//return eventfd
int AIORead(std::string path,void *buf,int epfd=-1)
{
m_filefd = openFile(path);
if (-1 == m_filefd) {
return -1;
}
m_ctx = 0;
if (io_setup(8192, &m_ctx)) {
perror("io_setup");
return -1;
}
if (posix_memalign(&buf, ALIGN_SIZE, RD_WR_SIZE)) {
perror("posix_memalign");
return 5;
}
printf("buf: %p\n", buf);
for (i = 0, iocbp = iocbs; i < NUM_EVENTS; ++i, ++iocbp) {
iocbps[i] = &iocbp->iocb;
io_prep_pread(&iocbp->iocb, fd, buf, RD_WR_SIZE, i * RD_WR_SIZE);
io_set_eventfd(&iocbp->iocb, efd);
io_set_callback(&iocbp->iocb, aio_callback);
iocbp->nth_request = i + 1;
}
if (io_submit(ctx, NUM_EVENTS, iocbps) != NUM_EVENTS) {
perror("io_submit");
return 6;
}
}
static int aio_submit(struct iothread *t, http_conn_t h)
{
struct iocb *iocb;
size_t data_len;
off_t data_off;
int ret, fd;
data_len = http_conn_data(h, &fd, &data_off);
assert(data_len);
iocb = hgang_alloc(aio_iocbs);
if ( NULL == iocb )
return 0;
io_prep_sendfile(iocb, fd, data_len, data_off, http_conn_socket(h));
iocb->data = h;
io_set_eventfd(iocb, efd->fd);
ret = io_submit(aio_ctx, 1, &iocb);
if ( ret <= 0 ) {
errno = -ret;
fprintf(stderr, "io_submit: %s\n", os_err());
return 0;
}
dprintf("io_submit: sendfile: %zu bytes\n", data_len);
in_flight++;
return 1;
}
int aio_preadv(io_context_t ctx, struct iocb *iocb, int fd, const struct iovec *iov, int iovcnt,
off_t offset, int ev, void *param)
{
struct iocb *ios[1] = { iocb };
io_prep_preadv(iocb, fd, iov, iovcnt, offset);
io_set_eventfd(iocb, ev);
iocb->data = param;
return io_submit(ctx, 1, ios);
}
int aio_read(int fd, void *buffer, size_t length, off_t offset, aio_callback_t cb){
struct iocb iocb;
struct iocb *piocb = &iocb;
int resp;
io_prep_pread(&iocb, fd, buffer, length, offset);
iocb.data = cb;
io_set_eventfd(&iocb, aio_fd);
resp = io_submit(aio_ctx,1, &piocb);
return resp;
}
void AsyncIO::submit(Op* op) {
CHECK_EQ(op->state(), Op::State::INITIALIZED);
CHECK_LT(pending_, capacity_) << "too many pending requests";
initializeContext(); // on demand
iocb* cb = &op->iocb_;
cb->data = nullptr; // unused
if (pollFd_ != -1) {
io_set_eventfd(cb, pollFd_);
}
int rc = io_submit(ctx_, 1, &cb);
checkKernelError(rc, "AsyncIO: io_submit failed");
DCHECK_EQ(rc, 1);
op->start();
++pending_;
}
/*
* Test whether counter overflow is detected and handled correctly.
*
* It is not possible to directly overflow the counter using the
* write() syscall. Overflows occur when the counter is incremented
* from kernel space, in an irq context, when it is not possible to
* block the calling thread of execution.
*
* The AIO subsystem internally uses eventfd mechanism for
* notification of completion of read or write requests. In this test
* we trigger a counter overflow, by setting the counter value to the
* max possible value initially. When the AIO subsystem notifies
* through the eventfd counter, the counter overflows.
*
* NOTE: If the the counter starts from an initial value of 0, it will
* take decades for an overflow to occur. But since we set the initial
* value to the max possible counter value, we are able to cause it to
* overflow with a single increment.
