[core] Add lock::observing module, for analyzing lock acquisition.

Add a new module `lock::observing`, enabled by the `observe-locks`
feature, that records all nested lock acquisitions in trace files.

Add a new utility to the workspace, `lock-analyzer`, that reads the
files written by the `observe-locks` feature and writes out a new
`define_lock_ranks!` macro invocation that covers all observed lock
usage, along with comments giving the held and acquired source
locations.

Author: jimblandy

Cargo.lock

@@ -7,6 +7,7 @@ members = [
     # default members
     "d3d12",
     "examples/",
+    "lock-analyzer",
     "naga-cli",
     "naga",
     "naga/fuzz",
@@ -24,6 +25,7 @@ exclude = []
 default-members = [
     "d3d12",
     "examples/",
+    "lock-analyzer",
     "naga-cli",
     "naga",
     "naga/fuzz",

Cargo.toml

@@ -0,0 +1,20 @@
+[package]
+name = "lock-analyzer"
+edition.workspace = true
+rust-version.workspace = true
+keywords.workspace = true
+license.workspace = true
+homepage.workspace = true
+repository.workspace = true
+version.workspace = true
+authors.workspace = true
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[dependencies]
+ron.workspace = true
+anyhow.workspace = true
+
+[dependencies.serde]
+workspace = true
+features = ["serde_derive"]

lock-analyzer/Cargo.toml

@@ -0,0 +1,215 @@
+//! Analyzer for data produced by `wgpu-core`'s `observe_locks` feature.
+//!
+//! When `wgpu-core`'s `observe_locks` feature is enabled, if the
+//! `WGPU_CORE_LOCK_OBSERVE_DIR` environment variable is set to the
+//! path of an existing directory, then every thread that acquires a
+//! lock in `wgpu-core` will write its own log file to that directory.
+//! You can then run this program to read those files and summarize
+//! the results.
+//!
+//! This program also consults the `WGPU_CORE_LOCK_OBSERVE_DIR`
+//! environment variable to find the log files written by `wgpu-core`.
+//!
+//! See `wgpu_core/src/lock/observing.rs` for a general explanation of
+//! this analysis.
+
+use std::sync::Arc;
+use std::{
+    collections::{btree_map::Entry, BTreeMap, BTreeSet, HashMap},
+    fmt,
+    path::PathBuf,
+};
+
+use anyhow::{Context, Result};
+
+fn main() -> Result<()> {
+    let mut ranks: BTreeMap<u32, Rank> = BTreeMap::default();
+
+    let Ok(dir) = std::env::var("WGPU_CORE_LOCK_OBSERVE_DIR") else {
+        eprintln!(
+            "Please set the `WGPU_CORE_LOCK_OBSERVE_DIR` environment variable\n\
+                   to the path of the directory containing the files written by\n\
+                   `wgpu-core`'s `observe_locks` feature."
+        );
+        anyhow::bail!("`WGPU_CORE_LOCK_OBSERVE_DIR` environment variable is not set");
+    };
+    for entry in std::fs::read_dir(dir)? {
+        let entry = entry?;
+        let name = PathBuf::from(&entry.file_name());
+        let Some(extension) = name.extension() else {
+            eprintln!("Ignoring {}", name.display());
+            continue;
+        };
+        if extension != "ron" {
+            eprintln!("Ignoring {}", name.display());
+            continue;
+        }
+
+        let contents = std::fs::read(entry.path())
+            .with_context(|| format!("reading lock observations from {}", name.display()))?;
+        // The addresses of `&'static Location<'static>` values could
+        // vary from run to run.
+        let mut locations: HashMap<u64, Arc<Location>> = HashMap::default();
+        for line in contents.split(|&b| b == b'\n') {
+            if line.is_empty() {
+                continue;
+            }
+            let action = ron::de::from_bytes::<Action>(line)
+                .with_context(|| format!("Error parsing action from {}", name.display()))?;
+            match action {
+                Action::Location {
+                    address,
+                    file,
+                    line,
+                    column,
+                } => {
+                    let file = match file.split_once("src/") {
+                        Some((_, after)) => after.to_string(),
+                        None => file,
+                    };
+                    assert!(locations
+                        .insert(address, Arc::new(Location { file, line, column }))
+                        .is_none());
+                }
+                Action::Rank {
+                    bit,
+                    member_name,
+                    const_name,
+                } => match ranks.entry(bit) {
+                    Entry::Occupied(occupied) => {
+                        let rank = occupied.get();
+                        assert_eq!(rank.member_name, member_name);
+                        assert_eq!(rank.const_name, const_name);
+                    }
+                    Entry::Vacant(vacant) => {
+                        vacant.insert(Rank {
+                            member_name,
+                            const_name,
+                            acquisitions: BTreeMap::default(),
+                        });
+                    }
+                },
+                Action::Acquisition {
+                    older_rank,
+                    older_location,
+                    newer_rank,
+                    newer_location,
+                } => {
+                    let older_location = locations[&older_location].clone();
+                    let newer_location = locations[&newer_location].clone();
+                    ranks
+                        .get_mut(&older_rank)
+                        .unwrap()
+                        .acquisitions
+                        .entry(newer_rank)
+                        .or_default()
+                        .entry(older_location)
+                        .or_default()
+                        .insert(newer_location);
+                }
+            }
+        }
+    }
+
+    for older_rank in ranks.values() {
+        if older_rank.acquisitions.is_empty() {
+            // We'll print leaf locks separately, below.
+            continue;
+        }
+        println!(
+            "    rank {} {:?} followed by {{",
+            older_rank.const_name, older_rank.member_name
+        );
+        let mut first_newer = true;
+        for (newer_rank, locations) in &older_rank.acquisitions {
+            if !first_newer {
+                println!();
+            }
+            for (older_location, newer_locations) in locations {
+                if newer_locations.len() == 1 {
+                    for newer_loc in newer_locations {
+                        println!("        // holding {older_location} while locking {newer_loc}");
+                    }
+                } else {
+                    println!("        // holding {older_location} while locking:");
+                    for newer_loc in newer_locations {
+                        println!("        //     {newer_loc}");
+                    }
+                }
+            }
+            println!("        {},", ranks[newer_rank].const_name);
+            first_newer = false;
+        }
+        println!("    }};");
+        println!();
+    }
+
+    for older_rank in ranks.values() {
+        if !older_rank.acquisitions.is_empty() {
+            continue;
+        }
+
+        println!(
+            "    rank {} {:?} followed by {{ }};",
+            older_rank.const_name, older_rank.member_name
+        );
+    }
+
+    Ok(())
+}
+
+#[derive(Debug, serde::Deserialize)]
+enum Action {
+    /// A location that we will refer to in later actions.
+    Location {
+        address: u64,
+        file: String,
+        line: u32,
+        column: u32,
+    },
+
+    /// A lock rank that we will refer to in later actions.
+    Rank {
+        bit: u32,
+        member_name: String,
+        const_name: String,
+    },
+
+    /// An attempt to acquire a lock while holding another lock.
+    Acquisition {
+        /// The number of the already acquired lock's rank.
+        older_rank: u32,
+
+        /// The source position at which we acquired it. Specifically,
+        /// its `Location`'s address, as an integer.
+        older_location: u64,
+
+        /// The number of the rank of the lock we are acquiring.
+        newer_rank: u32,
+
+        /// The source position at which we are acquiring it.
+        /// Specifically, its `Location`'s address, as an integer.
+        newer_location: u64,
+    },
+}
+
+struct Rank {
+    member_name: String,
+    const_name: String,
+    acquisitions: BTreeMap<u32, LocationSet>,
+}
+
+type LocationSet = BTreeMap<Arc<Location>, BTreeSet<Arc<Location>>>;
+
+#[derive(Eq, Ord, PartialEq, PartialOrd)]
+struct Location {
+    file: String,
+    line: u32,
+    column: u32,
+}
+
+impl fmt::Display for Location {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(f, "{}:{}", self.file, self.line)
+    }
+}

