blob: a27a6f4ea322eba18695cee811b94c25b3bbdc05 [file] [log] [blame]
use std::io::Write;
use std::ops::ControlFlow;
use std::sync::Arc;
use std::{iter, mem};
use rustc_data_structures::fx::{FxHashMap, FxHashSet};
use rustc_errors::{Diag, DiagCtxtHandle};
use rustc_hir::def::DefKind;
use rustc_middle::queries::TaggedQueryKey;
use rustc_middle::query::{Cycle, QueryJob, QueryJobId, QueryLatch, QueryStackFrame, QueryWaiter};
use rustc_middle::ty::TyCtxt;
use rustc_span::{DUMMY_SP, Span};
use crate::{CollectActiveJobsKind, collect_active_query_jobs};
/// Map from query job IDs to job information collected by
/// `collect_active_query_jobs`.
#[derive(Debug, Default)]
pub struct QueryJobMap<'tcx> {
map: FxHashMap<QueryJobId, QueryJobInfo<'tcx>>,
}
impl<'tcx> QueryJobMap<'tcx> {
/// Adds information about a job ID to the job map.
///
/// Should only be called by `collect_active_query_jobs_inner`.
pub(crate) fn insert(&mut self, id: QueryJobId, info: QueryJobInfo<'tcx>) {
self.map.insert(id, info);
}
fn tagged_key_of(&self, id: QueryJobId) -> TaggedQueryKey<'tcx> {
self.map[&id].tagged_key
}
fn span_of(&self, id: QueryJobId) -> Span {
self.map[&id].job.span
}
fn parent_of(&self, id: QueryJobId) -> Option<QueryJobId> {
self.map[&id].job.parent
}
fn latch_of(&self, id: QueryJobId) -> Option<&QueryLatch<'tcx>> {
self.map[&id].job.latch.as_ref()
}
}
#[derive(Debug)]
pub(crate) struct QueryJobInfo<'tcx> {
pub(crate) tagged_key: TaggedQueryKey<'tcx>,
pub(crate) job: QueryJob<'tcx>,
}
pub(crate) fn find_cycle_in_stack<'tcx>(
id: QueryJobId,
job_map: QueryJobMap<'tcx>,
current_job: &Option<QueryJobId>,
span: Span,
) -> Cycle<'tcx> {
// Find the waitee amongst `current_job` parents.
let mut frames = Vec::new();
let mut current_job = Option::clone(current_job);
while let Some(job) = current_job {
let info = &job_map.map[&job];
frames.push(QueryStackFrame { span: info.job.span, tagged_key: info.tagged_key });
if job == id {
frames.reverse();
// This is the end of the cycle. The span entry we included was for
// the usage of the cycle itself, and not part of the cycle.
// Replace it with the span which caused the cycle to form.
frames[0].span = span;
// Find out why the cycle itself was used.
let usage = try {
let parent = info.job.parent?;
QueryStackFrame { span: info.job.span, tagged_key: job_map.tagged_key_of(parent) }
};
return Cycle { usage, frames };
}
current_job = info.job.parent;
}
panic!("did not find a cycle")
}
/// Finds the query job closest to the root that is for the same query method as `id`
/// (but not necessarily the same query key), and returns information about it.
#[cold]
#[inline(never)]
pub(crate) fn find_dep_kind_root<'tcx>(
tcx: TyCtxt<'tcx>,
id: QueryJobId,
job_map: QueryJobMap<'tcx>,
) -> (Span, String, usize) {
let mut depth = 1;
let mut info = &job_map.map[&id];
// Two query jobs are for the same query method if they have the same
// `TaggedQueryKey` discriminant.
let expected_query = mem::discriminant::<TaggedQueryKey<'tcx>>(&info.tagged_key);
let mut last_info = info;
while let Some(id) = info.job.parent {
info = &job_map.map[&id];
if mem::discriminant(&info.tagged_key) == expected_query {
depth += 1;
last_info = info;
}
}
(last_info.job.span, last_info.tagged_key.description(tcx), depth)
}
/// The locaton of a resumable waiter. The usize is the index into waiters in the query's latch.
