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=335505#c18 Thomas Scholz changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW Info Provider|th.sch@gmx.net | --- Comment #18 from Thomas Scholz 2007-11-24 08:24:35 MST --- Created an attachment (id=184608) --> (https://bugzilla.novell.com/attachment.cgi?id=184608) smart output Hello and sorry for my delayed response. All disks I connected to the promise controller cause problems that lead in a system crash sooner or later. I tried out 3 (older) hitatchi- and 1 (new) segate disk. I provide you the smart output as requested for the last of the three hitachi disks that was connected to the promise controller. At the moment, the system works fine because I'm not using this disk. I use it only for testing because of this bug. Today I did the following: 1) - dd if=/dev/sdc of=/dev/null bs=1M as described in #12 - /dev/sdc is connected to pata_pdc202xx_old, the Drive Fitness Test shows no errors on this device - immediately after starting dd /var/log/messages shows the following: Nov 24 15:57:39 DePP2 sshd[5806]: Accepted publickey for tom from 192.168.0.79 port 50275 ssh2 Nov 24 15:57:46 DePP2 su: (to root) tom on /dev/pts/5 Nov 24 15:58:05 DePP2 kernel: ata4.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 Nov 24 15:58:05 DePP2 kernel: ata4.00: BMDMA stat 0x4 Nov 24 15:58:05 DePP2 kernel: ata4.00: cmd c8/00:f8:08:0f:00/00:00:00:00:00/e0 tag 0 cdb 0x0 data 126976 in Nov 24 15:58:05 DePP2 kernel: res 51/84:00:ff:0f:00/00:00:00:00:00/e0 Emask 0x10 (ATA bus error) Nov 24 15:58:05 DePP2 kernel: ata4: soft resetting link Nov 24 15:58:05 DePP2 kernel: ata4.00: configured for UDMA/100 Nov 24 15:58:05 DePP2 kernel: ata4: EH complete Nov 24 15:58:05 DePP2 kernel: sd 3:0:0:0: [sdc] 120103200 512-byte hardware sectors (61493 MB) Nov 24 15:58:05 DePP2 kernel: sd 3:0:0:0: [sdc] Write Protect is off Nov 24 15:58:05 DePP2 kernel: sd 3:0:0:0: [sdc] Mode Sense: 00 3a 00 00 Nov 24 15:58:05 DePP2 kernel: sd 3:0:0:0: [sdc] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA .. Nov 24 16:01:35 DePP2 kernel: ata4.00: limiting speed to PIO4 Nov 24 16:01:35 DePP2 kernel: ata4.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 Nov 24 16:01:35 DePP2 kernel: ata4.00: BMDMA stat 0x4 Nov 24 16:01:35 DePP2 kernel: ata4.00: cmd c8/00:08:00:32:04/00:00:00:00:00/e0 tag 0 cdb 0x0 data 4096 in Nov 24 16:01:35 DePP2 kernel: res 51/84:00:07:32:04/00:00:00:00:00/e0 Emask 0x10 (ATA bus error) Nov 24 16:01:35 DePP2 kernel: ata4: soft resetting link Nov 24 16:01:35 DePP2 kernel: ata4.00: configured for PIO4 Nov 24 16:01:35 DePP2 kernel: ata4: EH complete Nov 24 16:01:36 DePP2 kernel: sd 3:0:0:0: [sdc] 120103200 512-byte hardware sectors (61493 MB) Nov 24 16:01:36 DePP2 kernel: sd 3:0:0:0: [sdc] Write Protect is off Nov 24 16:01:36 DePP2 kernel: sd 3:0:0:0: [sdc] Mode Sense: 00 3a 00 00 Nov 24 16:01:36 DePP2 kernel: sd 3:0:0:0: [sdc] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA 2) The brokenmodules=pata_pdc202xx_old boot parameter seems to be ignored on my openSUSE 10.3 system. All harddrives appear as /dev/sdX. (In reply to comment #9 from Tejun Heo)

So, other disks connected to the promise controller don't cause such problems, right? Please do the followings.

1. Post the result of "smartctl -a /dev/sdX" where sdX is the offending device.

2. Swap the power and SATA connector of the offending drive and another drive on the promise controller and see whether the problem follows the drive or stays with the port.

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=335505#c20 --- Comment #20 from Robert Davies 2007-11-24 11:26:34 MST --- Rob said "Thomas, I had the same problem, so that "brokenmodules=" works I did the following :" That was unclear, I of course append kernel argument in the GRUB entry like this : kernel /vmlinuz-2.6.12-0.1-default root=/dev/disk/by-label/OS-Root ... splash=silent showopts brokenmodules=pata_pdc202xx_old -- 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=335505 User teheo@novell.com added comment https://bugzilla.novell.com/show_bug.cgi?id=335505#c22 Tejun Heo changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |LATER --- Comment #22 from Tejun Heo 2008-06-24 00:05:40 MDT --- Pinged upstream. I don't think I can do much here till the bug is fixed upstream (I don't have hardware or document access). I'm closing this one as LATER. 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.