fish-shell/src/io.rs

1043 lines
36 KiB
Rust
Raw Normal View History

2023-04-09 20:06:40 +08:00
use crate::builtins::shared::{STATUS_CMD_ERROR, STATUS_CMD_OK, STATUS_READ_TOO_MUCH};
use crate::common::{str2wcstring, wcs2string, EMPTY_STRING};
2024-01-14 04:35:55 +08:00
use crate::fd_monitor::{
Callback, FdMonitor, FdMonitorItem, FdMonitorItemId, ItemAction, ItemWakeReason,
};
use crate::fds::{
make_autoclose_pipes, make_fd_nonblocking, wopen_cloexec, AutoCloseFd, PIPE_ERROR,
};
2023-04-09 20:06:40 +08:00
use crate::flog::{should_flog, FLOG, FLOGF};
use crate::global_safety::RelaxedAtomicBool;
use crate::nix::isatty;
2023-04-09 20:06:40 +08:00
use crate::path::path_apply_working_directory;
use crate::proc::JobGroupRef;
2023-04-09 20:06:40 +08:00
use crate::redirection::{RedirectionMode, RedirectionSpecList};
use crate::signal::SigChecker;
2023-04-09 20:06:40 +08:00
use crate::topic_monitor::topic_t;
2023-08-09 06:16:04 +08:00
use crate::wchar::prelude::*;
use crate::wutil::{perror, perror_io, wdirname, wstat, wwrite_to_fd};
2023-04-09 20:06:40 +08:00
use errno::Errno;
2024-01-21 07:21:49 +08:00
use libc::{EAGAIN, EINTR, ENOENT, ENOTDIR, EPIPE, EWOULDBLOCK, STDOUT_FILENO};
use nix::fcntl::OFlag;
2024-01-21 06:57:34 +08:00
use nix::sys::stat::Mode;
use std::cell::{RefCell, UnsafeCell};
2024-03-23 13:44:27 +08:00
use std::fs::File;
2024-02-10 11:49:07 +08:00
use std::io;
2024-01-29 02:07:06 +08:00
use std::os::fd::{AsRawFd, IntoRawFd, OwnedFd, RawFd};
use std::sync::atomic::{AtomicU64, Ordering};
use std::sync::{Arc, Condvar, Mutex, MutexGuard};
2023-04-09 20:06:40 +08:00
/// separated_buffer_t represents a buffer of output from commands, prepared to be turned into a
/// variable. For example, command substitutions output into one of these. Most commands just
/// produce a stream of bytes, and those get stored directly. However other commands produce
/// explicitly separated output, in particular `string` like `string collect` and `string split0`.
/// The buffer tracks a sequence of elements. Some elements are explicitly separated and should not
/// be further split; other elements have inferred separation and may be split by IFS (or not,
/// depending on its value).
#[derive(Clone, Copy, PartialEq, Eq)]
pub enum SeparationType {
/// this element should be further separated by IFS
inferred,
/// this element is explicitly separated and should not be further split
explicitly,
}
pub struct BufferElement {
pub contents: Vec<u8>,
pub separation: SeparationType,
2023-04-09 20:06:40 +08:00
}
impl BufferElement {
pub fn new(contents: Vec<u8>, separation: SeparationType) -> Self {
BufferElement {
contents,
separation,
}
}
pub fn is_explicitly_separated(&self) -> bool {
self.separation == SeparationType::explicitly
}
}
/// A separated_buffer_t contains a list of elements, some of which may be separated explicitly and
/// others which must be separated further by the user (e.g. via IFS).
pub struct SeparatedBuffer {
/// Limit on how much data we'll buffer. Zero means no limit.
buffer_limit: usize,
/// Current size of all contents.
contents_size: usize,
/// List of buffer elements.
elements: Vec<BufferElement>,
/// True if we're discarding input because our buffer_limit has been exceeded.
discard: bool,
}
impl SeparatedBuffer {
pub fn new(limit: usize) -> Self {
SeparatedBuffer {
buffer_limit: limit,
contents_size: 0,
elements: vec![],
discard: false,
}
}
/// Return the buffer limit size, or 0 for no limit.
2023-04-09 20:06:40 +08:00
pub fn limit(&self) -> usize {
self.buffer_limit
}
/// Return the contents size.
2023-10-13 16:59:12 +08:00
pub fn len(&self) -> usize {
2023-04-09 20:06:40 +08:00
self.contents_size
}
/// Return whether the output has been discarded.
2023-04-09 20:06:40 +08:00
pub fn discarded(&self) -> bool {
self.discard
}
/// Serialize the contents to a single string, where explicitly separated elements have a
/// newline appended.
pub fn newline_serialized(&self) -> Vec<u8> {
2023-10-13 16:59:12 +08:00
let mut result = Vec::with_capacity(self.len());
2023-04-09 20:06:40 +08:00
for elem in &self.elements {
result.extend_from_slice(&elem.contents);
if elem.is_explicitly_separated() {
result.push(b'\n');
}
}
result
}
/// Return the list of elements.
