[Bug 214909] New: performance problems with ata_piix while installation on ATA disk
https://bugzilla.novell.com/show_bug.cgi?id=214909 Summary: performance problems with ata_piix while installation on ATA disk Product: openSUSE 10.2 Version: Beta 1 Platform: Other OS/Version: Other Status: NEW Severity: Critical Priority: P5 - None Component: Kernel AssignedTo: hare@novell.com ReportedBy: dkukawka@novell.com QAContact: qa@suse.de Looks as if there is a problem with the ata_piix kernel driver while install Beta1 on a ACER TM C300 with a ATA disk. The installation need hours and the disk performance is more than bad. Take a look at the attachments (hwinfo --storage and vmstat log) -- 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=214909 ------- Comment #1 from dkukawka@novell.com 2006-10-25 05:58 MST ------- Created an attachment (id=102558) --> (https://bugzilla.novell.com/attachment.cgi?id=102558&action=view) output of hwinfo --storage -- 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=214909 ------- Comment #2 from dkukawka@novell.com 2006-10-25 05:59 MST ------- Created an attachment (id=102559) --> (https://bugzilla.novell.com/attachment.cgi?id=102559&action=view) log from vmstat 1 Look as if the CPU is permanently waiting for IO. -- 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=214909 ------- Comment #3 from fseidel@novell.com 2006-10-25 07:05 MST ------- I have the same symptoms here with a IBM Thinkpad T41p, with a ICH4-M IDE Controller (0x8086/0x24ca) and a ST960812A disk. I played with some boot params often pointed to regarding ata_piix (like ata_piix.force_pcs={0,1,2} and noprobe=hda), but none seemd to help here. -- 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=214909 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Severity|Critical |Major ------- Comment #4 from hare@novell.com 2006-10-27 02:26 MST ------- You can always use 'brokenmodules=ata_piix' in the kernel commandline for installation. Then installation will be using the old IDE driver. Meanwhile I've updated the libata stack which should improve things quite a lot. -- 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=214909 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution| |FIXED ------- Comment #5 from hare@novell.com 2006-10-27 02:51 MST ------- Bonnie says something about 47MB/s for block write and 34MB/s for block read with the new stack. So I'd say this is solved. -- 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=214909 ------- Comment #6 from dkukawka@novell.com 2006-10-27 03:43 MST ------- And where can I get a package with the new stack? I have to test this on the machines from the report before we can say it's 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=214909 ------- Comment #7 from hare@novell.com 2006-10-27 04:51 MST ------- Kernel for i386 are at ~hare/Export/stable -- 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=214909 dkukawka@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |REOPENED Resolution|FIXED | ------- Comment #8 from dkukawka@novell.com 2006-10-27 06:58 MST ------- With this kernel we are a littlebit faster, but the boot needs to long. With ata_piix until kdm is ready: 4:37 with piix until kdm is ready: 1:03 min with ata_piix there is high disk activity, but no progress in boot: mdraid, creat boot.msg, start syslog, start nsc/powersaved, start smpppd, start postfix, start kdm. -- 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=214909 cthiel@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |cthiel@novell.com Severity|Major |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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=214909 ------- Comment #9 from dkukawka@novell.com 2006-10-30 11:22 MST ------- Btw. not sure if ata_piix is the problem, but there is also no device for the internal CD/DVD drive. Only ram*, loop* and sda for the harddisk. If I use piix the device is shown under /sys/block/hdc. -- 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=214909 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|REOPENED |NEEDINFO Info Provider| |dkukawka@novell.com ------- Comment #10 from hare@novell.com 2006-10-31 01:32 MST ------- Can you try to boot with 'combined_mode=libata' ? -- 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=214909 ------- Comment #11 from aj@novell.com 2006-11-03 08:42 MST ------- Danny, is this working better with beta1plus? -- 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=214909 hare@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |hare@novell.com AssignedTo|hare@novell.com |htejun@gmail.com Status|NEEDINFO |NEW Info Provider|dkukawka@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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=214909 htejun@gmail.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |htejun@gmail.com Status|NEW |ASSIGNED ------- Comment #14 from htejun@gmail.com 2006-11-06 07:04 MST ------- Hello, the ata_piix's long delay during boot is probably due to phantom device. Please post boot dmesg. 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=214909 ------- Comment #15 from dkukawka@novell.com 2006-11-06 07:31 MST ------- (In reply to comment #14)
Hello, the ata_piix's long delay during boot is probably due to phantom device. Please post boot dmesg. Thanks.
