diff options
author | Jinkun Jang <jinkun.jang@samsung.com> | 2013-03-12 15:13:12 +0900 |
---|---|---|
committer | Jinkun Jang <jinkun.jang@samsung.com> | 2013-03-12 15:13:12 +0900 |
commit | c913e4871de39ddb56024c0b507b3a4c0399eac1 (patch) | |
tree | 849b9ab7ca5883d7908b6a297023c95104dfb560 /daemon/load.cpp | |
parent | 8701a5293e4e52a995b1ee0b683f234848d8d745 (diff) | |
download | icecream-c913e4871de39ddb56024c0b507b3a4c0399eac1.tar.gz icecream-c913e4871de39ddb56024c0b507b3a4c0399eac1.tar.bz2 icecream-c913e4871de39ddb56024c0b507b3a4c0399eac1.zip |
Tizen 2.1 base
Diffstat (limited to 'daemon/load.cpp')
-rw-r--r-- | daemon/load.cpp | 373 |
1 files changed, 373 insertions, 0 deletions
diff --git a/daemon/load.cpp b/daemon/load.cpp new file mode 100644 index 0000000..ce64e23 --- /dev/null +++ b/daemon/load.cpp @@ -0,0 +1,373 @@ +/* + Copyright (c) 1999, 2000 Chris Schlaeger <cs@kde.org> + Copyright (c) 2003 Stephan Kulow <coolo@kde.org> + + This program is free software; you can redistribute it and/or + modify it under the terms of version 2 of the GNU General Public + License as published by the Free Software Foundation. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. +*/ + +#include "config.h" +#include "load.h" +#include <unistd.h> +#include <stdio.h> +#include <math.h> +#include <logging.h> +#include <sys/types.h> +#include <sys/stat.h> +#include <fcntl.h> +#ifdef HAVE_SYS_PARAM_H +#include <sys/param.h> +#endif + +#ifdef HAVE_MACH_HOST_INFO_H +#define USE_MACH 1 +#elif !defined( __linux__ ) && !defined(__CYGWIN__) +#define USE_SYSCTL +#endif + +#ifdef USE_MACH +#include <mach/host_info.h> +#include <mach/mach_host.h> +#include <mach/mach_init.h> +#endif + +#ifdef HAVE_KINFO_H +#include <kinfo.h> +#endif + +#ifdef HAVE_DEVSTAT_H +#include <sys/resource.h> +#include <sys/sysctl.h> +#include <devstat.h> +#endif + +using namespace std; + +// what the kernel puts as ticks in /proc/stat +typedef unsigned long long load_t; + + +struct CPULoadInfo +{ + /* A CPU can be loaded with user processes, reniced processes and + * system processes. Unused processing time is called idle load. + * These variable store the percentage of each load type. */ + int userLoad; + int niceLoad; + int sysLoad; + int idleLoad; + + /* To calculate the loads we need to remember the tick values for each + * load type. */ + load_t userTicks; + load_t niceTicks; + load_t sysTicks; + load_t idleTicks; + load_t waitTicks; + + CPULoadInfo() { + userTicks = 0; + niceTicks = 0; + sysTicks = 0; + idleTicks = 0; + waitTicks = 0; + } +}; + +static void updateCPULoad( CPULoadInfo* load ) +{ + load_t totalTicks; + load_t currUserTicks, currSysTicks, currNiceTicks, currIdleTicks, currWaitTicks; + +#if defined(USE_SYSCTL) && defined(__DragonFly__) + static struct kinfo_cputime cp_time; + + kinfo_get_sched_cputime(&cp_time); + /* There is one more load type exported via this interface in DragonFlyBSD - + * interrupt load. But I think that we can do without it for our needs. */ + currUserTicks = cp_time.cp_user; + currNiceTicks = cp_time.cp_nice; + currSysTicks = cp_time.cp_sys; + currIdleTicks = cp_time.cp_idle; + /* It doesn't exist in DragonFlyBSD. */ + currWaitTicks = 0; + +#elif defined (USE_SYSCTL) + static int mibs[4] = { 0,0,0,0 }; + static size_t mibsize = 4; + unsigned long ticks[CPUSTATES]; + size_t mibdatasize = sizeof(ticks); + + if (mibs[0]==0) { + if (sysctlnametomib("kern.cp_time",mibs,&mibsize) < 0) { + load->userTicks = load->sysTicks = load->niceTicks = load->idleTicks = 0; + load->userLoad = load->sysLoad = load->niceLoad = load->idleLoad = 0; + mibs[0]=0; + return; + } + } + if (sysctl(mibs,mibsize,&ticks,&mibdatasize,NULL,0) < 0) { + load->userTicks = load->sysTicks = load->niceTicks = load->idleTicks = 0; + load->userLoad = load->sysLoad = load->niceLoad = load->idleLoad = 0; + return; + } else { + currUserTicks = ticks[CP_USER]; + currNiceTicks = ticks[CP_NICE]; + currSysTicks = ticks[CP_SYS]; + currIdleTicks = ticks[CP_IDLE]; + } + +#elif defined( USE_MACH ) + host_cpu_load_info r_load; + + kern_return_t error; + mach_msg_type_number_t count; + + count = HOST_CPU_LOAD_INFO_COUNT; + mach_port_t port = mach_host_self(); + error = host_statistics(port, HOST_CPU_LOAD_INFO, + (host_info_t)&r_load, &count); + + if (error != KERN_SUCCESS) + return; + + currUserTicks = r_load.cpu_ticks[CPU_STATE_USER]; + currNiceTicks = r_load.cpu_ticks[CPU_STATE_NICE]; + currSysTicks = r_load.cpu_ticks[CPU_STATE_SYSTEM]; + currIdleTicks = r_load.cpu_ticks[CPU_STATE_IDLE]; + currWaitTicks = 0; + +#else + char buf[ 256 ]; + static int fd = -1; + + if ( fd < 0 ) { + if (( fd = open( "/proc/stat", O_RDONLY ) ) < 0 ) { + log_error() << "Cannot open file \'/proc/stat\'!\n" + "The kernel needs to be compiled with support\n" + "for /proc filesystem enabled!" << endl; + return; + } + fcntl(fd, F_SETFD, FD_CLOEXEC); + } + + lseek(fd, 0, SEEK_SET); + ssize_t n; + + while ( (n = read( fd, buf, sizeof(buf) -1 )) < 0 && errno == EINTR) + ; + + if ( n < 20 ) { + log_error() << "no enough data in /proc/stat?" << endl; + return; + } + buf[n] = 0; + + /* wait ticks only exist with Linux >= 2.6.0. treat as 0 otherwise */ + currWaitTicks = 0; + // sscanf( buf, "%*s %lu %lu %lu %lu %lu", &currUserTicks, &currNiceTicks, + sscanf( buf, "%*s %llu %llu %llu %llu %llu", &currUserTicks, &currNiceTicks, // RL modif + &currSysTicks, &currIdleTicks, &currWaitTicks ); +#endif + + totalTicks = ( currUserTicks - load->userTicks ) + + ( currSysTicks - load->sysTicks ) + + ( currNiceTicks - load->niceTicks ) + + ( currIdleTicks - load->idleTicks ) + + ( currWaitTicks - load->waitTicks ); + + if ( totalTicks > 10 ) { + load->userLoad = ( 1000 * ( currUserTicks - load->userTicks ) ) / totalTicks; + load->sysLoad = ( 1000 * ( currSysTicks - load->sysTicks ) ) / totalTicks; + load->niceLoad = ( 1000 * ( currNiceTicks - load->niceTicks ) ) / totalTicks; + load->idleLoad = ( 1000 - ( load->userLoad + load->sysLoad + load->niceLoad) ); + if ( load->idleLoad < 0 ) + load->idleLoad = 0; + } else { + load->userLoad = load->sysLoad = load->niceLoad = 0; + load->idleLoad = 1000; + } + + load->userTicks = currUserTicks; + load->sysTicks = currSysTicks; + load->niceTicks = currNiceTicks; + load->idleTicks = currIdleTicks; + load->waitTicks = currWaitTicks; +} + +#ifndef USE_SYSCTL +static unsigned long int scan_one( const char* buff, const char *key ) +{ + const char *b = strstr( buff, key ); + if ( !b ) + return 0; + unsigned long int val = 0; + if ( sscanf( b + strlen( key ), ": %lu", &val ) != 1 ) + return 0; + return val; +} +#endif + +static unsigned int calculateMemLoad( unsigned long int &NetMemFree ) +{ + unsigned long long MemFree = 0, Buffers = 0, Cached = 0; + +#ifdef USE_MACH + /* Get VM statistics. */ + vm_statistics_data_t vm_stat; + mach_msg_type_number_t count = sizeof(vm_stat) / sizeof(natural_t); + kern_return_t error = host_statistics(mach_host_self(), HOST_VM_INFO, + (host_info_t)&vm_stat, &count); + if (error != KERN_SUCCESS) + return 0; + + vm_size_t pagesize; + host_page_size(mach_host_self(), &pagesize); + + unsigned long long MemInactive = (unsigned long long) vm_stat.inactive_count * pagesize; + MemFree = (unsigned long long) vm_stat.free_count * pagesize; + + // blunt lie - but when's sche macht + Buffers = MemInactive; + +#elif defined( USE_SYSCTL ) + size_t len = sizeof (MemFree); + if ((sysctlbyname("vm.stats.vm.v_free_count", &MemFree, &len, NULL, 