2023-04-09 20:06:40 +08:00
pub fn elements(&self) -> &[BufferElement] {
&self.elements
}
/// Append the given data with separation type `sep`.
2023-04-09 20:06:40 +08:00
pub fn append(&mut self, data: &[u8], sep: SeparationType) -> bool {
if !self.try_add_size(data.len()) {
return false;
}
// Try merging with the last element.
if sep == SeparationType::inferred && self.last_inferred() {
self.elements
.last_mut()
.unwrap()
.contents
.extend_from_slice(data);
} else {
self.elements.push(BufferElement::new(data.to_vec(), sep));
}
true
}
/// Remove all elements and unset the discard flag.
pub fn clear(&mut self) {
self.elements.clear();
self.contents_size = 0;
self.discard = false;
}
/// Return true if our last element has an inferred separation type.
2023-04-09 20:06:40 +08:00
fn last_inferred(&self) -> bool {
!self.elements.is_empty() && !self.elements.last().unwrap().is_explicitly_separated()
}
/// Mark that we are about to add the given size `delta` to the buffer. Return true if we
2023-04-09 20:06:40 +08:00
/// succeed, false if we exceed buffer_limit.
fn try_add_size(&mut self, delta: usize) -> bool {
if self.discard {
return false;
}
let proposed_size = self.contents_size + delta;
if proposed_size < delta || (self.buffer_limit > 0 && proposed_size > self.buffer_limit) {
self.clear();
self.discard = true;
return false;
}
self.contents_size = proposed_size;
true
}
}
/// Describes what type of IO operation an io_data_t represents.
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
2023-04-09 20:06:40 +08:00
pub enum IoMode {
file,
pipe,
fd,
close,
bufferfill,
}
/// Represents a FD redirection.
pub trait IoData {
/// Type of redirect.
fn io_mode(&self) -> IoMode;
/// FD to redirect.
fn fd(&self) -> RawFd;
/// Source fd. This is dup2'd to fd, or if it is -1, then fd is closed.
/// That is, we call dup2(source_fd, fd).
fn source_fd(&self) -> RawFd;
fn print(&self);
// The address of the object, for comparison.
fn as_ptr(&self) -> *const ();
fn as_bufferfill(&self) -> Option<&IoBufferfill> {
None
}
2023-04-09 20:06:40 +08:00
}
// todo!("this should be safe because of how it's used. Rationalize this better.")
pub trait IoDataSync: IoData + Send + Sync {}
unsafe impl Send for IoClose {}
unsafe impl Send for IoFd {}
unsafe impl Send for IoFile {}
unsafe impl Send for IoPipe {}
unsafe impl Send for IoBufferfill {}
unsafe impl Sync for IoClose {}
unsafe impl Sync for IoFd {}
unsafe impl Sync for IoFile {}
unsafe impl Sync for IoPipe {}
unsafe impl Sync for IoBufferfill {}
impl IoDataSync for IoClose {}
impl IoDataSync for IoFd {}
impl IoDataSync for IoFile {}
impl IoDataSync for IoPipe {}
impl IoDataSync for IoBufferfill {}
2023-04-09 20:06:40 +08:00
pub struct IoClose {
fd: RawFd,
}
impl IoClose {
pub fn new(fd: RawFd) -> Self {
IoClose { fd }
}
}
impl IoData for IoClose {
fn io_mode(&self) -> IoMode {
IoMode::close
}
fn fd(&self) -> RawFd {
self.fd
}
fn source_fd(&self) -> RawFd {
-1
}
fn print(&self) {
eprintf!("close %d\n", self.fd)
2023-04-09 20:06:40 +08:00
}
fn as_ptr(&self) -> *const () {
(self as *const Self).cast()
}
}
pub struct IoFd {
fd: RawFd,
source_fd: RawFd,
}
impl IoFd {
/// fd to redirect specified fd to. For example, in 2>&1, source_fd is 1, and io_data_t::fd
/// is 2.
pub fn new(fd: RawFd, source_fd: RawFd) -> Self {
IoFd { fd, source_fd }
}
}
impl IoData for IoFd {
fn io_mode(&self) -> IoMode {
IoMode::fd
}
fn fd(&self) -> RawFd {
self.fd
}
fn source_fd(&self) -> RawFd {
self.source_fd
}
fn print(&self) {
eprintf!("FD map %d -> %d\n", self.source_fd, self.fd)
2023-04-09 20:06:40 +08:00
}
fn as_ptr(&self) -> *const () {
(self as *const Self).cast()
}
}
/// Represents a redirection to or from an opened file.
pub struct IoFile {
fd: RawFd,
2024-03-23 13:44:27 +08:00
// The file which we are writing to or reading from.
