openssl/ssl/quic/quic_reactor.c

/*
 * Copyright 2022-2023 The OpenSSL Project Authors. All Rights Reserved.
 *
 * Licensed under the Apache License 2.0 (the "License").  You may not use
 * this file except in compliance with the License.  You can obtain a copy
 * in the file LICENSE in the source distribution or at
 * https://www.openssl.org/source/license.html
 */
#include "internal/quic_reactor.h"
#include "internal/common.h"
#include "internal/thread_arch.h"
#include <assert.h>

/*
 * Core I/O Reactor Framework
 * ==========================
 */
static void rtor_notify_other_threads(QUIC_REACTOR *rtor);

int ossl_quic_reactor_init(QUIC_REACTOR *rtor,
                           void (*tick_cb)(QUIC_TICK_RESULT *res, void *arg,
                                           uint32_t flags),
                           void *tick_cb_arg,
                           CRYPTO_MUTEX *mutex,
                           OSSL_TIME initial_tick_deadline,
                           uint64_t flags)
{
    rtor->poll_r.type       = BIO_POLL_DESCRIPTOR_TYPE_NONE;
    rtor->poll_w.type       = BIO_POLL_DESCRIPTOR_TYPE_NONE;
    rtor->net_read_desired  = 0;
    rtor->net_write_desired = 0;
    rtor->can_poll_r        = 0;
    rtor->can_poll_w        = 0;
    rtor->tick_deadline     = initial_tick_deadline;

    rtor->tick_cb           = tick_cb;
    rtor->tick_cb_arg       = tick_cb_arg;
    rtor->mutex             = mutex;

    rtor->cur_blocking_waiters = 0;

    if ((flags & QUIC_REACTOR_FLAG_USE_NOTIFIER) != 0) {
        if (!ossl_rio_notifier_init(&rtor->notifier))
            return 0;

        if ((rtor->notifier_cv = ossl_crypto_condvar_new()) == NULL) {
            ossl_rio_notifier_cleanup(&rtor->notifier);
            return 0;
        }

        rtor->have_notifier = 1;
    } else {
        rtor->have_notifier = 0;
    }

    return 1;
}

void ossl_quic_reactor_cleanup(QUIC_REACTOR *rtor)
{
    if (rtor == NULL)
        return;

    if (rtor->have_notifier) {
        ossl_rio_notifier_cleanup(&rtor->notifier);
        rtor->have_notifier = 0;

        ossl_crypto_condvar_free(&rtor->notifier_cv);
    }
}

void ossl_quic_reactor_set_poll_r(QUIC_REACTOR *rtor, const BIO_POLL_DESCRIPTOR *r)
{
    if (r == NULL)
        rtor->poll_r.type = BIO_POLL_DESCRIPTOR_TYPE_NONE;
    else
        rtor->poll_r = *r;

    rtor->can_poll_r
        = ossl_quic_reactor_can_support_poll_descriptor(rtor, &rtor->poll_r);
}

void ossl_quic_reactor_set_poll_w(QUIC_REACTOR *rtor, const BIO_POLL_DESCRIPTOR *w)
{
    if (w == NULL)
        rtor->poll_w.type = BIO_POLL_DESCRIPTOR_TYPE_NONE;
    else
        rtor->poll_w = *w;

    rtor->can_poll_w
        = ossl_quic_reactor_can_support_poll_descriptor(rtor, &rtor->poll_w);
}

const BIO_POLL_DESCRIPTOR *ossl_quic_reactor_get_poll_r(const QUIC_REACTOR *rtor)
{
    return &rtor->poll_r;
}

const BIO_POLL_DESCRIPTOR *ossl_quic_reactor_get_poll_w(const QUIC_REACTOR *rtor)
{
    return &rtor->poll_w;
}

int ossl_quic_reactor_can_support_poll_descriptor(const QUIC_REACTOR *rtor,
                                                  const BIO_POLL_DESCRIPTOR *d)
{
    return d->type == BIO_POLL_DESCRIPTOR_TYPE_SOCK_FD;
}

int ossl_quic_reactor_can_poll_r(const QUIC_REACTOR *rtor)
{
    return rtor->can_poll_r;
}

int ossl_quic_reactor_can_poll_w(const QUIC_REACTOR *rtor)
{
    return rtor->can_poll_w;
}

int ossl_quic_reactor_net_read_desired(QUIC_REACTOR *rtor)
{
    return rtor->net_read_desired;
}

int ossl_quic_reactor_net_write_desired(QUIC_REACTOR *rtor)
{
    return rtor->net_write_desired;
}

