[dpdk-stable] [PATCH 2/2] power: check if userspace governor is available

Radoslaw Biernacki radoslaw.biernacki at linaro.org
Mon Oct 16 15:47:08 CEST 2017


Since for new Intel CPU's kernel use intel_pstate driver,
which does not offer userspace governor, it is vise to check
the userspace governor availability before trying to perform
governor switch. The outcome from this patch is direct
information for user about the root cause of failure during
the rte_power_acpi_cpufreq_init(). This patch also add the
/sys file name to error message as on some platforms some
files expected by this rte_power are not available. This is
also useful for pinning down the root cause of the problem.

Signed-off-by: Radoslaw Biernacki <radoslaw.biernacki at linaro.org>
---
 lib/librte_power/rte_power_acpi_cpufreq.c | 35 ++++++++++++++++++++++++-------
 1 file changed, 27 insertions(+), 8 deletions(-)

diff --git a/lib/librte_power/rte_power_acpi_cpufreq.c b/lib/librte_power/rte_power_acpi_cpufreq.c
index 8bf5685..342eb22 100644
--- a/lib/librte_power/rte_power_acpi_cpufreq.c
+++ b/lib/librte_power/rte_power_acpi_cpufreq.c
@@ -55,9 +55,9 @@
 #define POWER_DEBUG_TRACE(fmt, args...)
 #endif
 
-#define FOPEN_OR_ERR_RET(f, retval) do { \
+#define FOPEN_OR_ERR_RET(f, fullpath, retval) do { \
 		if ((f) == NULL) { \
-			RTE_LOG(ERR, POWER, "File not openned\n"); \
+			RTE_LOG(ERR, POWER, "File %s not openned\n", fullpath); \
 			return retval; \
 		} \
 } while (0)
@@ -80,6 +80,8 @@
 #define POWER_CONVERT_TO_DECIMAL 10
 
 #define POWER_GOVERNOR_USERSPACE "userspace"
+#define POWER_SYSFILE_AVAIL_GOVERNOR   \
+		"/sys/devices/system/cpu/cpu%u/cpufreq/scaling_available_governors"
 #define POWER_SYSFILE_GOVERNOR   \
 		"/sys/devices/system/cpu/cpu%u/cpufreq/scaling_governor"
 #define POWER_SYSFILE_AVAIL_FREQ \
@@ -170,10 +172,28 @@ power_set_governor_userspace(struct rte_power_info *pi)
 	char *s;
 	int val;
 
+	/* check if userspace governor is available */
+
+	snprintf(fullpath, sizeof(fullpath), POWER_SYSFILE_AVAIL_GOVERNOR,
+			pi->lcore_id);
+	f = fopen(fullpath, "r");
+	FOPEN_OR_ERR_RET(f, fullpath, ret);
+
+	s = fgets(buf, sizeof(buf), f);
+	FOPS_OR_NULL_GOTO(s, out);
+
+	if (!strstr(buf, POWER_GOVERNOR_USERSPACE)) {
+		RTE_LOG(ERR, POWER, "Userspace governor is not available\n");
+		goto out;
+	}
+	fclose(f);
+
+	/* store current governor and set userspace governor */
+
 	snprintf(fullpath, sizeof(fullpath), POWER_SYSFILE_GOVERNOR,
 			pi->lcore_id);
 	f = fopen(fullpath, "rw+");
-	FOPEN_OR_ERR_RET(f, ret);
+	FOPEN_OR_ERR_RET(f, fullpath, ret);
 
 	if (setvbuf(f, NULL, _IONBF, 0)) {
 		RTE_LOG(ERR, POWER, "Cannot set unbuffered mode\n");
@@ -184,8 +204,7 @@ power_set_governor_userspace(struct rte_power_info *pi)
 	FOPS_OR_NULL_GOTO(s, out);
 
 	/* Check if current governor is userspace */
-	if (strncmp(buf, POWER_GOVERNOR_USERSPACE,
-			sizeof(POWER_GOVERNOR_USERSPACE)) == 0) {
+	if (!strcmp(buf, POWER_GOVERNOR_USERSPACE)) {
 		ret = 0;
 		POWER_DEBUG_TRACE("Power management governor of lcore %u is "
 				"already userspace\n", pi->lcore_id);
@@ -228,7 +247,7 @@ power_get_available_freqs(struct rte_power_info *pi)
 	snprintf(fullpath, sizeof(fullpath), POWER_SYSFILE_AVAIL_FREQ,
 			pi->lcore_id);
 	f = fopen(fullpath, "r");
-	FOPEN_OR_ERR_RET(f, ret);
+	FOPEN_OR_ERR_RET(f, fullpath, ret);
 
 	s = fgets(buf, sizeof(buf), f);
 	FOPS_OR_NULL_GOTO(s, out);
@@ -296,7 +315,7 @@ power_init_for_setting_freq(struct rte_power_info *pi)
 	snprintf(fullpath, sizeof(fullpath), POWER_SYSFILE_SETSPEED,
 			pi->lcore_id);
 	f = fopen(fullpath, "rw+");
-	FOPEN_OR_ERR_RET(f, -1);
+	FOPEN_OR_ERR_RET(f, fullpath, -1);
 
 	if (setvbuf(f, NULL, _IONBF, 0)) {
 		RTE_LOG(ERR, POWER, "Cannot set unbuffered mode\n");
@@ -398,7 +417,7 @@ power_set_governor_original(struct rte_power_info *pi)
 	snprintf(fullpath, sizeof(fullpath), POWER_SYSFILE_GOVERNOR,
 			pi->lcore_id);
 	f = fopen(fullpath, "rw+");
-	FOPEN_OR_ERR_RET(f, ret);
+	FOPEN_OR_ERR_RET(f, fullpath, ret);
 
 	if (setvbuf(f, NULL, _IONBF, 0)) {
 		RTE_LOG(ERR, POWER, "Cannot set unbuffered mode\n");
-- 
2.7.4



More information about the stable mailing list