*
* When the counter overflows, the following are tested
* 1. Check whether POLLERR event occurs in poll() for the eventfd.
* 2. Check whether readfd_set/writefd_set is set in select() for the
eventfd.
* 3. The counter value is UINT64_MAX.
*/
static int trigger_eventfd_overflow(int evfd, int *fd, io_context_t * ctx)
{
int ret;
struct iocb iocb;
struct iocb *iocbap[1];
static char buf[4 * 1024];
*ctx = 0;
ret = io_setup(16, ctx);
if (ret < 0) {
errno = -ret;
tst_resm(TINFO | TERRNO, "io_setup error");
return -1;
}
*fd = open("testfile", O_RDWR | O_CREAT, 0644);
if (*fd == -1) {
tst_resm(TINFO | TERRNO, "open(testfile) failed");
goto err_io_destroy;
}
ret = set_counter(evfd, UINT64_MAX - 1);
if (ret == -1) {
tst_resm(TINFO, "error setting counter to UINT64_MAX-1");
goto err_close_file;
}
io_prep_pwrite(&iocb, *fd, buf, sizeof(buf), 0);
io_set_eventfd(&iocb, evfd);
iocbap[0] = &iocb;
ret = io_submit(*ctx, 1, iocbap);
if (ret < 0) {
errno = -ret;
tst_resm(TINFO | TERRNO, "error submitting iocb");
goto err_close_file;
}
return 0;
err_close_file:
close(*fd);
err_io_destroy:
io_destroy(*ctx);
return -1;
}
BlockDriverAIOCB *laio_submit(BlockDriverState *bs, void *aio_ctx, int fd,
int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
BlockDriverCompletionFunc *cb, void *opaque, int type)
{
struct qemu_laio_state *s = aio_ctx;
struct qemu_laiocb *laiocb;
struct iocb *iocbs;
off_t offset = sector_num * 512;
laiocb = qemu_aio_get(&laio_pool, bs, cb, opaque);
if (!laiocb)
return NULL;
laiocb->nbytes = nb_sectors * 512;
laiocb->ctx = s;
laiocb->ret = -EINPROGRESS;
laiocb->is_read = (type == QEMU_AIO_READ);
laiocb->qiov = qiov;
iocbs = &laiocb->iocb;
switch (type) {
case QEMU_AIO_WRITE:
io_prep_pwritev(iocbs, fd, qiov->iov, qiov->niov, offset);
break;
case QEMU_AIO_READ:
io_prep_preadv(iocbs, fd, qiov->iov, qiov->niov, offset);
break;
/* Currently Linux kernel does not support other operations */
default:
fprintf(stderr, "%s: invalid AIO request type 0x%x.\n",
__func__, type);
goto out_free_aiocb;
}
io_set_eventfd(&laiocb->iocb, s->efd);
s->count++;
if (io_submit(s->ctx, 1, &iocbs) < 0)
goto out_dec_count;
return &laiocb->common;
out_dec_count:
s->count--;
out_free_aiocb:
qemu_aio_release(laiocb);
return NULL;
}
BlockAIOCB *laio_submit(BlockDriverState *bs, void *aio_ctx, int fd,
int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
BlockCompletionFunc *cb, void *opaque, int type)
{
struct qemu_laio_state *s = aio_ctx;
struct qemu_laiocb *laiocb;
struct iocb *iocbs;
off_t offset = sector_num * 512;
laiocb = qemu_aio_get(&laio_aiocb_info, bs, cb, opaque);
laiocb->nbytes = nb_sectors * 512;
laiocb->ctx = s;
laiocb->ret = -EINPROGRESS;
laiocb->is_read = (type == QEMU_AIO_READ);
laiocb->qiov = qiov;
iocbs = &laiocb->iocb;
