Auto merge of #102655 - joboet:windows_tls_opt, r=ChrisDenton

Optimize TLS on Windows

This implements the suggestion in the current TLS code to embed the linked list of destructors in the `StaticKey` structure to save allocations. Additionally, locking is avoided when no destructor needs to be run. By using one Windows-provided `Once` per key instead of a global lock, locking is more finely-grained (this unblocks #100579).

2 files changed, 9 insertions, 24 deletions

diff --git a/library/std/src/sys_common/mod.rs b/library/std/src/sys_common/mod.rs
index e4dd0253668..8c19f9332dc 100644
--- a/library/std/src/sys_common/mod.rs
+++ b/library/std/src/sys_common/mod.rs
@@ -34,11 +34,18 @@ pub mod rwlock;
 pub mod thread;
 pub mod thread_info;
 pub mod thread_local_dtor;
-pub mod thread_local_key;
 pub mod thread_parker;
 pub mod wtf8;
 
 cfg_if::cfg_if! {
+    if #[cfg(target_os = "windows")] {
+        pub use crate::sys::thread_local_key;
+    } else {
+        pub mod thread_local_key;
+    }
+}
+
+cfg_if::cfg_if! {
     if #[cfg(any(target_os = "l4re",
                  target_os = "hermit",
                  feature = "restricted-std",
diff --git a/library/std/src/sys_common/thread_local_key.rs b/library/std/src/sys_common/thread_local_key.rs
index 032bf604d73..747579f1781 100644
--- a/library/std/src/sys_common/thread_local_key.rs
+++ b/library/std/src/sys_common/thread_local_key.rs
@@ -53,7 +53,6 @@ mod tests;
 
 use crate::sync::atomic::{self, AtomicUsize, Ordering};
 use crate::sys::thread_local_key as imp;
-use crate::sys_common::mutex::StaticMutex;
 
 /// A type for TLS keys that are statically allocated.
 ///
@@ -151,25 +150,6 @@ impl StaticKey {
     }
 
     unsafe fn lazy_init(&self) -> usize {
-        // Currently the Windows implementation of TLS is pretty hairy, and
-        // it greatly simplifies creation if we just synchronize everything.
-        //
-        // Additionally a 0-index of a tls key hasn't been seen on windows, so
-        // we just simplify the whole branch.
-        if imp::requires_synchronized_create() {
-            // We never call `INIT_LOCK.init()`, so it is UB to attempt to
-            // acquire this mutex reentrantly!
-            static INIT_LOCK: StaticMutex = StaticMutex::new();
-            let _guard = INIT_LOCK.lock();
-            let mut key = self.key.load(Ordering::SeqCst);
-            if key == 0 {
-                key = imp::create(self.dtor) as usize;
-                self.key.store(key, Ordering::SeqCst);
-            }
-            rtassert!(key != 0);
-            return key;
-        }
-
         // POSIX allows the key created here to be 0, but the compare_exchange
         // below relies on using 0 as a sentinel value to check who won the
         // race to set the shared TLS key. As far as I know, there is no
@@ -232,8 +212,6 @@ impl Key {
 
 impl Drop for Key {
     fn drop(&mut self) {
-        // Right now Windows doesn't support TLS key destruction, but this also
-        // isn't used anywhere other than tests, so just leak the TLS key.
-        // unsafe { imp::destroy(self.key) }
+        unsafe { imp::destroy(self.key) }
     }
 }