htop/dragonflybsd/DragonFlyBSDProcessList.c

534 lines
18 KiB
C
Raw Normal View History

/*
htop - DragonFlyBSDProcessList.c
(C) 2014 Hisham H. Muhammad
(C) 2017 Diederik de Groot
Released under the GNU GPLv2, see the COPYING file
in the source distribution for its full text.
*/
#include "ProcessList.h"
#include "DragonFlyBSDProcessList.h"
#include "DragonFlyBSDProcess.h"
2020-10-03 20:00:27 +00:00
#include "Macros.h"
#include <unistd.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/sysctl.h>
#include <sys/user.h>
#include <err.h>
#include <fcntl.h>
#include <limits.h>
#include <string.h>
#include <sys/param.h>
static int MIB_hw_physmem[2];
static int MIB_vm_stats_vm_v_page_count[4];
static int pageSize;
static int pageSizeKb;
static int MIB_vm_stats_vm_v_wire_count[4];
static int MIB_vm_stats_vm_v_active_count[4];
static int MIB_vm_stats_vm_v_cache_count[4];
static int MIB_vm_stats_vm_v_inactive_count[4];
static int MIB_vm_stats_vm_v_free_count[4];
static int MIB_vfs_bufspace[2];
static int MIB_kern_cp_time[2];
static int MIB_kern_cp_times[2];
static int kernelFScale;
ProcessList* ProcessList_new(UsersTable* usersTable, Hashtable* pidMatchList, uid_t userId) {
size_t len;
char errbuf[_POSIX2_LINE_MAX];
DragonFlyBSDProcessList* dfpl = xCalloc(1, sizeof(DragonFlyBSDProcessList));
ProcessList* pl = (ProcessList*) dfpl;
ProcessList_init(pl, Class(DragonFlyBSDProcess), usersTable, pidMatchList, userId);
// physical memory in system: hw.physmem
// physical page size: hw.pagesize
// usable pagesize : vm.stats.vm.v_page_size
len = 2; sysctlnametomib("hw.physmem", MIB_hw_physmem, &len);
len = sizeof(pageSize);
if (sysctlbyname("vm.stats.vm.v_page_size", &pageSize, &len, NULL, 0) == -1) {
pageSize = PAGE_SIZE;
pageSizeKb = PAGE_SIZE_KB;
} else {
pageSizeKb = pageSize / ONE_K;
}
// usable page count vm.stats.vm.v_page_count
// actually usable memory : vm.stats.vm.v_page_count * vm.stats.vm.v_page_size
len = 4; sysctlnametomib("vm.stats.vm.v_page_count", MIB_vm_stats_vm_v_page_count, &len);
len = 4; sysctlnametomib("vm.stats.vm.v_wire_count", MIB_vm_stats_vm_v_wire_count, &len);
len = 4; sysctlnametomib("vm.stats.vm.v_active_count", MIB_vm_stats_vm_v_active_count, &len);
len = 4; sysctlnametomib("vm.stats.vm.v_cache_count", MIB_vm_stats_vm_v_cache_count, &len);
len = 4; sysctlnametomib("vm.stats.vm.v_inactive_count", MIB_vm_stats_vm_v_inactive_count, &len);
len = 4; sysctlnametomib("vm.stats.vm.v_free_count", MIB_vm_stats_vm_v_free_count, &len);
len = 2; sysctlnametomib("vfs.bufspace", MIB_vfs_bufspace, &len);
int cpus = 1;
len = sizeof(cpus);
if (sysctlbyname("hw.ncpu", &cpus, &len, NULL, 0) != 0) {
cpus = 1;
}
size_t sizeof_cp_time_array = sizeof(unsigned long) * CPUSTATES;
len = 2; sysctlnametomib("kern.cp_time", MIB_kern_cp_time, &len);
dfpl->cp_time_o = xCalloc(cpus, sizeof_cp_time_array);
dfpl->cp_time_n = xCalloc(cpus, sizeof_cp_time_array);
len = sizeof_cp_time_array;
// fetch initial single (or average) CPU clicks from kernel
sysctl(MIB_kern_cp_time, 2, dfpl->cp_time_o, &len, NULL, 0);
// on smp box, fetch rest of initial CPU's clicks
if (cpus > 1) {
len = 2; sysctlnametomib("kern.cp_times", MIB_kern_cp_times, &len);
dfpl->cp_times_o = xCalloc(cpus, sizeof_cp_time_array);
dfpl->cp_times_n = xCalloc(cpus, sizeof_cp_time_array);
len = cpus * sizeof_cp_time_array;
sysctl(MIB_kern_cp_times, 2, dfpl->cp_times_o, &len, NULL, 0);
}
pl->cpuCount = MAXIMUM(cpus, 1);
if (cpus == 1 ) {
dfpl->cpus = xRealloc(dfpl->cpus, sizeof(CPUData));
} else {
// on smp we need CPUs + 1 to store averages too (as kernel kindly provides that as well)