Do you mean the delay from comment #8 ? -- 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=214909 ------- Comment #16 from htejun@gmail.com 2006-11-06 09:25 MST ------- Yeap, that one. -- 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=214909 ------- Comment #17 from dkukawka@novell.com 2006-11-06 16:33 MST ------- Only to be clear with comment #8: Not the detection of the devices need so long, but the complete boot and the system is after the boot also nearly unusable slow with disk IO. -- 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=214909 ------- Comment #18 from htejun@gmail.com 2006-11-06 18:18 MST ------- Oh.. I thought that was detection delay. In any case, boot dmesg should help diagnosing 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=214909 seife@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |seife@novell.com ------- Comment #19 from seife@novell.com 2006-11-09 14:58 MST ------- Still the same in beta2. remounting with "mount /dev/sda3 /mnt -oremount,barrier=0" during installation helped (and putting barrier=0 into /etc/fstab) -- 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=214909 htejun@gmail.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |seife@novell.com ------- Comment #20 from htejun@gmail.com 2006-11-09 19:31 MST ------- Stefan, please post 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=214909 seife@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED ------- Comment #21 from seife@novell.com 2006-11-10 00:43 MST ------- Created an attachment (id=104605) --> (https://bugzilla.novell.com/attachment.cgi?id=104605&action=view) dmesg of the running system, without barrier=0 -- 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=214909 ------- Comment #22 from htejun@gmail.com 2006-11-10 01:05 MST ------- Hmmm... Everything seems fine. No detection problems. The CD/DVD is available as /dev/sr0. I'm not sure whether the performance difference comes from. You're saying that booting takes more than three times longer w/ ata_piix than w/ piix. Right? Can you please report the followings? * With ata_piix loaded, 'hdparm -t /dev/sda' and 'hdparm -I /dev/sda'. * With piix loaded, dmesg, 'hdparm -t /dev/hda' and 'hdparm -I /dev/hda'. 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=214909 htejun@gmail.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |htejun@gmail.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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=214909 ------- Comment #23 from seife@novell.com 2006-11-10 01:49 MST ------- no. Booting / detection is pretty fast, but apparently barrier writes on ext3 take a long time, thus "barrier=0" mount option "fixes" it. During installation, after every rpm installed, there is a ~5 seconds stall. I guess that RPM is doing some flush after every package. I'm using only ext3 on this machine. d119:~ # hdparm -t /dev/sda /dev/sda: Timing buffered disk reads: 96 MB in 3.05 seconds = 31.47 MB/sec i can only boot with piix when i'm back in the office. -- 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=214909 seife@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jeffm@novell.com Status|NEEDINFO |ASSIGNED Info Provider|htejun@gmail.com | ------- Comment #24 from seife@novell.com 2006-11-10 01:57 MST ------- maybe this helps. different machines, but not _that_ different: piix: root@strolchi:/# dd if=/dev/zero of=/tmp/zero bs=1M count=100 oflag=sync 100+0 records in 100+0 records out 104857600 bytes (105 MB) copied, 13,0965 s, 8,0 MB/s ata_piix (only 10MB!): d119:~ # dd if=/dev/zero of=/tmp/zero bs=1M count=10 oflag=sync 10+0 records in 10+0 records out 10485760 bytes (10 MB) copied, 73,1469 s, 143 kB/s This would match the ~5 seconds pause after every RPM: 10 blocks are written, 10 sync() calls, every one of those takes ~7 seconds, and a little bit of time to actually write the data. Jeff, do these ~7 seconds barrier-timeout (?) on ext3 trigger something for you? -- 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=214909 ------- Comment #25 from htejun@gmail.com 2006-11-10 03:41 MST ------- I'm skeptical ext3 has anything to do with it. IDE and libata both export identical barrier capability to upper layer for the same drive. Can you please post the result of 'hdparm -I' of your disk? libata and IDE might use different command for cache flushing under certain circumstances and that could be causing the difference here. -- 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=214909 htejun@gmail.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |seife@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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=214909 ------- Comment #26 from seife@novell.com 2006-11-10 04:50 MST ------- Note: with beta1, this happened with all machines with piix when using the ata_piix driver, so i do not think it is an disk-specific issue. Will try the rescue system to boot the machine with piix. d119:~ # hdparm -I /dev/sda /dev/sda: ATA device, with non-removable media Model Number: ST960812A Serial Number: 3PJ00N95 Firmware Revision: 3.04 Standards: Used: ATA/ATAPI-6 T13 1410D revision 2 Supported: 6 5 4 Configuration: Logical max current cylinders 16383 17475 heads 16 15 sectors/track 63 63 -- CHS current addressable sectors: 16513875 LBA user addressable sectors: 117210240 