Skip to contentSkip to conversation

Intel Hardware P-States (HWP) support

PATCH Superseded Jason Andryuk 3 participants 27 messages Jul 26, 2023 – Jul 31, 2023

v7 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

v7 cover Intel Hardware P-States (HWP) support

Hi,

This patch series adds Hardware-Controlled Performance States (HWP) for Intel processors to Xen.

v2 was only partially reviewed, so v3 is mostly a reposting of v2. In v2 & v3, I think I addressed all comments for v1. I kept patch 11 "xenpm:
Factor out a non-fatal cpuid_parse variant", with a v2 comment explaining why I keep it.

v3 adds "xen/x86: Tweak PDC bits when using HWP". Qubes testing revealed an issue where enabling HWP can crash firwmare code (maybe SMM). This requires a Linux change to get the PDC bits from Xen and pass them to ACPI. Roger has a patch [0] to set the PDC bits. Roger's 3 patch series was tested with "xen/x86: Tweak PDC bits when using HWP" on affected hardware and allowed proper operation.

v4:
There is a large amount or renaming from HWP/hwp to CPPC/cppc in the series. The driver remains hwp_ prefixed since it is dealing with the hardware interface. The sysctl, xc and xenpm interfaces were renamed to cppc to be the generic ACPI CPPC (Collaborative Processor Performance Control) interface.

struct xen_get_cpufreq_para was re-organized in a binary compatible fashion to nest scaling governor options. This allows the cppc support to use uint32_t's for its parameters.

HWP is now enabled with a top-level cpufreq=hwp option. It will fallback to cpufreq=xen if hwp is unavailable. This seems like the most user-friendly option. Since the user was trying to specify *some* cpufreq, we should give them the best that we can instead of disabling the functionality.

"xenpm: Factor out a non-fatal cpuid_parse variant" was dropped. set-cpufreq-cppc expects either a cpu number or none specified, which implies all.

Some patches were re-arrange - "xen/x86: Tweak PDC bits when using HWP" now comes immediately after "cpufreq: Add Hardware P-State (HWP) driver"

The implementation of "cpufreq: Allow restricting to internal governors only " changed, so I removed Jan's Ack.

v5:
HWP is enabled with a toplevel cpufreq=hwp option. There is no fallback by default, but a cpufreq=hwp;xen syntax is now supported. That tries hwp first. If HWP registration is unsuccessful, then xen registration is performed as a fallback.

More changes from Jan's feedback. They are typically minor and documented in individual patches.

v6:
After some discussion, the cpufreq=hwp;xen syntax is retained, with the global applicability of "verbose" mention in docs. Minor tweaks elsewhere as documented in individual patches.

v7:
Minor tweaks from review feedback - documented in respective patches. Jan previously asked for a toolstack side review / ack of "[PATCH 04/15] xen/sysctl: Nest cpufreq scaling options"

Previous cover letter:

With HWP, the processor makes its own determinations for frequency selection, though users can set some parameters and preferences. There is also Turbo Boost which dynamically pushes the max frequency if possible.

The existing governors don't work with HWP since they select frequencies and HWP doesn't expose those. Therefore a dummy hwp-interal governor is used that doesn't do anything.

xenpm get-cpufreq-para is extended to show HWP parameters, and set-cpufreq-cppc is added to set them.

A lightly loaded OpenXT laptop showed ~1W power savings according to powertop. A mostly idle Fedora system (dom0 only) showed a more modest power savings.

This is for a 10th gen 6-core 1600 MHz base 4900 MHZ max cpu. In the default balance mode, Turbo Boost doesn't exceed 4GHz. Tweaking the energy_perf preference with xenpm set-cpufreq-para balance ene:64, I've seen the CPU hit 4.7GHz before throttling down and bouncing around between 4.3 and 4.5 GHz. Curiously the other cores read ~4GHz when turbo boost takes affect. This was done after pinning all dom0 cores, and using taskset to pin to vCPU/pCPU 11 and running a bash tightloop.

HWP defaults to disabled and running with the existing HWP configuration
- it doesn't reconfigure by default. It can be enabled with cpufreq=hwp.

Hardware Duty Cycling (HDC) is another feature to autonomously powerdown things. It defaults to enabled when HWP is enabled, but HDC can be disabled on the command line. cpufreq=xen:hwp,no-hdc

I've only tested on 8th gen and 10th gen systems with activity window and energy_perf support. So the pathes for CPUs lacking those features are untested.

Fast MSR support was removed in v2. The model specific checking was not done properly, and I don't have hardware to test with. Since writes are expected to be infrequent, I just removed the code.

This changes the systcl_pm_op hypercall, so that wants review.

Regards,
Jason

[0] https://lore.kernel.org/xen-devel/[email protected]/

Jason Andryuk (15):
cpufreq: Allow restricting to internal governors only
cpufreq: Add perf_freq to cpuinfo
cpufreq: Export intel_feature_detect xen/sysctl: Nest cpufreq scaling options pmstat&xenpm: Re-arrage for cpufreq union
cpufreq: Add Hardware P-State (HWP) driver xen/x86: Tweak PDC bits when using HWP
xenpm: Change get-cpufreq-para output for hwp
cpufreq: Export HWP parameters to userspace as CPPC
libxc: Include cppc_para in definitions
xenpm: Print HWP/CPPC parameters
xen: Add SET_CPUFREQ_HWP xen_sysctl_pm_op
libxc: Add xc_set_cpufreq_cppc
xenpm: Add set-cpufreq-cppc subcommand
CHANGELOG: Add Intel HWP entry

 CHANGELOG.md                                 |   2 +-
 docs/misc/xen-command-line.pandoc            |  20 +-
 tools/include/xenctrl.h                      |  28 +-
 tools/libs/ctrl/xc_pm.c                      |  42 +-
 tools/misc/xenpm.c                           | 392 ++++++++++-
 xen/arch/x86/acpi/cpufreq/Makefile           |   1 +
 xen/arch/x86/acpi/cpufreq/cpufreq.c          |  31 +-
 xen/arch/x86/acpi/cpufreq/hwp.c              | 658 +++++++++++++++++++
 xen/arch/x86/acpi/lib.c                      |   5 +
 xen/arch/x86/cpu/mcheck/mce_intel.c          |   6 +
 xen/arch/x86/include/asm/cpufeature.h        |  12 +-
 xen/arch/x86/include/asm/msr-index.h         |  16 +-
 xen/drivers/acpi/pmstat.c                    | 100 +--
 xen/drivers/cpufreq/cpufreq.c                |  60 +-
 xen/drivers/cpufreq/cpufreq_misc_governors.c |   9 +
 xen/drivers/cpufreq/cpufreq_ondemand.c       |   3 +
 xen/drivers/cpufreq/utility.c                |   1 +
 xen/include/acpi/cpufreq/cpufreq.h           |  22 +
 xen/include/acpi/pdc_intel.h                 |   1 +
 xen/include/public/sysctl.h                  | 144 +++-
 20 files changed, 1432 insertions(+), 121 deletions(-)
 create mode 100644 xen/arch/x86/acpi/cpufreq/hwp.c
Signature · 1 line
2.41.0
#1 lore raw

PATCH v7 1/15 cpufreq: Allow restricting to internal governors only

For hwp, the standard governors are not usable, and only the internal one is applicable. Add the cpufreq_governor_internal boolean to indicate when an internal governor, like hwp, will be used. This is set during presmp_initcall, and governor registration can be skipped when called during initcall.

This way unusable governors are not registered, and only compatible governors are advertised to userspace.

---
v6:
Add Jan's Ack

v5:
Check cpufreq_governor_internal and skip registration as applicable Remove internal flag

v4:
Rework to use an internal flag
Removed Jan's Ack since the approach is different.

v3:
Switch to initdata
Add Jan Acked-by
Commit message s/they/the/ typo
Don't register hwp-internal when running non-hwp - Marek

v2:
Switch to "-internal"
Add blank line in header

---
 xen/drivers/cpufreq/cpufreq.c                | 1 +
 xen/drivers/cpufreq/cpufreq_misc_governors.c | 9 +++++++++
 xen/drivers/cpufreq/cpufreq_ondemand.c       | 3 +++
 xen/include/acpi/cpufreq/cpufreq.h           | 2 ++
 4 files changed, 15 insertions(+)
xen/drivers/cpufreq/cpufreq.c +1 -0
index 2321c7dd07..67a58d409b 100644
@@ -56,6 +56,7 @@ struct cpufreq_dom {
 };
 static LIST_HEAD_READ_MOSTLY(cpufreq_dom_list_head);
 
+bool __initdata cpufreq_governor_internal;
 struct cpufreq_governor *__read_mostly cpufreq_opt_governor;
 LIST_HEAD_READ_MOSTLY(cpufreq_governor_list);
 
xen/drivers/cpufreq/cpufreq_misc_governors.c +9 -0
index f5571f5486..0327fad23b 100644
@@ -120,6 +120,9 @@ static int __init cf_check cpufreq_gov_userspace_init(void)
 {
     unsigned int cpu;
 
+    if ( cpufreq_governor_internal )
+        return 0;
+
     for_each_online_cpu(cpu)
         per_cpu(cpu_set_freq, cpu) = userspace_cmdline_freq;
     register_cpu_notifier(&cpufreq_userspace_cpu_nfb);
@@ -162,6 +165,9 @@ struct cpufreq_governor cpufreq_gov_performance = {
 
 static int __init cf_check cpufreq_gov_performance_init(void)
 {
+    if ( cpufreq_governor_internal )
+        return 0;
+
     return cpufreq_register_governor(&cpufreq_gov_performance);
 }
 __initcall(cpufreq_gov_performance_init);
@@ -201,6 +207,9 @@ struct cpufreq_governor cpufreq_gov_powersave = {
 
 static int __init cf_check cpufreq_gov_powersave_init(void)
 {
+    if ( cpufreq_governor_internal )
+        return 0;
+
     return cpufreq_register_governor(&cpufreq_gov_powersave);
 }
 __initcall(cpufreq_gov_powersave_init);
xen/drivers/cpufreq/cpufreq_ondemand.c +3 -0
index fbcd14d6c3..06cfc88d30 100644
@@ -360,6 +360,9 @@ struct cpufreq_governor cpufreq_gov_dbs = {
 
 static int __init cf_check cpufreq_gov_dbs_init(void)
 {
+    if ( cpufreq_governor_internal )
+        return 0;
+
     return cpufreq_register_governor(&cpufreq_gov_dbs);
 }
 __initcall(cpufreq_gov_dbs_init);
xen/include/acpi/cpufreq/cpufreq.h +2 -0
index 35dcf21e8f..44fc4c58fc 100644
@@ -116,6 +116,8 @@ extern struct cpufreq_governor cpufreq_gov_powersave;
 
 extern struct list_head cpufreq_governor_list;
 
+extern bool cpufreq_governor_internal;
+
 extern int cpufreq_register_governor(struct cpufreq_governor *governor);
 extern struct cpufreq_governor *__find_governor(const char *governor);
 #define CPUFREQ_DEFAULT_GOVERNOR &cpufreq_gov_dbs
Signature · 1 line
2.41.0
#2 lore raw

PATCH v7 2/15 cpufreq: Add perf_freq to cpuinfo

In reply to Jason Andryuk

acpi-cpufreq scales the aperf/mperf measurements by max_freq, but HWP needs to scale by base frequency. Settings max_freq to base_freq "works" but the code is not obvious, and returning values to userspace is tricky. Add an additonal perf_freq member which is used for scaling aperf/mperf measurements.

---
v3:
Add Jan's Ack

I don't like this, but it seems the best way to re-use the common aperf/mperf code. The other option would be to add wrappers that then do the acpi vs. hwp scaling.

---
 xen/arch/x86/acpi/cpufreq/cpufreq.c | 2 +-
 xen/drivers/cpufreq/utility.c       | 1 +
 xen/include/acpi/cpufreq/cpufreq.h  | 3 +++
 3 files changed, 5 insertions(+), 1 deletion(-)
xen/arch/x86/acpi/cpufreq/cpufreq.c +1 -1
index 2e0067fbe5..6c70d04395 100644
@@ -316,7 +316,7 @@ unsigned int get_measured_perf(unsigned int cpu, unsigned int flag)
     else
         perf_percent = 0;
 
-    return policy->cpuinfo.max_freq * perf_percent / 100;
+    return policy->cpuinfo.perf_freq * perf_percent / 100;
 }
 
 static unsigned int cf_check get_cur_freq_on_cpu(unsigned int cpu)
xen/drivers/cpufreq/utility.c +1 -0
index 9eb7ecedcd..6831f62851 100644
@@ -236,6 +236,7 @@ int cpufreq_frequency_table_cpuinfo(struct cpufreq_policy *policy,
 
     policy->min = policy->cpuinfo.min_freq = min_freq;
     policy->max = policy->cpuinfo.max_freq = max_freq;
+    policy->cpuinfo.perf_freq = max_freq;
     policy->cpuinfo.second_max_freq = second_max_freq;
 
     if (policy->min == ~0)
xen/include/acpi/cpufreq/cpufreq.h +3 -0
index 44fc4c58fc..1f1898d811 100644
@@ -37,6 +37,9 @@ extern struct acpi_cpufreq_data *cpufreq_drv_data[NR_CPUS];
 struct cpufreq_cpuinfo {
     unsigned int        max_freq;
     unsigned int        second_max_freq;    /* P1 if Turbo Mode is on */
+    unsigned int        perf_freq; /* Scaling freq for aperf/mpref.
+                                      acpi-cpufreq uses max_freq, but HWP uses
+                                      base_freq.*/
     unsigned int        min_freq;
     unsigned int        transition_latency; /* in 10^(-9) s = nanoseconds */
 };
Signature · 1 line
2.41.0
#3 lore raw

PATCH v7 3/15 cpufreq: Export intel_feature_detect

In reply to Jason Andryuk

Export feature_detect as intel_feature_detect so it can be re-used by HWP.

---
v4:
Add Jan's Ack

v3:
Remove void * cast when calling intel_feature_detect

v2:
export intel_feature_detect with typed pointer
Move intel_feature_detect to acpi/cpufreq/cpufreq.h since the declaration now contains struct cpufreq_policy *.

