This patchset updates cpupower to make it more accurate by removing
the userspace to kernel transition and read_msr initiated IPI delays.
The first patch does a little re-arrangement of variables in the
cpuidle_monitor struct to prepare for a new flag.
The second patch introduces a per_cpu_schedule flag which, when set,
will allow cpupower to move to each of the cpus in the system. The
advantage of this is that the IPI latency is removed when reading the
APERF/MPERF registers, since an IPI is not generate for rdmsrs when
the source and destination cpus are the same for the IPI.
The third patch introduces the RDPRU instruction, which will allow
cpupower to not use the msr module for APERF/MPERF register reads.
This will remove the userspace to kernel transition delays when
reading the APERF/MPERF registers.
The fourth patch updates the ToDo file with ideas for further improving
the code handling the per_cpu_schedule flag.
v1->v2:
* Added cover letter.
* Used bind_cpu instead of rewriting the same code.
* Move needs_root to flag sub-struct.
* Introduce per_cpu_schedule flag.
v2->v3:
* Added ToDo patch to this set.
* Fix checkpatch warnings.
Janakarajan Natarajan (4):
cpupower: Move needs_root variable into a sub-struct
cpupower: mperf_monitor: Introduce per_cpu_schedule flag
cpupower: mperf_monitor: Update cpupower to use the RDPRU instruction
cpupower: ToDo: Update ToDo with ideas for per_cpu_schedule handling
tools/power/cpupower/ToDo | 14 ++++
tools/power/cpupower/utils/helpers/cpuid.c | 4 ++
tools/power/cpupower/utils/helpers/helpers.h | 1 +
.../utils/idle_monitor/amd_fam14h_idle.c | 2 +-
.../utils/idle_monitor/cpuidle_sysfs.c | 2 +-
.../utils/idle_monitor/cpupower-monitor.c | 2 +-
.../utils/idle_monitor/cpupower-monitor.h | 5 +-
.../utils/idle_monitor/hsw_ext_idle.c | 2 +-
.../utils/idle_monitor/mperf_monitor.c | 64 +++++++++++++++----
.../cpupower/utils/idle_monitor/nhm_idle.c | 2 +-
.../cpupower/utils/idle_monitor/snb_idle.c | 2 +-
11 files changed, 82 insertions(+), 18 deletions(-)
--
2.17.1
From: Janakarajan Natarajan <[email protected]>
Move the needs_root variable into a sub-struct. This is in preparation
for adding a new flag for cpuidle_monitor.
Update all uses of the needs_root variable to reflect this change.
Signed-off-by: Janakarajan Natarajan <[email protected]>
---
tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c | 2 +-
tools/power/cpupower/utils/idle_monitor/cpuidle_sysfs.c | 2 +-
tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c | 2 +-
tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h | 4 +++-
tools/power/cpupower/utils/idle_monitor/hsw_ext_idle.c | 2 +-
tools/power/cpupower/utils/idle_monitor/mperf_monitor.c | 2 +-
tools/power/cpupower/utils/idle_monitor/nhm_idle.c | 2 +-
tools/power/cpupower/utils/idle_monitor/snb_idle.c | 2 +-
8 files changed, 10 insertions(+), 8 deletions(-)
diff --git a/tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c b/tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c
index 3f893b99b337..33dc34db4f3c 100644
--- a/tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c
+++ b/tools/power/cpupower/utils/idle_monitor/amd_fam14h_idle.c
@@ -328,7 +328,7 @@ struct cpuidle_monitor amd_fam14h_monitor = {
.stop = amd_fam14h_stop,
.do_register = amd_fam14h_register,
.unregister = amd_fam14h_unregister,
- .needs_root = 1,
+ .flags.needs_root = 1,
.overflow_s = OVERFLOW_MS / 1000,
};
#endif /* #if defined(__i386__) || defined(__x86_64__) */
diff --git a/tools/power/cpupower/utils/idle_monitor/cpuidle_sysfs.c b/tools/power/cpupower/utils/idle_monitor/cpuidle_sysfs.c
index f634aeb65c5f..3c4cee160b0e 100644
--- a/tools/power/cpupower/utils/idle_monitor/cpuidle_sysfs.c
+++ b/tools/power/cpupower/utils/idle_monitor/cpuidle_sysfs.c
@@ -207,6 +207,6 @@ struct cpuidle_monitor cpuidle_sysfs_monitor = {
.stop = cpuidle_stop,
.do_register = cpuidle_register,
.unregister = cpuidle_unregister,
- .needs_root = 0,
+ .flags.needs_root = 0,
.overflow_s = UINT_MAX,
};
diff --git a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c
