mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: K Prateek Nayak <kprateek.nayak@amd.com>
To: Mario Limonciello <mario.limonciello@amd.com>
Cc: Perry Yuan <perry.yuan@amd.com>,
	"open list:X86 ARCHITECTURE (32-BIT AND 64-BIT)"
	<linux-kernel@vger.kernel.org>,
	"open list:CPU FREQUENCY SCALING FRAMEWORK"
	<linux-pm@vger.kernel.org>
Subject: Re: [PATCH 3/4] cpufreq: amd-pstate-ut: Tolerate aliased EPP preferences in the EPP test
Date: Thu, 1 Oct 2026 22:22:00 +0530	[thread overview]
Message-ID: <b3ed027f-8c7b-4ac5-aac6-95d47ac0c41d@amd.com> (raw)
In-Reply-To: <20260921190235.3651688-3-mario.limonciello@amd.com>

Hello Mario,

On 9/22/2026 12:32 AM, Mario Limonciello wrote:
> @@ -385,12 +388,27 @@ static int amd_pstate_ut_epp(u32 index)
>  		strreplace(buf, '\n', '\0');
>  		/*
>  		 * "dynamic" mode reports the EPP as "dynamic(profile:X)"
> -		 * Trim at "(" and just compare tie the epp string.
> +		 * Trim at "(" and just keep the preference name.
>  		 */
>  		strreplace(buf, '(', '\0');
>  
> -		if (strcmp(buf, epp_strings[i])) {
> -			pr_err("String EPP value mismatch: %s != %s\n", buf, epp_strings[i]);
> +		/*
> +		 * The preference read back may legitimately differ from the one
> +		 * written: some platforms (e.g. Zen6) program the same raw EPP
> +		 * value for more than one named preference, so show() reports the
> +		 * first name that matches that value. Rather than compare the
> +		 * strings, re-write whatever name was reported and confirm it
> +		 * programs the same raw EPP value. This verifies the show()/store()
> +		 * mapping stays consistent while tolerating such aliasing.
> +		 */
> +		ret = store_energy_performance_preference(policy, buf, strlen(buf));
> +		if (ret < 0)
> +			goto out;
> +		got_epp = FIELD_GET(AMD_CPPC_EPP_PERF_MASK, cpudata->cppc_req_cached);
> +
> +		if (want_epp != got_epp) {


So this feels a bit hacky since it may be possible an epp
may be programmed incorrectly in hardware and as a result
it aliases with a different string.

Can we use amd_pstate_cpu_epp_values() as a source of truth
to double verify?

Something like below:

  (Only build tested on bleeding-edge with this patch reverted)

diff --git a/drivers/cpufreq/amd-pstate-ut.c b/drivers/cpufreq/amd-pstate-ut.c
index be10ab90b820..506d7bb2e660 100644
--- a/drivers/cpufreq/amd-pstate-ut.c
+++ b/drivers/cpufreq/amd-pstate-ut.c
@@ -304,12 +304,12 @@ static int amd_pstate_set_mode(enum amd_pstate_mode mode)
 
 static int amd_pstate_ut_epp(u32 index)
 {
-	static const char * const epp_strings[] = {
-		"dynamic",
-		"power",
-		"balance_power",
-		"balance_performance",
-		"performance",
+	enum energy_perf_value_index epp_modes[] = {
+		EPP_INDEX_DYNAMIC,
+		EPP_INDEX_POWERSAVE,
+		EPP_INDEX_BALANCE_POWERSAVE,
+		EPP_INDEX_BALANCE_PERFORMANCE,
+		EPP_INDEX_PERFORMANCE,
 	};
 	char *buf __free(cleanup_page) = NULL;
 	struct cpufreq_policy *policy = NULL;
@@ -378,9 +378,9 @@ static int amd_pstate_ut_epp(u32 index)
 		}
 	}
 
-	for (i = 0; i < ARRAY_SIZE(epp_strings); i++) {
+	for (i = 0; i < ARRAY_SIZE(epp_modes); i++) {
 		memset(buf, 0, PAGE_SIZE);
-		snprintf(buf, PAGE_SIZE, "%s", epp_strings[i]);
+		snprintf(buf, PAGE_SIZE, "%s", energy_perf_strings[epp_modes[i]]);
 		ret = store_energy_performance_preference(policy, buf, strlen(buf));
 		if (ret < 0)
 			goto out;
@@ -396,8 +396,24 @@ static int amd_pstate_ut_epp(u32 index)
 		 */
 		strreplace(buf, '(', '\0');
 
