diff options
Diffstat (limited to 'freebsd/FreeBSDProcessList.c')
-rw-r--r-- | freebsd/FreeBSDProcessList.c | 316 |
1 files changed, 287 insertions, 29 deletions
diff --git a/freebsd/FreeBSDProcessList.c b/freebsd/FreeBSDProcessList.c index f1b7a191..e8d6f137 100644 --- a/freebsd/FreeBSDProcessList.c +++ b/freebsd/FreeBSDProcessList.c @@ -23,53 +23,160 @@ in the source distribution for its full text. #include <sys/param.h> #include <sys/jail.h> #include <sys/uio.h> +#include <sys/resource.h> #define JAIL_ERRMSGLEN 1024 char jail_errmsg[JAIL_ERRMSGLEN]; typedef struct CPUData_ { - unsigned long long int totalTime; - unsigned long long int totalPeriod; + + double userPercent; + double nicePercent; + double systemPercent; + double irqPercent; + double idlePercent; + double systemAllPercent; + } CPUData; typedef struct FreeBSDProcessList_ { ProcessList super; kvm_t* kd; + int zfsArcEnabled; + + unsigned long long int memWire; + unsigned long long int memActive; + unsigned long long int memInactive; + unsigned long long int memFree; + unsigned long long int memZfsArc; + + CPUData* cpus; + unsigned long *cp_time_o; + unsigned long *cp_time_n; + + unsigned long *cp_times_o; + unsigned long *cp_times_n; + } FreeBSDProcessList; }*/ + +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_hw_physmem[2]; +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_kstat_zfs_misc_arcstats_size[5]; + +static int MIB_kern_cp_time[2]; +static int MIB_kern_cp_times[2]; +static int kernelFScale; -static int pageSizeKb; ProcessList* ProcessList_new(UsersTable* usersTable, Hashtable* pidWhiteList, uid_t userId) { FreeBSDProcessList* fpl = calloc(1, sizeof(FreeBSDProcessList)); ProcessList* pl = (ProcessList*) fpl; ProcessList_init(pl, Class(FreeBSDProcess), usersTable, pidWhiteList, userId); + size_t len; + + // 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); + + len = sizeof(fpl->memZfsArc); + if (sysctlbyname("kstat.zfs.misc.arcstats.size", &fpl->memZfsArc, &len, + NULL, 0) == 0 && fpl->memZfsArc != 0) { + sysctlnametomib("kstat.zfs.misc.arcstats.size", MIB_kstat_zfs_misc_arcstats_size, &len); + fpl->zfsArcEnabled = 1; + } else { + fpl->zfsArcEnabled = 0; + } + + + int smp = 0; + len = sizeof(smp); + + if (sysctlbyname("kern.smp.active", &smp, &len, NULL, 0) != 0 || len != sizeof(smp)) { + smp = 0; + } + int cpus = 1; - size_t sizeof_cpus = sizeof(cpus); - int err = sysctlbyname("hw.ncpu", &cpus, &sizeof_cpus, NULL, 0); - if (err) cpus = 1; + len = sizeof(cpus); + + if (smp) { + int err = sysctlbyname("kern.smp.cpus", &cpus, &len, NULL, 0); + if (err) cpus = 1; + } else { + cpus = 1; + } + + size_t sizeof_cp_time_array = sizeof(unsigned long) * CPUSTATES; + len = 2; sysctlnametomib("kern.cp_time", MIB_kern_cp_time, &len); + fpl->cp_time_o = calloc(cpus, sizeof_cp_time_array); + fpl->cp_time_n = calloc(cpus, sizeof_cp_time_array); + len = sizeof_cp_time_array; + + // fetch intial single (or average) CPU clicks from kernel + sysctl(MIB_kern_cp_time, 2, fpl->cp_time_o, &len, NULL, 0); + + // on smp box, fetch rest of intial CPU's clicks + if (cpus > 1) { + len = 2; sysctlnametomib("kern.cp_times", MIB_kern_cp_times, &len); + fpl->cp_times_o = calloc(cpus, sizeof_cp_time_array); + fpl->cp_times_n = calloc(cpus, sizeof_cp_time_array); + len = cpus * sizeof_cp_time_array; + sysctl(MIB_kern_cp_times, 2, fpl->cp_times_o, &len, NULL, 0); + } + pl->cpuCount = MAX(cpus, 1); - fpl->cpus = realloc(fpl->cpus, cpus * sizeof(CPUData)); - for (int i = 0; i < cpus; i++) { - fpl->cpus[i].totalTime = 1; - fpl->cpus[i].totalPeriod = 1; + if (cpus == 1 ) { + fpl->cpus = realloc(fpl->cpus, sizeof(CPUData)); + } else { + // on smp we need CPUs + 1 to store averages too (as kernel kindly provides that as well) + fpl->cpus = realloc(fpl->cpus, (pl->cpuCount + 1) * sizeof(CPUData)); } - size_t 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_cache_count", MIB_vm_stats_vm_v_cache_count, &len); - len = 2; sysctlnametomib("hw.physmem", MIB_hw_physmem, &len); - pageSizeKb = PAGE_SIZE_KB; + + len = sizeof(kernelFScale); + if (sysctlbyname("kern.fscale", &kernelFScale, &len, NULL, 0) == -1) { + //sane default for kernel provded CPU precentage scaling, at least on x86 machines, in case this sysctl call failed + kernelFScale = 2048; + } fpl->kd = kvm_open(NULL, "/dev/null", NULL, 0, NULL); assert(fpl->kd); @@ -81,22 +188,149 @@ void ProcessList_delete(ProcessList* this) { const FreeBSDProcessList* fpl = (FreeBSDProcessList*) this; if (fpl->kd) kvm_close(fpl->kd); + if (fpl->cp_time_o != NULL) free(fpl->cp_time_o); + if (fpl->cp_time_n != NULL) free(fpl->cp_time_n); + if (fpl->cp_times_o != NULL) free(fpl->cp_times_o); + if (fpl->cp_times_n != NULL) free(fpl->cp_times_n); + ProcessList_done(this); free(this); } -static inline void FreeBSDProcessList_scanMemoryInfo(ProcessList* pl) { +static inline void FreeBSDProcessList_scanCPUTime(ProcessList* pl) { const FreeBSDProcessList* fpl = (FreeBSDProcessList*) 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 long total_o = 0; + unsigned long long total_n = 0; + unsigned long long total_d = 0; + 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, fpl->cp_time_n, &sizeof_cp_time_array, NULL, 0); + + // get rest of CPUs + if (cpus > 1) { + // on smp systems FreeBSD kernel concats all CPU states into one long array in + // kern.cp_times sysctl OID + // we store averages in fpl->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, fpl->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 = fpl->cp_time_n; + cp_time_o = fpl->cp_time_o; + } else { + if (i == 0 ) { + // average + cp_time_n = fpl->cp_time_n; + cp_time_o = fpl->cp_time_o; + } else { + // specific smp cores + cp_times_offset = i - 1; + cp_time_n = fpl->cp_times_n + (cp_times_offset * CPUSTATES); + cp_time_o = fpl->cp_times_o + (cp_times_offset * CPUSTATES); + } + } + + // diff old vs new + 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 = &(fpl->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 FreeBSDProcessList_scanMemoryInfo(ProcessList* pl) { + FreeBSDProcessList* fpl = (FreeBSDProcessList*) 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 + // + // with ZFS ARC situation becomes bit muddled, as ARC behaves like "user_free" + // and belongs into cache, but is reported as wired by kernel + // + // htop_used = active + (wired - arc) + // htop_cache = buffers + cache + arc 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_wire_count, 4, &(pl->usedMem), &len, NULL, 0); - pl->usedMem *= pageSizeKb; - pl->freeMem = pl->totalMem - pl->usedMem; + + sysctl(MIB_vm_stats_vm_v_active_count, 4, &(fpl->memActive), &len, NULL, 0); + fpl->memActive *= pageSizeKb; + + sysctl(MIB_vm_stats_vm_v_wire_count, 4, &(fpl->memWire), &len, NULL, 0); + fpl->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; + if (fpl->zfsArcEnabled) { + len = sizeof(fpl->memZfsArc); + sysctl(MIB_kstat_zfs_misc_arcstats_size, 5, &(fpl->memZfsArc), &len , NULL, 0); + fpl->memZfsArc /= 1024; + fpl->memWire -= fpl->memZfsArc; + pl->cachedMem += fpl->memZfsArc; + // maybe when we learn how to make custom memory meter + // we could do custom arc breakdown? + } + + pl->usedMem = fpl->memActive + fpl->memWire; + + //currently unused, same as with arc, custom meter perhaps + //sysctl(MIB_vm_stats_vm_v_inactive_count, 4, &(fpl->memInactive), &len, NULL, 0); + //sysctl(MIB_vm_stats_vm_v_free_count, 4, &(fpl->memFree), &len, NULL, 0); + //pl->freeMem = fpl->memInactive + fpl->memFree; + //pl->freeMem *= pageSizeKb; + struct kvm_swap swap[16]; int nswap = kvm_getswapinfo(fpl->kd, swap, sizeof(swap)/sizeof(swap[0]), 0); pl->totalSwap = 0; @@ -109,7 +343,6 @@ static inline void FreeBSDProcessList_scanMemoryInfo(ProcessList* pl) { pl->usedSwap *= pageSizeKb; pl->sharedMem = 0; // currently unused - pl->buffersMem = 0; // not exposed to userspace } char* FreeBSDProcessList_readProcessName(kvm_t* kd, struct kinfo_proc* kproc, int* basenameEnd) { @@ -186,20 +419,21 @@ void ProcessList_goThroughEntries(ProcessList* this) { bool hideUserlandThreads = settings->hideUserlandThreads; FreeBSDProcessList_scanMemoryInfo(this); + FreeBSDProcessList_scanCPUTime(this); + int cpus = this->cpuCount; int count = 0; struct kinfo_proc* kprocs = kvm_getprocs(fpl->kd, KERN_PROC_ALL, 0, &count); for (int i = 0; i < count; i++) { struct kinfo_proc* kproc = &kprocs[i]; - bool preExisting = false; + bool isIdleProcess = false; Process* proc = ProcessList_getProcess(this, kproc->ki_pid, &preExisting, (Process_New) FreeBSDProcess_new); FreeBSDProcess* fp = (FreeBSDProcess*) proc; proc->show = ! ((hideKernelThreads && Process_isKernelThread(fp)) || (hideUserlandThreads && Process_isUserlandThread(proc))); - if (!preExisting) { fp->jid = kproc->ki_jid; proc->pid = kproc->ki_pid; @@ -207,6 +441,7 @@ void ProcessList_goThroughEntries(ProcessList* this) { fp->kernel = 1; else fp->kernel = 0; + proc->ppid = kproc->ki_ppid; proc->tpgid = kproc->ki_tpgid; proc->tgid = kproc->ki_pid; proc->session = kproc->ki_sid; @@ -220,27 +455,51 @@ void ProcessList_goThroughEntries(ProcessList* this) { fp->jname = FreeBSDProcessList_readJailName(kproc); } else { if(fp->jid != kproc->ki_jid) { + // proces can enter jail anytime fp->jid = kproc->ki_jid; free(fp->jname); fp->jname = FreeBSDProcessList_readJailName(kproc); } + if (proc->ppid != kproc->ki_ppid) { + // if there are reapers in the system, proces can get reparented anytime + proc->ppid = kproc->ki_ppid; + } + if(proc->st_uid != kproc->ki_uid) { + // some proceses change users (eg. to lower privs) + proc->st_uid = kproc->ki_uid; + proc->user = UsersTable_getRef(this->usersTable, proc->st_uid); + } if (settings->updateProcessNames) { free(proc->comm); proc->comm = FreeBSDProcessList_readProcessName(fpl->kd, kproc, &proc->basenameOffset); } } - proc->ppid = kproc->ki_ppid; - - proc->m_size = kproc->ki_size / pageSizeKb / 1000; - proc->m_resident = kproc->ki_rssize; // * pageSizeKb; + // from FreeBSD source /src/usr.bin/top/machine.c + proc->m_size = kproc->ki_size / 1024; + proc->m_resident = kproc->ki_rssize * pageSizeKb; proc->nlwp = kproc->ki_numthreads; proc->time = (kproc->ki_runtime + 5000) / 10000; + proc->percent_cpu = 100.0 * ((double)kproc->ki_pctcpu / (double)kernelFScale); + if (proc->percent_cpu > 0.1) { + // system idle process should own all CPU time left regardless of CPU count + if ( strcmp("idle", kproc->ki_comm) == 0 ) { + isIdleProcess = true; + } else { + if (cpus > 1) + proc->percent_cpu = proc->percent_cpu / (double) cpus; + } + } + if (isIdleProcess == false && proc->percent_cpu >= 99.8) { + // don't break formatting + proc->percent_cpu = 99.8; + } + proc->priority = kproc->ki_pri.pri_level - PZERO; if (strcmp("intr", kproc->ki_comm) == 0 && kproc->ki_flag & P_SYSTEM) { - proc->nice = 0; //@etosan: freebsd intr kernel process (not thread) has weird nice value + proc->nice = 0; //@etosan: intr kernel process (not thread) has weird nice value } else if (kproc->ki_pri.pri_class == PRI_TIMESHARE) { proc->nice = kproc->ki_nice - NZERO; } else if (PRI_IS_REALTIME(kproc->ki_pri.pri_class)) { @@ -249,7 +508,6 @@ void ProcessList_goThroughEntries(ProcessList* this) { proc->nice = PRIO_MAX + 1 + kproc->ki_pri.pri_level - PRI_MIN_IDLE; } - switch (kproc->ki_stat) { case SIDL: proc->state = 'I'; break; case SRUN: proc->state = 'R'; break; |