1 /* Copyright (c) Mark Harmstone 2016-17 2 * 3 * This file is part of WinBtrfs. 4 * 5 * WinBtrfs is free software: you can redistribute it and/or modify 6 * it under the terms of the GNU Lesser General Public Licence as published by 7 * the Free Software Foundation, either version 3 of the Licence, or 8 * (at your option) any later version. 9 * 10 * WinBtrfs is distributed in the hope that it will be useful, 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 * GNU Lesser General Public Licence for more details. 14 * 15 * You should have received a copy of the GNU Lesser General Public Licence 16 * along with WinBtrfs. If not, see <http://www.gnu.org/licenses/>. */ 17 18 #include "btrfs_drv.h" 19 #include <ata.h> 20 #include <ntddscsi.h> 21 #include <ntddstor.h> 22 23 #define MAX_CSUM_SIZE (4096 - sizeof(tree_header) - sizeof(leaf_node)) 24 25 // #define DEBUG_WRITE_LOOPS 26 27 typedef struct { 28 KEVENT Event; 29 IO_STATUS_BLOCK iosb; 30 } write_context; 31 32 typedef struct { 33 EXTENT_ITEM_TREE eit; 34 UINT8 type; 35 TREE_BLOCK_REF tbr; 36 } EXTENT_ITEM_TREE2; 37 38 typedef struct { 39 EXTENT_ITEM ei; 40 UINT8 type; 41 TREE_BLOCK_REF tbr; 42 } EXTENT_ITEM_SKINNY_METADATA; 43 44 static NTSTATUS create_chunk(device_extension* Vcb, chunk* c, PIRP Irp); 45 static NTSTATUS update_tree_extents(device_extension* Vcb, tree* t, PIRP Irp, LIST_ENTRY* rollback); 46 47 #ifndef _MSC_VER // not in mingw yet 48 #define DEVICE_DSM_FLAG_TRIM_NOT_FS_ALLOCATED 0x80000000 49 #endif 50 51 _Function_class_(IO_COMPLETION_ROUTINE) 52 #ifdef __REACTOS__ 53 static NTSTATUS NTAPI write_completion(PDEVICE_OBJECT DeviceObject, PIRP Irp, PVOID conptr) { 54 #else 55 static NTSTATUS write_completion(PDEVICE_OBJECT DeviceObject, PIRP Irp, PVOID conptr) { 56 #endif 57 write_context* context = conptr; 58 59 UNUSED(DeviceObject); 60 61 context->iosb = Irp->IoStatus; 62 KeSetEvent(&context->Event, 0, FALSE); 63 64 return STATUS_MORE_PROCESSING_REQUIRED; 65 } 66 67 NTSTATUS write_data_phys(_In_ PDEVICE_OBJECT device, _In_ UINT64 address, _In_reads_bytes_(length) void* data, _In_ UINT32 length) { 68 NTSTATUS Status; 69 LARGE_INTEGER offset; 70 PIRP Irp; 71 PIO_STACK_LOCATION IrpSp; 72 write_context context; 73 74 TRACE("(%p, %llx, %p, %x)\n", device, address, data, length); 75 76 RtlZeroMemory(&context, sizeof(write_context)); 77 78 KeInitializeEvent(&context.Event, NotificationEvent, FALSE); 79 80 offset.QuadPart = address; 81 82 Irp = IoAllocateIrp(device->StackSize, FALSE); 83 84 if (!Irp) { 85 ERR("IoAllocateIrp failed\n"); 86 return STATUS_INSUFFICIENT_RESOURCES; 87 } 88 89 IrpSp = IoGetNextIrpStackLocation(Irp); 90 IrpSp->MajorFunction = IRP_MJ_WRITE; 91 92 if (device->Flags & DO_BUFFERED_IO) { 93 Irp->AssociatedIrp.SystemBuffer = data; 94 95 Irp->Flags = IRP_BUFFERED_IO; 96 } else if (device->Flags & DO_DIRECT_IO) { 97 Irp->MdlAddress = IoAllocateMdl(data, length, FALSE, FALSE, NULL); 98 if (!Irp->MdlAddress) { 99 DbgPrint("IoAllocateMdl failed\n"); 100 Status = STATUS_INSUFFICIENT_RESOURCES; 101 goto exit; 102 } 103 104 Status = STATUS_SUCCESS; 105 106 _SEH2_TRY { 107 MmProbeAndLockPages(Irp->MdlAddress, KernelMode, IoReadAccess); 108 } _SEH2_EXCEPT (EXCEPTION_EXECUTE_HANDLER) { 109 Status = _SEH2_GetExceptionCode(); 110 } _SEH2_END; 111 112 if (!NT_SUCCESS(Status)) { 113 ERR("MmProbeAndLockPages threw exception %08x\n", Status); 114 IoFreeMdl(Irp->MdlAddress); 115 goto exit; 116 } 117 } else { 118 Irp->UserBuffer = data; 119 } 120 121 IrpSp->Parameters.Write.Length = length; 122 IrpSp->Parameters.Write.ByteOffset = offset; 123 124 Irp->UserIosb = &context.iosb; 125 126 Irp->UserEvent = &context.Event; 127 128 IoSetCompletionRoutine(Irp, write_completion, &context, TRUE, TRUE, TRUE); 129 130 Status = IoCallDriver(device, Irp); 131 132 if (Status == STATUS_PENDING) { 133 KeWaitForSingleObject(&context.Event, Executive, KernelMode, FALSE, NULL); 134 Status = context.iosb.Status; 135 } 136 137 if (!NT_SUCCESS(Status)) { 138 ERR("IoCallDriver returned %08x\n", Status); 139 } 140 141 if (device->Flags & DO_DIRECT_IO) { 142 MmUnlockPages(Irp->MdlAddress); 143 IoFreeMdl(Irp->MdlAddress); 144 } 145 146 exit: 147 IoFreeIrp(Irp); 148 149 return Status; 150 } 151 152 static void add_trim_entry(device* dev, UINT64 address, UINT64 size) { 153 space* s = ExAllocatePoolWithTag(PagedPool, sizeof(space), ALLOC_TAG); 154 if (!s) { 155 ERR("out of memory\n"); 156 return; 157 } 158 159 s->address = address; 160 s->size = size; 161 dev->num_trim_entries++; 162 163 InsertTailList(&dev->trim_list, &s->list_entry); 164 } 165 166 static void clean_space_cache_chunk(device_extension* Vcb, chunk* c) { 167 ULONG type; 168 169 if (Vcb->trim && !Vcb->options.no_trim) { 170 if (c->chunk_item->type & BLOCK_FLAG_DUPLICATE) 171 type = BLOCK_FLAG_DUPLICATE; 172 else if (c->chunk_item->type & BLOCK_FLAG_RAID0) 173 type = BLOCK_FLAG_RAID0; 174 else if (c->chunk_item->type & BLOCK_FLAG_RAID1) 175 type = BLOCK_FLAG_DUPLICATE; 176 else if (c->chunk_item->type & BLOCK_FLAG_RAID10) 177 type = BLOCK_FLAG_RAID10; 178 else if (c->chunk_item->type & BLOCK_FLAG_RAID5) 179 type = BLOCK_FLAG_RAID5; 180 else if (c->chunk_item->type & BLOCK_FLAG_RAID6) 181 type = BLOCK_FLAG_RAID6; 182 else // SINGLE 183 type = BLOCK_FLAG_DUPLICATE; 184 } 185 186 while (!IsListEmpty(&c->deleting)) { 187 space* s = CONTAINING_RECORD(c->deleting.Flink, space, list_entry); 188 189 if (Vcb->trim && !Vcb->options.no_trim && (!Vcb->options.no_barrier || !(c->chunk_item->type & BLOCK_FLAG_METADATA))) { 190 CHUNK_ITEM_STRIPE* cis = (CHUNK_ITEM_STRIPE*)&c->chunk_item[1]; 191 192 if (type == BLOCK_FLAG_DUPLICATE) { 193 UINT16 i; 194 195 for (i = 0; i < c->chunk_item->num_stripes; i++) { 196 if (c->devices[i] && c->devices[i]->devobj && !c->devices[i]->readonly && c->devices[i]->trim) 197 add_trim_entry(c->devices[i], s->address - c->offset + cis[i].offset, s->size); 198 } 199 } else if (type == BLOCK_FLAG_RAID0) { 200 UINT64 startoff, endoff; 201 UINT16 startoffstripe, endoffstripe, i; 202 203 get_raid0_offset(s->address - c->offset, c->chunk_item->stripe_length, c->chunk_item->num_stripes, &startoff, &startoffstripe); 204 get_raid0_offset(s->address - c->offset + s->size - 1, c->chunk_item->stripe_length, c->chunk_item->num_stripes, &endoff, &endoffstripe); 205 206 for (i = 0; i < c->chunk_item->num_stripes; i++) { 207 if (c->devices[i] && c->devices[i]->devobj && !c->devices[i]->readonly && c->devices[i]->trim) { 208 UINT64 stripestart, stripeend; 209 210 if (startoffstripe > i) 211 stripestart = startoff - (startoff % c->chunk_item->stripe_length) + c->chunk_item->stripe_length; 212 else if (startoffstripe == i) 213 stripestart = startoff; 214 else 215 stripestart = startoff - (startoff % c->chunk_item->stripe_length); 216 217 if (endoffstripe > i) 218 stripeend = endoff - (endoff % c->chunk_item->stripe_length) + c->chunk_item->stripe_length; 219 else if (endoffstripe == i) 220 stripeend = endoff + 1; 221 else 222 stripeend = endoff - (endoff % c->chunk_item->stripe_length); 223 224 if (stripestart != stripeend) 225 add_trim_entry(c->devices[i], stripestart + cis[i].offset, stripeend - stripestart); 226 } 227 } 228 } else if (type == BLOCK_FLAG_RAID10) { 229 UINT64 startoff, endoff; 230 UINT16 sub_stripes, startoffstripe, endoffstripe, i; 231 232 sub_stripes = max(1, c->chunk_item->sub_stripes); 233 234 get_raid0_offset(s->address - c->offset, c->chunk_item->stripe_length, c->chunk_item->num_stripes / sub_stripes, &startoff, &startoffstripe); 235 get_raid0_offset(s->address - c->offset + s->size - 1, c->chunk_item->stripe_length, c->chunk_item->num_stripes / sub_stripes, &endoff, &endoffstripe); 236 237 startoffstripe *= sub_stripes; 238 endoffstripe *= sub_stripes; 239 240 for (i = 0; i < c->chunk_item->num_stripes; i += sub_stripes) { 241 ULONG j; 242 UINT64 stripestart, stripeend; 243 244 if (startoffstripe > i) 245 stripestart = startoff - (startoff % c->chunk_item->stripe_length) + c->chunk_item->stripe_length; 246 else if (startoffstripe == i) 247 stripestart = startoff; 248 else 249 stripestart = startoff - (startoff % c->chunk_item->stripe_length); 250 251 if (endoffstripe > i) 252 stripeend = endoff - (endoff % c->chunk_item->stripe_length) + c->chunk_item->stripe_length; 253 else if (endoffstripe == i) 254 stripeend = endoff + 1; 255 else 256 stripeend = endoff - (endoff % c->chunk_item->stripe_length); 257 258 if (stripestart != stripeend) { 259 for (j = 0; j < sub_stripes; j++) { 260 if (c->devices[i+j] && c->devices[i+j]->devobj && !c->devices[i+j]->readonly && c->devices[i+j]->trim) 261 add_trim_entry(c->devices[i+j], stripestart + cis[i+j].offset, stripeend - stripestart); 262 } 263 } 264 } 265 } 266 // FIXME - RAID5(?), RAID6(?) 267 } 268 269 RemoveEntryList(&s->list_entry); 270 ExFreePool(s); 271 } 272 } 273 274 typedef struct { 275 DEVICE_MANAGE_DATA_SET_ATTRIBUTES* dmdsa; 276 ATA_PASS_THROUGH_EX apte; 277 PIRP Irp; 278 IO_STATUS_BLOCK iosb; 279 } ioctl_context_stripe; 280 281 typedef struct { 282 KEVENT Event; 283 LONG left; 284 ioctl_context_stripe* stripes; 285 } ioctl_context; 286 287 _Function_class_(IO_COMPLETION_ROUTINE) 288 #ifdef __REACTOS__ 289 static NTSTATUS NTAPI ioctl_completion(PDEVICE_OBJECT DeviceObject, PIRP Irp, PVOID conptr) { 290 #else 291 static NTSTATUS ioctl_completion(PDEVICE_OBJECT DeviceObject, PIRP Irp, PVOID conptr) { 292 #endif 293 ioctl_context* context = (ioctl_context*)conptr; 294 LONG left2 = InterlockedDecrement(&context->left); 295 296 UNUSED(DeviceObject); 297 UNUSED(Irp); 298 299 if (left2 == 0) 300 KeSetEvent(&context->Event, 0, FALSE); 301 302 return STATUS_MORE_PROCESSING_REQUIRED; 303 } 304 305 static void clean_space_cache(device_extension* Vcb) { 306 LIST_ENTRY* le; 307 chunk* c; 308 ULONG num; 309 310 TRACE("(%p)\n", Vcb); 311 312 ExAcquireResourceSharedLite(&Vcb->chunk_lock, TRUE); 313 314 le = Vcb->chunks.Flink; 315 while (le != &Vcb->chunks) { 316 c = CONTAINING_RECORD(le, chunk, list_entry); 317 318 if (c->space_changed) { 319 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 320 321 if (c->space_changed) 322 clean_space_cache_chunk(Vcb, c); 323 324 c->space_changed = FALSE; 325 326 ExReleaseResourceLite(&c->lock); 327 } 328 329 le = le->Flink; 330 } 331 332 ExReleaseResourceLite(&Vcb->chunk_lock); 333 334 if (Vcb->trim && !Vcb->options.no_trim) { 335 ioctl_context context; 336 ULONG total_num; 337 338 context.left = 0; 339 340 le = Vcb->devices.Flink; 341 while (le != &Vcb->devices) { 342 device* dev = CONTAINING_RECORD(le, device, list_entry); 343 344 if (dev->devobj && !dev->readonly && dev->trim && dev->num_trim_entries > 0) 345 context.left++; 346 347 le = le->Flink; 348 } 349 350 if (context.left == 0) 351 return; 352 353 total_num = context.left; 354 num = 0; 355 356 KeInitializeEvent(&context.Event, NotificationEvent, FALSE); 357 358 context.stripes = ExAllocatePoolWithTag(NonPagedPool, sizeof(ioctl_context_stripe) * context.left, ALLOC_TAG); 359 if (!context.stripes) { 360 ERR("out of memory\n"); 361 return; 362 } 363 364 RtlZeroMemory(context.stripes, sizeof(ioctl_context_stripe) * context.left); 365 366 le = Vcb->devices.Flink; 367 while (le != &Vcb->devices) { 368 device* dev = CONTAINING_RECORD(le, device, list_entry); 369 370 if (dev->devobj && !dev->readonly && dev->trim && dev->num_trim_entries > 0) { 371 LIST_ENTRY* le2; 372 ioctl_context_stripe* stripe = &context.stripes[num]; 373 DEVICE_DATA_SET_RANGE* ranges; 374 ULONG datalen = (ULONG)sector_align(sizeof(DEVICE_MANAGE_DATA_SET_ATTRIBUTES), sizeof(UINT64)) + (dev->num_trim_entries * sizeof(DEVICE_DATA_SET_RANGE)), i; 375 PIO_STACK_LOCATION IrpSp; 376 377 stripe->dmdsa = ExAllocatePoolWithTag(PagedPool, datalen, ALLOC_TAG); 378 if (!stripe->dmdsa) { 379 ERR("out of memory\n"); 380 goto nextdev; 381 } 382 383 stripe->dmdsa->Size = sizeof(DEVICE_MANAGE_DATA_SET_ATTRIBUTES); 384 stripe->dmdsa->Action = DeviceDsmAction_Trim; 385 stripe->dmdsa->Flags = DEVICE_DSM_FLAG_TRIM_NOT_FS_ALLOCATED; 386 stripe->dmdsa->ParameterBlockOffset = 0; 387 stripe->dmdsa->ParameterBlockLength = 0; 388 stripe->dmdsa->DataSetRangesOffset = (ULONG)sector_align(sizeof(DEVICE_MANAGE_DATA_SET_ATTRIBUTES), sizeof(UINT64)); 389 stripe->dmdsa->DataSetRangesLength = dev->num_trim_entries * sizeof(DEVICE_DATA_SET_RANGE); 390 391 ranges = (DEVICE_DATA_SET_RANGE*)((UINT8*)stripe->dmdsa + stripe->dmdsa->DataSetRangesOffset); 392 393 i = 0; 394 395 le2 = dev->trim_list.Flink; 396 while (le2 != &dev->trim_list) { 397 space* s = CONTAINING_RECORD(le2, space, list_entry); 398 399 ranges[i].StartingOffset = s->address; 400 ranges[i].LengthInBytes = s->size; 401 i++; 402 403 le2 = le2->Flink; 404 } 405 406 stripe->Irp = IoAllocateIrp(dev->devobj->StackSize, FALSE); 407 408 if (!stripe->Irp) { 409 ERR("IoAllocateIrp failed\n"); 410 goto nextdev; 411 } 412 413 IrpSp = IoGetNextIrpStackLocation(stripe->Irp); 414 IrpSp->MajorFunction = IRP_MJ_DEVICE_CONTROL; 415 416 IrpSp->Parameters.DeviceIoControl.IoControlCode = IOCTL_STORAGE_MANAGE_DATA_SET_ATTRIBUTES; 417 IrpSp->Parameters.DeviceIoControl.InputBufferLength = datalen; 418 IrpSp->Parameters.DeviceIoControl.OutputBufferLength = 0; 419 420 stripe->Irp->AssociatedIrp.SystemBuffer = stripe->dmdsa; 421 stripe->Irp->Flags |= IRP_BUFFERED_IO; 422 stripe->Irp->UserBuffer = NULL; 423 stripe->Irp->UserIosb = &stripe->iosb; 424 425 IoSetCompletionRoutine(stripe->Irp, ioctl_completion, &context, TRUE, TRUE, TRUE); 426 427 IoCallDriver(dev->devobj, stripe->Irp); 428 429 nextdev: 430 while (!IsListEmpty(&dev->trim_list)) { 431 space* s = CONTAINING_RECORD(RemoveHeadList(&dev->trim_list), space, list_entry); 432 ExFreePool(s); 433 } 434 435 dev->num_trim_entries = 0; 436 437 num++; 438 } 439 440 le = le->Flink; 441 } 442 443 KeWaitForSingleObject(&context.Event, Executive, KernelMode, FALSE, NULL); 444 445 for (num = 0; num < total_num; num++) { 446 if (context.stripes[num].dmdsa) 447 ExFreePool(context.stripes[num].dmdsa); 448 } 449 450 ExFreePool(context.stripes); 451 } 452 } 453 454 static BOOL trees_consistent(device_extension* Vcb) { 455 ULONG maxsize = Vcb->superblock.node_size - sizeof(tree_header); 456 LIST_ENTRY* le; 457 458 le = Vcb->trees.Flink; 459 while (le != &Vcb->trees) { 460 tree* t = CONTAINING_RECORD(le, tree, list_entry); 461 462 if (t->write) { 463 if (t->header.num_items == 0 && t->parent) { 464 #ifdef DEBUG_WRITE_LOOPS 465 ERR("empty tree found, looping again\n"); 466 #endif 467 return FALSE; 468 } 469 470 if (t->size > maxsize) { 471 #ifdef DEBUG_WRITE_LOOPS 472 ERR("overlarge tree found (%u > %u), looping again\n", t->size, maxsize); 473 #endif 474 return FALSE; 475 } 476 477 if (!t->has_new_address) { 478 #ifdef DEBUG_WRITE_LOOPS 479 ERR("tree found without new address, looping again\n"); 480 #endif 481 return FALSE; 482 } 483 } 484 485 le = le->Flink; 486 } 487 488 return TRUE; 489 } 490 491 static NTSTATUS add_parents(device_extension* Vcb, PIRP Irp) { 492 ULONG level; 493 LIST_ENTRY* le; 494 495 for (level = 0; level <= 255; level++) { 496 BOOL nothing_found = TRUE; 497 498 TRACE("level = %u\n", level); 499 500 le = Vcb->trees.Flink; 501 while (le != &Vcb->trees) { 502 tree* t = CONTAINING_RECORD(le, tree, list_entry); 503 504 if (t->write && t->header.level == level) { 505 TRACE("tree %p: root = %llx, level = %x, parent = %p\n", t, t->header.tree_id, t->header.level, t->parent); 506 507 nothing_found = FALSE; 508 509 if (t->parent) { 510 if (!t->parent->write) 511 TRACE("adding tree %p (level %x)\n", t->parent, t->header.level); 512 513 t->parent->write = TRUE; 514 } else if (t->root != Vcb->root_root && t->root != Vcb->chunk_root) { 515 KEY searchkey; 516 traverse_ptr tp; 517 NTSTATUS Status; 518 519 searchkey.obj_id = t->root->id; 520 searchkey.obj_type = TYPE_ROOT_ITEM; 521 searchkey.offset = 0xffffffffffffffff; 522 523 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 524 if (!NT_SUCCESS(Status)) { 525 ERR("error - find_item returned %08x\n", Status); 526 return Status; 527 } 528 529 if (tp.item->key.obj_id != searchkey.obj_id || tp.item->key.obj_type != searchkey.obj_type) { 530 ERR("could not find ROOT_ITEM for tree %llx\n", searchkey.obj_id); 531 return STATUS_INTERNAL_ERROR; 532 } 533 534 if (tp.item->size < sizeof(ROOT_ITEM)) { // if not full length, delete and create new entry 535 ROOT_ITEM* ri = ExAllocatePoolWithTag(PagedPool, sizeof(ROOT_ITEM), ALLOC_TAG); 536 537 if (!ri) { 538 ERR("out of memory\n"); 539 return STATUS_INSUFFICIENT_RESOURCES; 540 } 541 542 RtlCopyMemory(ri, &t->root->root_item, sizeof(ROOT_ITEM)); 543 544 Status = delete_tree_item(Vcb, &tp); 545 if (!NT_SUCCESS(Status)) { 546 ERR("delete_tree_item returned %08x\n", Status); 547 ExFreePool(ri); 548 return Status; 549 } 550 551 Status = insert_tree_item(Vcb, Vcb->root_root, tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, ri, sizeof(ROOT_ITEM), NULL, Irp); 552 if (!NT_SUCCESS(Status)) { 553 ERR("insert_tree_item returned %08x\n", Status); 554 ExFreePool(ri); 555 return Status; 556 } 557 } 558 } 559 } 560 561 le = le->Flink; 562 } 563 564 if (nothing_found) 565 break; 566 } 567 568 return STATUS_SUCCESS; 569 } 570 571 static void add_parents_to_cache(tree* t) { 572 while (t->parent) { 573 t = t->parent; 574 t->write = TRUE; 575 } 576 } 577 578 static BOOL insert_tree_extent_skinny(device_extension* Vcb, UINT8 level, UINT64 root_id, chunk* c, UINT64 address, PIRP Irp, LIST_ENTRY* rollback) { 579 NTSTATUS Status; 580 EXTENT_ITEM_SKINNY_METADATA* eism; 581 traverse_ptr insert_tp; 582 583 eism = ExAllocatePoolWithTag(PagedPool, sizeof(EXTENT_ITEM_SKINNY_METADATA), ALLOC_TAG); 584 if (!eism) { 585 ERR("out of memory\n"); 586 return FALSE; 587 } 588 589 eism->ei.refcount = 1; 590 eism->ei.generation = Vcb->superblock.generation; 591 eism->ei.flags = EXTENT_ITEM_TREE_BLOCK; 592 eism->type = TYPE_TREE_BLOCK_REF; 593 eism->tbr.offset = root_id; 594 595 Status = insert_tree_item(Vcb, Vcb->extent_root, address, TYPE_METADATA_ITEM, level, eism, sizeof(EXTENT_ITEM_SKINNY_METADATA), &insert_tp, Irp); 596 if (!NT_SUCCESS(Status)) { 597 ERR("insert_tree_item returned %08x\n", Status); 598 ExFreePool(eism); 599 return FALSE; 600 } 601 602 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 603 604 space_list_subtract(c, FALSE, address, Vcb->superblock.node_size, rollback); 605 606 ExReleaseResourceLite(&c->lock); 607 608 add_parents_to_cache(insert_tp.tree); 609 610 return TRUE; 611 } 612 613 BOOL find_metadata_address_in_chunk(device_extension* Vcb, chunk* c, UINT64* address) { 614 LIST_ENTRY* le; 615 space* s; 616 617 TRACE("(%p, %llx, %p)\n", Vcb, c->offset, address); 618 619 if (Vcb->superblock.node_size > c->chunk_item->size - c->used) 620 return FALSE; 621 622 if (!c->cache_loaded) { 623 NTSTATUS Status = load_cache_chunk(Vcb, c, NULL); 624 625 if (!NT_SUCCESS(Status)) { 626 ERR("load_cache_chunk returned %08x\n", Status); 627 return FALSE; 628 } 629 } 630 631 if (IsListEmpty(&c->space_size)) 632 return FALSE; 633 634 if (!c->last_alloc_set) { 635 s = CONTAINING_RECORD(c->space.Blink, space, list_entry); 636 637 c->last_alloc = s->address; 638 c->last_alloc_set = TRUE; 639 640 if (s->size >= Vcb->superblock.node_size) { 641 *address = s->address; 642 c->last_alloc += Vcb->superblock.node_size; 643 return TRUE; 644 } 645 } 646 647 le = c->space.Flink; 648 while (le != &c->space) { 649 s = CONTAINING_RECORD(le, space, list_entry); 650 651 if (s->address <= c->last_alloc && s->address + s->size >= c->last_alloc + Vcb->superblock.node_size) { 652 *address = c->last_alloc; 653 c->last_alloc += Vcb->superblock.node_size; 654 return TRUE; 655 } 656 657 le = le->Flink; 658 } 659 660 le = c->space_size.Flink; 661 while (le != &c->space_size) { 662 s = CONTAINING_RECORD(le, space, list_entry_size); 663 664 if (s->size == Vcb->superblock.node_size) { 665 *address = s->address; 666 c->last_alloc = s->address + Vcb->superblock.node_size; 667 return TRUE; 668 } else if (s->size < Vcb->superblock.node_size) { 669 if (le == c->space_size.Flink) 670 return FALSE; 671 672 s = CONTAINING_RECORD(le->Blink, space, list_entry_size); 673 674 *address = s->address; 675 c->last_alloc = s->address + Vcb->superblock.node_size; 676 677 return TRUE; 678 } 679 680 le = le->Flink; 681 } 682 683 s = CONTAINING_RECORD(c->space_size.Blink, space, list_entry_size); 684 685 if (s->size > Vcb->superblock.node_size) { 686 *address = s->address; 687 c->last_alloc = s->address + Vcb->superblock.node_size; 688 return TRUE; 689 } 690 691 return FALSE; 692 } 693 694 static BOOL insert_tree_extent(device_extension* Vcb, UINT8 level, UINT64 root_id, chunk* c, UINT64* new_address, PIRP Irp, LIST_ENTRY* rollback) { 695 NTSTATUS Status; 696 UINT64 address; 697 EXTENT_ITEM_TREE2* eit2; 698 traverse_ptr insert_tp; 699 700 TRACE("(%p, %x, %llx, %p, %p, %p, %p)\n", Vcb, level, root_id, c, new_address, rollback); 701 702 if (!find_metadata_address_in_chunk(Vcb, c, &address)) 703 return FALSE; 704 705 if (Vcb->superblock.incompat_flags & BTRFS_INCOMPAT_FLAGS_SKINNY_METADATA) { 706 BOOL b = insert_tree_extent_skinny(Vcb, level, root_id, c, address, Irp, rollback); 707 708 if (b) 709 *new_address = address; 710 711 return b; 712 } 713 714 eit2 = ExAllocatePoolWithTag(PagedPool, sizeof(EXTENT_ITEM_TREE2), ALLOC_TAG); 715 if (!eit2) { 716 ERR("out of memory\n"); 717 return FALSE; 718 } 719 720 eit2->eit.extent_item.refcount = 1; 721 eit2->eit.extent_item.generation = Vcb->superblock.generation; 722 eit2->eit.extent_item.flags = EXTENT_ITEM_TREE_BLOCK; 723 eit2->eit.level = level; 724 eit2->type = TYPE_TREE_BLOCK_REF; 725 eit2->tbr.offset = root_id; 726 727 Status = insert_tree_item(Vcb, Vcb->extent_root, address, TYPE_EXTENT_ITEM, Vcb->superblock.node_size, eit2, sizeof(EXTENT_ITEM_TREE2), &insert_tp, Irp); 728 if (!NT_SUCCESS(Status)) { 729 ERR("insert_tree_item returned %08x\n", Status); 730 ExFreePool(eit2); 731 return FALSE; 732 } 733 734 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 735 736 space_list_subtract(c, FALSE, address, Vcb->superblock.node_size, rollback); 737 738 ExReleaseResourceLite(&c->lock); 739 740 add_parents_to_cache(insert_tp.tree); 741 742 *new_address = address; 743 744 return TRUE; 745 } 746 747 NTSTATUS get_tree_new_address(device_extension* Vcb, tree* t, PIRP Irp, LIST_ENTRY* rollback) { 748 NTSTATUS Status; 749 chunk *origchunk = NULL, *c; 750 LIST_ENTRY* le; 751 UINT64 flags, addr; 752 753 if (t->root->id == BTRFS_ROOT_CHUNK) 754 flags = Vcb->system_flags; 755 else 756 flags = Vcb->metadata_flags; 757 758 if (t->has_address) { 759 origchunk = get_chunk_from_address(Vcb, t->header.address); 760 761 if (origchunk && !origchunk->readonly && !origchunk->reloc && origchunk->chunk_item->type == flags && 762 insert_tree_extent(Vcb, t->header.level, t->root->id, origchunk, &addr, Irp, rollback)) { 763 t->new_address = addr; 764 t->has_new_address = TRUE; 765 return STATUS_SUCCESS; 766 } 767 } 768 769 ExAcquireResourceExclusiveLite(&Vcb->chunk_lock, TRUE); 770 771 le = Vcb->chunks.Flink; 772 while (le != &Vcb->chunks) { 773 c = CONTAINING_RECORD(le, chunk, list_entry); 774 775 if (!c->readonly && !c->reloc) { 776 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 777 778 if (c != origchunk && c->chunk_item->type == flags && (c->chunk_item->size - c->used) >= Vcb->superblock.node_size) { 779 if (insert_tree_extent(Vcb, t->header.level, t->root->id, c, &addr, Irp, rollback)) { 780 ExReleaseResourceLite(&c->lock); 781 ExReleaseResourceLite(&Vcb->chunk_lock); 782 t->new_address = addr; 783 t->has_new_address = TRUE; 784 return STATUS_SUCCESS; 785 } 786 } 787 788 ExReleaseResourceLite(&c->lock); 789 } 790 791 le = le->Flink; 792 } 793 794 // allocate new chunk if necessary 795 796 Status = alloc_chunk(Vcb, flags, &c, FALSE); 797 798 if (!NT_SUCCESS(Status)) { 799 ERR("alloc_chunk returned %08x\n", Status); 800 ExReleaseResourceLite(&Vcb->chunk_lock); 801 return Status; 802 } 803 804 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 805 806 if ((c->chunk_item->size - c->used) >= Vcb->superblock.node_size) { 807 if (insert_tree_extent(Vcb, t->header.level, t->root->id, c, &addr, Irp, rollback)) { 808 ExReleaseResourceLite(&c->lock); 809 ExReleaseResourceLite(&Vcb->chunk_lock); 810 t->new_address = addr; 811 t->has_new_address = TRUE; 812 return STATUS_SUCCESS; 813 } 814 } 815 816 ExReleaseResourceLite(&c->lock); 817 818 ExReleaseResourceLite(&Vcb->chunk_lock); 819 820 ERR("couldn't find any metadata chunks with %x bytes free\n", Vcb->superblock.node_size); 821 822 return STATUS_DISK_FULL; 823 } 824 825 static NTSTATUS reduce_tree_extent(device_extension* Vcb, UINT64 address, tree* t, UINT64 parent_root, UINT8 level, PIRP Irp, LIST_ENTRY* rollback) { 826 NTSTATUS Status; 827 UINT64 rc, root; 828 829 TRACE("(%p, %llx, %p)\n", Vcb, address, t); 830 831 rc = get_extent_refcount(Vcb, address, Vcb->superblock.node_size, Irp); 832 if (rc == 0) { 833 ERR("error - refcount for extent %llx was 0\n", address); 834 return STATUS_INTERNAL_ERROR; 835 } 836 837 if (!t || t->parent) 838 root = parent_root; 839 else 840 root = t->header.tree_id; 841 842 Status = decrease_extent_refcount_tree(Vcb, address, Vcb->superblock.node_size, root, level, Irp); 843 if (!NT_SUCCESS(Status)) { 844 ERR("decrease_extent_refcount_tree returned %08x\n", Status); 845 return Status; 846 } 847 848 if (rc == 1) { 849 chunk* c = get_chunk_from_address(Vcb, address); 850 851 if (c) { 852 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 853 854 if (!c->cache_loaded) { 855 Status = load_cache_chunk(Vcb, c, NULL); 856 857 if (!NT_SUCCESS(Status)) { 858 ERR("load_cache_chunk returned %08x\n", Status); 859 ExReleaseResourceLite(&c->lock); 860 return Status; 861 } 862 } 863 864 c->used -= Vcb->superblock.node_size; 865 866 space_list_add(c, address, Vcb->superblock.node_size, rollback); 867 868 ExReleaseResourceLite(&c->lock); 869 } else 870 ERR("could not find chunk for address %llx\n", address); 871 } 872 873 return STATUS_SUCCESS; 874 } 875 876 static NTSTATUS add_changed_extent_ref_edr(changed_extent* ce, EXTENT_DATA_REF* edr, BOOL old) { 877 LIST_ENTRY *le2, *list; 878 changed_extent_ref* cer; 879 880 list = old ? &ce->old_refs : &ce->refs; 881 882 le2 = list->Flink; 883 while (le2 != list) { 884 cer = CONTAINING_RECORD(le2, changed_extent_ref, list_entry); 885 886 if (cer->type == TYPE_EXTENT_DATA_REF && cer->edr.root == edr->root && cer->edr.objid == edr->objid && cer->edr.offset == edr->offset) { 887 cer->edr.count += edr->count; 888 goto end; 889 } 890 891 le2 = le2->Flink; 892 } 893 894 cer = ExAllocatePoolWithTag(PagedPool, sizeof(changed_extent_ref), ALLOC_TAG); 895 if (!cer) { 896 ERR("out of memory\n"); 897 return STATUS_INSUFFICIENT_RESOURCES; 898 } 899 900 cer->type = TYPE_EXTENT_DATA_REF; 901 RtlCopyMemory(&cer->edr, edr, sizeof(EXTENT_DATA_REF)); 902 InsertTailList(list, &cer->list_entry); 903 904 end: 905 if (old) 906 ce->old_count += edr->count; 907 else 908 ce->count += edr->count; 909 910 return STATUS_SUCCESS; 911 } 912 913 static NTSTATUS add_changed_extent_ref_sdr(changed_extent* ce, SHARED_DATA_REF* sdr, BOOL old) { 914 LIST_ENTRY *le2, *list; 915 changed_extent_ref* cer; 916 917 list = old ? &ce->old_refs : &ce->refs; 918 919 le2 = list->Flink; 920 while (le2 != list) { 921 cer = CONTAINING_RECORD(le2, changed_extent_ref, list_entry); 922 923 if (cer->type == TYPE_SHARED_DATA_REF && cer->sdr.offset == sdr->offset) { 924 cer->sdr.count += sdr->count; 925 goto end; 926 } 927 928 le2 = le2->Flink; 929 } 930 931 cer = ExAllocatePoolWithTag(PagedPool, sizeof(changed_extent_ref), ALLOC_TAG); 932 if (!cer) { 933 ERR("out of memory\n"); 934 return STATUS_INSUFFICIENT_RESOURCES; 935 } 936 937 cer->type = TYPE_SHARED_DATA_REF; 938 RtlCopyMemory(&cer->sdr, sdr, sizeof(SHARED_DATA_REF)); 939 InsertTailList(list, &cer->list_entry); 940 941 end: 942 if (old) 943 ce->old_count += sdr->count; 944 else 945 ce->count += sdr->count; 946 947 return STATUS_SUCCESS; 948 } 949 950 static BOOL shared_tree_is_unique(device_extension* Vcb, tree* t, PIRP Irp, LIST_ENTRY* rollback) { 951 KEY searchkey; 952 traverse_ptr tp; 953 NTSTATUS Status; 954 955 if (!t->updated_extents && t->has_address) { 956 Status = update_tree_extents(Vcb, t, Irp, rollback); 957 if (!NT_SUCCESS(Status)) { 958 ERR("update_tree_extents returned %08x\n", Status); 959 return FALSE; 960 } 961 } 962 963 searchkey.obj_id = t->header.address; 964 searchkey.obj_type = Vcb->superblock.incompat_flags & BTRFS_INCOMPAT_FLAGS_SKINNY_METADATA ? TYPE_METADATA_ITEM : TYPE_EXTENT_ITEM; 965 searchkey.offset = 0xffffffffffffffff; 966 967 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 968 if (!NT_SUCCESS(Status)) { 969 ERR("error - find_item returned %08x\n", Status); 970 return FALSE; 971 } 972 973 if (tp.item->key.obj_id == t->header.address && (tp.item->key.obj_type == TYPE_METADATA_ITEM || tp.item->key.obj_type == TYPE_EXTENT_ITEM)) 974 return FALSE; 975 else 976 return TRUE; 977 } 978 979 static NTSTATUS update_tree_extents(device_extension* Vcb, tree* t, PIRP Irp, LIST_ENTRY* rollback) { 980 NTSTATUS Status; 981 UINT64 rc = get_extent_refcount(Vcb, t->header.address, Vcb->superblock.node_size, Irp); 982 UINT64 flags = get_extent_flags(Vcb, t->header.address, Irp); 983 984 if (rc == 0) { 985 ERR("refcount for extent %llx was 0\n", t->header.address); 986 return STATUS_INTERNAL_ERROR; 987 } 988 989 if (flags & EXTENT_ITEM_SHARED_BACKREFS || t->header.flags & HEADER_FLAG_SHARED_BACKREF || !(t->header.flags & HEADER_FLAG_MIXED_BACKREF)) { 990 TREE_BLOCK_REF tbr; 991 BOOL unique = rc > 1 ? FALSE : (t->parent ? shared_tree_is_unique(Vcb, t->parent, Irp, rollback) : FALSE); 992 993 if (t->header.level == 0) { 994 LIST_ENTRY* le; 995 996 le = t->itemlist.Flink; 997 while (le != &t->itemlist) { 998 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 999 1000 if (!td->inserted && td->key.obj_type == TYPE_EXTENT_DATA && td->size >= sizeof(EXTENT_DATA) - 1 + sizeof(EXTENT_DATA2)) { 1001 EXTENT_DATA* ed = (EXTENT_DATA*)td->data; 1002 1003 if (ed->type == EXTENT_TYPE_REGULAR || ed->type == EXTENT_TYPE_PREALLOC) { 1004 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ed->data; 1005 1006 if (ed2->size > 0) { 1007 EXTENT_DATA_REF edr; 1008 changed_extent* ce = NULL; 1009 chunk* c = get_chunk_from_address(Vcb, ed2->address); 1010 1011 if (c) { 1012 LIST_ENTRY* le2; 1013 1014 le2 = c->changed_extents.Flink; 1015 while (le2 != &c->changed_extents) { 1016 changed_extent* ce2 = CONTAINING_RECORD(le2, changed_extent, list_entry); 1017 1018 if (ce2->address == ed2->address) { 1019 ce = ce2; 1020 break; 1021 } 1022 1023 le2 = le2->Flink; 1024 } 1025 } 1026 1027 edr.root = t->root->id; 1028 edr.objid = td->key.obj_id; 1029 edr.offset = td->key.offset - ed2->offset; 1030 edr.count = 1; 1031 1032 if (ce) { 1033 Status = add_changed_extent_ref_edr(ce, &edr, TRUE); 1034 if (!NT_SUCCESS(Status)) { 1035 ERR("add_changed_extent_ref_edr returned %08x\n", Status); 1036 return Status; 1037 } 1038 1039 Status = add_changed_extent_ref_edr(ce, &edr, FALSE); 1040 if (!NT_SUCCESS(Status)) { 1041 ERR("add_changed_extent_ref_edr returned %08x\n", Status); 1042 return Status; 1043 } 1044 } 1045 1046 Status = increase_extent_refcount(Vcb, ed2->address, ed2->size, TYPE_EXTENT_DATA_REF, &edr, NULL, 0, Irp); 1047 if (!NT_SUCCESS(Status)) { 1048 ERR("increase_extent_refcount returned %08x\n", Status); 1049 return Status; 1050 } 1051 1052 if ((flags & EXTENT_ITEM_SHARED_BACKREFS && unique) || !