Solaris: Implement LWP enumeration (#768)

Squashed the following commits:

* Solaris: Get LWP enumeration working
* Solaris: Make showing and hiding of kernel threads behave
* Solaris: remove usage of lwpstatus that is no longer needed
* Solaris: no discrete access to parent proc structure needed
* Solaris: Restore runtime MaxPid detection after LWP changes
* Solaris: Workaround virtual PID signal issue by shadowing kill() with a macro
* Solaris: Fix unintention double-shifting of virtual PID for LWP enumeration
* Solaris: Add LWPID to default display since LWP enumeration is also default
* Solaris: use PAGE_SIZE_KB from Process.h instead of custom definition
* Solaris: stop LWP enumeration at 1023 LWPs per proc since that is all we can handle in the virtual PID

solaris/Platform.c

@@ -35,8 +35,11 @@ in the source distribution for its full text.
 #include "Action.h"
 #include "BatteryMeter.h"
 #include "SignalsPanel.h"
+#include <signal.h>
 #include <sys/mkdev.h>
 
+#define  kill(pid, signal) kill(pid / 1024, signal)
+
 extern ProcessFieldData Process_fields[];
 typedef struct var kvar_t;
 
@@ -91,7 +94,7 @@ const SignalItem Platform_signals[] = {
 
 const unsigned int Platform_numberOfSignals = sizeof(Platform_signals)/sizeof(SignalItem);
 
-ProcessField Platform_defaultFields[] = { PID, USER, PRIORITY, NICE, M_SIZE, M_RESIDENT, STATE, PERCENT_CPU, PERCENT_MEM, TIME, COMM, 0 };
+ProcessField Platform_defaultFields[] = { PID, LWPID, USER, PRIORITY, NICE, M_SIZE, M_RESIDENT, STATE, PERCENT_CPU, PERCENT_MEM, TIME, COMM, 0 };
 
 MeterClass* Platform_meterTypes[] = {
    &CPUMeter_class,

solaris/Platform.h

@@ -14,8 +14,11 @@ in the source distribution for its full text.
 #include "Action.h"
 #include "BatteryMeter.h"
 #include "SignalsPanel.h"
+#include <signal.h>
 #include <sys/mkdev.h>
 
+#define  kill(pid, signal) kill(pid / 1024, signal)
+
 extern ProcessFieldData Process_fields[];
 typedef struct var kvar_t;
 

solaris/SolarisProcess.c

@@ -29,7 +29,8 @@ typedef enum SolarisProcessFields {
    TASKID = 103,
    POOLID = 104,
    CONTID = 105,
-   LAST_PROCESSFIELD = 106,
+   LWPID = 106,
+   LAST_PROCESSFIELD = 107,
 } SolarisProcessField;
 
 
@@ -42,6 +43,10 @@ typedef struct SolarisProcess_ {
    projid_t   projid;
    poolid_t   poolid;
    ctid_t     contid;
+   bool       is_lwp;
+   pid_t      realpid;
+   pid_t      realppid;
+   pid_t      lwpid;
 } SolarisProcess;
 
 
@@ -96,6 +101,7 @@ ProcessFieldData Process_fields[] = {
    [TASKID] = { .name = "TASKID", .title = " TSKID ", .description = "Task ID", .flags = 0, },
    [POOLID] = { .name = "POOLID", .title = " POLID ", .description = "Pool ID", .flags = 0, },
    [CONTID] = { .name = "CONTID", .title = " CNTID ", .description = "Contract ID", .flags = 0, },
+   [LWPID] = { .name = "LWPID", .title = " LWPID ", .description = "LWP ID", .flags = 0, },
    [LAST_PROCESSFIELD] = { .name = "*** report bug! ***", .title = NULL, .description = NULL, .flags = 0, },
 };
 
