Distribution: Debian Hardware Environment: IBM Thinkpad R40e Problem Description: When something i reading /proc/acpi/battery/BAT0/info system is unusable. It's loosing keyboard & mouse events, and it is slow. Also, reading of that file takes about 4 seconds. Steps to reproduce: $ time cat /proc/acpi/battery/BAT0/info present: yes design capacity: 43200 mWh last full capacity: 26320 mWh battery technology: rechargeable design voltage: 10800 mV design capacity warning: 1316 mWh design capacity low: 200 mWh capacity granularity 1: 1 mWh capacity granularity 2: 1 mWh model number: IBM-92P0987 serial number: 20 battery type: LION OEM info: SANYO real 0m5.495s user 0m0.000s sys 0m0.004s Same with /proc/acpi/battery/BAT0/state With 2.6.11.11 + swsusp2 patch it works fine.
any difference with ec_intr=0?
Also, please try the latest mm* version.
2.6.18-rc2 is available now.
No response from bug submitter, please reopen if problem persists.
Sorry for delay, a have no time for experimenting... :( I've upgraded to vanilla 2.6.18.1 + swsusp2. And problem is still here. When I booted with ec_intr=0 and launch this: while cat /proc/acpi/battery/BAT0/info ; do : ; done >/dev/null Mouse pointer slows down at only 1/4 of it's normal speed and movement is not smooth. With /proc/acpi/battery/BAT0/state is only bit slower. With /proc/acpi/battery/BAT0/alarm or /proc/cpuinfo it is ok. With ec_intr=0 is kernel usable, but problem still exists.
Created attachment 9248 [details] /proc/config.gz of running 2.6.18.1 + swsusp2
Please attach acpidump and dmesg after 'cat /proc/acpi/battery/*/*'.
Created attachment 9252 [details] acpidump > acpidump.txt
Created attachment 9253 [details] dmesg > dmesg.txt
Ok. Here they are. Files has been created by this commands: cat /proc/acpi/battery/*/* acpidump > acpidump.txt dmesg > dmesg.txt
Please try the 2.6.18.1 without swsusp2 patch.
No change. Same behavior without swsusp2.
Please try the latest 2.6.19-rc-git* version with/without ec_intr=0. There are some important changes in these versions. 2.6.19-rc-git7 is available now.
No change :(
> With 2.6.11.11 + swsusp2 patch it works fine. Do you have 2.6.11.11? What is the time of 'cat /proc/acpi/battery/*/*' on 2.6.11.11? Do you know any other latest versions where the problem is absent?
time for n in $(seq 1 1000) do cat /proc/acpi/battery/*/* >/dev/null done times of this loop at 2.6.11.11 are: real 0m23.097s user 0m0.870s sys 0m21.470s similar results with 2.6.7, 2.6.13.4 and 2.6.15.1 But with 2.6.17.1 for 100 loops (not 1000 as above): real 11m15.629s user 0m0.000s sys 0m0.004s
Please check the problem on 2.6.17, try 2.6.16 if the problem persists on 2.6.17. We need to minimize the differences between failed and passed versions. It will be nice if you can find two 'near by' versions (2.6.X.n and 2.6.X.n+1) Also, please try to reproduce a failure without the ibm_acpi module (i.e. remove ibm_acpi module from the .config, etc ...).
Ok, but it takes a while...
Please try the latest 2.6.20-rc1-mm* version.