fd/eventfd: wake all registered pollers on write, not just one

[stlankes]

A write made the eventfd readable but only woke the oldest waiter (pop_front). With the poll(2)-based fd multiplexing re-registering a fresh waker on every poll, the read queue accumulates stale wakers, so the single wake could hit a dead waker and miss the real one — a cross-thread wakeup (e.g. tokio's I/O driver) could then be lost. Drain the queue and wake every registered poller, matching level-triggered readiness.

[github-actions]

Benchmark Results

Details

Benchmark	Current: 926a17a	Previous: f83a235	Performance Ratio
startup_benchmark Build Time	76.57 s	77.08 s	0.99 ❗
startup_benchmark File Size	0.73 MB	0.73 MB	1.00 ❗
Startup Time - 1 core	0.74 s (±0.02 s)	0.73 s (±0.02 s)	1.02
Startup Time - 2 cores	0.72 s (±0.02 s)	0.75 s (±0.02 s)	0.96
Startup Time - 4 cores	0.74 s (±0.02 s)	0.76 s (±0.02 s)	0.98
multithreaded_benchmark Build Time	76.92 s	80.66 s	0.95 ❗
multithreaded_benchmark File Size	0.86 MB	0.86 MB	1.00 ❗
Multithreaded Pi Efficiency - 2 Threads	90.43 % (±4.97 %)	89.44 % (±5.91 %)	1.01
Multithreaded Pi Efficiency - 4 Threads	43.63 % (±2.40 %)	44.06 % (±2.65 %)	0.99
Multithreaded Pi Efficiency - 8 Threads	25.72 % (±1.21 %)	25.56 % (±1.89 %)	1.01
micro_benchmarks Build Time	73.30 s	87.20 s	0.84 ❗
micro_benchmarks File Size	0.87 MB	0.87 MB	1.00 ❗
Scheduling time - 1 thread	75.19 ticks (±1.12 ticks)	67.04 ticks (±1.91 ticks)	1.12 ❗
Scheduling time - 2 threads	40.31 ticks (±3.78 ticks)	39.33 ticks (±3.42 ticks)	1.02
Micro - Time for syscall (getpid)	3.58 ticks (±0.35 ticks)	3.43 ticks (±0.24 ticks)	1.04
Memcpy speed - (built_in) block size 4096	83332.79 MByte/s (±57641.25 MByte/s)	84527.91 MByte/s (±58322.62 MByte/s)	0.99
Memcpy speed - (built_in) block size 1048576	30601.84 MByte/s (±24632.77 MByte/s)	30845.87 MByte/s (±25037.16 MByte/s)	0.99
Memcpy speed - (built_in) block size 16777216	29689.78 MByte/s (±24394.89 MByte/s)	27492.52 MByte/s (±22708.20 MByte/s)	1.08
Memset speed - (built_in) block size 4096	83461.91 MByte/s (±57733.48 MByte/s)	84666.80 MByte/s (±58421.97 MByte/s)	0.99
Memset speed - (built_in) block size 1048576	31334.89 MByte/s (±25053.50 MByte/s)	31593.14 MByte/s (±25470.12 MByte/s)	0.99
Memset speed - (built_in) block size 16777216	30456.42 MByte/s (±24833.78 MByte/s)	28269.66 MByte/s (±23193.64 MByte/s)	1.08
Memcpy speed - (rust) block size 4096	74417.37 MByte/s (±52137.63 MByte/s)	74722.10 MByte/s (±52209.99 MByte/s)	1.00
Memcpy speed - (rust) block size 1048576	30530.38 MByte/s (±24581.65 MByte/s)	30824.11 MByte/s (±25054.28 MByte/s)	0.99
Memcpy speed - (rust) block size 16777216	29665.68 MByte/s (±24368.16 MByte/s)	27597.67 MByte/s (±22766.03 MByte/s)	1.07
Memset speed - (rust) block size 4096	74973.55 MByte/s (±52546.16 MByte/s)	74865.22 MByte/s (±52302.23 MByte/s)	1.00
Memset speed - (rust) block size 1048576	31271.76 MByte/s (±25015.03 MByte/s)	31565.59 MByte/s (±25484.76 MByte/s)	0.99
Memset speed - (rust) block size 16777216	30433.51 MByte/s (±24807.93 MByte/s)	28374.92 MByte/s (±23250.83 MByte/s)	1.07
alloc_benchmarks Build Time	73.42 s	81.01 s	0.91 ❗
alloc_benchmarks File Size	0.81 MB	0.81 MB	1.00 ❗
Allocations - Allocation success	91.32 %	91.32 %	1
Allocations - Deallocation success	100.00 %	100.00 %	1
Allocations - Pre-fail Allocations	61.46 %	61.46 %	1
Allocations - Average Allocation time	5842.40 Ticks (±154.64 Ticks)	6133.98 Ticks (±129.62 Ticks)	0.95 ❗
Allocations - Average Allocation time (no fail)	6441.15 Ticks (±171.37 Ticks)	6784.80 Ticks (±184.79 Ticks)	0.95 ❗
Allocations - Average Deallocation time	1187.21 Ticks (±338.54 Ticks)	1292.85 Ticks (±344.36 Ticks)	0.92
mutex_benchmark Build Time	74.01 s	81.90 s	0.90 ❗
mutex_benchmark File Size	0.87 MB	0.87 MB	1.00 ❗
Mutex Stress Test Average Time per Iteration - 1 Threads	12.14 ns (±0.49 ns)	12.08 ns (±0.27 ns)	1.00
Mutex Stress Test Average Time per Iteration - 2 Threads	39.06 ns (±2.19 ns)	17.94 ns (±3.62 ns)	2.18 ❗

This comment was automatically generated by workflow using github-action-benchmark.