-
Notifications
You must be signed in to change notification settings - Fork 1
/
base.cpp
561 lines (500 loc) · 17.3 KB
/
base.cpp
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
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
/**
* @file base.cpp
*
* @copyright 2021-2022 extratype
*/
#include "base.hpp"
#include <filesystem>
using std::bad_alloc;
namespace chunkdisk
{
static constexpr auto MAX_CHUNK_PARTS = usize(16384);
ChunkRange ChunkDiskBase::BlockChunkRange(u64 block_addr, u64 count) const
{
auto start_idx = block_addr / chunk_length;
auto start_off = block_addr % chunk_length;
auto end_idx = start_idx;
// start_idx: [start_off, chunk_length)
if (count <= chunk_length - start_off)
{
return ChunkRange{ start_idx, start_off, end_idx, start_off + count };
}
// align to the next chunk
count -= chunk_length - start_off;
end_idx += 1 + (count / chunk_length);
auto end_off = count % chunk_length;
if (end_off == 0)
{
end_idx -= 1;
end_off = chunk_length;
}
return ChunkRange{ start_idx, start_off, end_idx, end_off };
}
PageRange ChunkDiskBase::BlockPageRange(u64 chunk_idx, u64 start_off, u64 end_off) const
{
auto base_idx = chunk_idx * (chunk_length / page_length);
auto count = end_off - start_off;
auto sidx = start_off / page_length;
auto soff = u32(start_off % page_length);
auto eidx = sidx;
// sidx: [soff, page_length)
if (count <= page_length - soff)
{
return PageRange{ base_idx, sidx, soff, eidx, u32(soff + count) };
}
// align to the next page
count -= page_length - soff;
eidx += 1 + (count / page_length);
auto eoff = u32(count % page_length);
if (eoff == 0)
{
eidx -= 1;
eoff = page_length;
}
return PageRange{ base_idx, sidx, soff, eidx, eoff };
}
template <class F>
DWORD ChunkDiskBase::IterPart(const usize part_idx, F&& func)
{
try
{
for (auto& p : std::filesystem::directory_iterator(part_dirname[part_idx] + L'\\'))
{
auto fname = p.path().filename().wstring();
if (_wcsnicmp(fname.data(), L"chunk", 5) != 0) continue;
auto* endp = PWSTR();
auto idx = wcstoull(fname.data() + 5, &endp, 10);
if (fname.data() + 5 == endp || *endp != L'\0'
|| errno == ERANGE || idx >= chunk_count)
{
continue;
}
auto err = DWORD(func(idx));
if (err != ERROR_SUCCESS) return err;
}
}
catch (const bad_alloc&)
{
return ERROR_NOT_ENOUGH_MEMORY;
}
catch (const std::system_error& e)
{
return e.code().value();
}
return ERROR_SUCCESS;
}
DWORD ChunkDiskBase::Start()
{
try
{
// make class movable
mutex_parts_ = std::make_unique<std::shared_mutex>();
if (read_only && move_enabled)
{
SpdLogErr(L"error: cannot specify -W 1 option with -M 1");
return ERROR_INVALID_PARAMETER;
}
// put a lock file to prevent mistakes
const auto num_parts = part_dirname.size();
auto part_lock = std::vector<FileHandle>(num_parts);
// base of a differential disk if read_only
// ERROR_SHARING_VIOLATION occurs when write access requested
const auto desired_access = GENERIC_READ | (read_only ? 0 : GENERIC_WRITE);
// base may be shared with others
const auto share_mode = read_only ? FILE_SHARE_READ : 0;
// .lock should be removed manually after merging
const auto cr_disp = read_only ? OPEN_ALWAYS : CREATE_NEW;
// no double mount for write: temporary .lock
// disallow writing on base until merged: persistent .lock
const auto flags_attrs = FILE_ATTRIBUTE_NORMAL | (read_only ? 0 : FILE_FLAG_DELETE_ON_CLOSE);
for (auto i = usize(0); i < num_parts; ++i)
{
auto path = part_dirname[i] + L"\\.lock";
auto h = FileHandle(CreateFileW(
path.data(), desired_access, share_mode, nullptr,
cr_disp, flags_attrs, nullptr));
if (!h)
{
auto err = GetLastError();
SpdLogErr(L"error: failed to create %s with code %lu", path.data(), err);
return err;
}
part_lock[i] = std::move(h);
}
part_lock_ = std::move(part_lock);