file: File,
2023-04-09 20:06:40 +08:00
}
impl IoFile {
2024-03-23 13:44:27 +08:00
pub fn new(fd: RawFd, file: File) -> Self {
IoFile { fd, file }
2023-04-09 20:06:40 +08:00
// Invalid file redirections are replaced with a closed fd, so the following
// assertion isn't guaranteed to pass:
// assert(file_fd_.valid() && "File is not valid");
}
}
impl IoData for IoFile {
fn io_mode(&self) -> IoMode {
IoMode::file
}
fn fd(&self) -> RawFd {
self.fd
}
fn source_fd(&self) -> RawFd {
2024-03-23 13:44:27 +08:00
self.file.as_raw_fd()
2023-04-09 20:06:40 +08:00
}
fn print(&self) {
2024-03-23 13:44:27 +08:00
eprintf!("file %d -> %d\n", self.file.as_raw_fd(), self.fd)
2023-04-09 20:06:40 +08:00
}
fn as_ptr(&self) -> *const () {
(self as *const Self).cast()
}
}
/// Represents (one end) of a pipe.
pub struct IoPipe {
fd: RawFd,
// The pipe's fd. Conceptually this is dup2'd to io_data_t::fd.
2024-01-29 02:07:06 +08:00
pipe_fd: OwnedFd,
2023-04-09 20:06:40 +08:00
/// Whether this is an input pipe. This is used only for informational purposes.
is_input: bool,
}
impl IoPipe {
2024-01-29 02:07:06 +08:00
pub fn new(fd: RawFd, is_input: bool, pipe_fd: OwnedFd) -> Self {
2023-04-09 20:06:40 +08:00
IoPipe {
fd,
pipe_fd,
is_input,
}
}
}
impl IoData for IoPipe {
fn io_mode(&self) -> IoMode {
IoMode::pipe
}
fn fd(&self) -> RawFd {
self.fd
}
fn source_fd(&self) -> RawFd {
2024-01-29 02:07:06 +08:00
self.pipe_fd.as_raw_fd()
2023-04-09 20:06:40 +08:00
}
fn print(&self) {
eprintf!(
2023-04-09 20:06:40 +08:00
"pipe {%d} (input: %s) -> %d\n",
self.source_fd(),
if self.is_input { "yes" } else { "no" },
self.fd
)
}
fn as_ptr(&self) -> *const () {
(self as *const Self).cast()
}
}
/// Represents filling an io_buffer_t. Very similar to io_pipe_t.
pub struct IoBufferfill {
target: RawFd,
/// Write end. The other end is connected to an io_buffer_t.
2024-01-29 02:07:06 +08:00
write_fd: OwnedFd,
2023-04-09 20:06:40 +08:00
/// The receiving buffer.
buffer: Arc<IoBuffer>,
2023-04-09 20:06:40 +08:00
}
impl IoBufferfill {
/// Create an io_bufferfill_t which, when written from, fills a buffer with the contents.
/// Returns an error on failure, e.g. too many open fds.
2024-02-10 11:49:07 +08:00
pub fn create() -> io::Result<Arc<IoBufferfill>> {
Self::create_opts(0, STDOUT_FILENO)
}
2023-04-09 20:06:40 +08:00
/// Create an io_bufferfill_t which, when written from, fills a buffer with the contents.
/// Returns an error on failure, e.g. too many open fds.
2023-04-09 20:06:40 +08:00
///
/// \param target the fd which this will be dup2'd to - typically stdout.
2024-02-10 11:49:07 +08:00
pub fn create_opts(buffer_limit: usize, target: RawFd) -> io::Result<Arc<IoBufferfill>> {
2023-04-09 20:06:40 +08:00
assert!(target >= 0, "Invalid target fd");
// Construct our pipes.
let pipes = make_autoclose_pipes()?;
// Our buffer will read from the read end of the pipe. This end must be non-blocking. This is
// because our fillthread needs to poll to decide if it should shut down, and also accept input
// from direct buffer transfers.
2024-01-29 02:07:06 +08:00
match make_fd_nonblocking(pipes.read.as_raw_fd()) {
Ok(_) => (),
Err(e) => {
FLOG!(warning, PIPE_ERROR);
perror_io("fcntl", &e);
2024-02-10 11:49:07 +08:00
return Err(e);
}
2023-04-09 20:06:40 +08:00
}
// Our fillthread gets the read end of the pipe; out_pipe gets the write end.
let buffer = Arc::new(IoBuffer::new(buffer_limit));
begin_filling(&buffer, pipes.read);
2024-02-10 11:49:07 +08:00
Ok(Arc::new(IoBufferfill {
2023-04-09 20:06:40 +08:00
target,
write_fd: pipes.write,
buffer,
}))
}
pub fn buffer_ref(&self) -> &Arc<IoBuffer> {
&self.buffer
}
pub fn buffer(&self) -> &IoBuffer {
&self.buffer
2023-04-09 20:06:40 +08:00
}
/// Reset the receiver (possibly closing the write end of the pipe), and complete the fillthread
/// of the buffer. Return the buffer.
pub fn finish(filler: Arc<IoBufferfill>) -> SeparatedBuffer {
2023-04-09 20:06:40 +08:00
// The io filler is passed in. This typically holds the only instance of the write side of the
// pipe used by the buffer's fillthread (except for that side held by other processes). Get the
// buffer out of the bufferfill and clear the shared_ptr; this will typically widow the pipe.
