1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
//! Customization of where and how the generation lock works. //! //! By default, all the [`ArcSwapAny`](../struct.ArcSwapAny.html) instances share the same //! generation lock. This is to save space in them (they have the same size as a single pointer), //! because the default lock is quite a large data structure (it's sharded, to prevent too much //! contention between different threads). This has the disadvantage that a lock on one instance //! influences another instance. //! //! The things in this module allow customizing how the lock behaves. The default one is //! [`Global`](struct.Global.html). If you want to use independent but unsharded lock, use the //! [`PrivateUnsharded`](struct.PrivateUnsharded.html) (or the //! [`IndependentArcSwap`](../type.IndependentArcSwap.html) type alias). //! //! Or you can implement your own lock, but you probably should study the internals of the library //! first. //! //! # Not Implemented Yet //! //! These variants would probably make sense, but haven't been written yet: //! //! * A lock storage that is shared, but only between a certain group of pointers. It could be //! either as a reference (but then each `ArcSwap` would get a bit bigger), or a macro that could //! generate an independent but global storage. use std::cell::Cell; use std::sync::atomic::{AtomicUsize, Ordering}; /// Number of shards (see [`Shard`]). const SHARD_CNT: usize = 9; /// How many generations we have in the lock. pub(crate) const GEN_CNT: usize = 2; /// A single shard. /// /// This is one copy of place where the library keeps tracks of generation locks. It consists of a /// pair of counters and allows double-buffering readers (therefore, even if there's a never-ending /// stream of readers coming in, writer will get through eventually). /// /// To avoid contention and sharing of the counters between readers, we don't have one pair of /// generation counters, but several. The reader picks one shard and uses that, while the writer /// looks through all of them. This is still not perfect (two threads may choose the same ID), but /// it helps. /// /// Each [`LockStorage`](trait.LockStorage.html) must provide a (non-empty) array of these. #[repr(align(64))] #[derive(Default)] pub struct Shard(pub(crate) [AtomicUsize; GEN_CNT]); impl Shard { /// Takes a snapshot of current values (with Acquire ordering) pub(crate) fn snapshot(&self) -> [usize; GEN_CNT] { [ self.0[0].load(Ordering::Acquire), self.0[1].load(Ordering::Acquire), ] } } /// Abstraction of the place where generation locks are stored. /// /// The trait is unsafe because if the trait messes up with the values stored in there in any way /// (or makes the values available to something else that messes them up), this can cause UB and /// daemons and discomfort to users and such. The library expects it is the only one storing values /// there. In other words, it is expected the trait is only a dumb storage and doesn't actively do /// anything. pub unsafe trait LockStorage: Default { /// The type for keeping several shards. /// /// In general, it is expected to be a fixed-size array, but different implementations can have /// different sizes. type Shards: AsRef<[Shard]>; /// Access to the generation index. /// /// Must return the same instance of the `AtomicUsize` for the lifetime of the storage, must /// start at `0` and the trait itself must not modify it. Must be async-signal-safe. fn gen_idx(&self) -> &AtomicUsize; /// Access to the shards storage. /// /// Must return the same instance of the shards for the lifetime of the storage. Must start /// zeroed-out and the trait itself must not modify it. fn shards(&self) -> &Self::Shards; /// Pick one shard of the all selected. /// /// Returns the index of one of the shards. The choice can be arbitrary, but it should be fast /// and avoid collisions. fn choose_shard(&self) -> usize; } static GEN_IDX: AtomicUsize = AtomicUsize::new(0); macro_rules! sh { () => { Shard([AtomicUsize::new(0), AtomicUsize::new(0)]) }; } type Shards = [Shard; SHARD_CNT]; /// The global shards. static SHARDS: [Shard; SHARD_CNT] = [ sh!(), sh!(), sh!(), sh!(), sh!(), sh!(), sh!(), sh!(), sh!