Distribution: Debian Hardware Environment: Toshiba Stalellite Pro M30 Software Environment: Problem Description: While reading entries from /proc/acpi/toshiba, kernel still allocates memory and doesn't free it until kernel kills everything because of no free memory. Steps to reproduce: This script shows allocations: grep size-64\ /proc/slabinfo sleep 10 grep size-64\ /proc/slabinfo for x in `seq 1 100` ; do cat /proc/acpi/toshiba/fan > /dev/null done grep size-64\ /proc/slabinfo sleep 10 grep size-64\ /proc/slabinfo Output looks like: size-64 201360 201361 64 61 1 : tunables 120 60 0 : slabdata 3301 3301 0 size-64 201360 201361 64 61 1 : tunables 120 60 0 : slabdata 3301 3301 0 size-64 201604 201605 64 61 1 : tunables 120 60 0 : slabdata 3305 3305 0 size-64 201604 201605 64 61 1 : tunables 120 60 0 : slabdata 3305 3305 0 It has been happening also in several older kernel versions.
Does reading other entry under acpi/toshiba have the same memory leak behavior? Could you please post your acpidmp output? A quick look at the toshiba driver shows there isn't memory leak there, so maybe it's ACPICA's issue.
Other entries in /proc/acpi/toshiba has this behaviour, I wasn't able to reproduce it using another files in /proc/acpi. I will post acpidmp output later, I don't have that computer right here.
Created attachment 5136 [details] acpidmp output Here it comes.
Is this problem still present in kernel 2.6.13-rc3?
Yes, still present.
Created attachment 5363 [details] acpidmp output I also updated BIOS meanwhile, so it's probably that ACPI tables changed. Here is current version.
This problem seems to be gone in 2.6.14.