---
 xen/arch/x86/acpi/cpufreq/cpufreq.c | 8 ++++++--
 xen/include/acpi/cpufreq/cpufreq.h  | 2 ++
 2 files changed, 8 insertions(+), 2 deletions(-)
xen/arch/x86/acpi/cpufreq/cpufreq.c +6 -2
index 6c70d04395..f1cc473b4f 100644
@@ -339,9 +339,8 @@ static unsigned int cf_check get_cur_freq_on_cpu(unsigned int cpu)
     return extract_freq(get_cur_val(cpumask_of(cpu)), data);
 }
 
-static void cf_check feature_detect(void *info)
+void intel_feature_detect(struct cpufreq_policy *policy)
 {
-    struct cpufreq_policy *policy = info;
     unsigned int eax;
 
     eax = cpuid_eax(6);
@@ -353,6 +352,11 @@ static void cf_check feature_detect(void *info)
     }
 }
 
+static void cf_check feature_detect(void *info)
+{
+    intel_feature_detect(info);
+}
+
 static unsigned int check_freqs(const cpumask_t *mask, unsigned int freq,
                                 struct acpi_cpufreq_data *data)
 {
xen/include/acpi/cpufreq/cpufreq.h +2 -0
index 1f1898d811..482ea5b0de 100644
@@ -243,4 +243,6 @@ int write_userspace_scaling_setspeed(unsigned int cpu, unsigned int freq);
 void cpufreq_dbs_timer_suspend(void);
 void cpufreq_dbs_timer_resume(void);
 
+void intel_feature_detect(struct cpufreq_policy *policy);
+
 #endif /* __XEN_CPUFREQ_PM_H__ */
Signature · 1 line
2.41.0
#4 lore raw

PATCH v7 4/15 xen/sysctl: Nest cpufreq scaling options

In reply to Jason Andryuk

Add a union and struct so that most of the scaling variables of struct xen_get_cpufreq_para are within in a binary-compatible layout. This allows cppc_para to live in the larger union and use uint32_ts - struct xen_cppc_para will be 10 uint32_t's.

The new scaling struct is 3 * uint32_t + 16 bytes CPUFREQ_NAME_LEN + 4 * uint32_t for xen_ondemand = 11 uint32_t. That means the old size is retained, int32_t turbo_enabled doesn't move and it's binary compatible.

The out-of-context memcpy() in xc_get_cpufreq_para() now handles the copying of the fields removed there.

---
NOTE: Jan would like a toolside review / ack because:
Nevertheless I continue to be uncertain about all of this: Parts of the struct can apparently go out of sync with the sysctl struct, but other parts have to remain in sync without there being an appropriate build-time check (checking merely sizes clearly isn't enough). Therefore I'd really like to have a toolstack side review / ack here as well.

v6:
Add Jan's Reviewed-by

v5:
Expand commit message
Change comment to driver/governor

---
 tools/include/xenctrl.h     | 22 +++++++++++++---------
 tools/libs/ctrl/xc_pm.c     |  7 +------
 tools/misc/xenpm.c          | 24 ++++++++++++------------
 xen/drivers/acpi/pmstat.c   | 27 ++++++++++++++-------------
 xen/include/public/sysctl.h | 22 +++++++++++++---------
 5 files changed, 53 insertions(+), 49 deletions(-)
tools/include/xenctrl.h +13 -9
index dba33d5d0f..8aedb952a0 100644
@@ -1909,16 +1909,20 @@ struct xc_get_cpufreq_para {
     uint32_t cpuinfo_cur_freq;
     uint32_t cpuinfo_max_freq;
     uint32_t cpuinfo_min_freq;
-    uint32_t scaling_cur_freq;
-
-    char scaling_governor[CPUFREQ_NAME_LEN];
-    uint32_t scaling_max_freq;
-    uint32_t scaling_min_freq;
-
-    /* for specific governor */
     union {
-        xc_userspace_t userspace;
-        xc_ondemand_t ondemand;
+        struct {
+            uint32_t scaling_cur_freq;
+
+            char scaling_governor[CPUFREQ_NAME_LEN];
+            uint32_t scaling_max_freq;
+            uint32_t scaling_min_freq;
+
+            /* for specific governor */
+            union {
+                xc_userspace_t userspace;
+                xc_ondemand_t ondemand;
+            } u;
+        } s;
     } u;
 
     int32_t turbo_enabled;
tools/libs/ctrl/xc_pm.c +1 -6
index c3a9864bf7..6e751e242f 100644
@@ -265,17 +265,12 @@ int xc_get_cpufreq_para(xc_interface *xch, int cpuid,
         user_para->cpuinfo_cur_freq = sys_para->cpuinfo_cur_freq;
         user_para->cpuinfo_max_freq = sys_para->cpuinfo_max_freq;
         user_para->cpuinfo_min_freq = sys_para->cpuinfo_min_freq;
-        user_para->scaling_cur_freq = sys_para->scaling_cur_freq;
-        user_para->scaling_max_freq = sys_para->scaling_max_freq;
-        user_para->scaling_min_freq = sys_para->scaling_min_freq;
         user_para->turbo_enabled    = sys_para->turbo_enabled;
 
         memcpy(user_para->scaling_driver,
                 sys_para->scaling_driver, CPUFREQ_NAME_LEN);
-        memcpy(user_para->scaling_governor,
-                sys_para->scaling_governor, CPUFREQ_NAME_LEN);
 
-        /* copy to user_para no matter what cpufreq governor */
+        /* copy to user_para no matter what cpufreq driver/governor */
         BUILD_BUG_ON(sizeof(((struct xc_get_cpufreq_para *)0)->u) !=
 		     sizeof(((struct xen_get_cpufreq_para *)0)->u));
 
tools/misc/xenpm.c +12 -12
index 1bb6187e56..ee8ce5d5f2 100644
@@ -730,39 +730,39 @@ static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
     printf("scaling_avail_gov    : %s\n",
            p_cpufreq->scaling_available_governors);
 
-    printf("current_governor     : %s\n", p_cpufreq->scaling_governor);
-    if ( !strncmp(p_cpufreq->scaling_governor,
+    printf("current_governor     : %s\n", p_cpufreq->u.s.scaling_governor);
+    if ( !strncmp(p_cpufreq->u.s.scaling_governor,
                   "userspace", CPUFREQ_NAME_LEN) )
     {
         printf("  userspace specific :\n");
         printf("    scaling_setspeed : %u\n",
-               p_cpufreq->u.userspace.scaling_setspeed);
+               p_cpufreq->u.s.u.userspace.scaling_setspeed);
     }
-    else if ( !strncmp(p_cpufreq->scaling_governor,
+    else if ( !strncmp(p_cpufreq->u.s.scaling_governor,
                        "ondemand", CPUFREQ_NAME_LEN) )
     {
         printf("  ondemand specific  :\n");
         printf("    sampling_rate    : max [%u] min [%u] cur [%u]\n",
-               p_cpufreq->u.ondemand.sampling_rate_max,
-               p_cpufreq->u.ondemand.sampling_rate_min,
-               p_cpufreq->u.ondemand.sampling_rate);
+               p_cpufreq->u.s.u.ondemand.sampling_rate_max,
+               p_cpufreq->u.s.u.ondemand.sampling_rate_min,
+               p_cpufreq->u.s.u.ondemand.sampling_rate);
         printf("    up_threshold     : %u\n",
-               p_cpufreq->u.ondemand.up_threshold);
+               p_cpufreq->u.s.u.ondemand.up_threshold);
     }
 
     printf("scaling_avail_freq   :");
     for ( i = 0; i < p_cpufreq->freq_num; i++ )
         if ( p_cpufreq->scaling_available_frequencies[i] ==
-             p_cpufreq->scaling_cur_freq )
+             p_cpufreq->u.s.scaling_cur_freq )
             printf(" *%d", p_cpufreq->scaling_available_frequencies[i]);
         else
             printf(" %d", p_cpufreq->scaling_available_frequencies[i]);
     printf("\n");
 
     printf("scaling frequency    : max [%u] min [%u] cur [%u]\n",
-           p_cpufreq->scaling_max_freq,
-           p_cpufreq->scaling_min_freq,
-           p_cpufreq->scaling_cur_freq);
+           p_cpufreq->u.s.scaling_max_freq,
+           p_cpufreq->u.s.scaling_min_freq,
+           p_cpufreq->u.s.scaling_cur_freq);
 
     printf("turbo mode           : %s\n",
            p_cpufreq->turbo_enabled ? "enabled" : "disabled or n/a");
xen/drivers/acpi/pmstat.c +14 -13
index 1bae635101..f5a9ac3f1a 100644
@@ -258,37 +258,38 @@ static int get_cpufreq_para(struct xen_sysctl_pm_op *op)
         cpufreq_driver.get ? cpufreq_driver.get(op->cpuid) : policy->cur;
     op->u.get_para.cpuinfo_max_freq = policy->cpuinfo.max_freq;
     op->u.get_para.cpuinfo_min_freq = policy->cpuinfo.min_freq;
-    op->u.get_para.scaling_cur_freq = policy->cur;
-    op->u.get_para.scaling_max_freq = policy->max;
-    op->u.get_para.scaling_min_freq = policy->min;
+
+    op->u.get_para.u.s.scaling_cur_freq = policy->cur;
+    op->u.get_para.u.s.scaling_max_freq = policy->max;
+    op->u.get_para.u.s.scaling_min_freq = policy->min;
 
     if ( cpufreq_driver.name[0] )
-        strlcpy(op->u.get_para.scaling_driver, 
+        strlcpy(op->u.get_para.scaling_driver,
             cpufreq_driver.name, CPUFREQ_NAME_LEN);
     else
         strlcpy(op->u.get_para.scaling_driver, "Unknown", CPUFREQ_NAME_LEN);
 
     if ( policy->governor->name[0] )
-        strlcpy(op->u.get_para.scaling_governor, 
+        strlcpy(op->u.get_para.u.s.scaling_governor,
             policy->governor->name, CPUFREQ_NAME_LEN);
     else
-        strlcpy(op->u.get_para.scaling_governor, "Unknown", CPUFREQ_NAME_LEN);
+        strlcpy(op->u.get_para.u.s.scaling_governor, "Unknown", CPUFREQ_NAME_LEN);
 
     /* governor specific para */
-    if ( !strncasecmp(op->u.get_para.scaling_governor,
+    if ( !strncasecmp(op->u.get_para.u.s.scaling_governor,
                       "userspace", CPUFREQ_NAME_LEN) )
     {
-        op->u.get_para.u.userspace.scaling_setspeed = policy->cur;
+        op->u.get_para.u.s.u.userspace.scaling_setspeed = policy->cur;
     }
 
-    if ( !strncasecmp(op->u.get_para.scaling_governor,
+    if ( !strncasecmp(op->u.get_para.u.s.scaling_governor,
                       "ondemand", CPUFREQ_NAME_LEN) )
     {
         ret = get_cpufreq_ondemand_para(
-            &op->u.get_para.u.ondemand.sampling_rate_max,
-            &op->u.get_para.u.ondemand.sampling_rate_min,
-            &op->u.get_para.u.ondemand.sampling_rate,
-            &op->u.get_para.u.ondemand.up_threshold);
+            &op->u.get_para.u.s.u.ondemand.sampling_rate_max,
+            &op->u.get_para.u.s.u.ondemand.sampling_rate_min,
+            &op->u.get_para.u.s.u.ondemand.sampling_rate,
+            &op->u.get_para.u.s.u.ondemand.up_threshold);
     }
     op->u.get_para.turbo_enabled = cpufreq_get_turbo_status(op->cpuid);
 
xen/include/public/sysctl.h +13 -9
index fa7147de47..c11c0b1a6c 100644
@@ -317,16 +317,20 @@ struct xen_get_cpufreq_para {
     uint32_t cpuinfo_cur_freq;
     uint32_t cpuinfo_max_freq;
     uint32_t cpuinfo_min_freq;
-    uint32_t scaling_cur_freq;
-
-    char scaling_governor[CPUFREQ_NAME_LEN];
-    uint32_t scaling_max_freq;
-    uint32_t scaling_min_freq;
-
-    /* for specific governor */
     union {
-        struct  xen_userspace userspace;
-        struct  xen_ondemand ondemand;
+        struct {
+            uint32_t scaling_cur_freq;
+
+            char scaling_governor[CPUFREQ_NAME_LEN];
+            uint32_t scaling_max_freq;
+            uint32_t scaling_min_freq;
+
+            /* for specific governor */
+            union {
+                struct  xen_userspace userspace;
+                struct  xen_ondemand ondemand;
+            } u;
+        } s;
     } u;
 
     int32_t turbo_enabled;
Signature · 1 line
2.41.0
Acked-by on v7 4/15
#5 lore raw

Re: [PATCH v7 04/15] xen/sysctl: Nest cpufreq scaling options

Quoted 14 lines · Jason Andryuk

Add a union and struct so that most of the scaling variables of struct xen_get_cpufreq_para are within in a binary-compatible layout. This allows cppc_para to live in the larger union and use uint32_ts - struct xen_cppc_para will be 10 uint32_t's.

The new scaling struct is 3 * uint32_t + 16 bytes CPUFREQ_NAME_LEN + 4 * uint32_t for xen_ondemand = 11 uint32_t. That means the old size is retained, int32_t turbo_enabled doesn't move and it's binary compatible.

The out-of-context memcpy() in xc_get_cpufreq_para() now handles the copying of the fields removed there.

Quoted 8 lines

---
NOTE: Jan would like a toolside review / ack because:
Nevertheless I continue to be uncertain about all of this: Parts of the struct can apparently go out of sync with the sysctl struct, but other parts have to remain in sync without there being an appropriate build-time check (checking merely sizes clearly isn't enough). Therefore I'd really like to have a toolstack side review / ack here as well.

I wish we could just use "struct xen_get_cpufreq_para" instead of having to make a copy to replace the XEN_GUEST_HANDLE_*() macro.

Next I guess we could try to have something like the compat layer in xen is doing, with plenty of CHECK_FIELD_ and other CHECK_* macro, but that would be a lot of work. (xen/include/xen/compat.h and how it's used in xen/include/compat/xlat.h)

Unless you feel like adding more build check, I guess the patch is good enough like that:

Thanks,

Signature · 1 line
Anthony PERARD
#6 lore raw

Re: [PATCH v7 04/15] xen/sysctl: Nest cpufreq scaling options

Quoted 25 lines · Anthony PERARD
Quoted 14 lines · Jason Andryuk

Add a union and struct so that most of the scaling variables of struct xen_get_cpufreq_para are within in a binary-compatible layout. This allows cppc_para to live in the larger union and use uint32_ts - struct xen_cppc_para will be 10 uint32_t's.

