cache2 0.2.3

Performance-first bounded RAM and NVMe hybrid cache
Documentation
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
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
// Copyright 2026 ScopeDB, Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

//! Owned-buffer record reads for the RegionStore data path.
//!
//! A packed record may start on a 32-byte boundary, while direct I/O requires
//! a 4 KiB-aligned address, offset, and length. The index stores a size-class
//! upper bound, so buffered I/O reads that bounded range and direct I/O expands
//! it once to 4 KiB boundaries. Region generation and the exact record envelope
//! are validated above without another lookup or read.

use std::io;
use std::ops::Range;

use crate::format::RECORD_ALIGNMENT;
use crate::index::IndexEntry;
use crate::io_engine::{
    BoundedIoRequest, IoBuffer, IoCompletion, IoDeadlineExceeded, IoEngine, IoOperation,
    OperationKind, ReadSlot, RequestId, submit_cache_read,
};
use crate::recovery::{DATA_REGION_AREA_OFFSET, DataGeometry};
use crate::resources::BufferLease;

pub(crate) const _READ_ALIGNMENT: usize = 4096;
pub(crate) const _MAX_READ_ALIGNMENT_OVERHEAD: usize = 2 * _READ_ALIGNMENT;

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub(crate) struct ReadCandidate {
    pub(crate) entry: IndexEntry,
    pub(crate) region_generation: u64,
}

#[derive(Clone, Debug, Eq, PartialEq)]
pub(crate) struct ReadPlan {
    pub(crate) hash: u64,
    pub(crate) entry: IndexEntry,
    pub(crate) region_generation: u64,
    pub(crate) absolute: u64,
    pub(crate) read_len: usize,
    pub(crate) record_range: Range<usize>,
}

pub(crate) struct PendingRead {
    plan: ReadPlan,
    request_id: RequestId,
    request: BoundedIoRequest,
}

pub(crate) struct ReadCompletion {
    pub(crate) plan: ReadPlan,
    pub(crate) result: io::Result<()>,
    pub(crate) buffer: Option<BufferLease>,
}

impl ReadCompletion {
    /// Returns the bounded candidate range only after every completion
    /// invariant has passed. Direct-I/O alignment bytes stay private.
    pub(crate) fn record_bytes(&self) -> Option<&[u8]> {
        if self.result.is_err() {
            return None;
        }
        self.buffer
            .as_ref()?
            .prepared(self.plan.read_len)
            .ok()?
            .get(self.plan.record_range.clone())
    }
}

impl PendingRead {
    #[cfg(test)]
    pub(crate) fn wait(self, engine: &dyn IoEngine) -> ReadCompletion {
        let Self {
            plan,
            request_id,
            request,
        } = self;
        let completion = request.wait(engine);
        Self::finish(plan, request_id, completion)
    }

    pub(crate) async fn wait_async(
        self,
        engine: std::sync::Arc<dyn IoEngine>,
        tokio_handle: &tokio::runtime::Handle,
    ) -> ReadCompletion {
        let Self {
            plan,
            request_id,
            request,
        } = self;
        let completion = request.wait_async(engine, tokio_handle).await;
        Self::finish(plan, request_id, completion)
    }

    fn finish(
        plan: ReadPlan,
        request_id: RequestId,
        completion: Result<IoCompletion, IoDeadlineExceeded>,
    ) -> ReadCompletion {
        let completion = match completion {
            Ok(completion) => completion,
            Err(timeout) => {
                let (error, buffer) = timeout.into_lease();
                return ReadCompletion {
                    plan,
                    result: Err(error),
                    buffer,
                };
            }
        };
        let identity_valid =
            completion.request_id == request_id && completion.kind == OperationKind::Read;
        let bytes_transferred = completion.bytes_transferred;
        let (io_result, buffer) = completion.into_lease();

        let protocol_error = if !identity_valid {
            Some(io::Error::new(
                io::ErrorKind::InvalidData,
                "Region record completion identity does not match its request",
            ))
        } else if buffer.is_none() {
            Some(io::Error::new(
                io::ErrorKind::InvalidData,
                "Region record completion did not return its owned buffer",
            ))
        } else if buffer
            .as_ref()
            .is_some_and(|buffer| buffer.prepared(plan.read_len).is_err())
        {
            Some(io::Error::new(
                io::ErrorKind::InvalidData,
                "Region record completion returned a short buffer",
            ))
        } else {
            None
        };