dfpl->cpus = xRealloc(dfpl->cpus, (pl->cpuCount + 1) * sizeof(CPUData));
}
len = sizeof(kernelFScale);
if (sysctlbyname("kern.fscale", &kernelFScale, &len, NULL, 0) == -1) {
//sane default for kernel provided CPU percentage scaling, at least on x86 machines, in case this sysctl call failed
kernelFScale = 2048;
}
dfpl->kd = kvm_openfiles(NULL, "/dev/null", NULL, 0, errbuf);
if (dfpl->kd == NULL) {
errx(1, "kvm_open: %s", errbuf);
}
return pl;
}
void ProcessList_delete(ProcessList* this) {
const DragonFlyBSDProcessList* dfpl = (DragonFlyBSDProcessList*) this;
if (dfpl->kd) kvm_close(dfpl->kd);
if (dfpl->jails) {
Hashtable_delete(dfpl->jails);
}
free(dfpl->cp_time_o);
free(dfpl->cp_time_n);
free(dfpl->cp_times_o);
free(dfpl->cp_times_n);
free(dfpl->cpus);
ProcessList_done(this);
free(this);
}
static inline void DragonFlyBSDProcessList_scanCPUTime(ProcessList* pl) {
const DragonFlyBSDProcessList* dfpl = (DragonFlyBSDProcessList*) pl;
int cpus = pl->cpuCount; // actual CPU count
int maxcpu = cpus; // max iteration (in case we have average + smp)
int cp_times_offset;
assert(cpus > 0);
size_t sizeof_cp_time_array;
unsigned long *cp_time_n; // old clicks state
unsigned long *cp_time_o; // current clicks state
unsigned long cp_time_d[CPUSTATES];
double cp_time_p[CPUSTATES];
// get averages or single CPU clicks
sizeof_cp_time_array = sizeof(unsigned long) * CPUSTATES;
sysctl(MIB_kern_cp_time, 2, dfpl->cp_time_n, &sizeof_cp_time_array, NULL, 0);
// get rest of CPUs
if (cpus > 1) {
// on smp systems DragonFlyBSD kernel concats all CPU states into one long array in
// kern.cp_times sysctl OID
// we store averages in dfpl->cpus[0], and actual cores after that
maxcpu = cpus + 1;
sizeof_cp_time_array = cpus * sizeof(unsigned long) * CPUSTATES;
sysctl(MIB_kern_cp_times, 2, dfpl->cp_times_n, &sizeof_cp_time_array, NULL, 0);
}
for (int i = 0; i < maxcpu; i++) {
if (cpus == 1) {
// single CPU box
cp_time_n = dfpl->cp_time_n;
cp_time_o = dfpl->cp_time_o;
} else {
if (i == 0 ) {
// average
cp_time_n = dfpl->cp_time_n;
cp_time_o = dfpl->cp_time_o;
} else {
// specific smp cores
cp_times_offset = i - 1;
cp_time_n = dfpl->cp_times_n + (cp_times_offset * CPUSTATES);
cp_time_o = dfpl->cp_times_o + (cp_times_offset * CPUSTATES);
}
}
// diff old vs new
unsigned long long total_o = 0;
unsigned long long total_n = 0;
unsigned long long total_d = 0;
for (int s = 0; s < CPUSTATES; s++) {
cp_time_d[s] = cp_time_n[s] - cp_time_o[s];
total_o += cp_time_o[s];
total_n += cp_time_n[s];
}
// totals
total_d = total_n - total_o;
if (total_d < 1 ) total_d = 1;
// save current state as old and calc percentages
for (int s = 0; s < CPUSTATES; ++s) {
cp_time_o[s] = cp_time_n[s];
cp_time_p[s] = ((double)cp_time_d[s]) / ((double)total_d) * 100;
}
CPUData* cpuData = &(dfpl->cpus[i]);
cpuData->userPercent = cp_time_p[CP_USER];
cpuData->nicePercent = cp_time_p[CP_NICE];
cpuData->systemPercent = cp_time_p[CP_SYS];
cpuData->irqPercent = cp_time_p[CP_INTR];
cpuData->systemAllPercent = cp_time_p[CP_SYS] + cp_time_p[CP_INTR];
// this one is not really used, but we store it anyway
cpuData->idlePercent = cp_time_p[CP_IDLE];
}
}
static inline void DragonFlyBSDProcessList_scanMemoryInfo(ProcessList* pl) {
DragonFlyBSDProcessList* dfpl = (DragonFlyBSDProcessList*) pl;
// @etosan:
// memory counter relationships seem to be these:
// total = active + wired + inactive + cache + free
// htop_used (unavail to anybody) = active + wired
// htop_cache (for cache meter) = buffers + cache
// user_free (avail to procs) = buffers + inactive + cache + free
size_t len = sizeof(pl->totalMem);
//disabled for now, as it is always smaller than phycal amount of memory...