(t->header.flags & HEADER_FLAG_MIXED_BACKREF)) { 1053 UINT64 sdrrc = find_extent_shared_data_refcount(Vcb, ed2->address, t->header.address, Irp); 1054 1055 if (sdrrc > 0) { 1056 SHARED_DATA_REF sdr; 1057 1058 sdr.offset = t->header.address; 1059 sdr.count = 1; 1060 1061 Status = decrease_extent_refcount(Vcb, ed2->address, ed2->size, TYPE_SHARED_DATA_REF, &sdr, NULL, 0, 1062 t->header.address, ce ? ce->superseded : FALSE, Irp); 1063 if (!NT_SUCCESS(Status)) { 1064 ERR("decrease_extent_refcount returned %08x\n", Status); 1065 return Status; 1066 } 1067 1068 if (ce) { 1069 LIST_ENTRY* le2; 1070 1071 le2 = ce->refs.Flink; 1072 while (le2 != &ce->refs) { 1073 changed_extent_ref* cer = CONTAINING_RECORD(le2, changed_extent_ref, list_entry); 1074 1075 if (cer->type == TYPE_SHARED_DATA_REF && cer->sdr.offset == sdr.offset) { 1076 ce->count--; 1077 cer->sdr.count--; 1078 break; 1079 } 1080 1081 le2 = le2->Flink; 1082 } 1083 1084 le2 = ce->old_refs.Flink; 1085 while (le2 != &ce->old_refs) { 1086 changed_extent_ref* cer = CONTAINING_RECORD(le2, changed_extent_ref, list_entry); 1087 1088 if (cer->type == TYPE_SHARED_DATA_REF && cer->sdr.offset == sdr.offset) { 1089 ce->old_count--; 1090 1091 if (cer->sdr.count > 1) 1092 cer->sdr.count--; 1093 else { 1094 RemoveEntryList(&cer->list_entry); 1095 ExFreePool(cer); 1096 } 1097 1098 break; 1099 } 1100 1101 le2 = le2->Flink; 1102 } 1103 } 1104 } 1105 } 1106 1107 // FIXME - clear shared flag if unique? 1108 } 1109 } 1110 } 1111 1112 le = le->Flink; 1113 } 1114 } else { 1115 LIST_ENTRY* le; 1116 1117 le = t->itemlist.Flink; 1118 while (le != &t->itemlist) { 1119 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 1120 1121 if (!td->inserted) { 1122 tbr.offset = t->root->id; 1123 1124 Status = increase_extent_refcount(Vcb, td->treeholder.address, Vcb->superblock.node_size, TYPE_TREE_BLOCK_REF, 1125 &tbr, &td->key, t->header.level - 1, Irp); 1126 if (!NT_SUCCESS(Status)) { 1127 ERR("increase_extent_refcount returned %08x\n", Status); 1128 return Status; 1129 } 1130 1131 if (unique || !(t->header.flags & HEADER_FLAG_MIXED_BACKREF)) { 1132 UINT64 sbrrc = find_extent_shared_tree_refcount(Vcb, td->treeholder.address, t->header.address, Irp); 1133 1134 if (sbrrc > 0) { 1135 SHARED_BLOCK_REF sbr; 1136 1137 sbr.offset = t->header.address; 1138 1139 Status = decrease_extent_refcount(Vcb, td->treeholder.address, Vcb->superblock.node_size, TYPE_SHARED_BLOCK_REF, &sbr, NULL, 0, 1140 t->header.address, FALSE, Irp); 1141 if (!NT_SUCCESS(Status)) { 1142 ERR("decrease_extent_refcount returned %08x\n", Status); 1143 return Status; 1144 } 1145 } 1146 } 1147 1148 // FIXME - clear shared flag if unique? 1149 } 1150 1151 le = le->Flink; 1152 } 1153 } 1154 1155 if (unique) { 1156 UINT64 sbrrc = find_extent_shared_tree_refcount(Vcb, t->header.address, t->parent->header.address, Irp); 1157 1158 if (sbrrc == 1) { 1159 SHARED_BLOCK_REF sbr; 1160 1161 sbr.offset = t->parent->header.address; 1162 1163 Status = decrease_extent_refcount(Vcb, t->header.address, Vcb->superblock.node_size, TYPE_SHARED_BLOCK_REF, &sbr, NULL, 0, 1164 t->parent->header.address, FALSE, Irp); 1165 if (!NT_SUCCESS(Status)) { 1166 ERR("decrease_extent_refcount returned %08x\n", Status); 1167 return Status; 1168 } 1169 } 1170 } 1171 1172 if (t->parent) 1173 tbr.offset = t->parent->header.tree_id; 1174 else 1175 tbr.offset = t->header.tree_id; 1176 1177 Status = increase_extent_refcount(Vcb, t->header.address, Vcb->superblock.node_size, TYPE_TREE_BLOCK_REF, &tbr, 1178 t->parent ? &t->paritem->key : NULL, t->header.level, Irp); 1179 if (!NT_SUCCESS(Status)) { 1180 ERR("increase_extent_refcount returned %08x\n", Status); 1181 return Status; 1182 } 1183 1184 // FIXME - clear shared flag if unique? 1185 1186 t->header.flags &= ~HEADER_FLAG_SHARED_BACKREF; 1187 } 1188 1189 if (rc > 1 || t->header.tree_id == t->root->id) { 1190 Status = reduce_tree_extent(Vcb, t->header.address, t, t->parent ? t->parent->header.tree_id : t->header.tree_id, t->header.level, Irp, rollback); 1191 1192 if (!NT_SUCCESS(Status)) { 1193 ERR("reduce_tree_extent returned %08x\n", Status); 1194 return Status; 1195 } 1196 } 1197 1198 t->has_address = FALSE; 1199 1200 if ((rc > 1 || t->header.tree_id != t->root->id) && !(flags & EXTENT_ITEM_SHARED_BACKREFS)) { 1201 if (t->header.tree_id == t->root->id) { 1202 flags |= EXTENT_ITEM_SHARED_BACKREFS; 1203 update_extent_flags(Vcb, t->header.address, flags, Irp); 1204 } 1205 1206 if (t->header.level > 0) { 1207 LIST_ENTRY* le; 1208 1209 le = t->itemlist.Flink; 1210 while (le != &t->itemlist) { 1211 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 1212 1213 if (!td->inserted) { 1214 if (t->header.tree_id == t->root->id) { 1215 SHARED_BLOCK_REF sbr; 1216 1217 sbr.offset = t->header.address; 1218 1219 Status = increase_extent_refcount(Vcb, td->treeholder.address, Vcb->superblock.node_size, TYPE_SHARED_BLOCK_REF, &sbr, &td->key, t->header.level - 1, Irp); 1220 } else { 1221 TREE_BLOCK_REF tbr; 1222 1223 tbr.offset = t->root->id; 1224 1225 Status = increase_extent_refcount(Vcb, td->treeholder.address, Vcb->superblock.node_size, TYPE_TREE_BLOCK_REF, &tbr, &td->key, t->header.level - 1, Irp); 1226 } 1227 1228 if (!NT_SUCCESS(Status)) { 1229 ERR("increase_extent_refcount returned %08x\n", Status); 1230 return Status; 1231 } 1232 } 1233 1234 le = le->Flink; 1235 } 1236 } else { 1237 LIST_ENTRY* le; 1238 1239 le = t->itemlist.Flink; 1240 while (le != &t->itemlist) { 1241 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 1242 1243 if (!td->inserted && td->key.obj_type == TYPE_EXTENT_DATA && td->size >= sizeof(EXTENT_DATA) - 1 + sizeof(EXTENT_DATA2)) { 1244 EXTENT_DATA* ed = (EXTENT_DATA*)td->data; 1245 1246 if (ed->type == EXTENT_TYPE_REGULAR || ed->type == EXTENT_TYPE_PREALLOC) { 1247 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ed->data; 1248 1249 if (ed2->size > 0) { 1250 changed_extent* ce = NULL; 1251 chunk* c = get_chunk_from_address(Vcb, ed2->address); 1252 1253 if (c) { 1254 LIST_ENTRY* le2; 1255 1256 le2 = c->changed_extents.Flink; 1257 while (le2 != &c->changed_extents) { 1258 changed_extent* ce2 = CONTAINING_RECORD(le2, changed_extent, list_entry); 1259 1260 if (ce2->address == ed2->address) { 1261 ce = ce2; 1262 break; 1263 } 1264 1265 le2 = le2->Flink; 1266 } 1267 } 1268 1269 if (t->header.tree_id == t->root->id) { 1270 SHARED_DATA_REF sdr; 1271 1272 sdr.offset = t->header.address; 1273 sdr.count = 1; 1274 1275 if (ce) { 1276 Status = add_changed_extent_ref_sdr(ce, &sdr, TRUE); 1277 if (!NT_SUCCESS(Status)) { 1278 ERR("add_changed_extent_ref_edr returned %08x\n", Status); 1279 return Status; 1280 } 1281 1282 Status = add_changed_extent_ref_sdr(ce, &sdr, FALSE); 1283 if (!NT_SUCCESS(Status)) { 1284 ERR("add_changed_extent_ref_edr returned %08x\n", Status); 1285 return Status; 1286 } 1287 } 1288 1289 Status = increase_extent_refcount(Vcb, ed2->address, ed2->size, TYPE_SHARED_DATA_REF, &sdr, NULL, 0, Irp); 1290 } else { 1291 EXTENT_DATA_REF edr; 1292 1293 edr.root = t->root->id; 1294 edr.objid = td->key.obj_id; 1295 edr.offset = td->key.offset - ed2->offset; 1296 edr.count = 1; 1297 1298 if (ce) { 1299 Status = add_changed_extent_ref_edr(ce, &edr, TRUE); 1300 if (!NT_SUCCESS(Status)) { 1301 ERR("add_changed_extent_ref_edr returned %08x\n", Status); 1302 return Status; 1303 } 1304 1305 Status = add_changed_extent_ref_edr(ce, &edr, FALSE); 1306 if (!NT_SUCCESS(Status)) { 1307 ERR("add_changed_extent_ref_edr returned %08x\n", Status); 1308 return Status; 1309 } 1310 } 1311 1312 Status = increase_extent_refcount(Vcb, ed2->address, ed2->size, TYPE_EXTENT_DATA_REF, &edr, NULL, 0, Irp); 1313 } 1314 1315 if (!NT_SUCCESS(Status)) { 1316 ERR("increase_extent_refcount returned %08x\n", Status); 1317 return Status; 1318 } 1319 } 1320 } 1321 } 1322 1323 le = le->Flink; 1324 } 1325 } 1326 } 1327 1328 t->updated_extents = TRUE; 1329 t->header.tree_id = t->root->id; 1330 1331 return STATUS_SUCCESS; 1332 } 1333 1334 static NTSTATUS allocate_tree_extents(device_extension* Vcb, PIRP Irp, LIST_ENTRY* rollback) { 1335 LIST_ENTRY* le; 1336 NTSTATUS Status; 1337 BOOL changed = FALSE; 1338 UINT8 max_level = 0, level; 1339 1340 TRACE("(%p)\n", Vcb); 1341 1342 le = Vcb->trees.Flink; 1343 while (le != &Vcb->trees) { 1344 tree* t = CONTAINING_RECORD(le, tree, list_entry); 1345 1346 if (t->write && !t->has_new_address) { 1347 chunk* c; 1348 1349 if (t->has_address) { 1350 c = get_chunk_from_address(Vcb, t->header.address); 1351 1352 if (c) { 1353 if (!c->cache_loaded) { 1354 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 1355 1356 if (!c->cache_loaded) { 1357 Status = load_cache_chunk(Vcb, c, NULL); 1358 1359 if (!NT_SUCCESS(Status)) { 1360 ERR("load_cache_chunk returned %08x\n", Status); 1361 ExReleaseResourceLite(&c->lock); 1362 return Status; 1363 } 1364 } 1365 1366 ExReleaseResourceLite(&c->lock); 1367 } 1368 } 1369 } 1370 1371 Status = get_tree_new_address(Vcb, t, Irp, rollback); 1372 if (!NT_SUCCESS(Status)) { 1373 ERR("get_tree_new_address returned %08x\n", Status); 1374 return Status; 1375 } 1376 1377 TRACE("allocated extent %llx\n", t->new_address); 1378 1379 c = get_chunk_from_address(Vcb, t->new_address); 1380 1381 if (c) 1382 c->used += Vcb->superblock.node_size; 1383 else { 1384 ERR("could not find chunk for address %llx\n", t->new_address); 1385 return STATUS_INTERNAL_ERROR; 1386 } 1387 1388 changed = TRUE; 1389 1390 if (t->header.level > max_level) 1391 max_level = t->header.level; 1392 } 1393 1394 le = le->Flink; 1395 } 1396 1397 if (!changed) 1398 return STATUS_SUCCESS; 1399 1400 level = max_level; 1401 do { 1402 le = Vcb->trees.Flink; 1403 while (le != &Vcb->trees) { 1404 tree* t = CONTAINING_RECORD(le, tree, list_entry); 1405 1406 if (t->write && !t->updated_extents && t->has_address && t->header.level == level) { 1407 Status = update_tree_extents(Vcb, t, Irp, rollback); 1408 if (!NT_SUCCESS(Status)) { 1409 ERR("update_tree_extents returned %08x\n", Status); 1410 return Status; 1411 } 1412 } 1413 1414 le = le->Flink; 1415 } 1416 1417 if (level == 0) 1418 break; 1419 1420 level--; 1421 } while (TRUE); 1422 1423 return STATUS_SUCCESS; 1424 } 1425 1426 static NTSTATUS update_root_root(device_extension* Vcb, BOOL no_cache, PIRP Irp, LIST_ENTRY* rollback) { 1427 LIST_ENTRY* le; 1428 NTSTATUS Status; 1429 1430 TRACE("(%p)\n", Vcb); 1431 1432 le = Vcb->trees.Flink; 1433 while (le != &Vcb->trees) { 1434 tree* t = CONTAINING_RECORD(le, tree, list_entry); 1435 1436 if (t->write && !t->parent) { 1437 if (t->root != Vcb->root_root && t->root != Vcb->chunk_root) { 1438 KEY searchkey; 1439 traverse_ptr tp; 1440 1441 searchkey.obj_id = t->root->id; 1442 searchkey.obj_type = TYPE_ROOT_ITEM; 1443 searchkey.offset = 0xffffffffffffffff; 1444 1445 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 1446 if (!NT_SUCCESS(Status)) { 1447 ERR("error - find_item returned %08x\n", Status); 1448 return Status; 1449 } 1450 1451 if (tp.item->key.obj_id != searchkey.obj_id || tp.item->key.obj_type != searchkey.obj_type) { 1452 ERR("could not find ROOT_ITEM for tree %llx\n", searchkey.obj_id); 1453 return STATUS_INTERNAL_ERROR; 1454 } 1455 1456 TRACE("updating the address for root %llx to %llx\n", searchkey.obj_id, t->new_address); 1457 1458 t->root->root_item.block_number = t->new_address; 1459 t->root->root_item.root_level = t->header.level; 1460 t->root->root_item.generation = Vcb->superblock.generation; 1461 t->root->root_item.generation2 = Vcb->superblock.generation; 1462 1463 // item is guaranteed to be at least sizeof(ROOT_ITEM), due to add_parents 1464 1465 RtlCopyMemory(tp.item->data, &t->root->root_item, sizeof(ROOT_ITEM)); 1466 } 1467 1468 t->root->treeholder.address = t->new_address; 1469 t->root->treeholder.generation = Vcb->superblock.generation; 1470 } 1471 1472 le = le->Flink; 1473 } 1474 1475 if (!no_cache && !(Vcb->superblock.compat_ro_flags & BTRFS_COMPAT_RO_FLAGS_FREE_SPACE_CACHE)) { 1476 ExAcquireResourceSharedLite(&Vcb->chunk_lock, TRUE); 1477 Status = update_chunk_caches(Vcb, Irp, rollback); 1478 ExReleaseResourceLite(&Vcb->chunk_lock); 1479 1480 if (!NT_SUCCESS(Status)) { 1481 ERR("update_chunk_caches returned %08x\n", Status); 1482 return Status; 1483 } 1484 } 1485 1486 return STATUS_SUCCESS; 1487 } 1488 1489 NTSTATUS do_tree_writes(device_extension* Vcb, LIST_ENTRY* tree_writes, BOOL no_free) { 1490 chunk* c; 1491 LIST_ENTRY* le; 1492 tree_write* tw; 1493 NTSTATUS Status; 1494 ULONG i, num_bits; 1495 write_data_context* wtc; 1496 ULONG bit_num = 0; 1497 BOOL raid56 = FALSE; 1498 1499 // merge together runs 1500 c = NULL; 1501 le = tree_writes->Flink; 1502 while (le != tree_writes) { 1503 tw = CONTAINING_RECORD(le, tree_write, list_entry); 1504 1505 if (!c || tw->address < c->offset || tw->address >= c->offset + c->chunk_item->size) 1506 c = get_chunk_from_address(Vcb, tw->address); 1507 else { 1508 tree_write* tw2 = CONTAINING_RECORD(le->Blink, tree_write, list_entry); 1509 1510 if (tw->address == tw2->address + tw2->length) { 1511 UINT8* data = ExAllocatePoolWithTag(NonPagedPool, tw2->length + tw->length, ALLOC_TAG); 1512 1513 if (!data) { 1514 ERR("out of memory\n"); 1515 return STATUS_INSUFFICIENT_RESOURCES; 1516 } 1517 1518 RtlCopyMemory(data, tw2->data, tw2->length); 1519 RtlCopyMemory(&data[tw2->length], tw->data, tw->length); 1520 1521 if (!no_free) 1522 ExFreePool(tw2->data); 1523 1524 tw2->data = data; 1525 tw2->length += tw->length; 1526 1527 if (!no_free) // FIXME - what if we allocated this just now? 1528 ExFreePool(tw->data); 1529 1530 RemoveEntryList(&tw->list_entry); 1531 ExFreePool(tw); 1532 1533 le = tw2->list_entry.Flink; 1534 continue; 1535 } 1536 } 1537 1538 tw->c = c; 1539 1540 if (c->chunk_item->type & (BLOCK_FLAG_RAID5 | BLOCK_FLAG_RAID6)) 1541 raid56 = TRUE; 1542 1543 le = le->Flink; 1544 } 1545 1546 num_bits = 0; 1547 1548 le = tree_writes->Flink; 1549 while (le != tree_writes) { 1550 tw = CONTAINING_RECORD(le, tree_write, list_entry); 1551 1552 num_bits++; 1553 1554 le = le->Flink; 1555 } 1556 1557 wtc = ExAllocatePoolWithTag(NonPagedPool, sizeof(write_data_context) * num_bits, ALLOC_TAG); 1558 if (!wtc) { 1559 ERR("out of memory\n"); 1560 return STATUS_INSUFFICIENT_RESOURCES; 1561 } 1562 1563 le = tree_writes->Flink; 1564 1565 while (le != tree_writes) { 1566 tw = CONTAINING_RECORD(le, tree_write, list_entry); 1567 1568 TRACE("address: %llx, size: %x\n", tw->address, tw->length); 1569 1570 KeInitializeEvent(&wtc[bit_num].Event, NotificationEvent, FALSE); 1571 InitializeListHead(&wtc[bit_num].stripes); 1572 wtc[bit_num].need_wait = FALSE; 1573 wtc[bit_num].stripes_left = 0; 1574 wtc[bit_num].parity1 = wtc[bit_num].parity2 = wtc[bit_num].scratch = NULL; 1575 wtc[bit_num].mdl = wtc[bit_num].parity1_mdl = wtc[bit_num].parity2_mdl = NULL; 1576 1577 Status = write_data(Vcb, tw->address, tw->data, tw->length, &wtc[bit_num], NULL, NULL, FALSE, 0, HighPagePriority); 1578 if (!NT_SUCCESS(Status)) { 1579 ERR("write_data returned %08x\n", Status); 1580 1581 for (i = 0; i < num_bits; i++) { 1582 free_write_data_stripes(&wtc[i]); 1583 } 1584 ExFreePool(wtc); 1585 1586 return Status; 1587 } 1588 1589 bit_num++; 1590 1591 le = le->Flink; 1592 } 1593 1594 for (i = 0; i < num_bits; i++) { 1595 if (wtc[i].stripes.Flink != &wtc[i].stripes) { 1596 // launch writes and wait 1597 le = wtc[i].stripes.Flink; 1598 while (le != &wtc[i].stripes) { 1599 write_data_stripe* stripe = CONTAINING_RECORD(le, write_data_stripe, list_entry); 1600 1601 if (stripe->status != WriteDataStatus_Ignore) { 1602 wtc[i].need_wait = TRUE; 1603 IoCallDriver(stripe->device->devobj, stripe->Irp); 1604 } 1605 1606 le = le->Flink; 1607 } 1608 } 1609 } 1610 1611 for (i = 0; i < num_bits; i++) { 1612 if (wtc[i].need_wait) 1613 KeWaitForSingleObject(&wtc[i].Event, Executive, KernelMode, FALSE, NULL); 1614 } 1615 1616 for (i = 0; i < num_bits; i++) { 1617 le = wtc[i].stripes.Flink; 1618 while (le != &wtc[i].stripes) { 1619 write_data_stripe* stripe = CONTAINING_RECORD(le, write_data_stripe, list_entry); 1620 1621 if (stripe->status != WriteDataStatus_Ignore && !NT_SUCCESS(stripe->iosb.Status)) { 1622 Status = stripe->iosb.Status; 1623 log_device_error(Vcb, stripe->device, BTRFS_DEV_STAT_WRITE_ERRORS); 1624 break; 1625 } 1626 1627 le = le->Flink; 1628 } 1629 1630 free_write_data_stripes(&wtc[i]); 1631 } 1632 1633 ExFreePool(wtc); 1634 1635 if (raid56) { 1636 c = NULL; 1637 1638 le = tree_writes->Flink; 1639 while (le != tree_writes) { 1640 tw = CONTAINING_RECORD(le, tree_write, list_entry); 1641 1642 if (tw->c != c) { 1643 c = tw->c; 1644 1645 ExAcquireResourceExclusiveLite(&c->partial_stripes_lock, TRUE); 1646 1647 while (!IsListEmpty(&c->partial_stripes)) { 1648 partial_stripe* ps = CONTAINING_RECORD(RemoveHeadList(&c->partial_stripes), partial_stripe, list_entry); 1649 1650 Status = flush_partial_stripe(Vcb, c, ps); 1651 1652 if (ps->bmparr) 1653 ExFreePool(ps->bmparr); 1654 1655 ExFreePool(ps); 1656 1657 if (!NT_SUCCESS(Status)) { 1658 ERR("flush_partial_stripe returned %08x\n", Status); 1659 ExReleaseResourceLite(&c->partial_stripes_lock); 1660 return Status; 1661 } 1662 } 1663 1664 ExReleaseResourceLite(&c->partial_stripes_lock); 1665 } 1666 1667 le = le->Flink; 1668 } 1669 } 1670 1671 return STATUS_SUCCESS; 1672 } 1673 1674 static NTSTATUS write_trees(device_extension* Vcb, PIRP Irp) { 1675 ULONG level; 1676 UINT8 *data, *body; 1677 UINT32 crc32; 1678 NTSTATUS Status; 1679 LIST_ENTRY* le; 1680 LIST_ENTRY tree_writes; 1681 tree_write* tw; 1682 1683 TRACE("(%p)\n", Vcb); 1684 1685 InitializeListHead(&tree_writes); 1686 1687 for (level = 0; level <= 255; level++) { 1688 BOOL nothing_found = TRUE; 1689 1690 TRACE("level = %u\n", level); 1691 1692 le = Vcb->trees.Flink; 1693 while (le != &Vcb->trees) { 1694 tree* t = CONTAINING_RECORD(le, tree, list_entry); 1695 1696 if (t->write && t->header.level == level) { 1697 KEY firstitem, searchkey; 1698 LIST_ENTRY* le2; 1699 traverse_ptr tp; 1700 1701 if (!t->has_new_address) { 1702 ERR("error - tried to write tree with no new address\n"); 1703 return STATUS_INTERNAL_ERROR; 1704 } 1705 1706 le2 = t->itemlist.Flink; 1707 while (le2 != &t->itemlist) { 1708 tree_data* td = CONTAINING_RECORD(le2, tree_data, list_entry); 1709 if (!td->ignore) { 1710 firstitem = td->key; 1711 break; 1712 } 1713 le2 = le2->Flink; 1714 } 1715 1716 if (t->parent) { 1717 t->paritem->key = firstitem; 1718 t->paritem->treeholder.address = t->new_address; 1719 t->paritem->treeholder.generation = Vcb->superblock.generation; 1720 } 1721 1722 if (!