view rust/hg-core/src/lock.rs @ 49491:c6a1beba27e9

bisect: avoid copying ancestor list for non-merge commits During a bisection, hg needs to compute a list of all ancestors for every candidate commit. This is accomplished via a bottom-up traversal of the set of candidates, during which each revision's ancestor list is populated using the ancestor list of its parent(s). Previously, this involved copying the entire list, which could be very long in if the bisection range was large. To help improve this, we can observe that each candidate commit is visited exactly once, at which point its ancestor list is copied into its children's lists and then dropped. In the case of non-merge commits, a commit's ancestor list consists exactly of its parent's list plus itself. This means that we can trivially reuse the parent's existing list for one of its non-merge children, which avoids copying entirely if that commit is the parent's only child. This makes bisections over linear ranges of commits much faster. During some informal testing in the large publicly-available `mozilla-central` repository, this noticeably sped up bisections over large ranges of history: Setup: $ cd mozilla-central $ hg bisect --reset $ hg bisect --good 0 $ hg log -r tip -T '{rev}\n' 628417 Test: $ time hg bisect --bad tip --noupdate Before: real 3m35.927s user 3m35.553s sys 0m0.319s After: real 1m41.142s user 1m40.810s sys 0m0.285s
author Arun Kulshreshtha <akulshreshtha@janestreet.com>
date Tue, 30 Aug 2022 15:29:55 -0400
parents 59be65b7cdfd
children 29cf3167e459
line wrap: on
line source

//! Filesystem-based locks for local repositories

use crate::errors::HgError;
use crate::errors::HgResultExt;
use crate::utils::StrExt;
use crate::vfs::Vfs;
use std::io;
use std::io::ErrorKind;

#[derive(derive_more::From)]
pub enum LockError {
    AlreadyHeld,
    #[from]
    Other(HgError),
}

/// Try to call `f` with the lock acquired, without waiting.
///
/// If the lock is aready held, `f` is not called and `LockError::AlreadyHeld`
/// is returned. `LockError::Io` is returned for any unexpected I/O error
/// accessing the lock file, including for removing it after `f` was called.
/// The return value of `f` is dropped in that case. If all is successful, the
/// return value of `f` is forwarded.
pub fn try_with_lock_no_wait<R>(
    hg_vfs: Vfs,
    lock_filename: &str,
    f: impl FnOnce() -> R,
) -> Result<R, LockError> {
    let our_lock_data = &*OUR_LOCK_DATA;
    for _retry in 0..5 {
        match make_lock(hg_vfs, lock_filename, our_lock_data) {
            Ok(()) => {
                let result = f();
                unlock(hg_vfs, lock_filename)?;
                return Ok(result);
            }
            Err(HgError::IoError { error, .. })
                if error.kind() == ErrorKind::AlreadyExists =>
            {
                let lock_data = read_lock(hg_vfs, lock_filename)?;
                if lock_data.is_none() {
                    // Lock was apparently just released, retry acquiring it
                    continue;
                }
                if !lock_should_be_broken(&lock_data) {
                    return Err(LockError::AlreadyHeld);
                }
                // The lock file is left over from a process not running
                // anymore. Break it, but with another lock to
                // avoid a race.
                break_lock(hg_vfs, lock_filename)?;

                // Retry acquiring
            }
            Err(error) => Err(error)?,
        }
    }
    Err(LockError::AlreadyHeld)
}

fn break_lock(hg_vfs: Vfs, lock_filename: &str) -> Result<(), LockError> {
    try_with_lock_no_wait(hg_vfs, &format!("{}.break", lock_filename), || {
        // Check again in case some other process broke and
        // acquired the lock in the meantime
        let lock_data = read_lock(hg_vfs, lock_filename)?;
        if !lock_should_be_broken(&lock_data) {
            return Err(LockError::AlreadyHeld);
        }
        Ok(hg_vfs.remove_file(lock_filename)?)
    })?
}

