The TPMI (Topology Aware Register and PM Capsule Interface) provides a
flexible, extendable and PCIe enumerable MMIO interface for PM features.
The TPMI documentation can be downloaded from:
https://github.com/intel/tpmi_power_management
Intel RAPL (Running Average Power Limit) is one of the features that
benefit from this. Using TPMI Interface has advantage over traditional MSR
(Model Specific Register) interface, where a thread needs to be scheduled
on the target CPU to read or write. Also the RAPL features vary between
CPU models, and hence lot of model specific code. Here TPMI provides an
architectural interface by providing hierarchical tables and fields,
which will not need any model specific implementation.
Given that there are some differences between RAPL TPMI Interface and the
existing RAPL MSR/MMIO Interface, this patch series improves the RAPL
common code to satisfy the new requirements from TPMI interface, and then
introduces the RAPL TPMI Interface driver.
Patch 1-4 cleanups and preparation work.
Patch 5 adds support for per Domain Unit register.
Patch 6-10 improves Power Limits handling, and provides support
for per Power Limit register, and per Power Limit Lock.
Patch 11-12 support rapl_package without online CPUs. So that TPMI
rapl_package still works with whole package offlined.
Patch 13-15 introduces RAPL Core support for TPMI Interface and the
RAPL TPMI Interface driver.
This series depends on the TPMI base driver which has been merged in 6.3-rc1.
thanks,
rui
---
Changes since v1:
- use set_defaults() and variable name 'defaults' for rapl_defaults
structure.
- use 'rpi_default' instead of 'rpis' for the default rapl primitive
information of MSR/MMIO Interface.
- rephase the changelog of patch 7/15.
- change the subject of patch 10/15 and use a helper for getting the
primitive for power limit LOCK bit control.
----------------------------------------------------------------
Zhang Rui (15):
powercap/intel_rapl: Remove unused field in struct rapl_if_priv
powercap/intel_rapl: Allow probing without CPUID match
powercap/intel_rapl: Support per Interface rapl_defaults
powercap/intel_rapl: Support per Interface primitive information
powercap/intel_rapl: Support per domain energy/power/time unit
powercap/intel_rapl: Use index to initialize primitive information
powercap/intel_rapl: Change primitive order
powercap/intel_rapl: Use bitmap for Power Limits
powercap/intel_rapl: Cleanup Power Limits support
powercap/intel_rapl: Introduce per Power Limit lock
powercap/intel_rapl: Remove redundant cpu parameter
powercap/intel_rapl: Make cpu optional for rapl_package
powercap/intel_rapl: Introduce RAPL I/F type
powercap/intel_rapl: Introduce core support for TPMI interface
powercap/intel_rapl_tpmi: Introduce RAPL TPMI interface driver
drivers/powercap/Kconfig | 14 +
drivers/powercap/Makefile | 1 +
drivers/powercap/intel_rapl_common.c | 868 ++++++++++++---------
drivers/powercap/intel_rapl_msr.c | 14 +-
drivers/powercap/intel_rapl_tpmi.c | 325 ++++++++
.../intel/int340x_thermal/processor_thermal_rapl.c | 11 +-
include/linux/intel_rapl.h | 40 +-
7 files changed, 875 insertions(+), 398 deletions(-)
create mode 100644 drivers/powercap/intel_rapl_tpmi.c
Currently, CPU model checks is used to
1. get proper rapl_defaults callbacks for RAPL MSR/MMIO Interface.
2. create a platform device node for the intel_rapl_msr driver to probe.
Both of these are only mandatory for the RAPL MSR/MMIO Interface.
Make the CPUID match optional.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 35 +++++++++++-----------------
1 file changed, 14 insertions(+), 21 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 8970c7b80884..2d2c15eea8eb 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -1528,32 +1528,25 @@ static int __init rapl_init(void)
int ret;
id = x86_match_cpu(rapl_ids);
- if (!id) {
- pr_err("driver does not support CPU family %d model %d\n",
- boot_cpu_data.x86, boot_cpu_data.x86_model);
+ if (id) {
+ rapl_defaults = (struct rapl_defaults *)id->driver_data;
- return -ENODEV;
- }
-
- rapl_defaults = (struct rapl_defaults *)id->driver_data;
-
- ret = register_pm_notifier(&rapl_pm_notifier);
- if (ret)
- return ret;
+ rapl_msr_platdev = platform_device_alloc("intel_rapl_msr", 0);
+ if (!rapl_msr_platdev)
+ return -ENOMEM;
- rapl_msr_platdev = platform_device_alloc("intel_rapl_msr", 0);
- if (!rapl_msr_platdev) {
- ret = -ENOMEM;
- goto end;
+ ret = platform_device_add(rapl_msr_platdev);
+ if (ret) {
+ platform_device_put(rapl_msr_platdev);
+ return ret;
+ }
}
- ret = platform_device_add(rapl_msr_platdev);
- if (ret)
+ ret = register_pm_notifier(&rapl_pm_notifier);
+ if (ret && rapl_msr_platdev) {
+ platform_device_del(rapl_msr_platdev);
platform_device_put(rapl_msr_platdev);
-
-end:
- if (ret)
- unregister_pm_notifier(&rapl_pm_notifier);
+ }
return ret;
}
--
2.25.1
RAPL primitive information is Interface specific.
Although current MSR and MMIO Interface share the same RAPL primitives,
new Interface like TPMI has its own RAPL primitive information.
Save the primitive information in the Interface private structure.
Plus, using variant name "rp" for struct rapl_primitive_info is
confusing because "rp" is also used for struct rapl_package.
Use "rpi" as the variant name for struct rapl_primitive_info, and rename
the previous rpi[] array to avoid conflict.
No functional change.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 50 ++++++++++++++++++----------
include/linux/intel_rapl.h | 2 ++
2 files changed, 35 insertions(+), 17 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 59e470a57342..267ec36b1649 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -654,7 +654,7 @@ static u64 rapl_unit_xlate(struct rapl_domain *rd, enum unit_type type,
}
/* in the order of enum rapl_primitives */
-static struct rapl_primitive_info rpi[] = {
+static struct rapl_primitive_info rpi_default[] = {
/* name, mask, shift, msr index, unit divisor */
PRIMITIVE_INFO_INIT(ENERGY_COUNTER, ENERGY_STATUS_MASK, 0,
RAPL_DOMAIN_REG_STATUS, ENERGY_UNIT, 0),
@@ -710,9 +710,20 @@ static struct rapl_primitive_info rpi[] = {
{NULL, 0, 0, 0},
};
+static struct rapl_primitive_info *get_rpi(struct rapl_package *rp, int prim)
+{
+ struct rapl_primitive_info *rpi = rp->priv->rpi;
+
+ if (prim < 0 || prim > NR_RAPL_PRIMITIVES || !rpi)
+ return NULL;
+
+ return &rpi[prim];
+}
+
static int rapl_config(struct rapl_package *rp)
{
rp->priv->defaults = (void *)rapl_defaults;
+ rp->priv->rpi = (void *)rpi_default;
return 0;
}
@@ -763,14 +774,14 @@ static int rapl_read_data_raw(struct rapl_domain *rd,
{
u64 value;
enum rapl_primitives prim_fixed = prim_fixups(rd, prim);
- struct rapl_primitive_info *rp = &rpi[prim_fixed];
+ struct rapl_primitive_info *rpi = get_rpi(rd->rp, prim_fixed);
struct reg_action ra;
int cpu;
- if (!rp->name || rp->flag & RAPL_PRIMITIVE_DUMMY)
+ if (!rpi || !rpi->name || rpi->flag & RAPL_PRIMITIVE_DUMMY)
return -EINVAL;
- ra.reg = rd->regs[rp->id];
+ ra.reg = rd->regs[rpi->id];
if (!ra.reg)
return -EINVAL;
@@ -778,26 +789,26 @@ static int rapl_read_data_raw(struct rapl_domain *rd,
/* domain with 2 limits has different bit */
if (prim == FW_LOCK && rd->rp->priv->limits[rd->id] == 2) {
- rp->mask = POWER_HIGH_LOCK;
- rp->shift = 63;
+ rpi->mask = POWER_HIGH_LOCK;
+ rpi->shift = 63;
}
/* non-hardware data are collected by the polling thread */
- if (rp->flag & RAPL_PRIMITIVE_DERIVED) {
+ if (rpi->flag & RAPL_PRIMITIVE_DERIVED) {
*data = rd->rdd.primitives[prim];
return 0;
}
- ra.mask = rp->mask;
+ ra.mask = rpi->mask;
if (rd->rp->priv->read_raw(cpu, &ra)) {
pr_debug("failed to read reg 0x%llx on cpu %d\n", ra.reg, cpu);
return -EIO;
}
- value = ra.value >> rp->shift;
+ value = ra.value >> rpi->shift;
if (xlate)
- *data = rapl_unit_xlate(rd, rp->unit, value, 0);
+ *data = rapl_unit_xlate(rd, rpi->unit, value, 0);
else
*data = value;
@@ -810,21 +821,24 @@ static int rapl_write_data_raw(struct rapl_domain *rd,
unsigned long long value)
{
enum rapl_primitives prim_fixed = prim_fixups(rd, prim);
- struct rapl_primitive_info *rp = &rpi[prim_fixed];
+ struct rapl_primitive_info *rpi = get_rpi(rd->rp, prim_fixed);
int cpu;
u64 bits;
struct reg_action ra;
int ret;
+ if (!rpi || !rpi->name || rpi->flag & RAPL_PRIMITIVE_DUMMY)
+ return -EINVAL;
+
cpu = rd->rp->lead_cpu;
- bits = rapl_unit_xlate(rd, rp->unit, value, 1);
- bits <<= rp->shift;
- bits &= rp->mask;
+ bits = rapl_unit_xlate(rd, rpi->unit, value, 1);
+ bits <<= rpi->shift;
+ bits &= rpi->mask;
memset(&ra, 0, sizeof(ra));
- ra.reg = rd->regs[rp->id];
- ra.mask = rp->mask;
+ ra.reg = rd->regs[rpi->id];
+ ra.mask = rpi->mask;
ra.value = bits;
ret = rd->rp->priv->write_raw(cpu, &ra);
@@ -1176,8 +1190,10 @@ static void rapl_update_domain_data(struct rapl_package *rp)
rp->domains[dmn].name);
/* exclude non-raw primitives */
for (prim = 0; prim < NR_RAW_PRIMITIVES; prim++) {
+ struct rapl_primitive_info *rpi = get_rpi(rp, prim);
+
if (!rapl_read_data_raw(&rp->domains[dmn], prim,
- rpi[prim].unit, &val))
+ rpi->unit, &val))
rp->domains[dmn].rdd.primitives[prim] = val;
}
}
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index ebd1cad78212..f51e2df7130e 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -122,6 +122,7 @@ struct reg_action {
* @write_raw: Callback for writing RAPL interface specific
* registers.
* @defaults: internal pointer to interface default settings
+ * @rpi: internal pointer to interface primitive info
*/
struct rapl_if_priv {
struct powercap_control_type *control_type;
@@ -132,6 +133,7 @@ struct rapl_if_priv {
int (*read_raw)(int cpu, struct reg_action *ra);
int (*write_raw)(int cpu, struct reg_action *ra);
void *defaults;
+ void *rpi;
};
/* maximum rapl package domain name: package-%d-die-%d */
--
2.25.1
After commit f1e8d7560d30 ("powercap/intel_rapl: enumerate Psys RAPL
domain together with package RAPL domain"), the platform_rapl_domain field
is not used anymore. Remove it from rapl_if_priv structure.
Fixes: f1e8d7560d30 ("powercap/intel_rapl: enumerate Psys RAPL domain together with package RAPL domain")
Signed-off-by: Zhang Rui <[email protected]>
---
include/linux/intel_rapl.h | 1 -
1 file changed, 1 deletion(-)
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index 9f4b6f5b822f..828557645770 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -124,7 +124,6 @@ struct reg_action {
*/
struct rapl_if_priv {
struct powercap_control_type *control_type;
- struct rapl_domain *platform_rapl_domain;
enum cpuhp_state pcap_rapl_online;
u64 reg_unit;
u64 regs[RAPL_DOMAIN_MAX][RAPL_DOMAIN_REG_MAX];
--
2.25.1
Currently, the RAPL primitive information array is required to be
initialized in the order of enum rapl_primitives.
This can break easily, especially when different RAPL Interfaces may
support different sets of primitives.
Convert the code to initialize the primitive information using array
index explicitly.
