//! Wait for events to trigger on specific file descriptors use std::os::unix::io::{AsFd, AsRawFd, BorrowedFd}; use crate::errno::Errno; use crate::Result; /// This is a wrapper around `libc::pollfd`. /// /// It's meant to be used as an argument to the [`poll`](fn.poll.html) and /// [`ppoll`](fn.ppoll.html) functions to specify the events of interest /// for a specific file descriptor. /// /// After a call to `poll` or `ppoll`, the events that occurred can be /// retrieved by calling [`revents()`](#method.revents) on the `PollFd`. #[repr(transparent)] #[derive(Clone, Copy, Debug, Eq, Hash, PartialEq)] pub struct PollFd<'fd> { pollfd: libc::pollfd, _fd: std::marker::PhantomData>, } impl<'fd> PollFd<'fd> { /// Creates a new `PollFd` specifying the events of interest /// for a given file descriptor. // // Different from other I/O-safe interfaces, here, we have to take `AsFd` // by reference to prevent the case where the `fd` is closed but it is // still in use. For example: // // ```rust // let (reader, _) = pipe().unwrap(); // // // If `PollFd::new()` takes `AsFd` by value, then `reader` will be consumed, // // but the file descriptor of `reader` will still be in use. // let pollfd = PollFd::new(reader, flag); // // // Do something with `pollfd`, which uses the CLOSED fd. // ``` pub fn new(fd: &'fd Fd, events: PollFlags) -> PollFd<'fd> { PollFd { pollfd: libc::pollfd { fd: fd.as_fd().as_raw_fd(), events: events.bits(), revents: PollFlags::empty().bits(), }, _fd: std::marker::PhantomData, } } /// Returns the events that occurred in the last call to `poll` or `ppoll`. Will only return /// `None` if the kernel provides status flags that Nix does not know about. pub fn revents(self) -> Option { PollFlags::from_bits(self.pollfd.revents) } /// Returns if any of the events of interest occured in the last call to `poll` or `ppoll`. Will /// only return `None` if the kernel provides status flags that Nix does not know about. /// /// Equivalent to `x.revents()? != PollFlags::empty()`. /// /// This is marginally more efficient than [`PollFd::all`]. pub fn any(self) -> Option { Some(self.revents()? != PollFlags::empty()) } /// Returns if all the events of interest occured in the last call to `poll` or `ppoll`. Will /// only return `None` if the kernel provides status flags that Nix does not know about. /// /// Equivalent to `x.revents()? & x.events() == x.events()`. /// /// This is marginally less efficient than [`PollFd::any`]. pub fn all(self) -> Option { Some(self.revents()? & self.events() == self.events()) } /// The events of interest for this `PollFd`. pub fn events(self) -> PollFlags { PollFlags::from_bits(self.pollfd.events).unwrap() } /// Modify the events of interest for this `PollFd`. pub fn set_events(&mut self, events: PollFlags) { self.pollfd.events = events.bits(); } } impl<'fd> AsFd for PollFd<'fd> { fn as_fd(&self) -> BorrowedFd<'_> { // Safety: // // BorrowedFd::borrow_raw(RawFd) requires that the raw fd being passed // must remain open for the duration of the returned BorrowedFd, this is // guaranteed as the returned BorrowedFd has the lifetime parameter same // as `self`: // "fn as_fd<'self>(&'self self) -> BorrowedFd<'self>" // which means that `self` (PollFd) is guaranteed to outlive the returned // BorrowedFd. (Lifetime: PollFd > BorrowedFd) // // And the lifetime parameter of PollFd::new(fd, ...) ensures that `fd` // (an owned file descriptor) must outlive the returned PollFd: // "pub fn new(fd: &'fd Fd, events: PollFlags) -> PollFd<'fd>" // (Lifetime: Owned fd > PollFd) // // With two above relationships, we can conclude that the `Owned file // descriptor` will outlive the returned BorrowedFd, // (Lifetime: Owned fd > BorrowedFd) // i.e., the raw fd being passed will remain valid for the lifetime of // the returned BorrowedFd. unsafe { BorrowedFd::borrow_raw(self.pollfd.fd) } } } libc_bitflags! { /// These flags define the different events that can be monitored by `poll` and `ppoll` pub struct PollFlags: libc::c_short { /// There is data to read. POLLIN; /// There is some exceptional condition on the file descriptor. /// /// Possibilities include: /// /// * There is out-of-band data on a TCP socket (see /// [tcp(7)](https://man7.org/linux/man-pages/man7/tcp.7.html)). /// * A