index d3c3e6e7aa26..6d44fec55ad5 100644
--- a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c
+++ b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c
@@ -408,7 +408,7 @@ int cmd_monitor(int argc, char **argv)
dprint("Try to register: %s\n", all_monitors[num]->name);
test_mon = all_monitors[num]->do_register();
if (test_mon) {
- if (test_mon->needs_root && !run_as_root) {
+ if (test_mon->flags.needs_root && !run_as_root) {
fprintf(stderr, _("Available monitor %s needs "
"root access\n"), test_mon->name);
continue;
diff --git a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
index a2d901d3bfaf..9b612d999660 100644
--- a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
+++ b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
@@ -60,7 +60,9 @@ struct cpuidle_monitor {
struct cpuidle_monitor* (*do_register) (void);
void (*unregister)(void);
unsigned int overflow_s;
- int needs_root;
+ struct {
+ unsigned int needs_root:1;
+ } flags;
};
extern long long timespec_diff_us(struct timespec start, struct timespec end);
diff --git a/tools/power/cpupower/utils/idle_monitor/hsw_ext_idle.c b/tools/power/cpupower/utils/idle_monitor/hsw_ext_idle.c
index 7c7451d3f494..885740a654e6 100644
--- a/tools/power/cpupower/utils/idle_monitor/hsw_ext_idle.c
+++ b/tools/power/cpupower/utils/idle_monitor/hsw_ext_idle.c
@@ -188,7 +188,7 @@ struct cpuidle_monitor intel_hsw_ext_monitor = {
.stop = hsw_ext_stop,
.do_register = hsw_ext_register,
.unregister = hsw_ext_unregister,
- .needs_root = 1,
+ .flags.needs_root = 1,
.overflow_s = 922000000 /* 922337203 seconds TSC overflow
at 20GHz */
};
diff --git a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
index 44806a6dae11..7cae74202a4d 100644
--- a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
+++ b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
@@ -333,7 +333,7 @@ struct cpuidle_monitor mperf_monitor = {
.stop = mperf_stop,
.do_register = mperf_register,
.unregister = mperf_unregister,
- .needs_root = 1,
+ .flags.needs_root = 1,
.overflow_s = 922000000 /* 922337203 seconds TSC overflow
at 20GHz */
};
diff --git a/tools/power/cpupower/utils/idle_monitor/nhm_idle.c b/tools/power/cpupower/utils/idle_monitor/nhm_idle.c
index be7256696a37..114271165182 100644
--- a/tools/power/cpupower/utils/idle_monitor/nhm_idle.c
+++ b/tools/power/cpupower/utils/idle_monitor/nhm_idle.c
@@ -208,7 +208,7 @@ struct cpuidle_monitor intel_nhm_monitor = {
.stop = nhm_stop,
.do_register = intel_nhm_register,
.unregister = intel_nhm_unregister,
- .needs_root = 1,
+ .flags.needs_root = 1,
.overflow_s = 922000000 /* 922337203 seconds TSC overflow
at 20GHz */
};
diff --git a/tools/power/cpupower/utils/idle_monitor/snb_idle.c b/tools/power/cpupower/utils/idle_monitor/snb_idle.c
index 968333571cad..df8b223cc096 100644
--- a/tools/power/cpupower/utils/idle_monitor/snb_idle.c
+++ b/tools/power/cpupower/utils/idle_monitor/snb_idle.c
@@ -192,7 +192,7 @@ struct cpuidle_monitor intel_snb_monitor = {
.stop = snb_stop,
.do_register = snb_register,
.unregister = snb_unregister,
- .needs_root = 1,
+ .flags.needs_root = 1,
.overflow_s = 922000000 /* 922337203 seconds TSC overflow
at 20GHz */
};
--
2.17.1
From: Janakarajan Natarajan <[email protected]>
The per_cpu_schedule flag is used to move the cpupower process to the cpu
on which we are looking to read the APERF/MPERF registers.
This prevents IPIs from being generated by read_msr()s as we are already
on the cpu of interest.
Ex: If cpupower is running on CPU 0 and we execute
read_msr(20, MSR_APERF, val) then,
read_msr(20, MSR_MPERF, val)
the msr module will generate an IPI from CPU 0 to CPU 20 to query
for the MSR_APERF and then the MSR_MPERF in separate IPIs.
This delay, caused by IPI latency, between reading the APERF and MPERF
registers may cause both of them to go out of sync.
The use of the per_cpu_schedule flag reduces the probability of this
from happening. It comes at the cost of a negligible increase in cpu
consumption caused by the migration of cpupower across each of the
cpus of the system.