switch (type) {
case QEMU_AIO_WRITE:
io_prep_pwritev(iocbs, fd, qiov->iov, qiov->niov, offset);
break;
case QEMU_AIO_READ:
io_prep_preadv(iocbs, fd, qiov->iov, qiov->niov, offset);
break;
/* Currently Linux kernel does not support other operations */
default:
fprintf(stderr, "%s: invalid AIO request type 0x%x.\n",
__func__, type);
goto out_free_aiocb;
}
io_set_eventfd(&laiocb->iocb, event_notifier_get_fd(&s->e));
if (!s->io_q.plugged) {
if (io_submit(s->ctx, 1, &iocbs) < 0) {
goto out_free_aiocb;
}
} else {
ioq_enqueue(s, iocbs);
}
return &laiocb->common;
out_free_aiocb:
qemu_aio_unref(laiocb);
return NULL;
}
int
aioperf_libaio_read(aioperf_io_task_t *io_task)
{
struct iocb *libaio_req = NULL;
aioperf_repository_t *repository = NULL;
aioperf_conf_info_t *conf_info = NULL;
unsigned long left_size = 0;
unsigned int count = 0;
repository = io_task->repository;
conf_info = repository->conf_info;
left_size = io_task->file_size - io_task->offset;
if (!(libaio_req = (struct iocb *)aioperf_memory_pool_alloc(repository->pool,
sizeof(struct iocb)))) {
printf("alloc error\n");
return AIOPERF_ERROR;
}
if (!(io_task->buf = (char *)aioperf_memory_memalign(512,
conf_info->buf_size))) {
printf("memory memalign error\n");
}
io_task->data = libaio_req;
if (left_size > conf_info->buf_size) {
count = conf_info->buf_size;
} else {
count = left_size;
}
io_prep_pread(libaio_req, io_task->fd, io_task->buf, count,
io_task->offset);
libaio_req->data = io_task;
io_set_eventfd(libaio_req, io_task->repository->signal_fd);
if (io_submit(*(io_context_t *)repository->mgr->data, 1, &libaio_req) < 0) {
printf("io_submit error\n");
}
return AIOPERF_OK;
}
static void do_io_async(void)
{
size_t n_files = sizeof(files) / sizeof(files[0]);
size_t i;
io_context_t ctx = 0;
struct iocb *iocb;
struct iocb **piocb;
int n_aio_ops, rc;
/* TODO 1 - allocate iocb and piocb */
iocb = (struct iocb *) malloc(n_files * sizeof(*iocb));
DIE(iocb == NULL, "malloc");
piocb = (struct iocb **) malloc(n_files * sizeof(*piocb));
DIE(piocb == NULL, "malloc");
/* TODO 1 - initialiaze iocb and piocb */
for (i = 0; i < n_files; i++) {
io_prep_pwrite(&iocb[i], fds[i], g_buffer, BUFSIZ, 0);
piocb[i] = &iocb[i];
#ifdef USE_EVENTFD
/* TODO 2 - set up eventfd notification */
io_set_eventfd(&iocb[i], efd);
#endif
}
/* TODO 1 - setup aio context */
rc = io_setup(n_files, &ctx);
DIE(rc < 0, "io_setup");
/* TODO 1 - submit aio */
n_aio_ops = io_submit(ctx, n_files, piocb);
DIE(n_aio_ops < 0, "io_submit");
/* wait for completion*/
wait_aio(ctx, n_files);
/* TODO 1 - destroy aio context */
io_destroy(ctx);
}
void AsyncIO::submit(Op* op) {
CHECK_EQ(op->state(), Op::State::INITIALIZED);
initializeContext(); // on demand
// We can increment past capacity, but we'll clean up after ourselves.
