[PATCH] Further improve child exit reaping

[fio.git] / fio.c

diff --git a/fio.c b/fio.c
index bce13a72306fccc62d84020ee4d57a3a2d8d8722..e386f53c0259726cabfb388909d0e85c6e6a8a49 100644 (file)
--- a/fio.c
+++ b/fio.c
@@ -26,12 +26,11 @@
 #include <string.h>
 #include <signal.h>
 #include <time.h>
-#include <assert.h>
+#include <locale.h>
 #include <sys/stat.h>
 #include <sys/wait.h>
 #include <sys/ipc.h>
 #include <sys/shm.h>
-#include <sys/ioctl.h>
 #include <sys/mman.h>
 
 #include "fio.h"
@@ -48,6 +47,7 @@ int temp_stall_ts;
 
 static volatile int startup_sem;
 static volatile int fio_abort;
+static int exit_value;
 
 struct io_log *agg_io_log[2];
 
@@ -82,11 +82,6 @@ static void sig_handler(int sig)
                        disk_util_timer_arm();
                        print_thread_status();
                        break;
-               case SIGSEGV:
-                       fprintf(stderr, "fio: got segfault, aborting\n");
-                       terminate_threads(TERMINATE_ALL, 1);
-                       fio_abort = 1;
-                       exit(0);
                default:
                        printf("\nfio: terminating on signal %d\n", sig);
                        fflush(stdout);
@@ -388,7 +383,7 @@ static void do_io(struct thread_data *td)
 
        td_set_runstate(td, TD_RUNNING);
 
-       while (td->this_io_bytes[td->ddir] < td->io_size) {
+       while ((td->this_io_bytes[0] + td->this_io_bytes[1]) < td->io_size) {
                struct timespec *timeout;
                int min_evts = 0;
                struct io_u *io_u;
@@ -405,12 +400,21 @@ static void do_io(struct thread_data *td)
                        break;
 
                memcpy(&s, &io_u->start_time, sizeof(s));
-
+requeue:
                ret = td_io_queue(td, io_u);
                if (ret) {
-                       td_verror(td, io_u->error);
-                       put_io_u(td, io_u);
-                       break;
+                       if (ret > 0 && (io_u->xfer_buflen != io_u->resid) &&
+                           io_u->resid) {
+                               /*
+                                * short read/write. requeue.
+                                */
+                               io_u->xfer_buflen = io_u->resid;
+                               io_u->xfer_buf += ret;
+                               goto requeue;
+                       } else {
+                               put_io_u(td, io_u);
+                               break;
+                       }
                }
 
                add_slat_sample(td, io_u->ddir, mtime_since(&io_u->start_time, &io_u->issue_time));
@@ -684,8 +688,10 @@ static void *thread_main(void *data)
        if (td_io_init(td))
                goto err;
 
-       if (td->exec_prerun)
-               system(td->exec_prerun);
+       if (td->exec_prerun) {
+               if (system(td->exec_prerun) < 0)
+                       goto err;
+       }
 
        fio_gettime(&td->epoch, NULL);
        getrusage(RUSAGE_SELF, &td->ru_start);
@@ -740,40 +746,45 @@ static void *thread_main(void *data)
                finish_log(td, td->clat_log, "clat");
        if (td->write_iolog_file)
                write_iolog_close(td);
-       if (td->exec_postrun)
-               system(td->exec_postrun);
+       if (td->exec_postrun) {
+               if (system(td->exec_postrun) < 0)
+                       log_err("fio: postrun %s failed\n", td->exec_postrun);
+       }
 
        if (exitall_on_terminate)
                terminate_threads(td->groupid, 0);
 
 err:
+       if (td->error)
+               printf("fio: pid=%d, err=%d/%s\n", td->pid, td->error, td->verror);
        close_files(td);
        close_ioengine(td);
        cleanup_io_u(td);
        td_set_runstate(td, TD_EXITED);
-       return NULL;
-
+       return (void *) td->error;
 }
 
 /*
  * We cannot pass the td data into a forked process, so attach the td and
  * pass it to the thread worker.
  */
-static void *fork_main(int shmid, int offset)
+static int fork_main(int shmid, int offset)
 {
        struct thread_data *td;
-       void *data;
+       void *data, *ret;
 
        data = shmat(shmid, NULL, 0);
        if (data == (void *) -1) {
+               int __err = errno;
+
                perror("shmat");
-               return NULL;
+               return __err;
        }
 
        td = data + offset * sizeof(struct thread_data);
-       thread_main(td);
+       ret = thread_main(td);
        shmdt(data);
-       return NULL;
+       return (int) ret;
 }
 
 /*
@@ -803,6 +814,9 @@ static void reap_threads(int *nr_running, int *t_rate, int *m_rate)
                        continue;
                }
 
+               if (td->error)
+                       exit_value++;
+
                td_set_runstate(td, TD_REAPED);
 
                if (td->use_thread) {
@@ -810,8 +824,15 @@ static void reap_threads(int *nr_running, int *t_rate, int *m_rate)
 
                        if (pthread_join(td->thread, (void *) &ret))
                                perror("thread_join");
-               } else
-                       waitpid(td->pid, NULL, 0);
+               } else {
+                       int status;
+
+                       waitpid(td->pid, &status, 0);
+                       if (WIFEXITED(status) && WEXITSTATUS(status)) {
+                               if (!exit_value)
+                                       exit_value++;
+                       }
+               }
 
                (*nr_running)--;
                (*m_rate) -= td->ratemin;
@@ -861,6 +882,9 @@ static void run_threads(void)
                 * client data interspersed on disk
                 */
                if (setup_files(td)) {
+                       exit_value++;
+                       if (td->error)
+                               log_err("fio: pid=%d, err=%d/%s\n", td->pid, td->error, td->verror);
                        td_set_runstate(td, TD_REAPED);
                        todo--;
                }
@@ -917,8 +941,9 @@ static void run_threads(void)
                                if (fork())
                                        fio_sem_down(&startup_sem);
                                else {
-                                       fork_main(shm_id, i);
-                                       exit(0);
+                                       int ret = fork_main(shm_id, i);
+
+                                       exit(ret);
                                }
                        }
                }
@@ -997,6 +1022,13 @@ int main(int argc, char *argv[])
 {
        long ps;
 
+       /*
+        * We need locale for number printing, if it isn't set then just
+        * go with the US format.
+        */
+       if (!getenv("LC_NUMERIC"))
+               setlocale(LC_NUMERIC, "en_US");
+
        if (parse_options(argc, argv))
                return 1;
 
@@ -1030,5 +1062,5 @@ int main(int argc, char *argv[])
                }
        }
 
-       return 0;
+       return exit_value;
 }