Signed-off-by: Janakarajan Natarajan <[email protected]>
---
.../utils/idle_monitor/cpupower-monitor.h | 1 +
.../utils/idle_monitor/mperf_monitor.c | 42 ++++++++++++++-----
2 files changed, 33 insertions(+), 10 deletions(-)
diff --git a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
index 9b612d999660..5b5eb1da0cce 100644
--- a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
+++ b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
@@ -62,6 +62,7 @@ struct cpuidle_monitor {
unsigned int overflow_s;
struct {
unsigned int needs_root:1;
+ unsigned int per_cpu_schedule:1;
} flags;
};
diff --git a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
index 7cae74202a4d..afb2e6f8edd3 100644
--- a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
+++ b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
@@ -86,15 +86,35 @@ static int mperf_get_tsc(unsigned long long *tsc)
return ret;
}
+static int get_aperf_mperf(int cpu, unsigned long long *aval,
+ unsigned long long *mval)
+{
+ int ret;
+
+ /*
+ * Running on the cpu from which we read the registers will
+ * prevent APERF/MPERF from going out of sync because of IPI
+ * latency introduced by read_msr()s.
+ */
+ if (mperf_monitor.flags.per_cpu_schedule) {
+ if (bind_cpu(cpu))
+ return 1;
+ }
+
+ ret = read_msr(cpu, MSR_APERF, aval);
+ ret |= read_msr(cpu, MSR_MPERF, mval);
+
+ return ret;
+}
+
static int mperf_init_stats(unsigned int cpu)
{
- unsigned long long val;
+ unsigned long long aval, mval;
int ret;
- ret = read_msr(cpu, MSR_APERF, &val);
- aperf_previous_count[cpu] = val;
- ret |= read_msr(cpu, MSR_MPERF, &val);
- mperf_previous_count[cpu] = val;
+ ret = get_aperf_mperf(cpu, &aval, &mval);
+ aperf_previous_count[cpu] = aval;
+ mperf_previous_count[cpu] = mval;
is_valid[cpu] = !ret;
return 0;
@@ -102,13 +122,12 @@ static int mperf_init_stats(unsigned int cpu)
static int mperf_measure_stats(unsigned int cpu)
{
- unsigned long long val;
+ unsigned long long aval, mval;
int ret;
- ret = read_msr(cpu, MSR_APERF, &val);
- aperf_current_count[cpu] = val;
- ret |= read_msr(cpu, MSR_MPERF, &val);
- mperf_current_count[cpu] = val;
+ ret = get_aperf_mperf(cpu, &aval, &mval);
+ aperf_current_count[cpu] = aval;
+ mperf_current_count[cpu] = mval;
is_valid[cpu] = !ret;
return 0;
@@ -305,6 +324,9 @@ struct cpuidle_monitor *mperf_register(void)
if (init_maxfreq_mode())
return NULL;
+ if (cpupower_cpu_info.vendor == X86_VENDOR_AMD)
+ mperf_monitor.flags.per_cpu_schedule = 1;
+
/* Free this at program termination */
is_valid = calloc(cpu_count, sizeof(int));
mperf_previous_count = calloc(cpu_count, sizeof(unsigned long long));
--
2.17.1
From: Janakarajan Natarajan <[email protected]>
Based on Thomas Renninger's feedback/ideas. Re-structure the code
to better handle the per_cpu_schedule mechanism which was introduced
when adding support for AMD Zen based processors.
Signed-off-by: Janakarajan Natarajan <[email protected]>
---
tools/power/cpupower/ToDo | 14 ++++++++++++++
1 file changed, 14 insertions(+)
diff --git a/tools/power/cpupower/ToDo b/tools/power/cpupower/ToDo
index 6e8b89f282e6..b196a139a3e4 100644
--- a/tools/power/cpupower/ToDo
+++ b/tools/power/cpupower/ToDo
@@ -8,3 +8,17 @@ ToDos sorted by priority:
- Add another c1e debug idle monitor
-> Is by design racy with BIOS, but could be added
with a --force option and some "be careful" messages
+- Add cpu_start()/cpu_stop() callbacks for monitor
+ -> This is to move the per_cpu logic from inside the
+ monitor to outside it. This can be given higher
+ priority in fork_it.
+- Fork as many processes as there are CPUs in case the
+ per_cpu_schedule flag is set.
+ -> Bind forked process to each cpu.
+ -> Execute start measures via the forked processes on
+ each cpu.
+ -> Run test executable in a forked process.
+ -> Execute stop measures via the forked processes on
+ each cpu.
+ This would be ideal as it will not introduce noise in the
+ tested executable.