No functional change.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 54 ++++++++++++++--------------
1 file changed, 26 insertions(+), 28 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 3625d4466cb3..97d9e1e628e2 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -629,61 +629,59 @@ static u64 rapl_unit_xlate(struct rapl_domain *rd, enum unit_type type,
return div64_u64(value, scale);
}
-/* in the order of enum rapl_primitives */
-static struct rapl_primitive_info rpi_default[] = {
+static struct rapl_primitive_info rpi_default[NR_RAPL_PRIMITIVES] = {
/* name, mask, shift, msr index, unit divisor */
- PRIMITIVE_INFO_INIT(ENERGY_COUNTER, ENERGY_STATUS_MASK, 0,
+ [ENERGY_COUNTER] = PRIMITIVE_INFO_INIT(ENERGY_COUNTER, ENERGY_STATUS_MASK, 0,
RAPL_DOMAIN_REG_STATUS, ENERGY_UNIT, 0),
- PRIMITIVE_INFO_INIT(POWER_LIMIT1, POWER_LIMIT1_MASK, 0,
+ [POWER_LIMIT1] = PRIMITIVE_INFO_INIT(POWER_LIMIT1, POWER_LIMIT1_MASK, 0,
RAPL_DOMAIN_REG_LIMIT, POWER_UNIT, 0),
- PRIMITIVE_INFO_INIT(POWER_LIMIT2, POWER_LIMIT2_MASK, 32,
+ [POWER_LIMIT2] = PRIMITIVE_INFO_INIT(POWER_LIMIT2, POWER_LIMIT2_MASK, 32,
RAPL_DOMAIN_REG_LIMIT, POWER_UNIT, 0),
- PRIMITIVE_INFO_INIT(POWER_LIMIT4, POWER_LIMIT4_MASK, 0,
+ [POWER_LIMIT4] = PRIMITIVE_INFO_INIT(POWER_LIMIT4, POWER_LIMIT4_MASK, 0,
RAPL_DOMAIN_REG_PL4, POWER_UNIT, 0),
- PRIMITIVE_INFO_INIT(FW_LOCK, POWER_LOW_LOCK, 31,
+ [FW_LOCK] = PRIMITIVE_INFO_INIT(FW_LOCK, POWER_LOW_LOCK, 31,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(PL1_ENABLE, POWER_LIMIT1_ENABLE, 15,
+ [PL1_ENABLE] = PRIMITIVE_INFO_INIT(PL1_ENABLE, POWER_LIMIT1_ENABLE, 15,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(PL1_CLAMP, POWER_LIMIT1_CLAMP, 16,
+ [PL1_CLAMP] = PRIMITIVE_INFO_INIT(PL1_CLAMP, POWER_LIMIT1_CLAMP, 16,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(PL2_ENABLE, POWER_LIMIT2_ENABLE, 47,
+ [PL2_ENABLE] = PRIMITIVE_INFO_INIT(PL2_ENABLE, POWER_LIMIT2_ENABLE, 47,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(PL2_CLAMP, POWER_LIMIT2_CLAMP, 48,
+ [PL2_CLAMP] = PRIMITIVE_INFO_INIT(PL2_CLAMP, POWER_LIMIT2_CLAMP, 48,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(PL4_ENABLE, POWER_LIMIT4_MASK, 0,
+ [PL4_ENABLE] = PRIMITIVE_INFO_INIT(PL4_ENABLE, POWER_LIMIT4_MASK, 0,
RAPL_DOMAIN_REG_PL4, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(TIME_WINDOW1, TIME_WINDOW1_MASK, 17,
+ [TIME_WINDOW1] = PRIMITIVE_INFO_INIT(TIME_WINDOW1, TIME_WINDOW1_MASK, 17,
RAPL_DOMAIN_REG_LIMIT, TIME_UNIT, 0),
- PRIMITIVE_INFO_INIT(TIME_WINDOW2, TIME_WINDOW2_MASK, 49,
+ [TIME_WINDOW2] = PRIMITIVE_INFO_INIT(TIME_WINDOW2, TIME_WINDOW2_MASK, 49,
RAPL_DOMAIN_REG_LIMIT, TIME_UNIT, 0),
- PRIMITIVE_INFO_INIT(THERMAL_SPEC_POWER, POWER_INFO_THERMAL_SPEC_MASK,
+ [THERMAL_SPEC_POWER] = PRIMITIVE_INFO_INIT(THERMAL_SPEC_POWER, POWER_INFO_THERMAL_SPEC_MASK,
0, RAPL_DOMAIN_REG_INFO, POWER_UNIT, 0),
- PRIMITIVE_INFO_INIT(MAX_POWER, POWER_INFO_MAX_MASK, 32,
+ [MAX_POWER] = PRIMITIVE_INFO_INIT(MAX_POWER, POWER_INFO_MAX_MASK, 32,
RAPL_DOMAIN_REG_INFO, POWER_UNIT, 0),
- PRIMITIVE_INFO_INIT(MIN_POWER, POWER_INFO_MIN_MASK, 16,
+ [MIN_POWER] = PRIMITIVE_INFO_INIT(MIN_POWER, POWER_INFO_MIN_MASK, 16,
RAPL_DOMAIN_REG_INFO, POWER_UNIT, 0),
- PRIMITIVE_INFO_INIT(MAX_TIME_WINDOW, POWER_INFO_MAX_TIME_WIN_MASK, 48,
+ [MAX_TIME_WINDOW] = PRIMITIVE_INFO_INIT(MAX_TIME_WINDOW, POWER_INFO_MAX_TIME_WIN_MASK, 48,
RAPL_DOMAIN_REG_INFO, TIME_UNIT, 0),
- PRIMITIVE_INFO_INIT(THROTTLED_TIME, PERF_STATUS_THROTTLE_TIME_MASK, 0,
+ [THROTTLED_TIME] = PRIMITIVE_INFO_INIT(THROTTLED_TIME, PERF_STATUS_THROTTLE_TIME_MASK, 0,
RAPL_DOMAIN_REG_PERF, TIME_UNIT, 0),
- PRIMITIVE_INFO_INIT(PRIORITY_LEVEL, PP_POLICY_MASK, 0,
+ [PRIORITY_LEVEL] = PRIMITIVE_INFO_INIT(PRIORITY_LEVEL, PP_POLICY_MASK, 0,
RAPL_DOMAIN_REG_POLICY, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(PSYS_POWER_LIMIT1, PSYS_POWER_LIMIT1_MASK, 0,
+ [PSYS_POWER_LIMIT1] = PRIMITIVE_INFO_INIT(PSYS_POWER_LIMIT1, PSYS_POWER_LIMIT1_MASK, 0,
RAPL_DOMAIN_REG_LIMIT, POWER_UNIT, 0),
- PRIMITIVE_INFO_INIT(PSYS_POWER_LIMIT2, PSYS_POWER_LIMIT2_MASK, 32,
+ [PSYS_POWER_LIMIT2] = PRIMITIVE_INFO_INIT(PSYS_POWER_LIMIT2, PSYS_POWER_LIMIT2_MASK, 32,
RAPL_DOMAIN_REG_LIMIT, POWER_UNIT, 0),
- PRIMITIVE_INFO_INIT(PSYS_PL1_ENABLE, PSYS_POWER_LIMIT1_ENABLE, 17,
+ [PSYS_PL1_ENABLE] = PRIMITIVE_INFO_INIT(PSYS_PL1_ENABLE, PSYS_POWER_LIMIT1_ENABLE, 17,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(PSYS_PL2_ENABLE, PSYS_POWER_LIMIT2_ENABLE, 49,
+ [PSYS_PL2_ENABLE] = PRIMITIVE_INFO_INIT(PSYS_PL2_ENABLE, PSYS_POWER_LIMIT2_ENABLE, 49,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
- PRIMITIVE_INFO_INIT(PSYS_TIME_WINDOW1, PSYS_TIME_WINDOW1_MASK, 19,
+ [PSYS_TIME_WINDOW1] = PRIMITIVE_INFO_INIT(PSYS_TIME_WINDOW1, PSYS_TIME_WINDOW1_MASK, 19,
RAPL_DOMAIN_REG_LIMIT, TIME_UNIT, 0),
- PRIMITIVE_INFO_INIT(PSYS_TIME_WINDOW2, PSYS_TIME_WINDOW2_MASK, 51,
+ [PSYS_TIME_WINDOW2] = PRIMITIVE_INFO_INIT(PSYS_TIME_WINDOW2, PSYS_TIME_WINDOW2_MASK, 51,
RAPL_DOMAIN_REG_LIMIT, TIME_UNIT, 0),
/* non-hardware */
- PRIMITIVE_INFO_INIT(AVERAGE_POWER, 0, 0, 0, POWER_UNIT,
+ [AVERAGE_POWER] = PRIMITIVE_INFO_INIT(AVERAGE_POWER, 0, 0, 0, POWER_UNIT,
RAPL_PRIMITIVE_DERIVED),
- {NULL, 0, 0, 0},
};
static struct rapl_primitive_info *get_rpi(struct rapl_package *rp, int prim)
--
2.25.1
RAPL MSR/MMIO Interface has package scope unit register but some RAPL
domains like Dram/Psys may use a fixed energy unit value instead of the
default unit value on certain platforms.
RAPL TPMI Interface supports per domain unit register.
For the above reasons, add support for per domain unit register and per
domain energy/power/time unit.
When per domain unit register is not available, use the package scope
unit register as the per domain unit register for each RAPL domain so
that this change is transparent to MSR/MMIO Interface.
No functional change intended.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 128 +++++++++++++++------------
include/linux/intel_rapl.h | 8 +-
2 files changed, 73 insertions(+), 63 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 267ec36b1649..3625d4466cb3 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -105,9 +105,9 @@ static const char pl4_name[] = "peak_power";
struct rapl_defaults {
u8 floor_freq_reg_addr;
- int (*check_unit)(struct rapl_package *rp, int cpu);
+ int (*check_unit)(struct rapl_domain *rd, int cpu);
void (*set_floor_freq)(struct rapl_domain *rd, bool mode);
- u64 (*compute_time_window)(struct rapl_package *rp, u64 val,
+ u64 (*compute_time_window)(struct rapl_domain *rd, u64 val,
bool to_raw);
unsigned int dram_domain_energy_unit;
unsigned int psys_domain_energy_unit;
@@ -557,7 +557,6 @@ static void rapl_init_domains(struct rapl_package *rp)
enum rapl_domain_type i;
enum rapl_domain_reg_id j;
struct rapl_domain *rd = rp->domains;
- struct rapl_defaults *defaults = get_defaults(rp);
for (i = 0; i < RAPL_DOMAIN_MAX; i++) {
unsigned int mask = rp->domain_map & (1 << i);
@@ -596,24 +595,6 @@ static void rapl_init_domains(struct rapl_package *rp)
for (j = 0; j < RAPL_DOMAIN_REG_MAX; j++)
rd->regs[j] = rp->priv->regs[i][j];
- switch (i) {
- case RAPL_DOMAIN_DRAM:
- rd->domain_energy_unit =
- defaults->dram_domain_energy_unit;
- if (rd->domain_energy_unit)
- pr_info("DRAM domain energy unit %dpj\n",
- rd->domain_energy_unit);
- break;
- case RAPL_DOMAIN_PLATFORM:
- rd->domain_energy_unit =
- defaults->psys_domain_energy_unit;
- if (rd->domain_energy_unit)
- pr_info("Platform domain energy unit %dpj\n",
- rd->domain_energy_unit);
- break;
- default:
- break;
- }
rd++;
}
}
@@ -622,24 +603,19 @@ static u64 rapl_unit_xlate(struct rapl_domain *rd, enum unit_type type,
u64 value, int to_raw)
{
u64 units = 1;
- struct rapl_package *rp = rd->rp;
- struct rapl_defaults *defaults = get_defaults(rp);
+ struct rapl_defaults *defaults = get_defaults(rd->rp);
u64 scale = 1;
switch (type) {
case POWER_UNIT:
- units = rp->power_unit;
+ units = rd->power_unit;
break;
case ENERGY_UNIT:
scale = ENERGY_UNIT_SCALE;
- /* per domain unit takes precedence */
- if (rd->domain_energy_unit)
- units = rd->domain_energy_unit;
- else
- units = rp->energy_unit;
+ units = rd->energy_unit;
break;
case TIME_UNIT:
- return defaults->compute_time_window(rp, value, to_raw);
+ return defaults->compute_time_window(rd, value, to_raw);
case ARBITRARY_UNIT:
default:
return value;
@@ -857,58 +833,58 @@ static int rapl_write_data_raw(struct rapl_domain *rd,
* power unit : microWatts : Represented in milliWatts by default
* time unit : microseconds: Represented in seconds by default
*/
-static int rapl_check_unit_core(struct rapl_package *rp, int cpu)
+static int rapl_check_unit_core(struct rapl_domain *rd, int cpu)
{
struct reg_action ra;
u32 value;
- ra.reg = rp->priv->reg_unit;
+ ra.reg = rd->regs[RAPL_DOMAIN_REG_UNIT];
ra.mask = ~0;
- if (rp->priv->read_raw(cpu, &ra)) {
+ if (rd->rp->priv->read_raw(cpu, &ra)) {
pr_err("Failed to read power unit REG 0x%llx on CPU %d, exit.\n",
- rp->priv->reg_unit, cpu);
+ ra.reg, cpu);
return -ENODEV;
}
value = (ra.value & ENERGY_UNIT_MASK) >> ENERGY_UNIT_OFFSET;
- rp->energy_unit = ENERGY_UNIT_SCALE * 1000000 / (1 << value);
+ rd->energy_unit = ENERGY_UNIT_SCALE * 1000000 / (1 << value);
value = (ra.value & POWER_UNIT_MASK) >> POWER_UNIT_OFFSET;
- rp->power_unit = 1000000 / (1 << value);
+ rd->power_unit = 1000000 / (1 << value);
value = (ra.value & TIME_UNIT_MASK) >> TIME_UNIT_OFFSET;
- rp->time_unit = 1000000 / (1 << value);
+ rd->time_unit = 1000000 / (1 << value);
- pr_debug("Core CPU %s energy=%dpJ, time=%dus, power=%duW\n",
- rp->name, rp->energy_unit, rp->time_unit, rp->power_unit);
+ pr_debug("Core CPU %s:%s energy=%dpJ, time=%dus, power=%duW\n",
+ rd->rp->name, rd->name, rd->energy_unit, rd->time_unit, rd->power_unit);
return 0;
}
-static int rapl_check_unit_atom(struct rapl_package *rp, int cpu)
+static int rapl_check_unit_atom(struct rapl_domain *rd, int cpu)
{
struct reg_action ra;
u32 value;
- ra.reg = rp->priv->reg_unit;
+ ra.reg = rd->regs[RAPL_DOMAIN_REG_UNIT];
ra.mask = ~0;
- if (rp->priv->read_raw(cpu, &ra)) {
+ if (rd->rp->priv->read_raw(cpu, &ra)) {
pr_err("Failed to read power unit REG 0x%llx on CPU %d, exit.\n",
- rp->priv->reg_unit, cpu);
+ ra.reg, cpu);
return -ENODEV;
}
value = (ra.value & ENERGY_UNIT_MASK) >> ENERGY_UNIT_OFFSET;
- rp->energy_unit = ENERGY_UNIT_SCALE * 1 << value;
+ rd->energy_unit = ENERGY_UNIT_SCALE * 1 << value;
value = (ra.value & POWER_UNIT_MASK) >> POWER_UNIT_OFFSET;
- rp->power_unit = (1 << value) * 1000;
+ rd->power_unit = (1 << value) * 1000;
value = (ra.value & TIME_UNIT_MASK) >> TIME_UNIT_OFFSET;
- rp->time_unit = 1000000 / (1 << value);
+ rd->time_unit = 1000000 / (1 << value);
- pr_debug("Atom %s energy=%dpJ, time=%dus, power=%duW\n",
- rp->name, rp->energy_unit, rp->time_unit, rp->power_unit);
+ pr_debug("Atom %s:%s energy=%dpJ, time=%dus, power=%duW\n",
+ rd->rp->name, rd->name, rd->energy_unit, rd->time_unit, rd->power_unit);
return 0;
}
@@ -1011,7 +987,7 @@ static void set_floor_freq_atom(struct rapl_domain *rd, bool enable)
defaults->floor_freq_reg_addr, mdata);
}
-static u64 rapl_compute_time_window_core(struct rapl_package *rp, u64 value,
+static u64 rapl_compute_time_window_core(struct rapl_domain *rd, u64 value,
bool to_raw)
{
u64 f, y; /* fraction and exp. used for time unit */
@@ -1023,12 +999,12 @@ static u64 rapl_compute_time_window_core(struct rapl_package *rp, u64 value,
if (!to_raw) {
f = (value & 0x60) >> 5;
y = value & 0x1f;
- value = (1 << y) * (4 + f) * rp->time_unit / 4;
+ value = (1 << y) * (4 + f) * rd->time_unit / 4;
} else {
- if (value < rp->time_unit)
+ if (value < rd->time_unit)
return 0;
- do_div(value, rp->time_unit);
+ do_div(value, rd->time_unit);
y = ilog2(value);
/*
@@ -1044,7 +1020,7 @@ static u64 rapl_compute_time_window_core(struct rapl_package *rp, u64 value,
return value;
}
-static u64 rapl_compute_time_window_atom(struct rapl_package *rp, u64 value,
+static u64 rapl_compute_time_window_atom(struct rapl_domain *rd, u64 value,
bool to_raw)
{
/*
@@ -1052,9 +1028,9 @@ static u64 rapl_compute_time_window_atom(struct rapl_package *rp, u64 value,
* where time_unit is default to 1 sec. Never 0.
*/
if (!to_raw)
- return (value) ? value * rp->time_unit : rp->time_unit;
+ return (value) ? value * rd->time_unit : rd->time_unit;
- value = div64_u64(value, rp->time_unit);
+ value = div64_u64(value, rd->time_unit);
return value;
}
@@ -1299,6 +1275,40 @@ static int rapl_check_domain(int cpu, int domain, struct rapl_package *rp)
return 0;
}
+/*
+ * Get per domain energy/power/time unit.
+ * RAPL Interfaces without per domain unit register will use the package
+ * scope unit register to set per domain units.
+ */
+static int rapl_get_domain_unit(struct rapl_domain *rd)
+{
+ struct rapl_defaults *defaults = get_defaults(rd->rp);
+ int ret;
+
+ if (!rd->regs[RAPL_DOMAIN_REG_UNIT]) {
+ if (!rd->rp->priv->reg_unit) {
+ pr_err("No valid Unit register found\n");
+ return -ENODEV;
+ }
+ rd->regs[RAPL_DOMAIN_REG_UNIT] = rd->rp->priv->reg_unit;
+ }
+
+ if (!defaults->check_unit) {
+ pr_err("missing .check_unit() callback\n");
+ return -ENODEV;
+ }
+
+ ret = defaults->check_unit(rd, rd->rp->lead_cpu);
+ if (ret)
+ return ret;
+
+ if (rd->id == RAPL_DOMAIN_DRAM && defaults->dram_domain_energy_unit)
+ rd->energy_unit = defaults->dram_domain_energy_unit;
+ if (rd->id == RAPL_DOMAIN_PLATFORM && defaults->psys_domain_energy_unit)
+ rd->energy_unit = defaults->psys_domain_energy_unit;
+ return 0;
+}
+
/*
* Check if power limits are available. Two cases when they are not available:
* 1. Locked by BIOS, in this case we still provide read-only access so that
@@ -1359,8 +1369,10 @@ static int rapl_detect_domains(struct rapl_package *rp, int cpu)
rapl_init_domains(rp);
- for (rd = rp->domains; rd < rp->domains + rp->nr_domains; rd++)
+ for (rd = rp->domains; rd < rp->domains + rp->nr_domains; rd++) {
+ rapl_get_domain_unit(rd);
rapl_detect_powerlimit(rd);
+ }
return 0;
}
@@ -1418,7 +1430,6 @@ struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv)
{
int id = topology_logical_die_id(cpu);
struct rapl_package *rp;
- struct rapl_defaults *defaults;
int ret;
rp = kzalloc(sizeof(struct rapl_package), GFP_KERNEL);
@@ -1442,9 +1453,8 @@ struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv)
snprintf(rp->name, PACKAGE_DOMAIN_NAME_LENGTH, "package-%d",
topology_physical_package_id(cpu));
- defaults = get_defaults(rp);
/* check if the package contains valid domains */
- if (rapl_detect_domains(rp, cpu) || defaults->check_unit(rp, cpu)) {
+ if (rapl_detect_domains(rp, cpu)) {
ret = -ENODEV;
goto err_free_package;
}
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index f51e2df7130e..936fb8c3082c 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -30,6 +30,7 @@ enum rapl_domain_reg_id {
RAPL_DOMAIN_REG_POLICY,
RAPL_DOMAIN_REG_INFO,
RAPL_DOMAIN_REG_PL4,
+ RAPL_DOMAIN_REG_UNIT,
RAPL_DOMAIN_REG_MAX,
};
@@ -96,7 +97,9 @@ struct rapl_domain {
struct rapl_power_limit rpl[NR_POWER_LIMITS];
u64 attr_map; /* track capabilities */
unsigned int state;
- unsigned int domain_energy_unit;
+ unsigned int power_unit;
+ unsigned int energy_unit;
+ unsigned int time_unit;
struct rapl_package *rp;
};
@@ -143,9 +146,6 @@ struct rapl_package {
unsigned int id; /* logical die id, equals physical 1-die systems */
unsigned int nr_domains;
unsigned long domain_map; /* bit map of active domains */
- unsigned int power_unit;
- unsigned int energy_unit;
- unsigned int time_unit;
struct rapl_domain *domains; /* array of domains, sized at runtime */
struct powercap_zone *power_zone; /* keep track of parent zone */
unsigned long power_limit_irq; /* keep track of package power limit
--
2.25.1
The same set of operations are shared by different Powert Limits,
including Power Limit get/set, Power Limit enable/disable, clamping
enable/disable, time window get/set, and max power get/set, etc.