0) == -1) || !len) + MemFree = 0; /* Doesn't work under FreeBSD v2.2.x */ + + + len = sizeof (Buffers); + if ((sysctlbyname("vfs.bufspace", &Buffers, &len, NULL, 0) == -1) || !len) + Buffers = 0; /* Doesn't work under FreeBSD v2.2.x */ + + len = sizeof (Cached); + if ((sysctlbyname("vm.stats.vm.v_cache_count", &Cached, &len, NULL, 0) == -1) || !len) + Cached = 0; /* Doesn't work under FreeBSD v2.2.x */ +#else + /* The interesting information is definitely within the first 256 bytes */ + char buf[256]; + static int fd = -1; + + if ( fd < 0 ) { + if ( ( fd = open( "/proc/meminfo", O_RDONLY ) ) < 0 ) { + log_error() << "Cannot open file \'/proc/meminfo\'!\n" + "The kernel needs to be compiled with support\n" + "for /proc filesystem enabled!" << endl; + return 0; + } + fcntl(fd, F_SETFD, FD_CLOEXEC); + } + lseek (fd, 0, SEEK_SET); + ssize_t n; + while ((n = read( fd, buf, sizeof( buf ) -1 )) < 0 && errno == EINTR) + ; + if (n < 20) + return 0; + + buf[n] = '\0'; + MemFree = scan_one( buf, "MemFree" ); + Buffers = scan_one( buf, "Buffers" ); + Cached = scan_one( buf, "Cached" ); +#endif + + if ( Buffers > 50 * 1024 ) + Buffers -= 50 * 1024; + else + Buffers /= 2; + + if ( Cached > 50 * 1024 ) + Cached -= 50 * 1024; + else + Cached /= 2; + + NetMemFree = MemFree + Cached + Buffers; + if ( NetMemFree > 128 * 1024 ) + return 0; + else + return 1000 - ( NetMemFree * 1000 / ( 128 * 1024 ) ); +} + +// Load average calculation based on CALC_LOAD(), in the 2.6 Linux kernel +// oldVal - previous load avg. +// numJobs - current number of active jobs +// rate - update rate, in seconds (usually 60, 300, or 900) +// delta_t - time since last update, in seconds +double compute_load( double oldVal, unsigned int currentJobs, unsigned int rate, double delta_t ) +{ + double weight = 1.0 / exp( delta_t / rate ); + return oldVal * weight + currentJobs * (1.0 - weight); +} + +double getEpocTime() +{ + timeval tv; + gettimeofday( &tv, NULL ); + return (double) tv.tv_sec + (double) tv.tv_usec / 1000000.0; +} + +// Simulates getloadavg(), but only for specified number of jobs +// Note: this is stateful and not thread-safe! +// Also, it differs from getloadavg() in that its notion of load +// is only updated as often as it's called. +int fakeloadavg( double *p_result, int resultEntries, unsigned int currentJobs ) +{ + // internal state + static const int numLoads = 3; + static double loads[numLoads] = { 0.0, 0.0, 0.0 }; + static unsigned int rates[numLoads] = { 60, 300, 900 }; + static double lastUpdate = getEpocTime(); + + // First, update all state + double now = getEpocTime(); + double delta_t = std::max( now - lastUpdate, 0.0 ); // guard against user changing system time backwards + lastUpdate = now; + for (int l = 0; l < numLoads; l++) { + loads[l] = compute_load( loads[0], currentJobs, rates[l], delta_t ); + } + + // Then, return requested values + int numFilled = std::min( std::max( resultEntries, 0 ), numLoads ); + for (int n = 0; n < numFilled; n++) p_result[n] = loads[n]; + return numFilled; +} + +bool fill_stats( unsigned long &myidleload, unsigned long &myniceload, unsigned int &memory_fillgrade, StatsMsg *msg, unsigned int hint ) +{ + static CPULoadInfo load; + + updateCPULoad( &load ); + + myidleload = load.idleLoad; + myniceload = load.niceLoad; + + if ( msg ) { + unsigned long int MemFree = 0; + + memory_fillgrade = calculateMemLoad( MemFree ); + + double avg[3]; +#if HAVE_GETLOADAVG + getloadavg( avg, 3 ); + (void) hint; +#else + fakeloadavg( avg, 3, hint ); +#endif + msg->loadAvg1 = (load_t)( avg[0] * 1000 ); + msg->loadAvg5 = (load_t)( avg[1] * 1000 ); + msg->loadAvg10 = (load_t)( avg[2] * 1000 ); + + msg->freeMem = (load_t)( MemFree / 1024.0 + 0.5 ); + + } + return true; +} |