Bug 484173
| Summary: | Intel Q33 SATA problem on libata and ata_piix | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Product: | Red Hat Enterprise Linux 5 | Reporter: | Hiroaki Nakano <nakano.hiroaki> | ||||||||||
| Component: | kernel | Assignee: | David Milburn <dmilburn> | ||||||||||
| Status: | CLOSED WONTFIX | QA Contact: | Mike Gahagan <mgahagan> | ||||||||||
| Severity: | medium | Docs Contact: | |||||||||||
| Priority: | low | ||||||||||||
| Version: | 5.2 | CC: | dzickus, maeno.masaki, t.h.amundsen | ||||||||||
| Target Milestone: | rc | Keywords: | Reopened | ||||||||||
| Target Release: | --- | ||||||||||||
| Hardware: | All | ||||||||||||
| OS: | Linux | ||||||||||||
| Whiteboard: | |||||||||||||
| Fixed In Version: | Doc Type: | Bug Fix | |||||||||||
| Doc Text: | Story Points: | --- | |||||||||||
| Clone Of: | Environment: | ||||||||||||
| Last Closed: | 2009-08-10 21:23:39 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
Hiroaki Nakano
2009-02-05 06:57:36 UTC
Created attachment 330964 [details]
/var/log/dmesg at this problem
Hi, This is a result of RHEL5 supporting combined mode where the IDE driver may claim the device(s) if the SATA controller is setup in IDE mode. And in this case, any devices controlled by the IDE driver will be in PIO mode which explains your slow throughput. In RHEL5, in order to have libata (ata_piix) claim the devices and run them in DMA mode, you will need to use the "ideX=noprobe" kernel parameter, or "combined_mode=libata" or if possible set the SATA controller in AHCI mode and let the ahci driver control the devices. The combined mode has been removed in the upstream kernels, but, we have to maintain it for RHEL5. The only kernel change we can make for RHEL5 is if the "combine_mode" parameter doesn't work, in that case re-open this BZ and supply your PCI device IDS and I can supply a test kernel. Thanks. Hi, I practiced the "combine_mode" parameter on RHEL5, but didn't change occuring. When I set the "ideX=noprobe" kernel parameter, it works to be good. I attach dmesg when I set the "combine_mode" parameter. Created attachment 332165 [details]
/var/log/dmesg file when I set the "combine_mode" parameter.
/var/log/dmesg file when I set the "combine_mode" parameter on RHEL5.
Following is /proc/cmdline:
ro root=LABEL=/ rhgb quiet elevator=deadline combined_mode=libata
Following is "uname -a":
Linux osspc015b 2.6.18-128.el5 #1 SMP Wed Dec 17 11:42:39 EST 2008 i686 i686 i386 GNU/Linux
Ok, thanks. Would you also attach the output of "lspci -vvxxx"? Created attachment 332521 [details]
lspci -vvxxx on Intel Q33 SATA
Hi,
Thanks you for the reaction.
I get the output of "lspci -vvxxx", so it attaches.
It shall be good if this output helps you.
Hi, Would you please boot kernel-2.6.18-131.el5.bz484173.1 with the "combined_mode=libata" kernel parameter and attach the dmesg output after booting? This test kernel will print out the combined mode register for your sata ide controller. Thank you. http://people.redhat.com/dmilburn/ Hi, Thanks you for test kernel. I will boot this on the reproducible environment, and report the result afresh. Created attachment 334596 [details]
dmesg on the test kernel
Hi,
This is boot dmesg on the test kernel 2.6.18-131.el5.
best regards,
Hi, Our previous quirks to handle this for ich6, ich7, and ich8 read the map address register (0x90) and look at bits 1:0 to determine if configured in combined mode. I was not expecting this register to return 0x0 for the 8086:2921 device in your configuration, looking at the ich9 documentation we may not be able to depend on that register value to setup a quirk, I will need to investigate furthur. Hi, I have looked thru the ICH9 documentation, the SATA controller has three different modes of operation - Native IDE, legacy, and ahci/raid. In order for us to quirk off the Map Address Register (0x90) your BIOS would have to configure the SATA controller in Legacy or Combined mode. If you have a system that you can experiment on, if the system BIOS will allow you to change the mode to legacy you could try the test kernel (Comment #7) with the kernel parameter "combined_mode=libata". Another possibility is changing the setting to ahci mode if possible, of course you would need to make sure ahci driver is loaded in your initrd. If neither of those options works out, then you will probably have to continue using the "ideX=noprobe" option. Please be careful changing any BIOS settings, you probably want to try this only on a test system at first. We inquired of the support charge of HP Desktop. HP's support said, "The BIOS setting of HP Compaq Business Desktop dx7400 MT (Intel Q33 Express (ICH9)) doesn't have AHCI or Legacy or Combined mode. It is not possible to correspond to various modes even if the BIOS is updated." Therefore, we hope to corrected to work normaly without "ide[01]=noprobe". It is the same matter as the content contributed of "NAKANO hiroaki <nakano.hiroaki.co.jp>". Sorry, we will not be able to change this behavior in RHEL5, please continue to use the "ideX=noprobe" option on the kernel command line. I see. We gave up fixing it this time. To execute the alternate solution (ide[01]=noprobe), we want to persuade our customer. |