// Then allow the buffer to finish.
filler
.buffer
.complete_background_fillthread_and_take_buffer()
}
}
impl IoData for IoBufferfill {
fn io_mode(&self) -> IoMode {
IoMode::bufferfill
}
fn fd(&self) -> RawFd {
self.target
}
fn source_fd(&self) -> RawFd {
2024-01-29 02:07:06 +08:00
self.write_fd.as_raw_fd()
2023-04-09 20:06:40 +08:00
}
fn print(&self) {
2024-01-29 02:07:06 +08:00
eprintf!(
"bufferfill %d -> %d\n",
self.write_fd.as_raw_fd(),
self.fd()
)
2023-04-09 20:06:40 +08:00
}
fn as_ptr(&self) -> *const () {
(self as *const Self).cast()
}
fn as_bufferfill(&self) -> Option<&IoBufferfill> {
Some(self)
}
2023-04-09 20:06:40 +08:00
}
/// An io_buffer_t is a buffer which can populate itself by reading from an fd.
/// It is not an io_data_t.
pub struct IoBuffer {
/// Buffer storing what we have read.
buffer: Mutex<SeparatedBuffer>,
/// Atomic flag indicating our fillthread should shut down.
shutdown_fillthread: RelaxedAtomicBool,
/// A promise, allowing synchronization with the background fill operation.
/// The operation has a reference to this as well, and fulfills this promise when it exits.
#[allow(clippy::type_complexity)]
fill_waiter: RefCell<Option<Arc<(Mutex<bool>, Condvar)>>>,
2023-04-09 20:06:40 +08:00
/// The item id of our background fillthread fd monitor item.
item_id: AtomicU64,
2023-04-09 20:06:40 +08:00
}
// safety: todo!("rationalize why fill_waiter is safe")
unsafe impl Send for IoBuffer {}
unsafe impl Sync for IoBuffer {}
2023-04-09 20:06:40 +08:00
impl IoBuffer {
pub fn new(limit: usize) -> Self {
IoBuffer {
buffer: Mutex::new(SeparatedBuffer::new(limit)),
shutdown_fillthread: RelaxedAtomicBool::new(false),
fill_waiter: RefCell::new(None),
item_id: AtomicU64::new(0),
2023-04-09 20:06:40 +08:00
}
}
/// Append a string to the buffer.
pub fn append(&self, data: &[u8], typ: SeparationType) -> bool {
2023-04-09 20:06:40 +08:00
self.buffer.lock().unwrap().append(data, typ)
}
/// Return true if output was discarded due to exceeding the read limit.
2023-04-09 20:06:40 +08:00
pub fn discarded(&self) -> bool {
self.buffer.lock().unwrap().discarded()
}
/// Read some, filling the buffer. The buffer is passed in to enforce that the append lock is
/// held. Return positive on success, 0 if closed, -1 on error (in which case errno will be
2023-04-09 20:06:40 +08:00
/// set).
pub fn read_once(fd: RawFd, buffer: &mut MutexGuard<'_, SeparatedBuffer>) -> isize {
assert!(fd >= 0, "Invalid fd");
errno::set_errno(Errno(0));
let mut bytes = [b'\0'; 4096 * 4];
// We want to swallow EINTR only; in particular EAGAIN needs to be returned back to the caller.
let amt = loop {
let amt = unsafe {
libc::read(
fd,
std::ptr::addr_of_mut!(bytes).cast(),
std::mem::size_of_val(&bytes),
)
};
if amt < 0 && errno::errno().0 == EINTR {
continue;
}
break amt;
};
if amt < 0 && ![EAGAIN, EWOULDBLOCK].contains(&errno::errno().0) {
perror("read");
} else if amt > 0 {
buffer.append(
&bytes[0..usize::try_from(amt).unwrap()],
SeparationType::inferred,
);
}
amt
}
/// End the background fillthread operation, and return the buffer, transferring ownership.
pub fn complete_background_fillthread_and_take_buffer(&self) -> SeparatedBuffer {
2023-04-09 20:06:40 +08:00
// Mark that our fillthread is done, then wake it up.
assert!(self.fillthread_running(), "Should have a fillthread");
assert!(
self.item_id.load(Ordering::SeqCst) != 0,
2023-04-09 20:06:40 +08:00
"Should have a valid item ID"
);
self.shutdown_fillthread.store(true);
fd_monitor().poke_item(FdMonitorItemId::from(self.item_id.load(Ordering::SeqCst)));
2023-04-09 20:06:40 +08:00
// Wait for the fillthread to fulfill its promise, and then clear the future so we know we no
// longer have one.
let mut promise = self.fill_waiter.borrow_mut();
let (mutex, condvar) = &**promise.as_ref().unwrap();
2023-04-09 20:06:40 +08:00
{
2023-12-10 17:34:52 +08:00
let done_guard = mutex.lock().unwrap();
let _done_guard = condvar.wait_while(done_guard, |done| !*done).unwrap();
2023-04-09 20:06:40 +08:00
}
*promise = None;
2023-04-09 20:06:40 +08:00
// Return our buffer, transferring ownership.
let mut locked_buff = self.buffer.lock().unwrap();
let mut result = SeparatedBuffer::new(locked_buff.limit());
std::mem::swap(&mut result, &mut locked_buff);
locked_buff.clear();
result
}
/// Helper to return whether the fillthread is running.
pub fn fillthread_running(&self) -> bool {
return self.fill_waiter.borrow().is_some();
2023-04-09 20:06:40 +08:00
}
}
/// Begin the fill operation, reading from the given fd in the background.