auto p = pending_.fetch_add(1, std::memory_order_acq_rel);
if (p >= capacity_) {
decrementPending();
throw std::range_error("AsyncIO: too many pending requests");
}
iocb* cb = &op->iocb_;
cb->data = nullptr; // unused
if (pollFd_ != -1) {
io_set_eventfd(cb, pollFd_);
}
int rc = io_submit(ctx_, 1, &cb);
if (rc < 0) {
decrementPending();
throwSystemErrorExplicit(-rc, "AsyncIO: io_submit failed");
}
submitted_++;
DCHECK_EQ(rc, 1);
op->start();
}
int main(int argc, char *argv[])
{
int efd, fd, epfd;
io_context_t ctx;
struct timespec tms;
struct io_event events[NUM_EVENTS];
struct custom_iocb iocbs[NUM_EVENTS];
struct iocb *iocbps[NUM_EVENTS];
struct custom_iocb *iocbp;
int i, j, r;
void *buf;
struct epoll_event epevent;
// TODO:Used to record event?
efd = eventfd(0, EFD_NONBLOCK | EFD_CLOEXEC);
if (efd == -1) {
perror("eventfd");
return 2;
}
// File Open Operation
fd = open(TEST_FILE, O_RDWR | O_CREAT | O_DIRECT, 0644);
if (fd == -1) {
perror("open");
return 3;
}
// Init the file size first.
ftruncate(fd, TEST_FILE_SIZE);
ctx = 0;
// 1. io_setup: Create an asynchronous I/O context capable of receiving
// at least nr_events.
// 2. 8192: It could receive at least 8192 events??
if (io_setup(8192, &ctx)) {
perror("io_setup");
return 4;
}
// Allocate RD_WR_SIZE bytes, and places the address of the allocated memory
// in *buf. The address of the allocated memory will be a multiple of ALIGN_SIZE,
if (posix_memalign(&buf, ALIGN_SIZE, RD_WR_SIZE)) {
perror("posix_memalign");
return 5;
}
printf("buf: %p\n", buf);
for (i = 0, iocbp = iocbs; i < NUM_EVENTS; ++i, ++iocbp) {
iocbps[i] = &iocbp->iocb;
// An inline convenience function designed to facilitate the initialization of the iocb for
// an asynchronous read operation. When the function finished, the iocbp->iocb will be:
// iocb->u.c.nbytes= RD_WR_SIZE
// iocb->aio_fildes = fd
// iocb->u.c.buf = buf
// iocb->u.c.offset = i * RD_WR_SIZE
io_prep_pread(&iocbp->iocb, fd, buf, RD_WR_SIZE, i * RD_WR_SIZE);
io_set_eventfd(&iocbp->iocb, efd);
// Set up io completion callback function.
io_set_callback(&iocbp->iocb, aio_callback);
iocbp->nth_request = i + 1;
}
// Submit asynchronous I/O blocks for processing. It will queue NUM_EVENTS
// I/O request blocks for processing in the AIO context ctx.
// iocbps is an array of nr AIO control blocks, which will be submitted to context ctx.
if (io_submit(ctx, NUM_EVENTS, iocbps) != NUM_EVENTS) {
perror("io_submit");
return 6;
}
// Open an epoll file descriptor ( epoll is I/O event notification facility).
// It will returns a file descriptor referring to the new epoll instance.
epfd = epoll_create(1);
if (epfd == -1) {
perror("epoll_create");
return 7;
}
epevent.events = EPOLLIN | EPOLLET;
epevent.data.ptr = NULL;
// epoll_ctl: Control interface for an epoll descriptor.
// EPOLL_CTL_ADD: the operation that will be performanced
// efd: The target file descriptor that the operation will be performed for.
// epevent: describle the object linked to the fle descriptor efd.
if (epoll_ctl(epfd, EPOLL_CTL_ADD, efd, &epevent)) {
perror("epoll_ctl");
return 8;
}
i = 0;
while (i < NUM_EVENTS) {
uint64_t finished_aio;
// epoll_wait: wait for an I/O event on an epoll file descriptor
// epfd: the file descriptor that the epoll instance refer to.
// epevent: points to the events that will be available for the caller.
// 1: at most 1 are returned by the function
// -1: specifies the minimum number of milliseconds that epoll_wait() will block.
