Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1314048 > unrolled thread

[PATCH v5 0/5] cpufreq: powernv: Redesign the presentation of throttle notification and solve bug-fixes in the driver

Back to article view | Back to linux.kernel

Contents

  [PATCH v5 0/5] cpufreq: powernv: Redesign the presentation of throttle notification and solve bug-fixes in the driver Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com> - 2016-01-21 10:50 +0100
    [PATCH v5 2/5] cpufreq: powernv: Remove cpu_to_chip_id() from hot-path Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com> - 2016-01-21 10:50 +0100
    [PATCH v5 3/5] cpufreq: powernv/tracing: Add powernv_throttle tracepoint Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com> - 2016-01-21 10:50 +0100

#1314048 — [PATCH v5 0/5] cpufreq: powernv: Redesign the presentation of throttle notification and solve bug-fixes in the driver

In POWER8, OCC(On-Chip-Controller) can throttle the frequency of the
CPU when the chip crosses its thermal and power limits. Currently,
powernv-cpufreq driver detects and reports this event as a console
message. Some machines may not sustain the max turbo frequency in all
conditions and can be throttled frequently. This can lead to the
flooding of console with throttle messages. So this patchset aims to
redesign the presentation of this event via sysfs counters and
tracepoints. And it also fixes couple of bugs reported in the driver.

- Patch [1] fixes the cpu hot-plug bug in powernv_cpufreq_work_fn().
- Patch [2] solves a bug in powernv_cpufreq_throttle_check(), which
  calls in to cpu_to_chip_id() in hot path which reads DT every time
  to find the chip id.
- Patches [3] to [5] will add a perf trace point
  "power:powernv_throttle" and sysfs throttle counter stats in
  /sys/devices/system/cpu/cpufreq/chipN.

Changes from v4:
- Fix a hot-plug bug in powernv_cpufreq_work_fn()
- Changes wrt Gautham's and Shreyas's comments 

Changes from v3:
- Add a fix to replace cpu_to_chip_id() with simpler PIR shift to 
  obtain the chip id.
- Break patch2 in to two patches separating the tracepoint and sysfs
  attribute changes.

Changes from v2:
- Fixed kbuild test warning.
drivers/cpufreq/powernv-cpufreq.c:609:2: warning: ignoring return
value of 'kstrtoint', declared with attribute warn_unused_result
[-Wunused-result]

Shilpasri G Bhat (5):
  cpufreq: powernv: Hot-plug safe the kworker thread
  cpufreq: powernv: Remove cpu_to_chip_id() from hot-path
  cpufreq: powernv/tracing: Add powernv_throttle tracepoint
  cpufreq: powernv: Replace pr_info with trace print for throttle event
  cpufreq: powernv: Add sysfs attributes to show throttle stats

 drivers/cpufreq/powernv-cpufreq.c | 312 +++++++++++++++++++++++++++++++-------
 include/trace/events/power.h      |  22 +++
 kernel/trace/power-traces.c       |   1 +
 3 files changed, 280 insertions(+), 55 deletions(-)

-- 
1.9.3

[toc] | [next] | [standalone]

#1314050 — [PATCH v5 2/5] cpufreq: powernv: Remove cpu_to_chip_id() from hot-path

cpu_to_chip_id() does a DT walk through to find out the chip id by
taking a contended device tree lock. This adds an unnecessary overhead
in a hot path. So instead of calling cpu_to_chip_id() everytime cache
the chip ids for all cores in the array 'core_to_chip_map' and use it
in the hotpath.

Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
---
Changes from v4:
- Taken care of Shreyas's comments to add a core_to_chip_map array to
  store the chip id.

 drivers/cpufreq/powernv-cpufreq.c | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)

diff --git a/drivers/cpufreq/powernv-cpufreq.c b/drivers/cpufreq/powernv-cpufreq.c
index 50a5b21..5ea103d 100644
--- a/drivers/cpufreq/powernv-cpufreq.c
+++ b/drivers/cpufreq/powernv-cpufreq.c
@@ -42,6 +42,7 @@
 
 static struct cpufreq_frequency_table powernv_freqs[POWERNV_MAX_PSTATES+1];
 static bool rebooting, throttled, occ_reset;
+static unsigned int *core_to_chip_map;
 
 static struct chip {
 	unsigned int id;
@@ -312,13 +313,14 @@ static inline unsigned int get_nominal_index(void)
 static void powernv_cpufreq_throttle_check(void *data)
 {
 	unsigned int cpu = smp_processor_id();
+	unsigned int chip_id = core_to_chip_map[cpu_core_index_of_thread(cpu)];
 	unsigned long pmsr;
 	int pmsr_pmax, i;
 
 	pmsr = get_pmspr(SPRN_PMSR);
 
 	for (i = 0; i < nr_chips; i++)
-		if (chips[i].id == cpu_to_chip_id(cpu))
+		if (chips[i].id == chip_id)
 			break;
 