2024-01-29 02:07:06 +08:00
fn begin_filling(iobuffer: &Arc<IoBuffer>, fd: OwnedFd) {
assert!(!iobuffer.fillthread_running(), "Already have a fillthread");
2023-04-09 20:06:40 +08:00
// We want to fill buffer_ by reading from fd. fd is the read end of a pipe; the write end is
// owned by another process, or something else writing in fish.
// Pass fd to an fd_monitor. It will add fd to its select() loop, and give us a callback when
// the fd is readable, or when our item is poked. The usual path is that we will get called
// back, read a bit from the fd, and append it to the buffer. Eventually the write end of the
// pipe will be closed - probably the other process exited - and fd will be widowed; read() will
// then return 0 and we will stop reading.
// In exotic circumstances the write end of the pipe will not be closed; this may happen in
// e.g.:
// cmd ( background & ; echo hi )
// Here the background process will inherit the write end of the pipe and hold onto it forever.
// In this case, when complete_background_fillthread() is called, the callback will be invoked
// with item_wake_reason_t::poke, and we will notice that the shutdown flag is set (this
// indicates that the command substitution is done); in this case we will read until we get
// EAGAIN and then give up.
// Construct a promise. We will fulfill it in our fill thread, and wait for it in
// complete_background_fillthread(). Note that TSan complains if the promise's dtor races with
// the future's call to wait(), so we store the promise, not just its future (#7681).
let promise = Arc::new((Mutex::new(false), Condvar::new()));
iobuffer.fill_waiter.replace(Some(promise.clone()));
2023-04-09 20:06:40 +08:00
// Run our function to read until the receiver is closed.
// It's OK to capture 'buffer' because 'this' waits for the promise in its dtor.
2024-01-14 04:10:24 +08:00
let item_callback: Callback = {
2023-04-09 20:06:40 +08:00
let iobuffer = iobuffer.clone();
2024-01-14 04:10:24 +08:00
Box::new(move |fd: &mut AutoCloseFd, reason: ItemWakeReason| {
// Only check the shutdown flag if we timed out or were poked.
// It's important that if select() indicated we were readable, that we call select() again
// allowing it to time out. Note the typical case is that the fd will be closed, in which
// case select will return immediately.
let mut done = false;
if reason == ItemWakeReason::Readable {
// select() reported us as readable; read a bit.
let mut buf = iobuffer.buffer.lock().unwrap();
2024-01-29 02:07:06 +08:00
let ret = IoBuffer::read_once(fd.as_raw_fd(), &mut buf);
2024-01-14 04:10:24 +08:00
done = ret == 0 || (ret < 0 && ![EAGAIN, EWOULDBLOCK].contains(&errno::errno().0));
} else if iobuffer.shutdown_fillthread.load() {
// Here our caller asked us to shut down; read while we keep getting data.
// This will stop when the fd is closed or if we get EAGAIN.
let mut buf = iobuffer.buffer.lock().unwrap();
loop {
2024-01-29 02:07:06 +08:00
let ret = IoBuffer::read_once(fd.as_raw_fd(), &mut buf);
2024-01-14 04:10:24 +08:00
if ret <= 0 {
break;
2023-04-09 20:06:40 +08:00
}
}
2024-01-14 04:10:24 +08:00
done = true;
}
2024-01-14 04:35:55 +08:00
if !done {
ItemAction::Retain
} else {
2024-01-14 04:10:24 +08:00
fd.close();
let (mutex, condvar) = &*promise;
{
let mut done = mutex.lock().unwrap();
*done = true;
2023-04-09 20:06:40 +08:00
}
2024-01-14 04:10:24 +08:00
condvar.notify_one();
2024-01-14 04:35:55 +08:00
ItemAction::Remove
2024-01-14 04:10:24 +08:00
}
})
2023-04-09 20:06:40 +08:00
};
2024-01-29 02:07:06 +08:00
let fd = AutoCloseFd::new(fd.into_raw_fd());
let item_id = fd_monitor().add(FdMonitorItem::new(fd, None, item_callback));
iobuffer.item_id.store(u64::from(item_id), Ordering::SeqCst);
2023-04-09 20:06:40 +08:00
}
pub type IoDataRef = Arc<dyn IoDataSync>;
2023-04-09 20:06:40 +08:00
#[derive(Clone, Default)]
2023-04-09 20:06:40 +08:00
pub struct IoChain(pub Vec<IoDataRef>);
impl IoChain {
pub fn new() -> Self {
Default::default()
}
pub fn remove(&mut self, element: &dyn IoDataSync) {
let element = element as *const _;
let element = element as *const ();
2023-04-09 20:06:40 +08:00
self.0.retain(|e| {
let e = Arc::as_ptr(e) as *const ();
2023-04-09 20:06:40 +08:00
!std::ptr::eq(e, element)
});
}
pub fn clear(&mut self) {
self.0.clear()
}
2023-04-09 20:06:40 +08:00
pub fn push(&mut self, element: IoDataRef) {
self.0.push(element);
}
pub fn append(&mut self, chain: &IoChain) -> bool {
self.0.extend_from_slice(&chain.0);
true
}
/// Return the last io redirection in the chain for the specified file descriptor, or nullptr
2023-04-09 20:06:40 +08:00
/// if none.