//...to avoid "where is my memory?" questions
//sysctl(MIB_vm_stats_vm_v_page_count, 4, &(pl->totalMem), &len, NULL, 0);
//pl->totalMem *= pageSizeKb;
sysctl(MIB_hw_physmem, 2, &(pl->totalMem), &len, NULL, 0);
pl->totalMem /= 1024;
sysctl(MIB_vm_stats_vm_v_active_count, 4, &(dfpl->memActive), &len, NULL, 0);
dfpl->memActive *= pageSizeKb;
sysctl(MIB_vm_stats_vm_v_wire_count, 4, &(dfpl->memWire), &len, NULL, 0);
dfpl->memWire *= pageSizeKb;
sysctl(MIB_vfs_bufspace, 2, &(pl->buffersMem), &len, NULL, 0);
pl->buffersMem /= 1024;
sysctl(MIB_vm_stats_vm_v_cache_count, 4, &(pl->cachedMem), &len, NULL, 0);
pl->cachedMem *= pageSizeKb;
pl->usedMem = dfpl->memActive + dfpl->memWire;
//currently unused, same as with arc, custom meter perhaps
//sysctl(MIB_vm_stats_vm_v_inactive_count, 4, &(dfpl->memInactive), &len, NULL, 0);
//sysctl(MIB_vm_stats_vm_v_free_count, 4, &(dfpl->memFree), &len, NULL, 0);
//pl->freeMem = dfpl->memInactive + dfpl->memFree;
//pl->freeMem *= pageSizeKb;
struct kvm_swap swap[16];
2020-10-03 20:00:27 +00:00
int nswap = kvm_getswapinfo(dfpl->kd, swap, ARRAYSIZE(swap), 0);
pl->totalSwap = 0;
pl->usedSwap = 0;
for (int i = 0; i < nswap; i++) {
pl->totalSwap += swap[i].ksw_total;
pl->usedSwap += swap[i].ksw_used;
}
pl->totalSwap *= pageSizeKb;
pl->usedSwap *= pageSizeKb;
pl->sharedMem = 0; // currently unused
}
char* DragonFlyBSDProcessList_readProcessName(kvm_t* kd, struct kinfo_proc* kproc, int* basenameEnd) {
char** argv = kvm_getargv(kd, kproc, 0);
if (!argv) {
return xStrdup(kproc->kp_comm);
}
int len = 0;
for (int i = 0; argv[i]; i++) {
len += strlen(argv[i]) + 1;
}
char* comm = xMalloc(len);
char* at = comm;
*basenameEnd = 0;
for (int i = 0; argv[i]; i++) {
at = stpcpy(at, argv[i]);
if (!*basenameEnd) {
*basenameEnd = at - comm;
}
*at = ' ';
at++;
}
at--;
*at = '\0';
return comm;
}
static inline void DragonFlyBSDProcessList_scanJails(DragonFlyBSDProcessList* dfpl) {
size_t len;
char *jls; /* Jail list */
char *curpos;
char *nextpos;
if (sysctlbyname("jail.list", NULL, &len, NULL, 0) == -1) {
fprintf(stderr, "initial sysctlbyname / jail.list failed\n");
exit(3);
}
retry:
if (len == 0)
return;
jls = xMalloc(len);
if (jls == NULL) {
fprintf(stderr, "xMalloc failed\n");
exit(4);
}
if (sysctlbyname("jail.list", jls, &len, NULL, 0) == -1) {
if (errno == ENOMEM) {
free(jls);
goto retry;
}
fprintf(stderr, "sysctlbyname / jail.list failed\n");
exit(5);
}
if (dfpl->jails) {
Hashtable_delete(dfpl->jails);
}
dfpl->jails = Hashtable_new(20, true);
curpos = jls;
while (curpos) {
int jailid;
char *str_hostname;
nextpos = strchr(curpos, '\n');
if (nextpos)
*nextpos++ = 0;
jailid = atoi(strtok(curpos, " "));
str_hostname = strtok(NULL, " ");
char *jname = (char *) (Hashtable_get(dfpl->jails, jailid));
if (jname == NULL) {
jname = xStrdup(str_hostname);
Hashtable_put(dfpl->jails, jailid, jname);
}
curpos = nextpos;
}
free(jls);
}
char* DragonFlyBSDProcessList_readJailName(DragonFlyBSDProcessList* dfpl, int jailid) {