The new scaling struct is 3 * uint32_t + 16 bytes CPUFREQ_NAME_LEN + 4 * uint32_t for xen_ondemand = 11 uint32_t. That means the old size is retained, int32_t turbo_enabled doesn't move and it's binary compatible.

The out-of-context memcpy() in xc_get_cpufreq_para() now handles the copying of the fields removed there.

Quoted 8 lines

---
NOTE: Jan would like a toolside review / ack because:
Nevertheless I continue to be uncertain about all of this: Parts of the struct can apparently go out of sync with the sysctl struct, but other parts have to remain in sync without there being an appropriate build-time check (checking merely sizes clearly isn't enough). Therefore I'd really like to have a toolstack side review / ack here as well.

Quoted 7 lines

I wish we could just use "struct xen_get_cpufreq_para" instead of having to make a copy to replace the XEN_GUEST_HANDLE_*() macro.

Next I guess we could try to have something like the compat layer in xen is doing, with plenty of CHECK_FIELD_ and other CHECK_* macro, but that would be a lot of work. (xen/include/xen/compat.h and how it's used in xen/include/compat/xlat.h)

I can add a set of BUILD_BUG_ON()s checking the offsets of the two structs' members. I think that would work and we don't need the complication of the compat code. The build of the library just deals with a single bit-width and needs to be consistent. The hypervisor needs to deal with receiving differing pointer sizes and layouts, but the userspace library just uses whatever it was compiled for. The preprocessor expands XEN_GUEST_HANDLE_64(uint32) -> "typedef struct { uint32_t *p; } __guest_handle_uint32", so it's just going to be a single pointer in size, which will match between the two.

Does that sound right, or am I missing something?

Quoted 3 lines

Unless you feel like adding more build check, I guess the patch is good enough like that:

If I am incorrect about the above... I'll just leave it as-is.

Thanks,
Jason

#7 lore raw

Re: [PATCH v7 04/15] xen/sysctl: Nest cpufreq scaling options

Quoted 30 lines · Jason Andryuk
Quoted 28 lines · Anthony PERARD
Quoted 14 lines · Jason Andryuk

Add a union and struct so that most of the scaling variables of struct xen_get_cpufreq_para are within in a binary-compatible layout. This allows cppc_para to live in the larger union and use uint32_ts - struct xen_cppc_para will be 10 uint32_t's.

The new scaling struct is 3 * uint32_t + 16 bytes CPUFREQ_NAME_LEN + 4 * uint32_t for xen_ondemand = 11 uint32_t. That means the old size is retained, int32_t turbo_enabled doesn't move and it's binary compatible.

The out-of-context memcpy() in xc_get_cpufreq_para() now handles the copying of the fields removed there.

Quoted 8 lines

---
NOTE: Jan would like a toolside review / ack because:
Nevertheless I continue to be uncertain about all of this: Parts of the struct can apparently go out of sync with the sysctl struct, but other parts have to remain in sync without there being an appropriate build-time check (checking merely sizes clearly isn't enough). Therefore I'd really like to have a toolstack side review / ack here as well.

I wish we could just use "struct xen_get_cpufreq_para" instead of having to make a copy to replace the XEN_GUEST_HANDLE_*() macro.

Quoted 7 lines
Quoted 4 lines

Next I guess we could try to have something like the compat layer in xen is doing, with plenty of CHECK_FIELD_ and other CHECK_* macro, but that would be a lot of work. (xen/include/xen/compat.h and how it's used in xen/include/compat/xlat.h)

I can add a set of BUILD_BUG_ON()s checking the offsets of the two structs' members.

Yes, that would be fine.

Quoted 10 lines

I think that would work and we don't need the
complication of the compat code. The build of the library just deals with a single bit-width and needs to be consistent. The hypervisor needs to deal with receiving differing pointer sizes and layouts, but the userspace library just uses whatever it was compiled for. The preprocessor expands XEN_GUEST_HANDLE_64(uint32) -> "typedef struct { uint32_t *p; } __guest_handle_uint32", so it's just going to be a single pointer in size, which will match between the two.

Does that sound right, or am I missing something?

That sounds about right.

Thanks,

Signature · 1 line
Anthony PERARD
#8 lore raw

PATCH v7 5/15 pmstat&xenpm: Re-arrage for cpufreq union

In reply to Jason Andryuk

Rearrange code now that xen_sysctl_pm_op's get_para fields has the nested union and struct. In particular, the scaling governor information like scaling_available_governors is inside the union, so it is not always available. Move those fields (op->u.get_para.u.s.u.*) together as well as the common fields (ones outside the union like op->u.get_para.turbo_enabled).

With that, gov_num may be 0, so bounce buffer handling needs to be modified.

scaling_governor and other fields inside op->u.get_para.u.s.u.* won't be used for hwp, so this will simplify the change when hwp support is introduced and re-indents these lines all together.

---
v6:
Add Jan's Ack

v5:
Remove some formatting changes
Expand commit message

---
 tools/libs/ctrl/xc_pm.c   | 12 ++++++++----
 tools/misc/xenpm.c        |  3 ++-
 xen/drivers/acpi/pmstat.c | 24 ++++++++++++------------
 3 files changed, 22 insertions(+), 17 deletions(-)
tools/libs/ctrl/xc_pm.c +8 -4
index 6e751e242f..cea3eab22e 100644
@@ -221,7 +221,7 @@ int xc_get_cpufreq_para(xc_interface *xch, int cpuid,
     {
         if ( (!user_para->affected_cpus)                    ||
              (!user_para->scaling_available_frequencies)    ||
-             (!user_para->scaling_available_governors) )
+             (user_para->gov_num && !user_para->scaling_available_governors) )
         {
             errno = EINVAL;
             return -1;
@@ -230,12 +230,15 @@ int xc_get_cpufreq_para(xc_interface *xch, int cpuid,
             goto unlock_1;
         if ( xc_hypercall_bounce_pre(xch, scaling_available_frequencies) )
             goto unlock_2;
-        if ( xc_hypercall_bounce_pre(xch, scaling_available_governors) )
+        if ( user_para->gov_num &&
+             xc_hypercall_bounce_pre(xch, scaling_available_governors) )
             goto unlock_3;
 
         set_xen_guest_handle(sys_para->affected_cpus, affected_cpus);
         set_xen_guest_handle(sys_para->scaling_available_frequencies, scaling_available_frequencies);
-        set_xen_guest_handle(sys_para->scaling_available_governors, scaling_available_governors);
+        if ( user_para->gov_num )
+            set_xen_guest_handle(sys_para->scaling_available_governors,
+                                 scaling_available_governors);
     }
 
     sysctl.cmd = XEN_SYSCTL_pm_op;
@@ -278,7 +281,8 @@ int xc_get_cpufreq_para(xc_interface *xch, int cpuid,
     }
 
 unlock_4:
-    xc_hypercall_bounce_post(xch, scaling_available_governors);
+    if ( user_para->gov_num )
+        xc_hypercall_bounce_post(xch, scaling_available_governors);
 unlock_3:
     xc_hypercall_bounce_post(xch, scaling_available_frequencies);
 unlock_2:
tools/misc/xenpm.c +2 -1
index ee8ce5d5f2..1c474c3b59 100644
@@ -811,7 +811,8 @@ static int show_cpufreq_para_by_cpuid(xc_interface *xc_handle, int cpuid)
             ret = -ENOMEM;
             goto out;
         }
-        if (!(p_cpufreq->scaling_available_governors =
+        if (p_cpufreq->gov_num &&
+            !(p_cpufreq->scaling_available_governors =
               malloc(p_cpufreq->gov_num * CPUFREQ_NAME_LEN * sizeof(char))))
         {
             fprintf(stderr,
xen/drivers/acpi/pmstat.c +12 -12
index f5a9ac3f1a..d67d99e62f 100644
@@ -239,6 +239,18 @@ static int get_cpufreq_para(struct xen_sysctl_pm_op *op)
     if ( ret )
         return ret;
 
+    op->u.get_para.cpuinfo_cur_freq =
+        cpufreq_driver.get ? cpufreq_driver.get(op->cpuid) : policy->cur;
+    op->u.get_para.cpuinfo_max_freq = policy->cpuinfo.max_freq;
+    op->u.get_para.cpuinfo_min_freq = policy->cpuinfo.min_freq;
+    op->u.get_para.turbo_enabled = cpufreq_get_turbo_status(op->cpuid);
+
+    if ( cpufreq_driver.name[0] )
+        strlcpy(op->u.get_para.scaling_driver,
+            cpufreq_driver.name, CPUFREQ_NAME_LEN);
+    else
+        strlcpy(op->u.get_para.scaling_driver, "Unknown", CPUFREQ_NAME_LEN);
+
     if ( !(scaling_available_governors =
            xzalloc_array(char, gov_num * CPUFREQ_NAME_LEN)) )
         return -ENOMEM;
@@ -254,21 +266,10 @@ static int get_cpufreq_para(struct xen_sysctl_pm_op *op)
     if ( ret )
         return ret;
 
-    op->u.get_para.cpuinfo_cur_freq =
-        cpufreq_driver.get ? cpufreq_driver.get(op->cpuid) : policy->cur;
-    op->u.get_para.cpuinfo_max_freq = policy->cpuinfo.max_freq;
-    op->u.get_para.cpuinfo_min_freq = policy->cpuinfo.min_freq;
-
     op->u.get_para.u.s.scaling_cur_freq = policy->cur;
     op->u.get_para.u.s.scaling_max_freq = policy->max;
     op->u.get_para.u.s.scaling_min_freq = policy->min;
 
-    if ( cpufreq_driver.name[0] )
-        strlcpy(op->u.get_para.scaling_driver,
-            cpufreq_driver.name, CPUFREQ_NAME_LEN);
-    else
-        strlcpy(op->u.get_para.scaling_driver, "Unknown", CPUFREQ_NAME_LEN);
-
     if ( policy->governor->name[0] )
         strlcpy(op->u.get_para.u.s.scaling_governor,
             policy->governor->name, CPUFREQ_NAME_LEN);
@@ -291,7 +292,6 @@ static int get_cpufreq_para(struct xen_sysctl_pm_op *op)
             &op->u.get_para.u.s.u.ondemand.sampling_rate,
             &op->u.get_para.u.s.u.ondemand.up_threshold);
     }
-    op->u.get_para.turbo_enabled = cpufreq_get_turbo_status(op->cpuid);
 
     return ret;
 }
Signature · 1 line
2.41.0
Acked-by on v7 5/15
#9 lore raw

Re: [PATCH v7 05/15] pmstat&xenpm: Re-arrage for cpufreq union

Quoted 10 lines · Jason Andryuk

Rearrange code now that xen_sysctl_pm_op's get_para fields has the nested union and struct. In particular, the scaling governor information like scaling_available_governors is inside the union, so it is not always available. Move those fields (op->u.get_para.u.s.u.*) together as well as the common fields (ones outside the union like op->u.get_para.turbo_enabled).

With that, gov_num may be 0, so bounce buffer handling needs to be modified.

Quoted 6 lines

scaling_governor and other fields inside op->u.get_para.u.s.u.* won't be used for hwp, so this will simplify the change when hwp support is introduced and re-indents these lines all together.

Thanks,

Signature · 1 line
Anthony PERARD
#10 lore raw

PATCH v7 6/15 cpufreq: Add Hardware P-State (HWP) driver

In reply to Jason Andryuk

From the Intel SDM: "Hardware-Controlled Performance States (HWP), which autonomously selects performance states while utilizing OS supplied performance guidance hints."

Enable HWP to run in autonomous mode by poking the correct MSRs. HWP is disabled by default, and cpufreq=hwp enables it.

cpufreq= parsing is expanded to allow cpufreq=hwp;xen. This allows trying HWP and falling back to xen if not available. Only hwp and xen are supported for this fallback feature. hdc is a sub-option under hwp (i.e. cpufreq=hwp,hdc=0) as is verbose.

There is no interface to configure - xen_sysctl_pm_op/xenpm will be extended to configure in subsequent patches. It will run with the default values, which should be the default 0x80 (out of 0x0-0xff) energy/performance preference.

Unscientific powertop measurement of an mostly idle, customized OpenXT install:
A 10th gen 6-core laptop showed battery discharge drop from ~9.x to ~7.x watts.
A 8th gen 4-core laptop dropped from ~10 to ~9

Power usage depends on many factors, especially display brightness, but this does show a power saving in balanced mode when CPU utilization is low.

HWP isn't compatible with an external governor - it doesn't take explicit frequency requests. Therefore a minimal internal governor, hwp, is also added as a placeholder.

While adding to the xen-command-line.pandoc entry, un-nest verbose from minfreq. They are independent.

With cpufreq=hwp,verbose, HWP prints processor capabilities that are not used by the code, like HW_FEEDBACK. This is done because otherwise there isn't a convenient way to query the information.

Xen doesn't use the HWP interrupt, so it is disabled like in the Linux pstate driver.

---
We disable on cpuid_level < 0x16. cpuid(0x16) is used to get the cpu frequencies for calculating the APERF/MPERF. Without it, things would still work, but the average cpu frequency output would be wrong.

My 8th & 10th gen test systems both report:
(XEN) HWP: 1 notify: 1 act_window: 1 energy_perf: 1 pkg_level: 0 peci: 0 (XEN) HWP: Hardware Duty Cycling (HDC) supported
(XEN) HWP: HW_FEEDBACK not supported

We can't use parse_boolean() since it requires a single name=val string and cpufreq_handle_common_option is provided two strings. Use parse_bool() and manual handle no-hwp.

FAST_IA32_HWP_REQUEST was removed in v2. The check in v1 was wrong, it's a model specific feature and the CPUID bit is only available after enabling via the MSR. Support was untested since I don't have hardware with the feature. Writes are expected to be infrequent, so just leave it out.