 	/* Check for Pmax Capping */
@@ -558,19 +560,29 @@ static int init_chip_info(void)
 	unsigned int chip[256];
 	unsigned int cpu, i;
 	unsigned int prev_chip_id = UINT_MAX;
+	cpumask_t cpu_mask;
+	int ret = -ENOMEM;
 
-	for_each_possible_cpu(cpu) {
+	cpumask_copy(&cpu_mask, cpu_possible_mask);
+	core_to_chip_map = kcalloc(cpu_nr_cores(), sizeof(unsigned int),
+				   GFP_KERNEL);
+	if (!core_to_chip_map)
+		goto out;
+
+	for_each_cpu(cpu, &cpu_mask) {
 		unsigned int id = cpu_to_chip_id(cpu);
 
 		if (prev_chip_id != id) {
 			prev_chip_id = id;
 			chip[nr_chips++] = id;
 		}
+		core_to_chip_map[cpu_core_index_of_thread(cpu)] = id;
+		cpumask_andnot(&cpu_mask, &cpu_mask, cpu_sibling_mask(cpu));
 	}
 
 	chips = kmalloc_array(nr_chips, sizeof(struct chip), GFP_KERNEL);
 	if (!chips)
-		return -ENOMEM;
+		goto free_chip_map;
 
 	for (i = 0; i < nr_chips; i++) {
 		chips[i].id = chip[i];
@@ -581,6 +593,10 @@ static int init_chip_info(void)
 	}
 
 	return 0;
+free_chip_map:
+	kfree(core_to_chip_map);
+out:
+	return ret;
 }
 
 static int __init powernv_cpufreq_init(void)
@@ -614,6 +630,8 @@ static void __exit powernv_cpufreq_exit(void)
 	unregister_reboot_notifier(&powernv_cpufreq_reboot_nb);
 	opal_message_notifier_unregister(OPAL_MSG_OCC,
 					 &powernv_cpufreq_opal_nb);
+	kfree(chips);
+	kfree(core_to_chip_map);
 	cpufreq_unregister_driver(&powernv_cpufreq_driver);
 }
 module_exit(powernv_cpufreq_exit);
-- 
1.9.3

[toc] | [prev] | [next] | [standalone]

#1314051 — [PATCH v5 3/5] cpufreq: powernv/tracing: Add powernv_throttle tracepoint

This patch adds the powernv_throttle tracepoint to trace the CPU
frequency throttling event, which is used by the powernv-cpufreq
driver in POWER8.

Signed-off-by: Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: Steven Rostedt <rostedt@goodmis.org>
---
No changes since v2.

 include/trace/events/power.h | 22 ++++++++++++++++++++++
 kernel/trace/power-traces.c  |  1 +
 2 files changed, 23 insertions(+)

diff --git a/include/trace/events/power.h b/include/trace/events/power.h
index 284244e..19e5030 100644
--- a/include/trace/events/power.h
+++ b/include/trace/events/power.h
@@ -38,6 +38,28 @@ DEFINE_EVENT(cpu, cpu_idle,
 	TP_ARGS(state, cpu_id)
 );
 
+TRACE_EVENT(powernv_throttle,
+
+	TP_PROTO(int chip_id, const char *reason, int pmax),
+
+	TP_ARGS(chip_id, reason, pmax),
+
+	TP_STRUCT__entry(
+		__field(int, chip_id)
+		__string(reason, reason)
+		__field(int, pmax)
+	),
+
+	TP_fast_assign(
+		__entry->chip_id = chip_id;
+		__assign_str(reason, reason);
+		__entry->pmax = pmax;
+	),
+
+	TP_printk("Chip %d Pmax %d %s", __entry->chip_id,
+		  __entry->pmax, __get_str(reason))
+);
+
 TRACE_EVENT(pstate_sample,
 
 	TP_PROTO(u32 core_busy,
diff --git a/kernel/trace/power-traces.c b/kernel/trace/power-traces.c
index eb4220a..81b8745 100644
--- a/kernel/trace/power-traces.c
+++ b/kernel/trace/power-traces.c
@@ -15,4 +15,5 @@
 
 EXPORT_TRACEPOINT_SYMBOL_GPL(suspend_resume);
 EXPORT_TRACEPOINT_SYMBOL_GPL(cpu_idle);
+EXPORT_TRACEPOINT_SYMBOL_GPL(powernv_throttle);
 
-- 
1.9.3

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web