LBA48 user addressable sectors: 117210240 device size with M = 1024*1024: 57231 MBytes device size with M = 1000*1000: 60011 MBytes (60 GB) Capabilities: LBA, IORDY(can be disabled) Standby timer values: spec'd by Standard, no device specific minimum R/W multiple sector transfer: Max = 16 Current = 16 Advanced power management level: unknown setting (0x8080) Recommended acoustic management value: 254, current value: 0 DMA: mdma0 mdma1 mdma2 udma0 udma1 udma2 udma3 udma4 *udma5 Cycle time: min=120ns recommended=120ns PIO: pio0 pio1 pio2 pio3 pio4 Cycle time: no flow control=240ns IORDY flow control=120ns Commands/features: Enabled Supported: * SMART feature set Security Mode feature set * Power Management feature set * Write cache * Look-ahead * Host Protected Area feature set * WRITE_BUFFER command * READ_BUFFER command * DOWNLOAD_MICROCODE * Advanced Power Management feature set SET_MAX security extension * 48-bit Address feature set Device Configuration Overlay feature set * Mandatory FLUSH_CACHE * FLUSH_CACHE_EXT * SMART error logging * SMART self-test * IDLE_IMMEDIATE with UNLOAD * SMART Command Transport (SCT) feature set Security: Master password revision code = 65534 supported not enabled not locked frozen not expired: security count not supported: enhanced erase HW reset results: CBLID- above Vih Device num = 0 determined by CSEL Checksum: correct -- 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=214909 seife@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED Info Provider|seife@novell.com | ------- Comment #27 from seife@novell.com 2006-11-10 05:02 MST ------- in the rescue system, with piix, i get 7.9MB/sec for the 100MB sync, 1MB blocksize write to the same disk, so it is comparable to the other machine. hdparm -I is exactly the same with piix. -- 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=214909 htejun@gmail.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |seife@novell.com ------- Comment #28 from htejun@gmail.com 2006-11-10 06:05 MST ------- I see, but just to be on the safe side, please test the following patch. Just one boot with the patched kernel should be enough. After booting (of course with ata_piix), do the sync write test. It should be slow as before. After that do the following. # echo 1 > /sys/module/libata/parameters/flush_ext_sel And, repeat the sync write test. Please report the resulting dmesg. TIA. -- 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=214909 ------- Comment #29 from htejun@gmail.com 2006-11-10 06:06 MST ------- Created an attachment (id=104658) --> (https://bugzilla.novell.com/attachment.cgi?id=104658&action=view) libata flush selection debug patch -- 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=214909 ------- Comment #30 from seife@novell.com 2006-11-10 06:55 MST ------- ok, will do. But it will only happen on monday (i'm out of office already) -- 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=214909 ------- Comment #31 from seife@novell.com 2006-11-10 10:01 MST ------- d119:~ # dd if=/dev/zero of=/tmp/zero bs=1M count=10 oflag=sync 10+0 records in 10+0 records out 10485760 bytes (10 MB) copied, 100,92 s, 104 kB/s d119:~ # echo 1 > /sys/module/libata/parameters/flush_ext_sel d119:~ # dd if=/dev/zero of=/tmp/zero bs=1M count=10 oflag=sync 10+0 records in 10+0 records out 10485760 bytes (10 MB) copied, 1,76438 s, 5,9 MB/s d119:~ # will attach 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=214909 seife@novell.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED ------- Comment #32 from seife@novell.com 2006-11-10 10:03 MST ------- Created an attachment (id=104716) --> (https://bugzilla.novell.com/attachment.cgi?id=104716&action=view) dmesg with Tejun's Patch This is with the patch applied and after doing the two dd commands in the previous comment. -- 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=214909 ------- Comment #33 from htejun@gmail.com 2006-11-10 11:39 MST ------- Explanation and patch have been posted to libata-dev#upstream. http://news.gmane.org/gmane.linux.ide I'm attaching patch against 10.2-beta1 and closing the bug. -- 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=214909 ------- Comment #34 from htejun@gmail.com 2006-11-10 11:39 MST ------- Created an attachment (id=104735) --> (https://bugzilla.novell.com/attachment.cgi?id=104735&action=view) fix FLUSH_EXT selection -- 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=214909 htejun@gmail.com changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED ------- Comment #35 from htejun@gmail.com 2006-11-10 11:40 MST ------- 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, or are watching someone who is.
https://bugzilla.novell.com/show_bug.cgi?id=214909 ------- Comment #36 from htejun@gmail.com 2006-11-10 11:41 MST ------- Oops, the URL is... http://article.gmane.org/gmane.linux.ide/13985 -- 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=214909 ------- Comment #37 from hare@novell.com 2006-11-13 00:46 MST ------- Seife, this is definitely worth a RACE award. Cool debugging, Tejun. -- 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=214909 ------- Comment #38 from hare@novell.com 2006-11-13 00:53 MST ------- Patch committed to the kernel CVS. -- 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.
participants (1)
-
bugzilla_noreply@novell.com