---
v2:
Alphabetize headers
Re-work driver registration
name hwp_drv_data anonymous union "hw"
Drop hwp_verbose_cont
style cleanups
Condense hwp_governor switch
hwp_cpufreq_target remove .raw from hwp_req assignment Use typed-pointer in a few functions
Pass type to xzalloc
Add HWP_ENERGY_PERF_BALANCE/IA32_ENERGY_BIAS_BALANCE defines Add XEN_HWP_GOVERNOR define for "hwp-internal" Capitalize CPUID and MSR defines
Change '_' to '-' for energy-perf & act-window Read-modify-write MSRs updates
Use FAST_IA32_HWP_REQUEST_MSR_ENABLE define
constify pointer in hwp_set_misc_turbo
Add space after non-fallthrough break in governor switch Add IA32_ENERGY_BIAS_MASK define
Check CPUID_PM_LEAK for energy bias when needed
Fail initialization with curr_req = -1
Fold hwp_read_capabilities into hwp_init_msrs
Add command line cpufreq=xen:hwp
Add command line cpufreq=xen:hdc
Use per_cpu for hwp_drv_data pointers
Move hwp_energy_perf_bias call into hwp_write_request energy_perf 0 is valid, so hwp_energy_perf_bias cannot be skipped Ensure we don't generate interrupts
Remove Fast Write of Uncore MSR
Initialize hwp_drv_data from curr_req
Use SPDX line instead of license text in hwp.c

v3:
Add cf_check to cpufreq_gov_hwp_init() - Marek
Print cpuid_level with %#x - Marek

v4:
Use BIT() for CPUID and MSR bits
Move __initdata after type
Add __ro_after_init to feature_*
Remove aperf/mperf comment
Move feature_hwp_energy_perf { to newline
Remove _IA32_ infix
Use unsigned int & bool for bitfields
Require energy perf pref (Remove ENERGY_PERF_BIAS support) Initialize activity_window
Return errors on wrmsr failure
Change command line to: cpufreq=xen:hwp
Move hdc into the hwp-specific handle_options
Drop feature_hwp_energy_perf, feature_hwp_pkg_level_ctl & feature_hwp_peci Print features before exiting when energy/performance preference isn't available Disable HWP MSR on initialization error
Change hwp_ print macros to add prefixes
Disable HDC when hdc=0 - (opt_hdc no longer initdata)
Mark hwp governor internal and use "hwp" name
Add XEN_HWP_DRIVER
Use top-level cpufreq=hwp command line option
Document that cpufreq=hwp falls back to cpufreq=xen without hardware Add SPDX suffix GPL-2.0-only

v5:
Use _AC() macro in MSR_IA32_MISC_ENABLE_TURBO_DISENGAGE definition hwp_err arg re-ordering
Use XEN_HWP_DRIVER_NAME
Use cpufreq.h for all declarations
Clear feature_hdc on failure and print a message
Use unnamed bitfields instead of reservered
Remove asm/io.h include
static DEFINE_PER_CPU_READ_MOSTLY(struct hwp_drv_data Remove some empty newlines
Align feature_hdc assignment
Remove feature_hwp
Remove unnecesary return at end of void hwp_init_msrs() BUILD_BUG_ON member variable
Reformat a compound if
Clear pre_cpu hwp_drv_data before xfree()
Only print HWP capabilities for CPU 0
Specify processor models in turbo comment
Use arg[1] in setup_cpufreq_option()
Remove some log messages
Drop double newline
Parse verbose as a boolean instead of the custom parsing. Support cpufreq=hwp;xen fallback
Call hwp_available() from hwp_register_driver()
Move cpufreq_govenor_internal setting to hwp_register_driver

v6:
docs: mention verbose applies to both.
Use cpufreq=hwp;xen,verbose example with verbose at then end. Declare induction i inside for loop
Add CPUFREQ_none
Only call cpufreq fallback for -ENODEV
Brace on newline fixes
hdc_set_pkg_hdc_ctl()/hdc_set_pm_ctl1() return boolean
Continue without hdc
Consolidate hdc if statement
Use hwp_req variable for BUILD_BUG_ON
Newline between declarations and statement
__initdata for cpufreq_xen_opts/cpufreq_xen_cnt
Only print HWP_REQUEST when differing from BSP.
Remove MSR_PM_ENABLE printing
Move hdc_set_pkg_hdc_ctl(), hdc_set_pm_ctl1(), hwp_get_cpu_speeds(), and
hwp_init_msrs() to before their sole caller, hwp_cpufreq_cpu_init()
Fix cpufreq= parsing without options (set ret=0)
Drop stray s NULL check
Return -EINVAL from hwp_cmdline_parse()
Latch first CPU HWP request and print differences
Use strpbrk to split on ',' or ':' for arg in xen/hwp do..while Limit hwp parsing error to just current option with %.*s and (end ?: e) - s
docs: Expand and clarify the handling of verbose with cpufreq fallbacks.

v7:
s/##_VA_ARGS/## args/ in macros
Add __init to hwp_handle_option
Add comment for 0 cpuid freqs
Reword intel_feature_detect comment.
Fix indent for HDC failure
Make first_run static and use for first cpu latching hwp_set_misc_turbo() return -EACCES on error s/end/*end/ in setup_cpufreq_option()

---
 docs/misc/xen-command-line.pandoc     |  20 +-
 xen/arch/x86/acpi/cpufreq/Makefile    |   1 +
 xen/arch/x86/acpi/cpufreq/cpufreq.c   |  21 +-
 xen/arch/x86/acpi/cpufreq/hwp.c       | 524 ++++++++++++++++++++++++++
 xen/arch/x86/include/asm/cpufeature.h |  12 +-
 xen/arch/x86/include/asm/msr-index.h  |  15 +-
 xen/drivers/cpufreq/cpufreq.c         |  59 ++-
 xen/include/acpi/cpufreq/cpufreq.h    |  10 +
 xen/include/public/sysctl.h           |   2 +
 9 files changed, 649 insertions(+), 15 deletions(-)
 create mode 100644 xen/arch/x86/acpi/cpufreq/hwp.c
docs/misc/xen-command-line.pandoc +19 -1
index 4060ebdc5d..f77542de47 100644
@@ -499,7 +499,7 @@ If set, force use of the performance counters for oprofile, rather than detectin
 available support.
 
 ### cpufreq
-> `= none | {{ <boolean> | xen } [:[powersave|performance|ondemand|userspace][,<maxfreq>][,[<minfreq>][,[verbose]]]]} | dom0-kernel`
+> `= none | {{ <boolean> | xen } { [:[powersave|performance|ondemand|userspace][,[<maxfreq>]][,[<minfreq>]]] } [,verbose]} | dom0-kernel | hwp[:[<hdc>][,verbose]]`
 
 > Default: `xen`
 
@@ -510,6 +510,24 @@ choice of `dom0-kernel` is deprecated and not supported by all Dom0 kernels.
 * `<maxfreq>` and `<minfreq>` are integers which represent max and min processor frequencies
   respectively.
 * `verbose` option can be included as a string or also as `verbose=<integer>`
+  for `xen`.  It is a boolean for `hwp`.
+* `hwp` selects Hardware-Controlled Performance States (HWP) on supported Intel
+  hardware.  HWP is a Skylake+ feature which provides better CPU power
+  management.  The default is disabled.  If `hwp` is selected, but hardware
+  support is not available, Xen will fallback to cpufreq=xen.
+* `<hdc>` is a boolean to enable Hardware Duty Cycling (HDC).  HDC enables the
+  processor to autonomously force physical package components into idle state.
+  The default is enabled, but the option only applies when `hwp` is enabled.
+
+There is also support for `;`-separated fallback options:
+`cpufreq=hwp;xen,verbose`.  This first tries `hwp` and falls back to `xen` if
+unavailable.  Note: The `verbose` suboption is handled globally.  Setting it
+for either the primary or fallback option applies to both irrespective of where
+it is specified.
+
+Note: grub2 requires to escape or quote ';', so `"cpufreq=hwp;xen"` should be
+specified within double quotes inside grub.cfg.  Refer to the grub2
+documentation for more information.
 
 ### cpuid (x86)
 > `= List of comma separated booleans`
xen/arch/x86/acpi/cpufreq/Makefile +1 -0
index f75da9b9ca..db83aa6b14 100644
@@ -1,2 +1,3 @@
 obj-y += cpufreq.o
+obj-y += hwp.o
 obj-y += powernow.o
xen/arch/x86/acpi/cpufreq/cpufreq.c +20 -1
index f1cc473b4f..61b62c370a 100644
@@ -642,7 +642,26 @@ static int __init cf_check cpufreq_driver_init(void)
         switch ( boot_cpu_data.x86_vendor )
         {
         case X86_VENDOR_INTEL:
-            ret = cpufreq_register_driver(&acpi_cpufreq_driver);
+            ret = -ENOENT;
+
+            for ( unsigned int i = 0; i < cpufreq_xen_cnt; i++ )
+            {
+                switch ( cpufreq_xen_opts[i] )
+                {
+                case CPUFREQ_xen:
+                    ret = cpufreq_register_driver(&acpi_cpufreq_driver);
+                    break;
+                case CPUFREQ_hwp:
+                    ret = hwp_register_driver();
+                    break;
+                case CPUFREQ_none:
+                    ret = 0;
+                    break;
+                }
+
+                if ( ret != -ENODEV )
+                    break;
+            }
             break;
 
         case X86_VENDOR_AMD:
xen/arch/x86/acpi/cpufreq/hwp.c new file +524 -0

524 lines not on this page · raw message

xen/arch/x86/include/asm/cpufeature.h +10 -2
index 64e1dad225..93466441f5 100644
@@ -52,8 +52,16 @@ static inline bool boot_cpu_has(unsigned int feat)
     return cpu_has(&boot_cpu_data, feat);
 }
 
-#define CPUID_PM_LEAF                    6
-#define CPUID6_ECX_APERFMPERF_CAPABILITY 0x1
+#define CPUID_PM_LEAF                                6
+#define CPUID6_EAX_HWP                               BIT(7, U)
+#define CPUID6_EAX_HWP_NOTIFICATION                  BIT(8, U)
+#define CPUID6_EAX_HWP_ACTIVITY_WINDOW               BIT(9, U)
+#define CPUID6_EAX_HWP_ENERGY_PERFORMANCE_PREFERENCE BIT(10, U)
+#define CPUID6_EAX_HWP_PACKAGE_LEVEL_REQUEST         BIT(11, U)
+#define CPUID6_EAX_HDC                               BIT(13, U)
+#define CPUID6_EAX_HWP_PECI                          BIT(16, U)
+#define CPUID6_EAX_HW_FEEDBACK                       BIT(19, U)
+#define CPUID6_ECX_APERFMPERF_CAPABILITY             BIT(0, U)
 
 /* CPUID level 0x00000001.edx */
 #define cpu_has_fpu             1
xen/arch/x86/include/asm/msr-index.h +14 -1
index 2382fc8e11..56f93cb8e2 100644
@@ -151,6 +151,13 @@
 
 #define MSR_PKRS                            0x000006e1
 
+#define MSR_PM_ENABLE                       0x00000770
+#define  PM_ENABLE_HWP_ENABLE               BIT(0, ULL)
+
+#define MSR_HWP_CAPABILITIES                0x00000771
+#define MSR_HWP_INTERRUPT                   0x00000773
+#define MSR_HWP_REQUEST                     0x00000774
+
 #define MSR_X2APIC_FIRST                    0x00000800
 #define MSR_X2APIC_LAST                     0x000008ff
 
@@ -165,6 +172,11 @@
 #define  PASID_PASID_MASK                   0x000fffff
 #define  PASID_VALID                        (_AC(1, ULL) << 31)
 
+#define MSR_PKG_HDC_CTL                     0x00000db0
+#define  PKG_HDC_CTL_HDC_PKG_ENABLE         BIT(0, ULL)
+#define MSR_PM_CTL1                         0x00000db1
+#define  PM_CTL1_HDC_ALLOW_BLOCK            BIT(0, ULL)
+
 #define MSR_UARCH_MISC_CTRL                 0x00001b01
 #define  UARCH_CTRL_DOITM                   (_AC(1, ULL) <<  0)
 
@@ -466,7 +478,8 @@
 #define MSR_IA32_MISC_ENABLE_MONITOR_ENABLE (1<<18)
 #define MSR_IA32_MISC_ENABLE_LIMIT_CPUID  (1<<22)
 #define MSR_IA32_MISC_ENABLE_XTPR_DISABLE (1<<23)
-#define MSR_IA32_MISC_ENABLE_XD_DISABLE   (_AC(1, ULL) << 34)
+#define MSR_IA32_MISC_ENABLE_XD_DISABLE      (_AC(1, ULL) << 34)
+#define MSR_IA32_MISC_ENABLE_TURBO_DISENGAGE (_AC(1, ULL) << 38)
 
 #define MSR_IA32_TSC_DEADLINE		0x000006E0
 #define MSR_IA32_ENERGY_PERF_BIAS	0x000001b0
xen/drivers/cpufreq/cpufreq.c +49 -10

89 lines not on this page · raw message

xen/include/acpi/cpufreq/cpufreq.h +10 -0
index 482ea5b0de..cfe975814e 100644
@@ -24,6 +24,13 @@ DECLARE_PER_CPU(spinlock_t, cpufreq_statistic_lock);
 
 extern bool_t cpufreq_verbose;
 
+enum cpufreq_xen_opt {
+    CPUFREQ_none,
+    CPUFREQ_xen,
+    CPUFREQ_hwp,
+};
+extern enum cpufreq_xen_opt cpufreq_xen_opts[2];
+extern unsigned int cpufreq_xen_cnt;
 struct cpufreq_governor;
 
 struct acpi_cpufreq_data {
@@ -245,4 +252,7 @@ void cpufreq_dbs_timer_resume(void);
 
 void intel_feature_detect(struct cpufreq_policy *policy);
 
+int hwp_cmdline_parse(const char *s, const char *e);
+int hwp_register_driver(void);
+
 #endif /* __XEN_CPUFREQ_PM_H__ */
xen/include/public/sysctl.h +2 -0
index c11c0b1a6c..5a30b16fcf 100644
@@ -296,6 +296,8 @@ struct xen_ondemand {
     uint32_t up_threshold;
 };
 
+#define XEN_HWP_DRIVER_NAME "hwp"
+
 /*
  * cpufreq para name of this structure named
  * same as sysfs file name of native linux
Signature · 1 line
2.41.0
Reviewed-by on v7 6/15
#11 lore raw

Re: [PATCH v7 06/15] cpufreq: Add Hardware P-State (HWP) driver

Quoted 34 lines · Jason Andryuk

From the Intel SDM: "Hardware-Controlled Performance States (HWP), which autonomously selects performance states while utilizing OS supplied performance guidance hints."

