Bug 2636
Summary: | HDD I/O drops to 229kb/sec if highmem is enabled | ||
---|---|---|---|
Product: | IO/Storage | Reporter: | Ludwig Ruderstaller (hostmaster) |
Component: | Other | Assignee: | Andrew Morton (akpm) |
Status: | CLOSED CODE_FIX | ||
Severity: | blocking | CC: | bunk, bzolnier, kernel.sensei, petar.forai, protasnb, rodd |
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.5 | Subsystem: | |
Regression: | --- | Bisected commit-id: | |
Attachments: |
additional Hardware Info and Kernel Config
A kernel profile capture |
Description
Ludwig Ruderstaller
2004-05-04 08:37:52 UTC
Created attachment 2785 [details]
additional Hardware Info and Kernel Config
Distro: Gentoo HW: P4 Prescott, HT enabled, 2x512MB RAM, ICH6 in AHCI mode, 1x WD Raptor 740GD i have exactly the same problem over here, posted it already in gentoo's bugzilla and forum, seems that a few ppl have related/same troubles with highmem enabled kernels regarding disc performance. with a highmem enabled kernel my boot time is ~10mins vs < 1min as compared to a highmem enabeld kernel booted with flag mem=880MB. app start up times are also in the same magnitude slower. general cpu /system performance seems unaffected. bonnie output: highmem, > 1000MB RAM booted Seeker 1...Seeker 2...Seeker 3...start 'em...done...done...done... -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 100 843 99.9 214295 78.5 766914 100.3 731 99.9 2364623 101.6 97878.5 185.9 highmem <= 880MB RAM booted Seeker 1...Seeker 2...Seeker 3...start 'em...done...done...done... -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 100 45898 99.7 393189 100.2 894995 94.4 50903 100.0 2453987 100.6 147552.5 184.4 Linux eistee 2.6.12-rc4 #2 SMP. Tried -mm, gentoo-dev-sources, love-sources, i can always reproduce it. how comes that this bug report is over a year old and nobody of the devs replied? Andrew, do you have any idea what might be causeing such problems? Nope. We'd need to see a kernel profile. If you tell them what exactly you want you might get it. ;-) *** Bug 4703 has been marked as a duplicate of this bug. *** just tell me what to do and i'll do it. i'm happy to be able to contribute and the problem to be solved is in my very personal interest. just tell me what to do and i'll do it. i'm happy to be able to contribute and the problem to be solved is in my very personal interest. Sorry for the late answer, please read Documentation/basic_profiling.txt in the kernel sources. Hi, I've exactly the same problem here on my 2GB RAM Notebook. When I boot with mem=1G the system speed is OK. I just read the basic_profile doc, it seems that there is 2 methods (readprofile and Oprofile)... which one should I use ? What is better ? Thanks in advance ! Boris. readprofile should be sufficient. oprofile is superior, but takes a bit more effort. Thanks. Created attachment 6058 [details]
A kernel profile capture
Hi,
Here is my kernel profile captured with : readprofile -m /boot/System.map >
captured_profile
HTH
Thanks.
hm, the cpu is almost totally idle. Bart, have you any ideas? I didn't mention that but : I just let the machine boot while I went to my work place and then log in via ssh, the machine does "nothing" actually ... can that explain why the CPU is idle ? Thanks, Boris. Any updates on this problem? Has it been resolved, or still happens with new kernels? Thanks. Since no activity for quite long time, I am closing the bug. Please reopen if it is confirmed with latest kernel. I still have this problem with the 2.6.25.6. Please Reopen this bug. Tell me if you want some logs / profiling and how to get these informations. Regards, Boris Fersing PS : I didn't change my hardware since the comment #10 I updated to the most latest stable kernel : 2.6.26.5 because I wanted to do a profiling and I found out that this version doesn't have this bug anymore. It seems to be fixed... I'm sorry for the reopen request, I should have tried the 2.6.26.5 before. regards, |