char* hostname;
char* jname;
if (jailid != 0 && dfpl->jails && (hostname = (char *)Hashtable_get(dfpl->jails, jailid))) {
jname = xStrdup(hostname);
} else {
jname = xStrdup("-");
}
return jname;
}
void ProcessList_goThroughEntries(ProcessList* this, bool pauseProcessUpdate) {
DragonFlyBSDProcessList* dfpl = (DragonFlyBSDProcessList*) this;
Settings* settings = this->settings;
bool hideKernelThreads = settings->hideKernelThreads;
bool hideUserlandThreads = settings->hideUserlandThreads;
DragonFlyBSDProcessList_scanMemoryInfo(this);
DragonFlyBSDProcessList_scanCPUTime(this);
DragonFlyBSDProcessList_scanJails(dfpl);
// in pause mode only gather global data for meters (CPU/memory/...)
if (pauseProcessUpdate)
return;
int count = 0;
// TODO Kernel Threads seem to be skipped, need to figure out the correct flag
struct kinfo_proc* kprocs = kvm_getprocs(dfpl->kd, KERN_PROC_ALL | (!hideUserlandThreads ? KERN_PROC_FLAG_LWP : 0), 0, &count);
for (int i = 0; i < count; i++) {
struct kinfo_proc* kproc = &kprocs[i];
bool preExisting = false;
bool ATTR_UNUSED isIdleProcess = false;
// note: dragonflybsd kernel processes all have the same pid, so we misuse the kernel thread address to give them a unique identifier
Process* proc = ProcessList_getProcess(this, kproc->kp_ktaddr ? (pid_t)kproc->kp_ktaddr : kproc->kp_pid, &preExisting, (Process_New) DragonFlyBSDProcess_new);
DragonFlyBSDProcess* dfp = (DragonFlyBSDProcess*) proc;
proc->show = ! ((hideKernelThreads && Process_isKernelThread(dfp)) || (hideUserlandThreads && Process_isUserlandThread(proc)));
if (!preExisting) {
dfp->jid = kproc->kp_jailid;
if (kproc->kp_ktaddr && kproc->kp_flags & P_SYSTEM) {
// dfb kernel threads all have the same pid, so we misuse the kernel thread address to give them a unique identifier
proc->pid = (pid_t)kproc->kp_ktaddr;
dfp->kernel = 1;
} else {
proc->pid = kproc->kp_pid; // process ID
dfp->kernel = 0;
}
proc->ppid = kproc->kp_ppid; // parent process id
proc->tpgid = kproc->kp_tpgid; // tty process group id
2017-04-21 14:34:40 +00:00
//proc->tgid = kproc->kp_lwp.kl_tid; // thread group id
proc->tgid = kproc->kp_pid; // thread group id
proc->pgrp = kproc->kp_pgid; // process group id
proc->session = kproc->kp_sid;
proc->tty_nr = kproc->kp_tdev; // control terminal device number
proc->st_uid = kproc->kp_uid; // user ID
proc->processor = kproc->kp_lwp.kl_origcpu;
proc->starttime_ctime = kproc->kp_start.tv_sec;
proc->user = UsersTable_getRef(this->usersTable, proc->st_uid);
ProcessList_add((ProcessList*)this, proc);
proc->comm = DragonFlyBSDProcessList_readProcessName(dfpl->kd, kproc, &proc->basenameOffset);
dfp->jname = DragonFlyBSDProcessList_readJailName(dfpl, kproc->kp_jailid);
} else {
proc->processor = kproc->kp_lwp.kl_cpuid;
if(dfp->jid != kproc->kp_jailid) { // process can enter jail anytime
dfp->jid = kproc->kp_jailid;
free(dfp->jname);
dfp->jname = DragonFlyBSDProcessList_readJailName(dfpl, kproc->kp_jailid);
}