// read parts to check chunks
auto part_current = std::vector<u64>(num_parts, 0);
auto chunk_parts = std::unordered_map<u64, usize>();
for (auto i = usize(0); i < num_parts; ++i)
{
auto err = IterPart(i, [this, i, &part_current, &chunk_parts](u64 idx) -> DWORD
{
auto [it, emplaced] = chunk_parts.emplace(idx, i);
if (!emplaced)
{
SpdLogErr(L"error: chunk%llu is duplicate in part #%llu and #$llu",
idx, it->first + 1, i + 1);
return ERROR_DUPLICATE_TAG;
}
if (++part_current[i] > part_max[i])
{
SpdLogErr(L"error: too many chunks in part #%llu", i + 1);
return ERROR_INVALID_PARAMETER;
}
return ERROR_SUCCESS;
});
if (err != ERROR_SUCCESS) return err;
}
part_current_ = std::move(part_current);
// initialize for AssignChunkPart()
// part_current_new_ is initially zero
for (auto new_part = part_current_new_; new_part < num_parts; ++new_part)
{
if (part_current_[new_part] < part_max[new_part])
{
part_current_new_ = new_part;
break;
}
}
}
catch (const bad_alloc&)
{
return ERROR_NOT_ENOUGH_MEMORY;
}
return ERROR_SUCCESS;
}
DWORD ChunkDiskBase::ChunkPath(const u64 chunk_idx, const usize part_idx, std::wstring& path) const
{
try
{
path = part_dirname[part_idx] + L"\\chunk" + std::to_wstring(chunk_idx);
return ERROR_SUCCESS;
}
catch (const bad_alloc&)
{
return ERROR_NOT_ENOUGH_MEMORY;
}
}
DWORD ChunkDiskBase::FindChunkPart(const u64 chunk_idx, usize& part_idx, SRWLock& lk)
{
if (!lk) lk = SRWLock(*mutex_parts_, false);
auto it = chunk_parts_.find(chunk_idx);
if (it != chunk_parts_.end())
{
part_idx = (*it).second;
return ERROR_SUCCESS;
}
// no reinsert_back() for performance; insertion order
if (!lk.is_exclusive())
{
lk.switch_lock();
it = chunk_parts_.find(chunk_idx);
if (it != chunk_parts_.end())
{
part_idx = (*it).second;
return ERROR_SUCCESS;
}
}
// lk is exclusive
auto err = [this, chunk_idx, &part_idx]() -> DWORD
{
const auto num_parts = part_dirname.size();
auto i = usize(0);
auto err = DWORD(ERROR_SUCCESS);
auto idx = usize(num_parts);
for (; i < num_parts; ++i)
{
auto path = std::wstring();
err = ChunkPath(chunk_idx, i, path);
if (err != ERROR_SUCCESS) break;
auto attrs = GetFileAttributesW(path.data());
if (attrs != INVALID_FILE_ATTRIBUTES)
{
idx = i;
++i;
break;
}
err = GetLastError();
// ERROR_PATH_NOT_FOUND if the parent directory does not exist
if (err != ERROR_FILE_NOT_FOUND) break;
err = ERROR_SUCCESS;
}
if (err != ERROR_SUCCESS) return err; // blame err
for (; i < num_parts; ++i)
{
auto path = std::wstring();
err = ChunkPath(chunk_idx, i, path);
if (err != ERROR_SUCCESS) break;
auto attrs = GetFileAttributesW(path.data());
if (attrs != INVALID_FILE_ATTRIBUTES)
{
// duplicate chunk
err = ERROR_DUPLICATE_TAG;
break;
}
err = GetLastError();
// ERROR_PATH_NOT_FOUND if the parent directory does not exist
if (err != ERROR_FILE_NOT_FOUND) break;
err = ERROR_SUCCESS;
}
if (err != ERROR_SUCCESS) return err; // blame err
part_idx = idx;
return ERROR_SUCCESS;
}();
if (err != ERROR_SUCCESS)
{
lk.unlock();
return err;
}
// cache result
try
{
if (chunk_parts_.size() >= MAX_CHUNK_PARTS) chunk_parts_.pop_front();
chunk_parts_.emplace(chunk_idx, part_idx);
return ERROR_SUCCESS;
}
catch (const bad_alloc&)
{
// failed to cache, successful anyway
return ERROR_SUCCESS;
}
}
bool ChunkDiskBase::CheckChunk(const u64 chunk_idx)
{
auto lk = SRWLock();
auto num_parts = part_dirname.size();
auto part_idx = num_parts;
FindChunkPart(chunk_idx, part_idx, lk);
return part_idx != num_parts;
}
DWORD ChunkDiskBase::AssignChunkPart()
{
const auto num_parts = part_dirname.size();
auto new_part = part_current_new_;
for (; new_part < num_parts; ++new_part)
{
if (part_current_[new_part] >= part_max[new_part] && move_enabled)
{
// refresh to get the actual value
