cpufreq: Revert commit a66b2e to fix suspend/resume regression
authorSrivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Thu, 11 Jul 2013 22:15:37 +0000 (03:45 +0530)
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>
Sun, 14 Jul 2013 23:31:36 +0000 (01:31 +0200)
commit a66b2e (cpufreq: Preserve sysfs files across suspend/resume)
has unfortunately caused several things in the cpufreq subsystem to
break subtly after a suspend/resume cycle.

The intention of that patch was to retain the file permissions of the
cpufreq related sysfs files across suspend/resume.  To achieve that,
the commit completely removed the calls to cpufreq_add_dev() and
__cpufreq_remove_dev() during suspend/resume transitions.  But the
problem is that those functions do 2 kinds of things:
  1. Low-level initialization/tear-down that are critical to the
     correct functioning of cpufreq-core.
  2. Kobject and sysfs related initialization/teardown.

Ideally we should have reorganized the code to cleanly separate these
two responsibilities, and skipped only the sysfs related parts during
suspend/resume.  Since we skipped the entire callbacks instead (which
also included some CPU and cpufreq-specific critical components),
cpufreq subsystem started behaving erratically after suspend/resume.

So revert the commit to fix the regression.  We'll revisit and address
the original goal of that commit separately, since it involves quite a
bit of careful code reorganization and appears to be non-trivial.

(While reverting the commit, note that another commit f51e1eb
 (cpufreq: Fix cpufreq regression after suspend/resume) already
 reverted part of the original set of changes.  So revert only the
 remaining ones).

Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Tested-by: Paul Bolle <pebolle@tiscali.nl>
Cc: 3.10+ <stable@vger.kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
drivers/cpufreq/cpufreq.c
drivers/cpufreq/cpufreq_stats.c

index 0937b8d6c2a4cebe904bbc9c5cdbf24afd2a681f..7dcfa6854833562f6194759bac5391379ce9fc18 100644 (file)
@@ -1942,13 +1942,15 @@ static int __cpuinit cpufreq_cpu_callback(struct notifier_block *nfb,
        if (dev) {
                switch (action) {
                case CPU_ONLINE:
+               case CPU_ONLINE_FROZEN:
                        cpufreq_add_dev(dev, NULL);
                        break;
                case CPU_DOWN_PREPARE:
-               case CPU_UP_CANCELED_FROZEN:
+               case CPU_DOWN_PREPARE_FROZEN:
                        __cpufreq_remove_dev(dev, NULL);
                        break;
                case CPU_DOWN_FAILED:
+               case CPU_DOWN_FAILED_FROZEN:
                        cpufreq_add_dev(dev, NULL);
                        break;
                }
index cd9e81713a71343b00f4c2e761f5f32c56824ab3..12225d19ffcbd0ec1cca2b21c46f384ea848f82b 100644 (file)
@@ -353,13 +353,11 @@ static int __cpuinit cpufreq_stat_cpu_callback(struct notifier_block *nfb,
                cpufreq_update_policy(cpu);
                break;
        case CPU_DOWN_PREPARE:
+       case CPU_DOWN_PREPARE_FROZEN:
                cpufreq_stats_free_sysfs(cpu);
                break;
        case CPU_DEAD:
-               cpufreq_stats_free_table(cpu);
-               break;
-       case CPU_UP_CANCELED_FROZEN:
-               cpufreq_stats_free_sysfs(cpu);
+       case CPU_DEAD_FROZEN:
                cpufreq_stats_free_table(cpu);
                break;
        }