On 03/11/2013 06:23 PM, Rafael J. Wysocki wrote:
On Friday, February 22, 2013 10:24:33 AM Nathan Zimmer wrote:
I am noticing the cpufreq_driver_lock is quite hot.
On an idle 512 system perf shows me most of the system time is spent on this
lock. This is quite significant as top shows 5% of time in system time.
My solution was to first convert the lock to a rwlock and then to the rcu.
v2: Rebase
v3: Read the RCU documentation instead of skimming it. Also I based on
git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm.git pm+acpi-3.9-rc1
I assumed that was what you would prefer Rafael.
v4: Removed an unnecessary syncronize_rcu().
Nathan Zimmer (2):
cpufreq: Convert the cpufreq_driver_lock to a rwlock
cpufreq: Convert the cpufreq_driver_lock to use the rcu
drivers/cpufreq/cpufreq.c | 286 ++++++++++++++++++++++++++++++++++------------
1 file changed, 211 insertions(+), 75 deletions(-)
I'm going to take patch [1/2] for v3.10, but patch [2/2] still needs some
work it seems. Is that correct? If so, are you going to send an update?
Rafael
Viresh pointed out that cpufreq_cpu_data still needs a lock.
This means placing a vanilla spinlock back into __cpufreq_cpu_get which
is what I need to avoid. I haven't had the time I should to sort that out.
Nate
--
To unsubscribe from this list: send the line "unsubscribe cpufreq" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