-		if (strcmp(buf, epp_strings[i])) {
-			pr_err("String EPP value mismatch: %s != %s\n", buf, epp_strings[i]);
+		if (strcmp(buf, energy_perf_strings[epp_modes[i]])) {
+			u8 epp = FIELD_GET(AMD_CPPC_EPP_PERF_MASK, cpudata->cppc_req_cached);
+
+			/*
+			 * The preference read back may legitimately differ from the one
+			 * written: some platforms (e.g. Zen6) program the same raw EPP
+			 * value for more than one named preference, so show() reports the
+			 * first name that matches that value.
+			 *
+			 * If strings mismatch, confirm the CPU is programmed with the correct
+			 * EPP value for the mode. This verifies hardware programming while
+			 * tolerating such aliasing in driver.
+			 */
+			if (epp == amd_pstate_cpu_epp_values(cpudata->cpu_type)[epp_modes[i]])
+				continue;
+
+			pr_err("String EPP value mismatch: %s != %s\n",
+			       buf, energy_perf_strings[epp_modes[i]]);
 			ret = -EINVAL;
 			goto out;
 		}
@@ -413,11 +429,12 @@ static int amd_pstate_ut_epp(u32 index)
 		 * restore it here before dropping policy reference.
 		 */
 		if (orig_dynamic_epp) {
+			const char * const dynamic_string = energy_perf_strings[epp_modes[0]];
 			int ret2;
 
 			ret2 = store_energy_performance_preference(policy,
-								  epp_strings[0],
-								  strlen(epp_strings[0]));
+								  dynamic_string,
+								  strlen(dynamic_string));
 			if (!ret && (ret2 < 0))
 				ret = ret2;
 		}
diff --git a/drivers/cpufreq/amd-pstate.c b/drivers/cpufreq/amd-pstate.c
index f27676e8e9ef..89f911d378ae 100644
--- a/drivers/cpufreq/amd-pstate.c
+++ b/drivers/cpufreq/amd-pstate.c
@@ -44,7 +44,6 @@
 #include <acpi/cppc_acpi.h>
 
 #include <asm/msr.h>
-#include <asm/processor.h>
 #include <asm/cpufeature.h>
 #include <asm/cpu_device_id.h>
 
@@ -89,46 +88,6 @@ static int cppc_state = AMD_PSTATE_UNDEFINED;
 static bool amd_pstate_prefcore = true;
 static struct quirk_entry *quirks;
 
-/*
- * AMD Energy Preference Performance (EPP)
- * The EPP is used in the CCLK DPM controller to drive
- * the frequency that a core is going to operate during
- * short periods of activity. EPP values will be utilized for
- * different OS profiles (balanced, performance, power savings)
- * display strings corresponding to EPP index in the
- * energy_perf_strings[]
- *	index		String
- *-------------------------------------
- *	0		default
- *	1		performance
- *	2		balance_performance
- *	3		balance_power
- *	4		power
- *	5		custom (for raw EPP values)
- *	6		dynamic (platform profile driven selection)
- */
-enum energy_perf_value_index {
-	EPP_INDEX_DEFAULT = 0,
-	EPP_INDEX_PERFORMANCE,
-	EPP_INDEX_BALANCE_PERFORMANCE,
-	EPP_INDEX_BALANCE_POWERSAVE,
-	EPP_INDEX_POWERSAVE,
-	EPP_INDEX_CUSTOM,
-	EPP_INDEX_DYNAMIC,
-	EPP_INDEX_MAX,
-};
-
-static const char * const energy_perf_strings[] = {
-	[EPP_INDEX_DEFAULT] = "default",
-	[EPP_INDEX_PERFORMANCE] = "performance",
-	[EPP_INDEX_BALANCE_PERFORMANCE] = "balance_performance",
-	[EPP_INDEX_BALANCE_POWERSAVE] = "balance_power",
-	[EPP_INDEX_POWERSAVE] = "power",
-	[EPP_INDEX_CUSTOM] = "custom",
-	[EPP_INDEX_DYNAMIC] = "dynamic",
-};
-static_assert(ARRAY_SIZE(energy_perf_strings) == EPP_INDEX_MAX);
-
 /*
  * The numeric EPP value programmed for each named preference. First dimension
  * is CPU type (TOPO_CPU_TYPE_ANY for non-hybrid, TOPO_CPU_TYPE_PERFORMANCE/
@@ -159,7 +118,7 @@ static_assert(ARRAY_SIZE(epp_values) == TOPO_CPU_TYPE_LOW_POWER + 1,
  * Non-hybrid systems use TOPO_CPU_TYPE_ANY; hybrid systems use the CPU's
  * actual type (PERFORMANCE/EFFICIENCY/LOW_POWER).
  */
-static inline u8 *amd_pstate_cpu_epp_values(enum x86_topology_cpu_type cpu_type)
+u8 *amd_pstate_cpu_epp_values(enum x86_topology_cpu_type cpu_type)
 {
 	switch (cpu_type) {
 	case TOPO_CPU_TYPE_PERFORMANCE:
@@ -170,6 +129,7 @@ static inline u8 *amd_pstate_cpu_epp_values(enum x86_topology_cpu_type cpu_type)
 		return epp_values[TOPO_CPU_TYPE_ANY];
 	}
 }
+EXPORT_SYMBOL_FOR_PSTATE_UT(amd_pstate_cpu_epp_values);
 