(Vcb->superblock.incompat_flags & BTRFS_INCOMPAT_FLAGS_SKINNY_METADATA)) { 1723 EXTENT_ITEM_TREE* eit; 1724 1725 searchkey.obj_id = t->new_address; 1726 searchkey.obj_type = TYPE_EXTENT_ITEM; 1727 searchkey.offset = Vcb->superblock.node_size; 1728 1729 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 1730 if (!NT_SUCCESS(Status)) { 1731 ERR("error - find_item returned %08x\n", Status); 1732 return Status; 1733 } 1734 1735 if (keycmp(searchkey, tp.item->key)) { 1736 ERR("could not find %llx,%x,%llx in extent_root (found %llx,%x,%llx instead)\n", searchkey.obj_id, searchkey.obj_type, searchkey.offset, tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset); 1737 return STATUS_INTERNAL_ERROR; 1738 } 1739 1740 if (tp.item->size < sizeof(EXTENT_ITEM_TREE)) { 1741 ERR("(%llx,%x,%llx) was %u bytes, expected at least %u\n", tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, tp.item->size, sizeof(EXTENT_ITEM_TREE)); 1742 return STATUS_INTERNAL_ERROR; 1743 } 1744 1745 eit = (EXTENT_ITEM_TREE*)tp.item->data; 1746 eit->firstitem = firstitem; 1747 } 1748 1749 nothing_found = FALSE; 1750 } 1751 1752 le = le->Flink; 1753 } 1754 1755 if (nothing_found) 1756 break; 1757 } 1758 1759 TRACE("allocated tree extents\n"); 1760 1761 le = Vcb->trees.Flink; 1762 while (le != &Vcb->trees) { 1763 tree* t = CONTAINING_RECORD(le, tree, list_entry); 1764 LIST_ENTRY* le2; 1765 #ifdef DEBUG_PARANOID 1766 UINT32 num_items = 0, size = 0; 1767 BOOL crash = FALSE; 1768 #endif 1769 1770 if (t->write) { 1771 #ifdef DEBUG_PARANOID 1772 BOOL first = TRUE; 1773 KEY lastkey; 1774 1775 le2 = t->itemlist.Flink; 1776 while (le2 != &t->itemlist) { 1777 tree_data* td = CONTAINING_RECORD(le2, tree_data, list_entry); 1778 if (!td->ignore) { 1779 num_items++; 1780 1781 if (!first) { 1782 if (keycmp(td->key, lastkey) == 0) { 1783 ERR("(%llx,%x,%llx): duplicate key\n", td->key.obj_id, td->key.obj_type, td->key.offset); 1784 crash = TRUE; 1785 } else if (keycmp(td->key, lastkey) == -1) { 1786 ERR("(%llx,%x,%llx): key out of order\n", td->key.obj_id, td->key.obj_type, td->key.offset); 1787 crash = TRUE; 1788 } 1789 } else 1790 first = FALSE; 1791 1792 lastkey = td->key; 1793 1794 if (t->header.level == 0) 1795 size += td->size; 1796 } 1797 le2 = le2->Flink; 1798 } 1799 1800 if (t->header.level == 0) 1801 size += num_items * sizeof(leaf_node); 1802 else 1803 size += num_items * sizeof(internal_node); 1804 1805 if (num_items != t->header.num_items) { 1806 ERR("tree %llx, level %x: num_items was %x, expected %x\n", t->root->id, t->header.level, num_items, t->header.num_items); 1807 crash = TRUE; 1808 } 1809 1810 if (size != t->size) { 1811 ERR("tree %llx, level %x: size was %x, expected %x\n", t->root->id, t->header.level, size, t->size); 1812 crash = TRUE; 1813 } 1814 1815 if (t->header.num_items == 0 && t->parent) { 1816 ERR("tree %llx, level %x: tried to write empty tree with parent\n", t->root->id, t->header.level); 1817 crash = TRUE; 1818 } 1819 1820 if (t->size > Vcb->superblock.node_size - sizeof(tree_header)) { 1821 ERR("tree %llx, level %x: tried to write overlarge tree (%x > %x)\n", t->root->id, t->header.level, t->size, Vcb->superblock.node_size - sizeof(tree_header)); 1822 crash = TRUE; 1823 } 1824 1825 if (crash) { 1826 ERR("tree %p\n", t); 1827 le2 = t->itemlist.Flink; 1828 while (le2 != &t->itemlist) { 1829 tree_data* td = CONTAINING_RECORD(le2, tree_data, list_entry); 1830 if (!td->ignore) { 1831 ERR("%llx,%x,%llx inserted=%u\n", td->key.obj_id, td->key.obj_type, td->key.offset, td->inserted); 1832 } 1833 le2 = le2->Flink; 1834 } 1835 int3; 1836 } 1837 #endif 1838 t->header.address = t->new_address; 1839 t->header.generation = Vcb->superblock.generation; 1840 t->header.tree_id = t->root->id; 1841 t->header.flags |= HEADER_FLAG_MIXED_BACKREF; 1842 t->header.fs_uuid = Vcb->superblock.uuid; 1843 t->has_address = TRUE; 1844 1845 data = ExAllocatePoolWithTag(NonPagedPool, Vcb->superblock.node_size, ALLOC_TAG); 1846 if (!data) { 1847 ERR("out of memory\n"); 1848 Status = STATUS_INSUFFICIENT_RESOURCES; 1849 goto end; 1850 } 1851 1852 body = data + sizeof(tree_header); 1853 1854 RtlCopyMemory(data, &t->header, sizeof(tree_header)); 1855 RtlZeroMemory(body, Vcb->superblock.node_size - sizeof(tree_header)); 1856 1857 if (t->header.level == 0) { 1858 leaf_node* itemptr = (leaf_node*)body; 1859 int i = 0; 1860 UINT8* dataptr = data + Vcb->superblock.node_size; 1861 1862 le2 = t->itemlist.Flink; 1863 while (le2 != &t->itemlist) { 1864 tree_data* td = CONTAINING_RECORD(le2, tree_data, list_entry); 1865 if (!td->ignore) { 1866 dataptr = dataptr - td->size; 1867 1868 itemptr[i].key = td->key; 1869 itemptr[i].offset = (UINT32)((UINT8*)dataptr - (UINT8*)body); 1870 itemptr[i].size = td->size; 1871 i++; 1872 1873 if (td->size > 0) 1874 RtlCopyMemory(dataptr, td->data, td->size); 1875 } 1876 1877 le2 = le2->Flink; 1878 } 1879 } else { 1880 internal_node* itemptr = (internal_node*)body; 1881 int i = 0; 1882 1883 le2 = t->itemlist.Flink; 1884 while (le2 != &t->itemlist) { 1885 tree_data* td = CONTAINING_RECORD(le2, tree_data, list_entry); 1886 if (!td->ignore) { 1887 itemptr[i].key = td->key; 1888 itemptr[i].address = td->treeholder.address; 1889 itemptr[i].generation = td->treeholder.generation; 1890 i++; 1891 } 1892 1893 le2 = le2->Flink; 1894 } 1895 } 1896 1897 crc32 = calc_crc32c(0xffffffff, (UINT8*)&((tree_header*)data)->fs_uuid, Vcb->superblock.node_size - sizeof(((tree_header*)data)->csum)); 1898 crc32 = ~crc32; 1899 *((UINT32*)data) = crc32; 1900 TRACE("setting crc32 to %08x\n", crc32); 1901 1902 tw = ExAllocatePoolWithTag(PagedPool, sizeof(tree_write), ALLOC_TAG); 1903 if (!tw) { 1904 ERR("out of memory\n"); 1905 ExFreePool(data); 1906 Status = STATUS_INSUFFICIENT_RESOURCES; 1907 goto end; 1908 } 1909 1910 tw->address = t->new_address; 1911 tw->length = Vcb->superblock.node_size; 1912 tw->data = data; 1913 1914 if (IsListEmpty(&tree_writes)) 1915 InsertTailList(&tree_writes, &tw->list_entry); 1916 else { 1917 BOOL inserted = FALSE; 1918 1919 le2 = tree_writes.Flink; 1920 while (le2 != &tree_writes) { 1921 tree_write* tw2 = CONTAINING_RECORD(le2, tree_write, list_entry); 1922 1923 if (tw2->address > tw->address) { 1924 InsertHeadList(le2->Blink, &tw->list_entry); 1925 inserted = TRUE; 1926 break; 1927 } 1928 1929 le2 = le2->Flink; 1930 } 1931 1932 if (!inserted) 1933 InsertTailList(&tree_writes, &tw->list_entry); 1934 } 1935 } 1936 1937 le = le->Flink; 1938 } 1939 1940 Status = do_tree_writes(Vcb, &tree_writes, FALSE); 1941 if (!NT_SUCCESS(Status)) { 1942 ERR("do_tree_writes returned %08x\n", Status); 1943 goto end; 1944 } 1945 1946 Status = STATUS_SUCCESS; 1947 1948 end: 1949 while (!IsListEmpty(&tree_writes)) { 1950 le = RemoveHeadList(&tree_writes); 1951 tw = CONTAINING_RECORD(le, tree_write, list_entry); 1952 1953 if (tw->data) 1954 ExFreePool(tw->data); 1955 1956 ExFreePool(tw); 1957 } 1958 1959 return Status; 1960 } 1961 1962 static void update_backup_superblock(device_extension* Vcb, superblock_backup* sb, PIRP Irp) { 1963 KEY searchkey; 1964 traverse_ptr tp; 1965 1966 RtlZeroMemory(sb, sizeof(superblock_backup)); 1967 1968 sb->root_tree_addr = Vcb->superblock.root_tree_addr; 1969 sb->root_tree_generation = Vcb->superblock.generation; 1970 sb->root_level = Vcb->superblock.root_level; 1971 1972 sb->chunk_tree_addr = Vcb->superblock.chunk_tree_addr; 1973 sb->chunk_tree_generation = Vcb->superblock.chunk_root_generation; 1974 sb->chunk_root_level = Vcb->superblock.chunk_root_level; 1975 1976 searchkey.obj_id = BTRFS_ROOT_EXTENT; 1977 searchkey.obj_type = TYPE_ROOT_ITEM; 1978 searchkey.offset = 0xffffffffffffffff; 1979 1980 if (NT_SUCCESS(find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp))) { 1981 if (tp.item->key.obj_id == searchkey.obj_id && tp.item->key.obj_type == searchkey.obj_type && tp.item->size >= sizeof(ROOT_ITEM)) { 1982 ROOT_ITEM* ri = (ROOT_ITEM*)tp.item->data; 1983 1984 sb->extent_tree_addr = ri->block_number; 1985 sb->extent_tree_generation = ri->generation; 1986 sb->extent_root_level = ri->root_level; 1987 } 1988 } 1989 1990 searchkey.obj_id = BTRFS_ROOT_FSTREE; 1991 1992 if (NT_SUCCESS(find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp))) { 1993 if (tp.item->key.obj_id == searchkey.obj_id && tp.item->key.obj_type == searchkey.obj_type && tp.item->size >= sizeof(ROOT_ITEM)) { 1994 ROOT_ITEM* ri = (ROOT_ITEM*)tp.item->data; 1995 1996 sb->fs_tree_addr = ri->block_number; 1997 sb->fs_tree_generation = ri->generation; 1998 sb->fs_root_level = ri->root_level; 1999 } 2000 } 2001 2002 searchkey.obj_id = BTRFS_ROOT_DEVTREE; 2003 2004 if (NT_SUCCESS(find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp))) { 2005 if (tp.item->key.obj_id == searchkey.obj_id && tp.item->key.obj_type == searchkey.obj_type && tp.item->size >= sizeof(ROOT_ITEM)) { 2006 ROOT_ITEM* ri = (ROOT_ITEM*)tp.item->data; 2007 2008 sb->dev_root_addr = ri->block_number; 2009 sb->dev_root_generation = ri->generation; 2010 sb->dev_root_level = ri->root_level; 2011 } 2012 } 2013 2014 searchkey.obj_id = BTRFS_ROOT_CHECKSUM; 2015 2016 if (NT_SUCCESS(find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp))) { 2017 if (tp.item->key.obj_id == searchkey.obj_id && tp.item->key.obj_type == searchkey.obj_type && tp.item->size >= sizeof(ROOT_ITEM)) { 2018 ROOT_ITEM* ri = (ROOT_ITEM*)tp.item->data; 2019 2020 sb->csum_root_addr = ri->block_number; 2021 sb->csum_root_generation = ri->generation; 2022 sb->csum_root_level = ri->root_level; 2023 } 2024 } 2025 2026 sb->total_bytes = Vcb->superblock.total_bytes; 2027 sb->bytes_used = Vcb->superblock.bytes_used; 2028 sb->num_devices = Vcb->superblock.num_devices; 2029 } 2030 2031 typedef struct { 2032 void* context; 2033 UINT8* buf; 2034 PMDL mdl; 2035 device* device; 2036 NTSTATUS Status; 2037 PIRP Irp; 2038 LIST_ENTRY list_entry; 2039 } write_superblocks_stripe; 2040 2041 typedef struct _write_superblocks_context { 2042 KEVENT Event; 2043 LIST_ENTRY stripes; 2044 LONG left; 2045 } write_superblocks_context; 2046 2047 _Function_class_(IO_COMPLETION_ROUTINE) 2048 #ifdef __REACTOS__ 2049 static NTSTATUS NTAPI write_superblock_completion(PDEVICE_OBJECT DeviceObject, PIRP Irp, PVOID conptr) { 2050 #else 2051 static NTSTATUS write_superblock_completion(PDEVICE_OBJECT DeviceObject, PIRP Irp, PVOID conptr) { 2052 #endif 2053 write_superblocks_stripe* stripe = conptr; 2054 write_superblocks_context* context = stripe->context; 2055 2056 UNUSED(DeviceObject); 2057 2058 stripe->Status = Irp->IoStatus.Status; 2059 2060 if (InterlockedDecrement(&context->left) == 0) 2061 KeSetEvent(&context->Event, 0, FALSE); 2062 2063 return STATUS_MORE_PROCESSING_REQUIRED; 2064 } 2065 2066 static NTSTATUS write_superblock(device_extension* Vcb, device* device, write_superblocks_context* context) { 2067 unsigned int i = 0; 2068 2069 // All the documentation says that the Linux driver only writes one superblock 2070 // if it thinks a disk is an SSD, but this doesn't seem to be the case! 2071 2072 while (superblock_addrs[i] > 0 && device->devitem.num_bytes >= superblock_addrs[i] + sizeof(superblock)) { 2073 ULONG sblen = (ULONG)sector_align(sizeof(superblock), Vcb->superblock.sector_size); 2074 superblock* sb; 2075 UINT32 crc32; 2076 write_superblocks_stripe* stripe; 2077 PIO_STACK_LOCATION IrpSp; 2078 2079 sb = ExAllocatePoolWithTag(NonPagedPool, sblen, ALLOC_TAG); 2080 if (!sb) { 2081 ERR("out of memory\n"); 2082 return STATUS_INSUFFICIENT_RESOURCES; 2083 } 2084 2085 RtlCopyMemory(sb, &Vcb->superblock, sizeof(superblock)); 2086 2087 if (sblen > sizeof(superblock)) 2088 RtlZeroMemory((UINT8*)sb + sizeof(superblock), sblen - sizeof(superblock)); 2089 2090 RtlCopyMemory(&sb->dev_item, &device->devitem, sizeof(DEV_ITEM)); 2091 sb->sb_phys_addr = superblock_addrs[i]; 2092 2093 crc32 = ~calc_crc32c(0xffffffff, (UINT8*)&sb->uuid, (ULONG)sizeof(superblock) - sizeof(sb->checksum)); 2094 RtlCopyMemory(&sb->checksum, &crc32, sizeof(UINT32)); 2095 2096 stripe = ExAllocatePoolWithTag(NonPagedPool, sizeof(write_superblocks_stripe), ALLOC_TAG); 2097 if (!stripe) { 2098 ERR("out of memory\n"); 2099 ExFreePool(sb); 2100 return STATUS_INSUFFICIENT_RESOURCES; 2101 } 2102 2103 stripe->buf = (UINT8*)sb; 2104 2105 stripe->Irp = IoAllocateIrp(device->devobj->StackSize, FALSE); 2106 if (!stripe->Irp) { 2107 ERR("IoAllocateIrp failed\n"); 2108 ExFreePool(stripe); 2109 ExFreePool(sb); 2110 return STATUS_INSUFFICIENT_RESOURCES; 2111 } 2112 2113 IrpSp = IoGetNextIrpStackLocation(stripe->Irp); 2114 IrpSp->MajorFunction = IRP_MJ_WRITE; 2115 2116 if (i == 0) 2117 IrpSp->Flags |= SL_WRITE_THROUGH; 2118 2119 if (device->devobj->Flags & DO_BUFFERED_IO) { 2120 stripe->Irp->AssociatedIrp.SystemBuffer = sb; 2121 stripe->mdl = NULL; 2122 2123 stripe->Irp->Flags = IRP_BUFFERED_IO; 2124 } else if (device->devobj->Flags & DO_DIRECT_IO) { 2125 stripe->mdl = IoAllocateMdl(sb, sblen, FALSE, FALSE, NULL); 2126 if (!stripe->mdl) { 2127 ERR("IoAllocateMdl failed\n"); 2128 IoFreeIrp(stripe->Irp); 2129 ExFreePool(stripe); 2130 ExFreePool(sb); 2131 return STATUS_INSUFFICIENT_RESOURCES; 2132 } 2133 2134 stripe->Irp->MdlAddress = stripe->mdl; 2135 2136 MmBuildMdlForNonPagedPool(stripe->mdl); 2137 } else { 2138 stripe->Irp->UserBuffer = sb; 2139 stripe->mdl = NULL; 2140 } 2141 2142 IrpSp->Parameters.Write.Length = sblen; 2143 IrpSp->Parameters.Write.ByteOffset.QuadPart = superblock_addrs[i]; 2144 2145 IoSetCompletionRoutine(stripe->Irp, write_superblock_completion, stripe, TRUE, TRUE, TRUE); 2146 2147 stripe->context = context; 2148 stripe->device = device; 2149 InsertTailList(&context->stripes, &stripe->list_entry); 2150 2151 context->left++; 2152 2153 i++; 2154 } 2155 2156 if (i == 0) 2157 ERR("no superblocks written!