Bug 248819

Summary: Hard Drive speed with 2.6.22.1-15 test is 25x slower
Product: [Fedora] Fedora Reporter: James Hubbard <jameshubbard>
Component: kernelAssignee: Kernel Maintainer List <kernel-maint>
Status: CLOSED ERRATA QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: high Docs Contact:
Priority: low    
Version: 6CC: alan
Target Milestone: ---   
Target Release: ---   
Hardware: i686   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2007-09-05 21:19:14 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
output of hdparm, lsmod, dmesg for 6.20 and 6.22.1-15 none

Description James Hubbard 2007-07-18 23:19:36 UTC 
Description of problem:
Using the new 2.6.22.1-15 kernel makes my machine very slow.  Using hdparm, I
discovered the speed ~1.6 MB/sec vs ~49MB/sec with 2.6.20.

Version-Release number of selected component (if applicable):
kernel-2.6.22.1-15 Dell M65 

How reproducible:
Happens every time I boot the 2.6.22.1-15 kernel.

Steps to Reproduce:
1.Boot System
  
Actual results:
Under kernel 2.6.22.1-15 using:
hdparm -tT /dev/hda
/dev/hda:
 Timing cached reads:   11792 MB in  1.99 seconds = 5916.75 MB/sec
 Timing buffered disk reads:    6 MB in  3.54 seconds =   1.69 MB/sec

Expected results:
Under kernel 2.6.20-1.2962.fc6 using 
hdparm -tT /dev/sda
/dev/sda:
 Timing cached reads:   14492 MB in  1.99 seconds = 7273.03 MB/sec
 Timing buffered disk reads:  148 MB in  3.00 seconds =  49.27 MB/sec


Additional info:
I've tried the 2.6.22 kernel with  nohz=off highres=off, which doesn't make a
difference. 

I'm attaching the output of hdparm -tT /dev/xda, lsmod, and dmesg for both kernels. 

Here's the email to the test list where I reported the problem:
https://www.redhat.com/archives/fedora-test-list/2007-July/msg00372.html
Comment 1 James Hubbard 2007-07-18 23:19:37 UTC 
Created attachment 159566 [details]
output of hdparm, lsmod, dmesg for 6.20 and 6.22.1-15
Comment 2 Chuck Ebbert 2007-07-19 15:26:14 UTC 
This is really strange. In the FC6 2.6.20 kernels, the IDE driver does
not claim the drive; later on ata_piix gets control:

 Uniform Multi-Platform E-IDE driver Revision: 7.00alpha2
 ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
 ide0: I/O resource 0x1F0-0x1F7 not free.
 ide0: ports already in use, skipping probe

In 2.6.22, IDE does:

 Uniform Multi-Platform E-IDE driver Revision: 7.00alpha2
 ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
 Probing IDE interface ide0...
 Clocksource tsc unstable (delta = -101028827 ns)
 hda: Hitachi HTS721010G9SA00, ATA DISK drive

So what changed? The configuration looks the same, with IDE built in
and the ata driver as a module.
Comment 3 Alan Cox 2007-07-19 15:37:27 UTC 
2.6.22 no longer includes the rather horrible combined mode hacks to allow split
driver support so you must use libata for both in those cases. With old IDE
loaded and libata modular old IDE will claim the device as a legacy device and
stop libata claiming it
Comment 4 Chuck Ebbert 2007-09-05 21:19:14 UTC 
This should be fixed an all recent FC6 2.6.22 kernels. (The combined_mode code
was backported.)