But the same operation for different Power Limit has different
primitives because they use different registers/register bits.
A lot of dirty/duplicate code was introduced to handle this difference.
Instead of using hardcoded primitive name directly, using Power Limit id
+ operation type is much cleaner.
For this sense, move POWER_LIMIT1/POWER_LIMIT2/POWER_LIMIT4 to the
beginning of enum rapl_primitives so that they can be reused as
Power Limit ids.
No functional change.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 4 ++--
include/linux/intel_rapl.h | 5 +++--
2 files changed, 5 insertions(+), 4 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 97d9e1e628e2..154f93b3dee5 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -631,14 +631,14 @@ static u64 rapl_unit_xlate(struct rapl_domain *rd, enum unit_type type,
static struct rapl_primitive_info rpi_default[NR_RAPL_PRIMITIVES] = {
/* name, mask, shift, msr index, unit divisor */
- [ENERGY_COUNTER] = PRIMITIVE_INFO_INIT(ENERGY_COUNTER, ENERGY_STATUS_MASK, 0,
- RAPL_DOMAIN_REG_STATUS, ENERGY_UNIT, 0),
[POWER_LIMIT1] = PRIMITIVE_INFO_INIT(POWER_LIMIT1, POWER_LIMIT1_MASK, 0,
RAPL_DOMAIN_REG_LIMIT, POWER_UNIT, 0),
[POWER_LIMIT2] = PRIMITIVE_INFO_INIT(POWER_LIMIT2, POWER_LIMIT2_MASK, 32,
RAPL_DOMAIN_REG_LIMIT, POWER_UNIT, 0),
[POWER_LIMIT4] = PRIMITIVE_INFO_INIT(POWER_LIMIT4, POWER_LIMIT4_MASK, 0,
RAPL_DOMAIN_REG_PL4, POWER_UNIT, 0),
+ [ENERGY_COUNTER] = PRIMITIVE_INFO_INIT(ENERGY_COUNTER, ENERGY_STATUS_MASK, 0,
+ RAPL_DOMAIN_REG_STATUS, ENERGY_UNIT, 0),
[FW_LOCK] = PRIMITIVE_INFO_INIT(FW_LOCK, POWER_LOW_LOCK, 31,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
[PL1_ENABLE] = PRIMITIVE_INFO_INIT(PL1_ENABLE, POWER_LIMIT1_ENABLE, 15,
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index 936fb8c3082c..bbd03b17dc8d 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -37,10 +37,10 @@ enum rapl_domain_reg_id {
struct rapl_domain;
enum rapl_primitives {
- ENERGY_COUNTER,
POWER_LIMIT1,
POWER_LIMIT2,
POWER_LIMIT4,
+ ENERGY_COUNTER,
FW_LOCK,
PL1_ENABLE, /* power limit 1, aka long term */
@@ -75,7 +75,8 @@ struct rapl_domain_data {
unsigned long timestamp;
};
-#define NR_POWER_LIMITS (3)
+#define NR_POWER_LIMITS (POWER_LIMIT4 + 1)
+
struct rapl_power_limit {
struct powercap_zone_constraint *constraint;
int prim_id; /* primitive ID used to enable */
--
2.25.1
Currently, a RAPL package is registered with the number of Power Limits
supported in each RAPL domain. But this doesn't tell which Power Limits
are available. Using the number of Power Limits supported to guess the
availability of each Power Limit is fragile.
Use bitmap to represent the availability of each Power Limit.
Note that PL1 is mandatory thus it does not need to be set explicitly by
the RAPL Interface drivers.
No functional change intended.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 14 ++++++--------
drivers/powercap/intel_rapl_msr.c | 6 +++---
.../intel/int340x_thermal/processor_thermal_rapl.c | 4 ++--
3 files changed, 11 insertions(+), 13 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 154f93b3dee5..8e77df42257a 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -574,20 +574,18 @@ static void rapl_init_domains(struct rapl_package *rp)
rapl_domain_names[i]);
rd->id = i;
+
+ /* PL1 is supported by default */
+ rp->priv->limits[i] |= BIT(POWER_LIMIT1);
rd->rpl[0].prim_id = PL1_ENABLE;
rd->rpl[0].name = pl1_name;
- /*
- * The PL2 power domain is applicable for limits two
- * and limits three
- */
- if (rp->priv->limits[i] >= 2) {
+ if (rp->priv->limits[i] & BIT(POWER_LIMIT2)) {
rd->rpl[1].prim_id = PL2_ENABLE;
rd->rpl[1].name = pl2_name;
}
- /* Enable PL4 domain if the total power limits are three */
- if (rp->priv->limits[i] == 3) {
+ if (rp->priv->limits[i] & BIT(POWER_LIMIT4)) {
rd->rpl[2].prim_id = PL4_ENABLE;
rd->rpl[2].name = pl4_name;
}
@@ -762,7 +760,7 @@ static int rapl_read_data_raw(struct rapl_domain *rd,
cpu = rd->rp->lead_cpu;
/* domain with 2 limits has different bit */
- if (prim == FW_LOCK && rd->rp->priv->limits[rd->id] == 2) {
+ if (prim == FW_LOCK && (rd->rp->priv->limits[rd->id] & BIT(POWER_LIMIT2))) {
rpi->mask = POWER_HIGH_LOCK;
rpi->shift = 63;
}
diff --git a/drivers/powercap/intel_rapl_msr.c b/drivers/powercap/intel_rapl_msr.c
index a27673706c3d..6fe5e556aa51 100644
--- a/drivers/powercap/intel_rapl_msr.c
+++ b/drivers/powercap/intel_rapl_msr.c
@@ -45,8 +45,8 @@ static struct rapl_if_priv rapl_msr_priv_intel = {
MSR_DRAM_POWER_LIMIT, MSR_DRAM_ENERGY_STATUS, MSR_DRAM_PERF_STATUS, 0, MSR_DRAM_POWER_INFO },
.regs[RAPL_DOMAIN_PLATFORM] = {
MSR_PLATFORM_POWER_LIMIT, MSR_PLATFORM_ENERGY_STATUS, 0, 0, 0},
- .limits[RAPL_DOMAIN_PACKAGE] = 2,
- .limits[RAPL_DOMAIN_PLATFORM] = 2,
+ .limits[RAPL_DOMAIN_PACKAGE] = BIT(POWER_LIMIT2),
+ .limits[RAPL_DOMAIN_PLATFORM] = BIT(POWER_LIMIT2),
};
static struct rapl_if_priv rapl_msr_priv_amd = {
@@ -169,7 +169,7 @@ static int rapl_msr_probe(struct platform_device *pdev)
rapl_msr_priv->write_raw = rapl_msr_write_raw;
if (id) {
- rapl_msr_priv->limits[RAPL_DOMAIN_PACKAGE] = 3;
+ rapl_msr_priv->limits[RAPL_DOMAIN_PACKAGE] |= BIT(POWER_LIMIT4);
rapl_msr_priv->regs[RAPL_DOMAIN_PACKAGE][RAPL_DOMAIN_REG_PL4] =
MSR_VR_CURRENT_CONFIG;
pr_info("PL4 support detected.\n");
diff --git a/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c b/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
index a205221ec8df..e070239106f5 100644
--- a/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
+++ b/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
@@ -15,8 +15,8 @@ static const struct rapl_mmio_regs rapl_mmio_default = {
.reg_unit = 0x5938,
.regs[RAPL_DOMAIN_PACKAGE] = { 0x59a0, 0x593c, 0x58f0, 0, 0x5930},
.regs[RAPL_DOMAIN_DRAM] = { 0x58e0, 0x58e8, 0x58ec, 0, 0},
- .limits[RAPL_DOMAIN_PACKAGE] = 2,
- .limits[RAPL_DOMAIN_DRAM] = 2,
+ .limits[RAPL_DOMAIN_PACKAGE] = BIT(POWER_LIMIT2),
+ .limits[RAPL_DOMAIN_DRAM] = BIT(POWER_LIMIT2),
};
static int rapl_mmio_cpu_online(unsigned int cpu)
--
2.25.1
rapl_defaults is Interface specific.
Although current MSR and MMIO Interface share the same rapl_defaults,
new Interface like TPMI need its own rapl_defaults callbacks.
Save the rapl_defaults information in the Interface private structure.
No functional change.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 46 ++++++++++++++++++++--------
include/linux/intel_rapl.h | 2 ++
2 files changed, 35 insertions(+), 13 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 2d2c15eea8eb..59e470a57342 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -115,6 +115,11 @@ struct rapl_defaults {
};
static struct rapl_defaults *rapl_defaults;
+static struct rapl_defaults *get_defaults(struct rapl_package *rp)
+{
+ return rp->priv->defaults;
+}
+
/* Sideband MBI registers */
#define IOSF_CPU_POWER_BUDGET_CTL_BYT (0x2)
#define IOSF_CPU_POWER_BUDGET_CTL_TNG (0xdf)
@@ -227,14 +232,15 @@ static int find_nr_power_limit(struct rapl_domain *rd)
static int set_domain_enable(struct powercap_zone *power_zone, bool mode)
{
struct rapl_domain *rd = power_zone_to_rapl_domain(power_zone);
+ struct rapl_defaults *defaults = get_defaults(rd->rp);
if (rd->state & DOMAIN_STATE_BIOS_LOCKED)
return -EACCES;
cpus_read_lock();
rapl_write_data_raw(rd, PL1_ENABLE, mode);
- if (rapl_defaults->set_floor_freq)
- rapl_defaults->set_floor_freq(rd, mode);
+ if (defaults->set_floor_freq)
+ defaults->set_floor_freq(rd, mode);
cpus_read_unlock();
return 0;
@@ -551,6 +557,7 @@ static void rapl_init_domains(struct rapl_package *rp)
enum rapl_domain_type i;
enum rapl_domain_reg_id j;
struct rapl_domain *rd = rp->domains;
+ struct rapl_defaults *defaults = get_defaults(rp);
for (i = 0; i < RAPL_DOMAIN_MAX; i++) {
unsigned int mask = rp->domain_map & (1 << i);
@@ -592,14 +599,14 @@ static void rapl_init_domains(struct rapl_package *rp)
switch (i) {
case RAPL_DOMAIN_DRAM:
rd->domain_energy_unit =
- rapl_defaults->dram_domain_energy_unit;
+ defaults->dram_domain_energy_unit;
if (rd->domain_energy_unit)
pr_info("DRAM domain energy unit %dpj\n",
rd->domain_energy_unit);
break;
case RAPL_DOMAIN_PLATFORM:
rd->domain_energy_unit =
- rapl_defaults->psys_domain_energy_unit;
+ defaults->psys_domain_energy_unit;
if (rd->domain_energy_unit)
pr_info("Platform domain energy unit %dpj\n",
rd->domain_energy_unit);
@@ -616,6 +623,7 @@ static u64 rapl_unit_xlate(struct rapl_domain *rd, enum unit_type type,
{
u64 units = 1;
struct rapl_package *rp = rd->rp;
+ struct rapl_defaults *defaults = get_defaults(rp);
u64 scale = 1;
switch (type) {
@@ -631,7 +639,7 @@ static u64 rapl_unit_xlate(struct rapl_domain *rd, enum unit_type type,
units = rp->energy_unit;
break;
case TIME_UNIT:
- return rapl_defaults->compute_time_window(rp, value, to_raw);
+ return defaults->compute_time_window(rp, value, to_raw);
case ARBITRARY_UNIT:
default:
return value;
@@ -702,10 +710,18 @@ static struct rapl_primitive_info rpi[] = {
{NULL, 0, 0, 0},
};
+static int rapl_config(struct rapl_package *rp)
+{
+ rp->priv->defaults = (void *)rapl_defaults;
+ return 0;
+}
+
static enum rapl_primitives
prim_fixups(struct rapl_domain *rd, enum rapl_primitives prim)
{
- if (!rapl_defaults->spr_psys_bits)
+ struct rapl_defaults *defaults = get_defaults(rd->rp);
+
+ if (!defaults->spr_psys_bits)
return prim;
if (rd->id != RAPL_DOMAIN_PLATFORM)
@@ -960,16 +976,17 @@ static void set_floor_freq_default(struct rapl_domain *rd, bool mode)
static void set_floor_freq_atom(struct rapl_domain *rd, bool enable)
{
static u32 power_ctrl_orig_val;
+ struct rapl_defaults *defaults = get_defaults(rd->rp);
u32 mdata;
- if (!rapl_defaults->floor_freq_reg_addr) {
+ if (!defaults->floor_freq_reg_addr) {
pr_err("Invalid floor frequency config register\n");
return;
}
if (!power_ctrl_orig_val)
iosf_mbi_read(BT_MBI_UNIT_PMC, MBI_CR_READ,
- rapl_defaults->floor_freq_reg_addr,
+ defaults->floor_freq_reg_addr,
&power_ctrl_orig_val);
mdata = power_ctrl_orig_val;
if (enable) {
@@ -977,7 +994,7 @@ static void set_floor_freq_atom(struct rapl_domain *rd, bool enable)
mdata |= 1 << 8;
}
iosf_mbi_write(BT_MBI_UNIT_PMC, MBI_CR_WRITE,
- rapl_defaults->floor_freq_reg_addr, mdata);
+ defaults->floor_freq_reg_addr, mdata);
}
static u64 rapl_compute_time_window_core(struct rapl_package *rp, u64 value,
@@ -1385,11 +1402,9 @@ struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv)
{
int id = topology_logical_die_id(cpu);
struct rapl_package *rp;
+ struct rapl_defaults *defaults;
int ret;
- if (!rapl_defaults)
- return ERR_PTR(-ENODEV);
-
rp = kzalloc(sizeof(struct rapl_package), GFP_KERNEL);
if (!rp)
return ERR_PTR(-ENOMEM);
@@ -1399,6 +1414,10 @@ struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv)
rp->lead_cpu = cpu;
rp->priv = priv;
+ ret = rapl_config(rp);
+ if (ret)
+ goto err_free_package;
+
if (topology_max_die_per_package() > 1)
snprintf(rp->name, PACKAGE_DOMAIN_NAME_LENGTH,
"package-%d-die-%d",
@@ -1407,8 +1426,9 @@ struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv)
snprintf(rp->name, PACKAGE_DOMAIN_NAME_LENGTH, "package-%d",
topology_physical_package_id(cpu));
+ defaults = get_defaults(rp);
/* check if the package contains valid domains */
- if (rapl_detect_domains(rp, cpu) || rapl_defaults->check_unit(rp, cpu)) {
+ if (rapl_detect_domains(rp, cpu) || defaults->check_unit(rp, cpu)) {
ret = -ENODEV;
goto err_free_package;
}
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index 828557645770..ebd1cad78212 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -121,6 +121,7 @@ struct reg_action {
* registers.
* @write_raw: Callback for writing RAPL interface specific
* registers.
+ * @defaults: internal pointer to interface default settings
*/
struct rapl_if_priv {
struct powercap_control_type *control_type;
@@ -130,6 +131,7 @@ struct rapl_if_priv {
int limits[RAPL_DOMAIN_MAX];
int (*read_raw)(int cpu, struct reg_action *ra);
int (*write_raw)(int cpu, struct reg_action *ra);
+ void *defaults;
};
/* maximum rapl package domain name: package-%d-die-%d */
--
2.25.1
The same set of operations are shared by different Powert Limits,
including Power Limit get/set, Power Limit enable/disable, clamping
enable/disable, time window get/set, and max power get/set, etc.
But the same operation for different Power Limit has different
primitives because they use different registers/register bits.
A lot of dirty/duplicate code was introduced to handle this difference.
Introduce a universal way to issue Power Limit operations.