 /**
  * struct amd_pstate_epp_values - EPP values for the four named preferences
diff --git a/drivers/cpufreq/amd-pstate.h b/drivers/cpufreq/amd-pstate.h
index c7189d177b81..422188c639b1 100644
--- a/drivers/cpufreq/amd-pstate.h
+++ b/drivers/cpufreq/amd-pstate.h
@@ -11,6 +11,8 @@
 #include <linux/pm_qos.h>
 #include <linux/platform_profile.h>
 
+#include <asm/processor.h>
+
 #if IS_MODULE(CONFIG_X86_AMD_PSTATE_UT)
 #define EXPORT_SYMBOL_FOR_PSTATE_UT(symbol) \
 	EXPORT_SYMBOL_FOR_MODULES(symbol, "amd-pstate-ut")
@@ -162,6 +164,48 @@ enum amd_pstate_mode {
 	AMD_PSTATE_MAX,
 };
 
+/*
+ * AMD Energy Preference Performance (EPP)
+ * The EPP is used in the CCLK DPM controller to drive
+ * the frequency that a core is going to operate during
+ * short periods of activity. EPP values will be utilized for
+ * different OS profiles (balanced, performance, power savings)
+ * display strings corresponding to EPP index in the
+ * energy_perf_strings[]
+ *	index		String
+ *-------------------------------------
+ *	0		default
+ *	1		performance
+ *	2		balance_performance
+ *	3		balance_power
+ *	4		power
+ *	5		custom (for raw EPP values)
+ *	6		dynamic (platform profile driven selection)
+ */
+enum energy_perf_value_index {
+	EPP_INDEX_DEFAULT = 0,
+	EPP_INDEX_PERFORMANCE,
+	EPP_INDEX_BALANCE_PERFORMANCE,
+	EPP_INDEX_BALANCE_POWERSAVE,
+	EPP_INDEX_POWERSAVE,
+	EPP_INDEX_CUSTOM,
+	EPP_INDEX_DYNAMIC,
+	EPP_INDEX_MAX,
+};
+
+static const char * const energy_perf_strings[] = {
+	[EPP_INDEX_DEFAULT] = "default",
+	[EPP_INDEX_PERFORMANCE] = "performance",
+	[EPP_INDEX_BALANCE_PERFORMANCE] = "balance_performance",
+	[EPP_INDEX_BALANCE_POWERSAVE] = "balance_power",
+	[EPP_INDEX_POWERSAVE] = "power",
+	[EPP_INDEX_CUSTOM] = "custom",
+	[EPP_INDEX_DYNAMIC] = "dynamic",
+};
+static_assert(ARRAY_SIZE(energy_perf_strings) == EPP_INDEX_MAX);
+
+u8 *amd_pstate_cpu_epp_values(enum x86_topology_cpu_type cpu_type);
+
 static inline u8 freq_to_perf(union perf_cached perf, u32 nominal_freq, unsigned int freq_val)
 {
 	u32 perf_val = DIV_ROUND_UP_ULL((u64)freq_val * perf.nominal_perf, nominal_freq);
---

Thoughts?

> +			pr_err("EPP mapping inconsistent: %s programmed %u but %s programmed %u\n",
> +			       epp_strings[i], want_epp, buf, got_epp);
>  			ret = -EINVAL;
>  			goto out;
>  		}

-- 
Thanks and Regards,
Prateek


  reply	other threads:[~2026-10-01 16:52 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-21 19:02 [PATCH 1/4] cpufreq: amd-pstate: Restore previous mode when changing driver mode fails Mario Limonciello
2026-09-21 19:02 ` [PATCH 2/4] cpufreq: amd-pstate: Propagate cppc_set_auto_sel() errors on mode change Mario Limonciello
2026-09-21 19:02 ` [PATCH 3/4] cpufreq: amd-pstate-ut: Tolerate aliased EPP preferences in the EPP test Mario Limonciello
2026-10-01 16:52   ` K Prateek Nayak [this message]
2026-10-01 17:20     ` Mario Limonciello
2026-09-21 19:02 ` [PATCH 4/4] cpufreq: amd-pstate-ut: Don't require max_freq >= nominal_freq Mario Limonciello

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=b3ed027f-8c7b-4ac5-aac6-95d47ac0c41d@amd.com \
    --to=kprateek.nayak@amd.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pm@vger.kernel.org \
    --cc=mario.limonciello@amd.com \
    --cc=perry.yuan@amd.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®