if (proc->ppid != kproc->kp_ppid) { // if there are reapers in the system, process can get reparented anytime
proc->ppid = kproc->kp_ppid;
}
if(proc->st_uid != kproc->kp_uid) { // some processes change users (eg. to lower privs)
proc->st_uid = kproc->kp_uid;
proc->user = UsersTable_getRef(this->usersTable, proc->st_uid);
}
if (settings->updateProcessNames) {
free(proc->comm);
proc->comm = DragonFlyBSDProcessList_readProcessName(dfpl->kd, kproc, &proc->basenameOffset);
}
}
proc->m_size = kproc->kp_vm_map_size / 1024 / pageSizeKb;
proc->m_resident = kproc->kp_vm_rssize;
proc->percent_mem = (proc->m_resident * PAGE_SIZE_KB) / (double)(this->totalMem) * 100.0;
proc->nlwp = kproc->kp_nthreads; // number of lwp thread
proc->time = (kproc->kp_swtime + 5000) / 10000;
proc->percent_cpu = 100.0 * ((double)kproc->kp_lwp.kl_pctcpu / (double)kernelFScale);
proc->percent_mem = 100.0 * (proc->m_resident * PAGE_SIZE_KB) / (double)(this->totalMem);
if (proc->percent_cpu > 0.1) {
// system idle process should own all CPU time left regardless of CPU count
if ( strcmp("idle", kproc->kp_comm) == 0 ) {
isIdleProcess = true;
}
}
if (kproc->kp_lwp.kl_pid != -1)
proc->priority = kproc->kp_lwp.kl_prio;
else
proc->priority = -kproc->kp_lwp.kl_tdprio;
switch(kproc->kp_lwp.kl_rtprio.type) {
case RTP_PRIO_REALTIME:
proc->nice = PRIO_MIN - 1 - RTP_PRIO_MAX + kproc->kp_lwp.kl_rtprio.prio;
break;
case RTP_PRIO_IDLE:
proc->nice = PRIO_MAX + 1 + kproc->kp_lwp.kl_rtprio.prio;
break;
case RTP_PRIO_THREAD:
proc->nice = PRIO_MIN - 1 - RTP_PRIO_MAX - kproc->kp_lwp.kl_rtprio.prio;
break;
default:
proc->nice = kproc->kp_nice;
break;
}
// would be nice if we could store multiple states in proc->state (as enum) and have writeField render them
switch (kproc->kp_stat) {
case SIDL: proc->state = 'I'; isIdleProcess = true; break;
case SACTIVE:
switch (kproc->kp_lwp.kl_stat) {
case LSSLEEP:
if (kproc->kp_lwp.kl_flags & LWP_SINTR) // interruptable wait short/long
if (kproc->kp_lwp.kl_slptime >= MAXSLP) {
proc->state = 'I';
isIdleProcess = true;
} else {
proc->state = 'S';
}
else if (kproc->kp_lwp.kl_tdflags & TDF_SINTR) // interruptable lwkt wait
proc->state = 'S';
else if (kproc->kp_paddr) // uninterruptable wait
proc->state = 'D';
else // uninterruptable lwkt wait
proc->state = 'B';
break;
case LSRUN:
if (kproc->kp_lwp.kl_stat == LSRUN) {
if (!(kproc->kp_lwp.kl_tdflags & (TDF_RUNNING | TDF_RUNQ)))
proc->state = 'Q';
else
proc->state = 'R';
}
break;
case LSSTOP:
proc->state = 'T';
break;
default:
proc->state = 'A';
break;
}
break;
case SSTOP: proc->state = 'T'; break;
case SZOMB: proc->state = 'Z'; break;
case SCORE: proc->state = 'C'; break;
default: proc->state = '?';
}
if (kproc->kp_flags & P_SWAPPEDOUT) {
proc->state = 'W';
}
if (kproc->kp_flags & P_TRACED) {
proc->state = 'T';
}
if (kproc->kp_flags & P_JAILED) {
proc->state = 'J';
}
2019-10-31 16:39:12 +00:00
if (Process_isKernelThread(dfp)) {
this->kernelThreads++;
}
this->totalTasks++;
if (proc->state == 'R')
this->runningTasks++;
proc->updated = true;
}
}