/// We'll use this to remove the waiter using `QueryLatch::extract_waiter` if we're waking it up.
type ResumableWaiterLocation = (QueryJobId, usize);
/// This abstracts over non-resumable waiters which are found in `QueryJob`'s `parent` field
/// and resumable waiters are in `latch` field.
struct AbstractedWaiter {
/// The span corresponding to the reason for why we're waiting on this query.
span: Span,
/// The query which we are waiting from, if none the waiter is from a compiler root.
parent: Option<QueryJobId>,
resumable: Option<ResumableWaiterLocation>,
}
/// Returns all the non-resumable and resumable waiters of a query.
/// This is used so we can uniformly loop over both non-resumable and resumable waiters.
fn abstracted_waiters_of(job_map: &QueryJobMap<'_>, query: QueryJobId) -> Vec<AbstractedWaiter> {
let mut result = Vec::new();
// Add the parent which is a non-resumable waiter since it's on the same stack
result.push(AbstractedWaiter {
span: job_map.span_of(query),
parent: job_map.parent_of(query),
resumable: None,
});
// Add the explicit waiters which use condvars and are resumable
if let Some(latch) = job_map.latch_of(query) {
for (i, waiter) in latch.waiters.lock().as_ref().unwrap().iter().enumerate() {
result.push(AbstractedWaiter {
span: waiter.span,
parent: waiter.parent,
resumable: Some((query, i)),
});
}
}
result
}
/// Looks for a query cycle by doing a depth first search starting at `query`.
/// `span` is the reason for the `query` to execute. This is initially DUMMY_SP.
/// If a cycle is detected, this initial value is replaced with the span causing
/// the cycle. `stack` will contain just the cycle on return if detected.
fn find_cycle<'tcx>(
job_map: &QueryJobMap<'tcx>,
query: QueryJobId,
span: Span,
stack: &mut Vec<(Span, QueryJobId)>,
visited: &mut FxHashSet<QueryJobId>,
) -> ControlFlow<Option<ResumableWaiterLocation>> {
if !visited.insert(query) {
return if let Some(pos) = stack.iter().position(|q| q.1 == query) {
// We detected a query cycle, fix up the initial span and return Some
// Remove previous stack entries
stack.drain(0..pos);
// Replace the span for the first query with the cycle cause
stack[0].0 = span;
ControlFlow::Break(None)
} else {
ControlFlow::Continue(())
};
}
// Query marked as visited is added it to the stack
stack.push((span, query));
// Visit all the waiters
for abstracted_waiter in abstracted_waiters_of(job_map, query) {
let Some(parent) = abstracted_waiter.parent else {
// Skip waiters which are not queries
continue;
};
if let ControlFlow::Break(maybe_resumable) =
find_cycle(job_map, parent, abstracted_waiter.span, stack, visited)
{
// Return the resumable waiter in `waiter.resumable` if present
return ControlFlow::Break(abstracted_waiter.resumable.or(maybe_resumable));
}
}
// Remove the entry in our stack since we didn't find a cycle
stack.pop();
ControlFlow::Continue(())
}
/// Finds out if there's a path to the compiler root (aka. code which isn't in a query)
/// from `query` without going through any of the queries in `visited`.
/// This is achieved with a depth first search.
fn connected_to_root<'tcx>(
job_map: &QueryJobMap<'tcx>,
query: QueryJobId,
visited: &mut FxHashSet<QueryJobId>,
) -> bool {
// We already visited this or we're deliberately ignoring it
if !visited.insert(query) {
return false;
}
// Visit all the waiters
for abstracted_waiter in abstracted_waiters_of(job_map, query) {
match abstracted_waiter.parent {
// This query is connected to the root
None => return true,
Some(parent) => {
if connected_to_root(job_map, parent, visited) {
return true;
}
}
}
}
false
}
/// Looks for a query cycle using the last query in `jobs`.
/// If a cycle is found, all queries in the cycle is removed from `jobs` and
/// the function return true.