@@ -107,6 +113,7 @@ ProcessPidColumn Process_pidColumns[] = {
    { .id = CONTID, .label = "CNTID" },
    { .id = PID, .label = "PID" },
    { .id = PPID, .label = "PPID" },
+   { .id = LWPID, .label = "LWPID" },
    { .id = TPGID, .label = "TPGID" },
    { .id = TGID, .label = "TGID" },
    { .id = PGRP, .label = "PGRP" },
@@ -148,6 +155,15 @@ void SolarisProcess_writeField(Process* this, RichString* str, ProcessField fiel
       }
       break;
    }
+   case PID: xSnprintf(buffer, n, Process_pidFormat, sp->realpid); break;
+   case PPID: xSnprintf(buffer, n, Process_pidFormat, sp->realppid); break;
+   case LWPID:{
+      if (sp->lwpid <= 0) {
+         xSnprintf(buffer, n, "    - ");   
+      } else {
+         xSnprintf(buffer, n, Process_pidFormat, sp->lwpid); break;
+      }
+   }; break;
    default:
       Process_writeField(this, str, field);
       return;
@@ -178,6 +194,12 @@ long SolarisProcess_compare(const void* v1, const void* v2) {
       return (p1->contid - p2->contid);
    case ZONE:
       return strcmp(p1->zname ? p1->zname : "global", p2->zname ? p2->zname : "global");
+   case PID:
+      return (p1->realpid - p2->realpid);
+   case PPID:
+      return (p1->realppid - p2->realppid);
+   case LWPID:
+      return (p1->lwpid - p2->lwpid);
    default:
       return Process_compare(v1, v2);
    }
@@ -186,8 +208,11 @@ long SolarisProcess_compare(const void* v1, const void* v2) {
 bool Process_isThread(Process* this) {
    SolarisProcess* fp = (SolarisProcess*) this;
 
-   if (fp->kernel == 1 )
+   if (fp->kernel == 1 ) {
       return 1;
-   else
+   } else if (fp->is_lwp) {
+      return 1;
+   } else {
       return 0;
    }
+}

solaris/SolarisProcess.h

@@ -21,7 +21,8 @@ typedef enum SolarisProcessFields {
    TASKID = 103,
    POOLID = 104,
    CONTID = 105,
-   LAST_PROCESSFIELD = 106,
+   LWPID = 106,
+   LAST_PROCESSFIELD = 107,
 } SolarisProcessField;
 
 
@@ -34,6 +35,10 @@ typedef struct SolarisProcess_ {
    projid_t   projid;
    poolid_t   poolid;
    ctid_t     contid;
+   bool       is_lwp;
+   pid_t      realpid;
+   pid_t      realppid;
+   pid_t      lwpid;
 } SolarisProcess;
 
 

solaris/SolarisProcessList.c

@@ -243,6 +243,144 @@ void ProcessList_delete(ProcessList* this) {
    free(this);
 }
 
