b=21571 stacksize and locking fixes for loadgen patch from umka

[fs/lustre-release.git] / lustre / utils / loadgen.c

diff --git a/lustre/utils/loadgen.c b/lustre/utils/loadgen.c
index 0f5f75b..20f51c5 100644 (file)
--- a/lustre/utils/loadgen.c
+++ b/lustre/utils/loadgen.c
@@ -16,8 +16,8 @@
  * in the LICENSE file that accompanied this code).
  *
  * You should have received a copy of the GNU General Public License
- * version 2 along with this program; If not, see [sun.com URL with a
- * copy of GPLv2].
+ * version 2 along with this program; If not, see
+ * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
  *
  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  * CA 95054 USA or visit www.sun.com if you need additional information or
@@ -58,13 +58,12 @@
 #include <sys/time.h>
 
 #include <lnet/lnetctl.h>
-#include "parser.h"
+#include <libcfs/libcfsutil.h>
 #include "obdctl.h"
 
 static char cmdname[512];
 static char target[64] = "";
 char nid[64] = "";
-static int live_threads = 0;
 static int sig_received = 0;
 static int o_verbose = 4; /* 0-5 */
 static int my_oss = 0;
@@ -78,7 +77,7 @@ static int jt_quit(int argc, char **argv) {
 static int loadgen_usage(int argc, char **argv)
 {
         if (argc == 1) {
-                fprintf(stderr, 
+                fprintf(stderr,
         "This is a test program used to simulate large numbers of\n"
         "clients.  The echo obds are used, so the obdecho module must\n"
         "be loaded.\n"
@@ -145,7 +144,9 @@ struct kid_t {
         int              k_dev;
 };
 
+static int live_threads = 0;
 static struct kid_t *kid_list = NULL;
+pthread_mutex_t m_list = PTHREAD_MUTEX_INITIALIZER;
 
 static struct kid_t *push_kid(int tnum)
 {
@@ -156,10 +157,12 @@ static struct kid_t *push_kid(int tnum)
                 return NULL;
         }
         kid->k_pthread = pthread_self();
+        pthread_mutex_lock(&m_list);
         kid->k_next = kid_list;
         kid->k_id = tnum;
         kid_list = kid;
         live_threads++;
+        pthread_mutex_unlock(&m_list);
         return kid;
 }
 
@@ -667,7 +670,7 @@ static void *run_one_child(void *threadvp)
 {
         struct kid_t *kid;
         char oname[10], ename[10];
-        int thread = (long)threadvp, dev;
+        int thread = (long)threadvp, dev = 0;
         int rc = 0, err;
 
         if (o_verbose > 2)
@@ -691,12 +694,14 @@ static void *run_one_child(void *threadvp)
 
         while(!(rc || sig_received)) {
                 pthread_mutex_lock(&m_trigger);
+               pthread_mutex_lock(&m_list);
                 waiting_count++;
                 if ((waiting_count == live_threads) && timer_on) {
                         report_perf();
                         timer_on = 0;
                         all_done = 1;
                 }
+               pthread_mutex_unlock(&m_list);
                 pthread_cond_wait(&cv_trigger, &m_trigger);
                 waiting_count--;
                 all_done = 0;
@@ -732,6 +737,19 @@ out:
         pthread_exit((void *)(long)rc);
 }
 
+/* 
+ * PTHREAD_STACK_MIN is 16K minimal stack for threads. This
+ * is stack consumed by one thread, which executes NULL procedure.
+ * We need some more here and 20k stack for one client thread
+ * is enough to not overflow. In same time it does not consume
+ * a lot of memory for large number of threads.
+ *
+ * 20K virtual clients will only consume 320M + 400M. Still to
+ * create this number of virtual clients we need to fix 8192
+ * OBDs limit.
+ */
+#define CLIENT_THREAD_STACK_SIZE (PTHREAD_STACK_MIN + (20 * 1024))
+
 static int loadgen_start_clients(int argc, char **argv)
 {
         int rc = 0, i, numt;
@@ -757,7 +775,7 @@ static int loadgen_start_clients(int argc, char **argv)
                         cmdname, rc, strerror(errno));
                 return -errno;
         }
-        pthread_attr_setstacksize (&attr, PTHREAD_STACK_MIN);
+        pthread_attr_setstacksize (&attr, CLIENT_THREAD_STACK_SIZE);
         pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE);
 
         numt += live_threads;
@@ -862,13 +880,17 @@ static int loadgen_write(int argc, char **argv)
         if (argc < 3 || argc > 4)
                 return CMD_HELP;
         threads = atoi(argv[1]);
+        pthread_mutex_lock(&m_list);
         if (threads > live_threads) {
+               pthread_mutex_unlock(&m_list);
                 fprintf(stderr, "requested %d threads but only %d are running. "
                         "Use 'start' to start some more.\n",
                         threads, live_threads);
                 return -EOVERFLOW;
+        } else {
+               pthread_mutex_unlock(&m_list);
         }
-        trigger(C_WRITE, threads, atoi(argv[2]), 
+        trigger(C_WRITE, threads, atoi(argv[2]),
                 (argc == 4) ? atoi(argv[3]) : 0);
         return 0;
 }
@@ -917,7 +939,7 @@ static int loadgen_start_echosrv(int argc, char **argv)
                         cmdname, rc);
                 goto clean;
         }
-  
+
         /* Create an OSS to handle the communications */
         /* attach ost OSS OSS_UUID */
         args[1] = "ost";
@@ -956,7 +978,7 @@ clean:
 
 static int loadgen_wait(int argc, char **argv)
 {
-        /* Give scripts a chance to start some threads */   
+        /* Give scripts a chance to start some threads */
         sleep(1);
         while (!all_done) {
                 sleep(1);
@@ -966,7 +988,7 @@ static int loadgen_wait(int argc, char **argv)
 
 static int loadgen_init(int argc, char **argv)
 {
-        char *args[3];
+        char *args[4];
         int rc;
 
         sprintf(cmdname, "%s", argv[0]);
@@ -996,7 +1018,7 @@ static int loadgen_init(int argc, char **argv)
 static int loadgen_exit()
 {
         int rc;
-        
+
         printf("stopping %d children\n", live_threads);
         kill_kids();
         rc = wait_for_threads();
@@ -1038,7 +1060,7 @@ static int loadgen_main(int argc, char **argv)
 
 out:
         obd_finalize(argc, argv);
-        return rc;
+        return rc < 0 ? -rc : rc;
 }
 
 #ifndef LIBLUSTRE_TEST