Enable HWP to run in autonomous mode by poking the correct MSRs. HWP is disabled by default, and cpufreq=hwp enables it.

cpufreq= parsing is expanded to allow cpufreq=hwp;xen. This allows trying HWP and falling back to xen if not available. Only hwp and xen are supported for this fallback feature. hdc is a sub-option under hwp (i.e. cpufreq=hwp,hdc=0) as is verbose.

There is no interface to configure - xen_sysctl_pm_op/xenpm will be extended to configure in subsequent patches. It will run with the default values, which should be the default 0x80 (out of 0x0-0xff) energy/performance preference.

Unscientific powertop measurement of an mostly idle, customized OpenXT install:
A 10th gen 6-core laptop showed battery discharge drop from ~9.x to ~7.x watts.
A 8th gen 4-core laptop dropped from ~10 to ~9

Power usage depends on many factors, especially display brightness, but this does show a power saving in balanced mode when CPU utilization is low.

HWP isn't compatible with an external governor - it doesn't take explicit frequency requests. Therefore a minimal internal governor, hwp, is also added as a placeholder.

While adding to the xen-command-line.pandoc entry, un-nest verbose from minfreq. They are independent.

Quoted 8 lines

With cpufreq=hwp,verbose, HWP prints processor capabilities that are not used by the code, like HW_FEEDBACK. This is done because otherwise there isn't a convenient way to query the information.

Xen doesn't use the HWP interrupt, so it is disabled like in the Linux pstate driver.

xen/arch/x86/acpi/cpufreq/hwp.c line 408

Quoted 6 lines
xen/arch/x86/acpi/cpufreq/hwp.c new file +6 -0
+static int cf_check hwp_cpufreq_cpu_init(struct cpufreq_policy *policy)
+{
+    static union hwp_request initial_req;
+    unsigned int cpu = policy->cpu;
+    struct hwp_drv_data *data;
+    static bool first_run = true;

This would probably better be __read_mostly. Iirc there were some small adjustments requested elsewhere, so I guess I'll wait for v8 before checking to see how much of this series is ready to go in.

Jan

#12 lore raw

PATCH v7 7/15 xen/x86: Tweak PDC bits when using HWP

In reply to Jason Andryuk

Qubes testing of HWP support had a report of a laptop, Thinkpad X1 Carbon Gen 4 with a Skylake processor, locking up during boot when HWP is enabled. A user found a kernel bug that seems to be the same issue:
https://bugzilla.kernel.org/show_bug.cgi?id=110941.

That bug was fixed by Linux commit a21211672c9a ("ACPI / processor:
Request native thermal interrupt handling via _OSC"). The tl;dr is SMM crashes when it receives thermal interrupts, so Linux calls the ACPI _OSC method to take over interrupt handling.

The Linux fix looks at the CPU features to decide whether or not to call _OSC with bit 12 set to take over native interrupt handling. Xen needs some way to communicate HWP to Dom0 for making an equivalent call.

Xen exposes modified PDC bits via the platform_op set_pminfo hypercall. Expand that to set bit 12 when HWP is present and in use.

Any generated interrupt would be handled by Xen's thermal drive, which clears the status.

Bit 12 isn't named in the linux header and is open coded in Linux's usage. Name it ACPI_PDC_CPPC_NATIVE_INTR.

This will need a corresponding linux patch to pick up and apply the PDC bits.

---
v5:
Minor fixup for feature_hwp removal
Use cpurfreq.h for declaration

v4:
Added __ro_after_init s/ACPI_PDC_CPPC_NTV_INT/ACPI_PDC_CPPC_NATIVE_INTR/ Remove _IA32_
Fixup for opt_cpufreq_hwp removal
Add Jan Reviewed-by

v3:
New

---
 xen/arch/x86/acpi/cpufreq/hwp.c      | 9 +++++++++
 xen/arch/x86/acpi/lib.c              | 5 +++++
 xen/arch/x86/cpu/mcheck/mce_intel.c  | 6 ++++++
 xen/arch/x86/include/asm/msr-index.h | 1 +
 xen/include/acpi/cpufreq/cpufreq.h   | 1 +
 xen/include/acpi/pdc_intel.h         | 1 +
 6 files changed, 23 insertions(+)
xen/arch/x86/acpi/cpufreq/hwp.c +9 -0
index e9a6d23e38..33039ad39e 100644
@@ -12,6 +12,8 @@
 #include <asm/msr.h>
 #include <acpi/cpufreq/cpufreq.h>
 
+static bool __ro_after_init hwp_in_use;
+
 static bool __ro_after_init feature_hwp_notification;
 static bool __ro_after_init feature_hwp_activity_window;
 
@@ -152,6 +154,11 @@ static int __init cf_check cpufreq_gov_hwp_init(void)
 }
 __initcall(cpufreq_gov_hwp_init);
 
+bool hwp_active(void)
+{
+    return hwp_in_use;
+}
+
 static bool __init hwp_available(void)
 {
     unsigned int eax;
@@ -204,6 +211,8 @@ static bool __init hwp_available(void)
     hwp_verbose("HW_FEEDBACK %ssupported\n",
                 (eax & CPUID6_EAX_HW_FEEDBACK) ? "" : "not ");
 
+    hwp_in_use = true;
+
     hwp_info("Using HWP for cpufreq\n");
 
     return true;
xen/arch/x86/acpi/lib.c +5 -0
index 43831b92d1..51cb082ca0 100644
@@ -26,6 +26,8 @@
 #include <asm/fixmap.h>
 #include <asm/mwait.h>
 
+#include <acpi/cpufreq/cpufreq.h>
+
 u32 __read_mostly acpi_smi_cmd;
 u8 __read_mostly acpi_enable_value;
 u8 __read_mostly acpi_disable_value;
@@ -140,5 +142,8 @@ int arch_acpi_set_pdc_bits(u32 acpi_id, u32 *pdc, u32 mask)
 	    !(ecx & CPUID5_ECX_INTERRUPT_BREAK))
 		pdc[2] &= ~(ACPI_PDC_C_C1_FFH | ACPI_PDC_C_C2C3_FFH);
 
+	if (hwp_active())
+		pdc[2] |= ACPI_PDC_CPPC_NATIVE_INTR;
+
 	return 0;
 }
xen/arch/x86/cpu/mcheck/mce_intel.c +6 -0
index 2f23f02923..4045c6591d 100644
@@ -15,6 +15,9 @@
 #include <asm/p2m.h>
 #include <asm/mce.h>
 #include <asm/apic.h>
+
+#include <acpi/cpufreq/cpufreq.h>
+
 #include "mce.h"
 #include "x86_mca.h"
 #include "barrier.h"
@@ -64,6 +67,9 @@ static void cf_check intel_thermal_interrupt(struct cpu_user_regs *regs)
 
     ack_APIC_irq();
 
+    if ( hwp_active() )
+        wrmsr_safe(MSR_HWP_STATUS, 0);
+
     if ( NOW() < per_cpu(next, cpu) )
         return;
 
xen/arch/x86/include/asm/msr-index.h +1 -0
index 56f93cb8e2..f2410fc8c8 100644
@@ -157,6 +157,7 @@
 #define MSR_HWP_CAPABILITIES                0x00000771
 #define MSR_HWP_INTERRUPT                   0x00000773
 #define MSR_HWP_REQUEST                     0x00000774
+#define MSR_HWP_STATUS                      0x00000777
 
 #define MSR_X2APIC_FIRST                    0x00000800
 #define MSR_X2APIC_LAST                     0x000008ff
xen/include/acpi/cpufreq/cpufreq.h +1 -0
index cfe975814e..6737798e25 100644
@@ -254,5 +254,6 @@ void intel_feature_detect(struct cpufreq_policy *policy);
 
 int hwp_cmdline_parse(const char *s, const char *e);
 int hwp_register_driver(void);
+bool hwp_active(void);
 
 #endif /* __XEN_CPUFREQ_PM_H__ */
xen/include/acpi/pdc_intel.h +1 -0
index 4fb719d6f5..abaa098b51 100644
@@ -17,6 +17,7 @@
 #define ACPI_PDC_C_C1_FFH		(0x0100)
 #define ACPI_PDC_C_C2C3_FFH		(0x0200)
 #define ACPI_PDC_SMP_P_HWCOORD		(0x0800)
+#define ACPI_PDC_CPPC_NATIVE_INTR	(0x1000)
 
 #define ACPI_PDC_EST_CAPABILITY_SMP	(ACPI_PDC_SMP_C1PT | \
 					 ACPI_PDC_C_C1_HALT | \
Signature · 1 line
2.41.0
#13 lore raw

PATCH v7 8/15 xenpm: Change get-cpufreq-para output for hwp

In reply to Jason Andryuk

When using HWP, some of the returned data is not applicable. In that case, we should just omit it to avoid confusing the user. So switch to printing the base and max frequencies since those are relevant to HWP. Similarly, stop printing the CPU frequencies since those do not apply. The scaling fields are also no longer printed.

---
v5:
Use XEN_HWP_DRIVER_NAME
Add Jan's Ack

v4:
s/turbo/max/
Check for XEN_HWP_DRIVER driver instead of "-internal"

v2:
Use full governor name XEN_HWP_GOVERNOR to change output Style fixes

---
 tools/misc/xenpm.c | 83 +++++++++++++++++++++++++---------------------
 1 file changed, 46 insertions(+), 37 deletions(-)
tools/misc/xenpm.c +46 -37
index 1c474c3b59..21c93386de 100644
@@ -711,6 +711,7 @@ void start_gather_func(int argc, char *argv[])
 /* print out parameters about cpu frequency */
 static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
 {
+    bool hwp = strcmp(p_cpufreq->scaling_driver, XEN_HWP_DRIVER_NAME) == 0;
     int i;
 
     printf("cpu id               : %d\n", cpuid);
@@ -720,49 +721,57 @@ static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
         printf(" %d", p_cpufreq->affected_cpus[i]);
     printf("\n");
 
-    printf("cpuinfo frequency    : max [%u] min [%u] cur [%u]\n",
-           p_cpufreq->cpuinfo_max_freq,
-           p_cpufreq->cpuinfo_min_freq,
-           p_cpufreq->cpuinfo_cur_freq);
+    if ( hwp )
+        printf("cpuinfo frequency    : base [%u] max [%u]\n",
+               p_cpufreq->cpuinfo_min_freq,
+               p_cpufreq->cpuinfo_max_freq);
+    else
+        printf("cpuinfo frequency    : max [%u] min [%u] cur [%u]\n",
+               p_cpufreq->cpuinfo_max_freq,
+               p_cpufreq->cpuinfo_min_freq,
+               p_cpufreq->cpuinfo_cur_freq);
 
     printf("scaling_driver       : %s\n", p_cpufreq->scaling_driver);
 
-    printf("scaling_avail_gov    : %s\n",
-           p_cpufreq->scaling_available_governors);
-
-    printf("current_governor     : %s\n", p_cpufreq->u.s.scaling_governor);
-    if ( !strncmp(p_cpufreq->u.s.scaling_governor,
-                  "userspace", CPUFREQ_NAME_LEN) )
-    {
-        printf("  userspace specific :\n");
-        printf("    scaling_setspeed : %u\n",
-               p_cpufreq->u.s.u.userspace.scaling_setspeed);
-    }
-    else if ( !strncmp(p_cpufreq->u.s.scaling_governor,
-                       "ondemand", CPUFREQ_NAME_LEN) )
+    if ( !hwp )
     {
-        printf("  ondemand specific  :\n");
-        printf("    sampling_rate    : max [%u] min [%u] cur [%u]\n",
-               p_cpufreq->u.s.u.ondemand.sampling_rate_max,
-               p_cpufreq->u.s.u.ondemand.sampling_rate_min,
-               p_cpufreq->u.s.u.ondemand.sampling_rate);
-        printf("    up_threshold     : %u\n",
-               p_cpufreq->u.s.u.ondemand.up_threshold);
-    }
+        printf("scaling_avail_gov    : %s\n",
+               p_cpufreq->scaling_available_governors);
 
-    printf("scaling_avail_freq   :");
-    for ( i = 0; i < p_cpufreq->freq_num; i++ )
-        if ( p_cpufreq->scaling_available_frequencies[i] ==
-             p_cpufreq->u.s.scaling_cur_freq )
-            printf(" *%d", p_cpufreq->scaling_available_frequencies[i]);
-        else
-            printf(" %d", p_cpufreq->scaling_available_frequencies[i]);
-    printf("\n");
+        printf("current_governor     : %s\n", p_cpufreq->u.s.scaling_governor);
+        if ( !strncmp(p_cpufreq->u.s.scaling_governor,
+                      "userspace", CPUFREQ_NAME_LEN) )
+        {
+            printf("  userspace specific :\n");
+            printf("    scaling_setspeed : %u\n",
+                   p_cpufreq->u.s.u.userspace.scaling_setspeed);
+        }
+        else if ( !strncmp(p_cpufreq->u.s.scaling_governor,
+                           "ondemand", CPUFREQ_NAME_LEN) )
+        {
+            printf("  ondemand specific  :\n");
+            printf("    sampling_rate    : max [%u] min [%u] cur [%u]\n",
+                   p_cpufreq->u.s.u.ondemand.sampling_rate_max,
+                   p_cpufreq->u.s.u.ondemand.sampling_rate_min,
+                   p_cpufreq->u.s.u.ondemand.sampling_rate);
+            printf("    up_threshold     : %u\n",
+                   p_cpufreq->u.s.u.ondemand.up_threshold);
+        }
+
+        printf("scaling_avail_freq   :");
+        for ( i = 0; i < p_cpufreq->freq_num; i++ )
+            if ( p_cpufreq->scaling_available_frequencies[i] ==
+                 p_cpufreq->u.s.scaling_cur_freq )
+                printf(" *%d", p_cpufreq->scaling_available_frequencies[i]);
+            else
+                printf(" %d", p_cpufreq->scaling_available_frequencies[i]);
+        printf("\n");
 
-    printf("scaling frequency    : max [%u] min [%u] cur [%u]\n",
-           p_cpufreq->u.s.scaling_max_freq,
-           p_cpufreq->u.s.scaling_min_freq,
-           p_cpufreq->u.s.scaling_cur_freq);
+        printf("scaling frequency    : max [%u] min [%u] cur [%u]\n",
+               p_cpufreq->u.s.scaling_max_freq,
+               p_cpufreq->u.s.scaling_min_freq,
+               p_cpufreq->u.s.scaling_cur_freq);
+    }
 
     printf("turbo mode           : %s\n",
            p_cpufreq->turbo_enabled ? "enabled" : "disabled or n/a");
Signature · 1 line
2.41.0
#14 lore raw

