linux/drivers/cpufreq, branch v3.9.5

cpufreq / intel_pstate: fix ffmpeg regression

2013-05-19T18:38:42Z

commit ca182aee389f8026401510f4c63841cb02c820e8 upstream. The ffmpeg benchmark in the phoronix test suite has threads on multiple cores that rely on the progress on of threads on other cores and ping pong back and forth fast enough to make the core appear less busy than it "should" be. If the core has been at minimum p-state for a while bump the pstate up to kick the core to see if it is in this ping pong state. If the core is truly idle the p-state will be reduced at the next sample time. If the core makes more progress it will send more work to the thread bringing both threads out of the ping pong scenario and the p-state will be selected normally. This fixes a performance regression of approximately 30% Signed-off-by: Dirk Brandewie Signed-off-by: Rafael J. Wysocki Signed-off-by: Greg Kroah-Hartman

cpufreq / intel_pstate: use lowest requested max performance

2013-05-19T18:38:42Z

commit d8f469e9cff3bc4a6317d923e9506be046aa7bdc upstream. There are two ways that the maximum p-state can be clamped, via a policy change and via the sysfs file. The acpi-thermal driver adjusts the p-state policy in response to thermal events. These changes override the users settings at the moment. Use the lowest of the two requested values this ensures that we will not exceed the requested pstate from either mechanism. Reported-by: Srinivas Pandruvada Signed-off-by: Dirk Brandewie Signed-off-by: Rafael J. Wysocki Signed-off-by: Greg Kroah-Hartman

cpufreq / intel_pstate: remove idle time and duration from sample and calculations

2013-05-19T18:38:42Z

commit 1abc4b20b85b42e8573957e54b193385cf48b0d6 upstream. Idle time is taken into account in the APERF/MPERF ratio calculation there is no reason for the driver to track it seperately. This reduces the work in the driver and makes the code more readable. Removal of the tracking of sample duration removes the possibility of the divide by zero exception when the duration is sub 1us References: https://bugzilla.kernel.org/show_bug.cgi?id=56691 Reported-by: Mike Lothian Signed-off-by: Dirk Brandewie Signed-off-by: Rafael J. Wysocki Signed-off-by: Greg Kroah-Hartman

cpufreq / intel_pstate: Set timer timeout correctly

2013-04-08T20:09:23Z

The current calculation of the delay time is wrong and a cut and paste error from a previous experimental driver. This can result in the timeout being set to jiffies + 1 which setup the driver to race with itself if the APIC timer interrupt happens at just the right time. References: https://bugzilla.redhat.com/show_bug.cgi?id=920289 Reported-by: Adam Williamson Reported-and-tested-by: Parag Warudkar Signed-off-by: Dirk Brandewie Signed-off-by: Rafael J. Wysocki

cpufreq: Correct header guards typo

2013-04-02T13:14:29Z

It should be "governor". Signed-off-by: Borislav Petkov Acked-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki

cpufreq: check OF node /cpus presence before dereferencing it

2013-04-01T23:36:08Z

Check for the presence of the '/cpus' OF node before dereferencing it blindly: [ 4.181793] Unable to handle kernel NULL pointer dereference at virtual address 0000001c [ 4.181793] pgd = c0004000 [ 4.181823] [0000001c] *pgd=00000000 [ 4.181823] Internal error: Oops: 5 [#1] SMP ARM [ 4.181823] Modules linked in: [ 4.181823] CPU: 1 Tainted: G W (3.8.0-15-generic #25~hbankD) [ 4.181854] PC is at of_get_next_child+0x64/0x70 [ 4.181854] LR is at of_get_next_child+0x24/0x70 [ 4.181854] pc : [] lr : [] psr: 60000113 [ 4.181854] sp : ed891ec0 ip : ed891ec0 fp : ed891ed4 [ 4.181884] r10: c04dafd0 r9 : c098690c r8 : c0936208 [ 4.181884] r7 : ed890000 r6 : c0a63d00 r5 : 00000000 r4 : 00000000 [ 4.181884] r3 : 00000000 r2 : 00000000 r1 : 00000000 r0 : c0b2acc8 [ 4.181884] Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment kernel [ 4.181884] Control: 10c5387d Table: adcb804a DAC: 00000015 [ 4.181915] Process swapper/0 (pid: 1, stack limit = 0xed890238) [ 4.181915] Stack: (0xed891ec0 to 0xed892000) [ 4.181915] 1ec0: c09b7b70 00000007 ed891efc ed891ed8 c04daff4 c04fd9c0 00000000 c09b7b70 [ 4.181915] 1ee0: 00000007 c0a63d00 ed890000 c0936208 ed891f54 ed891f00 c00088e0 c04dafdc [ 4.181945] 1f00: ed891f54 ed891f10 c006e940 00000000 00000000 00000007 00000007 c08a4914 [ 4.181945] 1f20: 00000000 c07dbd30 c0a63d00 c09b7b70 00000007 c0a63d00 000000bc c0936208 [ 4.181945] 1f40: c098690c c0986914 ed891f94 ed891f58 c0936a40 c00087bc 00000007 00000007 [ 4.181976] 1f60: c0936208 be8bda20 b6eea010 c0a63d00 c064547c 00000000 00000000 00000000 [ 4.181976] 1f80: 00000000 00000000 ed891fac ed891f98 c0645498 c09368c8 00000000 00000000 [ 4.181976] 1fa0: 00000000 ed891fb0 c0014658 c0645488 00000000 00000000 00000000 00000000 [ 4.182006] 1fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000 [ 4.182006] 1fe0: 00000000 00000000 00000000 00000000 00000013 00000000 00000000 00000000 [ 4.182037] [] (of_get_next_child+0x64/0x70) from [] (cpu0_cpufreq_driver_init+0x24/0x284) [ 4.182067] [] (cpu0_cpufreq_driver_init+0x24/0x284) from [] (do_one_initcall+0x130/0x1b0) [ 4.182067] [] (do_one_initcall+0x130/0x1b0) from [] (kernel_init_freeable+0x184/0x24c) [ 4.182098] [] (kernel_init_freeable+0x184/0x24c) from [] (kernel_init+0x1c/0xf4) [ 4.182128] [] (kernel_init+0x1c/0xf4) from [] (ret_from_fork+0x14/0x20) [ 4.182128] Code: f57ff04f e320f004 e89da830 e89da830 (e595001c) [ 4.182128] ---[ end trace 634903a22e8609cb ]--- [ 4.182189] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b [ 4.182189] [ 4.642395] CPU0: stopping [rjw: Changelog] Signed-off-by: Paolo Pisati Acked-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki

cpufreq: acpi-cpufreq: Don't set policy->related_cpus from .init()

2013-03-25T14:13:32Z

With the addition of following patch: fcf8058 cpufreq: Simplify cpufreq_add_dev() cpufreq driver's .init() routine must initialize policy->cpus with mask of all possible CPUs (Online + Offline) that share the clock. Then the core would copy this mask onto policy->related_cpus and will reset policy->cpus to carry only online cpus. acpi-cpufreq driver wasn't updated with this assumption and so sometimes when we try to hot[un]plug CPUs at run time, sysfs directories get corrupted. This patch fixes acpi-cpufreq driver against this corruption. Reported-and-tested-by: Maciej Rutecki Tested-by: Borislav Petkov Signed-off-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki

cpufreq: stats: do cpufreq_cpu_put() corresponding to cpufreq_cpu_get()

2013-03-25T14:13:16Z

In cpufreq_stats_free_sysfs() we aren't balancing calls to cpufreq_cpu_get() with cpufreq_cpu_put(). This will never let us have ref count to policy->kobj as zero. We will get a hang if somehow cpufreq_driver_unregister() is called. And that can happen when we compile our driver as module and insmod/rmmod it. Signed-off-by: Viresh Kumar Acked-by: Amit Kucheria Signed-off-by: Rafael J. Wysocki

intel-pstate: Use #defines instead of hard-coded values.

2013-03-25T14:13:07Z

They are defined in coreboot (MSR_PLATFORM) and the other one is already defined in msr-index.h. Let's use those. Signed-off-by: Konrad Rzeszutek Wilk Acked-by: Viresh Kumar Acked-by: Dirk Brandewie Signed-off-by: Rafael J. Wysocki

cpufreq / intel_pstate: Fix calculation of current frequency

2013-03-25T14:12:59Z

Use the correct pstate value to calculate the effective frequency. References: https://bugzilla.redhat.com/show_bug.cgi?id=923942 Reported-by: Satish Balay Signed-off-by: Dirk Brandewie Signed-off-by: Rafael J. Wysocki