\n"); 2158 2159 return STATUS_SUCCESS; 2160 } 2161 2162 static NTSTATUS write_superblocks(device_extension* Vcb, PIRP Irp) { 2163 UINT64 i; 2164 NTSTATUS Status; 2165 LIST_ENTRY* le; 2166 write_superblocks_context context; 2167 2168 TRACE("(%p)\n", Vcb); 2169 2170 le = Vcb->trees.Flink; 2171 while (le != &Vcb->trees) { 2172 tree* t = CONTAINING_RECORD(le, tree, list_entry); 2173 2174 if (t->write && !t->parent) { 2175 if (t->root == Vcb->root_root) { 2176 Vcb->superblock.root_tree_addr = t->new_address; 2177 Vcb->superblock.root_level = t->header.level; 2178 } else if (t->root == Vcb->chunk_root) { 2179 Vcb->superblock.chunk_tree_addr = t->new_address; 2180 Vcb->superblock.chunk_root_generation = t->header.generation; 2181 Vcb->superblock.chunk_root_level = t->header.level; 2182 } 2183 } 2184 2185 le = le->Flink; 2186 } 2187 2188 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS - 1; i++) { 2189 RtlCopyMemory(&Vcb->superblock.backup[i], &Vcb->superblock.backup[i+1], sizeof(superblock_backup)); 2190 } 2191 2192 update_backup_superblock(Vcb, &Vcb->superblock.backup[BTRFS_NUM_BACKUP_ROOTS - 1], Irp); 2193 2194 KeInitializeEvent(&context.Event, NotificationEvent, FALSE); 2195 InitializeListHead(&context.stripes); 2196 context.left = 0; 2197 2198 le = Vcb->devices.Flink; 2199 while (le != &Vcb->devices) { 2200 device* dev = CONTAINING_RECORD(le, device, list_entry); 2201 2202 if (dev->devobj && !dev->readonly) { 2203 Status = write_superblock(Vcb, dev, &context); 2204 if (!NT_SUCCESS(Status)) { 2205 ERR("write_superblock returned %08x\n", Status); 2206 goto end; 2207 } 2208 } 2209 2210 le = le->Flink; 2211 } 2212 2213 if (IsListEmpty(&context.stripes)) { 2214 ERR("error - not writing any superblocks\n"); 2215 Status = STATUS_INTERNAL_ERROR; 2216 goto end; 2217 } 2218 2219 le = context.stripes.Flink; 2220 while (le != &context.stripes) { 2221 write_superblocks_stripe* stripe = CONTAINING_RECORD(le, write_superblocks_stripe, list_entry); 2222 2223 IoCallDriver(stripe->device->devobj, stripe->Irp); 2224 2225 le = le->Flink; 2226 } 2227 2228 KeWaitForSingleObject(&context.Event, Executive, KernelMode, FALSE, NULL); 2229 2230 le = context.stripes.Flink; 2231 while (le != &context.stripes) { 2232 write_superblocks_stripe* stripe = CONTAINING_RECORD(le, write_superblocks_stripe, list_entry); 2233 2234 if (!NT_SUCCESS(stripe->Status)) { 2235 ERR("device %llx returned %08x\n", stripe->device->devitem.dev_id, stripe->Status); 2236 log_device_error(Vcb, stripe->device, BTRFS_DEV_STAT_WRITE_ERRORS); 2237 Status = stripe->Status; 2238 goto end; 2239 } 2240 2241 le = le->Flink; 2242 } 2243 2244 Status = STATUS_SUCCESS; 2245 2246 end: 2247 while (!IsListEmpty(&context.stripes)) { 2248 write_superblocks_stripe* stripe = CONTAINING_RECORD(RemoveHeadList(&context.stripes), write_superblocks_stripe, list_entry); 2249 2250 if (stripe->mdl) { 2251 if (stripe->mdl->MdlFlags & MDL_PAGES_LOCKED) 2252 MmUnlockPages(stripe->mdl); 2253 2254 IoFreeMdl(stripe->mdl); 2255 } 2256 2257 if (stripe->Irp) 2258 IoFreeIrp(stripe->Irp); 2259 2260 if (stripe->buf) 2261 ExFreePool(stripe->buf); 2262 2263 ExFreePool(stripe); 2264 } 2265 2266 return Status; 2267 } 2268 2269 static NTSTATUS flush_changed_extent(device_extension* Vcb, chunk* c, changed_extent* ce, PIRP Irp, LIST_ENTRY* rollback) { 2270 LIST_ENTRY *le, *le2; 2271 NTSTATUS Status; 2272 UINT64 old_size; 2273 2274 if (ce->count == 0 && ce->old_count == 0) { 2275 while (!IsListEmpty(&ce->refs)) { 2276 changed_extent_ref* cer = CONTAINING_RECORD(RemoveHeadList(&ce->refs), changed_extent_ref, list_entry); 2277 ExFreePool(cer); 2278 } 2279 2280 while (!IsListEmpty(&ce->old_refs)) { 2281 changed_extent_ref* cer = CONTAINING_RECORD(RemoveHeadList(&ce->old_refs), changed_extent_ref, list_entry); 2282 ExFreePool(cer); 2283 } 2284 2285 goto end; 2286 } 2287 2288 le = ce->refs.Flink; 2289 while (le != &ce->refs) { 2290 changed_extent_ref* cer = CONTAINING_RECORD(le, changed_extent_ref, list_entry); 2291 UINT32 old_count = 0; 2292 2293 if (cer->type == TYPE_EXTENT_DATA_REF) { 2294 le2 = ce->old_refs.Flink; 2295 while (le2 != &ce->old_refs) { 2296 changed_extent_ref* cer2 = CONTAINING_RECORD(le2, changed_extent_ref, list_entry); 2297 2298 if (cer2->type == TYPE_EXTENT_DATA_REF && cer2->edr.root == cer->edr.root && cer2->edr.objid == cer->edr.objid && cer2->edr.offset == cer->edr.offset) { 2299 old_count = cer2->edr.count; 2300 break; 2301 } 2302 2303 le2 = le2->Flink; 2304 } 2305 2306 old_size = ce->old_count > 0 ? ce->old_size : ce->size; 2307 2308 if (cer->edr.count > old_count) { 2309 Status = increase_extent_refcount_data(Vcb, ce->address, old_size, cer->edr.root, cer->edr.objid, cer->edr.offset, cer->edr.count - old_count, Irp); 2310 2311 if (!NT_SUCCESS(Status)) { 2312 ERR("increase_extent_refcount_data returned %08x\n", Status); 2313 return Status; 2314 } 2315 } 2316 } else if (cer->type == TYPE_SHARED_DATA_REF) { 2317 le2 = ce->old_refs.Flink; 2318 while (le2 != &ce->old_refs) { 2319 changed_extent_ref* cer2 = CONTAINING_RECORD(le2, changed_extent_ref, list_entry); 2320 2321 if (cer2->type == TYPE_SHARED_DATA_REF && cer2->sdr.offset == cer->sdr.offset) { 2322 RemoveEntryList(&cer2->list_entry); 2323 ExFreePool(cer2); 2324 break; 2325 } 2326 2327 le2 = le2->Flink; 2328 } 2329 } 2330 2331 le = le->Flink; 2332 } 2333 2334 le = ce->refs.Flink; 2335 while (le != &ce->refs) { 2336 changed_extent_ref* cer = CONTAINING_RECORD(le, changed_extent_ref, list_entry); 2337 LIST_ENTRY* le3 = le->Flink; 2338 UINT32 old_count = 0; 2339 2340 if (cer->type == TYPE_EXTENT_DATA_REF) { 2341 le2 = ce->old_refs.Flink; 2342 while (le2 != &ce->old_refs) { 2343 changed_extent_ref* cer2 = CONTAINING_RECORD(le2, changed_extent_ref, list_entry); 2344 2345 if (cer2->type == TYPE_EXTENT_DATA_REF && cer2->edr.root == cer->edr.root && cer2->edr.objid == cer->edr.objid && cer2->edr.offset == cer->edr.offset) { 2346 old_count = cer2->edr.count; 2347 2348 RemoveEntryList(&cer2->list_entry); 2349 ExFreePool(cer2); 2350 break; 2351 } 2352 2353 le2 = le2->Flink; 2354 } 2355 2356 old_size = ce->old_count > 0 ? ce->old_size : ce->size; 2357 2358 if (cer->edr.count < old_count) { 2359 Status = decrease_extent_refcount_data(Vcb, ce->address, old_size, cer->edr.root, cer->edr.objid, cer->edr.offset, 2360 old_count - cer->edr.count, ce->superseded, Irp); 2361 2362 if (!NT_SUCCESS(Status)) { 2363 ERR("decrease_extent_refcount_data returned %08x\n", Status); 2364 return Status; 2365 } 2366 } 2367 2368 if (ce->size != ce->old_size && ce->old_count > 0) { 2369 KEY searchkey; 2370 traverse_ptr tp; 2371 void* data; 2372 2373 searchkey.obj_id = ce->address; 2374 searchkey.obj_type = TYPE_EXTENT_ITEM; 2375 searchkey.offset = ce->old_size; 2376 2377 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 2378 if (!NT_SUCCESS(Status)) { 2379 ERR("error - find_item returned %08x\n", Status); 2380 return Status; 2381 } 2382 2383 if (keycmp(searchkey, tp.item->key)) { 2384 ERR("could not find (%llx,%x,%llx) in extent tree\n", searchkey.obj_id, searchkey.obj_type, searchkey.offset); 2385 return STATUS_INTERNAL_ERROR; 2386 } 2387 2388 if (tp.item->size > 0) { 2389 data = ExAllocatePoolWithTag(PagedPool, tp.item->size, ALLOC_TAG); 2390 2391 if (!data) { 2392 ERR("out of memory\n"); 2393 return STATUS_INSUFFICIENT_RESOURCES; 2394 } 2395 2396 RtlCopyMemory(data, tp.item->data, tp.item->size); 2397 } else 2398 data = NULL; 2399 2400 Status = insert_tree_item(Vcb, Vcb->extent_root, ce->address, TYPE_EXTENT_ITEM, ce->size, data, tp.item->size, NULL, Irp); 2401 if (!NT_SUCCESS(Status)) { 2402 ERR("insert_tree_item returned %08x\n", Status); 2403 if (data) ExFreePool(data); 2404 return Status; 2405 } 2406 2407 Status = delete_tree_item(Vcb, &tp); 2408 if (!NT_SUCCESS(Status)) { 2409 ERR("delete_tree_item returned %08x\n", Status); 2410 return Status; 2411 } 2412 } 2413 } 2414 2415 RemoveEntryList(&cer->list_entry); 2416 ExFreePool(cer); 2417 2418 le = le3; 2419 } 2420 2421 #ifdef DEBUG_PARANOID 2422 if (!IsListEmpty(&ce->old_refs)) 2423 WARN("old_refs not empty\n"); 2424 #endif 2425 2426 end: 2427 if (ce->count == 0 && !ce->superseded) { 2428 c->used -= ce->size; 2429 space_list_add(c, ce->address, ce->size, rollback); 2430 } 2431 2432 RemoveEntryList(&ce->list_entry); 2433 ExFreePool(ce); 2434 2435 return STATUS_SUCCESS; 2436 } 2437 2438 void add_checksum_entry(device_extension* Vcb, UINT64 address, ULONG length, UINT32* csum, PIRP Irp) { 2439 KEY searchkey; 2440 traverse_ptr tp, next_tp; 2441 NTSTATUS Status; 2442 UINT64 startaddr, endaddr; 2443 ULONG len; 2444 UINT32* checksums; 2445 RTL_BITMAP bmp; 2446 ULONG* bmparr; 2447 ULONG runlength, index; 2448 2449 searchkey.obj_id = EXTENT_CSUM_ID; 2450 searchkey.obj_type = TYPE_EXTENT_CSUM; 2451 searchkey.offset = address; 2452 2453 // FIXME - create checksum_root if it doesn't exist at all 2454 2455 Status = find_item(Vcb, Vcb->checksum_root, &tp, &searchkey, FALSE, Irp); 2456 if (Status == STATUS_NOT_FOUND) { // tree is completely empty 2457 if (csum) { // not deleted 2458 ULONG length2 = length; 2459 UINT64 off = address; 2460 UINT32* data = csum; 2461 2462 do { 2463 UINT16 il = (UINT16)min(length2, MAX_CSUM_SIZE / sizeof(UINT32)); 2464 2465 checksums = ExAllocatePoolWithTag(PagedPool, il * sizeof(UINT32), ALLOC_TAG); 2466 if (!checksums) { 2467 ERR("out of memory\n"); 2468 return; 2469 } 2470 2471 RtlCopyMemory(checksums, data, il * sizeof(UINT32)); 2472 2473 Status = insert_tree_item(Vcb, Vcb->checksum_root, EXTENT_CSUM_ID, TYPE_EXTENT_CSUM, off, checksums, 2474 il * sizeof(UINT32), NULL, Irp); 2475 if (!NT_SUCCESS(Status)) { 2476 ERR("insert_tree_item returned %08x\n", Status); 2477 ExFreePool(checksums); 2478 return; 2479 } 2480 2481 length2 -= il; 2482 2483 if (length2 > 0) { 2484 off += il * Vcb->superblock.sector_size; 2485 data += il; 2486 } 2487 } while (length2 > 0); 2488 } 2489 } else if (!NT_SUCCESS(Status)) { 2490 ERR("find_item returned %08x\n", Status); 2491 return; 2492 } else { 2493 UINT32 tplen; 2494 2495 // FIXME - check entry is TYPE_EXTENT_CSUM? 2496 2497 if (tp.item->key.offset < address && tp.item->key.offset + (tp.item->size * Vcb->superblock.sector_size / sizeof(UINT32)) >= address) 2498 startaddr = tp.item->key.offset; 2499 else 2500 startaddr = address; 2501 2502 searchkey.obj_id = EXTENT_CSUM_ID; 2503 searchkey.obj_type = TYPE_EXTENT_CSUM; 2504 searchkey.offset = address + (length * Vcb->superblock.sector_size); 2505 2506 Status = find_item(Vcb, Vcb->checksum_root, &tp, &searchkey, FALSE, Irp); 2507 if (!NT_SUCCESS(Status)) { 2508 ERR("find_item returned %08x\n", Status); 2509 return; 2510 } 2511 2512 tplen = tp.item->size / sizeof(UINT32); 2513 2514 if (tp.item->key.offset + (tplen * Vcb->superblock.sector_size) >= address + (length * Vcb->superblock.sector_size)) 2515 endaddr = tp.item->key.offset + (tplen * Vcb->superblock.sector_size); 2516 else 2517 endaddr = address + (length * Vcb->superblock.sector_size); 2518 2519 TRACE("cs starts at %llx (%x sectors)\n", address, length); 2520 TRACE("startaddr = %llx\n", startaddr); 2521 TRACE("endaddr = %llx\n", endaddr); 2522 2523 len = (ULONG)((endaddr - startaddr) / Vcb->superblock.sector_size); 2524 2525 checksums = ExAllocatePoolWithTag(PagedPool, sizeof(UINT32) * len, ALLOC_TAG); 2526 if (!checksums) { 2527 ERR("out of memory\n"); 2528 return; 2529 } 2530 2531 bmparr = ExAllocatePoolWithTag(PagedPool, sizeof(ULONG) * ((len/8)+1), ALLOC_TAG); 2532 if (!bmparr) { 2533 ERR("out of memory\n"); 2534 ExFreePool(checksums); 2535 return; 2536 } 2537 2538 RtlInitializeBitMap(&bmp, bmparr, len); 2539 RtlSetAllBits(&bmp); 2540 2541 searchkey.obj_id = EXTENT_CSUM_ID; 2542 searchkey.obj_type = TYPE_EXTENT_CSUM; 2543 searchkey.offset = address; 2544 2545 Status = find_item(Vcb, Vcb->checksum_root, &tp, &searchkey, FALSE, Irp); 2546 if (!NT_SUCCESS(Status)) { 2547 ERR("find_item returned %08x\n", Status); 2548 ExFreePool(checksums); 2549 ExFreePool(bmparr); 2550 return; 2551 } 2552 2553 // set bit = free space, cleared bit = allocated sector 2554 2555 while (tp.item->key.offset < endaddr) { 2556 if (tp.item->key.offset >= startaddr) { 2557 if (tp.item->size > 0) { 2558 ULONG itemlen = (ULONG)min((len - (tp.item->key.offset - startaddr) / Vcb->superblock.sector_size) * sizeof(UINT32), tp.item->size); 2559 2560 RtlCopyMemory(&checksums[(tp.item->key.offset - startaddr) / Vcb->superblock.sector_size], tp.item->data, itemlen); 2561 RtlClearBits(&bmp, (ULONG)((tp.item->key.offset - startaddr) / Vcb->superblock.sector_size), itemlen / sizeof(UINT32)); 2562 } 2563 2564 Status = delete_tree_item(Vcb, &tp); 2565 if (!NT_SUCCESS(Status)) { 2566 ERR("delete_tree_item returned %08x\n", Status); 2567 ExFreePool(checksums); 2568 ExFreePool(bmparr); 2569 return; 2570 } 2571 } 2572 2573 if (find_next_item(Vcb, &tp, &next_tp, FALSE, Irp)) { 2574 tp = next_tp; 2575 } else 2576 break; 2577 } 2578 2579 if (!csum) { // deleted 2580 RtlSetBits(&bmp, (ULONG)((address - startaddr) / Vcb->superblock.sector_size), length); 2581 } else { 2582 RtlCopyMemory(&checksums[(address - startaddr) / Vcb->superblock.sector_size], csum, length * sizeof(UINT32)); 2583 RtlClearBits(&bmp, (ULONG)((address - startaddr) / Vcb->superblock.sector_size), length); 2584 } 2585 2586 runlength = RtlFindFirstRunClear(&bmp, &index); 2587 2588 while (runlength != 0) { 2589 do { 2590 UINT16 rl; 2591 UINT64 off; 2592 UINT32* data; 2593 2594 if (runlength * sizeof(UINT32) > MAX_CSUM_SIZE) 2595 rl = MAX_CSUM_SIZE / sizeof(UINT32); 2596 else 2597 rl = (UINT16)runlength; 2598 2599 data = ExAllocatePoolWithTag(PagedPool, sizeof(UINT32) * rl, ALLOC_TAG); 2600 if (!data) { 2601 ERR("out of memory\n"); 2602 ExFreePool(bmparr); 2603 ExFreePool(checksums); 2604 return; 2605 } 2606 2607 RtlCopyMemory(data, &checksums[index], sizeof(UINT32) * rl); 2608 2609 off = startaddr + UInt32x32To64(index, Vcb->superblock.sector_size); 2610 2611 Status = insert_tree_item(Vcb, Vcb->checksum_root, EXTENT_CSUM_ID, TYPE_EXTENT_CSUM, off, data, sizeof(UINT32) * rl, NULL, Irp); 2612 if (!NT_SUCCESS(Status)) { 2613 ERR("insert_tree_item returned %08x\n", Status); 2614 ExFreePool(data); 2615 ExFreePool(bmparr); 2616 ExFreePool(checksums); 2617 return; 2618 } 2619 2620 runlength -= rl; 2621 index += rl; 2622 } while (runlength > 0); 2623 2624 runlength = RtlFindNextForwardRunClear(&bmp, index, &index); 2625 } 2626 2627 ExFreePool(bmparr); 2628 ExFreePool(checksums); 2629 } 2630 } 2631 2632 static NTSTATUS update_chunk_usage(device_extension* Vcb, PIRP Irp, LIST_ENTRY* rollback) { 2633 LIST_ENTRY *le = Vcb->chunks.Flink, *le2; 2634 chunk* c; 2635 KEY searchkey; 2636 traverse_ptr tp; 2637 BLOCK_GROUP_ITEM* bgi; 2638 NTSTATUS Status; 2639 2640 TRACE("(%p)\n", Vcb); 2641 2642 ExAcquireResourceSharedLite(&Vcb->chunk_lock, TRUE); 2643 2644 while (le != &Vcb->chunks) { 2645 c = CONTAINING_RECORD(le, chunk, list_entry); 2646 2647 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 2648 2649 if (!c->cache_loaded && (!IsListEmpty(&c->changed_extents) || c->used != c->oldused)) { 2650 Status = load_cache_chunk(Vcb, c, NULL); 2651 2652 if (!NT_SUCCESS(Status)) { 2653 ERR("load_cache_chunk returned %08x\n", Status); 2654 ExReleaseResourceLite(&c->lock); 2655 goto end; 2656 } 2657 } 2658 2659 le2 = c->changed_extents.Flink; 2660 while (le2 != &c->changed_extents) { 2661 LIST_ENTRY* le3 = le2->Flink; 2662 changed_extent* ce = CONTAINING_RECORD(le2, changed_extent, list_entry); 2663 2664 Status = flush_changed_extent(Vcb, c, ce, Irp, rollback); 2665 if (!NT_SUCCESS(Status)) { 2666 ERR("flush_changed_extent returned %08x\n", Status); 2667 ExReleaseResourceLite(&c->lock); 2668 goto end; 2669 } 2670 2671 le2 = le3; 2672 } 2673 2674 // This is usually done by update_chunks, but we have to check again in case any new chunks 2675 // have been allocated since. 