Re: [PATCH v7 08/15] xenpm: Change get-cpufreq-para output for hwp

Quoted 13 lines · Jason Andryuk
tools/misc/xenpm.c +1 -0
index 1c474c3b59..21c93386de 100644
@@ -711,6 +711,7 @@ void start_gather_func(int argc, char *argv[])
 /* print out parameters about cpu frequency */
 static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
 {
+    bool hwp = strcmp(p_cpufreq->scaling_driver, XEN_HWP_DRIVER_NAME) == 0;
     int i;
 
     printf("cpu id               : %d\n", cpuid);

tools/misc/xenpm.c static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq) · line 723

Quoted 3 lines
tools/misc/xenpm.c +1 -0
@@ -720,49 +721,57 @@ static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
     printf("scaling_driver       : %s\n", p_cpufreq->scaling_driver);
 
+    if ( !hwp )

This test kind of feels wrong. Should we test instead the thing we want to print? Maybe declaring another bool, something like "bool scaling_governor = !hwp" just below the declaration of "bool hwp"? Otherwise, if there's another technology that comes along that isn't "hwp" and not something that can be printed with this, there's going to be some kind of hidden bug (even if probably easy to spot during development).

tools/misc/xenpm.c static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq) · line 739

Quoted 3 lines
tools/misc/xenpm.c +2 -0
     {
+        printf("scaling_avail_gov    : %s\n",
+               p_cpufreq->scaling_available_governors);

Thanks,

Signature · 1 line
Anthony PERARD
#15 lore raw

Re: [PATCH v7 08/15] xenpm: Change get-cpufreq-para output for hwp

Quoted 16 lines · Anthony PERARD
Quoted 15 lines · Jason Andryuk
tools/misc/xenpm.c +1 -0
index 1c474c3b59..21c93386de 100644
@@ -711,6 +711,7 @@ void start_gather_func(int argc, char *argv[])
 /* print out parameters about cpu frequency */
 static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
 {
+    bool hwp = strcmp(p_cpufreq->scaling_driver, XEN_HWP_DRIVER_NAME) == 0;
     int i;
 
     printf("cpu id               : %d\n", cpuid);
@@ -720,49 +721,57 @@ static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
     printf("scaling_driver       : %s\n", p_cpufreq->scaling_driver);
 
Quoted 9 lines

tools/misc/xenpm.c static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq) · line 736

Quoted 1 line
tools/misc/xenpm.c +1 -0
+    if ( !hwp )

This test kind of feels wrong. Should we test instead the thing we want to print? Maybe declaring another bool, something like "bool scaling_governor = !hwp" just below the declaration of "bool hwp"? Otherwise, if there's another technology that comes along that isn't "hwp" and not something that can be printed with this, there's going to be some kind of hidden bug (even if probably easy to spot during development).

We agreed in patch 11 that this test would be temporary, and changed in patch 11, so:

Thanks,

Signature · 1 line
Anthony PERARD
#16 lore raw

PATCH v7 9/15 cpufreq: Export HWP parameters to userspace as CPPC

In reply to Jason Andryuk

Extend xen_get_cpufreq_para to return hwp parameters. HWP is an implementation of ACPI CPPC (Collaborative Processor Performance Control). Use the CPPC name since that might be useful in the future for AMD P-state.

We need the features bitmask to indicate fields supported by the actual hardware - this only applies to activity window for the time being.

The HWP most_efficient is mapped to CPPC lowest_nonlinear, and guaranteed is mapped to nominal. CPPC has a guaranteed that is optional while nominal is required. ACPI spec says "If this register is not implemented, OSPM assumes guaranteed performance is always equal to nominal performance."

---
v2:
Style fixes
Don't bump XEN_SYSCTL_INTERFACE_VERSION
Drop cpufreq.h comment divider
Expand xen_hwp_para comment
Add HWP activity window mantissa/exponent defines
Handle union rename
Add const to get_hwp_para
Remove hw_ prefix from xen_hwp_para members
Use XEN_HWP_GOVERNOR
Use per_cpu for hwp_drv_data

v4:
Fixup for opt_cpufreq_hwp/hdc removal
get_hwp_para() takes cpu as arg
XEN_ prefix HWP_ACT_WINDOW_*
Drop HWP_ACT_WINDOW_EXPONENT_SHIFT - shift MASK
Remove Energy Bias (0-15) EPP fallback
Rename xen_hwp_para to xen_cppc_para
s/hwp/cppc/
Use scaling driver to switch output

v5:
Use XEN_HWP_DRIVER_NAME
Use cpufreq.h for declarations
Fixup some comments
Drop const from unsigned int cpu
Drop some unnecessary { }
Use strncmp
Switch sizeof(char) to sizeof(*scaling_available_governors) Reindent copy_to_guest call
Add "HWP: " prefix to sysctl comments for cppc->hwp mapping

v6:
Drop commit message uint8_t paragraph
sysctl.h s/They/They're/
Fix indent and add parens for read_scaling_available_governors() args Add Jan's Ack

---
 xen/arch/x86/acpi/cpufreq/hwp.c    | 23 ++++++++++
 xen/drivers/acpi/pmstat.c          | 74 ++++++++++++++++--------------
 xen/include/acpi/cpufreq/cpufreq.h |  2 +
 xen/include/public/sysctl.h        | 56 ++++++++++++++++++++++
 4 files changed, 121 insertions(+), 34 deletions(-)
xen/arch/x86/acpi/cpufreq/hwp.c +23 -0
index 33039ad39e..2cbfa1ce33 100644
@@ -519,6 +519,29 @@ static const struct cpufreq_driver __initconstrel hwp_cpufreq_driver =
     .update = hwp_cpufreq_update,
 };
 
+int get_hwp_para(unsigned int cpu,
+                 struct xen_cppc_para *cppc_para)
+{
+    const struct hwp_drv_data *data = per_cpu(hwp_drv_data, cpu);
+
+    if ( data == NULL )
+        return -ENODATA;
+
+    cppc_para->features         =
+        (feature_hwp_activity_window ? XEN_SYSCTL_CPPC_FEAT_ACT_WINDOW : 0);
+    cppc_para->lowest           = data->hw.lowest;
+    cppc_para->lowest_nonlinear = data->hw.most_efficient;
+    cppc_para->nominal          = data->hw.guaranteed;
+    cppc_para->highest          = data->hw.highest;
+    cppc_para->minimum          = data->minimum;
+    cppc_para->maximum          = data->maximum;
+    cppc_para->desired          = data->desired;
+    cppc_para->energy_perf      = data->energy_perf;
+    cppc_para->activity_window  = data->activity_window;
+
+    return 0;
+}
+
 int __init hwp_register_driver(void)
 {
     int ret;
xen/drivers/acpi/pmstat.c +40 -34

86 lines not on this page · raw message

xen/include/acpi/cpufreq/cpufreq.h +2 -0
index 6737798e25..f90e9d3a6c 100644
@@ -255,5 +255,7 @@ void intel_feature_detect(struct cpufreq_policy *policy);
 int hwp_cmdline_parse(const char *s, const char *e);
 int hwp_register_driver(void);
 bool hwp_active(void);
+int get_hwp_para(unsigned int cpu,
+                 struct xen_cppc_para *cppc_para);
 
 #endif /* __XEN_CPUFREQ_PM_H__ */
xen/include/public/sysctl.h +56 -0

68 lines not on this page · raw message

Signature · 1 line
2.41.0
#17 lore raw

PATCH v7 10/15 libxc: Include cppc_para in definitions

In reply to Jason Andryuk

Expose the cppc_para fields through libxc.

---
v4:
Rename hwp to cppc
Add Anthony's Ack

---
 tools/include/xenctrl.h | 2 ++
 1 file changed, 2 insertions(+)
tools/include/xenctrl.h +2 -0
index 8aedb952a0..2092632296 100644
@@ -1892,6 +1892,7 @@ int xc_smt_disable(xc_interface *xch);
  */
 typedef struct xen_userspace xc_userspace_t;
 typedef struct xen_ondemand xc_ondemand_t;
+typedef struct xen_cppc_para xc_cppc_para_t;
 
 struct xc_get_cpufreq_para {
     /* IN/OUT variable */
@@ -1923,6 +1924,7 @@ struct xc_get_cpufreq_para {
                 xc_ondemand_t ondemand;
             } u;
         } s;
+        xc_cppc_para_t cppc_para;
     } u;
 
     int32_t turbo_enabled;
Signature · 1 line
2.41.0
#18 lore raw

PATCH v7 11/15 xenpm: Print HWP/CPPC parameters

In reply to Jason Andryuk

Print HWP-specific parameters. Some are always present, but others depend on hardware support.

---
v2:
Style fixes
Declare i outside loop
Replace repearted hardware/configured limits with spaces Fixup for hw_ removal
Use XEN_HWP_GOVERNOR
Use HWP_ACT_WINDOW_EXPONENT_*
Remove energy_perf hw autonomous - 0 doesn't mean autonomous

v4:
Return activity_window from calculate_hwp_activity_window Use blanks instead of _ in output
Use MASK_EXTR
Check XEN_HWP_DRIVER name since governor is no longer returned s/hwp/cppc

v5:
Add Jan's Reviewed-by

---
 tools/misc/xenpm.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 66 insertions(+)
tools/misc/xenpm.c +66 -0
index 21c93386de..3abd99fd20 100644
@@ -708,6 +708,46 @@ void start_gather_func(int argc, char *argv[])
     pause();
 }
 
+static unsigned int calculate_activity_window(const xc_cppc_para_t *cppc,
+                                              const char **units)
+{
+    unsigned int mantissa = MASK_EXTR(cppc->activity_window,
+                                      XEN_CPPC_ACT_WINDOW_MANTISSA_MASK);
+    unsigned int exponent = MASK_EXTR(cppc->activity_window,
+                                      XEN_CPPC_ACT_WINDOW_EXPONENT_MASK);
+    unsigned int multiplier = 1;
+    unsigned int i;
+
+    /*
+     * SDM only states a 0 register is hardware selected, and doesn't mention
+     * a 0 mantissa with a non-0 exponent.  Only special case a 0 register.
+     */
+    if ( cppc->activity_window == 0 )
+    {
+        *units = "hardware selected";
+
+        return 0;
+    }
+
+    if ( exponent >= 6 )
+    {
+        *units = "s";
+        exponent -= 6;
+    }
+    else if ( exponent >= 3 )
+    {
+        *units = "ms";
+        exponent -= 3;
+    }
+    else
+        *units = "us";
+
+    for ( i = 0; i < exponent; i++ )
+        multiplier *= 10;
+
+    return mantissa * multiplier;
+}
+
 /* print out parameters about cpu frequency */
 static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
 {
@@ -772,6 +812,32 @@ static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
                p_cpufreq->u.s.scaling_min_freq,
                p_cpufreq->u.s.scaling_cur_freq);
     }
+    else
+    {
+        const xc_cppc_para_t *cppc = &p_cpufreq->u.cppc_para;
+
+        printf("cppc variables       :\n");
+        printf("  hardware limits    : lowest [%u] lowest nonlinear [%u]\n",
+               cppc->lowest, cppc->lowest_nonlinear);
+        printf("                     : nominal [%u] highest [%u]\n",
+               cppc->nominal, cppc->highest);
+        printf("  configured limits  : min [%u] max [%u] energy perf [%u]\n",
+               cppc->minimum, cppc->maximum, cppc->energy_perf);
+
+        if ( cppc->features & XEN_SYSCTL_CPPC_FEAT_ACT_WINDOW )
+        {
+            unsigned int activity_window;
+            const char *units;
+
+            activity_window = calculate_activity_window(cppc, &units);
+            printf("                     : activity_window [%u %s]\n",
+                   activity_window, units);
+        }
+
+        printf("                     : desired [%u%s]\n",
+               cppc->desired,
+               cppc->desired ? "" : " hw autonomous");
+    }
 
     printf("turbo mode           : %s\n",
            p_cpufreq->turbo_enabled ? "enabled" : "disabled or n/a");
Signature · 1 line
2.41.0
#19 lore raw

Re: [PATCH v7 11/15] xenpm: Print HWP/CPPC parameters

tools/misc/xenpm.c static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq) · line 816

Quoted 6 lines · Jason Andryuk
tools/misc/xenpm.c +2 -0
@@ -772,6 +812,32 @@ static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
                p_cpufreq->u.s.scaling_min_freq,
                p_cpufreq->u.s.scaling_cur_freq);
     }
+    else
+    {

I feel like this could be confusing. In this function, we have both:

    if ( hwp ) { this; } else { that; }
and
    if ( !hwp ) { that; } else { this; }

If we could have the condition in the same order, or use the same condition for both "true" blocks, that would be nice.

tools/misc/xenpm.c static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq) · line 821

Quoted 5 lines
tools/misc/xenpm.c +5 -0
+        const xc_cppc_para_t *cppc = &p_cpufreq->u.cppc_para;
+
+        printf("cppc variables       :\n");
+        printf("  hardware limits    : lowest [%u] lowest nonlinear [%u]\n",
+               cppc->lowest, cppc->lowest_nonlinear);

All these %u should be %"PRIu32", right? Even if the rest of the function is bogus... and even if it's probably be a while before %PRIu32 is different from %u.

Thanks,

Signature · 1 line
Anthony PERARD
#20 lore raw

Re: [PATCH v7 11/15] xenpm: Print HWP/CPPC parameters

Quoted 10 lines · Anthony PERARD

tools/misc/xenpm.c static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq) · line 816

Quoted 6 lines · Jason Andryuk
tools/misc/xenpm.c +2 -0
@@ -772,6 +812,32 @@ static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
                p_cpufreq->u.s.scaling_min_freq,
                p_cpufreq->u.s.scaling_cur_freq);
     }
+    else
+    {

I feel like this could be confusing. In this function, we have both:

Quoted 6 lines
    if ( hwp ) { this; } else { that; }
and
    if ( !hwp ) { that; } else { this; }

If we could have the condition in the same order, or use the same condition for both "true" blocks, that would be nice.

Makes sense. From your comment on patch 8, I was thinking of switching to a bool scaling_governor and using that. But now I think hwp is better since it's the specific governor - not the default one. In light of that, I would just re-organize everything to be "if ( hwp )".