/// If a cycle was not found, the starting query is removed from `jobs` and
/// the function returns false.
fn remove_cycle<'tcx>(
job_map: &QueryJobMap<'tcx>,
jobs: &mut Vec<QueryJobId>,
wakelist: &mut Vec<Arc<QueryWaiter<'tcx>>>,
) -> bool {
let mut visited = FxHashSet::default();
let mut stack = Vec::new();
// Look for a cycle starting with the last query in `jobs`
if let ControlFlow::Break(resumable) =
find_cycle(job_map, jobs.pop().unwrap(), DUMMY_SP, &mut stack, &mut visited)
{
// The stack is a vector of pairs of spans and queries; reverse it so that
// the earlier entries require later entries
let (mut spans, queries): (Vec<_>, Vec<_>) = stack.into_iter().rev().unzip();
// Shift the spans so that queries are matched with the span for their waitee
spans.rotate_right(1);
// Zip them back together
let mut stack: Vec<_> = iter::zip(spans, queries).collect();
// Remove the queries in our cycle from the list of jobs to look at
for r in &stack {
if let Some(pos) = jobs.iter().position(|j| j == &r.1) {
jobs.remove(pos);
}
}
struct EntryPoint {
query_in_cycle: QueryJobId,
query_waiting_on_cycle: Option<(Span, QueryJobId)>,
}
// Find the queries in the cycle which are
// connected to queries outside the cycle
let entry_points = stack
.iter()
.filter_map(|&(_, query_in_cycle)| {
let mut entrypoint = false;
let mut query_waiting_on_cycle = None;
// Find a direct waiter who leads to the root
for abstracted_waiter in abstracted_waiters_of(job_map, query_in_cycle) {
let Some(parent) = abstracted_waiter.parent else {
// The query in the cycle is directly connected to root.
entrypoint = true;
continue;
};
// Mark all the other queries in the cycle as already visited,
// so paths to the root through the cycle itself won't count.
let mut visited = FxHashSet::from_iter(stack.iter().map(|q| q.1));
if connected_to_root(job_map, parent, &mut visited) {
query_waiting_on_cycle = Some((abstracted_waiter.span, parent));
entrypoint = true;
break;
}
}
entrypoint.then_some(EntryPoint { query_in_cycle, query_waiting_on_cycle })
})
.collect::<Vec<EntryPoint>>();
// Pick an entry point, preferring ones with waiters
let entry_point = entry_points
.iter()
.find(|entry_point| entry_point.query_waiting_on_cycle.is_some())
.unwrap_or(&entry_points[0]);
// Shift the stack so that our entry point is first
let entry_point_pos =
stack.iter().position(|(_, query)| *query == entry_point.query_in_cycle);
if let Some(pos) = entry_point_pos {
stack.rotate_left(pos);
}
let usage = entry_point
.query_waiting_on_cycle
.map(|(span, job)| QueryStackFrame { span, tagged_key: job_map.tagged_key_of(job) });
// Create the cycle error
let error = Cycle {
usage,
frames: stack
.iter()
.map(|&(span, job)| QueryStackFrame {
span,
tagged_key: job_map.tagged_key_of(job),
})
.collect(),
};
// We unwrap `resumable` here since there must always be one
// edge which is resumable / waited using a query latch
let (waitee_query, waiter_idx) = resumable.unwrap();
// Extract the waiter we want to resume
let waiter = job_map.latch_of(waitee_query).unwrap().extract_waiter(waiter_idx);
// Set the cycle error so it will be picked up when resumed
*waiter.cycle.lock() = Some(error);
// Put the waiter on the list of things to resume
wakelist.push(waiter);
true
} else {
false
}
}
/// Detects query cycles by using depth first search over all active query jobs.
/// If a query cycle is found it will break the cycle by finding an edge which
/// uses a query latch and then resuming that waiter.
/// There may be multiple cycles involved in a deadlock, so this searches
/// all active queries for cycles before finally resuming all the waiters at once.
pub fn break_query_cycles<'tcx>(
job_map: QueryJobMap<'tcx>,
registry: &rustc_thread_pool::Registry,
) {
let mut wakelist = Vec::new();
// It is OK per the comments:
// - https://github.com/rust-lang/rust/pull/131200#issuecomment-2798854932
// - https://github.com/rust-lang/rust/pull/131200#issuecomment-2798866392
#[allow(rustc::potential_query_instability)]
let mut jobs: Vec<QueryJobId> = job_map.map.keys().copied().collect();
let mut found_cycle = false;
while jobs.len() > 0 {
if remove_cycle(&job_map, &mut jobs, &mut wakelist) {
found_cycle = true;
}
}
// Check that a cycle was found. It is possible for a deadlock to occur without
// a query cycle if a query which can be waited on uses Rayon to do multithreading
// internally. Such a query (X) may be executing on 2 threads (A and B) and A may
// wait using Rayon on B. Rayon may then switch to executing another query (Y)
// which in turn will wait on X causing a deadlock. We have a false dependency from
// X to Y due to Rayon waiting and a true dependency from Y to X. The algorithm here
// only considers the true dependency and won't detect a cycle.