2676 if (c->created) { 2677 Status = create_chunk(Vcb, c, Irp); 2678 if (!NT_SUCCESS(Status)) { 2679 ERR("create_chunk returned %08x\n", Status); 2680 ExReleaseResourceLite(&c->lock); 2681 goto end; 2682 } 2683 } 2684 2685 if (c->old_cache) { 2686 if (c->old_cache->dirty) { 2687 LIST_ENTRY batchlist; 2688 2689 InitializeListHead(&batchlist); 2690 2691 Status = flush_fcb(c->old_cache, FALSE, &batchlist, Irp); 2692 if (!NT_SUCCESS(Status)) { 2693 ERR("flush_fcb returned %08x\n", Status); 2694 ExReleaseResourceLite(&c->lock); 2695 clear_batch_list(Vcb, &batchlist); 2696 goto end; 2697 } 2698 2699 Status = commit_batch_list(Vcb, &batchlist, Irp); 2700 if (!NT_SUCCESS(Status)) { 2701 ERR("commit_batch_list returned %08x\n", Status); 2702 ExReleaseResourceLite(&c->lock); 2703 goto end; 2704 } 2705 } 2706 2707 free_fcb(Vcb, c->old_cache); 2708 c->old_cache = NULL; 2709 } 2710 2711 if (c->used != c->oldused) { 2712 searchkey.obj_id = c->offset; 2713 searchkey.obj_type = TYPE_BLOCK_GROUP_ITEM; 2714 searchkey.offset = c->chunk_item->size; 2715 2716 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 2717 if (!NT_SUCCESS(Status)) { 2718 ERR("error - find_item returned %08x\n", Status); 2719 ExReleaseResourceLite(&c->lock); 2720 goto end; 2721 } 2722 2723 if (keycmp(searchkey, tp.item->key)) { 2724 ERR("could not find (%llx,%x,%llx) in extent_root\n", searchkey.obj_id, searchkey.obj_type, searchkey.offset); 2725 Status = STATUS_INTERNAL_ERROR; 2726 ExReleaseResourceLite(&c->lock); 2727 goto end; 2728 } 2729 2730 if (tp.item->size < sizeof(BLOCK_GROUP_ITEM)) { 2731 ERR("(%llx,%x,%llx) was %u bytes, expected %u\n", tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, tp.item->size, sizeof(BLOCK_GROUP_ITEM)); 2732 Status = STATUS_INTERNAL_ERROR; 2733 ExReleaseResourceLite(&c->lock); 2734 goto end; 2735 } 2736 2737 bgi = ExAllocatePoolWithTag(PagedPool, tp.item->size, ALLOC_TAG); 2738 if (!bgi) { 2739 ERR("out of memory\n"); 2740 Status = STATUS_INSUFFICIENT_RESOURCES; 2741 ExReleaseResourceLite(&c->lock); 2742 goto end; 2743 } 2744 2745 RtlCopyMemory(bgi, tp.item->data, tp.item->size); 2746 bgi->used = c->used; 2747 2748 TRACE("adjusting usage of chunk %llx to %llx\n", c->offset, c->used); 2749 2750 Status = delete_tree_item(Vcb, &tp); 2751 if (!NT_SUCCESS(Status)) { 2752 ERR("delete_tree_item returned %08x\n", Status); 2753 ExFreePool(bgi); 2754 ExReleaseResourceLite(&c->lock); 2755 goto end; 2756 } 2757 2758 Status = insert_tree_item(Vcb, Vcb->extent_root, searchkey.obj_id, searchkey.obj_type, searchkey.offset, bgi, tp.item->size, NULL, Irp); 2759 if (!NT_SUCCESS(Status)) { 2760 ERR("insert_tree_item returned %08x\n", Status); 2761 ExFreePool(bgi); 2762 ExReleaseResourceLite(&c->lock); 2763 goto end; 2764 } 2765 2766 TRACE("bytes_used = %llx\n", Vcb->superblock.bytes_used); 2767 2768 Vcb->superblock.bytes_used += c->used - c->oldused; 2769 2770 TRACE("bytes_used = %llx\n", Vcb->superblock.bytes_used); 2771 2772 c->oldused = c->used; 2773 } 2774 2775 ExReleaseResourceLite(&c->lock); 2776 2777 le = le->Flink; 2778 } 2779 2780 Status = STATUS_SUCCESS; 2781 2782 end: 2783 ExReleaseResourceLite(&Vcb->chunk_lock); 2784 2785 return Status; 2786 } 2787 2788 static void get_first_item(tree* t, KEY* key) { 2789 LIST_ENTRY* le; 2790 2791 le = t->itemlist.Flink; 2792 while (le != &t->itemlist) { 2793 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 2794 2795 *key = td->key; 2796 return; 2797 } 2798 } 2799 2800 static NTSTATUS split_tree_at(device_extension* Vcb, tree* t, tree_data* newfirstitem, UINT32 numitems, UINT32 size) { 2801 tree *nt, *pt; 2802 tree_data* td; 2803 tree_data* oldlastitem; 2804 2805 TRACE("splitting tree in %llx at (%llx,%x,%llx)\n", t->root->id, newfirstitem->key.obj_id, newfirstitem->key.obj_type, newfirstitem->key.offset); 2806 2807 nt = ExAllocatePoolWithTag(PagedPool, sizeof(tree), ALLOC_TAG); 2808 if (!nt) { 2809 ERR("out of memory\n"); 2810 return STATUS_INSUFFICIENT_RESOURCES; 2811 } 2812 2813 RtlCopyMemory(&nt->header, &t->header, sizeof(tree_header)); 2814 nt->header.address = 0; 2815 nt->header.generation = Vcb->superblock.generation; 2816 nt->header.num_items = t->header.num_items - numitems; 2817 nt->header.flags = HEADER_FLAG_MIXED_BACKREF | HEADER_FLAG_WRITTEN; 2818 2819 nt->has_address = FALSE; 2820 nt->Vcb = Vcb; 2821 nt->parent = t->parent; 2822 2823 #ifdef DEBUG_PARANOID 2824 if (nt->parent && nt->parent->header.level <= nt->header.level) int3; 2825 #endif 2826 2827 nt->root = t->root; 2828 nt->new_address = 0; 2829 nt->has_new_address = FALSE; 2830 nt->updated_extents = FALSE; 2831 nt->uniqueness_determined = TRUE; 2832 nt->is_unique = TRUE; 2833 nt->list_entry_hash.Flink = NULL; 2834 nt->buf = NULL; 2835 InitializeListHead(&nt->itemlist); 2836 2837 oldlastitem = CONTAINING_RECORD(newfirstitem->list_entry.Blink, tree_data, list_entry); 2838 2839 nt->itemlist.Flink = &newfirstitem->list_entry; 2840 nt->itemlist.Blink = t->itemlist.Blink; 2841 nt->itemlist.Flink->Blink = &nt->itemlist; 2842 nt->itemlist.Blink->Flink = &nt->itemlist; 2843 2844 t->itemlist.Blink = &oldlastitem->list_entry; 2845 t->itemlist.Blink->Flink = &t->itemlist; 2846 2847 nt->size = t->size - size; 2848 t->size = size; 2849 t->header.num_items = numitems; 2850 nt->write = TRUE; 2851 2852 InsertTailList(&Vcb->trees, &nt->list_entry); 2853 2854 if (nt->header.level > 0) { 2855 LIST_ENTRY* le = nt->itemlist.Flink; 2856 2857 while (le != &nt->itemlist) { 2858 tree_data* td2 = CONTAINING_RECORD(le, tree_data, list_entry); 2859 2860 if (td2->treeholder.tree) { 2861 td2->treeholder.tree->parent = nt; 2862 #ifdef DEBUG_PARANOID 2863 if (td2->treeholder.tree->parent && td2->treeholder.tree->parent->header.level <= td2->treeholder.tree->header.level) int3; 2864 #endif 2865 } 2866 2867 le = le->Flink; 2868 } 2869 } else { 2870 LIST_ENTRY* le = nt->itemlist.Flink; 2871 2872 while (le != &nt->itemlist) { 2873 tree_data* td2 = CONTAINING_RECORD(le, tree_data, list_entry); 2874 2875 if (!td2->inserted && td2->data) { 2876 UINT8* data = ExAllocatePoolWithTag(PagedPool, td2->size, ALLOC_TAG); 2877 2878 if (!data) { 2879 ERR("out of memory\n"); 2880 return STATUS_INSUFFICIENT_RESOURCES; 2881 } 2882 2883 RtlCopyMemory(data, td2->data, td2->size); 2884 td2->data = data; 2885 td2->inserted = TRUE; 2886 } 2887 2888 le = le->Flink; 2889 } 2890 } 2891 2892 if (nt->parent) { 2893 td = ExAllocateFromPagedLookasideList(&Vcb->tree_data_lookaside); 2894 if (!td) { 2895 ERR("out of memory\n"); 2896 return STATUS_INSUFFICIENT_RESOURCES; 2897 } 2898 2899 td->key = newfirstitem->key; 2900 2901 InsertHeadList(&t->paritem->list_entry, &td->list_entry); 2902 2903 td->ignore = FALSE; 2904 td->inserted = TRUE; 2905 td->treeholder.tree = nt; 2906 nt->paritem = td; 2907 2908 nt->parent->header.num_items++; 2909 nt->parent->size += sizeof(internal_node); 2910 2911 goto end; 2912 } 2913 2914 TRACE("adding new tree parent\n"); 2915 2916 if (nt->header.level == 255) { 2917 ERR("cannot add parent to tree at level 255\n"); 2918 return STATUS_INTERNAL_ERROR; 2919 } 2920 2921 pt = ExAllocatePoolWithTag(PagedPool, sizeof(tree), ALLOC_TAG); 2922 if (!pt) { 2923 ERR("out of memory\n"); 2924 return STATUS_INSUFFICIENT_RESOURCES; 2925 } 2926 2927 RtlCopyMemory(&pt->header, &nt->header, sizeof(tree_header)); 2928 pt->header.address = 0; 2929 pt->header.num_items = 2; 2930 pt->header.level = nt->header.level + 1; 2931 pt->header.flags = HEADER_FLAG_MIXED_BACKREF | HEADER_FLAG_WRITTEN; 2932 2933 pt->has_address = FALSE; 2934 pt->Vcb = Vcb; 2935 pt->parent = NULL; 2936 pt->paritem = NULL; 2937 pt->root = t->root; 2938 pt->new_address = 0; 2939 pt->has_new_address = FALSE; 2940 pt->updated_extents = FALSE; 2941 pt->size = pt->header.num_items * sizeof(internal_node); 2942 pt->uniqueness_determined = TRUE; 2943 pt->is_unique = TRUE; 2944 pt->list_entry_hash.Flink = NULL; 2945 pt->buf = NULL; 2946 InitializeListHead(&pt->itemlist); 2947 2948 InsertTailList(&Vcb->trees, &pt->list_entry); 2949 2950 td = ExAllocateFromPagedLookasideList(&Vcb->tree_data_lookaside); 2951 if (!td) { 2952 ERR("out of memory\n"); 2953 return STATUS_INSUFFICIENT_RESOURCES; 2954 } 2955 2956 get_first_item(t, &td->key); 2957 td->ignore = FALSE; 2958 td->inserted = FALSE; 2959 td->treeholder.address = 0; 2960 td->treeholder.generation = Vcb->superblock.generation; 2961 td->treeholder.tree = t; 2962 InsertTailList(&pt->itemlist, &td->list_entry); 2963 t->paritem = td; 2964 2965 td = ExAllocateFromPagedLookasideList(&Vcb->tree_data_lookaside); 2966 if (!td) { 2967 ERR("out of memory\n"); 2968 return STATUS_INSUFFICIENT_RESOURCES; 2969 } 2970 2971 td->key = newfirstitem->key; 2972 td->ignore = FALSE; 2973 td->inserted = FALSE; 2974 td->treeholder.address = 0; 2975 td->treeholder.generation = Vcb->superblock.generation; 2976 td->treeholder.tree = nt; 2977 InsertTailList(&pt->itemlist, &td->list_entry); 2978 nt->paritem = td; 2979 2980 pt->write = TRUE; 2981 2982 t->root->treeholder.tree = pt; 2983 2984 t->parent = pt; 2985 nt->parent = pt; 2986 2987 #ifdef DEBUG_PARANOID 2988 if (t->parent && t->parent->header.level <= t->header.level) int3; 2989 if (nt->parent && nt->parent->header.level <= nt->header.level) int3; 2990 #endif 2991 2992 end: 2993 t->root->root_item.bytes_used += Vcb->superblock.node_size; 2994 2995 return STATUS_SUCCESS; 2996 } 2997 2998 static NTSTATUS split_tree(device_extension* Vcb, tree* t) { 2999 LIST_ENTRY* le; 3000 UINT32 size, ds, numitems; 3001 3002 size = 0; 3003 numitems = 0; 3004 3005 // FIXME - naïve implementation: maximizes number of filled trees 3006 3007 le = t->itemlist.Flink; 3008 while (le != &t->itemlist) { 3009 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 3010 3011 if (!td->ignore) { 3012 if (t->header.level == 0) 3013 ds = sizeof(leaf_node) + td->size; 3014 else 3015 ds = sizeof(internal_node); 3016 3017 if (numitems == 0 && ds > Vcb->superblock.node_size - sizeof(tree_header)) { 3018 ERR("(%llx,%x,%llx) in tree %llx is too large (%x > %x)\n", 3019 td->key.obj_id, td->key.obj_type, td->key.offset, t->root->id, 3020 ds, Vcb->superblock.node_size - sizeof(tree_header)); 3021 return STATUS_INTERNAL_ERROR; 3022 } 3023 3024 // FIXME - move back if previous item was deleted item with same key 3025 if (size + ds > Vcb->superblock.node_size - sizeof(tree_header)) 3026 return split_tree_at(Vcb, t, td, numitems, size); 3027 3028 size += ds; 3029 numitems++; 3030 } 3031 3032 le = le->Flink; 3033 } 3034 3035 return STATUS_SUCCESS; 3036 } 3037 3038 BOOL is_tree_unique(device_extension* Vcb, tree* t, PIRP Irp) { 3039 KEY searchkey; 3040 traverse_ptr tp; 3041 NTSTATUS Status; 3042 BOOL ret = FALSE; 3043 EXTENT_ITEM* ei; 3044 UINT8* type; 3045 3046 if (t->uniqueness_determined) 3047 return t->is_unique; 3048 3049 if (t->parent && !is_tree_unique(Vcb, t->parent, Irp)) 3050 goto end; 3051 3052 if (t->has_address) { 3053 searchkey.obj_id = t->header.address; 3054 searchkey.obj_type = Vcb->superblock.incompat_flags & BTRFS_INCOMPAT_FLAGS_SKINNY_METADATA ? TYPE_METADATA_ITEM : TYPE_EXTENT_ITEM; 3055 searchkey.offset = 0xffffffffffffffff; 3056 3057 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 3058 if (!NT_SUCCESS(Status)) { 3059 ERR("error - find_item returned %08x\n", Status); 3060 goto end; 3061 } 3062 3063 if (tp.item->key.obj_id != t->header.address || (tp.item->key.obj_type != TYPE_METADATA_ITEM && tp.item->key.obj_type != TYPE_EXTENT_ITEM)) 3064 goto end; 3065 3066 if (tp.item->key.obj_type == TYPE_EXTENT_ITEM && tp.item->size == sizeof(EXTENT_ITEM_V0)) 3067 goto end; 3068 3069 if (tp.item->size < sizeof(EXTENT_ITEM)) 3070 goto end; 3071 3072 ei = (EXTENT_ITEM*)tp.item->data; 3073 3074 if (ei->refcount > 1) 3075 goto end; 3076 3077 if (tp.item->key.obj_type == TYPE_EXTENT_ITEM && ei->flags & EXTENT_ITEM_TREE_BLOCK) { 3078 EXTENT_ITEM2* ei2; 3079 3080 if (tp.item->size < sizeof(EXTENT_ITEM) + sizeof(EXTENT_ITEM2)) 3081 goto end; 3082 3083 ei2 = (EXTENT_ITEM2*)&ei[1]; 3084 type = (UINT8*)&ei2[1]; 3085 } else 3086 type = (UINT8*)&ei[1]; 3087 3088 if (type >= tp.item->data + tp.item->size || *type != TYPE_TREE_BLOCK_REF) 3089 goto end; 3090 } 3091 3092 ret = TRUE; 3093 3094 end: 3095 t->is_unique = ret; 3096 t->uniqueness_determined = TRUE; 3097 3098 return ret; 3099 } 3100 3101 static NTSTATUS try_tree_amalgamate(device_extension* Vcb, tree* t, BOOL* done, BOOL* done_deletions, PIRP Irp, LIST_ENTRY* rollback) { 3102 LIST_ENTRY* le; 3103 tree_data* nextparitem = NULL; 3104 NTSTATUS Status; 3105 tree *next_tree, *par; 3106 BOOL loaded; 3107 3108 *done = FALSE; 3109 3110 TRACE("trying to amalgamate tree in root %llx, level %x (size %u)\n", t->root->id, t->header.level, t->size); 3111 3112 // FIXME - doesn't capture everything, as it doesn't ascend 3113 le = t->paritem->list_entry.Flink; 3114 while (le != &t->parent->itemlist) { 3115 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 3116 3117 if (!td->ignore) { 3118 nextparitem = td; 3119 break; 3120 } 3121 3122 le = le->Flink; 3123 } 3124 3125 if (!nextparitem) 3126 return STATUS_SUCCESS; 3127 3128 TRACE("nextparitem: key = %llx,%x,%llx\n", nextparitem->key.obj_id, nextparitem->key.obj_type, nextparitem->key.offset); 3129 3130 Status = do_load_tree(Vcb, &nextparitem->treeholder, t->root, t->parent, nextparitem, &loaded, NULL); 3131 if (!NT_SUCCESS(Status)) { 3132 ERR("do_load_tree returned %08x\n", Status); 3133 return Status; 3134 } 3135 3136 if (!is_tree_unique(Vcb, nextparitem->treeholder.tree, Irp)) 3137 return STATUS_SUCCESS; 3138 3139 next_tree = nextparitem->treeholder.tree; 3140 3141 if (!next_tree->updated_extents && next_tree->has_address) { 3142 Status = update_tree_extents(Vcb, next_tree, Irp, rollback); 3143 if (!NT_SUCCESS(Status)) { 3144 ERR("update_tree_extents returned %08x\n", Status); 3145 return Status; 3146 } 3147 } 3148 3149 if (t->size + next_tree->size <= Vcb->superblock.node_size - sizeof(tree_header)) { 3150 // merge two trees into one 3151 3152 t->header.num_items += next_tree->header.num_items; 3153 t->size += next_tree->size; 3154 3155 if (next_tree->header.level > 0) { 3156 le = next_tree->itemlist.Flink; 3157 3158 while (le != &next_tree->itemlist) { 3159 tree_data* td2 = CONTAINING_RECORD(le, tree_data, list_entry); 3160 3161 if (td2->treeholder.tree) { 3162 td2->treeholder.tree->parent = t; 3163 #ifdef DEBUG_PARANOID 3164 if (td2->treeholder.tree->parent && td2->treeholder.tree->parent->header.level <= td2->treeholder.tree->header.level) int3; 3165 #endif 3166 } 3167 3168 td2->inserted = TRUE; 3169 le = le->Flink; 3170 } 3171 } else { 3172 le = next_tree->itemlist.Flink; 3173 3174 while (le != &next_tree->itemlist) { 3175 tree_data* td2 = CONTAINING_RECORD(le, tree_data, list_entry); 3176 3177 if (!td2->inserted && td2->data) { 3178 UINT8* data = ExAllocatePoolWithTag(PagedPool, td2->size, ALLOC_TAG); 3179 3180 if (!data) { 3181 ERR("out of memory\n"); 3182 return STATUS_INSUFFICIENT_RESOURCES; 3183 } 3184 3185 RtlCopyMemory(data, td2->data, td2->size); 3186 td2->data = data; 3187 td2->inserted = TRUE; 3188 } 3189 3190 le = le->Flink; 3191 } 3192 } 3193 3194 t->itemlist.Blink->Flink = next_tree->itemlist.Flink; 3195 t->itemlist.Blink->Flink->Blink = t->itemlist.Blink; 3196 t->itemlist.Blink = next_tree->itemlist.Blink; 3197 t->itemlist.Blink->Flink = &t->itemlist; 3198 3199 next_tree->itemlist.Flink = next_tree->itemlist.Blink = &next_tree->itemlist; 3200 3201 next_tree->header.num_items = 0; 3202 next_tree->size = 0; 3203 3204 if (next_tree->has_new_address) { // delete associated EXTENT_ITEM 3205 Status = reduce_tree_extent(Vcb, next_tree->new_address, next_tree, next_tree->parent->header.tree_id, next_tree->header.level, Irp, rollback); 3206 3207 if (!NT_SUCCESS(Status)) { 3208 ERR("reduce_tree_extent returned %08x\n", Status); 3209 return Status; 3210 } 3211 } else if (next_tree->has_address) { 3212 Status = reduce_tree_extent(Vcb, next_tree->header.address, next_tree, next_tree->parent->header.tree_id, next_tree->header.level, Irp, rollback); 3213 3214 if (!NT_SUCCESS(Status)) { 3215 ERR("reduce_tree_extent returned %08x\n", Status); 3216 return Status; 3217 } 3218 } 3219 3220 if (!nextparitem->ignore) { 3221 nextparitem->ignore = TRUE; 3222 next_tree->parent->header.num_items--; 3223 next_tree->parent->size -= sizeof(internal_node); 3224 3225 *done_deletions = TRUE; 3226 } 3227 3228 par = next_tree->parent; 3229 while (par) { 3230 par->write = TRUE; 3231 par = par->parent; 3232 } 3233 3234 RemoveEntryList(&nextparitem->list_entry); 