OSSL_TIME ossl_quic_reactor_get_tick_deadline(QUIC_REACTOR *rtor)
{
    return rtor->tick_deadline;
}

int ossl_quic_reactor_tick(QUIC_REACTOR *rtor, uint32_t flags)
{
    QUIC_TICK_RESULT res = {0};

    /*
     * Note that the tick callback cannot fail; this is intentional. Arguably it
     * does not make that much sense for ticking to 'fail' (in the sense of an
     * explicit error indicated to the user) because ticking is by its nature
     * best effort. If something fatal happens with a connection we can report
     * it on the next actual application I/O call.
     */
    rtor->tick_cb(&res, rtor->tick_cb_arg, flags);

    rtor->net_read_desired  = res.net_read_desired;
    rtor->net_write_desired = res.net_write_desired;
    rtor->tick_deadline     = res.tick_deadline;
    if (res.notify_other_threads)
        rtor_notify_other_threads(rtor);

    return 1;
}

RIO_NOTIFIER *ossl_quic_reactor_get0_notifier(QUIC_REACTOR *rtor)
{
    return rtor->have_notifier ? &rtor->notifier : NULL;
}

/*
 * Blocking I/O Adaptation Layer
 * =============================
 */

/*
 * Utility which can be used to poll on up to two FDs. This is designed to
 * support use of split FDs (e.g. with SSL_set_rfd and SSL_set_wfd where
 * different FDs are used for read and write).
 *
 * Generally use of poll(2) is preferred where available. Windows, however,
 * hasn't traditionally offered poll(2), only select(2). WSAPoll() was
 * introduced in Vista but has seemingly been buggy until relatively recent
 * versions of Windows 10. Moreover we support XP so this is not a suitable
 * target anyway. However, the traditional issues with select(2) turn out not to
 * be an issue on Windows; whereas traditional *NIX select(2) uses a bitmap of
 * FDs (and thus is limited in the magnitude of the FDs expressible), Windows
 * select(2) is very different. In Windows, socket handles are not allocated
 * contiguously from zero and thus this bitmap approach was infeasible. Thus in
 * adapting the Berkeley sockets API to Windows a different approach was taken
 * whereby the fd_set contains a fixed length array of socket handles and an
 * integer indicating how many entries are valid; thus Windows select()
 * ironically is actually much more like *NIX poll(2) than *NIX select(2). In
 * any case, this means that the relevant limit for Windows select() is the
 * number of FDs being polled, not the magnitude of those FDs. Since we only
 * poll for two FDs here, this limit does not concern us.
 *
 * Usage: rfd and wfd may be the same or different. Either or both may also be
 * -1. If rfd_want_read is 1, rfd is polled for readability, and if
 * wfd_want_write is 1, wfd is polled for writability. Note that since any
 * passed FD is always polled for error conditions, setting rfd_want_read=0 and
 * wfd_want_write=0 is not the same as passing -1 for both FDs.
 *
 * deadline is a timestamp to return at. If it is ossl_time_infinite(), the call
 * never times out.
 *
 * Returns 0 on error and 1 on success. Timeout expiry is considered a success
 * condition. We don't elaborate our return values here because the way we are
 * actually using this doesn't currently care.
 *
 * If mutex is non-NULL, it is assumed to be held for write and is unlocked for
 * the duration of the call.
 *
 * Precondition:   mutex is NULL or is held for write (unchecked)
 * Postcondition:  mutex is NULL or is held for write (unless
 *                   CRYPTO_THREAD_write_lock fails)
 */
static int poll_two_fds(int rfd, int rfd_want_read,
                        int wfd, int wfd_want_write,
                        int notify_rfd,
                        OSSL_TIME deadline,
                        CRYPTO_MUTEX *mutex)
{
#if defined(OPENSSL_SYS_WINDOWS) || !defined(POLLIN)
    fd_set rfd_set, wfd_set, efd_set;
    OSSL_TIME now, timeout;
    struct timeval tv, *ptv;
    int maxfd, pres;

