Fix usleep() bug causing excessive host CPU consumption:

The root cause of the bug was in my incorrect #define of MAX_TOD_UPDATE_USECS in featall.h, which was #defined to 1000000 (one million) instead of 999999, thereby causing the TXF rubato_thread to do a usleep for 1000000 (one million) microseconds, which, being invalid, was causing it to not sleep at all and instead return immediately with an error, causing it to loop continuously without sleeping, consuming all available host CPU due to the timer threads being the highest priority thread in Hercules.

comm3705.c also had a similar bug wherein it was using "min(1000000,delay)" for one of its usleep().

In addition to fixing the previously mentioned bug, I have also introduced a new USLEEP() macro which calls a new "herc_usleep()" function that will report any usleep() error that might occur. (I purposely did not code it to fix (hide) the problem but rather to just report it so that the caller can thus be fixed to not make invalid usleep calls.)

This should close GitHub Issue #589 "Auxiliary threads eating up CPU time" as well as the long running "CPU utilization on Raspberry Pi" thread that took place in the main Hercules group recently.

Author: Fish-Git

cckddasd.c

@@ -310,7 +310,7 @@ int             rc, i;                  /* Return code, Loop index   */
     while (cckd->ras)
     {
         release_lock(&cckdblk.ralock);
-        usleep(1);
+        USLEEP(1);
         obtain_lock(&cckdblk.ralock);
     }
     release_lock(&cckdblk.ralock);

cckddasd64.c

@@ -140,7 +140,7 @@ int             rc, i;                  /* Return code, Loop index   */
     while (cckd->ras)
     {
         release_lock(&cckdblk.ralock);
-        usleep(1);
+        USLEEP(1);
         obtain_lock(&cckdblk.ralock);
     }
     release_lock(&cckdblk.ralock);

cgibin.c

@@ -280,7 +280,7 @@ int     msgcount = 22;
         panel_command(command);
         // Wait a bit before proceeding in case
         // the command issues a lot of messages
-        usleep(50000);
+        USLEEP(50000);
     }
 
     if((value = cgi_variable(webblk,"msgcount")))
@@ -1623,7 +1623,7 @@ void cgibin_api_v1_syslog(WEBBLK *webblk)
         panel_command(command);
         // Wait a bit before proceeding in case
         // the command issues a lot of messages
-        usleep(50000);
+        USLEEP(50000);
         hprintf(webblk->sock,"{\"command\": \"%s\",",command);
     } 
     else

channel.c

@@ -338,7 +338,7 @@ typedef struct PREFETCH PREFETCH;
 #define IODELAY(_dev) \
 do { \
   if (sysblk.iodelay > 0 && (_dev)->devchar[10] == 0x20) \
-    usleep(sysblk.iodelay); \
+    USLEEP(sysblk.iodelay); \
 } while(0)
 #else
 #define IODELAY(_dev)
@@ -4151,7 +4151,7 @@ do {                                                                   \
 #ifdef FEATURE_S370_CHANNEL
             if (dev->devtype == 0x2703)
                 if (dev->commadpt->lnctl == COMMADPT_LNCTL_ASYNC)
-                    usleep(5000);
+                    USLEEP(5000);
 #endif
 
 #if DEBUG_DUMP

cmdtab.h

@@ -1828,7 +1828,8 @@
   "When the z/Arch Transactional-Execution Facility *is* installed and\n"       \
   "enabled the minimum and default intervals are 200 and 400 microseconds.\n"   \
   "\n"                                                                          \
-  "The maximum allowed interval is 1000000 microseconds (one second).\n"        \
+  "The maximum allowed interval is "QSTR( MAX_TOD_UPDATE_USECS )" microseconds (one microsecond\n" \
+  "less than one second).\n"                                        \
   "\n"                                                                          \
   "Also note that due to host system limitations and/or design, some\n"         \
   "hosts may round up and/or coalesce short microsecond intervals to a\n"       \

