view src/os/unix/ngx_recv.c @ 8018:5119c8150478

Fixed runtime handling of systems without EPOLLRDHUP support. In 7583:efd71d49bde0 (nginx 1.17.5) along with introduction of the ioctl(FIONREAD) support proper handling of systems without EPOLLRDHUP support in the kernel (but with EPOLLRDHUP in headers) was broken. Before the change, rev->available was never set to 0 unless ngx_use_epoll_rdhup was also set (that is, runtime test for EPOLLRDHUP introduced in 6536:f7849bfb6d21 succeeded). After the change, rev->available might reach 0 on systems without runtime EPOLLRDHUP support, stopping further reading in ngx_readv_chain() and ngx_unix_recv(). And, if EOF happened to be already reported along with the last event, it is not reported again by epoll_wait(), leading to connection hangs and timeouts on such systems. This affects Linux kernels before 2.6.17 if nginx was compiled with newer headers, and, more importantly, emulation layers, such as DigitalOcean's App Platform's / gVisor's epoll emulation layer. Fix is to explicitly check ngx_use_epoll_rdhup before the corresponding rev->pending_eof tests in ngx_readv_chain() and ngx_unix_recv().
author Marcus Ball <marcus.ball@live.com>
date Mon, 30 May 2022 02:38:07 +0300
parents efd71d49bde0
children
line wrap: on
line source


/*
 * Copyright (C) Igor Sysoev
 * Copyright (C) Nginx, Inc.
 */


#include <ngx_config.h>
#include <ngx_core.h>
#include <ngx_event.h>


ssize_t
ngx_unix_recv(ngx_connection_t *c, u_char *buf, size_t size)
{
    ssize_t       n;
    ngx_err_t     err;
    ngx_event_t  *rev;

    rev = c->read;

#if (NGX_HAVE_KQUEUE)

    if (ngx_event_flags & NGX_USE_KQUEUE_EVENT) {
        ngx_log_debug3(NGX_LOG_DEBUG_EVENT, c->log, 0,
                       "recv: eof:%d, avail:%d, err:%d",
                       rev->pending_eof, rev->available, rev->kq_errno);

        if (rev->available == 0) {
            if (rev->pending_eof) {
                rev->ready = 0;
                rev->eof = 1;

                if (rev->kq_errno) {
                    rev->error = 1;
                    ngx_set_socket_errno(rev->kq_errno);

                    return ngx_connection_error(c, rev->kq_errno,
                               "kevent() reported about an closed connection");
                }

                return 0;

            } else {
                rev->ready = 0;
                return NGX_AGAIN;
            }
        }
    }

#endif

#if (NGX_HAVE_EPOLLRDHUP)

    if ((ngx_event_flags & NGX_USE_EPOLL_EVENT)
        && ngx_use_epoll_rdhup)
    {
        ngx_log_debug2(NGX_LOG_DEBUG_EVENT, c->log, 0,
                       "recv: eof:%d, avail:%d",
                       rev->pending_eof, rev->available);

        if (rev->available == 0 && !rev->pending_eof) {
            rev->ready = 0;
            return NGX_AGAIN;
        }
    }

#endif

    do {
        n = recv(c->fd, buf, size, 0);

        ngx_log_debug3(NGX_LOG_DEBUG_EVENT, c->log, 0,
                       "recv: fd:%d %z of %uz", c->fd, n, size);

        if (n == 0) {
            rev->ready = 0;
            rev->eof = 1;

#if (NGX_HAVE_KQUEUE)

            /*
             * on FreeBSD recv() may return 0 on closed socket
             * even if kqueue reported about available data
             */

            if (ngx_event_flags & NGX_USE_KQUEUE_EVENT) {
                rev->available = 0;
            }

#endif

            return 0;
        }

        if (n > 0) {

#if (NGX_HAVE_KQUEUE)

            if (ngx_event_flags & NGX_USE_KQUEUE_EVENT) {
                rev->available -= n;

                /*
                 * rev->available may be negative here because some additional
                 * bytes may be received between kevent() and recv()
                 */

                if (rev->available <= 0) {
                    if (!rev->pending_eof) {
                        rev->ready = 0;
                    }

                    rev->available = 0;
                }

                return n;
            }

#endif

#if (NGX_HAVE_FIONREAD)

            if (rev->available >= 0) {
                rev->available -= n;

                /*
                 * negative rev->available means some additional bytes
                 * were received between kernel notification and recv(),
                 * and therefore ev->ready can be safely reset even for
                 * edge-triggered event methods
                 */

                if (rev->available < 0) {
                    rev->available = 0;
                    rev->ready = 0;
                }

                ngx_log_debug1(NGX_LOG_DEBUG_EVENT, c->log, 0,
                               "recv: avail:%d", rev->available);

            } else if ((size_t) n == size) {

                if (ngx_socket_nread(c->fd, &rev->available) == -1) {
                    n = ngx_connection_error(c, ngx_socket_errno,
                                             ngx_socket_nread_n " failed");
                    break;
                }

                ngx_log_debug1(NGX_LOG_DEBUG_EVENT, c->log, 0,
                               "recv: avail:%d", rev->available);
            }

#endif

#if (NGX_HAVE_EPOLLRDHUP)

            if ((ngx_event_flags & NGX_USE_EPOLL_EVENT)
                && ngx_use_epoll_rdhup)
            {
                if ((size_t) n < size) {
                    if (!rev->pending_eof) {
                        rev->ready = 0;
                    }

                    rev->available = 0;
                }

                return n;
            }

#endif

            if ((size_t) n < size
                && !(ngx_event_flags & NGX_USE_GREEDY_EVENT))
            {
                rev->ready = 0;
            }

            return n;
        }

        err = ngx_socket_errno;

        if (err == NGX_EAGAIN || err == NGX_EINTR) {
            ngx_log_debug0(NGX_LOG_DEBUG_EVENT, c->log, err,
                           "recv() not ready");
            n = NGX_AGAIN;

        } else {
            n = ngx_connection_error(c, err, "recv() failed");
            break;
        }

    } while (err == NGX_EINTR);

    rev->ready = 0;

    if (n == NGX_ERROR) {
        rev->error = 1;
    }

    return n;
}