auto new_count = 0;
auto err = IterPart(new_part, [&new_count](u64 idx) -> DWORD
{
++new_count;
return ERROR_SUCCESS;
});
if (err != ERROR_SUCCESS) return err;
part_current_[new_part] = new_count;
}
if (part_current_[new_part] < part_max[new_part])
{
if (part_current_new_ != new_part) part_current_new_ = new_part;
return ERROR_SUCCESS;
}
}
for (new_part = 0; new_part < part_current_new_; ++new_part)
{
if (part_current_[new_part] >= part_max[new_part] && move_enabled)
{
// refresh to get the actual value
auto new_count = 0;
auto err = IterPart(new_part, [&new_count](u64 idx) -> DWORD
{
++new_count;
return ERROR_SUCCESS;
});
if (err != ERROR_SUCCESS) return err;
part_current_[new_part] = new_count;
}
if (part_current_[new_part] < part_max[new_part])
{
if (part_current_new_ != new_part) part_current_new_ = new_part;
return ERROR_SUCCESS;
}
}
// this branch is not reachable because
// ReadChunkDiskFile() checks total part_max and
// WinSpd checks requested addresses
part_current_new_ = 0;
return ERROR_SUCCESS;
}
DWORD ChunkDiskBase::CreateChunk(const u64 chunk_idx, FileHandle& handle_out,
const bool is_write, const bool is_locked, const bool retrying)
{
if (read_only && is_write) return ERROR_ACCESS_DENIED;
const auto num_parts = part_dirname.size();
auto part_found = false;
auto part_idx = num_parts;
auto lk = SRWLock(*mutex_parts_, retrying);
if (retrying) chunk_parts_.erase(chunk_idx);
auto err = FindChunkPart(chunk_idx, part_idx, lk);
if (err != ERROR_SUCCESS) return err;
if (part_idx != num_parts)
{
part_found = true;
}
else if (is_write)
{
// assign part, will create chunk file
if (!lk.is_exclusive())
{
lk.switch_lock();
err = FindChunkPart(chunk_idx, part_idx, lk);
if (err != ERROR_SUCCESS) return err;
// lk is kept exclusive
}
// lk is exclusive
if (part_idx != num_parts)
{
part_found = true;
}
else
{
err = AssignChunkPart();
if (err != ERROR_SUCCESS) return err;
part_idx = part_current_new_;
}
}
if (!is_write && !part_found)
{
// not present -> empty handle
if (is_locked)
{
return ERROR_FILE_NOT_FOUND;
}
else
{
handle_out = FileHandle();
return ERROR_SUCCESS;
}
}
// !is_write -> part_found
// is_write -> part_found or assigned
auto path = std::wstring();
err = ChunkPath(chunk_idx, part_idx, path);
if (err != ERROR_SUCCESS) return err;
// GENERIC_READ means FILE_GENERIC_READ
// GENERIC_WRITE means FILE_GENERIC_WRITE | FILE_READ_ATTRIBUTES
// Note that a file can still be extended with FILE_APPEND_DATA flag unset
// https://docs.microsoft.com/en-us/windows/win32/fileio/file-security-and-access-rights
//
// Always set GENERIC_READ to make written data readable
// DELETE required for FILE_DISPOSITION_INFO{TRUE}
const auto desired_access = GENERIC_READ | (is_write ? GENERIC_WRITE : 0)
| ((is_write && is_locked) ? DELETE : 0);
// exclusive if is_locked and is_write
// may be shared by multiple threads if not is_locked
const auto share_mode = is_locked ? (is_write ? 0 : FILE_SHARE_READ)
: (FILE_SHARE_READ | FILE_SHARE_WRITE);
// unbuffered asynchronous I/O if not is_locked
// buffered synchronous I/O if is_locked
const auto flags_attrs = FILE_ATTRIBUTE_NORMAL |
(is_locked ? 0 : (FILE_FLAG_NO_BUFFERING | FILE_FLAG_OVERLAPPED));
// chunk file size in bytes
// integer overflow checked in ReadChunkDiskFile()
const auto chunk_bytes = LARGE_INTEGER{.QuadPart = LONGLONG(BlockBytes(ChunkBlocks(1)))};
if (is_write && !part_found)
{
// create non-empty chunk file or nothing
// file should not exist
auto h_locked = FileHandle(CreateFileW(
path.data(), desired_access | DELETE, 0, nullptr,
CREATE_NEW, FILE_ATTRIBUTE_NORMAL, nullptr));
if (!h_locked) return GetLastError();
// This just reserves disk space and sets file length on NTFS.