pseudoterminal master in packet mode has seen a state /// change on the slave (see /// [ioctl_tty(2)](https://man7.org/linux/man-pages/man2/ioctl_tty.2.html)). /// * A cgroup.events file has been modified (see /// [cgroups(7)](https://man7.org/linux/man-pages/man7/cgroups.7.html)). POLLPRI; /// Writing is now possible, though a write larger that the /// available space in a socket or pipe will still block (unless /// `O_NONBLOCK` is set). POLLOUT; /// Equivalent to [`POLLIN`](constant.POLLIN.html) #[cfg(not(target_os = "redox"))] #[cfg_attr(docsrs, doc(cfg(all())))] POLLRDNORM; #[cfg(not(target_os = "redox"))] #[cfg_attr(docsrs, doc(cfg(all())))] /// Equivalent to [`POLLOUT`](constant.POLLOUT.html) POLLWRNORM; /// Priority band data can be read (generally unused on Linux). #[cfg(not(target_os = "redox"))] #[cfg_attr(docsrs, doc(cfg(all())))] POLLRDBAND; /// Priority data may be written. #[cfg(not(target_os = "redox"))] #[cfg_attr(docsrs, doc(cfg(all())))] POLLWRBAND; /// Error condition (only returned in /// [`PollFd::revents`](struct.PollFd.html#method.revents); /// ignored in [`PollFd::new`](struct.PollFd.html#method.new)). /// This bit is also set for a file descriptor referring to the /// write end of a pipe when the read end has been closed. POLLERR; /// Hang up (only returned in [`PollFd::revents`](struct.PollFd.html#method.revents); /// ignored in [`PollFd::new`](struct.PollFd.html#method.new)). /// Note that when reading from a channel such as a pipe or a stream /// socket, this event merely indicates that the peer closed its /// end of the channel. Subsequent reads from the channel will /// return 0 (end of file) only after all outstanding data in the /// channel has been consumed. POLLHUP; /// Invalid request: `fd` not open (only returned in /// [`PollFd::revents`](struct.PollFd.html#method.revents); /// ignored in [`PollFd::new`](struct.PollFd.html#method.new)). POLLNVAL; } } /// `poll` waits for one of a set of file descriptors to become ready to perform I/O. /// ([`poll(2)`](https://pubs.opengroup.org/onlinepubs/9699919799/functions/poll.html)) /// /// `fds` contains all [`PollFd`](struct.PollFd.html) to poll. /// The function will return as soon as any event occur for any of these `PollFd`s. /// /// The `timeout` argument specifies the number of milliseconds that `poll()` /// should block waiting for a file descriptor to become ready. The call /// will block until either: /// /// * a file descriptor becomes ready; /// * the call is interrupted by a signal handler; or /// * the timeout expires. /// /// Note that the timeout interval will be rounded up to the system clock /// granularity, and kernel scheduling delays mean that the blocking /// interval may overrun by a small amount. Specifying a negative value /// in timeout means an infinite timeout. Specifying a timeout of zero /// causes `poll()` to return immediately, even if no file descriptors are /// ready. pub fn poll(fds: &mut [PollFd], timeout: libc::c_int) -> Result { let res = unsafe { libc::poll( fds.as_mut_ptr() as *mut libc::pollfd, fds.len() as libc::nfds_t, timeout, ) }; Errno::result(res) } feature! { #![feature = "signal"] /// `ppoll()` allows an application to safely wait until either a file /// descriptor becomes ready or until a signal is caught. /// ([`poll(2)`](https://man7.org/linux/man-pages/man2/poll.2.html)) /// /// `ppoll` behaves like `poll`, but let you specify what signals may interrupt it /// with the `sigmask` argument. If you want `ppoll` to block indefinitely, /// specify `None` as `timeout` (it is like `timeout = -1` for `poll`). /// If `sigmask` is `None`, then no signal mask manipulation is performed, /// so in that case `ppoll` differs from `poll` only in the precision of the /// timeout argument. /// #[cfg(any(target_os = "android", target_os = "dragonfly", target_os = "freebsd", target_os = "linux"))] pub fn ppoll( fds: &mut [PollFd], timeout: Option, sigmask: Option ) -> Result { let timeout = timeout.as_ref().map_or(core::ptr::null(), |r| r.as_ref()); let sigmask = sigmask.as_ref().map_or(core::ptr::null(), |r| r.as_ref()); let res = unsafe { libc::ppoll(fds.as_mut_ptr() as *mut libc::pollfd, fds.len() as libc::nfds_t, timeout, sigmask) }; Errno::result(res) } }