        let result = match protocol_error {
            Some(error) => Err(error),
            None => io_result.and_then(|completed| {
                if completed != plan.read_len || bytes_transferred != plan.read_len {
                    return Err(io::Error::new(
                        io::ErrorKind::UnexpectedEof,
                        "Region record read completed with the wrong byte count",
                    ));
                }
                Ok(())
            }),
        };
        ReadCompletion {
            plan,
            result,
            buffer,
        }
    }
}

/// Submits one bounded planned range using an owned alignment-rounded buffer.
///
/// Validation happens before the lease is prepared or submitted. Any rejected
/// operation drops its lease immediately.
pub(crate) fn submit_read(
    engine: &dyn IoEngine,
    slot: ReadSlot,
    plan: ReadPlan,
    buffer: BufferLease,
) -> io::Result<PendingRead> {
    let buffer = IoBuffer::for_read(buffer, plan.read_len).map_err(|error| error.error)?;
    let request = submit_cache_read(engine, slot, IoOperation::read(buffer, plan.absolute))
        .map_err(|error| error.into_lease().0)?;
    Ok(PendingRead {
        plan,
        request_id: request.id(),
        request,
    })
}

pub(crate) fn plan_read(
    geometry: DataGeometry,
    hash: u64,
    candidate: ReadCandidate,
    align_for_direct_io: bool,
) -> io::Result<ReadPlan> {
    let ReadCandidate {
        entry,
        region_generation,
    } = candidate;
    let location = entry.location;
    let offset = u64::from(location.offset());
    let record_len_upper = usize::try_from(location.record_len()).map_err(|_| {
        io::Error::new(
            io::ErrorKind::InvalidInput,
            "Region record length does not fit this platform",
        )
    })?;
    let record_len_upper_u64 = u64::try_from(record_len_upper).map_err(|_| {
        io::Error::new(
            io::ErrorKind::InvalidInput,
            "Region record length does not fit the data geometry",
        )
    })?;
    if !geometry.is_valid()
        || region_generation == 0
        || location.region_id() >= geometry.region_count
        || offset % u64::from(RECORD_ALIGNMENT) != 0
        || offset >= geometry.region_size
        || record_len_upper == 0
        || record_len_upper % RECORD_ALIGNMENT as usize != 0
    {
        return Err(io::Error::new(
            io::ErrorKind::InvalidInput,
            "invalid durable Region record location",
        ));
    }
    // A valid final record may end exactly at the Region boundary while its
    // size class extends beyond it. Never read into the next Region.
    let record_len_u64 = record_len_upper_u64.min(geometry.region_size - offset);
    let record_len = usize::try_from(record_len_u64).map_err(|_| {
        io::Error::new(
            io::ErrorKind::InvalidInput,
            "bounded Region record length does not fit this platform",
        )
    })?;

    let record_absolute = u64::from(location.region_id())
        .checked_mul(geometry.region_size)
        .and_then(|base| DATA_REGION_AREA_OFFSET.checked_add(base))
        .and_then(|base| base.checked_add(offset))
        .ok_or_else(|| io::Error::new(io::ErrorKind::InvalidInput, "read offset overflow"))?;
    let record_absolute_end = record_absolute
        .checked_add(record_len_u64)
        .filter(|end| *end <= geometry.data_file_len)
        .ok_or_else(|| {
            io::Error::new(
                io::ErrorKind::InvalidInput,
                "Region record exceeds the data file",
            )
        })?;