#[cfg(unix)]
fn make_lock(
    hg_vfs: Vfs,
    lock_filename: &str,
    data: &str,
) -> Result<(), HgError> {
    // Use a symbolic link because creating it is atomic.
    // The link’s "target" contains data not representing any path.
    let fake_symlink_target = data;
    hg_vfs.create_symlink(lock_filename, fake_symlink_target)
}

fn read_lock(
    hg_vfs: Vfs,
    lock_filename: &str,
) -> Result<Option<String>, HgError> {
    let link_target =
        hg_vfs.read_link(lock_filename).io_not_found_as_none()?;
    if let Some(target) = link_target {
        let data = target
            .into_os_string()
            .into_string()
            .map_err(|_| HgError::corrupted("non-UTF-8 lock data"))?;
        Ok(Some(data))
    } else {
        Ok(None)
    }
}

fn unlock(hg_vfs: Vfs, lock_filename: &str) -> Result<(), HgError> {
    hg_vfs.remove_file(lock_filename)
}

/// Return whether the process that is/was holding the lock is known not to be
/// running anymore.
fn lock_should_be_broken(data: &Option<String>) -> bool {
    (|| -> Option<bool> {
        let (prefix, pid) = data.as_ref()?.split_2(':')?;
        if prefix != &*LOCK_PREFIX {
            return Some(false);
        }
        let process_is_running;

        #[cfg(unix)]
        {
            let pid: libc::pid_t = pid.parse().ok()?;
            unsafe {
                let signal = 0; // Test if we could send a signal, without sending
                let result = libc::kill(pid, signal);
                if result == 0 {
                    process_is_running = true
                } else {
                    let errno =
                        io::Error::last_os_error().raw_os_error().unwrap();
                    process_is_running = errno != libc::ESRCH
                }
            }
        }

        Some(!process_is_running)
    })()
    .unwrap_or(false)
}

lazy_static::lazy_static! {
    /// A string which is used to differentiate pid namespaces
    ///
    /// It's useful to detect "dead" processes and remove stale locks with
    /// confidence. Typically it's just hostname. On modern linux, we include an
    /// extra Linux-specific pid namespace identifier.
    static ref LOCK_PREFIX: String = {
        // Note: this must match the behavior of `_getlockprefix` in `mercurial/lock.py`

        /// Same as https://github.com/python/cpython/blob/v3.10.0/Modules/socketmodule.c#L5414
        const BUFFER_SIZE: usize = 1024;
        let mut buffer = [0 as libc::c_char; BUFFER_SIZE];
        let hostname_bytes = unsafe {
            let result = libc::gethostname(buffer.as_mut_ptr(), BUFFER_SIZE);
            if result != 0 {
                panic!("gethostname: {}", io::Error::last_os_error())
            }
            std::ffi::CStr::from_ptr(buffer.as_mut_ptr()).to_bytes()
        };
        let hostname =
            std::str::from_utf8(hostname_bytes).expect("non-UTF-8 hostname");

        #[cfg(target_os = "linux")]
        {
            use std::os::linux::fs::MetadataExt;
            match std::fs::metadata("/proc/self/ns/pid") {
                Ok(meta) => {
                    return format!("{}/{:x}", hostname, meta.st_ino())
                }
                Err(error) => {
                    // TODO: match on `error.kind()` when `NotADirectory`
                    // is available on all supported Rust versions:
                    // https://github.com/rust-lang/rust/issues/86442
                    use libc::{
                        ENOENT, // ErrorKind::NotFound
                        ENOTDIR, // ErrorKind::NotADirectory
                        EACCES, // ErrorKind::PermissionDenied
                    };
                    match error.raw_os_error() {
                        Some(ENOENT) | Some(ENOTDIR) | Some(EACCES) => {}
                        _ => panic!("stat /proc/self/ns/pid: {}", error),
                    }
                }
            }
        }

        hostname.to_owned()
    };

    static ref OUR_LOCK_DATA: String = format!("{}:{}", &*LOCK_PREFIX, std::process::id());
}