// -1 causes epoll_wait to block indefinitely.
if (epoll_wait(epfd, &epevent, 1, -1) != 1) {
perror("epoll_wait");
return 9;
}
if (read(efd, &finished_aio, sizeof(finished_aio)) != sizeof(finished_aio)) {
perror("read");
return 10;
}
printf("finished io number: %"PRIu64"\n", finished_aio);
while (finished_aio > 0) {
tms.tv_sec = 0;
tms.tv_nsec = 0;
// read asynchronous I/O events from the completion queue.
r = io_getevents(ctx, 1, NUM_EVENTS, events, &tms);
if (r > 0) {
for (j = 0; j < r; ++j) {
((io_callback_t)(events[j].data))(ctx, events[j].obj, events[j].res, events[j].res2);
}
i += r;
finished_aio -= r;
}
}
}
close(epfd);
free(buf);
// destroy an asynchronous I/O context
io_destroy(ctx);
close(fd);
close(efd);
remove(TEST_FILE);
return 0;
}
int main(int argc, char *argv[])
{
int efd, fd, epfd;
io_context_t ctx;
struct timespec tms;
struct io_event events[NUM_EVENTS];
struct custom_iocb iocbs[NUM_EVENTS];
struct iocb *iocbps[NUM_EVENTS];
struct custom_iocb *iocbp;
int i, j, r;
void *buf;
struct epoll_event epevent;
efd = eventfd(0, EFD_NONBLOCK | EFD_CLOEXEC);
if (efd == -1) {
perror("eventfd");
return 2;
}
fd = open(TEST_FILE, O_RDWR | O_CREAT | O_DIRECT, 0644);
if (fd == -1) {
perror("open");
return 3;
}
ftruncate(fd, TEST_FILE_SIZE);
ctx = 0;
if (io_setup(8192, &ctx)) {
perror("io_setup");
return 4;
}
if (posix_memalign(&buf, ALIGN_SIZE, RD_WR_SIZE)) {
perror("posix_memalign");
return 5;
}
printf("buf: %p\n", buf);
for (i = 0, iocbp = iocbs; i < NUM_EVENTS; ++i, ++iocbp) {
iocbps[i] = &iocbp->iocb;
io_prep_pread(&iocbp->iocb, fd, buf, RD_WR_SIZE, i * RD_WR_SIZE);
io_set_eventfd(&iocbp->iocb, efd);
io_set_callback(&iocbp->iocb, aio_callback);
iocbp->nth_request = i + 1;
}
if (io_submit(ctx, NUM_EVENTS, iocbps) != NUM_EVENTS) {
perror("io_submit");
return 6;
}
epfd = epoll_create(1);
if (epfd == -1) {
perror("epoll_create");
return 7;
}
epevent.events = EPOLLIN | EPOLLET;
epevent.data.ptr = NULL;
if (epoll_ctl(epfd, EPOLL_CTL_ADD, efd, &epevent)) {
perror("epoll_ctl");
return 8;
}
i = 0;
while (i < NUM_EVENTS) {
uint64_t finished_aio;
if (epoll_wait(epfd, &epevent, 1, -1) != 1) {
perror("epoll_wait");
return 9;
}
if (read(efd, &finished_aio, sizeof(finished_aio)) != sizeof(finished_aio)) {
perror("read");
return 10;
}
printf("finished io number: %"PRIu64"\n", finished_aio);
while (finished_aio > 0) {
tms.tv_sec = 0;
tms.tv_nsec = 0;
r = io_getevents(ctx, 1, NUM_EVENTS, events, &tms);
if (r > 0) {
for (j = 0; j < r; ++j) {
((io_callback_t)(events[j].data))(ctx, events[j].obj, events[j].res, events[j].res2);
}
i += r;
finished_aio -= r;
}
}
}
close(epfd);
free(buf);
io_destroy(ctx);
close(fd);
close(efd);
remove(TEST_FILE);
return 0;
}