    let alignment = _READ_ALIGNMENT as u64;
    let (absolute, io_end) = if align_for_direct_io {
        (
            record_absolute / alignment * alignment,
            align_up(record_absolute_end, alignment).ok_or_else(|| {
                io::Error::new(io::ErrorKind::InvalidInput, "aligned read overflow")
            })?,
        )
    } else {
        (record_absolute, record_absolute_end)
    };
    let read_len = io_end
        .checked_sub(absolute)
        .and_then(|length| usize::try_from(length).ok())
        .filter(|length| *length != 0 && (!align_for_direct_io || *length % _READ_ALIGNMENT == 0))
        .ok_or_else(|| io::Error::new(io::ErrorKind::InvalidInput, "invalid Region record read"))?;
    let record_start = record_absolute
        .checked_sub(absolute)
        .and_then(|start| usize::try_from(start).ok())
        .ok_or_else(|| io::Error::new(io::ErrorKind::InvalidInput, "record slice overflow"))?;
    let record_range = record_start
        ..record_start.checked_add(record_len).ok_or_else(|| {
            io::Error::new(io::ErrorKind::InvalidInput, "record slice end overflow")
        })?;
    let overhead = read_len.checked_sub(record_len).ok_or_else(|| {
        io::Error::new(
            io::ErrorKind::InvalidInput,
            "Region read is shorter than its candidate range",
        )
    })?;
    if record_range.end > read_len
        || (align_for_direct_io
            && (!absolute.is_multiple_of(alignment) || overhead >= _MAX_READ_ALIGNMENT_OVERHEAD))
        || io_end > geometry.data_file_len
    {
        return Err(io::Error::new(
            io::ErrorKind::InvalidInput,
            "Region read exceeds its bounds",
        ));
    }

    Ok(ReadPlan {
        hash,
        entry,
        region_generation,
        absolute,
        read_len,
        record_range,
    })
}

fn align_up(value: u64, alignment: u64) -> Option<u64> {
    value
        .checked_add(alignment.checked_sub(1)?)
        .map(|end| end / alignment * alignment)
}

#[cfg(test)]
mod tests {
    use std::sync::{Arc, Mutex};

    use super::*;
    use crate::index::PackedLocation;
    use crate::io_backend::{IoBackend, SyncMode, SyncPoint, WritePoint};
    use crate::io_engine::BackendIoEngine;
    use crate::resources::{ResourceController, ResourceLimits};

    #[derive(Default)]
    struct RecordingBackend {
        reads: Mutex<Vec<(u64, usize)>>,
    }

    impl IoBackend for RecordingBackend {
        fn len(&self) -> io::Result<u64> {
            Ok(u64::MAX)
        }

        fn set_len(&self, _len: u64) -> io::Result<()> {
            Ok(())
        }

        fn read_at(&self, buffer: &mut [u8], offset: u64) -> io::Result<usize> {
            self.reads
                .lock()
                .unwrap_or_else(|poisoned| poisoned.into_inner())
                .push((offset, buffer.len()));
            for (index, byte) in buffer.iter_mut().enumerate() {
                *byte = offset.wrapping_add(index as u64) as u8;
            }
            Ok(buffer.len())
        }

        fn write_at(&self, _point: WritePoint, _buffer: &[u8], _offset: u64) -> io::Result<usize> {
            Err(io::Error::new(io::ErrorKind::Unsupported, "write unused"))
        }

        fn sync(&self, _point: SyncPoint, _mode: SyncMode) -> io::Result<()> {
            Ok(())
        }

        fn try_lock_exclusive(&self) -> io::Result<()> {
            Ok(())
        }

        fn unlock(&self) -> io::Result<()> {
            Ok(())
        }
    }

    fn geometry() -> DataGeometry {
        let region_size = 64 * 1024;
        let region_count = 2;
        DataGeometry {
            data_file_len: DataGeometry::expected_file_len(region_size, region_count).unwrap(),
            region_size,
            region_count,
        }
    }

    fn entry(location: crate::index::PackedLocation) -> IndexEntry {
        IndexEntry { location }
    }

    fn candidate(entry: IndexEntry) -> ReadCandidate {
        ReadCandidate {
            entry,
            region_generation: 1,
        }
    }