With that, patch 8 is more of a transitive step, and I would leave the "if ( !hwp )". Then here in patch 11 the HWP code would be added first inside "if ( hwp )". Does that sound good?

Quoted 9 lines

tools/misc/xenpm.c static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq) · line 821

Quoted 5 lines
tools/misc/xenpm.c +5 -0
+        const xc_cppc_para_t *cppc = &p_cpufreq->u.cppc_para;
+
+        printf("cppc variables       :\n");
+        printf("  hardware limits    : lowest [%u] lowest nonlinear [%u]\n",
+               cppc->lowest, cppc->lowest_nonlinear);

All these %u should be %"PRIu32", right? Even if the rest of the function is bogus... and even if it's probably be a while before %PRIu32 is different from %u.

Yes, you are correct. In patch 8 "xenpm: Change get-cpufreq-para output for hwp", some existing %u-s are moved and a few more added. Do you want all of those converted to %PRIu32?

Thanks,
Jason

#21 lore raw

Re: [PATCH v7 11/15] xenpm: Print HWP/CPPC parameters

Quoted 25 lines · Jason Andryuk
Quoted 10 lines · Anthony PERARD

tools/misc/xenpm.c static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq) · line 816

Quoted 6 lines · Jason Andryuk
tools/misc/xenpm.c +2 -0
@@ -772,6 +812,32 @@ static void print_cpufreq_para(int cpuid, struct xc_get_cpufreq_para *p_cpufreq)
                p_cpufreq->u.s.scaling_min_freq,
                p_cpufreq->u.s.scaling_cur_freq);
     }
+    else
+    {

I feel like this could be confusing. In this function, we have both:

Quoted 6 lines
    if ( hwp ) { this; } else { that; }
and
    if ( !hwp ) { that; } else { this; }

If we could have the condition in the same order, or use the same condition for both "true" blocks, that would be nice.

Makes sense. From your comment on patch 8, I was thinking of switching to a bool scaling_governor and using that. But now I think hwp is better since it's the specific governor - not the default one. In light of that, I would just re-organize everything to be "if ( hwp )".

Quoted 3 lines

With that, patch 8 is more of a transitive step, and I would leave the "if ( !hwp )". Then here in patch 11 the HWP code would be added first inside "if ( hwp )". Does that sound good?

Yes, that sounds fine.

Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
tools/misc/xenpm.c +5 -0
+        const xc_cppc_para_t *cppc = &p_cpufreq->u.cppc_para;
+
+        printf("cppc variables       :\n");
+        printf("  hardware limits    : lowest [%u] lowest nonlinear [%u]\n",
+               cppc->lowest, cppc->lowest_nonlinear);
Quoted 7 lines
Quoted 3 lines

All these %u should be %"PRIu32", right? Even if the rest of the function is bogus... and even if it's probably be a while before %PRIu32 is different from %u.

Yes, you are correct. In patch 8 "xenpm: Change get-cpufreq-para output for hwp", some existing %u-s are moved and a few more added. Do you want all of those converted to %PRIu32?

For the newly added %u, yes, that would be nice.

As for the existing one, maybe as a separated patch, if you wish. At the moment, patch 8 is easy to review with "--ignore-space-change".

Cheers,

Signature · 1 line
Anthony PERARD
#22 lore raw

PATCH v7 12/15 xen: Add SET_CPUFREQ_HWP xen_sysctl_pm_op

In reply to Jason Andryuk

Add SET_CPUFREQ_HWP xen_sysctl_pm_op to set HWP parameters. The sysctl supports setting multiple values simultaneously as indicated by the set_params bits. This allows atomically applying new HWP configuration via a single wrmsr.

XEN_SYSCTL_HWP_SET_PRESET_BALANCE/PERFORMANCE/POWERSAVE provide three common presets. Setting them depends on hardware limits which the hypervisor is already caching. So using them allows skipping a hypercall to query the limits (lowest/highest) to then set those same values. The code is organized to allow a preset to be refined with additional parameters if desired.

"most_efficient" and "guaranteed" could be additional presets in the future, but the are not added now. Those levels can change at runtime, but we don't have code in place to monitor and update for those events.

Since activity window may not be supported by all hardware, omit writing it when not supported, and return that fact to userspace by updating set_params.

CPPC parameter checking disallows setting reserved bytes and ensure values are only non-zero when the corresponding set_params bit is set. There is no range checking (0-255 is allowed) since hardware is documented to clip internally.

---
v7:
Fix brace placement
Add Jan's R-b

v6:
Add () around bit test
Add U suffix to XEN_SYSCTL_ constants
Change -EINVAL returns to -ENOENT/-EOPNOTSUPP
Return success for no-op set_hwp_para()

v5:
Use cpufreq.h for include
Add () around bit test
Fix Configure typo
Drop duplicated energy_perf comment
Define XEN_SYSCTL_CPPC_ACT_WINDOW_MASK from *_MANTISSA_MASK and *_EXPONENT_MASK Re-order XEN_SYSCTL_CPPC_SET_* flags to match field and hardware order. Remove const from set_cppc param to update set_params
Skip Activity Window if not supported by hardware and clear set_params Make parameter parsing consistent
Add an exit path when there are no parameters to write.
Expand the header file to cover the IN/OUT set_params.
Remove the "desired" lowest/highest checking as hardware clips internally

v4:
Remove IA32_ENERGY_BIAS support
Validate parameters don't exceed 255
Use CPPC/cppc name
set_cppc_para() add const
set_cppc_para() return hwp_cpufreq_target()
Expand sysctl comments

v3:
Remove cpufreq_governor_internal from set_cpufreq_hwp

v2:
Update for naming anonymous union
Drop hwp_err for invalid input in set_hwp_para() Drop uint16_t cast in XEN_SYSCTL_HWP_SET_PARAM_MASK Drop parens for HWP_SET_PRESET defines
Reference activity_window format comment
Place SET_CPUFREQ_HWP after SET_CPUFREQ_PARA
Add {HWP,IA32}_ENERGY_PERF_MAX_{PERFORMANCE,POWERSAVE} defines Order defines before fields in sysctl.h
Use XEN_HWP_GOVERNOR
Use per_cpu for hwp_drv_data

---
 xen/arch/x86/acpi/cpufreq/hwp.c    | 102 +++++++++++++++++++++++++++++
 xen/drivers/acpi/pmstat.c          |  17 +++++
 xen/include/acpi/cpufreq/cpufreq.h |   2 +
 xen/include/public/sysctl.h        |  64 ++++++++++++++++++
 4 files changed, 185 insertions(+)
xen/arch/x86/acpi/cpufreq/hwp.c +102 -0
index 2cbfa1ce33..f6d445a8eb 100644
@@ -21,6 +21,10 @@ static bool __ro_after_init feature_hdc;
 
 static bool __ro_after_init opt_cpufreq_hdc = true;
 
+#define HWP_ENERGY_PERF_MAX_PERFORMANCE 0
+#define HWP_ENERGY_PERF_BALANCE         0x80
+#define HWP_ENERGY_PERF_MAX_POWERSAVE   0xff
+
 union hwp_request
 {
     struct
@@ -542,6 +546,104 @@ int get_hwp_para(unsigned int cpu,
     return 0;
 }
 
+int set_hwp_para(struct cpufreq_policy *policy,
+                 struct xen_set_cppc_para *set_cppc)
+{
+    unsigned int cpu = policy->cpu;
+    struct hwp_drv_data *data = per_cpu(hwp_drv_data, cpu);
+    bool cleared_act_window = false;
+
+    if ( data == NULL )
+        return -ENOENT;
+
+    /* Validate all parameters - Disallow reserved bits. */
+    if ( set_cppc->minimum > 255 ||
+         set_cppc->maximum > 255 ||
+         set_cppc->desired > 255 ||
+         set_cppc->energy_perf > 255 ||
+         (set_cppc->set_params & ~XEN_SYSCTL_CPPC_SET_PARAM_MASK) ||
+         (set_cppc->activity_window & ~XEN_SYSCTL_CPPC_ACT_WINDOW_MASK) )
+        return -EINVAL;
+
+    /* Only allow values if params bit is set. */
+    if ( (!(set_cppc->set_params & XEN_SYSCTL_CPPC_SET_DESIRED) &&
+          set_cppc->desired) ||
+         (!(set_cppc->set_params & XEN_SYSCTL_CPPC_SET_MINIMUM) &&
+          set_cppc->minimum) ||
+         (!(set_cppc->set_params & XEN_SYSCTL_CPPC_SET_MAXIMUM) &&
+          set_cppc->maximum) ||
+         (!(set_cppc->set_params & XEN_SYSCTL_CPPC_SET_ENERGY_PERF) &&
+          set_cppc->energy_perf) ||
+         (!(set_cppc->set_params & XEN_SYSCTL_CPPC_SET_ACT_WINDOW) &&
+          set_cppc->activity_window) )
+        return -EINVAL;
+
+    /* Clear out activity window if lacking HW supported. */
+    if ( (set_cppc->set_params & XEN_SYSCTL_CPPC_SET_ACT_WINDOW) &&
+         !feature_hwp_activity_window )
+    {
+        set_cppc->set_params &= ~XEN_SYSCTL_CPPC_SET_ACT_WINDOW;
+        cleared_act_window = true;
+    }
+
+    /* Return if there is nothing to do. */
+    if ( set_cppc->set_params == 0 )
+        return 0;
+
+    /* Apply presets */
+    switch ( set_cppc->set_params & XEN_SYSCTL_CPPC_SET_PRESET_MASK )
+    {
+    case XEN_SYSCTL_CPPC_SET_PRESET_POWERSAVE:
+        data->minimum = data->hw.lowest;
+        data->maximum = data->hw.lowest;
+        data->activity_window = 0;
+        data->energy_perf = HWP_ENERGY_PERF_MAX_POWERSAVE;
+        data->desired = 0;
+        break;
+
+    case XEN_SYSCTL_CPPC_SET_PRESET_PERFORMANCE:
+        data->minimum = data->hw.highest;
+        data->maximum = data->hw.highest;
+        data->activity_window = 0;
+        data->energy_perf = HWP_ENERGY_PERF_MAX_PERFORMANCE;
+        data->desired = 0;
+        break;
+
+    case XEN_SYSCTL_CPPC_SET_PRESET_BALANCE:
+        data->minimum = data->hw.lowest;
+        data->maximum = data->hw.highest;
+        data->activity_window = 0;
+        data->energy_perf = HWP_ENERGY_PERF_BALANCE;
+        data->desired = 0;
+        break;
+
+    case XEN_SYSCTL_CPPC_SET_PRESET_NONE:
+        break;
+
+    default:
+        return -EINVAL;
+    }
+
+    /* Further customize presets if needed */
+    if ( set_cppc->set_params & XEN_SYSCTL_CPPC_SET_MINIMUM )
+        data->minimum = set_cppc->minimum;
+
+    if ( set_cppc->set_params & XEN_SYSCTL_CPPC_SET_MAXIMUM )
+        data->maximum = set_cppc->maximum;
+
+    if ( set_cppc->set_params & XEN_SYSCTL_CPPC_SET_ENERGY_PERF )
+        data->energy_perf = set_cppc->energy_perf;
+
+    if ( set_cppc->set_params & XEN_SYSCTL_CPPC_SET_DESIRED )
+        data->desired = set_cppc->desired;
+
+    if ( set_cppc->set_params & XEN_SYSCTL_CPPC_SET_ACT_WINDOW )
+        data->activity_window = set_cppc->activity_window &
+                                XEN_SYSCTL_CPPC_ACT_WINDOW_MASK;
+
+    return hwp_cpufreq_target(policy, 0, 0);
+}
+
 int __init hwp_register_driver(void)
 {
     int ret;
xen/drivers/acpi/pmstat.c +17 -0
index e4e2df94a7..85097d463c 100644
@@ -400,6 +400,19 @@ static int set_cpufreq_para(struct xen_sysctl_pm_op *op)
     return ret;
 }
 
+static int set_cpufreq_cppc(struct xen_sysctl_pm_op *op)
+{
+    struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_policy, op->cpuid);
+
+    if ( !policy || !policy->governor )
+        return -ENOENT;
+
+    if ( !hwp_active() )
+        return -EOPNOTSUPP;
+
+    return set_hwp_para(policy, &op->u.set_cppc);
+}
+
 int do_pm_op(struct xen_sysctl_pm_op *op)
 {
     int ret = 0;
@@ -472,6 +485,10 @@ int do_pm_op(struct xen_sysctl_pm_op *op)
         break;
     }
 
+    case SET_CPUFREQ_CPPC:
+        ret = set_cpufreq_cppc(op);
+        break;
+
     case GET_CPUFREQ_AVGFREQ:
     {
         op->u.get_avgfreq = cpufreq_driver_getavg(op->cpuid, USR_GETAVG);
xen/include/acpi/cpufreq/cpufreq.h +2 -0
index f90e9d3a6c..281e3f513d 100644
@@ -257,5 +257,7 @@ int hwp_register_driver(void);
 bool hwp_active(void);
 int get_hwp_para(unsigned int cpu,
                  struct xen_cppc_para *cppc_para);
+int set_hwp_para(struct cpufreq_policy *policy,
+                 struct xen_set_cppc_para *set_cppc);
 
 #endif /* __XEN_CPUFREQ_PM_H__ */
xen/include/public/sysctl.h +64 -0

82 lines not on this page · raw message

Signature · 1 line
2.41.0
#23 lore raw

PATCH v7 13/15 libxc: Add xc_set_cpufreq_cppc

In reply to Jason Andryuk

Add xc_set_cpufreq_cppc to allow calling xen_systctl_pm_op SET_CPUFREQ_CPPC.