Instead of using hardcoded primitive name directly, use Power Limit id
+ operation type, and hide all the Power Limit difference details in a
central place, get_pl_prim(). Two helpers, rapl_read_pl_data() and
rapl_write_pl_data(), are introduced at the same time to simplify the
code for issuing Power Limit operations.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 343 ++++++++++++---------------
include/linux/intel_rapl.h | 1 -
2 files changed, 146 insertions(+), 198 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 8e77df42257a..7f80c35e5c86 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -96,9 +96,67 @@ enum unit_type {
#define DOMAIN_STATE_POWER_LIMIT_SET BIT(1)
#define DOMAIN_STATE_BIOS_LOCKED BIT(2)
-static const char pl1_name[] = "long_term";
-static const char pl2_name[] = "short_term";
-static const char pl4_name[] = "peak_power";
+static const char *pl_names[NR_POWER_LIMITS] = {
+ [POWER_LIMIT1] = "long_term",
+ [POWER_LIMIT2] = "short_term",
+ [POWER_LIMIT4] = "peak_power",
+};
+
+enum pl_prims {
+ PL_ENABLE,
+ PL_CLAMP,
+ PL_LIMIT,
+ PL_TIME_WINDOW,
+ PL_MAX_POWER,
+};
+
+static bool is_pl_valid(struct rapl_domain *rd, int pl)
+{
+ if (pl < POWER_LIMIT1 || pl > POWER_LIMIT4)
+ return false;
+ return rd->rpl[pl].name ? true : false;
+}
+
+static int get_pl_prim(int pl, enum pl_prims prim)
+{
+ switch (pl) {
+ case POWER_LIMIT1:
+ if (prim == PL_ENABLE)
+ return PL1_ENABLE;
+ if (prim == PL_CLAMP)
+ return PL1_CLAMP;
+ if (prim == PL_LIMIT)
+ return POWER_LIMIT1;
+ if (prim == PL_TIME_WINDOW)
+ return TIME_WINDOW1;
+ if (prim == PL_MAX_POWER)
+ return THERMAL_SPEC_POWER;
+ return -EINVAL;
+ case POWER_LIMIT2:
+ if (prim == PL_ENABLE)
+ return PL2_ENABLE;
+ if (prim == PL_CLAMP)
+ return PL2_CLAMP;
+ if (prim == PL_LIMIT)
+ return POWER_LIMIT2;
+ if (prim == PL_TIME_WINDOW)
+ return TIME_WINDOW2;
+ if (prim == PL_MAX_POWER)
+ return MAX_POWER;
+ return -EINVAL;
+ case POWER_LIMIT4:
+ if (prim == PL_LIMIT)
+ return POWER_LIMIT4;
+ if (prim == PL_ENABLE)
+ return PL4_ENABLE;
+ /* PL4 would be around two times PL2, use same prim as PL2. */
+ if (prim == PL_MAX_POWER)
+ return MAX_POWER;
+ return -EINVAL;
+ default:
+ return -EINVAL;
+ }
+}
#define power_zone_to_rapl_domain(_zone) \
container_of(_zone, struct rapl_domain, power_zone)
@@ -155,6 +213,12 @@ static int rapl_read_data_raw(struct rapl_domain *rd,
static int rapl_write_data_raw(struct rapl_domain *rd,
enum rapl_primitives prim,
unsigned long long value);
+static int rapl_read_pl_data(struct rapl_domain *rd, int pl,
+ enum pl_prims pl_prim,
+ bool xlate, u64 *data);
+static int rapl_write_pl_data(struct rapl_domain *rd, int pl,
+ enum pl_prims pl_prim,
+ unsigned long long value);
static u64 rapl_unit_xlate(struct rapl_domain *rd,
enum unit_type type, u64 value, int to_raw);
static void package_power_limit_irq_save(struct rapl_package *rp);
@@ -222,7 +286,7 @@ static int find_nr_power_limit(struct rapl_domain *rd)
int i, nr_pl = 0;
for (i = 0; i < NR_POWER_LIMITS; i++) {
- if (rd->rpl[i].name)
+ if (is_pl_valid(rd, i))
nr_pl++;
}
@@ -233,37 +297,34 @@ static int set_domain_enable(struct powercap_zone *power_zone, bool mode)
{
struct rapl_domain *rd = power_zone_to_rapl_domain(power_zone);
struct rapl_defaults *defaults = get_defaults(rd->rp);
-
- if (rd->state & DOMAIN_STATE_BIOS_LOCKED)
- return -EACCES;
+ int ret;
cpus_read_lock();
- rapl_write_data_raw(rd, PL1_ENABLE, mode);
- if (defaults->set_floor_freq)
+ ret = rapl_write_pl_data(rd, POWER_LIMIT1, PL_ENABLE, mode);
+ if (!ret && defaults->set_floor_freq)
defaults->set_floor_freq(rd, mode);
cpus_read_unlock();
- return 0;
+ return ret;
}
static int get_domain_enable(struct powercap_zone *power_zone, bool *mode)
{
struct rapl_domain *rd = power_zone_to_rapl_domain(power_zone);
u64 val;
+ int ret;
if (rd->state & DOMAIN_STATE_BIOS_LOCKED) {
*mode = false;
return 0;
}
cpus_read_lock();
- if (rapl_read_data_raw(rd, PL1_ENABLE, true, &val)) {
- cpus_read_unlock();
- return -EIO;
- }
- *mode = val;
+ ret = rapl_read_pl_data(rd, POWER_LIMIT1, PL_ENABLE, true, &val);
+ if (!ret)
+ *mode = val;
cpus_read_unlock();
- return 0;
+ return ret;
}
/* per RAPL domain ops, in the order of rapl_domain_type */
@@ -319,8 +380,8 @@ static int contraint_to_pl(struct rapl_domain *rd, int cid)
{
int i, j;
- for (i = 0, j = 0; i < NR_POWER_LIMITS; i++) {
- if ((rd->rpl[i].name) && j++ == cid) {
+ for (i = POWER_LIMIT1, j = 0; i < NR_POWER_LIMITS; i++) {
+ if (is_pl_valid(rd, i) && j++ == cid) {
pr_debug("%s: index %d\n", __func__, i);
return i;
}
@@ -341,36 +402,11 @@ static int set_power_limit(struct powercap_zone *power_zone, int cid,
cpus_read_lock();
rd = power_zone_to_rapl_domain(power_zone);
id = contraint_to_pl(rd, cid);
- if (id < 0) {
- ret = id;
- goto set_exit;
- }
-
rp = rd->rp;
- if (rd->state & DOMAIN_STATE_BIOS_LOCKED) {
- dev_warn(&power_zone->dev,
- "%s locked by BIOS, monitoring only\n", rd->name);
- ret = -EACCES;
- goto set_exit;
- }
-
- switch (rd->rpl[id].prim_id) {
- case PL1_ENABLE:
- rapl_write_data_raw(rd, POWER_LIMIT1, power_limit);
- break;
- case PL2_ENABLE:
- rapl_write_data_raw(rd, POWER_LIMIT2, power_limit);
- break;
- case PL4_ENABLE:
- rapl_write_data_raw(rd, POWER_LIMIT4, power_limit);
- break;
- default:
- ret = -EINVAL;
- }
+ ret = rapl_write_pl_data(rd, id, PL_LIMIT, power_limit);
if (!ret)
package_power_limit_irq_save(rp);
-set_exit:
cpus_read_unlock();
return ret;
}
@@ -380,38 +416,17 @@ static int get_current_power_limit(struct powercap_zone *power_zone, int cid,
{
struct rapl_domain *rd;
u64 val;
- int prim;
int ret = 0;
int id;
cpus_read_lock();
rd = power_zone_to_rapl_domain(power_zone);
id = contraint_to_pl(rd, cid);
- if (id < 0) {
- ret = id;
- goto get_exit;
- }
- switch (rd->rpl[id].prim_id) {
- case PL1_ENABLE:
- prim = POWER_LIMIT1;
- break;
- case PL2_ENABLE:
- prim = POWER_LIMIT2;
- break;
- case PL4_ENABLE:
- prim = POWER_LIMIT4;
- break;
- default:
- cpus_read_unlock();
- return -EINVAL;
- }
- if (rapl_read_data_raw(rd, prim, true, &val))
- ret = -EIO;
- else
+ ret = rapl_read_pl_data(rd, id, PL_LIMIT, true, &val);
+ if (!ret)
*data = val;
-get_exit:
cpus_read_unlock();
return ret;
@@ -427,23 +442,9 @@ static int set_time_window(struct powercap_zone *power_zone, int cid,
cpus_read_lock();
rd = power_zone_to_rapl_domain(power_zone);
id = contraint_to_pl(rd, cid);
- if (id < 0) {
- ret = id;
- goto set_time_exit;
- }
- switch (rd->rpl[id].prim_id) {
- case PL1_ENABLE:
- rapl_write_data_raw(rd, TIME_WINDOW1, window);
- break;
- case PL2_ENABLE:
- rapl_write_data_raw(rd, TIME_WINDOW2, window);
- break;
- default:
- ret = -EINVAL;
- }
+ ret = rapl_write_pl_data(rd, id, PL_TIME_WINDOW, window);
-set_time_exit:
cpus_read_unlock();
return ret;
}
@@ -459,33 +460,11 @@ static int get_time_window(struct powercap_zone *power_zone, int cid,
cpus_read_lock();
rd = power_zone_to_rapl_domain(power_zone);
id = contraint_to_pl(rd, cid);
- if (id < 0) {
- ret = id;
- goto get_time_exit;
- }
- switch (rd->rpl[id].prim_id) {
- case PL1_ENABLE:
- ret = rapl_read_data_raw(rd, TIME_WINDOW1, true, &val);
- break;
- case PL2_ENABLE:
- ret = rapl_read_data_raw(rd, TIME_WINDOW2, true, &val);
- break;
- case PL4_ENABLE:
- /*
- * Time window parameter is not applicable for PL4 entry
- * so assigining '0' as default value.
- */
- val = 0;
- break;
- default:
- cpus_read_unlock();
- return -EINVAL;
- }
+ ret = rapl_read_pl_data(rd, id, PL_TIME_WINDOW, true, &val);
if (!ret)
*data = val;
-get_time_exit:
cpus_read_unlock();
return ret;
@@ -505,36 +484,23 @@ static const char *get_constraint_name(struct powercap_zone *power_zone,
return NULL;
}
-static int get_max_power(struct powercap_zone *power_zone, int id, u64 *data)
+static int get_max_power(struct powercap_zone *power_zone, int cid, u64 *data)
{
struct rapl_domain *rd;
u64 val;
- int prim;
int ret = 0;
+ int id;
cpus_read_lock();
rd = power_zone_to_rapl_domain(power_zone);
- switch (rd->rpl[id].prim_id) {
- case PL1_ENABLE:
- prim = THERMAL_SPEC_POWER;
- break;
- case PL2_ENABLE:
- prim = MAX_POWER;
- break;
- case PL4_ENABLE:
- prim = MAX_POWER;
- break;
- default:
- cpus_read_unlock();
- return -EINVAL;
- }
- if (rapl_read_data_raw(rd, prim, true, &val))
- ret = -EIO;
- else
+ id = contraint_to_pl(rd, cid);
+
+ ret = rapl_read_pl_data(rd, id, PL_MAX_POWER, true, &val);
+ if (!ret)
*data = val;
/* As a generalization rule, PL4 would be around two times PL2. */
- if (rd->rpl[id].prim_id == PL4_ENABLE)
+ if (id == POWER_LIMIT4)
*data = *data * 2;
cpus_read_unlock();
@@ -560,6 +526,7 @@ static void rapl_init_domains(struct rapl_package *rp)
for (i = 0; i < RAPL_DOMAIN_MAX; i++) {
unsigned int mask = rp->domain_map & (1 << i);
+ int t;
if (!mask)
continue;
@@ -577,17 +544,10 @@ static void rapl_init_domains(struct rapl_package *rp)
/* PL1 is supported by default */
rp->priv->limits[i] |= BIT(POWER_LIMIT1);
- rd->rpl[0].prim_id = PL1_ENABLE;
- rd->rpl[0].name = pl1_name;
- if (rp->priv->limits[i] & BIT(POWER_LIMIT2)) {
- rd->rpl[1].prim_id = PL2_ENABLE;
- rd->rpl[1].name = pl2_name;
- }
-
- if (rp->priv->limits[i] & BIT(POWER_LIMIT4)) {
- rd->rpl[2].prim_id = PL4_ENABLE;
- rd->rpl[2].name = pl4_name;
+ for (t = POWER_LIMIT1; t < NR_POWER_LIMITS; t++) {
+ if (rp->priv->limits[i] & BIT(t))
+ rd->rpl[t].name = pl_names[t];
}
for (j = 0; j < RAPL_DOMAIN_REG_MAX; j++)
@@ -818,6 +778,33 @@ static int rapl_write_data_raw(struct rapl_domain *rd,
return ret;
}
+static int rapl_read_pl_data(struct rapl_domain *rd, int pl,
+ enum pl_prims pl_prim, bool xlate, u64 *data)
+{
+ enum rapl_primitives prim = get_pl_prim(pl, pl_prim);
+
+ if (!is_pl_valid(rd, pl))
+ return -EINVAL;
+
+ return rapl_read_data_raw(rd, prim, xlate, data);
+}
+
+static int rapl_write_pl_data(struct rapl_domain *rd, int pl,
+ enum pl_prims pl_prim,
+ unsigned long long value)
+{
+ enum rapl_primitives prim = get_pl_prim(pl, pl_prim);
+
+ if (!is_pl_valid(rd, pl))
+ return -EINVAL;
+
+ if (rd->state & DOMAIN_STATE_BIOS_LOCKED) {
+ pr_warn("%s:%s:%s locked by BIOS\n", rd->rp->name, rd->name, pl_names[pl]);
+ return -EACCES;
+ }
+
+ return rapl_write_data_raw(rd, prim, value);
+}
/*
* Raw RAPL data stored in MSRs are in certain scales. We need to
* convert them into standard units based on the units reported in
@@ -945,17 +932,16 @@ static void package_power_limit_irq_restore(struct rapl_package *rp)
static void set_floor_freq_default(struct rapl_domain *rd, bool mode)
{
- int nr_powerlimit = find_nr_power_limit(rd);
+ int i;
/* always enable clamp such that p-state can go below OS requested
* range. power capping priority over guranteed frequency.
*/
- rapl_write_data_raw(rd, PL1_CLAMP, mode);
+ rapl_write_pl_data(rd, POWER_LIMIT1, PL_CLAMP, mode);
- /* some domains have pl2 */
- if (nr_powerlimit > 1) {
- rapl_write_data_raw(rd, PL2_ENABLE, mode);
- rapl_write_data_raw(rd, PL2_CLAMP, mode);
+ for (i = POWER_LIMIT2; i < NR_POWER_LIMITS; i++) {
+ rapl_write_pl_data(rd, i, PL_ENABLE, mode);
+ rapl_write_pl_data(rd, i, PL_CLAMP, mode);
}
}
@@ -1327,11 +1313,10 @@ static void rapl_detect_powerlimit(struct rapl_domain *rd)
rd->state |= DOMAIN_STATE_BIOS_LOCKED;
}
}
- /* check if power limit MSR exists, otherwise domain is monitoring only */
- for (i = 0; i < NR_POWER_LIMITS; i++) {
- int prim = rd->rpl[i].prim_id;
- if (rapl_read_data_raw(rd, prim, false, &val64))
+ /* check if power limit exists, otherwise domain is monitoring only */
+ for (i = POWER_LIMIT1; i < NR_POWER_LIMITS; i++) {
+ if (rapl_read_pl_data(rd, i, PL_ENABLE, false, &val64))
rd->rpl[i].name = NULL;
}
}
@@ -1381,13 +1366,13 @@ void rapl_remove_package(struct rapl_package *rp)
package_power_limit_irq_restore(rp);
for (rd = rp->domains; rd < rp->domains + rp->nr_domains; rd++) {
- rapl_write_data_raw(rd, PL1_ENABLE, 0);
- rapl_write_data_raw(rd, PL1_CLAMP, 0);
- if (find_nr_power_limit(rd) > 1) {
- rapl_write_data_raw(rd, PL2_ENABLE, 0);
- rapl_write_data_raw(rd, PL2_CLAMP, 0);
- rapl_write_data_raw(rd, PL4_ENABLE, 0);
+ int i;
+
+ for (i = POWER_LIMIT1; i < NR_POWER_LIMITS; i++) {
+ rapl_write_pl_data(rd, i, PL_ENABLE, 0);
+ rapl_write_pl_data(rd, i, PL_CLAMP, 0);
}
+
if (rd->id == RAPL_DOMAIN_PACKAGE) {
rd_package = rd;
continue;
@@ -1472,38 +1457,18 @@ static void power_limit_state_save(void)
{
struct rapl_package *rp;
struct rapl_domain *rd;
- int nr_pl, ret, i;
+ int ret, i;
cpus_read_lock();
list_for_each_entry(rp, &rapl_packages, plist) {
if (!rp->power_zone)
continue;
rd = power_zone_to_rapl_domain(rp->power_zone);
- nr_pl = find_nr_power_limit(rd);
- for (i = 0; i < nr_pl; i++) {
- switch (rd->rpl[i].prim_id) {
- case PL1_ENABLE:
- ret = rapl_read_data_raw(rd,
- POWER_LIMIT1, true,
+ for (i = POWER_LIMIT1; i < NR_POWER_LIMITS; i++) {
+ ret = rapl_read_pl_data(rd, i, PL_LIMIT, true,
&rd->rpl[i].last_power_limit);
- if (ret)
- rd->rpl[i].last_power_limit = 0;
- break;
- case PL2_ENABLE:
- ret = rapl_read_data_raw(rd,
- POWER_LIMIT2, true,
- &rd->rpl[i].last_power_limit);
- if (ret)
- rd->rpl[i].last_power_limit = 0;
- break;
- case PL4_ENABLE:
- ret = rapl_read_data_raw(rd,
- POWER_LIMIT4, true,
- &rd->rpl[i].last_power_limit);
- if (ret)
- rd->rpl[i].last_power_limit = 0;
- break;
- }
+ if (ret)
+ rd->rpl[i].last_power_limit = 0;
}
}
cpus_read_unlock();
@@ -1513,33 +1478,17 @@ static void power_limit_state_restore(void)
{
struct rapl_package *rp;
struct rapl_domain *rd;
- int nr_pl, i;
+ int i;
cpus_read_lock();
list_for_each_entry(rp, &rapl_packages, plist) {
if (!rp->power_zone)
continue;
rd = power_zone_to_rapl_domain(rp->power_zone);
- nr_pl = find_nr_power_limit(rd);
- for (i = 0; i < nr_pl; i++) {
- switch (rd->rpl[i].prim_id) {
- case PL1_ENABLE:
- if (rd->rpl[i].last_power_limit)
- rapl_write_data_raw(rd, POWER_LIMIT1,
- rd->rpl[i].last_power_limit);
- break;
- case PL2_ENABLE:
- if (rd->rpl[i].last_power_limit)
- rapl_write_data_raw(rd, POWER_LIMIT2,
- rd->rpl[i].last_power_limit);
- break;
- case PL4_ENABLE:
- if (rd->rpl[i].last_power_limit)
- rapl_write_data_raw(rd, POWER_LIMIT4,
- rd->rpl[i].last_power_limit);
- break;
- }
- }
+ for (i = POWER_LIMIT1; i < NR_POWER_LIMITS; i++)
+ if (rd->rpl[i].last_power_limit)
+ rapl_write_pl_data(rd, i, PL_LIMIT,
+ rd->rpl[i].last_power_limit);
}
cpus_read_unlock();
}
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index bbd03b17dc8d..df17f4e51dbf 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -79,7 +79,6 @@ struct rapl_domain_data {
struct rapl_power_limit {
struct powercap_zone_constraint *constraint;
- int prim_id; /* primitive ID used to enable */
struct rapl_domain *domain;
const char *name;
u64 last_power_limit;
--
2.25.1
With RAPL MSR/MMIO Interface, each RAPL domain has one Power Limit
register. Each Power Limit register has one lock bit which tells the OS
if the power limit register can be used or not.
