X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=fio.c;h=a394ff8c0d3045ad33e3fdc24f7e022a3e13bed0;hp=9594f80582266ce8a8a5794ea9c52fb02a6adb1d;hb=50070b5a3b37f013a2b0b968e1b8b34bffff8589;hpb=069c29183fb55f467ba841959ba6731f3e4f2132 diff --git a/fio.c b/fio.c index 9594f805..a394ff8c 100644 --- a/fio.c +++ b/fio.c @@ -160,9 +160,19 @@ static void cleanup_pending_aio(struct thread_data *td) list_for_each_safe(entry, n, &td->io_u_busylist) { io_u = list_entry(entry, struct io_u, list); - r = td->io_ops->cancel(td, io_u); - if (!r) + /* + * if the io_u isn't in flight, then that generally + * means someone leaked an io_u. complain but fix + * it up, so we don't stall here. + */ + if ((io_u->flags & IO_U_F_FLIGHT) == 0) { + log_err("fio: non-busy IO on busy list\n"); put_io_u(td, io_u); + } else { + r = td->io_ops->cancel(td, io_u); + if (!r) + put_io_u(td, io_u); + } } } @@ -553,9 +563,12 @@ static int init_io_u(struct thread_data *td) fill_rand_buf(io_u, max_bs); io_u->index = i; + io_u->flags = IO_U_F_FREE; list_add(&io_u->list, &td->io_u_freelist); } + io_u_init_timeout(); + return 0; } @@ -704,6 +717,7 @@ static void *thread_main(void *data) } fio_gettime(&td->epoch, NULL); + memcpy(&td->timeout_end, &td->epoch, sizeof(td->epoch)); getrusage(RUSAGE_SELF, &td->ts.ru_start); runtime[0] = runtime[1] = 0; @@ -810,6 +824,8 @@ static void reap_threads(int *nr_running, int *t_rate, int *m_rate) */ pending = cputhreads = 0; for_each_td(td, i) { + int flags; + /* * ->io_ops is NULL for a thread that has closed its * io engine @@ -817,55 +833,57 @@ static void reap_threads(int *nr_running, int *t_rate, int *m_rate) if (td->io_ops && td->io_ops->flags & FIO_CPUIO) cputhreads++; - if (td->runstate < TD_EXITED) { - /* - * check if someone quit or got killed in an unusual way - */ - ret = waitpid(td->pid, &status, WNOHANG); - if (ret < 0) - perror("waitpid"); - else if ((ret == td->pid) && WIFSIGNALED(status)) { + if (!td->pid || td->runstate == TD_REAPED) + continue; + + flags = WNOHANG; + if (td->runstate == TD_EXITED) + flags = 0; + + /* + * check if someone quit or got killed in an unusual way + */ + ret = waitpid(td->pid, &status, flags); + if (ret < 0 && !td->use_thread) { + if (errno == ECHILD) { + log_err("fio: pid=%d disappeared %d\n", td->pid, td->runstate); + td_set_runstate(td, TD_REAPED); + goto reaped; + } + perror("waitpid"); + } else if (ret == td->pid) { + if (WIFSIGNALED(status)) { int sig = WTERMSIG(status); log_err("fio: pid=%d, got signal=%d\n", td->pid, sig); td_set_runstate(td, TD_REAPED); goto reaped; } - } - - if (td->runstate != TD_EXITED) { - if (td->runstate < TD_RUNNING) - pending++; - - continue; - } - - if (td->error) - exit_value++; - - td_set_runstate(td, TD_REAPED); - - if (td->use_thread) { - long ret; - - if (pthread_join(td->thread, (void *) &ret)) - perror("thread_join"); - } else { - int status; - - ret = waitpid(td->pid, &status, 0); - if (ret < 0) - perror("waitpid"); - else if (WIFEXITED(status) && WEXITSTATUS(status)) { - if (!exit_value) - exit_value++; + if (WIFEXITED(status)) { + if (WEXITSTATUS(status) && !td->error) + td->error = WEXITSTATUS(status); + if (td->use_thread) { + long ret; + + if (pthread_join(td->thread, (void *) &ret)) + perror("pthread_join"); + } + td_set_runstate(td, TD_REAPED); + goto reaped; } } + /* + * thread is not dead, continue + */ + continue; reaped: (*nr_running)--; (*m_rate) -= td->ratemin; (*t_rate) -= td->rate; + + if (td->error) + exit_value++; } if (*nr_running == cputhreads && !pending) @@ -921,6 +939,8 @@ static void run_threads(void) init_disk_util(td); } + set_genesis_time(); + while (todo) { struct thread_data *map[MAX_JOBS]; struct timeval this_start; @@ -1079,6 +1099,8 @@ int main(int argc, char *argv[]) setup_log(&agg_io_log[DDIR_WRITE]); } + set_genesis_time(); + disk_util_timer_arm(); run_threads();