pub fn io_for_fd(&self, fd: RawFd) -> Option<IoDataRef> {
self.0.iter().rev().find(|data| data.fd() == fd).cloned()
}
/// Attempt to resolve a list of redirection specs to IOs, appending to 'this'.
/// Return true on success, false on error, in which case an error will have been printed.
#[allow(clippy::collapsible_else_if)]
2023-04-09 20:06:40 +08:00
pub fn append_from_specs(&mut self, specs: &RedirectionSpecList, pwd: &wstr) -> bool {
let mut have_error = false;
let print_error = |err, target: &wstr| {
// If the error is that the file doesn't exist
// or there's a non-directory component,
// find the first problematic component for a better message.
if [ENOENT, ENOTDIR].contains(&err) {
FLOGF!(warning, FILE_ERROR, target);
let mut dname: &wstr = target;
while !dname.is_empty() {
let next: &wstr = wdirname(dname);
if let Ok(md) = wstat(next) {
if !md.is_dir() {
FLOGF!(warning, "Path '%ls' is not a directory", next);
} else {
FLOGF!(warning, "Path '%ls' does not exist", dname);
}
break;
}
dname = next;
}
} else if err != EINTR {
// If we get EINTR we had a cancel signal.
// That's expected (ctrl-c on the commandline),
// so no warning.
FLOGF!(warning, FILE_ERROR, target);
perror("open");
}
};
2023-04-09 20:06:40 +08:00
for spec in specs {
match spec.mode {
RedirectionMode::fd => {
if spec.is_close() {
self.push(Arc::new(IoClose::new(spec.fd)));
2023-04-09 20:06:40 +08:00
} else {
let target_fd = spec
.get_target_as_fd()
.expect("fd redirection should have been validated already");
self.push(Arc::new(IoFd::new(spec.fd, target_fd)));
2023-04-09 20:06:40 +08:00
}
}
_ => {
// We have a path-based redirection. Resolve it to a file.
2023-04-09 20:06:40 +08:00
// Mark it as CLO_EXEC because we don't want it to be open in any child.
let path = path_apply_working_directory(&spec.target, pwd);
let oflags = spec.oflags();
2024-01-21 06:13:13 +08:00
match wopen_cloexec(&path, oflags, OPEN_MASK) {
Ok(file) => {
2024-03-23 13:44:27 +08:00
self.push(Arc::new(IoFile::new(spec.fd, file)));
2024-01-21 06:13:13 +08:00
}
Err(err) => {
if oflags.contains(OFlag::O_EXCL) && err == nix::Error::EEXIST {
2024-01-21 06:13:13 +08:00
FLOGF!(warning, NOCLOB_ERROR, spec.target);
} else if spec.mode != RedirectionMode::try_input {
2024-01-21 06:13:13 +08:00
if should_flog!(warning) {
print_error(errno::errno().0, &spec.target);
2023-04-09 20:06:40 +08:00
}
}
2024-01-21 06:13:13 +08:00
// If opening a file fails, insert a closed FD instead of the file redirection
// and return false. This lets execution potentially recover and at least gives
// the shell a chance to gracefully regain control of the shell (see #7038).
if spec.mode != RedirectionMode::try_input {
self.push(Arc::new(IoClose::new(spec.fd)));
have_error = true;
continue;
} else {
// If we're told to try via `<?`, we use /dev/null
match wopen_cloexec(L!("/dev/null"), oflags, OPEN_MASK) {
Ok(fd) => {
self.push(Arc::new(IoFile::new(spec.fd, fd)));
}
_ => {
// /dev/null can't be opened???
if should_flog!(warning) {
print_error(errno::errno().0, L!("/dev/null"));
}
self.push(Arc::new(IoClose::new(spec.fd)));
have_error = true;
continue;
}
}
}
2023-04-09 20:06:40 +08:00
}
}
}
}
}
!have_error
}
/// Output debugging information to stderr.
pub fn print(&self) {
if self.0.is_empty() {
eprintf!(
2023-04-09 20:06:40 +08:00
"Empty chain %s\n",
format!("{:p}", std::ptr::addr_of!(self))
);
return;
}
eprintf!(
2023-04-09 20:06:40 +08:00
"Chain %s (%ld items):\n",
format!("{:p}", std::ptr::addr_of!(self)),
self.0.len()
);
for (i, io) in self.0.iter().enumerate() {
eprintf!("\t%lu: fd:%d, ", i, io.fd());
2023-04-09 20:06:40 +08:00
io.print();
}
}
}
/// Base class representing the output that a builtin can generate.