+void ProcessList_enumerateLWPs(Process* proc, char* name, ProcessList* pl, struct timeval tv) {
+   Process *lwp;
+   SolarisProcess *slwp;
+   SolarisProcess *sproc = (SolarisProcess*) proc;
+   Settings* settings = pl->settings;
+   bool hideKernelThreads = settings->hideKernelThreads;
+   bool hideUserlandThreads = settings->hideUserlandThreads;
+   char lwpdir[MAX_NAME+1];
+   DIR* dir = NULL;
+   FILE* fp = NULL;
+   pid_t lwpid;
+   bool preExisting = false;   
+   char filename[MAX_NAME+1];
+   lwpsinfo_t _lwpsinfo;
+   prusage_t _lwprusage;
+   struct tm date;
+   xSnprintf(lwpdir, MAX_NAME, "%s/%s/lwp", PROCDIR, name);
+   struct dirent* entry;
+   char* lwpname;
+   bool haveUsage = false;
+
+   dir = opendir(lwpdir);
+   if (!dir) return;
+   while ((entry = readdir(dir)) != NULL) {
+      lwpname = entry->d_name;
+      // With 10 bits to spare, we can only list up to 1023 unique LWPs per process
+      if (atoi(lwpname) > 1023) break;
+      lwpid   = proc->pid + atoi(lwpname);
+      lwp     = ProcessList_getProcess(pl, lwpid, &preExisting, (Process_New) SolarisProcess_new);
+      slwp    = (SolarisProcess*) lwp;
+      xSnprintf(filename, MAX_NAME, "%s/%s/lwp/%s/lwpsinfo", PROCDIR, name, lwpname);
+      fp   = fopen(filename, "r");
+      if ( fp == NULL ) continue;
+      fread(&_lwpsinfo,sizeof(lwpsinfo_t),1,fp);
+      fclose(fp);
+      xSnprintf(filename, MAX_NAME, "%s/%s/lwp/%s/lwpusage", PROCDIR, name, lwpname);
+      fp   = fopen(filename, "r");
+      if ( fp != NULL ) {
+         haveUsage = true;
+         fread(&_lwprusage,sizeof(prusage_t),1,fp);
+         fclose(fp);
+      }
+      slwp->is_lwp = TRUE;
+
+      if (!preExisting) {
+         lwp->basenameOffset  = -1;
+         slwp->kernel         = sproc->kernel;
+         // Fake values used for sorting
+         lwp->pid             = lwpid;
+         lwp->ppid            = proc->pid;
+         lwp->tgid            = proc->pid;
+         // Corresponding real values used for display
+         slwp->realpid        = sproc->realpid;
+         slwp->realppid       = sproc->realpid;
+         slwp->lwpid          = atoi(lwpname);
+         slwp->zoneid         = sproc->zoneid;
+         lwp->tty_nr          = proc->tty_nr;
+         lwp->pgrp            = proc->pgrp;
+         lwp->percent_cpu     = ((uint16_t)_lwpsinfo.pr_pctcpu/(double)32768)*(double)100.0;
+         // Not tracked per thread
+         lwp->percent_mem     = (double)0.0;
+         lwp->st_uid          = proc->st_uid;
+         lwp->user            = UsersTable_getRef(pl->usersTable, lwp->st_uid);
+         lwp->nlwp            = 0;
+         lwp->session         = proc->session;
+         lwp->comm            = xStrdup(proc->comm);
+         lwp->commLen         = strnlen(proc->comm,PRFNSZ);
+         slwp->zname          = sproc->zname;
+         if (haveUsage) {
+            lwp->majflt       = _lwprusage.pr_majf;
+            lwp->minflt       = _lwprusage.pr_minf;
+         } else {
+            lwp->majflt       = 0;
+            lwp->minflt       = 0;
+         }
+         lwp->m_resident      = 0;
+         lwp->m_size          = 0;
+         lwp->priority        = _lwpsinfo.pr_pri;
+         lwp->nice            = _lwpsinfo.pr_nice;
+         lwp->processor       = _lwpsinfo.pr_onpro;
+         lwp->state           = _lwpsinfo.pr_sname;