if !found_cycle {
panic!(
"deadlock detected as we're unable to find a query cycle to break\n\
current query map:\n{job_map:#?}",
);
}
// Mark all the thread we're about to wake up as unblocked. This needs to be done before
// we wake the threads up as otherwise Rayon could detect a deadlock if a thread we
// resumed fell asleep and this thread had yet to mark the remaining threads as unblocked.
for _ in 0..wakelist.len() {
rustc_thread_pool::mark_unblocked(registry);
}
for waiter in wakelist.into_iter() {
waiter.condvar.notify_one();
}
}
pub fn print_query_stack<'tcx>(
tcx: TyCtxt<'tcx>,
mut current_query: Option<QueryJobId>,
dcx: DiagCtxtHandle<'_>,
limit_frames: Option<usize>,
mut file: Option<std::fs::File>,
) -> usize {
// Be careful relying on global state here: this code is called from
// a panic hook, which means that the global `DiagCtxt` may be in a weird
// state if it was responsible for triggering the panic.
let mut count_printed = 0;
let mut count_total = 0;
// Make use of a partial query job map if we fail to take locks collecting active queries.
let job_map = collect_active_query_jobs(tcx, CollectActiveJobsKind::PartialAllowed);
if let Some(ref mut file) = file {
let _ = writeln!(file, "\n\nquery stack during panic:");
}
while let Some(query) = current_query {
let Some(query_info) = job_map.map.get(&query) else {
break;
};
let description = query_info.tagged_key.description(tcx);
if Some(count_printed) < limit_frames || limit_frames.is_none() {
// Only print to stderr as many stack frames as `num_frames` when present.
dcx.struct_failure_note(format!(
"#{count_printed} [{query_name}] {description}",
query_name = query_info.tagged_key.query_name(),
))
.with_span(query_info.job.span)
.emit();
count_printed += 1;
}
if let Some(ref mut file) = file {
let _ = writeln!(
file,
"#{count_total} [{query_name}] {description}",
query_name = query_info.tagged_key.query_name(),
);
}
current_query = query_info.job.parent;
count_total += 1;
}
if let Some(ref mut file) = file {
let _ = writeln!(file, "end of query stack");
}
count_total
}
#[inline(never)]
#[cold]
pub(crate) fn create_cycle_error<'tcx>(
tcx: TyCtxt<'tcx>,
Cycle { usage, frames }: &Cycle<'tcx>,
) -> Diag<'tcx> {
assert!(!frames.is_empty());
let span = frames[0].tagged_key.default_span(tcx, frames[1 % frames.len()].span);
let mut cycle_stack = Vec::new();
use crate::error::StackCount;
let stack_bottom = frames[0].tagged_key.description(tcx);
let stack_count = if frames.len() == 1 {
StackCount::Single { stack_bottom: stack_bottom.clone() }
} else {
StackCount::Multiple { stack_bottom: stack_bottom.clone() }
};
for i in 1..frames.len() {
let frame = &frames[i];
let span = frame.tagged_key.default_span(tcx, frames[(i + 1) % frames.len()].span);
cycle_stack
.push(crate::error::CycleStack { span, desc: frame.tagged_key.description(tcx) });
}
let cycle_usage = usage.as_ref().map(|usage| crate::error::CycleUsage {
span: usage.tagged_key.default_span(tcx, usage.span),
usage: usage.tagged_key.description(tcx),
});
let alias = if frames
.iter()
.all(|frame| frame.tagged_key.def_kind(tcx) == Some(DefKind::TyAlias))
{
Some(crate::error::Alias::Ty)
} else if frames.iter().all(|frame| frame.tagged_key.def_kind(tcx) == Some(DefKind::TraitAlias))
{
Some(crate::error::Alias::Trait)
} else {
None
};
let cycle_diag = crate::error::Cycle {
span,
cycle_stack,
stack_bottom,
alias,
cycle_usage,
stack_count,
note_span: (),
};
tcx.sess.dcx().create_err(cycle_diag)
}