[Bug 267444] New: Thinkpad Z61p S-ATA disk very slow with openSuSE-10.2-64bit
https://bugzilla.novell.com/show_bug.cgi?id=267444 Summary: Thinkpad Z61p S-ATA disk very slow with openSuSE-10.2- 64bit Product: openSUSE 10.2 Version: Final Platform: x86-64 OS/Version: openSUSE 10.2 Status: NEW Severity: Critical Priority: P5 - None Component: Kernel AssignedTo: teheo@novell.com ReportedBy: ffiene@veka.com QAContact: qa@suse.de My brand new Thinkpad Z61p is slower than my "old" T42p. Z61p: T7200 Dual2 Core 2GHz, 4GB RAM, 100GB S-ATA T42p: Centrino (Dothan?) 2GHz, 2GB RAM, 100GB IDE I found a possible issue: accessing S-ATA HD slows down the system. If i try "find / -name <bla>" the system is turned into a brick until find has been completed. "top" shows jumping 100% WA (wait IO?) from one CPU to the other. First i tried to find out if XFS or 64bit is the problem. 1.) XFS is not the problem, doing the same find on a LUKS-encrypted, XFS formatted USB HD is much more faster and "top" shows no WA. 2.) 64bit seems also not the problem, starting Ubuntu-7.04-64bit from CD, mounting HD with same mount options (noatime, nobarrier) shows a great performance! 3.) I've tried the kernel of the day, it was 2.6.21-rc7-git5 but no success, it has the same 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 teheo@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 teheo@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |ffiene@veka.com ------- Comment #1 from teheo@novell.com 2007-04-24 07:17 MST ------- Please post /var/log/boot.msg and the result of 'hdparm -I /dev/sda', 'hdparm -t /dev/sda', 'smartctl -d ata -a /dev/sda' and 'hwinfo --all'. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #2 from ffiene@veka.com 2007-04-24 07:51 MST ------- Created an attachment (id=133729) --> (https://bugzilla.novell.com/attachment.cgi?id=133729&action=view) /var/log/boot.msg -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #3 from ffiene@veka.com 2007-04-24 07:51 MST ------- Created an attachment (id=133731) --> (https://bugzilla.novell.com/attachment.cgi?id=133731&action=view) Output of hdparm -I /dev/sda -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #4 from ffiene@veka.com 2007-04-24 07:52 MST ------- Created an attachment (id=133733) --> (https://bugzilla.novell.com/attachment.cgi?id=133733&action=view) Output of hwinfo -all -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #5 from ffiene@veka.com 2007-04-24 07:52 MST ------- Created an attachment (id=133736) --> (https://bugzilla.novell.com/attachment.cgi?id=133736&action=view) Output of smartctl -d ata -a /dev/sda -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ffiene@veka.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|ffiene@veka.com | ------- Comment #6 from ffiene@veka.com 2007-04-24 07:58 MST ------- Created an attachment (id=133742) --> (https://bugzilla.novell.com/attachment.cgi?id=133742&action=view) Output of hdparm -t /dev/sda Process hdparm freezes after this output and i am not able to kill it. System is running as slow as before but it is running. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 teheo@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |ffiene@veka.com ------- Comment #7 from teheo@novell.com 2007-04-24 08:03 MST ------- Okay, your kernel is probably trying to recover from IO error. Please post the result of 'dmesg'. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ffiene@veka.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|ffiene@veka.com | ------- Comment #8 from ffiene@veka.com 2007-04-24 08:20 MST ------- Created an attachment (id=133766) --> (https://bugzilla.novell.com/attachment.cgi?id=133766&action=view) Output of dmesg -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #9 from ffiene@veka.com 2007-04-24 08:32 MST ------- Created an attachment (id=133773) --> (https://bugzilla.novell.com/attachment.cgi?id=133773&action=view) 4 times output of hdparm -t That's not too fast! ;-) Same result with the T42p, same disk but IDE. Shouldn't this be around 60-70MByte/sec? Dunno. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 teheo@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |INVALID ------- Comment #10 from teheo@novell.com 2007-04-24 08:41 MST ------- If you're seeing 30-40MB/s, there isn't any IO error. Dunno about that specific model's expected performance but I doubt it's a driver or kernel issue. If you've got a different OS (bsd, windows, whatever) installed on that machine, performing similar read throughput test on it would be a nice cross-check. If you see similar number on different OS and it's very different from your expectation of the drive, it might be that the specific drive is faulty or the model is hyped really well. :-) I'm resolving the bug as INVALID for now. Please re-open if you think it's not a hardware but kernel problem. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #11 from teheo@novell.com 2007-04-24 08:44 MST ------- Ah, one more thing to note. It might be that beagle or other auto-indexing is running, testing in single user mode will give you reliable numbers. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ffiene@veka.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |REOPENED Component|Kernel |Basesystem Priority|P5 - None |P1 - Urgent Resolution|INVALID | ------- Comment #12 from ffiene@veka.com 2007-04-24 12:49 MST ------- OK, you're right, starting with Ubuntu Live CD shows the same hdparm -t results, but as i've said in my first post: a simple "find / -name <bla>" turn my notebook into a brick, if i start Firefox or Konqueror directly after "find", for example, my notebook waits until "find" command is completed (after minutes), then Firefox appears immediatly. Ubuntu does not show this issue, but i want really want to keep SuSE. So please help!!! The same find on an external USB Disk encrypted with LUKS (must be slower than internal S-ATA without encryption) doesn't disturb the other applications. Also beagle cannot do this, the systems maybe slows down, but is usable. On the t42p, yaou cannot feel the difference. Both notebooks has the same disk, same data. I am not sure how to go on! Mybe i will try bonnie++. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 teheo@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |NEEDINFO Info Provider| |ffiene@veka.com ------- Comment #13 from teheo@novell.com 2007-04-24 12:57 MST ------- Hello, Frank. I think you might be seeing the same problem as the following one. http://bugzilla.kernel.org/show_bug.cgi?id=7590 It's a yet unknown upstream problem. I don't really think this is caused by driver layer. Recent posts indicate that changing thread scheduler behavior alleviates the problem. Or, maybe it's an artifact of cfq. Does doing 'echo deadline > /sys/block/sda/queue/scheduler' make any difference? -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #14 from teheo@novell.com 2007-04-24 12:59 MST ------- Oops, wrong link. The correct one is http://bugzilla.kernel.org/show_bug.cgi?id=7372 -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #15 from ffiene@veka.com 2007-04-24 14:20 MST ------- You made my day. Which schedule algorithm to choose? sa or deadline, they are working both fine! My notebook is flying! What's wrong with cfq? By the way, no scheduler was registered before "echo deadline ..:"m even not cfq -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #16 from teheo@novell.com 2007-04-24 20:00 MST ------- Okay, good to know. I have no idea why cfq is acting that way tho. I'll ask Jens whether he can take a look. I personally prefer deadline mainly because its code is simpler, which BTW doesn't have much bearing on its actual effectiveness. Most desktop workloads are probably better served by as. Hmm.. What do you mean by 'no scheduler was registered? What made you think that? -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #17 from teheo@novell.com 2007-04-25 00:17 MST ------- One thing to try. With cfq selected, please do 'echo 0 > /sys/block/sda/queue/iosched/slice_idle' and see whether things get better. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #18 from ffiene@veka.com 2007-04-25 00:29 MST ------- OK, in /proc/config.gz i see that cfq is the default scheduler, next time i boot Ubuntu i will have a look which scheduler is active there. Or does only SuSE compile only 64bit versions with cfq? Before 'echo deadline > /sys/block/sda/queue/scheduler' or put elevator=deadline in grub: cat /sys/block/sda/queue/scheduler noop anticipatory deadline cfq Here is nothing selected! After changes to scheduler (cat or grub): cat /sys/block/sda/queue/scheduler noop anticipatory [deadline] cfq So here is deadline assigned?! I've seen your last posting right a second ago and now it's getting funny. I made an "echo cfq > /sys/block/sda/queue/scheduler" then cat /sys/block/sda/queue/scheduler noop anticipatory deadline [cfq] So here is really cfq selected! My machine is reacting normal, so very fast! Maybe the default selection at kernel compile time has a problem? What scheduler is active without setting this in sysfs? -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 teheo@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|ffiene@veka.com | ------- Comment #19 from teheo@novell.com 2007-04-25 00:36 MST ------- That's really weird. Maybe something got screwed up in initrd or something. I'll try to find what's going on. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #20 from ffiene@veka.com 2007-04-25 02:22 MST ------- Now it's getting crazy. Booting Ubuntu from CD: cat /sys/block/sda/queue/scheduler noop anticipatory deadline [cfq] OK, the reason why ubuntu is working fine. But now: i've rebooted SuSE without elevator options now and tataaaa: cat /sys/block/sda/queue/scheduler noop anticipatory deadline [cfq] So at what time this setting has been lost and why it it present now? I didn't make a mkinitrd between this reboots! I saw a mkinitrd update last days. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444 ------- Comment #21 from ffiene@veka.com 2007-04-25 02:24 MST ------- Sorry, i forgot to make the stress test, find is as slow as the beginning! So i put back the elevator option into grubs kernel boot options. -- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=267444#c22
Christoph Resch
https://bugzilla.novell.com/show_bug.cgi?id=267444#c23
Christoph Resch
https://bugzilla.novell.com/show_bug.cgi?id=267444
Christoph Resch
https://bugzilla.novell.com/show_bug.cgi?id=267444#c24
Tejun Heo
https://bugzilla.novell.com/show_bug.cgi?id=267444
Tejun Heo
https://bugzilla.novell.com/show_bug.cgi?id=267444#c25
--- Comment #25 from Christoph Resch
https://bugzilla.novell.com/show_bug.cgi?id=267444#c26
--- Comment #26 from Tejun Heo
https://bugzilla.novell.com/show_bug.cgi?id=267444
Tejun Heo
https://bugzilla.novell.com/show_bug.cgi?id=267444#c27
--- Comment #27 from Tejun Heo
https://bugzilla.novell.com/show_bug.cgi?id=267444
User teheo@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=267444#c28
Tejun Heo
https://bugzilla.novell.com/show_bug.cgi?id=267444
User coolo@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=267444#c29
Stephan Kulow
https://bugzilla.novell.com/show_bug.cgi?id=267444
User coolo@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=267444#c30
Stephan Kulow
participants (1)
-
bugzilla_noreply＠novell.com