3235 ExFreePool(next_tree->paritem); 3236 next_tree->paritem = NULL; 3237 3238 next_tree->root->root_item.bytes_used -= Vcb->superblock.node_size; 3239 3240 free_tree(next_tree); 3241 3242 *done = TRUE; 3243 } else { 3244 // rebalance by moving items from second tree into first 3245 ULONG avg_size = (t->size + next_tree->size) / 2; 3246 KEY firstitem = {0, 0, 0}; 3247 BOOL changed = FALSE; 3248 3249 TRACE("attempting rebalance\n"); 3250 3251 le = next_tree->itemlist.Flink; 3252 while (le != &next_tree->itemlist && t->size < avg_size && next_tree->header.num_items > 1) { 3253 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 3254 ULONG size; 3255 3256 if (!td->ignore) { 3257 if (next_tree->header.level == 0) 3258 size = sizeof(leaf_node) + td->size; 3259 else 3260 size = sizeof(internal_node); 3261 } else 3262 size = 0; 3263 3264 if (t->size + size < Vcb->superblock.node_size - sizeof(tree_header)) { 3265 RemoveEntryList(&td->list_entry); 3266 InsertTailList(&t->itemlist, &td->list_entry); 3267 3268 if (next_tree->header.level > 0 && td->treeholder.tree) { 3269 td->treeholder.tree->parent = t; 3270 #ifdef DEBUG_PARANOID 3271 if (td->treeholder.tree->parent && td->treeholder.tree->parent->header.level <= td->treeholder.tree->header.level) int3; 3272 #endif 3273 } else if (next_tree->header.level == 0 && !td->inserted && td->size > 0) { 3274 UINT8* data = ExAllocatePoolWithTag(PagedPool, td->size, ALLOC_TAG); 3275 3276 if (!data) { 3277 ERR("out of memory\n"); 3278 return STATUS_INSUFFICIENT_RESOURCES; 3279 } 3280 3281 RtlCopyMemory(data, td->data, td->size); 3282 td->data = data; 3283 } 3284 3285 td->inserted = TRUE; 3286 3287 if (!td->ignore) { 3288 next_tree->size -= size; 3289 t->size += size; 3290 next_tree->header.num_items--; 3291 t->header.num_items++; 3292 } 3293 3294 changed = TRUE; 3295 } else 3296 break; 3297 3298 le = next_tree->itemlist.Flink; 3299 } 3300 3301 le = next_tree->itemlist.Flink; 3302 while (le != &next_tree->itemlist) { 3303 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 3304 3305 if (!td->ignore) { 3306 firstitem = td->key; 3307 break; 3308 } 3309 3310 le = le->Flink; 3311 } 3312 3313 // FIXME - once ascension is working, make this work with parent's parent, etc. 3314 if (next_tree->paritem) 3315 next_tree->paritem->key = firstitem; 3316 3317 par = next_tree; 3318 while (par) { 3319 par->write = TRUE; 3320 par = par->parent; 3321 } 3322 3323 if (changed) 3324 *done = TRUE; 3325 } 3326 3327 return STATUS_SUCCESS; 3328 } 3329 3330 static NTSTATUS update_extent_level(device_extension* Vcb, UINT64 address, tree* t, UINT8 level, PIRP Irp) { 3331 KEY searchkey; 3332 traverse_ptr tp; 3333 NTSTATUS Status; 3334 3335 if (Vcb->superblock.incompat_flags & BTRFS_INCOMPAT_FLAGS_SKINNY_METADATA) { 3336 searchkey.obj_id = address; 3337 searchkey.obj_type = TYPE_METADATA_ITEM; 3338 searchkey.offset = t->header.level; 3339 3340 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 3341 if (!NT_SUCCESS(Status)) { 3342 ERR("error - find_item returned %08x\n", Status); 3343 return Status; 3344 } 3345 3346 if (!keycmp(tp.item->key, searchkey)) { 3347 EXTENT_ITEM_SKINNY_METADATA* eism; 3348 3349 if (tp.item->size > 0) { 3350 eism = ExAllocatePoolWithTag(PagedPool, tp.item->size, ALLOC_TAG); 3351 3352 if (!eism) { 3353 ERR("out of memory\n"); 3354 return STATUS_INSUFFICIENT_RESOURCES; 3355 } 3356 3357 RtlCopyMemory(eism, tp.item->data, tp.item->size); 3358 } else 3359 eism = NULL; 3360 3361 Status = delete_tree_item(Vcb, &tp); 3362 if (!NT_SUCCESS(Status)) { 3363 ERR("delete_tree_item returned %08x\n", Status); 3364 if (eism) ExFreePool(eism); 3365 return Status; 3366 } 3367 3368 Status = insert_tree_item(Vcb, Vcb->extent_root, address, TYPE_METADATA_ITEM, level, eism, tp.item->size, NULL, Irp); 3369 if (!NT_SUCCESS(Status)) { 3370 ERR("insert_tree_item returned %08x\n", Status); 3371 if (eism) ExFreePool(eism); 3372 return Status; 3373 } 3374 3375 return STATUS_SUCCESS; 3376 } 3377 } 3378 3379 searchkey.obj_id = address; 3380 searchkey.obj_type = TYPE_EXTENT_ITEM; 3381 searchkey.offset = 0xffffffffffffffff; 3382 3383 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 3384 if (!NT_SUCCESS(Status)) { 3385 ERR("error - find_item returned %08x\n", Status); 3386 return Status; 3387 } 3388 3389 if (tp.item->key.obj_id == searchkey.obj_id && tp.item->key.obj_type == searchkey.obj_type) { 3390 EXTENT_ITEM_TREE* eit; 3391 3392 if (tp.item->size < sizeof(EXTENT_ITEM_TREE)) { 3393 ERR("(%llx,%x,%llx) was %u bytes, expected at least %u\n", tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, tp.item->size, sizeof(EXTENT_ITEM_TREE)); 3394 return STATUS_INTERNAL_ERROR; 3395 } 3396 3397 eit = ExAllocatePoolWithTag(PagedPool, tp.item->size, ALLOC_TAG); 3398 3399 if (!eit) { 3400 ERR("out of memory\n"); 3401 return STATUS_INSUFFICIENT_RESOURCES; 3402 } 3403 3404 RtlCopyMemory(eit, tp.item->data, tp.item->size); 3405 3406 Status = delete_tree_item(Vcb, &tp); 3407 if (!NT_SUCCESS(Status)) { 3408 ERR("delete_tree_item returned %08x\n", Status); 3409 ExFreePool(eit); 3410 return Status; 3411 } 3412 3413 eit->level = level; 3414 3415 Status = insert_tree_item(Vcb, Vcb->extent_root, tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, eit, tp.item->size, NULL, Irp); 3416 if (!NT_SUCCESS(Status)) { 3417 ERR("insert_tree_item returned %08x\n", Status); 3418 ExFreePool(eit); 3419 return Status; 3420 } 3421 3422 return STATUS_SUCCESS; 3423 } 3424 3425 ERR("could not find EXTENT_ITEM for address %llx\n", address); 3426 3427 return STATUS_INTERNAL_ERROR; 3428 } 3429 3430 static NTSTATUS update_tree_extents_recursive(device_extension* Vcb, tree* t, PIRP Irp, LIST_ENTRY* rollback) { 3431 NTSTATUS Status; 3432 3433 if (t->parent && !t->parent->updated_extents && t->parent->has_address) { 3434 Status = update_tree_extents_recursive(Vcb, t->parent, Irp, rollback); 3435 if (!NT_SUCCESS(Status)) 3436 return Status; 3437 } 3438 3439 Status = update_tree_extents(Vcb, t, Irp, rollback); 3440 if (!NT_SUCCESS(Status)) { 3441 ERR("update_tree_extents returned %08x\n", Status); 3442 return Status; 3443 } 3444 3445 return STATUS_SUCCESS; 3446 } 3447 3448 static NTSTATUS do_splits(device_extension* Vcb, PIRP Irp, LIST_ENTRY* rollback) { 3449 ULONG level, max_level; 3450 UINT32 min_size; 3451 BOOL empty, done_deletions = FALSE; 3452 NTSTATUS Status; 3453 tree* t; 3454 3455 TRACE("(%p)\n", Vcb); 3456 3457 max_level = 0; 3458 3459 for (level = 0; level <= 255; level++) { 3460 LIST_ENTRY *le, *nextle; 3461 3462 empty = TRUE; 3463 3464 TRACE("doing level %u\n", level); 3465 3466 le = Vcb->trees.Flink; 3467 3468 while (le != &Vcb->trees) { 3469 t = CONTAINING_RECORD(le, tree, list_entry); 3470 3471 nextle = le->Flink; 3472 3473 if (t->write && t->header.level == level) { 3474 empty = FALSE; 3475 3476 if (t->header.num_items == 0) { 3477 if (t->parent) { 3478 done_deletions = TRUE; 3479 3480 TRACE("deleting tree in root %llx\n", t->root->id); 3481 3482 t->root->root_item.bytes_used -= Vcb->superblock.node_size; 3483 3484 if (t->has_new_address) { // delete associated EXTENT_ITEM 3485 Status = reduce_tree_extent(Vcb, t->new_address, t, t->parent->header.tree_id, t->header.level, Irp, rollback); 3486 3487 if (!NT_SUCCESS(Status)) { 3488 ERR("reduce_tree_extent returned %08x\n", Status); 3489 return Status; 3490 } 3491 3492 t->has_new_address = FALSE; 3493 } else if (t->has_address) { 3494 Status = reduce_tree_extent(Vcb,t->header.address, t, t->parent->header.tree_id, t->header.level, Irp, rollback); 3495 3496 if (!NT_SUCCESS(Status)) { 3497 ERR("reduce_tree_extent returned %08x\n", Status); 3498 return Status; 3499 } 3500 3501 t->has_address = FALSE; 3502 } 3503 3504 if (!t->paritem->ignore) { 3505 t->paritem->ignore = TRUE; 3506 t->parent->header.num_items--; 3507 t->parent->size -= sizeof(internal_node); 3508 } 3509 3510 RemoveEntryList(&t->paritem->list_entry); 3511 ExFreePool(t->paritem); 3512 t->paritem = NULL; 3513 3514 free_tree(t); 3515 } else if (t->header.level != 0) { 3516 if (t->has_new_address) { 3517 Status = update_extent_level(Vcb, t->new_address, t, 0, Irp); 3518 3519 if (!NT_SUCCESS(Status)) { 3520 ERR("update_extent_level returned %08x\n", Status); 3521 return Status; 3522 } 3523 } 3524 3525 t->header.level = 0; 3526 } 3527 } else if (t->size > Vcb->superblock.node_size - sizeof(tree_header)) { 3528 TRACE("splitting overlarge tree (%x > %x)\n", t->size, Vcb->superblock.node_size - sizeof(tree_header)); 3529 3530 if (!t->updated_extents && t->has_address) { 3531 Status = update_tree_extents_recursive(Vcb, t, Irp, rollback); 3532 if (!NT_SUCCESS(Status)) { 3533 ERR("update_tree_extents_recursive returned %08x\n", Status); 3534 return Status; 3535 } 3536 } 3537 3538 Status = split_tree(Vcb, t); 3539 3540 if (!NT_SUCCESS(Status)) { 3541 ERR("split_tree returned %08x\n", Status); 3542 return Status; 3543 } 3544 } 3545 } 3546 3547 le = nextle; 3548 } 3549 3550 if (!empty) { 3551 max_level = level; 3552 } else { 3553 TRACE("nothing found for level %u\n", level); 3554 break; 3555 } 3556 } 3557 3558 min_size = (Vcb->superblock.node_size - sizeof(tree_header)) / 2; 3559 3560 for (level = 0; level <= max_level; level++) { 3561 LIST_ENTRY* le; 3562 3563 le = Vcb->trees.Flink; 3564 3565 while (le != &Vcb->trees) { 3566 t = CONTAINING_RECORD(le, tree, list_entry); 3567 3568 if (t->write && t->header.level == level && t->header.num_items > 0 && t->parent && t->size < min_size && 3569 t->root->id != BTRFS_ROOT_FREE_SPACE && is_tree_unique(Vcb, t, Irp)) { 3570 BOOL done; 3571 3572 do { 3573 Status = try_tree_amalgamate(Vcb, t, &done, &done_deletions, Irp, rollback); 3574 if (!NT_SUCCESS(Status)) { 3575 ERR("try_tree_amalgamate returned %08x\n", Status); 3576 return Status; 3577 } 3578 } while (done && t->size < min_size); 3579 } 3580 3581 le = le->Flink; 3582 } 3583 } 3584 3585 // simplify trees if top tree only has one entry 3586 3587 if (done_deletions) { 3588 for (level = max_level; level > 0; level--) { 3589 LIST_ENTRY *le, *nextle; 3590 3591 le = Vcb->trees.Flink; 3592 while (le != &Vcb->trees) { 3593 nextle = le->Flink; 3594 t = CONTAINING_RECORD(le, tree, list_entry); 3595 3596 if (t->write && t->header.level == level) { 3597 if (!t->parent && t->header.num_items == 1) { 3598 LIST_ENTRY* le2 = t->itemlist.Flink; 3599 tree_data* td = NULL; 3600 tree* child_tree = NULL; 3601 3602 while (le2 != &t->itemlist) { 3603 td = CONTAINING_RECORD(le2, tree_data, list_entry); 3604 if (!td->ignore) 3605 break; 3606 le2 = le2->Flink; 3607 } 3608 3609 TRACE("deleting top-level tree in root %llx with one item\n", t->root->id); 3610 3611 if (t->has_new_address) { // delete associated EXTENT_ITEM 3612 Status = reduce_tree_extent(Vcb, t->new_address, t, t->header.tree_id, t->header.level, Irp, rollback); 3613 3614 if (!NT_SUCCESS(Status)) { 3615 ERR("reduce_tree_extent returned %08x\n", Status); 3616 return Status; 3617 } 3618 3619 t->has_new_address = FALSE; 3620 } else if (t->has_address) { 3621 Status = reduce_tree_extent(Vcb,t->header.address, t, t->header.tree_id, t->header.level, Irp, rollback); 3622 3623 if (!NT_SUCCESS(Status)) { 3624 ERR("reduce_tree_extent returned %08x\n", Status); 3625 return Status; 3626 } 3627 3628 t->has_address = FALSE; 3629 } 3630 3631 if (!td->treeholder.tree) { // load first item if not already loaded 3632 KEY searchkey = {0,0,0}; 3633 traverse_ptr tp; 3634 3635 Status = find_item(Vcb, t->root, &tp, &searchkey, FALSE, Irp); 3636 if (!NT_SUCCESS(Status)) { 3637 ERR("error - find_item returned %08x\n", Status); 3638 return Status; 3639 } 3640 } 3641 3642 child_tree = td->treeholder.tree; 3643 3644 if (child_tree) { 3645 child_tree->parent = NULL; 3646 child_tree->paritem = NULL; 3647 } 3648 3649 t->root->root_item.bytes_used -= Vcb->superblock.node_size; 3650 3651 free_tree(t); 3652 3653 if (child_tree) 3654 child_tree->root->treeholder.tree = child_tree; 3655 } 3656 } 3657 3658 le = nextle; 3659 } 3660 } 3661 } 3662 3663 return STATUS_SUCCESS; 3664 } 3665 3666 static NTSTATUS remove_root_extents(device_extension* Vcb, root* r, tree_holder* th, UINT8 level, tree* parent, PIRP Irp, LIST_ENTRY* rollback) { 3667 NTSTATUS Status; 3668 3669 if (!th->tree) { 3670 Status = load_tree(Vcb, th->address, r, &th->tree, th->generation, NULL); 3671 3672 if (!NT_SUCCESS(Status)) { 3673 ERR("load_tree(%llx) returned %08x\n", th->address, Status); 3674 return Status; 3675 } 3676 } 3677 3678 if (level > 0) { 3679 LIST_ENTRY* le = th->tree->itemlist.Flink; 3680 3681 while (le != &th->tree->itemlist) { 3682 tree_data* td = CONTAINING_RECORD(le, tree_data, list_entry); 3683 3684 if (!td->ignore) { 3685 Status = remove_root_extents(Vcb, r, &td->treeholder, th->tree->header.level - 1, th->tree, Irp, rollback); 3686 3687 if (!NT_SUCCESS(Status)) { 3688 ERR("remove_root_extents returned %08x\n", Status); 3689 return Status; 3690 } 3691 } 3692 3693 le = le->Flink; 3694 } 3695 } 3696 3697 if (th->tree && !th->tree->updated_extents && th->tree->has_address) { 3698 Status = update_tree_extents(Vcb, th->tree, Irp, rollback); 3699 if (!NT_SUCCESS(Status)) { 3700 ERR("update_tree_extents returned %08x\n", Status); 3701 return Status; 3702 } 3703 } 3704 3705 if (!th->tree || th->tree->has_address) { 3706 Status = reduce_tree_extent(Vcb, th->address, NULL, parent ? parent->header.tree_id : r->id, level, Irp, rollback); 3707 3708 if (!NT_SUCCESS(Status)) { 3709 ERR("reduce_tree_extent(%llx) returned %08x\n", th->address, Status); 3710 return Status; 3711 } 3712 } 3713 3714 return STATUS_SUCCESS; 3715 } 3716 3717 static NTSTATUS drop_root(device_extension* Vcb, root* r, PIRP Irp, LIST_ENTRY* rollback) { 3718 NTSTATUS Status; 3719 KEY searchkey; 3720 traverse_ptr tp; 3721 3722 Status = remove_root_extents(Vcb, r, &r->treeholder, r->root_item.root_level, NULL, Irp, rollback); 3723 if (!NT_SUCCESS(Status)) { 3724 ERR("remove_root_extents returned %08x\n", Status); 3725 return Status; 3726 } 3727 3728 // remove entries in uuid root (tree 9) 3729 if (Vcb->uuid_root) { 3730 RtlCopyMemory(&searchkey.obj_id, &r->root_item.uuid.uuid[0], sizeof(UINT64)); 3731 searchkey.obj_type = TYPE_SUBVOL_UUID; 3732 RtlCopyMemory(&searchkey.offset, &r->root_item.uuid.uuid[sizeof(UINT64)], sizeof(UINT64)); 3733 3734 if (searchkey.obj_id != 0 || searchkey.offset != 0) { 3735 Status = find_item(Vcb, Vcb->uuid_root, &tp, &searchkey, FALSE, Irp); 3736 if (!NT_SUCCESS(Status)) { 3737 WARN("find_item returned %08x\n", Status); 3738 } else { 3739 if (!keycmp(tp.item->key, searchkey)) { 3740 Status = delete_tree_item(Vcb, &tp); 3741 if (!NT_SUCCESS(Status)) { 3742 ERR("delete_tree_item returned %08x\n", Status); 3743 return Status; 3744 } 3745 } else 3746 WARN("could not find (%llx,%x,%llx) in uuid tree\n", searchkey.obj_id, searchkey.obj_type, searchkey.offset); 3747 } 3748 } 3749 3750 if (r->root_item.rtransid > 0) { 3751 RtlCopyMemory(&searchkey.obj_id, &r->root_item.received_uuid.uuid[0], sizeof(UINT64)); 3752 searchkey.obj_type = TYPE_SUBVOL_REC_UUID; 3753 RtlCopyMemory(&searchkey.offset, &r->root_item.received_uuid.uuid[sizeof(UINT64)], sizeof(UINT64)); 3754 3755 Status = find_item(Vcb, Vcb->uuid_root, &tp, &searchkey, FALSE, Irp); 3756 if (!NT_SUCCESS(Status)) 3757 WARN("find_item returned %08x\n", Status); 3758 else { 3759 if (!keycmp(tp.item->key, searchkey)) { 3760 if (tp.item->size == sizeof(UINT64)) { 3761 UINT64* id = (UINT64*)tp.item->data; 3762 3763 if (*id == r->id) { 3764 Status = delete_tree_item(Vcb, &tp); 3765 if (!NT_SUCCESS(Status)) { 3766 ERR("delete_tree_item returned %08x\n", Status); 3767 return Status; 3768 } 3769 } 3770 } else if (tp.item->size > sizeof(UINT64)) { 3771 ULONG i; 3772 UINT64* ids = (UINT64*)tp.item->data; 3773 3774 for (i = 0; i < tp.item->size / sizeof(UINT64); i++) { 3775 if (ids[i] == r->id) { 3776 UINT64* ne; 3777 3778 ne = ExAllocatePoolWithTag(PagedPool, tp.item->size - sizeof(UINT64), ALLOC_TAG); 3779 if (!ne) { 3780 ERR("out of memory\n"); 3781 return STATUS_INSUFFICIENT_RESOURCES; 