(), ]; /// Global counter of threads. /// /// We specifically don't use ThreadId here, because it is opaque and doesn't give us a number :-(. static THREAD_ID_GEN: AtomicUsize = AtomicUsize::new(0); thread_local! { /// A shard a thread has chosen. /// /// The default value is just a marker it hasn't been set. static THREAD_SHARD: Cell<usize> = Cell::new(SHARD_CNT); } /// The default, global lock. /// /// The lock is stored out-of-band, globally. This means that one `ArcSwap` with this lock storage /// is only one machine word large, but a lock on one instance blocks the other, independent ones. /// /// It has several shards so threads are less likely to collide (HW-contend) on them. #[derive(Default)] pub struct Global; unsafe impl LockStorage for Global { type Shards = Shards; #[inline] fn gen_idx(&self) -> &AtomicUsize { &GEN_IDX } #[inline] fn shards(&self) -> &Shards { &SHARDS } #[inline] fn choose_shard(&self) -> usize { THREAD_SHARD .try_with(|ts| { let mut val = ts.get(); if val >= SHARD_CNT { val = THREAD_ID_GEN.fetch_add(1, Ordering::Relaxed) % SHARD_CNT; ts.set(val); } val }) .unwrap_or(0) } } /// A single „shard“ that is stored inline, inside the corresponding `ArcSwap`. Therefore, locks on /// each instance won't influence any other instances. On the other hand, the `ArcSwap` itself gets /// bigger and doesn't have multiple shards, so concurrent uses might contend each other a bit. /// /// ```rust /// # use std::sync::Arc; /// # use arc_swap::{ArcSwap, ArcSwapAny}; /// # use arc_swap::gen_lock::PrivateUnsharded; /// // This one shares locks with others. /// let shared = ArcSwap::from_pointee(42); /// // But this one has an independent lock. /// let independent = ArcSwapAny::<Arc<usize>, PrivateUnsharded>::from_pointee(42); /// /// // This'll hold a lock so any writers there wouldn't complete /// let l = independent.load_signal_safe(); /// // But the lock doesn't influence the shared one, so this goes through just fine /// shared.store(Arc::new(43)); /// /// assert_eq!(42, **l); /// ``` /// /// Note that there`s a type alias [`IndependentArcSwap`](../type.IndependentArcSwap.html) that can /// be used instead. #[derive(Default)] pub struct PrivateUnsharded { gen_idx: AtomicUsize, shard: [Shard; 1], } unsafe impl LockStorage for PrivateUnsharded { type Shards = [Shard; 1]; #[inline] fn gen_idx(&self) -> &AtomicUsize { &self.gen_idx } #[inline] fn shards(&self) -> &[Shard; 1] { &self.shard } #[inline] fn choose_shard(&self) -> usize { 0 } } /// An alternative to [`PrivateUnsharded`], but with configurable number of shards. /// /// The [`PrivateUnsharded`] is almost identical to `PrivateSharded<[Shard; 1]>` (the /// implementation takes advantage of some details to avoid a little bit of overhead). It allows /// the user to choose the trade-of between contention during locking and size of the pointer and /// speed during writes. /// /// [`PrivateUnsharded`]: struct.PrivateUnsharded.html /// /// # Note on `AsRef<[Shard]>` /// /// Rust provides the `AsRef` trait (or, actually any trait) up to arrays of 32 elements. If you /// need something bigger, you have to work around it with a newtype. #[derive(Default)] pub struct PrivateSharded<S> { gen_idx: AtomicUsize, shards: S, } /// Global counter of threads. /// /// We specifically don't use ThreadId here, because it is opaque and doesn't give us a number :-(. static PRIV_THREAD_ID_GEN: AtomicUsize = AtomicUsize::new(0); thread_local! { /// A shard a thread has chosen. static PRIV_THREAD_ID: usize = PRIV_THREAD_ID_GEN.fetch_add(1, Ordering::Relaxed); } unsafe impl<S: AsRef<[Shard]> + Default> LockStorage for PrivateSharded<S> { type Shards = S; #[inline] fn gen_idx(&self) -> &AtomicUsize { &self.gen_idx } #[inline] fn shards(&self) -> &Self::Shards { &self.shards } #[inline] fn choose_shard(&self) -> usize { PRIV_THREAD_ID .try_with(|id| id % self.shards.as_ref().len()) .unwrap_or(0) } } #[cfg(test)] mod tests { extern crate crossbeam_utils; use std::sync::Arc; use self::crossbeam_utils::thread; use super::super::{ArcSwapAny, SignalSafety}; use super::*; const ITERATIONS: usize = 100; // Does a kind of ping-pong between two threads, torturing the arc-swap somewhat. fn basic_check<S: LockStorage + Send + Sync>() { for _ in 0..ITERATIONS { let shared = ArcSwapAny::<_, S>::from(Arc::new(usize::max_value())); thread::scope(|scope| { for i in 0..2 { let shared = &shared; scope.spawn(move |_| { for j in 0..50 { if j % 2 == i { while **shared.lock_internal(SignalSafety::Unsafe) != j {} } else { shared.store(Arc::new(j)); } } }); } }) .unwrap(); } } #[test] fn basic_check_global() { basic_check::<Global>(); } #[test] fn basic_check_private_unsharded() { basic_check::<PrivateUnsharded>(); } #[test] fn basic_check_private_sharded_2() { basic_check::<PrivateSharded<[Shard; 2]>>(); } #[test] fn basic_check_private_sharded_63() { basic_check::<PrivateSharded<[Shard; 31]>>(); } }