# ifndef OPENSSL_SYS_WINDOWS
    /*
     * On Windows there is no relevant limit to the magnitude of a fd value (see
     * above). On *NIX the fd_set uses a bitmap and we must check the limit.
     */
    if (rfd >= FD_SETSIZE || wfd >= FD_SETSIZE)
        return 0;
# endif

    FD_ZERO(&rfd_set);
    FD_ZERO(&wfd_set);
    FD_ZERO(&efd_set);

    if (rfd != INVALID_SOCKET && rfd_want_read)
        openssl_fdset(rfd, &rfd_set);
    if (wfd != INVALID_SOCKET && wfd_want_write)
        openssl_fdset(wfd, &wfd_set);

    /* Always check for error conditions. */
    if (rfd != INVALID_SOCKET)
        openssl_fdset(rfd, &efd_set);
    if (wfd != INVALID_SOCKET)
        openssl_fdset(wfd, &efd_set);

    /* Check for notifier FD readability. */
    if (notify_rfd != INVALID_SOCKET) {
        openssl_fdset(notify_rfd, &rfd_set);
        openssl_fdset(notify_rfd, &efd_set);
    }

    maxfd = rfd;
    if (wfd > maxfd)
        maxfd = wfd;
    if (notify_rfd > maxfd)
        maxfd = notify_rfd;

    if (!ossl_assert(rfd != INVALID_SOCKET || wfd != INVALID_SOCKET
                     || !ossl_time_is_infinite(deadline)))
        /* Do not block forever; should not happen. */
        return 0;

    /*
     * The mutex dance (unlock/re-locak after poll/seclect) is
     * potentially problematic. This may create a situation when
     * two threads arrive to select/poll with the same file
     * descriptors. We just need to be aware of this.
     */
# if defined(OPENSSL_THREADS)
    if (mutex != NULL)
        ossl_crypto_mutex_unlock(mutex);
# endif

    do {
        /*
         * select expects a timeout, not a deadline, so do the conversion.
         * Update for each call to ensure the correct value is used if we repeat
         * due to EINTR.
         */
        if (ossl_time_is_infinite(deadline)) {
            ptv = NULL;
        } else {
            now = ossl_time_now();
            /*
             * ossl_time_subtract saturates to zero so we don't need to check if
             * now > deadline.
             */
            timeout = ossl_time_subtract(deadline, now);
            tv      = ossl_time_to_timeval(timeout);
            ptv     = &tv;
        }

        pres = select(maxfd + 1, &rfd_set, &wfd_set, &efd_set, ptv);
    } while (pres == -1 && get_last_socket_error_is_eintr());

# if defined(OPENSSL_THREADS)
    if (mutex != NULL)
        ossl_crypto_mutex_lock(mutex);
# endif

    return pres < 0 ? 0 : 1;
#else
    int pres, timeout_ms;
    OSSL_TIME now, timeout;
    struct pollfd pfds[3] = {0};
    size_t npfd = 0;

    if (rfd == wfd) {
        pfds[npfd].fd = rfd;
        pfds[npfd].events = (rfd_want_read  ? POLLIN  : 0)
                          | (wfd_want_write ? POLLOUT : 0);
        if (rfd >= 0 && pfds[npfd].events != 0)
            ++npfd;
    } else {
        pfds[npfd].fd     = rfd;
        pfds[npfd].events = (rfd_want_read ? POLLIN : 0);
        if (rfd >= 0 && pfds[npfd].events != 0)
            ++npfd;

        pfds[npfd].fd     = wfd;
        pfds[npfd].events = (wfd_want_write ? POLLOUT : 0);
        if (wfd >= 0 && pfds[npfd].events != 0)
            ++npfd;
    }

    if (notify_rfd >= 0) {
        pfds[npfd].fd       = notify_rfd;
        pfds[npfd].events   = POLLIN;
        ++npfd;
    }

    if (!ossl_assert(npfd != 0 || !ossl_time_is_infinite(deadline)))
        /* Do not block forever; should not happen. */
        return 0;

# if defined(OPENSSL_THREADS)
    if (mutex != NULL)
        ossl_crypto_mutex_unlock(mutex);
# endif

    do {
        if (ossl_time_is_infinite(deadline)) {
            timeout_ms = -1;
        } else {
            now         = ossl_time_now();
            timeout     = ossl_time_subtract(deadline, now);
            timeout_ms  = ossl_time2ms(timeout);
        }

        pres = poll(pfds, npfd, timeout_ms);
    } while (pres == -1 && get_last_socket_error_is_eintr());