// Writing to the file actually extends the physical data, but synchronously.
// See https://devblogs.microsoft.com/oldnewthing/20150710-00/?p=45171.
err = SetFilePointerEx(h_locked.get(), chunk_bytes, nullptr, FILE_BEGIN)
? ERROR_SUCCESS : GetLastError();
if (err == ERROR_SUCCESS) err = SetEndOfFile(h_locked.get()) ? ERROR_SUCCESS : GetLastError();
if (err == ERROR_SUCCESS)
{
try
{
// lk was switched to exclusive
auto [part_it, emplaced] = chunk_parts_.emplace(chunk_idx, part_idx);
if (!emplaced) (*part_it).second = part_idx;
if (chunk_parts_.size() > MAX_CHUNK_PARTS) chunk_parts_.pop_front();
++part_current_[part_idx];
}
catch (const bad_alloc&)
{
err = ERROR_NOT_ENOUGH_MEMORY;
}
}
if (err != ERROR_SUCCESS)
{
auto disp = FILE_DISPOSITION_INFO{TRUE};
SetFileInformationByHandle(h_locked.get(), FileDispositionInfo, &disp, sizeof(disp));
h_locked.reset(); // will remove the file
return err;
}
if (is_locked)
{
if (SetFilePointerEx(h_locked.get(), LARGE_INTEGER{.QuadPart = 0}, nullptr, FILE_BEGIN))
{
// same argument for CreateFileW()
handle_out = std::move(h_locked);
return ERROR_SUCCESS;
}
}
}
// file should exist
auto h = FileHandle(CreateFileW(
path.data(), desired_access, share_mode, nullptr,
OPEN_EXISTING, flags_attrs, nullptr));
if (!h)
{
err = GetLastError();
if (err != ERROR_FILE_NOT_FOUND || !move_enabled || retrying) return err;
// rescan for chunk gone missing
lk.unlock();
return CreateChunk(chunk_idx, handle_out, is_write, is_locked, true);
}
if (!(is_write && !part_found))
{
// check existing chunk file
auto file_size = LARGE_INTEGER();
if (!GetFileSizeEx(h.get(), &file_size)) return GetLastError();
if (!is_write && file_size.QuadPart == 0)
{
// empty chunk, nothing to read -> return empty handle
if (!is_locked) h.reset();
}
else
{
if (file_size.QuadPart != 0 && file_size.QuadPart != chunk_bytes.QuadPart)
{
return ERROR_INCORRECT_SIZE;
}
if (is_write && file_size.QuadPart == 0)
{
// chunk will be non-empty, extend size
if (!SetFilePointerEx(h.get(), chunk_bytes, nullptr, FILE_BEGIN)) return GetLastError();
// This just reserves disk space and sets file length on NTFS.
if (!SetEndOfFile(h.get())) return GetLastError();
if (!SetFilePointerEx(h.get(), LARGE_INTEGER{.QuadPart = 0}, nullptr, FILE_BEGIN))
{
return GetLastError();
}
}
}
}
handle_out = std::move(h);
return ERROR_SUCCESS;
}
void ChunkDiskBase::RemoveChunkLocked(const u64 chunk_idx, FileHandle handle)
{
auto num_parts = usize(part_dirname.size());
auto part_idx = num_parts;
auto lk = SRWLock(*mutex_parts_, true);
auto err = FindChunkPart(chunk_idx, part_idx, lk);
if (err != ERROR_SUCCESS) return;
// lk is kept exclusive
if (part_idx == num_parts) return;
auto [part_it, emplaced] = chunk_parts_.emplace(chunk_idx, num_parts);
if (!emplaced) (*part_it).second = num_parts;
if (chunk_parts_.size() > MAX_CHUNK_PARTS) chunk_parts_.pop_front();
--part_current_[part_idx];
handle.reset();
}
}