/// This has various subclasses depending on the ultimate output destination.
pub enum OutputStream {
/// A null output stream which ignores all writes.
Null,
Fd(FdOutputStream),
String(StringOutputStream),
Buffered(BufferedOutputStream),
}
2023-04-09 20:06:40 +08:00
impl OutputStream {
/// Return any internally buffered contents.
2023-04-09 20:06:40 +08:00
/// This is only implemented for a string_output_stream; others flush data to their underlying
/// receiver (fd, or separated buffer) immediately and so will return an empty string here.
pub fn contents(&self) -> &wstr {
match self {
OutputStream::String(stream) => stream.contents(),
OutputStream::Null | OutputStream::Fd(_) | OutputStream::Buffered(_) => &EMPTY_STRING,
}
2023-04-09 20:06:40 +08:00
}
/// Flush any unwritten data to the underlying device, and return an error code.
/// A 0 code indicates success. The base implementation returns 0.
pub fn flush_and_check_error(&mut self) -> libc::c_int {
match self {
OutputStream::Fd(stream) => stream.flush_and_check_error(),
OutputStream::Buffered(stream) => stream.flush_and_check_error(),
OutputStream::Null | OutputStream::String(_) => STATUS_CMD_OK.unwrap(),
}
}
/// Append a &wstr or WString.
pub fn append<Str: AsRef<wstr>>(&mut self, s: Str) -> bool {
let s = &s.as_ref();
match self {
OutputStream::Null => true,
OutputStream::Fd(stream) => stream.append(s),
OutputStream::String(stream) => stream.append(s),
OutputStream::Buffered(stream) => stream.append(s),
}
2023-04-09 20:06:40 +08:00
}
/// An optional override point. This is for explicit separation.
/// \param want_newline this is true if the output item should be ended with a newline. This
/// is only relevant if we are printing the output to a stream,
pub fn append_with_separation(
2023-04-09 20:06:40 +08:00
&mut self,
s: &wstr,
typ: SeparationType,
want_newline: bool,
) -> bool {
match self {
OutputStream::Buffered(stream) => stream.append_with_separation(s, typ, want_newline),
OutputStream::Fd(_) | OutputStream::Null | OutputStream::String(_) => {
if typ == SeparationType::explicitly && want_newline {
// Try calling "append" less - it might write() to an fd
let mut buf = s.to_owned();
buf.push('\n');
self.append(buf)
} else {
self.append(s)
}
}
2023-04-09 20:06:40 +08:00
}
}
/// Append a &wstr or WString with a newline
pub fn appendln(&mut self, s: impl Into<WString>) -> bool {
let s = s.into() + L!("\n");
self.append(s)
}
pub fn append_char(&mut self, c: char) -> bool {
2023-04-09 20:06:40 +08:00
self.append(wstr::from_char_slice(&[c]))
}
pub fn append1(&mut self, c: char) -> bool {
self.append_char(c)
}
pub fn push_back(&mut self, c: char) -> bool {
2023-04-09 20:06:40 +08:00
self.append_char(c)
}
pub fn push(&mut self, c: char) -> bool {
2023-04-09 20:06:40 +08:00
self.append(wstr::from_char_slice(&[c]))
}
// Append data from a narrow buffer, widening it.
pub fn append_narrow_buffer(&mut self, buffer: &SeparatedBuffer) -> bool {
2023-04-09 20:06:40 +08:00
for rhs_elem in buffer.elements() {
if !self.append_with_separation(
&str2wcstring(&rhs_elem.contents),
rhs_elem.separation,
false,
) {
return false;
}
}
true
}
}
/// An output stream for builtins which outputs to an fd.
/// Note the fd may be something like stdout; there is no ownership implied here.
pub struct FdOutputStream {
/// The file descriptor to write to.
fd: RawFd,
/// Used to check if a SIGINT has been received when EINTR is encountered
sigcheck: SigChecker,
2023-04-09 20:06:40 +08:00
/// Whether we have received an error.
errored: bool,
}
impl FdOutputStream {
/// Construct from a file descriptor, which must be nonegative.
pub fn new(fd: RawFd) -> Self {
assert!(fd >= 0, "Invalid fd");
FdOutputStream {
fd,
sigcheck: SigChecker::new(topic_t::sighupint),
2023-04-09 20:06:40 +08:00
errored: false,
}
}
2023-04-09 20:06:40 +08:00
fn append(&mut self, s: &wstr) -> bool {
if self.errored {
return false;
}
if wwrite_to_fd(s, self.fd).is_none() {
// Some of our builtins emit multiple screens worth of data sent to a pager (the primary
// example being the `history` builtin) and receiving SIGINT should be considered normal and
// non-exceptional (user request to abort via Ctrl-C), meaning we shouldn't print an error.
if errno::errno().0 == EINTR && self.sigcheck.check() {
// We have two options here: we can either return false without setting errored_ to
// true (*this* write will be silently aborted but the onus is on the caller to check
// the return value and skip future calls to `append()`) or we can flag the entire
// output stream as errored, causing us to both return false and skip any future writes.