lock-analyzer/src/main.rs

@@ -53,6 +53,10 @@ serde = ["dep:serde", "wgt/serde", "arrayvec/serde"]
 ## Enable API tracing.
 trace = ["ron", "serde", "naga/serialize"]
 
+## Enable lock order observation.
+## See `src/lock/observing.rs`.
+observe_locks = ["ron", "serde/serde_derive"]
+
 ## Enable API replaying
 replay = ["serde", "naga/deserialize"]
 

wgpu-core/Cargo.toml

@@ -31,11 +31,17 @@ pub mod rank;
 #[cfg_attr(not(wgpu_validate_locks), allow(dead_code))]
 mod ranked;
 
-#[cfg_attr(wgpu_validate_locks, allow(dead_code))]
+#[cfg(feature = "observe_locks")]
+mod observing;
+
+#[cfg_attr(any(wgpu_validate_locks, feature = "observe_locks"), allow(dead_code))]
 mod vanilla;
 
 #[cfg(wgpu_validate_locks)]
 pub use ranked::{Mutex, MutexGuard, RwLock, RwLockReadGuard, RwLockWriteGuard};
 
-#[cfg(not(wgpu_validate_locks))]
+#[cfg(feature = "observe_locks")]
+pub use observing::{Mutex, MutexGuard, RwLock, RwLockReadGuard, RwLockWriteGuard};
+
+#[cfg(not(any(wgpu_validate_locks, feature = "observe_locks")))]
 pub use vanilla::{Mutex, MutexGuard, RwLock, RwLockReadGuard, RwLockWriteGuard};