# HG changeset patch # User Keir Fraser # Date 1228490612 0 # Node ID de7fd862ada2ed079d0a5c407508eb63bb936992 # Parent 3905cbf523b2550f5025df6cc31ac60e48c1706f cpufreq: allow customization of some parameters Short of having a way for powersaved to dynamically adjust these values, at least allow specifying them on the command line. In particular, always running at an up-threshold of 80% is perhaps nice for laptop use, but certainly not desirable on servers. On shell scripts invoking large numbers of short-lived processes I noticed a 50% performance degradation on a dual-socket quad-core Barcelona just because of the load of an individual core never crossing the 80% boundary that would have resulted in increasing the frequency. (Powersaved on SLE10 sets this on native kernels to 60% or 80%, depending on whether performance or power reduction is preferred, *divided* by the number of CPUs, but capped at the lower limit of 20%.) Signed-off-by: Jan Beulich # HG changeset patch # User Keir Fraser # Date 1230557866 0 # Node ID 4035ea96ae2fafba7a5a4c1e810aa7d591758e8c # Parent 0af9fbf3f05306d4972cf05e4b6d7be2199a41cb cpufreq: Fix a cpufreq cmdline parse bug, and change sample_rate unit Signed-off-by: Liu Jinsong --- a/xen/arch/x86/acpi/cpufreq/cpufreq_ondemand.c +++ b/xen/arch/x86/acpi/cpufreq/cpufreq_ondemand.c @@ -22,15 +22,22 @@ #include #define DEF_FREQUENCY_UP_THRESHOLD (80) +#define MIN_FREQUENCY_UP_THRESHOLD (11) +#define MAX_FREQUENCY_UP_THRESHOLD (100) #define MIN_DBS_INTERVAL (MICROSECS(100)) -#define MIN_SAMPLING_MILLISECS (20) -#define MIN_STAT_SAMPLING_RATE \ +#define MIN_SAMPLING_RATE_RATIO (2) +#define MIN_SAMPLING_MILLISECS (MIN_SAMPLING_RATE_RATIO * 10) +#define MIN_STAT_SAMPLING_RATE \ (MIN_SAMPLING_MILLISECS * MILLISECS(1)) +#define MIN_SAMPLING_RATE \ + (def_sampling_rate / MIN_SAMPLING_RATE_RATIO) +#define MAX_SAMPLING_RATE (500 * def_sampling_rate) #define DEF_SAMPLING_RATE_LATENCY_MULTIPLIER (1000) #define TRANSITION_LATENCY_LIMIT (10 * 1000 ) static uint64_t def_sampling_rate; +static uint64_t usr_sampling_rate; /* Sampling types */ enum {DBS_NORMAL_SAMPLE, DBS_SUB_SAMPLE}; @@ -42,11 +49,9 @@ static unsigned int dbs_enable; /* nu static struct dbs_tuners { uint64_t sampling_rate; unsigned int up_threshold; - unsigned int ignore_nice; unsigned int powersave_bias; } dbs_tuners_ins = { .up_threshold = DEF_FREQUENCY_UP_THRESHOLD, - .ignore_nice = 0, .powersave_bias = 0, }; @@ -216,7 +221,20 @@ int cpufreq_governor_dbs(struct cpufreq_ if (def_sampling_rate < MIN_STAT_SAMPLING_RATE) def_sampling_rate = MIN_STAT_SAMPLING_RATE; - dbs_tuners_ins.sampling_rate = def_sampling_rate; + if (!usr_sampling_rate) + dbs_tuners_ins.sampling_rate = def_sampling_rate; + else if (usr_sampling_rate < MIN_SAMPLING_RATE) { + printk(KERN_WARNING "cpufreq/ondemand: " + "specified sampling rate too low, using %"PRIu64"\n", + MIN_SAMPLING_RATE); + dbs_tuners_ins.sampling_rate = MIN_SAMPLING_RATE; + } else if (usr_sampling_rate > MAX_SAMPLING_RATE) { + printk(KERN_WARNING "cpufreq/ondemand: " + "specified sampling rate too high, using %"PRIu64"\n", + MAX_SAMPLING_RATE); + dbs_tuners_ins.sampling_rate = MAX_SAMPLING_RATE; + } else + dbs_tuners_ins.sampling_rate = usr_sampling_rate; } dbs_timer_init(this_dbs_info); @@ -239,3 +257,55 @@ int cpufreq_governor_dbs(struct cpufreq_ } return 0; } + +void __init cpufreq_cmdline_parse(char *str) +{ + do { + char *val, *end = strchr(str, ','); + + if ( end ) + *end++ = '\0'; + val = strchr(str, '='); + if ( val ) + *val++ = '\0'; + + if ( !strcmp(str, "rate") && val ) + { + usr_sampling_rate = simple_strtoull(val, NULL, 0) * MICROSECS(1); + } + else if ( !strcmp(str, "threshold") && val ) + { + unsigned long tmp = simple_strtoul(val, NULL, 0); + + if ( tmp < MIN_FREQUENCY_UP_THRESHOLD ) + { + printk(XENLOG_WARNING "cpufreq/ondemand: " + "specified threshold too low, using %d\n", + MIN_FREQUENCY_UP_THRESHOLD); + tmp = MIN_FREQUENCY_UP_THRESHOLD; + } + else if ( tmp > MAX_FREQUENCY_UP_THRESHOLD ) + { + printk(XENLOG_WARNING "cpufreq/ondemand: " + "specified threshold too high, using %d\n", + MAX_FREQUENCY_UP_THRESHOLD); + tmp = MAX_FREQUENCY_UP_THRESHOLD; + } + dbs_tuners_ins.up_threshold = tmp; + } + else if ( !strcmp(str, "bias") && val ) + { + unsigned long tmp = simple_strtoul(val, NULL, 0); + + if ( tmp > 1000 ) + { + printk(XENLOG_WARNING "cpufreq/ondemand: " + "specified bias too high, using 1000\n"); + tmp = 1000; + } + dbs_tuners_ins.powersave_bias = tmp; + } + + str = end; + } while ( str ); +} --- a/xen/common/domain.c +++ b/xen/common/domain.c @@ -25,6 +25,7 @@ #include #include #include +#include #include #include #include @@ -41,16 +42,25 @@ boolean_param("dom0_vcpus_pin", opt_dom0 enum cpufreq_controller cpufreq_controller; static void __init setup_cpufreq_option(char *str) { + char *arg; + if ( !strcmp(str, "dom0-kernel") ) { xen_processor_pmbits &= ~XEN_PROCESSOR_PM_PX; cpufreq_controller = FREQCTL_dom0_kernel; opt_dom0_vcpus_pin = 1; + return; } - else if ( !strcmp(str, "xen") ) + + if ( (arg = strpbrk(str, ",:")) != NULL ) + *arg++ = '\0'; + + if ( !strcmp(str, "xen") ) { xen_processor_pmbits |= XEN_PROCESSOR_PM_PX; cpufreq_controller = FREQCTL_xen; + if ( arg && *arg ) + cpufreq_cmdline_parse(arg); } } custom_param("cpufreq", setup_cpufreq_option); --- a/xen/include/acpi/cpufreq/cpufreq.h +++ b/xen/include/acpi/cpufreq/cpufreq.h @@ -41,6 +41,8 @@ struct cpufreq_policy { }; extern struct cpufreq_policy xen_px_policy[NR_CPUS]; +void cpufreq_cmdline_parse(char *); + #define CPUFREQ_SHARED_TYPE_NONE (0) /* None */ #define CPUFREQ_SHARED_TYPE_HW (1) /* HW does needed coordination */ #define CPUFREQ_SHARED_TYPE_ALL (2) /* All dependent CPUs should set freq */