vpoll_create, vpoll_ctl, vpoll_close

#include <vpoll.h>

int vpoll_create(uint32_t init_events, int flags);

int vpoll_ctl(int fd, int op, uint32_t events);

int vpoll_close(int fd);

This library permits one to create a vpoll file descriptor "vpollfd" that can be used in poll/select/ppoll/pselect/epoll(2) system calls. The events reported by a can be controlled by vpoll_ctl. vpoll encodes the events using the same flags EPOLL* defined in epoll_ctl(2).

The library uses the vpoll extension for eventfd if the kernel provides it otherwise it (partially) emulates the feature using socketpair(2). The emulation supports only EPOLLIN, EPOLLOUT flags and a non standard version of EPOLLHUP/EPOLLRDHUP.

vpoll_create

This function creates a "vpollfd". The argument init_events is used to set the initial state of events. The following value can be included in flags:

FD_CLOEXEC: Set the close-on-exec flag on the new file descriptor. See the description of the O_CLOEXEC flag in open(2) for reasons why this may be useful.

vpoll_ctl

This function changes the set of pending events reported by a "vpollfd". The argument op can take the following values:

VPOLL_CTL_ADDEVENTS: the events set in the argument events are added to the set of pending events.

VPOLL_CTL_DELEVENTS: the events set in the argument events are deleted from the set of pending events.

VPOLL_CTL_SETEVENTS: the value of the argument events is assigned to the set of pending events.

vpoll_close

This function closes the vpoll file descritor.

vpoll_create returns the new file descriptor, or -1 if an error occurred (in which case, errno is set appropriately)

vpoll_ctl and vpoll_close return zero in case of success. On error, -1 is returned, and errno is set appropriately.

#define _GNU_SOURCE
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <errno.h>
#include <sys/epoll.h>
#include <vpoll.h>
int main(int argc, char *argv[]) {
    int vpollfd = vpoll_create(0, FD_CLOEXEC);
    int epfd = epoll_create1(EPOLL_CLOEXEC);
    struct epoll_event reqevents={EPOLLIN | EPOLLRDHUP | EPOLLERR |
        EPOLLOUT | EPOLLHUP | EPOLLPRI};
    epoll_ctl(epfd,  EPOLL_CTL_ADD, vpollfd, &reqevents);
    switch (fork()) {
        case 0:
            sleep(3);
            vpoll_ctl(vpollfd, VPOLL_CTL_ADDEVENTS,  EPOLLIN);
            sleep(3);
            vpoll_ctl(vpollfd, VPOLL_CTL_ADDEVENTS,  EPOLLIN);
            sleep(3);
            vpoll_ctl(vpollfd, VPOLL_CTL_ADDEVENTS,  EPOLLOUT);
            sleep(3);
            vpoll_ctl(vpollfd, VPOLL_CTL_ADDEVENTS,  EPOLLHUP);
            sleep(3);
            exit(0);
        default:
            while (1) {
                struct epoll_event ev;
                int n = epoll_wait(epfd, &ev, 1, 1000);
                if (n < 0) {
                    perror("epoll_wait");
                    break;
                }
                if (n > 0) {
                    printf("GOT event %x\n", ev.events);
                    vpoll_ctl(vpollfd, VPOLL_CTL_DELEVENTS, ev.events);
                    if (ev.events & EPOLLHUP)
                            break;
                } else {
                    printf("timeout\n");
                }
            }
            break;
        case -1:
            printf("fork error\n");
    }
    vpoll_close(vpollfd);
    close(epfd);
    return 0;
}

On a machine running a Linux Kernel providing eventfd/vpoll the output of this program is:

timeout
timeout
GOT event 1
timeout
timeout
GOT event 1
timeout
timeout
GOT event 4
timeout
timeout
GOT event 10

Instead when the demo program runs using the emulation layer the output is:

timeout
timeout
GOT event 1
timeout
timeout
GOT event 1
timeout
timeout
GOT event 4
timeout
timeout
GOT event 2011

In fact, the emulator uses a socketpair to generate the events. Hangup is emulated by closing the other end of the socketpair: this generates EPOLLHUP as well as EPOLLIN and EPOLLRDHUP.

VirtualSquare. Project leader: Renzo Davoli.