--
2.17.1
From: Janakarajan Natarajan <[email protected]>
AMD Zen 2 introduces the RDPRU instruction which can be used to access some
processor registers which are typically only accessible in privilege level
0. ECX specifies the register to read and EDX:EAX will contain the value read.
ECX: 0 - Register MPERF
1 - Register APERF
This has the added advantage of not having to use the msr module, since the
userspace to kernel transitions which occur during each read_msr() might
cause APERF and MPERF to go out of sync.
Signed-off-by: Janakarajan Natarajan <[email protected]>
---
tools/power/cpupower/utils/helpers/cpuid.c | 4 ++++
tools/power/cpupower/utils/helpers/helpers.h | 1 +
.../utils/idle_monitor/mperf_monitor.c | 20 +++++++++++++++++++
3 files changed, 25 insertions(+)
diff --git a/tools/power/cpupower/utils/helpers/cpuid.c b/tools/power/cpupower/utils/helpers/cpuid.c
index 5cc39d4e23ed..73bfafc60e9b 100644
--- a/tools/power/cpupower/utils/helpers/cpuid.c
+++ b/tools/power/cpupower/utils/helpers/cpuid.c
@@ -131,6 +131,10 @@ int get_cpu_info(struct cpupower_cpu_info *cpu_info)
if (ext_cpuid_level >= 0x80000007 &&
(cpuid_edx(0x80000007) & (1 << 9)))
cpu_info->caps |= CPUPOWER_CAP_AMD_CBP;
+
+ if (ext_cpuid_level >= 0x80000008 &&
+ cpuid_ebx(0x80000008) & (1 << 4))
+ cpu_info->caps |= CPUPOWER_CAP_AMD_RDPRU;
}
if (cpu_info->vendor == X86_VENDOR_INTEL) {
diff --git a/tools/power/cpupower/utils/helpers/helpers.h b/tools/power/cpupower/utils/helpers/helpers.h
index 357b19bb136e..c258eeccd05f 100644
--- a/tools/power/cpupower/utils/helpers/helpers.h
+++ b/tools/power/cpupower/utils/helpers/helpers.h
@@ -69,6 +69,7 @@ enum cpupower_cpu_vendor {X86_VENDOR_UNKNOWN = 0, X86_VENDOR_INTEL,
#define CPUPOWER_CAP_HAS_TURBO_RATIO 0x00000010
#define CPUPOWER_CAP_IS_SNB 0x00000020
#define CPUPOWER_CAP_INTEL_IDA 0x00000040
+#define CPUPOWER_CAP_AMD_RDPRU 0x00000080
#define CPUPOWER_AMD_CPBDIS 0x02000000
diff --git a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
index afb2e6f8edd3..e7d48cb563c0 100644
--- a/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
+++ b/tools/power/cpupower/utils/idle_monitor/mperf_monitor.c
@@ -19,6 +19,10 @@
#define MSR_APERF 0xE8
#define MSR_MPERF 0xE7
+#define RDPRU ".byte 0x0f, 0x01, 0xfd"
+#define RDPRU_ECX_MPERF 0
+#define RDPRU_ECX_APERF 1
+
#define MSR_TSC 0x10
#define MSR_AMD_HWCR 0xc0010015
@@ -89,6 +93,8 @@ static int mperf_get_tsc(unsigned long long *tsc)
static int get_aperf_mperf(int cpu, unsigned long long *aval,
unsigned long long *mval)
{
+ unsigned long low_a, high_a;
+ unsigned long low_m, high_m;
int ret;
/*
@@ -101,6 +107,20 @@ static int get_aperf_mperf(int cpu, unsigned long long *aval,
return 1;
}
+ if (cpupower_cpu_info.caps & CPUPOWER_CAP_AMD_RDPRU) {
+ asm volatile(RDPRU
+ : "=a" (low_a), "=d" (high_a)
+ : "c" (RDPRU_ECX_APERF));
+ asm volatile(RDPRU
+ : "=a" (low_m), "=d" (high_m)
+ : "c" (RDPRU_ECX_MPERF));
+
+ *aval = ((low_a) | (high_a) << 32);
+ *mval = ((low_m) | (high_m) << 32);
+
+ return 0;
+ }
+
ret = read_msr(cpu, MSR_APERF, aval);
ret |= read_msr(cpu, MSR_MPERF, mval);
--
2.17.1
Thanks Natarajan,
Shuah: I would be ok to queue this up...
Thanks,
Thomas
On 11/6/19 4:32 AM, Thomas Renninger wrote:
> Thanks Natarajan,
>
> Shuah: I would be ok to queue this up...
>
> Thanks,
>
> Thomas
>
>
>
Thomas,
Already queued up in
https://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux.git/log/?h=cpupower
Will send pull request o Rafael today to get these into pm tree.
thanks,
-- Shuah