Depending on the number of power limits supported by the power limit
register, the lock bit may apply to one or more power limits.
With RAPL TPMI Interface, each RAPL domain has multiple Power Limits,
and each Power Limit has its own register, with a lock bit.
To handle this, introduce support for lock bit per Power Limit.
For existing RAPL MSR/MMIO Interfaces, the lock bit in the Power Limit
register applies to all the Power Limits controlled by this register.
Remove the per domain DOMAIN_STATE_BIOS_LOCKED flag at the same time
because it can be replaced by the per Power Limit lock.
No functional change intended.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 50 +++++++++++++++++-----------
include/linux/intel_rapl.h | 2 ++
2 files changed, 32 insertions(+), 20 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index 7f80c35e5c86..d88008308d7a 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -94,7 +94,6 @@ enum unit_type {
#define DOMAIN_STATE_INACTIVE BIT(0)
#define DOMAIN_STATE_POWER_LIMIT_SET BIT(1)
-#define DOMAIN_STATE_BIOS_LOCKED BIT(2)
static const char *pl_names[NR_POWER_LIMITS] = {
[POWER_LIMIT1] = "long_term",
@@ -108,6 +107,7 @@ enum pl_prims {
PL_LIMIT,
PL_TIME_WINDOW,
PL_MAX_POWER,
+ PL_LOCK,
};
static bool is_pl_valid(struct rapl_domain *rd, int pl)
@@ -117,7 +117,18 @@ static bool is_pl_valid(struct rapl_domain *rd, int pl)
return rd->rpl[pl].name ? true : false;
}
-static int get_pl_prim(int pl, enum pl_prims prim)
+static int get_pl_lock_prim(struct rapl_domain *rd, int pl)
+{
+ /*
+ * Power Limit register that supports two power limits has a different
+ * bit position for the Lock bit.
+ */
+ if (rd->rp->priv->limits[rd->id] & BIT(POWER_LIMIT2))
+ return FW_HIGH_LOCK;
+ return FW_LOCK;
+}
+
+static int get_pl_prim(struct rapl_domain *rd, int pl, enum pl_prims prim)
{
switch (pl) {
case POWER_LIMIT1:
@@ -131,6 +142,8 @@ static int get_pl_prim(int pl, enum pl_prims prim)
return TIME_WINDOW1;
if (prim == PL_MAX_POWER)
return THERMAL_SPEC_POWER;
+ if (prim == PL_LOCK)
+ return get_pl_lock_prim(rd, pl);
return -EINVAL;
case POWER_LIMIT2:
if (prim == PL_ENABLE)
@@ -143,6 +156,8 @@ static int get_pl_prim(int pl, enum pl_prims prim)
return TIME_WINDOW2;
if (prim == PL_MAX_POWER)
return MAX_POWER;
+ if (prim == PL_LOCK)
+ return get_pl_lock_prim(rd, pl);
return -EINVAL;
case POWER_LIMIT4:
if (prim == PL_LIMIT)
@@ -314,7 +329,7 @@ static int get_domain_enable(struct powercap_zone *power_zone, bool *mode)
u64 val;
int ret;
- if (rd->state & DOMAIN_STATE_BIOS_LOCKED) {
+ if (rd->rpl[POWER_LIMIT1].locked) {
*mode = false;
return 0;
}
@@ -599,6 +614,8 @@ static struct rapl_primitive_info rpi_default[NR_RAPL_PRIMITIVES] = {
RAPL_DOMAIN_REG_STATUS, ENERGY_UNIT, 0),
[FW_LOCK] = PRIMITIVE_INFO_INIT(FW_LOCK, POWER_LOW_LOCK, 31,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
+ [FW_HIGH_LOCK] = PRIMITIVE_INFO_INIT(FW_LOCK, POWER_HIGH_LOCK, 63,
+ RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
[PL1_ENABLE] = PRIMITIVE_INFO_INIT(PL1_ENABLE, POWER_LIMIT1_ENABLE, 15,
RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
[PL1_CLAMP] = PRIMITIVE_INFO_INIT(PL1_CLAMP, POWER_LIMIT1_CLAMP, 16,
@@ -719,11 +736,6 @@ static int rapl_read_data_raw(struct rapl_domain *rd,
cpu = rd->rp->lead_cpu;
- /* domain with 2 limits has different bit */
- if (prim == FW_LOCK && (rd->rp->priv->limits[rd->id] & BIT(POWER_LIMIT2))) {
- rpi->mask = POWER_HIGH_LOCK;
- rpi->shift = 63;
- }
/* non-hardware data are collected by the polling thread */
if (rpi->flag & RAPL_PRIMITIVE_DERIVED) {
*data = rd->rdd.primitives[prim];
@@ -781,7 +793,7 @@ static int rapl_write_data_raw(struct rapl_domain *rd,
static int rapl_read_pl_data(struct rapl_domain *rd, int pl,
enum pl_prims pl_prim, bool xlate, u64 *data)
{
- enum rapl_primitives prim = get_pl_prim(pl, pl_prim);
+ enum rapl_primitives prim = get_pl_prim(rd, pl, pl_prim);
if (!is_pl_valid(rd, pl))
return -EINVAL;
@@ -793,12 +805,12 @@ static int rapl_write_pl_data(struct rapl_domain *rd, int pl,
enum pl_prims pl_prim,
unsigned long long value)
{
- enum rapl_primitives prim = get_pl_prim(pl, pl_prim);
+ enum rapl_primitives prim = get_pl_prim(rd, pl, pl_prim);
if (!is_pl_valid(rd, pl))
return -EINVAL;
- if (rd->state & DOMAIN_STATE_BIOS_LOCKED) {
+ if (rd->rpl[pl].locked) {
pr_warn("%s:%s:%s locked by BIOS\n", rd->rp->name, rd->name, pl_names[pl]);
return -EACCES;
}
@@ -1305,17 +1317,15 @@ static void rapl_detect_powerlimit(struct rapl_domain *rd)
u64 val64;
int i;
- /* check if the domain is locked by BIOS, ignore if MSR doesn't exist */
- if (!rapl_read_data_raw(rd, FW_LOCK, false, &val64)) {
- if (val64) {
- pr_info("RAPL %s domain %s locked by BIOS\n",
- rd->rp->name, rd->name);
- rd->state |= DOMAIN_STATE_BIOS_LOCKED;
+ for (i = POWER_LIMIT1; i < NR_POWER_LIMITS; i++) {
+ if (!rapl_read_pl_data(rd, i, PL_LOCK, false, &val64)) {
+ if (val64) {
+ rd->rpl[i].locked = true;
+ pr_info("%s:%s:%s locked by BIOS\n",
+ rd->rp->name, rd->name, pl_names[i]);
+ }
}
- }
- /* check if power limit exists, otherwise domain is monitoring only */
- for (i = POWER_LIMIT1; i < NR_POWER_LIMITS; i++) {
if (rapl_read_pl_data(rd, i, PL_ENABLE, false, &val64))
rd->rpl[i].name = NULL;
}
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index df17f4e51dbf..d07b460bac3b 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -42,6 +42,7 @@ enum rapl_primitives {
POWER_LIMIT4,
ENERGY_COUNTER,
FW_LOCK,
+ FW_HIGH_LOCK,
PL1_ENABLE, /* power limit 1, aka long term */
PL1_CLAMP, /* allow frequency to go below OS request */
@@ -81,6 +82,7 @@ struct rapl_power_limit {
struct powercap_zone_constraint *constraint;
struct rapl_domain *domain;
const char *name;
+ bool locked;
u64 last_power_limit;
};
--
2.25.1
For rapl_packages that rely on online CPUs to work, rp->lead_cpu always
has a valid CPU id.
Remove the redundant cpu parameter in rapl_check_domain(),
rapl_detect_domains() and .check_unit() callbacks.
No functional change.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 26 +++++++++++++-------------
1 file changed, 13 insertions(+), 13 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index d88008308d7a..fcb92f1f4bfc 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -178,7 +178,7 @@ static int get_pl_prim(struct rapl_domain *rd, int pl, enum pl_prims prim)
struct rapl_defaults {
u8 floor_freq_reg_addr;
- int (*check_unit)(struct rapl_domain *rd, int cpu);
+ int (*check_unit)(struct rapl_domain *rd);
void (*set_floor_freq)(struct rapl_domain *rd, bool mode);
u64 (*compute_time_window)(struct rapl_domain *rd, u64 val,
bool to_raw);
@@ -828,16 +828,16 @@ static int rapl_write_pl_data(struct rapl_domain *rd, int pl,
* power unit : microWatts : Represented in milliWatts by default
* time unit : microseconds: Represented in seconds by default
*/
-static int rapl_check_unit_core(struct rapl_domain *rd, int cpu)
+static int rapl_check_unit_core(struct rapl_domain *rd)
{
struct reg_action ra;
u32 value;
ra.reg = rd->regs[RAPL_DOMAIN_REG_UNIT];
ra.mask = ~0;
- if (rd->rp->priv->read_raw(cpu, &ra)) {
+ if (rd->rp->priv->read_raw(rd->rp->lead_cpu, &ra)) {
pr_err("Failed to read power unit REG 0x%llx on CPU %d, exit.\n",
- ra.reg, cpu);
+ ra.reg, rd->rp->lead_cpu);
return -ENODEV;
}
@@ -856,16 +856,16 @@ static int rapl_check_unit_core(struct rapl_domain *rd, int cpu)
return 0;
}
-static int rapl_check_unit_atom(struct rapl_domain *rd, int cpu)
+static int rapl_check_unit_atom(struct rapl_domain *rd)
{
struct reg_action ra;
u32 value;
ra.reg = rd->regs[RAPL_DOMAIN_REG_UNIT];
ra.mask = ~0;
- if (rd->rp->priv->read_raw(cpu, &ra)) {
+ if (rd->rp->priv->read_raw(rd->rp->lead_cpu, &ra)) {
pr_err("Failed to read power unit REG 0x%llx on CPU %d, exit.\n",
- ra.reg, cpu);
+ ra.reg, rd->rp->lead_cpu);
return -ENODEV;
}
@@ -1242,7 +1242,7 @@ static int rapl_package_register_powercap(struct rapl_package *rp)
return ret;
}
-static int rapl_check_domain(int cpu, int domain, struct rapl_package *rp)
+static int rapl_check_domain(int domain, struct rapl_package *rp)
{
struct reg_action ra;
@@ -1263,7 +1263,7 @@ static int rapl_check_domain(int cpu, int domain, struct rapl_package *rp)
*/
ra.mask = ENERGY_STATUS_MASK;
- if (rp->priv->read_raw(cpu, &ra) || !ra.value)
+ if (rp->priv->read_raw(rp->lead_cpu, &ra) || !ra.value)
return -ENODEV;
return 0;
@@ -1292,7 +1292,7 @@ static int rapl_get_domain_unit(struct rapl_domain *rd)
return -ENODEV;
}
- ret = defaults->check_unit(rd, rd->rp->lead_cpu);
+ ret = defaults->check_unit(rd);
if (ret)
return ret;
@@ -1334,14 +1334,14 @@ static void rapl_detect_powerlimit(struct rapl_domain *rd)
/* Detect active and valid domains for the given CPU, caller must
* ensure the CPU belongs to the targeted package and CPU hotlug is disabled.
*/
-static int rapl_detect_domains(struct rapl_package *rp, int cpu)
+static int rapl_detect_domains(struct rapl_package *rp)
{
struct rapl_domain *rd;
int i;
for (i = 0; i < RAPL_DOMAIN_MAX; i++) {
/* use physical package id to read counters */
- if (!rapl_check_domain(cpu, i, rp)) {
+ if (!rapl_check_domain(i, rp)) {
rp->domain_map |= 1 << i;
pr_info("Found RAPL domain %s\n", rapl_domain_names[i]);
}
@@ -1445,7 +1445,7 @@ struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv)
topology_physical_package_id(cpu));
/* check if the package contains valid domains */
- if (rapl_detect_domains(rp, cpu)) {
+ if (rapl_detect_domains(rp)) {
ret = -ENODEV;
goto err_free_package;
}
--
2.25.1
MSR RAPL Interface always removes a rapl_package when all the CPUs in
that rapl_package are offlined. This is because it relies on an online
CPU to access the MSR.
But for RAPL Interface using MMIO registers, when all the cpus within
the rapl_package are offlined,
1. the register can still be accessed
2. monitoring and setting the Power Pimits for the rapl_package is still
meaningful because of uncore power.
This means that, a valid rapl_package doesn't rely on one or more cpus
being onlined.
For this sense, make cpu optional for rapl_package. A rapl_package can
be registered either using a CPU id to represent the physical
package/die, or using the physical package id directly.
Note that, the thermal throttling interrupt is not disabled via
MSR_IA32_PACKAGE_THERM_INTERRUPT for such rapl_package at the moment.
If it is still needed in the future, this can be achieved by selecting
an onlined CPU using the physical package id.
Note that, processor_thermal_rapl, the current MMIO RAPL Interface
driver, can also be converted to register using a package id instead.
