Summary: | degraded performance after 2.6.16 | ||
---|---|---|---|
Product: | Gentoo Infrastructure | Reporter: | x.para |
Component: | Bugzilla | Assignee: | Gentoo Kernel Bug Wranglers and Kernel Maintainers <kernel> |
Status: | RESOLVED INVALID | ||
Severity: | normal | ||
Priority: | High | ||
Version: | unspecified | ||
Hardware: | x86 | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
Used for 2.6.16-gentoo-sources compilation
Used for 2.6.16-suspen2-r4 |
Description
x.para
2006-05-19 11:33:25 UTC
Please use "New -> Gentoo Linux -> Component: Ebuilds" in future. Infrastructure is not the correct addressee. Please try to reproduce with gentoo-sources-2.6.16-r7 (and if problem persists, vanilla-sources-2.6.17-rc4). I reproduced with gentoo-sources-2.6.16-r7 with the same - same performance problems results. that I tried vanilla-sources-2.6.17-rc4 and also faced same prerformace issues. Than I tried 2.6.16 vanila from kernel.org and everything looks ok. In the future, please reopen a bug when you provide the requested information. We need to figure out which kernel version introduces the bug then. Please start by seeing if you can reproduce the issue on gentoo-sources-2.6.16 (the first 2.6.16 release). Configure it as close to vanilla as possible, e.g. do not include vesafb-tng and fbsplash. OK. Here we go with some results. As per your advice I used gentoo-sources-2.6.16 and completely removed vesafb-tng and fbsplash. So no frame buffer console was used and I faced the same performance issues. Yes I think there must be something in gentoo patchset because 2.6.16 vanila does not have this problems. Please let me know what more I could try to remove from the kernel. I don't have an idea what to try now. Odd. Are you definitely using the same .config file for each kernel that you try? The next thing to investigate is which exact patch is causing the problem. The patches applied to gentoo-sources-2.6.16 (initial release) are found in these tarballs: http://dev.gentoo.org/~dsd/gentoo-sources/tarballs/genpatches-2.6.16-1.base.tar.bz2 http://dev.gentoo.org/~dsd/gentoo-sources/tarballs/genpatches-2.6.16-1.extras.tar.bz2 Apply them in ascending numeric order to a clean and fast 2.6.16 kernel, a few at a time, until you figure out which one causes the slowdown. If you need to step back at any time, you can revert patches in descending numeric order using the -R argument to patch. If you need more specific instructions, just say. Sorry I have to admit, that there is something wrong in my .config. I am still not sure what it is but when I started with the patching I found that I have performance issues also on vanilla 2.6.16 without any patch. As soon as I discovered it, I tried config from 2.6.16-r1 kernel with natural performance, applied it to 2.6.16 and I achived normal results. Than I tried to apply it to 2.6.16-suspend2-sources and it works too. So my conclusion is that I have migration problems with .config file between 2.6.15 and 2.6.16. There is many new definitions and if some of them is used in default it can make conflict, that makes the system performing like this. I know I am not the one having these problems, so I attached two config /working and non working/ files for your awareness but I am not sure if you are interested in it. If not plase close this case. Sorry I have to admit, that there is something wrong in my .config. I am still not sure what it is but when I started with the patching I found that I have performance issues also on vanilla 2.6.16 without any patch. As soon as I discovered it, I tried config from 2.6.16-r1 kernel with natural performance, applied it to 2.6.16 and I achived normal results. Than I tried to apply it to 2.6.16-suspend2-sources and it works too. So my conclusion is that I have migration problems with .config file between 2.6.15 and 2.6.16. There is many new definitions and if some of them is used in default it can make conflict, that makes the system performing like this. I know I am not the one having these problems, so I attached two config /working and non working/ files for your awareness but I am not sure if you are interested in it. If not plase close this case. Created attachment 89296 [details]
Used for 2.6.16-gentoo-sources compilation
Created attachment 89297 [details]
Used for 2.6.16-suspen2-r4
This config gives the kernel normal performance.
Here are some suspects from the bad config: These were enabled in the bad config but not in the good: CONFIG_SLOB: This is only for embedded systems. It does not scale to normal system use. CONFIG_BLK_DEV_GENERIC: The generic (slow) IDE driver may be claiming your hardware I do not recommend reusing old configs for new kernels. It is fairly common for them to bring up problems like this one. Reusing configs is OK for minor upgrades (e.g. 2.6.16-r2 to 2.6.16-r3) but not for actual kernel upgrades (e.g. 2.6.15 to 2.6.16) -- too much changes each time for this to be viable for the majority of users. Thank you. CONFIG_SLOB was the root of all this. Using config from 2.6.15 made CONFIG_SLOB enabled and CONFIG_SLAB disabled. Just to correct this gave me normal performance back. Thanks for bug learning lesson. |