    #[test]
    fn unaligned_record_uses_one_aligned_read_and_returns_its_exact_slice() {
        let backend = Arc::new(RecordingBackend::default());
        let engine = BackendIoEngine::new(backend.clone(), 1).unwrap();
        let resources = ResourceController::try_new(ResourceLimits {
            memory_limit_bytes: _READ_ALIGNMENT,
            reserved_memory_bytes: 0,
        })
        .unwrap();
        let location = PackedLocation::new(1, 32, 64).unwrap();
        let entry = entry(location);
        let plan = plan_read(geometry(), 7, candidate(entry), true).unwrap();

        let slot = engine.try_reserve_read().unwrap();
        let completion = submit_read(
            &engine,
            slot,
            plan,
            resources.try_read_buffer(_READ_ALIGNMENT).unwrap(),
        )
        .unwrap()
        .wait(&engine);
        assert!(completion.result.is_ok());
        assert_eq!(completion.plan.record_range, 32..96);
        assert_eq!(completion.record_bytes().unwrap().len(), 64);

        let record_absolute = DATA_REGION_AREA_OFFSET + geometry().region_size + 32;
        let reads = backend
            .reads
            .lock()
            .unwrap_or_else(|poisoned| poisoned.into_inner());
        assert_eq!(reads.as_slice(), &[(record_absolute - 32, 4096)]);
        let expected = (0..64)
            .map(|index| record_absolute.wrapping_add(index) as u8)
            .collect::<Vec<_>>();
        assert_eq!(completion.record_bytes().unwrap(), expected);
        drop(reads);
        drop(completion.buffer);
        engine.shutdown().unwrap();
        assert_eq!(resources.managed_memory_snapshot().current_bytes, 0);
    }

    #[test]
    fn buffered_record_uses_one_size_class_upper_bound_read() {
        let backend = Arc::new(RecordingBackend::default());
        let engine = BackendIoEngine::new(backend.clone(), 1).unwrap();
        let resources = ResourceController::try_new(ResourceLimits {
            memory_limit_bytes: _READ_ALIGNMENT,
            reserved_memory_bytes: 0,
        })
        .unwrap();
        let entry = entry(PackedLocation::new(1, 32, 1120).unwrap());
        let plan = plan_read(geometry(), 7, candidate(entry), false).unwrap();
        let record_absolute = DATA_REGION_AREA_OFFSET + geometry().region_size + 32;

        let completion = submit_read(
            &engine,
            engine.try_reserve_read().unwrap(),
            plan,
            resources.try_read_buffer(1120).unwrap(),
        )
        .unwrap()
        .wait(&engine);
        assert!(completion.result.is_ok());
        assert_eq!(completion.plan.record_range, 0..1120);
        assert_eq!(completion.record_bytes().unwrap().len(), 1120);
        assert_eq!(
            backend
                .reads
                .lock()
                .unwrap_or_else(|poisoned| poisoned.into_inner())
                .as_slice(),
            &[(record_absolute, 1120)]
        );
        drop(completion.buffer);
        engine.shutdown().unwrap();
        assert_eq!(resources.managed_memory_snapshot().current_bytes, 0);
    }

    #[test]
    fn final_record_size_class_is_clamped_to_its_region() {
        let record_len = 1056;
        let offset = geometry().region_size as u32 - record_len;
        let entry = entry(PackedLocation::new(1, offset, 1120).unwrap());
        let plan = plan_read(geometry(), 7, candidate(entry), false).unwrap();

        assert_eq!(plan.record_range, 0..record_len as usize);
        assert_eq!(plan.read_len, record_len as usize);
        assert_eq!(
            plan.absolute,
            DATA_REGION_AREA_OFFSET + geometry().region_size + u64::from(offset)
        );
    }

    #[test]
    fn invalid_entry_is_rejected_before_allocating_or_issuing_io() {
        let backend = Arc::new(RecordingBackend::default());
        let engine = BackendIoEngine::new(backend.clone(), 1).unwrap();
        let invalid = entry(PackedLocation::new(geometry().region_count, 0, 32).unwrap());

        let error = plan_read(geometry(), 7, candidate(invalid), true).unwrap_err();
        assert_eq!(error.kind(), io::ErrorKind::InvalidInput);
        assert!(
            backend
                .reads
                .lock()
                .unwrap_or_else(|poisoned| poisoned.into_inner())
                .is_empty()
        );

        let generation_zero = ReadCandidate {
            entry: entry(PackedLocation::new(0, 0, 32).unwrap()),
            region_generation: 0,
        };
        assert_eq!(
            plan_read(geometry(), 7, generation_zero, true)
                .unwrap_err()
                .kind(),
            io::ErrorKind::InvalidInput
        );
        engine.shutdown().unwrap();
    }
}