+         lwp->time            = _lwpsinfo.pr_time.tv_sec;
+         slwp->taskid         = sproc->taskid;
+         slwp->projid         = sproc->projid;
+         slwp->poolid         = sproc->poolid;
+         slwp->contid         = sproc->contid;
+         lwp->starttime_ctime = _lwpsinfo.pr_start.tv_sec;
+         (void) localtime_r((time_t*) &lwp->starttime_ctime, &date);
+         strftime(lwp->starttime_show, 7, ((lwp->starttime_ctime > tv.tv_sec - 86400) ? "%R " : "%b%d "), &date);
+         ProcessList_add(pl, lwp);
+      } else {
+         slwp->zoneid         = sproc->zoneid;
+         lwp->pgrp            = proc->pgrp;
+         lwp->percent_cpu     = ((uint16_t)_lwpsinfo.pr_pctcpu/(double)32768)*(double)100.0;
+         // Not tracked per thread
+         lwp->percent_mem     = (double)0.0;
+         lwp->st_uid          = proc->st_uid;
+         lwp->user            = UsersTable_getRef(pl->usersTable, lwp->st_uid);
+         lwp->nlwp            = 0;
+         lwp->session         = proc->session;
+         lwp->comm            = xStrdup(proc->comm);
+         lwp->commLen         = strnlen(proc->comm,PRFNSZ);
+         slwp->zname          = sproc->zname;
+         if (haveUsage) {
+            lwp->majflt       = _lwprusage.pr_majf;
+            lwp->minflt       = _lwprusage.pr_minf;
+         }
+         lwp->m_resident      = 0;
+         lwp->m_size          = 0;
+         lwp->priority        = _lwpsinfo.pr_pri;
+         lwp->nice            = _lwpsinfo.pr_nice;
+         lwp->processor       = _lwpsinfo.pr_onpro;
+         lwp->state           = _lwpsinfo.pr_sname;
+         lwp->time            = _lwpsinfo.pr_time.tv_sec;
+         slwp->taskid         = sproc->taskid;
+         slwp->projid         = sproc->projid;
+         slwp->poolid         = sproc->poolid;
+         slwp->contid         = sproc->contid;
+      }
+      if (slwp->kernel) {
+         if(!hideKernelThreads) {
+            lwp->show = true;
+         } else {
+            lwp->show = false;
+         }
+      } else {
+         if(!hideUserlandThreads) {
+            lwp->show = true;
+         } else {
+            lwp->show = false;
+         }
+      }
+      lwp->updated = true;
+   }
+   closedir(dir);
+}
+
+
 void ProcessList_goThroughEntries(ProcessList* this) {
    SolarisProcessList* spl = (SolarisProcessList*) this;
    Settings* settings = this->settings;
@@ -254,7 +392,6 @@ void ProcessList_goThroughEntries(ProcessList* this) {
    int    pid;
    bool   preExisting = false;
    Process* proc = NULL;
-   Process* parent = NULL;
    SolarisProcess* sproc = NULL;
    psinfo_t _psinfo;
    pstatus_t _pstatus;
@@ -280,7 +417,6 @@ void ProcessList_goThroughEntries(ProcessList* this) {
       name = entry->d_name;
       pid = atoi(name);
       proc = ProcessList_getProcess(this, pid, &preExisting, (Process_New) SolarisProcess_new);
-      proc->tgid = parent ? parent->pid : pid;
       sproc = (SolarisProcess *) proc;
       xSnprintf(filename, MAX_NAME, "%s/%s/psinfo", PROCDIR, name);
       fp = fopen(filename, "r");
@@ -298,14 +434,17 @@ void ProcessList_goThroughEntries(ProcessList* this) {
       if ( fp == NULL ) continue;
       fread(&_prusage,sizeof(prusage_t),1,fp);
       fclose(fp);
-
+      sproc->is_lwp = FALSE;
-      double kb_per_page = ((double)PAGE_SIZE / (double)1024.0);
 