comm3705.c

@@ -1158,7 +1158,7 @@ static void *telnet_thread(void *vca)
         ca->hangup = 0;
         for (;;)
         {
-            usleep(50000);
+            USLEEP(50000);
             if (ca->hangup)
                 break;
             /* read_socket has changed from 3.04 to 3.06 - we need old way */
@@ -1224,7 +1224,7 @@ static void *commadpt_thread(void *vca)
         release_lock(&ca->lock);
         if(ca->ackspeed == 0) delay = 50000 + (ca->unack_attn_count * 100000);         /* Max's reliable algorithm      */
         else delay = (ca->unack_attn_count * ca->unack_attn_count + 1) * ca->ackspeed; /* much faster but TCAM hates it */
-        usleep(min(1000000,delay));                                                    /* go to sleep, max. 1 second    */
+        USLEEP(min((int)(ONE_MILLION-1),delay));                                       /* go to sleep, max. 1 second    */
         obtain_lock(&ca->lock);
         make_sna_requests2(ca);
         make_sna_requests3(ca);

commadpt.c

@@ -1474,7 +1474,7 @@ static void *commadpt_thread(void *vca)
                 if(ca->inbfr.havedata || ca->eol_flag)
                 {
                     if (ca->term == COMMADPT_TERM_2741) {
-                        usleep(10000);
+                        USLEEP(10000);
                     }
                     ca->curpending=COMMADPT_PEND_IDLE;
                     signal_condition(&ca->ipc);

console.c

@@ -3632,7 +3632,7 @@ int prev_rlen3270;
             {
                 // "COMM: pselect() failed: %s"
                 CONERROR( HHC90508, "D", strerror( select_errno ));
-                usleep( 50000 ); // (wait a bit; maybe it'll fix itself??)
+                USLEEP( 50000 ); // (wait a bit; maybe it'll fix itself??)
             }
             continue;
 
@@ -3696,7 +3696,7 @@ int prev_rlen3270;
                 {
                     // "COMM: accept() failed: %s"
                     CONERROR( HHC90509, "D", strerror( accept_errno ));
-                    usleep( 50000 ); // (wait a bit; maybe it'll fix itself??)
+                    USLEEP( 50000 ); // (wait a bit; maybe it'll fix itself??)
                 }
                 continue;
             }

ctc_ctci.c

@@ -1092,7 +1092,7 @@ static void*  CTCI_ReadThread( void* arg )
             // Don't use sched_yield() here; use an actual non-dispatchable
             // delay instead so as to allow another [possibly lower priority]
             // thread to 'read' (remove) some packet(s) from our frame buffer.
-            usleep( CTC_DELAY_USECS );  // (wait a bit before retrying...)
+            USLEEP( CTC_DELAY_USECS );  // (wait a bit before retrying...)
         }
     }
 

ctc_lcs.c

@@ -1691,7 +1691,7 @@ static void  UpdatePortStarted( int bStarted, DEVBLK* pDEVBLK, PLCSPORT pLCSPORT
     release_lock( &pLCSPORT->PortEventLock );
 
     PTT_DEBUG( "UPDTPORT pause 150", 000, pDEVBLK->devnum, pLCSPORT->bPort );
-    usleep( 150*1000 );
+    USLEEP( 150*1000 );
 }
 
 // ====================================================================
@@ -2191,7 +2191,7 @@ static void LCS_EnqueueReplyFrame( PLCSDEV pLCSDEV, PLCSCMDHDR pReply, size_t iS
         // Wait for LCS_Read to empty the buffer...
 
         ASSERT( ENOBUFS == errno );
-        usleep( CTC_DELAY_USECS );
+        USLEEP( CTC_DELAY_USECS );
     }
     PTT_TIMING( "af repNQ", 0, iSize, 0 );
     PTT_DEBUG( "ENQ RepFrame EXIT ", pReply->bCmdCode, pDEVBLK->devnum, bPort );
@@ -2798,7 +2798,7 @@ static void LCS_EnqueueEthFrame( PLCSPORT pLCSPORT, PLCSDEV pLCSDEV, BYTE* pData
         // Wait for LCS_Read to empty the buffer...
 
         ASSERT( ENOBUFS == errno );
-        usleep( CTC_DELAY_USECS );
+        USLEEP( CTC_DELAY_USECS );
     }
     PTT_TIMING( "af enqueue", 0, iSize, 0 );
     PTT_DEBUG( "ENQ EthFrame EXIT ", 000, pDEVBLK->devnum, bPort );
@@ -4150,7 +4150,7 @@ static void*  LCS_AttnThread( void* arg)
                 {
 
                     // Wait a small (but increasing) amount of time.
-                    usleep(interval);
+                    USLEEP(interval);
 
 //??                // is there still something in our frame buffer?
 //??                if (!pLCSDEV->fDataPending && !pLCSDEV->fReplyPending)