But this is not done right now because processor_thermal_rapl driver
works on single-package systems only, and offlining the only package
will not happen. So keep the previous logic.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 84 +++++++++++--------
drivers/powercap/intel_rapl_msr.c | 6 +-
.../int340x_thermal/processor_thermal_rapl.c | 6 +-
include/linux/intel_rapl.h | 8 +-
4 files changed, 60 insertions(+), 44 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index fcb92f1f4bfc..d337270b65d4 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -532,6 +532,12 @@ static const struct powercap_zone_constraint_ops constraint_ops = {
.get_name = get_constraint_name,
};
+/* Return the id used for read_raw/write_raw callback */
+static int get_rid(struct rapl_package *rp)
+{
+ return rp->lead_cpu >= 0 ? rp->lead_cpu : rp->id;
+}
+
/* called after domain detection and package level data are set */
static void rapl_init_domains(struct rapl_package *rp)
{
@@ -550,10 +556,12 @@ static void rapl_init_domains(struct rapl_package *rp)
if (i == RAPL_DOMAIN_PLATFORM && rp->id > 0) {
snprintf(rd->name, RAPL_DOMAIN_NAME_LENGTH, "psys-%d",
- topology_physical_package_id(rp->lead_cpu));
- } else
+ rp->lead_cpu >= 0 ? topology_physical_package_id(rp->lead_cpu) :
+ rp->id);
+ } else {
snprintf(rd->name, RAPL_DOMAIN_NAME_LENGTH, "%s",
rapl_domain_names[i]);
+ }
rd->id = i;
@@ -725,7 +733,6 @@ static int rapl_read_data_raw(struct rapl_domain *rd,
enum rapl_primitives prim_fixed = prim_fixups(rd, prim);
struct rapl_primitive_info *rpi = get_rpi(rd->rp, prim_fixed);
struct reg_action ra;
- int cpu;
if (!rpi || !rpi->name || rpi->flag & RAPL_PRIMITIVE_DUMMY)
return -EINVAL;
@@ -734,8 +741,6 @@ static int rapl_read_data_raw(struct rapl_domain *rd,
if (!ra.reg)
return -EINVAL;
- cpu = rd->rp->lead_cpu;
-
/* non-hardware data are collected by the polling thread */
if (rpi->flag & RAPL_PRIMITIVE_DERIVED) {
*data = rd->rdd.primitives[prim];
@@ -744,8 +749,8 @@ static int rapl_read_data_raw(struct rapl_domain *rd,
ra.mask = rpi->mask;
- if (rd->rp->priv->read_raw(cpu, &ra)) {
- pr_debug("failed to read reg 0x%llx on cpu %d\n", ra.reg, cpu);
+ if (rd->rp->priv->read_raw(get_rid(rd->rp), &ra)) {
+ pr_debug("failed to read reg 0x%llx for %s:%s\n", ra.reg, rd->rp->name, rd->name);
return -EIO;
}
@@ -766,7 +771,6 @@ static int rapl_write_data_raw(struct rapl_domain *rd,
{
enum rapl_primitives prim_fixed = prim_fixups(rd, prim);
struct rapl_primitive_info *rpi = get_rpi(rd->rp, prim_fixed);
- int cpu;
u64 bits;
struct reg_action ra;
int ret;
@@ -774,7 +778,6 @@ static int rapl_write_data_raw(struct rapl_domain *rd,
if (!rpi || !rpi->name || rpi->flag & RAPL_PRIMITIVE_DUMMY)
return -EINVAL;
- cpu = rd->rp->lead_cpu;
bits = rapl_unit_xlate(rd, rpi->unit, value, 1);
bits <<= rpi->shift;
bits &= rpi->mask;
@@ -785,7 +788,7 @@ static int rapl_write_data_raw(struct rapl_domain *rd,
ra.mask = rpi->mask;
ra.value = bits;
- ret = rd->rp->priv->write_raw(cpu, &ra);
+ ret = rd->rp->priv->write_raw(get_rid(rd->rp), &ra);
return ret;
}
@@ -835,9 +838,9 @@ static int rapl_check_unit_core(struct rapl_domain *rd)
ra.reg = rd->regs[RAPL_DOMAIN_REG_UNIT];
ra.mask = ~0;
- if (rd->rp->priv->read_raw(rd->rp->lead_cpu, &ra)) {
- pr_err("Failed to read power unit REG 0x%llx on CPU %d, exit.\n",
- ra.reg, rd->rp->lead_cpu);
+ if (rd->rp->priv->read_raw(get_rid(rd->rp), &ra)) {
+ pr_err("Failed to read power unit REG 0x%llx on %s:%s, exit.\n",
+ ra.reg, rd->rp->name, rd->name);
return -ENODEV;
}
@@ -863,9 +866,9 @@ static int rapl_check_unit_atom(struct rapl_domain *rd)
ra.reg = rd->regs[RAPL_DOMAIN_REG_UNIT];
ra.mask = ~0;
- if (rd->rp->priv->read_raw(rd->rp->lead_cpu, &ra)) {
- pr_err("Failed to read power unit REG 0x%llx on CPU %d, exit.\n",
- ra.reg, rd->rp->lead_cpu);
+ if (rd->rp->priv->read_raw(get_rid(rd->rp), &ra)) {
+ pr_err("Failed to read power unit REG 0x%llx on %s:%s, exit.\n",
+ ra.reg, rd->rp->name, rd->name);
return -ENODEV;
}
@@ -911,6 +914,9 @@ static void power_limit_irq_save_cpu(void *info)
static void package_power_limit_irq_save(struct rapl_package *rp)
{
+ if (rp->lead_cpu < 0)
+ return;
+
if (!boot_cpu_has(X86_FEATURE_PTS) || !boot_cpu_has(X86_FEATURE_PLN))
return;
@@ -925,6 +931,9 @@ static void package_power_limit_irq_restore(struct rapl_package *rp)
{
u32 l, h;
+ if (rp->lead_cpu < 0)
+ return;
+
if (!boot_cpu_has(X86_FEATURE_PTS) || !boot_cpu_has(X86_FEATURE_PLN))
return;
@@ -1263,7 +1272,7 @@ static int rapl_check_domain(int domain, struct rapl_package *rp)
*/
ra.mask = ENERGY_STATUS_MASK;
- if (rp->priv->read_raw(rp->lead_cpu, &ra) || !ra.value)
+ if (rp->priv->read_raw(get_rid(rp), &ra) || !ra.value)
return -ENODEV;
return 0;
@@ -1401,13 +1410,18 @@ void rapl_remove_package(struct rapl_package *rp)
EXPORT_SYMBOL_GPL(rapl_remove_package);
/* caller to ensure CPU hotplug lock is held */
-struct rapl_package *rapl_find_package_domain(int cpu, struct rapl_if_priv *priv)
+struct rapl_package *rapl_find_package_domain(int id, struct rapl_if_priv *priv, bool id_is_cpu)
{
- int id = topology_logical_die_id(cpu);
struct rapl_package *rp;
+ int uid;
+
+ if (id_is_cpu)
+ uid = topology_logical_die_id(id);
+ else
+ uid = id;
list_for_each_entry(rp, &rapl_packages, plist) {
- if (rp->id == id
+ if (rp->id == uid
&& rp->priv->control_type == priv->control_type)
return rp;
}
@@ -1417,9 +1431,8 @@ struct rapl_package *rapl_find_package_domain(int cpu, struct rapl_if_priv *priv
EXPORT_SYMBOL_GPL(rapl_find_package_domain);
/* called from CPU hotplug notifier, hotplug lock held */
-struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv)
+struct rapl_package *rapl_add_package(int id, struct rapl_if_priv *priv, bool id_is_cpu)
{
- int id = topology_logical_die_id(cpu);
struct rapl_package *rp;
int ret;
@@ -1427,23 +1440,26 @@ struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv)
if (!rp)
return ERR_PTR(-ENOMEM);
- /* add the new package to the list */
- rp->id = id;
- rp->lead_cpu = cpu;
- rp->priv = priv;
+ if (id_is_cpu) {
+ rp->id = topology_logical_die_id(id);
+ rp->lead_cpu = id;
+ if (topology_max_die_per_package() > 1)
+ snprintf(rp->name, PACKAGE_DOMAIN_NAME_LENGTH, "package-%d-die-%d",
+ topology_physical_package_id(id), topology_die_id(id));
+ else
+ snprintf(rp->name, PACKAGE_DOMAIN_NAME_LENGTH, "package-%d",
+ topology_physical_package_id(id));
+ } else {
+ rp->id = id;
+ rp->lead_cpu = -1;
+ snprintf(rp->name, PACKAGE_DOMAIN_NAME_LENGTH, "package-%d", id);
+ }
+ rp->priv = priv;
ret = rapl_config(rp);
if (ret)
goto err_free_package;
- if (topology_max_die_per_package() > 1)
- snprintf(rp->name, PACKAGE_DOMAIN_NAME_LENGTH,
- "package-%d-die-%d",
- topology_physical_package_id(cpu), topology_die_id(cpu));
- else
- snprintf(rp->name, PACKAGE_DOMAIN_NAME_LENGTH, "package-%d",
- topology_physical_package_id(cpu));
-
/* check if the package contains valid domains */
if (rapl_detect_domains(rp)) {
ret = -ENODEV;
diff --git a/drivers/powercap/intel_rapl_msr.c b/drivers/powercap/intel_rapl_msr.c
index 6fe5e556aa51..fff1a49d3b6c 100644
--- a/drivers/powercap/intel_rapl_msr.c
+++ b/drivers/powercap/intel_rapl_msr.c
@@ -68,9 +68,9 @@ static int rapl_cpu_online(unsigned int cpu)
{
struct rapl_package *rp;
- rp = rapl_find_package_domain(cpu, rapl_msr_priv);
+ rp = rapl_find_package_domain(cpu, rapl_msr_priv, true);
if (!rp) {
- rp = rapl_add_package(cpu, rapl_msr_priv);
+ rp = rapl_add_package(cpu, rapl_msr_priv, true);
if (IS_ERR(rp))
return PTR_ERR(rp);
}
@@ -83,7 +83,7 @@ static int rapl_cpu_down_prep(unsigned int cpu)
struct rapl_package *rp;
int lead_cpu;
- rp = rapl_find_package_domain(cpu, rapl_msr_priv);
+ rp = rapl_find_package_domain(cpu, rapl_msr_priv, true);
if (!rp)
return 0;
diff --git a/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c b/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
index e070239106f5..140fb85cfa7b 100644
--- a/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
+++ b/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
@@ -27,9 +27,9 @@ static int rapl_mmio_cpu_online(unsigned int cpu)
if (topology_physical_package_id(cpu))
return 0;
- rp = rapl_find_package_domain(cpu, &rapl_mmio_priv);
+ rp = rapl_find_package_domain(cpu, &rapl_mmio_priv, true);
if (!rp) {
- rp = rapl_add_package(cpu, &rapl_mmio_priv);
+ rp = rapl_add_package(cpu, &rapl_mmio_priv, true);
if (IS_ERR(rp))
return PTR_ERR(rp);
}
@@ -42,7 +42,7 @@ static int rapl_mmio_cpu_down_prep(unsigned int cpu)
struct rapl_package *rp;
int lead_cpu;
- rp = rapl_find_package_domain(cpu, &rapl_mmio_priv);
+ rp = rapl_find_package_domain(cpu, &rapl_mmio_priv, true);
if (!rp)
return 0;
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index d07b460bac3b..51509f35027b 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -135,8 +135,8 @@ struct rapl_if_priv {
u64 reg_unit;
u64 regs[RAPL_DOMAIN_MAX][RAPL_DOMAIN_REG_MAX];
int limits[RAPL_DOMAIN_MAX];
- int (*read_raw)(int cpu, struct reg_action *ra);
- int (*write_raw)(int cpu, struct reg_action *ra);
+ int (*read_raw)(int id, struct reg_action *ra);
+ int (*write_raw)(int id, struct reg_action *ra);
void *defaults;
void *rpi;
};
@@ -161,8 +161,8 @@ struct rapl_package {
struct rapl_if_priv *priv;
};
-struct rapl_package *rapl_find_package_domain(int cpu, struct rapl_if_priv *priv);
-struct rapl_package *rapl_add_package(int cpu, struct rapl_if_priv *priv);
+struct rapl_package *rapl_find_package_domain(int id, struct rapl_if_priv *priv, bool id_is_cpu);
+struct rapl_package *rapl_add_package(int id, struct rapl_if_priv *priv, bool id_is_cpu);
void rapl_remove_package(struct rapl_package *rp);
#endif /* __INTEL_RAPL_H__ */
--
2.25.1
Different RAPL Interfaces may have different primitive information and
rapl_defaults calls.
To better distinguish this difference in the RAPL framework code,
introduce a new enum to represent different types of RAPL Interfaces.
No functional change.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 18 +++++++++++++-----
drivers/powercap/intel_rapl_msr.c | 2 ++
.../int340x_thermal/processor_thermal_rapl.c | 1 +
include/linux/intel_rapl.h | 6 ++++++
4 files changed, 22 insertions(+), 5 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index d337270b65d4..b70d6915a6ba 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -186,7 +186,7 @@ struct rapl_defaults {
unsigned int psys_domain_energy_unit;
bool spr_psys_bits;
};
-static struct rapl_defaults *rapl_defaults;
+static struct rapl_defaults *defaults_msr;
static struct rapl_defaults *get_defaults(struct rapl_package *rp)
{
@@ -610,7 +610,7 @@ static u64 rapl_unit_xlate(struct rapl_domain *rd, enum unit_type type,
return div64_u64(value, scale);
}
-static struct rapl_primitive_info rpi_default[NR_RAPL_PRIMITIVES] = {
+static struct rapl_primitive_info rpi_msr[NR_RAPL_PRIMITIVES] = {
/* name, mask, shift, msr index, unit divisor */
[POWER_LIMIT1] = PRIMITIVE_INFO_INIT(POWER_LIMIT1, POWER_LIMIT1_MASK, 0,
RAPL_DOMAIN_REG_LIMIT, POWER_UNIT, 0),
@@ -679,8 +679,16 @@ static struct rapl_primitive_info *get_rpi(struct rapl_package *rp, int prim)
static int rapl_config(struct rapl_package *rp)
{
- rp->priv->defaults = (void *)rapl_defaults;
- rp->priv->rpi = (void *)rpi_default;
+ switch (rp->priv->type) {
+ /* MMIO I/F shares the same register layout as MSR registers */
+ case RAPL_IF_MMIO:
+ case RAPL_IF_MSR:
+ rp->priv->defaults = (void *)defaults_msr;
+ rp->priv->rpi = (void *)rpi_msr;
+ break;
+ default:
+ return -EINVAL;
+ }
return 0;
}
@@ -1546,7 +1554,7 @@ static int __init rapl_init(void)
id = x86_match_cpu(rapl_ids);
if (id) {
- rapl_defaults = (struct rapl_defaults *)id->driver_data;
+ defaults_msr = (struct rapl_defaults *)id->driver_data;
rapl_msr_platdev = platform_device_alloc("intel_rapl_msr", 0);
if (!rapl_msr_platdev)
diff --git a/drivers/powercap/intel_rapl_msr.c b/drivers/powercap/intel_rapl_msr.c
index fff1a49d3b6c..cff5c6e8d570 100644
--- a/drivers/powercap/intel_rapl_msr.c
+++ b/drivers/powercap/intel_rapl_msr.c
@@ -34,6 +34,7 @@
static struct rapl_if_priv *rapl_msr_priv;
static struct rapl_if_priv rapl_msr_priv_intel = {
+ .type = RAPL_IF_MSR,
.reg_unit = MSR_RAPL_POWER_UNIT,
.regs[RAPL_DOMAIN_PACKAGE] = {
MSR_PKG_POWER_LIMIT, MSR_PKG_ENERGY_STATUS, MSR_PKG_PERF_STATUS, 0, MSR_PKG_POWER_INFO },
@@ -50,6 +51,7 @@ static struct rapl_if_priv rapl_msr_priv_intel = {
};
static struct rapl_if_priv rapl_msr_priv_amd = {
+ .type = RAPL_IF_MSR,
.reg_unit = MSR_AMD_RAPL_POWER_UNIT,
.regs[RAPL_DOMAIN_PACKAGE] = {
0, MSR_AMD_PKG_ENERGY_STATUS, 0, 0, 0 },
diff --git a/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c b/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
index 140fb85cfa7b..013f1633f082 100644
--- a/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
+++ b/drivers/thermal/intel/int340x_thermal/processor_thermal_rapl.c
@@ -97,6 +97,7 @@ int proc_thermal_rapl_add(struct pci_dev *pdev, struct proc_thermal_device *proc
rapl_regs->regs[domain][reg];
rapl_mmio_priv.limits[domain] = rapl_regs->limits[domain];
}
+ rapl_mmio_priv.type = RAPL_IF_MMIO;
rapl_mmio_priv.reg_unit = (u64)proc_priv->mmio_base + rapl_regs->reg_unit;
rapl_mmio_priv.read_raw = rapl_mmio_read_raw;
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index 51509f35027b..65f358b64096 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -14,6 +14,11 @@
#include <linux/powercap.h>
#include <linux/cpuhotplug.h>
+enum rapl_if_type {
+ RAPL_IF_MSR, /* RAPL I/F using MSR registers */
+ RAPL_IF_MMIO, /* RAPL I/F using MMIO registers */
+};
+
enum rapl_domain_type {
RAPL_DOMAIN_PACKAGE, /* entire package/socket */
RAPL_DOMAIN_PP0, /* core power plane */
@@ -130,6 +135,7 @@ struct reg_action {
* @rpi: internal pointer to interface primitive info
*/
struct rapl_if_priv {
+ enum rapl_if_type type;
struct powercap_control_type *control_type;
enum cpuhp_state pcap_rapl_online;
u64 reg_unit;
--
2.25.1
Compared with existing RAPL MSR/MMIO Interface, the RAPL TPMI Interface
1. has per Power Limit register, thus has per Power Limit Lock and
Enable bit.