// We're currently going with the latter, especially seeing as no callers currently
// check the result of `append()` (since it was always a void function before).
} else if errno::errno().0 != EPIPE {
perror("write");
}
self.errored = true;
}
!self.errored
}
fn flush_and_check_error(&mut self) -> libc::c_int {
// Return a generic 1 on any write failure.
if self.errored {
STATUS_CMD_ERROR
} else {
STATUS_CMD_OK
}
.unwrap()
}
}
/// A simple output stream which buffers into a wcstring.
#[derive(Default)]
pub struct StringOutputStream {
contents: WString,
}
impl StringOutputStream {
pub fn new() -> Self {
Default::default()
}
2023-04-09 20:06:40 +08:00
fn append(&mut self, s: &wstr) -> bool {
self.contents.push_utfstr(s);
true
}
/// Return the wcstring containing the output.
2023-04-09 20:06:40 +08:00
fn contents(&self) -> &wstr {
&self.contents
}
}
/// An output stream for builtins which writes into a separated buffer.
pub struct BufferedOutputStream {
/// The buffer we are filling.
buffer: Arc<IoBuffer>,
2023-04-09 20:06:40 +08:00
}
impl BufferedOutputStream {
pub fn new(buffer: Arc<IoBuffer>) -> Self {
2023-04-09 20:06:40 +08:00
Self { buffer }
}
fn append(&mut self, s: &wstr) -> bool {
self.buffer.append(&wcs2string(s), SeparationType::inferred)
2023-04-09 20:06:40 +08:00
}
fn append_with_separation(
&mut self,
s: &wstr,
typ: SeparationType,
_want_newline: bool,
) -> bool {
self.buffer.append(&wcs2string(s), typ)
2023-04-09 20:06:40 +08:00
}
fn flush_and_check_error(&mut self) -> libc::c_int {
if self.buffer.discarded() {
2023-04-09 20:06:40 +08:00
return STATUS_READ_TOO_MUCH.unwrap();
}
0
}
}
pub struct IoStreams<'a> {
2023-04-09 20:06:40 +08:00
// Streams for out and err.
pub out: &'a mut OutputStream,
pub err: &'a mut OutputStream,
2023-04-09 20:06:40 +08:00
// fd representing stdin. This is not closed by the destructor.
// Note: if stdin is explicitly closed by `<&-` then this is -1!
pub stdin_fd: RawFd,
// Whether stdin is "directly redirected," meaning it is the recipient of a pipe (foo | cmd) or
// direct redirection (cmd < foo.txt). An "indirect redirection" would be e.g.
// begin ; cmd ; end < foo.txt
// If stdin is closed (cmd <&-) this is false.
pub stdin_is_directly_redirected: bool,
// Indicates whether stdout and stderr are specifically piped.
// If this is set, then the is_redirected flags must also be set.
pub out_is_piped: bool,
pub err_is_piped: bool,
// Indicates whether stdout and stderr are at all redirected (e.g. to a file or piped).
pub out_is_redirected: bool,
pub err_is_redirected: bool,
// Actual IO redirections. This is only used by the source builtin.
pub io_chain: &'a IoChain,
2023-04-09 20:06:40 +08:00
// The job group of the job, if any. This enables builtins which run more code like eval() to
// share pgid.
// FIXME: this is awkwardly placed.
pub job_group: Option<JobGroupRef>,
}
impl<'a> IoStreams<'a> {
pub fn new(
out: &'a mut OutputStream,
err: &'a mut OutputStream,
io_chain: &'a IoChain,
) -> Self {
IoStreams {
2023-04-09 20:06:40 +08:00
out,
err,
stdin_fd: -1,
stdin_is_directly_redirected: false,
out_is_piped: false,
err_is_piped: false,
out_is_redirected: false,
err_is_redirected: false,
io_chain,
2023-04-09 20:06:40 +08:00
job_group: None,
}
}
pub fn out_is_terminal(&self) -> bool {
!self.out_is_redirected && isatty(STDOUT_FILENO)
}
2023-04-09 20:06:40 +08:00
}
/// File redirection error message.
const FILE_ERROR: &wstr = L!("An error occurred while redirecting file '%ls'");
const NOCLOB_ERROR: &wstr = L!("The file '%ls' already exists");
/// Base open mode to pass to calls to open.
const OPEN_MASK: Mode = Mode::from_bits_truncate(0o666);
2023-04-09 20:06:40 +08:00
/// Provide the fd monitor used for background fillthread operations.
fn fd_monitor() -> &'static mut FdMonitor {
// Deliberately leaked to avoid shutdown dtors.
static mut FDM: *const UnsafeCell<FdMonitor> = std::ptr::null();
unsafe {
if FDM.is_null() {
FDM = Box::into_raw(Box::new(UnsafeCell::new(FdMonitor::new())))
}
}
let ptr: *mut FdMonitor = unsafe { (*FDM).get() };
unsafe { &mut *ptr }
}