# if defined(OPENSSL_THREADS)
    if (mutex != NULL)
        ossl_crypto_mutex_lock(mutex);
# endif

    return pres < 0 ? 0 : 1;
#endif
}

static int poll_descriptor_to_fd(const BIO_POLL_DESCRIPTOR *d, int *fd)
{
    if (d == NULL || d->type == BIO_POLL_DESCRIPTOR_TYPE_NONE) {
        *fd = INVALID_SOCKET;
        return 1;
    }

    if (d->type != BIO_POLL_DESCRIPTOR_TYPE_SOCK_FD
            || d->value.fd == INVALID_SOCKET)
        return 0;

    *fd = d->value.fd;
    return 1;
}

/*
 * Poll up to two abstract poll descriptors, as well as an optional notify FD.
 * Currently we only support poll descriptors which represent FDs.
 *
 * If mutex is non-NULL, it is assumed be a lock currently held for write and is
 * unlocked for the duration of any wait.
 *
 * Precondition:   mutex is NULL or is held for write (unchecked)
 * Postcondition:  mutex is NULL or is held for write (unless
 *                   CRYPTO_THREAD_write_lock fails)
 */
static int poll_two_descriptors(const BIO_POLL_DESCRIPTOR *r, int r_want_read,
                                const BIO_POLL_DESCRIPTOR *w, int w_want_write,
                                int notify_rfd,
                                OSSL_TIME deadline,
                                CRYPTO_MUTEX *mutex)
{
    int rfd, wfd;

    if (!poll_descriptor_to_fd(r, &rfd)
        || !poll_descriptor_to_fd(w, &wfd))
        return 0;

    return poll_two_fds(rfd, r_want_read, wfd, w_want_write,
                        notify_rfd, deadline, mutex);
}

/*
 * Notify other threads currently blocking in
 * ossl_quic_reactor_block_until_pred() calls that a predicate they are using
 * might now be met due to state changes.
 *
 * This function must be called after state changes which might cause a
 * predicate in another thread to now be met (i.e., ticking). It is a no-op if
 * inter-thread notification is not being used.
 *
 * The reactor mutex must be held while calling this function.
 */
static void rtor_notify_other_threads(QUIC_REACTOR *rtor)
{
    if (!rtor->have_notifier)
        return;

    /*
     * This function is called when we have done anything on this thread which
     * might allow a predicate for a block_until_pred call on another thread to
     * now be met.
     *
     * When this happens, we need to wake those threads using the notifier.
     * However, we do not want to wake *this* thread (if/when it subsequently
     * enters block_until_pred) due to the notifier FD becoming readable.
     * Therefore, signal the notifier, and use a CV to detect when all other
     * threads have woken.
     */

   if (rtor->cur_blocking_waiters == 0)
       /* Nothing to do in this case. */
       return;

   /* Signal the notifier to wake up all threads. */
   if (!rtor->signalled_notifier) {
       ossl_rio_notifier_signal(&rtor->notifier);
       rtor->signalled_notifier = 1;
   }

   /*
    * Wait on the CV until all threads have finished the first phase of the
    * wakeup process and the last thread out has taken responsibility for
    * unsignalling the notifier.
    */
    while (rtor->signalled_notifier)
        ossl_crypto_condvar_wait(rtor->notifier_cv, rtor->mutex);
}

/*
 * Block until a predicate function evaluates to true.
 *
 * If mutex is non-NULL, it is assumed be a lock currently held for write and is
 * unlocked for the duration of any wait.
 *
 * Precondition:   Must hold channel write lock (unchecked)
 * Precondition:   mutex is NULL or is held for write (unchecked)
 * Postcondition:  mutex is NULL or is held for write (unless
 *                   CRYPTO_THREAD_write_lock fails)
 */
int ossl_quic_reactor_block_until_pred(QUIC_REACTOR *rtor,
                                       int (*pred)(void *arg), void *pred_arg,
                                       uint32_t flags)
{
    int res, net_read_desired, net_write_desired, notifier_fd;
    OSSL_TIME tick_deadline;

    notifier_fd
        = (rtor->have_notifier ? ossl_rio_notifier_as_fd(&rtor->notifier)
                               : INVALID_SOCKET);