       if(!preExisting) {
-         sproc->kernel         = false;
+         // Fake PID values used for sorting, since Solaris LWPs lack unique PIDs
-         proc->pid             = _psinfo.pr_pid;
+         proc->pid             = (_psinfo.pr_pid * 1024);
-         proc->ppid            = _psinfo.pr_ppid;
+         proc->ppid            = (_psinfo.pr_ppid * 1024); 
-         proc->tgid            = _psinfo.pr_pid;
+         proc->tgid            = (_psinfo.pr_ppid * 1024);
+         // Corresponding real values used for display
+         sproc->realpid        = _psinfo.pr_pid;
+         sproc->realppid       = _psinfo.pr_ppid;
+         sproc->lwpid          = 0; 
          sproc->zoneid         = _psinfo.pr_zoneid;
          proc->tty_nr          = _psinfo.pr_ttydev;
          proc->pgrp            = _psinfo.pr_pgid;
@@ -323,8 +462,8 @@ void ProcessList_goThroughEntries(ProcessList* this) {
          sproc->zname          = SolarisProcessList_readZoneName(spl->kd,sproc);
          proc->majflt          = _prusage.pr_majf;
          proc->minflt          = _prusage.pr_minf; 
-         proc->m_resident      = (long)(((double)_psinfo.pr_rssize)/kb_per_page);
+         proc->m_resident      = _psinfo.pr_rssize/PAGE_SIZE_KB;
-         proc->m_size          = (long)(((double)_psinfo.pr_size)/kb_per_page);
+         proc->m_size          = _psinfo.pr_size/PAGE_SIZE_KB;
          proc->priority        = _psinfo.pr_lwp.pr_pri;
          proc->nice            = _psinfo.pr_lwp.pr_nice;
          proc->processor       = _psinfo.pr_lwp.pr_onpro;
@@ -335,11 +474,19 @@ void ProcessList_goThroughEntries(ProcessList* this) {
          sproc->poolid         = _psinfo.pr_poolid;
          sproc->contid         = _psinfo.pr_contract;
          proc->starttime_ctime = _psinfo.pr_start.tv_sec;
+         if ((sproc->realppid <= 0) && !(sproc->realpid <= 1)) {
+            sproc->kernel = true;
+         } else {
+            sproc->kernel = false;
+         }
          (void) localtime_r((time_t*) &proc->starttime_ctime, &date);
          strftime(proc->starttime_show, 7, ((proc->starttime_ctime > tv.tv_sec - 86400) ? "%R " : "%b%d "), &date); 
          ProcessList_add(this, proc);
       } else {
-         proc->ppid            = _psinfo.pr_ppid;
+         proc->ppid            = (_psinfo.pr_ppid * 1024);
+         proc->tgid            = (_psinfo.pr_ppid * 1024);
+         sproc->realppid       = _psinfo.pr_ppid;
+         sproc->lwpid          = 0; 
          sproc->zoneid         = _psinfo.pr_zoneid;
          // See note above about these percentages
          proc->percent_cpu     = ((uint16_t)_psinfo.pr_pctcpu/(double)32768)*(double)100.0;
@@ -353,8 +500,8 @@ void ProcessList_goThroughEntries(ProcessList* this) {
          sproc->zname          = SolarisProcessList_readZoneName(spl->kd,sproc);
          proc->majflt          = _prusage.pr_majf;
          proc->minflt          = _prusage.pr_minf;
-         proc->m_resident      = (long)(((double)_psinfo.pr_rssize)/kb_per_page); 
+         proc->m_resident      = _psinfo.pr_rssize/PAGE_SIZE_KB; 
-         proc->m_size          = (long)(((double)_psinfo.pr_size)/kb_per_page); 
+         proc->m_size          = _psinfo.pr_size/PAGE_SIZE_KB; 
          proc->priority        = _psinfo.pr_lwp.pr_pri;
          proc->nice            = _psinfo.pr_lwp.pr_nice;
          proc->processor       = _psinfo.pr_lwp.pr_onpro;
@@ -365,8 +512,11 @@ void ProcessList_goThroughEntries(ProcessList* this) {
          sproc->poolid         = _psinfo.pr_poolid;
          sproc->contid         = _psinfo.pr_contract;
       }
-      proc->show = !(hideKernelThreads && (_pstatus.pr_flags & PR_ISSYS));
+      if (proc->nlwp > 1) {
-      if (_pstatus.pr_flags & PR_ISSYS) {
+         ProcessList_enumerateLWPs(proc, name, this, tv);
+      }
+      proc->show = !(hideKernelThreads && sproc->kernel);
+      if (_pstatus.pr_flags & sproc->kernel) {
          if (hideKernelThreads) {
             addRunning = 0;
             addTotal   = 0;

solaris/SolarisProcessList.h

@@ -51,6 +51,8 @@ ProcessList* ProcessList_new(UsersTable* usersTable, Hashtable* pidWhiteList, ui
 
 void ProcessList_delete(ProcessList* this);
 
+void ProcessList_enumerateLWPs(Process* proc, char* name, ProcessList* pl, struct timeval tv);
+
 void ProcessList_goThroughEntries(ProcessList* this);