2. doesn't have Power Limit Clamp bit.
3. the Power Limit Lock and Enable bits have different bit offsets.
These mean RAPL TPMI Interface needs its own primitive information.
RAPL TPMI Interface also has per domain unit register but with a
different register layout. This requires a TPMI specific rapl_defaults
call to decode the unit register.
Introduce the RAPL core support for TPMI Interface.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/intel_rapl_common.c | 119 ++++++++++++++++++++++++++-
include/linux/intel_rapl.h | 5 ++
2 files changed, 122 insertions(+), 2 deletions(-)
diff --git a/drivers/powercap/intel_rapl_common.c b/drivers/powercap/intel_rapl_common.c
index b70d6915a6ba..4e646e5e48f6 100644
--- a/drivers/powercap/intel_rapl_common.c
+++ b/drivers/powercap/intel_rapl_common.c
@@ -75,6 +75,15 @@
#define PSYS_TIME_WINDOW1_MASK (0x7FULL<<19)
#define PSYS_TIME_WINDOW2_MASK (0x7FULL<<51)
+/* bitmasks for RAPL TPMI, used by primitive access functions */
+#define TPMI_POWER_LIMIT_MASK 0x3FFFF
+#define TPMI_POWER_LIMIT_ENABLE BIT_ULL(62)
+#define TPMI_TIME_WINDOW_MASK (0x7FULL<<18)
+#define TPMI_INFO_SPEC_MASK 0x3FFFF
+#define TPMI_INFO_MIN_MASK (0x3FFFFULL << 18)
+#define TPMI_INFO_MAX_MASK (0x3FFFFULL << 36)
+#define TPMI_INFO_MAX_TIME_WIN_MASK (0x7FULL << 54)
+
/* Non HW constants */
#define RAPL_PRIMITIVE_DERIVED BIT(1) /* not from raw data */
#define RAPL_PRIMITIVE_DUMMY BIT(2)
@@ -119,6 +128,19 @@ static bool is_pl_valid(struct rapl_domain *rd, int pl)
static int get_pl_lock_prim(struct rapl_domain *rd, int pl)
{
+ if (rd->rp->priv->type == RAPL_IF_TPMI) {
+ if (pl == POWER_LIMIT1)
+ return PL1_LOCK;
+ if (pl == POWER_LIMIT2)
+ return PL2_LOCK;
+ if (pl == POWER_LIMIT4)
+ return PL4_LOCK;
+ }
+
+ /* MSR/MMIO Interface doesn't have Lock bit for PL4 */
+ if (pl == POWER_LIMIT4)
+ return -EINVAL;
+
/*
* Power Limit register that supports two power limits has a different
* bit position for the Lock bit.
@@ -134,7 +156,7 @@ static int get_pl_prim(struct rapl_domain *rd, int pl, enum pl_prims prim)
case POWER_LIMIT1:
if (prim == PL_ENABLE)
return PL1_ENABLE;
- if (prim == PL_CLAMP)
+ if (prim == PL_CLAMP && rd->rp->priv->type != RAPL_IF_TPMI)
return PL1_CLAMP;
if (prim == PL_LIMIT)
return POWER_LIMIT1;
@@ -148,7 +170,7 @@ static int get_pl_prim(struct rapl_domain *rd, int pl, enum pl_prims prim)
case POWER_LIMIT2:
if (prim == PL_ENABLE)
return PL2_ENABLE;
- if (prim == PL_CLAMP)
+ if (prim == PL_CLAMP && rd->rp->priv->type != RAPL_IF_TPMI)
return PL2_CLAMP;
if (prim == PL_LIMIT)
return POWER_LIMIT2;
@@ -167,6 +189,8 @@ static int get_pl_prim(struct rapl_domain *rd, int pl, enum pl_prims prim)
/* PL4 would be around two times PL2, use same prim as PL2. */
if (prim == PL_MAX_POWER)
return MAX_POWER;
+ if (prim == PL_LOCK)
+ return get_pl_lock_prim(rd, pl);
return -EINVAL;
default:
return -EINVAL;
@@ -187,6 +211,7 @@ struct rapl_defaults {
bool spr_psys_bits;
};
static struct rapl_defaults *defaults_msr;
+static const struct rapl_defaults defaults_tpmi;
static struct rapl_defaults *get_defaults(struct rapl_package *rp)
{
@@ -610,6 +635,7 @@ static u64 rapl_unit_xlate(struct rapl_domain *rd, enum unit_type type,
return div64_u64(value, scale);
}
+/* RAPL primitives for MSR and MMIO I/F */
static struct rapl_primitive_info rpi_msr[NR_RAPL_PRIMITIVES] = {
/* name, mask, shift, msr index, unit divisor */
[POWER_LIMIT1] = PRIMITIVE_INFO_INIT(POWER_LIMIT1, POWER_LIMIT1_MASK, 0,
@@ -667,6 +693,48 @@ static struct rapl_primitive_info rpi_msr[NR_RAPL_PRIMITIVES] = {
RAPL_PRIMITIVE_DERIVED),
};
+/* RAPL primitives for TPMI I/F */
+static struct rapl_primitive_info rpi_tpmi[NR_RAPL_PRIMITIVES] = {
+ /* name, mask, shift, msr index, unit divisor */
+ [POWER_LIMIT1] = PRIMITIVE_INFO_INIT(POWER_LIMIT1, TPMI_POWER_LIMIT_MASK, 0,
+ RAPL_DOMAIN_REG_LIMIT, POWER_UNIT, 0),
+ [POWER_LIMIT2] = PRIMITIVE_INFO_INIT(POWER_LIMIT2, TPMI_POWER_LIMIT_MASK, 0,
+ RAPL_DOMAIN_REG_PL2, POWER_UNIT, 0),
+ [POWER_LIMIT4] = PRIMITIVE_INFO_INIT(POWER_LIMIT4, TPMI_POWER_LIMIT_MASK, 0,
+ RAPL_DOMAIN_REG_PL4, POWER_UNIT, 0),
+ [ENERGY_COUNTER] = PRIMITIVE_INFO_INIT(ENERGY_COUNTER, ENERGY_STATUS_MASK, 0,
+ RAPL_DOMAIN_REG_STATUS, ENERGY_UNIT, 0),
+ [PL1_LOCK] = PRIMITIVE_INFO_INIT(PL1_LOCK, POWER_HIGH_LOCK, 63,
+ RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
+ [PL2_LOCK] = PRIMITIVE_INFO_INIT(PL2_LOCK, POWER_HIGH_LOCK, 63,
+ RAPL_DOMAIN_REG_PL2, ARBITRARY_UNIT, 0),
+ [PL4_LOCK] = PRIMITIVE_INFO_INIT(PL4_LOCK, POWER_HIGH_LOCK, 63,
+ RAPL_DOMAIN_REG_PL4, ARBITRARY_UNIT, 0),
+ [PL1_ENABLE] = PRIMITIVE_INFO_INIT(PL1_ENABLE, TPMI_POWER_LIMIT_ENABLE, 62,
+ RAPL_DOMAIN_REG_LIMIT, ARBITRARY_UNIT, 0),
+ [PL2_ENABLE] = PRIMITIVE_INFO_INIT(PL2_ENABLE, TPMI_POWER_LIMIT_ENABLE, 62,
+ RAPL_DOMAIN_REG_PL2, ARBITRARY_UNIT, 0),
+ [PL4_ENABLE] = PRIMITIVE_INFO_INIT(PL4_ENABLE, TPMI_POWER_LIMIT_ENABLE, 62,
+ RAPL_DOMAIN_REG_PL4, ARBITRARY_UNIT, 0),
+ [TIME_WINDOW1] = PRIMITIVE_INFO_INIT(TIME_WINDOW1, TPMI_TIME_WINDOW_MASK, 18,
+ RAPL_DOMAIN_REG_LIMIT, TIME_UNIT, 0),
+ [TIME_WINDOW2] = PRIMITIVE_INFO_INIT(TIME_WINDOW2, TPMI_TIME_WINDOW_MASK, 18,
+ RAPL_DOMAIN_REG_PL2, TIME_UNIT, 0),
+ [THERMAL_SPEC_POWER] = PRIMITIVE_INFO_INIT(THERMAL_SPEC_POWER, TPMI_INFO_SPEC_MASK, 0,
+ RAPL_DOMAIN_REG_INFO, POWER_UNIT, 0),
+ [MAX_POWER] = PRIMITIVE_INFO_INIT(MAX_POWER, TPMI_INFO_MAX_MASK, 36,
+ RAPL_DOMAIN_REG_INFO, POWER_UNIT, 0),
+ [MIN_POWER] = PRIMITIVE_INFO_INIT(MIN_POWER, TPMI_INFO_MIN_MASK, 18,
+ RAPL_DOMAIN_REG_INFO, POWER_UNIT, 0),
+ [MAX_TIME_WINDOW] = PRIMITIVE_INFO_INIT(MAX_TIME_WINDOW, TPMI_INFO_MAX_TIME_WIN_MASK, 54,
+ RAPL_DOMAIN_REG_INFO, TIME_UNIT, 0),
+ [THROTTLED_TIME] = PRIMITIVE_INFO_INIT(THROTTLED_TIME, PERF_STATUS_THROTTLE_TIME_MASK, 0,
+ RAPL_DOMAIN_REG_PERF, TIME_UNIT, 0),
+ /* non-hardware */
+ [AVERAGE_POWER] = PRIMITIVE_INFO_INIT(AVERAGE_POWER, 0, 0, 0,
+ POWER_UNIT, RAPL_PRIMITIVE_DERIVED),
+};
+
static struct rapl_primitive_info *get_rpi(struct rapl_package *rp, int prim)
{
struct rapl_primitive_info *rpi = rp->priv->rpi;
@@ -686,6 +754,10 @@ static int rapl_config(struct rapl_package *rp)
rp->priv->defaults = (void *)defaults_msr;
rp->priv->rpi = (void *)rpi_msr;
break;
+ case RAPL_IF_TPMI:
+ rp->priv->defaults = (void *)&defaults_tpmi;
+ rp->priv->rpi = (void *)rpi_tpmi;
+ break;
default:
return -EINVAL;
}
@@ -1046,6 +1118,49 @@ static u64 rapl_compute_time_window_atom(struct rapl_domain *rd, u64 value,
return value;
}
+/* TPMI Unit register has different layout */
+#define TPMI_POWER_UNIT_OFFSET POWER_UNIT_OFFSET
+#define TPMI_POWER_UNIT_MASK POWER_UNIT_MASK
+#define TPMI_ENERGY_UNIT_OFFSET 0x06
+#define TPMI_ENERGY_UNIT_MASK 0x7C0
+#define TPMI_TIME_UNIT_OFFSET 0x0C
+#define TPMI_TIME_UNIT_MASK 0xF000
+
+static int rapl_check_unit_tpmi(struct rapl_domain *rd)
+{
+ struct reg_action ra;
+ u32 value;
+
+ ra.reg = rd->regs[RAPL_DOMAIN_REG_UNIT];
+ ra.mask = ~0;
+ if (rd->rp->priv->read_raw(get_rid(rd->rp), &ra)) {
+ pr_err("Failed to read power unit REG 0x%llx on %s:%s, exit.\n",
+ ra.reg, rd->rp->name, rd->name);
+ return -ENODEV;
+ }
+
+ value = (ra.value & TPMI_ENERGY_UNIT_MASK) >> TPMI_ENERGY_UNIT_OFFSET;
+ rd->energy_unit = ENERGY_UNIT_SCALE * 1000000 / (1 << value);
+
+ value = (ra.value & TPMI_POWER_UNIT_MASK) >> TPMI_POWER_UNIT_OFFSET;
+ rd->power_unit = 1000000 / (1 << value);
+
+ value = (ra.value & TPMI_TIME_UNIT_MASK) >> TPMI_TIME_UNIT_OFFSET;
+ rd->time_unit = 1000000 / (1 << value);
+
+ pr_debug("Core CPU %s:%s energy=%dpJ, time=%dus, power=%duW\n",
+ rd->rp->name, rd->name, rd->energy_unit, rd->time_unit, rd->power_unit);
+
+ return 0;
+}
+
+static const struct rapl_defaults defaults_tpmi = {
+ .check_unit = rapl_check_unit_tpmi,
+ /* Reuse existing logic, ignore the PL_CLAMP failures and enable all Power Limits */
+ .set_floor_freq = set_floor_freq_default,
+ .compute_time_window = rapl_compute_time_window_core,
+};
+
static const struct rapl_defaults rapl_defaults_core = {
.floor_freq_reg_addr = 0,
.check_unit = rapl_check_unit_core,
diff --git a/include/linux/intel_rapl.h b/include/linux/intel_rapl.h
index 65f358b64096..e6936cb25047 100644
--- a/include/linux/intel_rapl.h
+++ b/include/linux/intel_rapl.h
@@ -17,6 +17,7 @@
enum rapl_if_type {
RAPL_IF_MSR, /* RAPL I/F using MSR registers */
RAPL_IF_MMIO, /* RAPL I/F using MMIO registers */
+ RAPL_IF_TPMI, /* RAPL I/F using TPMI registers */
};
enum rapl_domain_type {
@@ -36,6 +37,7 @@ enum rapl_domain_reg_id {
RAPL_DOMAIN_REG_INFO,
RAPL_DOMAIN_REG_PL4,
RAPL_DOMAIN_REG_UNIT,
+ RAPL_DOMAIN_REG_PL2,
RAPL_DOMAIN_REG_MAX,
};
@@ -48,6 +50,9 @@ enum rapl_primitives {
ENERGY_COUNTER,
FW_LOCK,
FW_HIGH_LOCK,
+ PL1_LOCK,
+ PL2_LOCK,
+ PL4_LOCK,
PL1_ENABLE, /* power limit 1, aka long term */
PL1_CLAMP, /* allow frequency to go below OS request */
--
2.25.1
The TPMI (Topology Aware Register and PM Capsule Interface) provides a
flexible, extendable and PCIe enumerable MMIO interface for PM features.
Intel RAPL (Running Average Power Limit) is one of the features that
benefit from this. Using TPMI Interface has advantage over traditional MSR
(Model Specific Register) interface, where a thread needs to be scheduled
on the target CPU to read or write. Also the RAPL features vary between
CPU models, and hence lot of model specific code. Here TPMI provides an
architectural interface by providing hierarchical tables and fields,
which will not need any model specific implementation.
TPMI interface uses a PCI VSEC structure to expose the location of MMIO
interface for PM feature enumeration and control.
The Intel VSEC driver parses VSEC structures present in the PCI
configuration space of the given device and creates an auxiliary device
object for each of them. In particular, it creates an auxiliary device
object representing TPMI that can be bound to by an auxiliary driver.
Then the TPMI enumeration driver binds to the TPMI auxiliary device
object created by the Intel VSEC driver, parses the PM Feature Structure
(PFS) present in the TPMI MMIO region and creates device nodes for PM
features described in the PFS.
This RAPL TPMI Interface driver binds the RAPL auxiliary device created
by the TPMI enumeration driver and expose the RAPL control to userspace
via powercap sysfs class.
RAPL TPMI details are published in the following document:
https://github.com/intel/tpmi_power_management/blob/main/RAPL_TPMI_public_disclosure_FINAL.docx
Note, for now, the RAPL TPMI Interface and RAPL MSR Interface cannot
co-exists on the same platform (RAPL TPMI Interface is not supported on
any platforms in the CPU model list for RAPL MSR Interface). Thus
register the RAPL TPMI powercap control type with name "intel-rapl",
the same as RAPL MSR Interface, so that it is transparent to userspace.
Signed-off-by: Zhang Rui <[email protected]>
Tested-by: Wang Wendy <[email protected]>
---
drivers/powercap/Kconfig | 14 ++
drivers/powercap/Makefile | 1 +
drivers/powercap/intel_rapl_tpmi.c | 325 +++++++++++++++++++++++++++++
3 files changed, 340 insertions(+)
create mode 100644 drivers/powercap/intel_rapl_tpmi.c
diff --git a/drivers/powercap/Kconfig b/drivers/powercap/Kconfig
index 90d33cd1b670..e71399804c14 100644
--- a/drivers/powercap/Kconfig
+++ b/drivers/powercap/Kconfig
@@ -33,6 +33,20 @@ config INTEL_RAPL
controller, CPU core (Power Plane 0), graphics uncore (Power Plane
1), etc.
+config INTEL_RAPL_TPMI
+ tristate "Intel RAPL Support via TPMI Interface"
+ depends on X86
+ depends on INTEL_TPMI
+ select INTEL_RAPL_CORE
+ help
+ This enables support for the Intel Running Average Power Limit (RAPL)
+ technology via TPMI interface, which allows power limits to be enforced
+ and monitored.
+
+ In RAPL, the platform level settings are divided into domains for
+ fine grained control. These domains include processor package, DRAM
+ controller, platform, etc.