---
v2:
Mark xc_set_hwp_para_t const

v4:
s/hwp/cppc/
Add Anthony's Ack

v5:
Remove const and copy back result

---
 tools/include/xenctrl.h |  4 ++++
 tools/libs/ctrl/xc_pm.c | 23 +++++++++++++++++++++++
 2 files changed, 27 insertions(+)
tools/include/xenctrl.h +4 -0
index 2092632296..52f42fb5b6 100644
@@ -1930,11 +1930,15 @@ struct xc_get_cpufreq_para {
     int32_t turbo_enabled;
 };
 
+typedef struct xen_set_cppc_para xc_set_cppc_para_t;
+
 int xc_get_cpufreq_para(xc_interface *xch, int cpuid,
                         struct xc_get_cpufreq_para *user_para);
 int xc_set_cpufreq_gov(xc_interface *xch, int cpuid, char *govname);
 int xc_set_cpufreq_para(xc_interface *xch, int cpuid,
                         int ctrl_type, int ctrl_value);
+int xc_set_cpufreq_cppc(xc_interface *xch, int cpuid,
+                        xc_set_cppc_para_t *set_cppc);
 int xc_get_cpufreq_avgfreq(xc_interface *xch, int cpuid, int *avg_freq);
 
 int xc_set_sched_opt_smt(xc_interface *xch, uint32_t value);
tools/libs/ctrl/xc_pm.c +23 -0
index cea3eab22e..1f267147f6 100644
@@ -329,6 +329,29 @@ int xc_set_cpufreq_para(xc_interface *xch, int cpuid,
     return xc_sysctl(xch, &sysctl);
 }
 
+int xc_set_cpufreq_cppc(xc_interface *xch, int cpuid,
+                        xc_set_cppc_para_t *set_cppc)
+{
+    DECLARE_SYSCTL;
+    int ret;
+
+    if ( !xch )
+    {
+        errno = EINVAL;
+        return -1;
+    }
+    sysctl.cmd = XEN_SYSCTL_pm_op;
+    sysctl.u.pm_op.cmd = SET_CPUFREQ_CPPC;
+    sysctl.u.pm_op.cpuid = cpuid;
+    sysctl.u.pm_op.u.set_cppc = *set_cppc;
+
+    ret = xc_sysctl(xch, &sysctl);
+
+    *set_cppc = sysctl.u.pm_op.u.set_cppc;
+
+    return ret;
+}
+
 int xc_get_cpufreq_avgfreq(xc_interface *xch, int cpuid, int *avg_freq)
 {
     int ret = 0;
Signature · 1 line
2.41.0
#24 lore raw

PATCH v7 14/15 xenpm: Add set-cpufreq-cppc subcommand

In reply to Jason Andryuk

set-cpufreq-cppc allows setting the Hardware P-State (HWP) parameters.

It can be run on all or just a single cpu. There are presets of balance, powersave & performance. Those can be further tweaked by param:val arguments as explained in the usage description.

Parameter names are just checked to the first 3 characters to shorten typing.

Some options are hardware dependent, and ranges can be found in get-cpufreq-para.

---
v5:
Make i unsigned int
Use local max_cpuid instead of max_cpu_nr
Add Jan's Ack
Check set_params and print a message if activity window wasn't set

v4:
Remove energy bias 0-15 & 7 references
Use MASK_INSR
Fixup { placement
Drop extra case in parse_activity_window
strcmp suffix
Expand help text
s/hwp/cppc/
Use isdigit() to check cpuid - otherwise run on all CPUs.

v2:
Compare provided parameter name and not just 3 characters. Use "-" in parameter names
Remove hw_
Replace sscanf with strchr & strtoul.
Remove toplevel error message with lower level ones. Help text s/127/128/
Help text mention truncation.
Avoid some truncation rounding down by adding 5 before division. Help test mention default microseconds
Also comment the limit check written to avoid overflow.

---
 tools/misc/xenpm.c | 244 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 244 insertions(+)
tools/misc/xenpm.c +244 -0
index 3abd99fd20..0877f43946 100644
@@ -16,6 +16,8 @@
  */
 #define MAX_NR_CPU 512
 
+#include <ctype.h>
+#include <limits.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <unistd.h>
@@ -67,6 +69,30 @@ void show_help(void)
             " set-max-cstate        <num>|'unlimited' [<num2>|'unlimited']\n"
             "                                     set the C-State limitation (<num> >= 0) and\n"
             "                                     optionally the C-sub-state limitation (<num2> >= 0)\n"
+            " set-cpufreq-cppc      [cpuid] [balance|performance|powersave] <param:val>*\n"
+            "                                     set Hardware P-State (HWP) parameters\n"
+            "                                     on CPU <cpuid> or all if omitted.\n"
+            "                                     optionally a preset of one of:\n"
+            "                                       balance|performance|powersave\n"
+            "                                     an optional list of param:val arguments\n"
+            "                                       minimum:N (0-255)\n"
+            "                                       maximum:N (0-255)\n"
+            "                                           get-cpufreq-para lowest/highest\n"
+            "                                           values are limits for\n"
+            "                                           minumum/maximum.\n"
+            "                                       desired:N (0-255)\n"
+            "                                           set explicit performance target.\n"
+            "                                           non-zero disables auto-HWP mode.\n"
+            "                                       energy-perf:N (0-255)\n"
+            "                                                   energy/performance hint\n"
+            "                                                   lower - favor performance\n"
+            "                                                   higher - favor powersave\n"
+            "                                                   128 - balance\n"
+            "                                       act-window:N{,m,u}s range 1us-1270s\n"
+            "                                           window for internal calculations.\n"
+            "                                           units default to \"us\" if unspecified.\n"
+            "                                           truncates un-representable values.\n"
+            "                                           0 lets the hardware decide.\n"
             " start [seconds]                     start collect Cx/Px statistics,\n"
             "                                     output after CTRL-C or SIGINT or several seconds.\n"
             " enable-turbo-mode     [cpuid]       enable Turbo Mode for processors that support it.\n"
@@ -1292,6 +1318,223 @@ void disable_turbo_mode(int argc, char *argv[])
                 errno, strerror(errno));
 }
 
+/*
+ * Parse activity_window:NNN{us,ms,s} and validate range.
+ *
+ * Activity window is a 7bit mantissa (0-127) with a 3bit exponent (0-7) base
+ * 10 in microseconds.  So the range is 1 microsecond to 1270 seconds.  A value
+ * of 0 lets the hardware autonomously select the window.
+ *
+ * Return 0 on success
+ *       -1 on error
+ */
+static int parse_activity_window(xc_set_cppc_para_t *set_cppc, unsigned long u,
+                                 const char *suffix)
+{
+    unsigned int exponent = 0;
+    unsigned int multiplier = 1;
+
+    if ( suffix && suffix[0] )
+    {
+        if ( strcmp(suffix, "s") == 0 )
+        {
+            multiplier = 1000 * 1000;
+            exponent = 6;
+        }
+        else if ( strcmp(suffix, "ms") == 0 )
+        {
+            multiplier = 1000;
+            exponent = 3;
+        }
+        else if ( strcmp(suffix, "us") != 0 )
+        {
+            fprintf(stderr, "invalid activity window units: \"%s\"\n", suffix);
+
+            return -1;
+        }
+    }
+
+    /* u * multipler > 1270 * 1000 * 1000 transformed to avoid overflow. */
+    if ( u > 1270 * 1000 * 1000 / multiplier )
+    {
+        fprintf(stderr, "activity window is too large\n");
+
+        return -1;
+    }
+
+    /* looking for 7 bits of mantissa and 3 bits of exponent */
+    while ( u > 127 )
+    {
+        u += 5; /* Round up to mitigate truncation rounding down
+                   e.g. 128 -> 120 vs 128 -> 130. */
+        u /= 10;
+        exponent += 1;
+    }
+
+    set_cppc->activity_window =
+        MASK_INSR(exponent, XEN_CPPC_ACT_WINDOW_EXPONENT_MASK) |
+        MASK_INSR(u, XEN_CPPC_ACT_WINDOW_MANTISSA_MASK);
+    set_cppc->set_params |= XEN_SYSCTL_CPPC_SET_ACT_WINDOW;
+
+    return 0;
+}
+
+static int parse_cppc_opts(xc_set_cppc_para_t *set_cppc, int *cpuid,
+                           int argc, char *argv[])
+{
+    int i = 0;
+
+    if ( argc < 1 )
+    {
+        fprintf(stderr, "Missing arguments\n");
+        return -1;
+    }
+
+    if ( isdigit(argv[i][0]) )
+    {
+        if ( sscanf(argv[i], "%d", cpuid) != 1 || *cpuid < 0 )
+        {
+            fprintf(stderr, "Could not parse cpuid \"%s\"\n", argv[i]);
+            return -1;
+        }
+
+        i++;
+    }
+
+    if ( i == argc )
+    {
+        fprintf(stderr, "Missing arguments\n");
+        return -1;
+    }
+
+    if ( strcasecmp(argv[i], "powersave") == 0 )
+    {
+        set_cppc->set_params = XEN_SYSCTL_CPPC_SET_PRESET_POWERSAVE;
+        i++;
+    }
+    else if ( strcasecmp(argv[i], "performance") == 0 )
+    {
+        set_cppc->set_params = XEN_SYSCTL_CPPC_SET_PRESET_PERFORMANCE;
+        i++;
+    }
+    else if ( strcasecmp(argv[i], "balance") == 0 )
+    {
+        set_cppc->set_params = XEN_SYSCTL_CPPC_SET_PRESET_BALANCE;
+        i++;
+    }
+
+    for ( ; i < argc; i++)
+    {
+        unsigned long val;
+        char *param = argv[i];
+        char *value;
+        char *suffix;
+        int ret;
+
+        value = strchr(param, ':');
+        if ( value == NULL )
+        {
+            fprintf(stderr, "\"%s\" is an invalid cppc parameter\n", argv[i]);
+            return -1;
+        }
+
+        value[0] = '\0';
+        value++;
+
+        errno = 0;
+        val = strtoul(value, &suffix, 10);
+        if ( (errno && val == ULONG_MAX) || value == suffix )
+        {
+            fprintf(stderr, "Could not parse number \"%s\"\n", value);
+            return -1;
+        }
+
+        if ( strncasecmp(param, "act-window", strlen(param)) == 0 )
+        {
+            ret = parse_activity_window(set_cppc, val, suffix);
+            if (ret)
+                return -1;
+
+            continue;
+        }
+
+        if ( val > 255 )
+        {
+            fprintf(stderr, "\"%s\" value \"%lu\" is out of range\n", param,
+                    val);
+            return -1;
+        }
+
+        if ( suffix && suffix[0] )
+        {
+            fprintf(stderr, "Suffix \"%s\" is invalid\n", suffix);
+            return -1;
+        }
+
+        if ( strncasecmp(param, "minimum", MAX(2, strlen(param))) == 0 )
+        {
+            set_cppc->minimum = val;
+            set_cppc->set_params |= XEN_SYSCTL_CPPC_SET_MINIMUM;
+        }
+        else if ( strncasecmp(param, "maximum", MAX(2, strlen(param))) == 0 )
+        {
+            set_cppc->maximum = val;
+            set_cppc->set_params |= XEN_SYSCTL_CPPC_SET_MAXIMUM;
+        }
+        else if ( strncasecmp(param, "desired", strlen(param)) == 0 )
+        {
+            set_cppc->desired = val;
+            set_cppc->set_params |= XEN_SYSCTL_CPPC_SET_DESIRED;
+        }
+        else if ( strncasecmp(param, "energy-perf", strlen(param)) == 0 )
+        {
+            set_cppc->energy_perf = val;
+            set_cppc->set_params |= XEN_SYSCTL_CPPC_SET_ENERGY_PERF;
+        }
+        else
+        {
+            fprintf(stderr, "\"%s\" is an invalid parameter\n", param);
+            return -1;
+        }
+    }
+
+    if ( set_cppc->set_params == 0 )
+    {
+        fprintf(stderr, "No parameters set in request\n");
+        return -1;
+    }
+
+    return 0;
+}
+
+static void cppc_set_func(int argc, char *argv[])
+{
+    xc_set_cppc_para_t set_cppc = {};
+    unsigned int max_cpuid = max_cpu_nr;
+    int cpuid = -1;
+    unsigned int i = 0;
+    uint32_t set_params;
+
+    if ( parse_cppc_opts(&set_cppc, &cpuid, argc, argv) )
+        exit(EINVAL);
+
+    if ( cpuid != -1 )
+    {
+        i = cpuid;
+        max_cpuid = i + 1;
+    }
+
+    set_params = set_cppc.set_params;
+    for ( ; i < max_cpuid; i++ ) {
+        if ( xc_set_cpufreq_cppc(xc_handle, i, &set_cppc) )
+            fprintf(stderr, "[CPU%d] failed to set cppc params (%d - %s)\n",
+                    i, errno, strerror(errno));
+    }
+
+    if ( (set_params ^ set_cppc.set_params) & XEN_SYSCTL_CPPC_SET_ACT_WINDOW )
+        printf("Activity window not supported and omitted\n");
+}
+
 struct {
     const char *name;
     void (*function)(int argc, char *argv[]);
@@ -1302,6 +1545,7 @@ struct {
     { "get-cpufreq-average", cpufreq_func },
     { "start", start_gather_func },
     { "get-cpufreq-para", cpufreq_para_func },
+    { "set-cpufreq-cppc", cppc_set_func },
     { "set-scaling-maxfreq", scaling_max_freq_func },
     { "set-scaling-minfreq", scaling_min_freq_func },
     { "set-scaling-governor", scaling_governor_func },
Signature · 1 line
2.41.0
Acked-by on v7 14/15
#25 lore raw

Re: [PATCH v7 14/15] xenpm: Add set-cpufreq-cppc subcommand

Quoted 6 lines · Jason Andryuk

set-cpufreq-cppc allows setting the Hardware P-State (HWP) parameters.

It can be run on all or just a single cpu. There are presets of balance, powersave & performance. Those can be further tweaked by param:val arguments as explained in the usage description.

Quoted 8 lines

Parameter names are just checked to the first 3 characters to shorten typing.

Some options are hardware dependent, and ranges can be found in get-cpufreq-para.

Thanks,

Signature · 1 line
Anthony PERARD
#26 lore raw

PATCH v7 15/15 CHANGELOG: Add Intel HWP entry

In reply to Jason Andryuk

---
v3:
Position under existing Added section
Add Henry's Ack

v2:
Add blank line

---
 CHANGELOG.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
CHANGELOG.md +1 -1
index 7d7e0590f8..8d6e6c3088 100644
@@ -24,7 +24,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
  - xl/libxl can customize SMBIOS strings for HVM guests.
  - Add support for AVX512-FP16 on x86.
  - On Arm, Xen supports guests running SVE/SVE2 instructions. (Tech Preview)
-
+ - Add Intel Hardware P-States (HWP) cpufreq driver.
 
 ## [4.17.0](https://xenbits.xen.org/gitweb/?p=xen.git;a=shortlog;h=RELEASE-4.17.0) - 2022-12-12
 
Signature · 1 line
2.41.0