    for (;;) {
        if ((flags & SKIP_FIRST_TICK) != 0)
            flags &= ~SKIP_FIRST_TICK;
        else
            /* best effort */
            ossl_quic_reactor_tick(rtor, 0);

        if ((res = pred(pred_arg)) != 0)
            return res;

        net_read_desired  = ossl_quic_reactor_net_read_desired(rtor);
        net_write_desired = ossl_quic_reactor_net_write_desired(rtor);
        tick_deadline     = ossl_quic_reactor_get_tick_deadline(rtor);
        if (!net_read_desired && !net_write_desired
            && ossl_time_is_infinite(tick_deadline))
            /* Can't wait if there is nothing to wait for. */
            return 0;

        ++rtor->cur_blocking_waiters;

        res = poll_two_descriptors(ossl_quic_reactor_get_poll_r(rtor),
                                   net_read_desired,
                                   ossl_quic_reactor_get_poll_w(rtor),
                                   net_write_desired,
                                   notifier_fd,
                                   tick_deadline,
                                   rtor->mutex);

        assert(rtor->cur_blocking_waiters > 0);
        --rtor->cur_blocking_waiters;

        /*
         * We have now exited the OS poller call. We may have
         * (rtor->signalled_notifier), and other threads may still be blocking.
         * This means that cur_blocking_waiters may still be non-zero. As such,
         * we cannot unsignal the notifier until all threads have had an
         * opportunity to wake up.
         *
         * At the same time, we cannot unsignal in the case where
         * cur_blocking_waiters is now zero because this condition may not occur
         * reliably. Consider the following scenario:
         *
         *   T1 enters block_until_pred, cur_blocking_waiters -> 1
         *   T2 enters block_until_pred, cur_blocking_waiters -> 2
         *   T3 enters block_until_pred, cur_blocking_waiters -> 3
         *
         *   T4 enters block_until_pred, does not block, ticks,
         *     sees that cur_blocking_waiters > 0 and signals the notifier
         *
         *   T3 wakes, cur_blocking_waiters -> 2
         *   T3 predicate is not satisfied, cur_blocking_waiters -> 3, block again
         *
         *   Notifier is still signalled, so T3 immediately wakes again
         *   and is stuck repeating the above steps.
         *
         *   T1, T2 are also woken by the notifier but never see
         *   cur_blocking_waiters drop to 0, so never unsignal the notifier.
         *
         * As such, a two phase approach is chosen when designalling the
         * notifier:
         *
         *   First, all of the poll_two_descriptor calls on all threads are
         *   allowed to exit due to the notifier being signalled.
         *
         *   Second, the thread which happened to be the one which decremented
         *   cur_blocking_waiters to 0 unsignals the notifier and is then
         *   responsible for broadcasting to a CV to indicate to the other
         *   threads that the synchronised wakeup has been completed. Other
         *   threads wait for this CV to be signalled.
         *
         */
        if (rtor->have_notifier && rtor->signalled_notifier) {
            if (rtor->cur_blocking_waiters == 0) {
                ossl_rio_notifier_unsignal(&rtor->notifier);
                rtor->signalled_notifier = 0;

                /*
                 * Release the other threads which have woken up (and possibly
                 * rtor_notify_other_threads as well).
                 */
                ossl_crypto_condvar_broadcast(rtor->notifier_cv);
            } else {
                /* We are not the last waiter out - so wait for that one. */
                while (rtor->signalled_notifier)
                    ossl_crypto_condvar_wait(rtor->notifier_cv, rtor->mutex);
            }
        }

        if (!res)
            /*
             * We don't actually care why the call succeeded (timeout, FD
             * readiness), we just call reactor_tick and start trying to do I/O
             * things again. If poll_two_fds returns 0, this is some other
             * non-timeout failure and we should stop here.
             *
             * TODO(QUIC FUTURE): In the future we could avoid unnecessary
             * syscalls by not retrying network I/O that isn't ready based
             * on the result of the poll call. However this might be difficult
             * because it requires we do the call to poll(2) or equivalent
             * syscall ourselves, whereas in the general case the application
             * does the polling and just calls SSL_handle_events().
             * Implementing this optimisation in the future will probably
             * therefore require API changes.
             */
            return 0;
    }

    return res;
}