+
config IDLE_INJECT
bool "Idle injection framework"
depends on CPU_IDLE
diff --git a/drivers/powercap/Makefile b/drivers/powercap/Makefile
index 4474201b4aa7..5ab0dce565b9 100644
--- a/drivers/powercap/Makefile
+++ b/drivers/powercap/Makefile
@@ -5,5 +5,6 @@ obj-$(CONFIG_DTPM_DEVFREQ) += dtpm_devfreq.o
obj-$(CONFIG_POWERCAP) += powercap_sys.o
obj-$(CONFIG_INTEL_RAPL_CORE) += intel_rapl_common.o
obj-$(CONFIG_INTEL_RAPL) += intel_rapl_msr.o
+obj-$(CONFIG_INTEL_RAPL_TPMI) += intel_rapl_tpmi.o
obj-$(CONFIG_IDLE_INJECT) += idle_inject.o
obj-$(CONFIG_ARM_SCMI_POWERCAP) += arm_scmi_powercap.o
diff --git a/drivers/powercap/intel_rapl_tpmi.c b/drivers/powercap/intel_rapl_tpmi.c
new file mode 100644
index 000000000000..c016127b3497
--- /dev/null
+++ b/drivers/powercap/intel_rapl_tpmi.c
@@ -0,0 +1,325 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * intel_rapl_tpmi: Intel RAPL driver via TPMI interface
+ *
+ * Copyright (c) 2023, Intel Corporation.
+ * All Rights Reserved.
+ *
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/auxiliary_bus.h>
+#include <linux/io.h>
+#include <linux/intel_tpmi.h>
+#include <linux/intel_rapl.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+
+#define TPMI_RAPL_VERSION 1
+
+/* 1 header + 10 registers + 5 reserved. 8 bytes for each. */
+#define TPMI_RAPL_DOMAIN_SIZE 128
+
+enum tpmi_rapl_domain_type {
+ TPMI_RAPL_DOMAIN_INVALID,
+ TPMI_RAPL_DOMAIN_SYSTEM,
+ TPMI_RAPL_DOMAIN_PACKAGE,
+ TPMI_RAPL_DOMAIN_RESERVED,
+ TPMI_RAPL_DOMAIN_MEMORY,
+ TPMI_RAPL_DOMAIN_MAX,
+};
+
+enum tpmi_rapl_register {
+ TPMI_RAPL_REG_HEADER,
+ TPMI_RAPL_REG_UNIT,
+ TPMI_RAPL_REG_PL1,
+ TPMI_RAPL_REG_PL2,
+ TPMI_RAPL_REG_PL3,
+ TPMI_RAPL_REG_PL4,
+ TPMI_RAPL_REG_RESERVED,
+ TPMI_RAPL_REG_ENERGY_STATUS,
+ TPMI_RAPL_REG_PERF_STATUS,
+ TPMI_RAPL_REG_POWER_INFO,
+ TPMI_RAPL_REG_INTERRUPT,
+ TPMI_RAPL_REG_MAX = 15,
+};
+
+struct tpmi_rapl_package {
+ struct rapl_if_priv priv;
+ struct intel_tpmi_plat_info *tpmi_info;
+ struct rapl_package *rp;
+ void __iomem *base;
+ struct list_head node;
+};
+
+static LIST_HEAD(tpmi_rapl_packages);
+static DEFINE_MUTEX(tpmi_rapl_lock);
+
+static struct powercap_control_type *tpmi_control_type;
+
+static int tpmi_rapl_read_raw(int id, struct reg_action *ra)
+{
+ if (!ra->reg)
+ return -EINVAL;
+
+ ra->value = readq((void __iomem *)ra->reg);
+
+ ra->value &= ra->mask;
+ return 0;
+}
+
+static int tpmi_rapl_write_raw(int id, struct reg_action *ra)
+{
+ u64 val;
+
+ if (!ra->reg)
+ return -EINVAL;
+
+ val = readq((void __iomem *)ra->reg);
+
+ val &= ~ra->mask;
+ val |= ra->value;
+
+ writeq(val, (void __iomem *)ra->reg);
+ return 0;
+}
+
+static struct tpmi_rapl_package *trp_alloc(int pkg_id)
+{
+ struct tpmi_rapl_package *trp;
+ int ret;
+
+ mutex_lock(&tpmi_rapl_lock);
+
+ if (list_empty(&tpmi_rapl_packages)) {
+ tpmi_control_type = powercap_register_control_type(NULL, "intel-rapl", NULL);
+ if (IS_ERR(tpmi_control_type)) {
+ ret = PTR_ERR(tpmi_control_type);
+ goto err_unlock;
+ }
+ }
+
+ trp = kzalloc(sizeof(*trp), GFP_KERNEL);
+ if (!trp) {
+ ret = -ENOMEM;
+ goto err_del_powercap;
+ }
+
+ list_add(&trp->node, &tpmi_rapl_packages);
+
+ mutex_unlock(&tpmi_rapl_lock);
+ return trp;
+
+err_del_powercap:
+ if (list_empty(&tpmi_rapl_packages))
+ powercap_unregister_control_type(tpmi_control_type);
+err_unlock:
+ mutex_unlock(&tpmi_rapl_lock);
+ return ERR_PTR(ret);
+}
+
+static void trp_release(struct tpmi_rapl_package *trp)
+{
+ mutex_lock(&tpmi_rapl_lock);
+ list_del(&trp->node);
+
+ if (list_empty(&tpmi_rapl_packages))
+ powercap_unregister_control_type(tpmi_control_type);
+
+ kfree(trp);
+ mutex_unlock(&tpmi_rapl_lock);
+}
+
+static int parse_one_domain(struct tpmi_rapl_package *trp, u32 offset)
+{
+ u8 tpmi_domain_version;
+ enum rapl_domain_type domain_type;
+ enum tpmi_rapl_domain_type tpmi_domain_type;
+ enum tpmi_rapl_register reg_index;
+ enum rapl_domain_reg_id reg_id;
+ int tpmi_domain_size, tpmi_domain_flags;
+ u64 *tpmi_rapl_regs = trp->base + offset;
+ u64 tpmi_domain_header = readq((void __iomem *)tpmi_rapl_regs);
+
+ /* Domain Parent bits are ignored for now */
+ tpmi_domain_version = tpmi_domain_header & 0xff;
+ tpmi_domain_type = tpmi_domain_header >> 8 & 0xff;
+ tpmi_domain_size = tpmi_domain_header >> 16 & 0xff;
+ tpmi_domain_flags = tpmi_domain_header >> 32 & 0xffff;
+
+ if (tpmi_domain_version != TPMI_RAPL_VERSION) {
+ pr_warn(FW_BUG "Unsupported version:%d\n", tpmi_domain_version);
+ return -ENODEV;
+ }
+
+ /* Domain size: in unit of 128 Bytes */
+ if (tpmi_domain_size != 1) {
+ pr_warn(FW_BUG "Invalid Domain size %d\n", tpmi_domain_size);
+ return -EINVAL;
+ }
+
+ /* Unit register and Energy Status register are mandatory for each domain */
+ if (!(tpmi_domain_flags & BIT(TPMI_RAPL_REG_UNIT)) ||
+ !(tpmi_domain_flags & BIT(TPMI_RAPL_REG_ENERGY_STATUS))) {
+ pr_warn(FW_BUG "Invalid Domain flag 0x%x\n", tpmi_domain_flags);
+ return -EINVAL;
+ }
+
+ switch (tpmi_domain_type) {
+ case TPMI_RAPL_DOMAIN_PACKAGE:
+ domain_type = RAPL_DOMAIN_PACKAGE;
+ break;
+ case TPMI_RAPL_DOMAIN_SYSTEM:
+ domain_type = RAPL_DOMAIN_PLATFORM;
+ break;
+ case TPMI_RAPL_DOMAIN_MEMORY:
+ domain_type = RAPL_DOMAIN_DRAM;
+ break;
+ default:
+ pr_warn(FW_BUG "Unsupported Domain type %d\n", tpmi_domain_type);
+ return -EINVAL;
+ }
+
+ if (trp->priv.regs[domain_type][RAPL_DOMAIN_REG_UNIT]) {
+ pr_warn(FW_BUG "Duplicate Domain type %d\n", tpmi_domain_type);
+ return -EINVAL;
+ }
+
+ reg_index = TPMI_RAPL_REG_HEADER;
+ while (++reg_index != TPMI_RAPL_REG_MAX) {
+ if (!(tpmi_domain_flags & BIT(reg_index)))
+ continue;
+
+ switch (reg_index) {
+ case TPMI_RAPL_REG_UNIT:
+ reg_id = RAPL_DOMAIN_REG_UNIT;
+ break;
+ case TPMI_RAPL_REG_PL1:
+ reg_id = RAPL_DOMAIN_REG_LIMIT;
+ trp->priv.limits[domain_type] |= BIT(POWER_LIMIT1);
+ break;
+ case TPMI_RAPL_REG_PL2:
+ reg_id = RAPL_DOMAIN_REG_PL2;
+ trp->priv.limits[domain_type] |= BIT(POWER_LIMIT2);
+ break;
+ case TPMI_RAPL_REG_PL4:
+ reg_id = RAPL_DOMAIN_REG_PL4;
+ trp->priv.limits[domain_type] |= BIT(POWER_LIMIT4);
+ break;
+ case TPMI_RAPL_REG_ENERGY_STATUS:
+ reg_id = RAPL_DOMAIN_REG_STATUS;
+ break;
+ case TPMI_RAPL_REG_PERF_STATUS:
+ reg_id = RAPL_DOMAIN_REG_PERF;
+ break;
+ case TPMI_RAPL_REG_POWER_INFO:
+ reg_id = RAPL_DOMAIN_REG_INFO;
+ break;
+ default:
+ continue;
+ }
+ trp->priv.regs[domain_type][reg_id] = (u64)&tpmi_rapl_regs[reg_index];
+ }
+
+ return 0;
+}
+
+static int intel_rapl_tpmi_probe(struct auxiliary_device *auxdev,
+ const struct auxiliary_device_id *id)
+{
+ struct tpmi_rapl_package *trp;
+ struct intel_tpmi_plat_info *info;
+ struct resource *res;
+ u32 offset;
+ int ret;
+
+ info = tpmi_get_platform_data(auxdev);
+ if (!info)
+ return -ENODEV;
+
+ trp = trp_alloc(info->package_id);
+ if (IS_ERR(trp))
+ return PTR_ERR(trp);
+
+ if (tpmi_get_resource_count(auxdev) > 1) {
+ dev_err(&auxdev->dev, "does not support multiple resources\n");
+ ret = -EINVAL;
+ goto err;
+ }
+
+ res = tpmi_get_resource_at_index(auxdev, 0);
+ if (!res) {
+ dev_err(&auxdev->dev, "can't fetch device resource info\n");
+ ret = -EIO;
+ goto err;
+ }
+
+ trp->base = devm_ioremap_resource(&auxdev->dev, res);
+ if (!trp->base) {
+ ret = -ENOMEM;
+ goto err;
+ }
+
+ for (offset = 0; offset < resource_size(res); offset += TPMI_RAPL_DOMAIN_SIZE) {
+ ret = parse_one_domain(trp, offset);
+ if (ret)
+ goto err;
+ }
+
+ trp->tpmi_info = info;
+ trp->priv.type = RAPL_IF_TPMI;
+ trp->priv.read_raw = tpmi_rapl_read_raw;
+ trp->priv.write_raw = tpmi_rapl_write_raw;
+ trp->priv.control_type = tpmi_control_type;
+
+ /* RAPL TPMI I/F is per physical package */
+ trp->rp = rapl_find_package_domain(info->package_id, &trp->priv, false);
+ if (trp->rp) {
+ dev_err(&auxdev->dev, "Domain for Package%d already exists\n", info->package_id);
+ ret = -EEXIST;
+ goto err;
+ }
+
+ trp->rp = rapl_add_package(info->package_id, &trp->priv, false);
+ if (IS_ERR(trp->rp)) {
+ dev_err(&auxdev->dev, "Failed to add RAPL Domain for Package%d, %ld\n",
+ info->package_id, PTR_ERR(trp->rp));
+ ret = PTR_ERR(trp->rp);
+ goto err;
+ }
+
+ auxiliary_set_drvdata(auxdev, trp);
+
+ return 0;
+err:
+ trp_release(trp);
+ return ret;
+}
+
+static void intel_rapl_tpmi_remove(struct auxiliary_device *auxdev)
+{
+ struct tpmi_rapl_package *trp = auxiliary_get_drvdata(auxdev);
+
+ rapl_remove_package(trp->rp);
+ trp_release(trp);
+}
+
+static const struct auxiliary_device_id intel_rapl_tpmi_ids[] = {
+ {.name = "intel_vsec.tpmi-rapl" },
+ { }
+};
+
+MODULE_DEVICE_TABLE(auxiliary, intel_rapl_tpmi_ids);
+
+static struct auxiliary_driver intel_rapl_tpmi_driver = {
+ .probe = intel_rapl_tpmi_probe,
+ .remove = intel_rapl_tpmi_remove,
+ .id_table = intel_rapl_tpmi_ids,
+};
+
+module_auxiliary_driver(intel_rapl_tpmi_driver)
+
+MODULE_IMPORT_NS(INTEL_TPMI);
+
+MODULE_DESCRIPTION("Intel RAPL TPMI Driver");
+MODULE_LICENSE("GPL");
--
2.25.1
On Wed, Apr 19, 2023 at 4:44 AM Zhang Rui <[email protected]> wrote:
>
> The TPMI (Topology Aware Register and PM Capsule Interface) provides a
> flexible, extendable and PCIe enumerable MMIO interface for PM features.
>
> The TPMI documentation can be downloaded from:
> https://github.com/intel/tpmi_power_management
>
> Intel RAPL (Running Average Power Limit) is one of the features that
> benefit from this. Using TPMI Interface has advantage over traditional MSR
> (Model Specific Register) interface, where a thread needs to be scheduled
> on the target CPU to read or write. Also the RAPL features vary between
> CPU models, and hence lot of model specific code. Here TPMI provides an
> architectural interface by providing hierarchical tables and fields,
> which will not need any model specific implementation.
>
> Given that there are some differences between RAPL TPMI Interface and the
> existing RAPL MSR/MMIO Interface, this patch series improves the RAPL
> common code to satisfy the new requirements from TPMI interface, and then
> introduces the RAPL TPMI Interface driver.
>
> Patch 1-4 cleanups and preparation work.
> Patch 5 adds support for per Domain Unit register.
> Patch 6-10 improves Power Limits handling, and provides support
> for per Power Limit register, and per Power Limit Lock.
> Patch 11-12 support rapl_package without online CPUs. So that TPMI
> rapl_package still works with whole package offlined.
> Patch 13-15 introduces RAPL Core support for TPMI Interface and the
> RAPL TPMI Interface driver.
>
> This series depends on the TPMI base driver which has been merged in 6.3-rc1.
>
> thanks,
> rui
> ---
> Changes since v1:
> - use set_defaults() and variable name 'defaults' for rapl_defaults
> structure.
> - use 'rpi_default' instead of 'rpis' for the default rapl primitive
> information of MSR/MMIO Interface.
> - rephase the changelog of patch 7/15.
> - change the subject of patch 10/15 and use a helper for getting the
> primitive for power limit LOCK bit control.
>
> ----------------------------------------------------------------
> Zhang Rui (15):
> powercap/intel_rapl: Remove unused field in struct rapl_if_priv
> powercap/intel_rapl: Allow probing without CPUID match
> powercap/intel_rapl: Support per Interface rapl_defaults
> powercap/intel_rapl: Support per Interface primitive information
> powercap/intel_rapl: Support per domain energy/power/time unit
> powercap/intel_rapl: Use index to initialize primitive information
> powercap/intel_rapl: Change primitive order
> powercap/intel_rapl: Use bitmap for Power Limits
> powercap/intel_rapl: Cleanup Power Limits support
> powercap/intel_rapl: Introduce per Power Limit lock
> powercap/intel_rapl: Remove redundant cpu parameter
> powercap/intel_rapl: Make cpu optional for rapl_package
> powercap/intel_rapl: Introduce RAPL I/F type
> powercap/intel_rapl: Introduce core support for TPMI interface
> powercap/intel_rapl_tpmi: Introduce RAPL TPMI interface driver
>
> drivers/powercap/Kconfig | 14 +
> drivers/powercap/Makefile | 1 +
> drivers/powercap/intel_rapl_common.c | 868 ++++++++++++---------
> drivers/powercap/intel_rapl_msr.c | 14 +-
> drivers/powercap/intel_rapl_tpmi.c | 325 ++++++++
> .../intel/int340x_thermal/processor_thermal_rapl.c | 11 +-
> include/linux/intel_rapl.h | 40 +-
> 7 files changed, 875 insertions(+), 398 deletions(-)
> create mode 100644 drivers/powercap/intel_rapl_tpmi.c
The whole series has been applied as 6.5 material, thanks!