X-Git-Url: https://git.kernel.dk/?a=blobdiff_plain;f=t%2Fio_uring.c;h=363cba3eba86c9c2c80098edaa9c9791a93145a9;hb=7f125e7f3879d23e79bc2ef5eed678ddab3b5c70;hp=0c8058ba089ed997274005edb871e00e4324be90;hpb=48e698fa59d6258f989d56890851d7658566aefc;p=fio.git diff --git a/t/io_uring.c b/t/io_uring.c index 0c8058ba..363cba3e 100644 --- a/t/io_uring.c +++ b/t/io_uring.c @@ -22,9 +22,7 @@ #include "../arch/arch.h" #include "../lib/types.h" -#include "../os/io_uring.h" - -#define barrier() __asm__ __volatile__("": : :"memory") +#include "../os/linux/io_uring.h" #define min(a, b) ((a < b) ? (a) : (b)) @@ -46,9 +44,8 @@ struct io_cq_ring { }; #define DEPTH 128 - -#define BATCH_SUBMIT 64 -#define BATCH_COMPLETE 64 +#define BATCH_SUBMIT 32 +#define BATCH_COMPLETE 32 #define BS 4096 @@ -69,7 +66,6 @@ struct submitter { struct drand48_data rand; struct io_sq_ring sq_ring; struct io_uring_sqe *sqes; - struct iovec iovecs[DEPTH]; struct io_cq_ring cq_ring; int inflight; unsigned long reaps; @@ -83,43 +79,47 @@ struct submitter { struct file files[MAX_FDS]; unsigned nr_files; unsigned cur_file; + struct iovec iovecs[]; }; -static struct submitter submitters[1]; +static struct submitter *submitter; static volatile int finish; +static int depth = DEPTH; +static int batch_submit = BATCH_SUBMIT; +static int batch_complete = BATCH_COMPLETE; static int polled = 1; /* use IO polling */ static int fixedbufs = 1; /* use fixed user buffers */ +static int register_files = 1; /* use fixed files */ static int buffered = 0; /* use buffered IO, not O_DIRECT */ static int sq_thread_poll = 0; /* use kernel submission/poller thread */ static int sq_thread_cpu = -1; /* pin above thread to this CPU */ +static int do_nop = 0; /* no-op SQ ring commands */ static int io_uring_register_buffers(struct submitter *s) { - struct io_uring_register_buffers reg = { - .iovecs = s->iovecs, - .nr_iovecs = DEPTH - }; + if (do_nop) + return 0; return syscall(__NR_sys_io_uring_register, s->ring_fd, - IORING_REGISTER_BUFFERS, ®); + IORING_REGISTER_BUFFERS, s->iovecs, depth); } static int io_uring_register_files(struct submitter *s) { - struct io_uring_register_files reg; int i; + if (do_nop) + return 0; + s->fds = calloc(s->nr_files, sizeof(__s32)); for (i = 0; i < s->nr_files; i++) { s->fds[i] = s->files[i].real_fd; s->files[i].fixed_fd = i; } - reg.fds = s->fds; - reg.nr_fds = s->nr_files; return syscall(__NR_sys_io_uring_register, s->ring_fd, - IORING_REGISTER_FILES, ®); + IORING_REGISTER_FILES, s->fds, s->nr_files); } static int io_uring_setup(unsigned entries, struct io_uring_params *p) @@ -131,7 +131,7 @@ static int io_uring_enter(struct submitter *s, unsigned int to_submit, unsigned int min_complete, unsigned int flags) { return syscall(__NR_sys_io_uring_enter, s->ring_fd, to_submit, - min_complete, flags); + min_complete, flags, NULL, 0); } static int gettid(void) @@ -141,7 +141,7 @@ static int gettid(void) static unsigned file_depth(struct submitter *s) { - return (DEPTH + s->nr_files - 1) / s->nr_files; + return (depth + s->nr_files - 1) / s->nr_files; } static void init_io(struct submitter *s, unsigned index) @@ -151,6 +151,11 @@ static void init_io(struct submitter *s, unsigned index) struct file *f; long r; + if (do_nop) { + sqe->opcode = IORING_OP_NOP; + return; + } + if (s->nr_files == 1) { f = &s->files[0]; } else { @@ -159,6 +164,7 @@ static void init_io(struct submitter *s, unsigned index) s->cur_file++; if (s->cur_file == s->nr_files) s->cur_file = 0; + f = &s->files[s->cur_file]; } } f->pending_ios++; @@ -166,20 +172,25 @@ static void init_io(struct submitter *s, unsigned index) lrand48_r(&s->rand, &r); offset = (r % (f->max_blocks - 1)) * BS; - sqe->flags = IOSQE_FIXED_FILE; + if (register_files) { + sqe->flags = IOSQE_FIXED_FILE; + sqe->fd = f->fixed_fd; + } else { + sqe->flags = 0; + sqe->fd = f->real_fd; + } if (fixedbufs) { sqe->opcode = IORING_OP_READ_FIXED; - sqe->addr = s->iovecs[index].iov_base; + sqe->addr = (unsigned long) s->iovecs[index].iov_base; sqe->len = BS; sqe->buf_index = index; } else { sqe->opcode = IORING_OP_READV; - sqe->addr = &s->iovecs[index]; + sqe->addr = (unsigned long) &s->iovecs[index]; sqe->len = 1; sqe->buf_index = 0; } sqe->ioprio = 0; - sqe->fd = f->fixed_fd; sqe->off = offset; sqe->user_data = (unsigned long) f; } @@ -192,7 +203,7 @@ static int prep_more_ios(struct submitter *s, int max_ios) next_tail = tail = *ring->tail; do { next_tail++; - barrier(); + read_barrier(); if (next_tail == *ring->head) break; @@ -205,9 +216,9 @@ static int prep_more_ios(struct submitter *s, int max_ios) if (*ring->tail != tail) { /* order tail store with writes to sqes above */ - barrier(); + write_barrier(); *ring->tail = tail; - barrier(); + write_barrier(); } return prepped; } @@ -244,15 +255,19 @@ static int reap_events(struct submitter *s) do { struct file *f; - barrier(); + read_barrier(); if (head == *ring->tail) break; cqe = &ring->cqes[head & cq_ring_mask]; - f = (struct file *) cqe->user_data; - f->pending_ios--; - if (cqe->res != BS) { - printf("io: unexpected ret=%d\n", cqe->res); - return -1; + if (!do_nop) { + f = (struct file *) (uintptr_t) cqe->user_data; + f->pending_ios--; + if (cqe->res != BS) { + printf("io: unexpected ret=%d\n", cqe->res); + if (polled && cqe->res == -EOPNOTSUPP) + printf("Your filesystem doesn't support poll\n"); + return -1; + } } if (cqe->flags & IOCQE_FLAG_CACHEHIT) s->cachehit++; @@ -264,7 +279,7 @@ static int reap_events(struct submitter *s) s->inflight -= reaped; *ring->head = head; - barrier(); + write_barrier(); return reaped; } @@ -282,26 +297,31 @@ static void *submitter_fn(void *data) do { int to_wait, to_submit, this_reap, to_prep; - if (!prepped && s->inflight < DEPTH) { - to_prep = min(DEPTH - s->inflight, BATCH_SUBMIT); + if (!prepped && s->inflight < depth) { + to_prep = min(depth - s->inflight, batch_submit); prepped = prep_more_ios(s, to_prep); } s->inflight += prepped; submit_more: to_submit = prepped; submit: - if (s->inflight + BATCH_SUBMIT < DEPTH) + if (to_submit && (s->inflight + to_submit <= depth)) to_wait = 0; else - to_wait = min(s->inflight + to_submit, BATCH_COMPLETE); + to_wait = min(s->inflight + to_submit, batch_complete); /* * Only need to call io_uring_enter if we're not using SQ thread * poll, or if IORING_SQ_NEED_WAKEUP is set. */ if (!sq_thread_poll || (*ring->flags & IORING_SQ_NEED_WAKEUP)) { - ret = io_uring_enter(s, to_submit, to_wait, - IORING_ENTER_GETEVENTS); + unsigned flags = 0; + + if (to_wait) + flags = IORING_ENTER_GETEVENTS; + if ((*ring->flags & IORING_SQ_NEED_WAKEUP)) + flags |= IORING_ENTER_SQ_WAKEUP; + ret = io_uring_enter(s, to_submit, to_wait, flags); s->calls++; } @@ -314,9 +334,10 @@ submit: do { int r; r = reap_events(s); - if (r == -1) + if (r == -1) { + s->finish = 1; break; - else if (r > 0) + } else if (r > 0) this_reap += r; } while (sq_thread_poll && this_reap < to_wait); s->reaps += this_reap; @@ -358,7 +379,7 @@ submit: static void sig_int(int sig) { printf("Exiting on signal %d\n", sig); - submitters[0].finish = 1; + submitter->finish = 1; finish = 1; } @@ -382,15 +403,17 @@ static int setup_ring(struct submitter *s) memset(&p, 0, sizeof(p)); - if (polled) + if (polled && !do_nop) p.flags |= IORING_SETUP_IOPOLL; if (sq_thread_poll) { p.flags |= IORING_SETUP_SQPOLL; - if (sq_thread_cpu != -1) + if (sq_thread_cpu != -1) { p.flags |= IORING_SETUP_SQ_AFF; + p.sq_thread_cpu = sq_thread_cpu; + } } - fd = io_uring_setup(DEPTH, &p); + fd = io_uring_setup(depth, &p); if (fd < 0) { perror("io_uring_setup"); return 1; @@ -405,10 +428,12 @@ static int setup_ring(struct submitter *s) } } - ret = io_uring_register_files(s); - if (ret < 0) { - perror("io_uring_register_files"); - return 1; + if (register_files) { + ret = io_uring_register_files(s); + if (ret < 0) { + perror("io_uring_register_files"); + return 1; + } } ptr = mmap(0, p.sq_off.array + p.sq_entries * sizeof(__u32), @@ -441,32 +466,89 @@ static int setup_ring(struct submitter *s) return 0; } +static void file_depths(char *buf) +{ + struct submitter *s = submitter; + char *p; + int i; + + buf[0] = '\0'; + p = buf; + for (i = 0; i < s->nr_files; i++) { + struct file *f = &s->files[i]; + + if (i + 1 == s->nr_files) + p += sprintf(p, "%d", f->pending_ios); + else + p += sprintf(p, "%d, ", f->pending_ios); + } +} + +static void usage(char *argv) +{ + printf("%s [options] -- [filenames]\n" + " -d : IO Depth, default %d\n" + " -s : Batch submit, default %d\n" + " -c : Batch complete, default %d\n", + argv, DEPTH, BATCH_SUBMIT, BATCH_COMPLETE); + exit(0); +} + int main(int argc, char *argv[]) { - struct submitter *s = &submitters[0]; + struct submitter *s; unsigned long done, calls, reap, cache_hit, cache_miss; - int err, i, flags, fd; - struct rlimit rlim; + int err, i, flags, fd, opt; + char *fdepths; void *ret; - if (argc < 2) { - printf("%s: filename\n", argv[0]); + if (!do_nop && argc < 2) { + printf("%s: filename [options]\n", argv[0]); return 1; } + while ((opt = getopt(argc, argv, "d:s:c:h?")) != -1) { + switch (opt) { + case 'd': + depth = atoi(optarg); + break; + case 's': + batch_submit = atoi(optarg); + break; + case 'c': + batch_complete = atoi(optarg); + break; + case 'h': + case '?': + default: + usage(argv[0]); + break; + } + } + + submitter = malloc(sizeof(*submitter) + depth * sizeof(struct iovec)); + memset(submitter, 0, sizeof(*submitter) + depth * sizeof(struct iovec)); + s = submitter; + flags = O_RDONLY | O_NOATIME; if (!buffered) flags |= O_DIRECT; - i = 1; - while (i < argc) { - struct file *f = &s->files[s->nr_files]; + i = optind; + while (!do_nop && i < argc) { + struct file *f; + if (s->nr_files == MAX_FDS) { + printf("Max number of files (%d) reached\n", MAX_FDS); + break; + } fd = open(argv[i], flags); if (fd < 0) { perror("open"); return 1; } + + f = &s->files[s->nr_files]; f->real_fd = fd; if (get_file_size(f)) { printf("failed getting size of device/file\n"); @@ -483,16 +565,20 @@ int main(int argc, char *argv[]) i++; } - rlim.rlim_cur = RLIM_INFINITY; - rlim.rlim_max = RLIM_INFINITY; - if (setrlimit(RLIMIT_MEMLOCK, &rlim) < 0) { - perror("setrlimit"); - return 1; + if (fixedbufs) { + struct rlimit rlim; + + rlim.rlim_cur = RLIM_INFINITY; + rlim.rlim_max = RLIM_INFINITY; + if (setrlimit(RLIMIT_MEMLOCK, &rlim) < 0) { + perror("setrlimit"); + return 1; + } } arm_sig_int(); - for (i = 0; i < DEPTH; i++) { + for (i = 0; i < depth; i++) { void *buf; if (posix_memalign(&buf, BS, BS)) { @@ -509,10 +595,11 @@ int main(int argc, char *argv[]) return 1; } printf("polled=%d, fixedbufs=%d, buffered=%d", polled, fixedbufs, buffered); - printf(" QD=%d, sq_ring=%d, cq_ring=%d\n", DEPTH, *s->sq_ring.ring_entries, *s->cq_ring.ring_entries); + printf(" QD=%d, sq_ring=%d, cq_ring=%d\n", depth, *s->sq_ring.ring_entries, *s->cq_ring.ring_entries); pthread_create(&s->thread, NULL, submitter_fn, s); + fdepths = malloc(8 * s->nr_files); cache_hit = cache_miss = reap = calls = done = 0; do { unsigned long this_done = 0; @@ -542,9 +629,10 @@ int main(int argc, char *argv[]) ipc = (this_reap - reap) / (this_call - calls); } else rpc = ipc = -1; - printf("IOPS=%lu, IOS/call=%ld/%ld, inflight=%u (head=%u tail=%u), Cachehit=%0.2f%%\n", + file_depths(fdepths); + printf("IOPS=%lu, IOS/call=%ld/%ld, inflight=%u (%s), Cachehit=%0.2f%%\n", this_done - done, rpc, ipc, s->inflight, - *s->cq_ring.head, *s->cq_ring.tail, hit); + fdepths, hit); done = this_done; calls = this_call; reap = this_reap; @@ -554,5 +642,6 @@ int main(int argc, char *argv[]) pthread_join(s->thread, &ret); close(s->ring_fd); + free(fdepths); return 0; }