3782 } 3783 3784 if (i > 0) 3785 RtlCopyMemory(ne, ids, sizeof(UINT64) * i); 3786 3787 if ((i + 1) * sizeof(UINT64) < tp.item->size) 3788 RtlCopyMemory(&ne[i], &ids[i + 1], tp.item->size - ((i + 1) * sizeof(UINT64))); 3789 3790 Status = delete_tree_item(Vcb, &tp); 3791 if (!NT_SUCCESS(Status)) { 3792 ERR("delete_tree_item returned %08x\n", Status); 3793 ExFreePool(ne); 3794 return Status; 3795 } 3796 3797 Status = insert_tree_item(Vcb, Vcb->uuid_root, tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, 3798 ne, tp.item->size - sizeof(UINT64), NULL, Irp); 3799 if (!NT_SUCCESS(Status)) { 3800 ERR("insert_tree_item returned %08x\n", Status); 3801 ExFreePool(ne); 3802 return Status; 3803 } 3804 3805 break; 3806 } 3807 } 3808 } 3809 } else 3810 WARN("could not find (%llx,%x,%llx) in uuid tree\n", searchkey.obj_id, searchkey.obj_type, searchkey.offset); 3811 } 3812 } 3813 } 3814 3815 // delete ROOT_ITEM 3816 3817 searchkey.obj_id = r->id; 3818 searchkey.obj_type = TYPE_ROOT_ITEM; 3819 searchkey.offset = 0xffffffffffffffff; 3820 3821 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 3822 if (!NT_SUCCESS(Status)) { 3823 ERR("find_item returned %08x\n", Status); 3824 return Status; 3825 } 3826 3827 if (tp.item->key.obj_id == searchkey.obj_id && tp.item->key.obj_type == searchkey.obj_type) { 3828 Status = delete_tree_item(Vcb, &tp); 3829 3830 if (!NT_SUCCESS(Status)) { 3831 ERR("delete_tree_item returned %08x\n", Status); 3832 return Status; 3833 } 3834 } else 3835 WARN("could not find (%llx,%x,%llx) in root_root\n", searchkey.obj_id, searchkey.obj_type, searchkey.offset); 3836 3837 // delete items in tree cache 3838 3839 free_trees_root(Vcb, r); 3840 3841 return STATUS_SUCCESS; 3842 } 3843 3844 static NTSTATUS drop_roots(device_extension* Vcb, PIRP Irp, LIST_ENTRY* rollback) { 3845 LIST_ENTRY *le = Vcb->drop_roots.Flink, *le2; 3846 NTSTATUS Status; 3847 3848 while (le != &Vcb->drop_roots) { 3849 root* r = CONTAINING_RECORD(le, root, list_entry); 3850 3851 le2 = le->Flink; 3852 3853 Status = drop_root(Vcb, r, Irp, rollback); 3854 if (!NT_SUCCESS(Status)) { 3855 ERR("drop_root(%llx) returned %08x\n", r->id, Status); 3856 return Status; 3857 } 3858 3859 le = le2; 3860 } 3861 3862 return STATUS_SUCCESS; 3863 } 3864 3865 NTSTATUS update_dev_item(device_extension* Vcb, device* device, PIRP Irp) { 3866 KEY searchkey; 3867 traverse_ptr tp; 3868 DEV_ITEM* di; 3869 NTSTATUS Status; 3870 3871 searchkey.obj_id = 1; 3872 searchkey.obj_type = TYPE_DEV_ITEM; 3873 searchkey.offset = device->devitem.dev_id; 3874 3875 Status = find_item(Vcb, Vcb->chunk_root, &tp, &searchkey, FALSE, Irp); 3876 if (!NT_SUCCESS(Status)) { 3877 ERR("error - find_item returned %08x\n", Status); 3878 return Status; 3879 } 3880 3881 if (keycmp(tp.item->key, searchkey)) { 3882 ERR("error - could not find DEV_ITEM for device %llx\n", device->devitem.dev_id); 3883 return STATUS_INTERNAL_ERROR; 3884 } 3885 3886 Status = delete_tree_item(Vcb, &tp); 3887 if (!NT_SUCCESS(Status)) { 3888 ERR("delete_tree_item returned %08x\n", Status); 3889 return Status; 3890 } 3891 3892 di = ExAllocatePoolWithTag(PagedPool, sizeof(DEV_ITEM), ALLOC_TAG); 3893 if (!di) { 3894 ERR("out of memory\n"); 3895 return STATUS_INSUFFICIENT_RESOURCES; 3896 } 3897 3898 RtlCopyMemory(di, &device->devitem, sizeof(DEV_ITEM)); 3899 3900 Status = insert_tree_item(Vcb, Vcb->chunk_root, 1, TYPE_DEV_ITEM, device->devitem.dev_id, di, sizeof(DEV_ITEM), NULL, Irp); 3901 if (!NT_SUCCESS(Status)) { 3902 ERR("insert_tree_item returned %08x\n", Status); 3903 ExFreePool(di); 3904 return Status; 3905 } 3906 3907 return STATUS_SUCCESS; 3908 } 3909 3910 static void regen_bootstrap(device_extension* Vcb) { 3911 sys_chunk* sc2; 3912 USHORT i = 0; 3913 LIST_ENTRY* le; 3914 3915 i = 0; 3916 le = Vcb->sys_chunks.Flink; 3917 while (le != &Vcb->sys_chunks) { 3918 sc2 = CONTAINING_RECORD(le, sys_chunk, list_entry); 3919 3920 TRACE("%llx,%x,%llx\n", sc2->key.obj_id, sc2->key.obj_type, sc2->key.offset); 3921 3922 RtlCopyMemory(&Vcb->superblock.sys_chunk_array[i], &sc2->key, sizeof(KEY)); 3923 i += sizeof(KEY); 3924 3925 RtlCopyMemory(&Vcb->superblock.sys_chunk_array[i], sc2->data, sc2->size); 3926 i += sc2->size; 3927 3928 le = le->Flink; 3929 } 3930 } 3931 3932 static NTSTATUS add_to_bootstrap(device_extension* Vcb, UINT64 obj_id, UINT8 obj_type, UINT64 offset, void* data, UINT16 size) { 3933 sys_chunk* sc; 3934 LIST_ENTRY* le; 3935 3936 if (Vcb->superblock.n + sizeof(KEY) + size > SYS_CHUNK_ARRAY_SIZE) { 3937 ERR("error - bootstrap is full\n"); 3938 return STATUS_INTERNAL_ERROR; 3939 } 3940 3941 sc = ExAllocatePoolWithTag(PagedPool, sizeof(sys_chunk), ALLOC_TAG); 3942 if (!sc) { 3943 ERR("out of memory\n"); 3944 return STATUS_INSUFFICIENT_RESOURCES; 3945 } 3946 3947 sc->key.obj_id = obj_id; 3948 sc->key.obj_type = obj_type; 3949 sc->key.offset = offset; 3950 sc->size = size; 3951 sc->data = ExAllocatePoolWithTag(PagedPool, sc->size, ALLOC_TAG); 3952 if (!sc->data) { 3953 ERR("out of memory\n"); 3954 ExFreePool(sc); 3955 return STATUS_INSUFFICIENT_RESOURCES; 3956 } 3957 3958 RtlCopyMemory(sc->data, data, sc->size); 3959 3960 le = Vcb->sys_chunks.Flink; 3961 while (le != &Vcb->sys_chunks) { 3962 sys_chunk* sc2 = CONTAINING_RECORD(le, sys_chunk, list_entry); 3963 3964 if (keycmp(sc2->key, sc->key) == 1) 3965 break; 3966 3967 le = le->Flink; 3968 } 3969 InsertTailList(le, &sc->list_entry); 3970 3971 Vcb->superblock.n += sizeof(KEY) + size; 3972 3973 regen_bootstrap(Vcb); 3974 3975 return STATUS_SUCCESS; 3976 } 3977 3978 static NTSTATUS create_chunk(device_extension* Vcb, chunk* c, PIRP Irp) { 3979 CHUNK_ITEM* ci; 3980 CHUNK_ITEM_STRIPE* cis; 3981 BLOCK_GROUP_ITEM* bgi; 3982 UINT16 i, factor; 3983 NTSTATUS Status; 3984 3985 ci = ExAllocatePoolWithTag(PagedPool, c->size, ALLOC_TAG); 3986 if (!ci) { 3987 ERR("out of memory\n"); 3988 return STATUS_INSUFFICIENT_RESOURCES; 3989 } 3990 3991 RtlCopyMemory(ci, c->chunk_item, c->size); 3992 3993 Status = insert_tree_item(Vcb, Vcb->chunk_root, 0x100, TYPE_CHUNK_ITEM, c->offset, ci, c->size, NULL, Irp); 3994 if (!NT_SUCCESS(Status)) { 3995 ERR("insert_tree_item failed\n"); 3996 ExFreePool(ci); 3997 return Status; 3998 } 3999 4000 if (c->chunk_item->type & BLOCK_FLAG_SYSTEM) { 4001 Status = add_to_bootstrap(Vcb, 0x100, TYPE_CHUNK_ITEM, c->offset, ci, c->size); 4002 if (!NT_SUCCESS(Status)) { 4003 ERR("add_to_bootstrap returned %08x\n", Status); 4004 return Status; 4005 } 4006 } 4007 4008 // add BLOCK_GROUP_ITEM to tree 2 4009 4010 bgi = ExAllocatePoolWithTag(PagedPool, sizeof(BLOCK_GROUP_ITEM), ALLOC_TAG); 4011 if (!bgi) { 4012 ERR("out of memory\n"); 4013 return STATUS_INSUFFICIENT_RESOURCES; 4014 } 4015 4016 bgi->used = c->used; 4017 bgi->chunk_tree = 0x100; 4018 bgi->flags = c->chunk_item->type; 4019 4020 Status = insert_tree_item(Vcb, Vcb->extent_root, c->offset, TYPE_BLOCK_GROUP_ITEM, c->chunk_item->size, bgi, sizeof(BLOCK_GROUP_ITEM), NULL, Irp); 4021 if (!NT_SUCCESS(Status)) { 4022 ERR("insert_tree_item failed\n"); 4023 ExFreePool(bgi); 4024 return Status; 4025 } 4026 4027 if (c->chunk_item->type & BLOCK_FLAG_RAID0) 4028 factor = c->chunk_item->num_stripes; 4029 else if (c->chunk_item->type & BLOCK_FLAG_RAID10) 4030 factor = c->chunk_item->num_stripes / c->chunk_item->sub_stripes; 4031 else if (c->chunk_item->type & BLOCK_FLAG_RAID5) 4032 factor = c->chunk_item->num_stripes - 1; 4033 else if (c->chunk_item->type & BLOCK_FLAG_RAID6) 4034 factor = c->chunk_item->num_stripes - 2; 4035 else // SINGLE, DUPLICATE, RAID1 4036 factor = 1; 4037 4038 // add DEV_EXTENTs to tree 4 4039 4040 cis = (CHUNK_ITEM_STRIPE*)&c->chunk_item[1]; 4041 4042 for (i = 0; i < c->chunk_item->num_stripes; i++) { 4043 DEV_EXTENT* de; 4044 4045 de = ExAllocatePoolWithTag(PagedPool, sizeof(DEV_EXTENT), ALLOC_TAG); 4046 if (!de) { 4047 ERR("out of memory\n"); 4048 return STATUS_INSUFFICIENT_RESOURCES; 4049 } 4050 4051 de->chunktree = Vcb->chunk_root->id; 4052 de->objid = 0x100; 4053 de->address = c->offset; 4054 de->length = c->chunk_item->size / factor; 4055 de->chunktree_uuid = Vcb->chunk_root->treeholder.tree->header.chunk_tree_uuid; 4056 4057 Status = insert_tree_item(Vcb, Vcb->dev_root, c->devices[i]->devitem.dev_id, TYPE_DEV_EXTENT, cis[i].offset, de, sizeof(DEV_EXTENT), NULL, Irp); 4058 if (!NT_SUCCESS(Status)) { 4059 ERR("insert_tree_item returned %08x\n", Status); 4060 ExFreePool(de); 4061 return Status; 4062 } 4063 4064 // FIXME - no point in calling this twice for the same device 4065 Status = update_dev_item(Vcb, c->devices[i], Irp); 4066 if (!NT_SUCCESS(Status)) { 4067 ERR("update_dev_item returned %08x\n", Status); 4068 return Status; 4069 } 4070 } 4071 4072 c->created = FALSE; 4073 4074 return STATUS_SUCCESS; 4075 } 4076 4077 static void remove_from_bootstrap(device_extension* Vcb, UINT64 obj_id, UINT8 obj_type, UINT64 offset) { 4078 sys_chunk* sc2; 4079 LIST_ENTRY* le; 4080 4081 le = Vcb->sys_chunks.Flink; 4082 while (le != &Vcb->sys_chunks) { 4083 sc2 = CONTAINING_RECORD(le, sys_chunk, list_entry); 4084 4085 if (sc2->key.obj_id == obj_id && sc2->key.obj_type == obj_type && sc2->key.offset == offset) { 4086 RemoveEntryList(&sc2->list_entry); 4087 4088 Vcb->superblock.n -= sizeof(KEY) + sc2->size; 4089 4090 ExFreePool(sc2->data); 4091 ExFreePool(sc2); 4092 regen_bootstrap(Vcb); 4093 return; 4094 } 4095 4096 le = le->Flink; 4097 } 4098 } 4099 4100 static NTSTATUS set_xattr(device_extension* Vcb, LIST_ENTRY* batchlist, root* subvol, UINT64 inode, char* name, UINT16 namelen, 4101 UINT32 crc32, UINT8* data, UINT16 datalen) { 4102 NTSTATUS Status; 4103 UINT16 xasize; 4104 DIR_ITEM* xa; 4105 4106 TRACE("(%p, %llx, %llx, %.*s, %08x, %p, %u)\n", Vcb, subvol->id, inode, namelen, name, crc32, data, datalen); 4107 4108 xasize = (UINT16)offsetof(DIR_ITEM, name[0]) + namelen + datalen; 4109 4110 xa = ExAllocatePoolWithTag(PagedPool, xasize, ALLOC_TAG); 4111 if (!xa) { 4112 ERR("out of memory\n"); 4113 return STATUS_INSUFFICIENT_RESOURCES; 4114 } 4115 4116 xa->key.obj_id = 0; 4117 xa->key.obj_type = 0; 4118 xa->key.offset = 0; 4119 xa->transid = Vcb->superblock.generation; 4120 xa->m = datalen; 4121 xa->n = namelen; 4122 xa->type = BTRFS_TYPE_EA; 4123 RtlCopyMemory(xa->name, name, namelen); 4124 RtlCopyMemory(xa->name + namelen, data, datalen); 4125 4126 Status = insert_tree_item_batch(batchlist, Vcb, subvol, inode, TYPE_XATTR_ITEM, crc32, xa, xasize, Batch_SetXattr); 4127 if (!NT_SUCCESS(Status)) { 4128 ERR("insert_tree_item_batch returned %08x\n", Status); 4129 ExFreePool(xa); 4130 return Status; 4131 } 4132 4133 return STATUS_SUCCESS; 4134 } 4135 4136 static NTSTATUS delete_xattr(device_extension* Vcb, LIST_ENTRY* batchlist, root* subvol, UINT64 inode, char* name, 4137 UINT16 namelen, UINT32 crc32) { 4138 NTSTATUS Status; 4139 UINT16 xasize; 4140 DIR_ITEM* xa; 4141 4142 TRACE("(%p, %llx, %llx, %.*s, %08x)\n", Vcb, subvol->id, inode, namelen, name, crc32); 4143 4144 xasize = (UINT16)offsetof(DIR_ITEM, name[0]) + namelen; 4145 4146 xa = ExAllocatePoolWithTag(PagedPool, xasize, ALLOC_TAG); 4147 if (!xa) { 4148 ERR("out of memory\n"); 4149 return STATUS_INSUFFICIENT_RESOURCES; 4150 } 4151 4152 xa->key.obj_id = 0; 4153 xa->key.obj_type = 0; 4154 xa->key.offset = 0; 4155 xa->transid = Vcb->superblock.generation; 4156 xa->m = 0; 4157 xa->n = namelen; 4158 xa->type = BTRFS_TYPE_EA; 4159 RtlCopyMemory(xa->name, name, namelen); 4160 4161 Status = insert_tree_item_batch(batchlist, Vcb, subvol, inode, TYPE_XATTR_ITEM, crc32, xa, xasize, Batch_DeleteXattr); 4162 if (!NT_SUCCESS(Status)) { 4163 ERR("insert_tree_item_batch returned %08x\n", Status); 4164 ExFreePool(xa); 4165 return Status; 4166 } 4167 4168 return STATUS_SUCCESS; 4169 } 4170 4171 static NTSTATUS insert_sparse_extent(fcb* fcb, LIST_ENTRY* batchlist, UINT64 start, UINT64 length) { 4172 NTSTATUS Status; 4173 EXTENT_DATA* ed; 4174 EXTENT_DATA2* ed2; 4175 4176 TRACE("((%llx, %llx), %llx, %llx)\n", fcb->subvol->id, fcb->inode, start, length); 4177 4178 ed = ExAllocatePoolWithTag(PagedPool, sizeof(EXTENT_DATA) - 1 + sizeof(EXTENT_DATA2), ALLOC_TAG); 4179 if (!ed) { 4180 ERR("out of memory\n"); 4181 return STATUS_INSUFFICIENT_RESOURCES; 4182 } 4183 4184 ed->generation = fcb->Vcb->superblock.generation; 4185 ed->decoded_size = length; 4186 ed->compression = BTRFS_COMPRESSION_NONE; 4187 ed->encryption = BTRFS_ENCRYPTION_NONE; 4188 ed->encoding = BTRFS_ENCODING_NONE; 4189 ed->type = EXTENT_TYPE_REGULAR; 4190 4191 ed2 = (EXTENT_DATA2*)ed->data; 4192 ed2->address = 0; 4193 ed2->size = 0; 4194 ed2->offset = 0; 4195 ed2->num_bytes = length; 4196 4197 Status = insert_tree_item_batch(batchlist, fcb->Vcb, fcb->subvol, fcb->inode, TYPE_EXTENT_DATA, start, ed, sizeof(EXTENT_DATA) - 1 + sizeof(EXTENT_DATA2), Batch_Insert); 4198 if (!NT_SUCCESS(Status)) { 4199 ERR("insert_tree_item_batch returned %08x\n", Status); 4200 ExFreePool(ed); 4201 return Status; 4202 } 4203 4204 return STATUS_SUCCESS; 4205 } 4206 4207 #ifdef _MSC_VER 4208 #pragma warning(push) 4209 #pragma warning(suppress: 28194) 4210 #endif 4211 NTSTATUS insert_tree_item_batch(LIST_ENTRY* batchlist, device_extension* Vcb, root* r, UINT64 objid, UINT8 objtype, UINT64 offset, 4212 _In_opt_ _When_(return >= 0, __drv_aliasesMem) void* data, UINT16 datalen, enum batch_operation operation) { 4213 LIST_ENTRY* le; 4214 batch_root* br = NULL; 4215 batch_item* bi; 4216 4217 le = batchlist->Flink; 4218 while (le != batchlist) { 4219 batch_root* br2 = CONTAINING_RECORD(le, batch_root, list_entry); 4220 4221 if (br2->r == r) { 4222 br = br2; 4223 break; 4224 } 4225 4226 le = le->Flink; 4227 } 4228 4229 if (!br) { 4230 br = ExAllocatePoolWithTag(PagedPool, sizeof(batch_root), ALLOC_TAG); 4231 if (!br) { 4232 ERR("out of memory\n"); 4233 return STATUS_INSUFFICIENT_RESOURCES; 4234 } 4235 4236 br->r = r; 4237 InitializeListHead(&br->items); 4238 InsertTailList(batchlist, &br->list_entry); 4239 } 4240 4241 bi = ExAllocateFromPagedLookasideList(&Vcb->batch_item_lookaside); 4242 if (!bi) { 4243 ERR("out of memory\n"); 4244 return STATUS_INSUFFICIENT_RESOURCES; 4245 } 4246 4247 bi->key.obj_id = objid; 4248 bi->key.obj_type = objtype; 4249 bi->key.offset = offset; 4250 bi->data = data; 4251 bi->datalen = datalen; 4252 bi->operation = operation; 4253 4254 le = br->items.Blink; 4255 while (le != &br->items) { 4256 batch_item* bi2 = CONTAINING_RECORD(le, batch_item, list_entry); 4257 4258 if (keycmp(bi2->key, bi->key) != 1) { 4259 InsertHeadList(&bi2->list_entry, &bi->list_entry); 4260 return STATUS_SUCCESS; 4261 } 4262 4263 le = le->Blink; 4264 } 4265 4266 InsertHeadList(&br->items, &bi->list_entry); 4267 4268 return STATUS_SUCCESS; 4269 } 4270 #ifdef _MSC_VER 4271 #pragma warning(pop) 4272 #endif 4273 4274 typedef struct { 4275 UINT64 address; 4276 UINT64 length; 4277 UINT64 offset; 4278 BOOL changed; 4279 chunk* chunk; 4280 UINT64 skip_start; 4281 UINT64 skip_end; 4282 LIST_ENTRY list_entry; 4283 } extent_range; 4284 4285 static void rationalize_extents(fcb* fcb, PIRP Irp) { 4286 LIST_ENTRY* le; 4287 LIST_ENTRY extent_ranges; 4288 extent_range* er; 4289 BOOL changed = FALSE, truncating = FALSE; 4290 UINT32 num_extents = 0; 4291 4292 InitializeListHead(&extent_ranges); 4293 4294 le = fcb->extents.Flink; 4295 while (le != &fcb->extents) { 4296 extent* ext = CONTAINING_RECORD(le, extent, list_entry); 4297 4298 if ((ext->extent_data.type == EXTENT_TYPE_REGULAR || ext->extent_data.type == EXTENT_TYPE_PREALLOC) && ext->extent_data.compression == BTRFS_COMPRESSION_NONE && ext->unique) { 4299 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ext->extent_data.data; 4300 4301 if (ed2->size != 0) { 4302 LIST_ENTRY* le2; 4303 4304 le2 = extent_ranges.Flink; 4305 while (le2 != &extent_ranges) { 4306 extent_range* er2 = CONTAINING_RECORD(le2, extent_range, list_entry); 4307 4308 if (er2->address == ed2->address) { 4309 er2->skip_start = min(er2->skip_start, ed2->offset); 4310 er2->skip_end = min(er2->skip_end, ed2->size - ed2->offset - ed2->num_bytes); 4311 goto cont; 4312 } else if (er2->address > ed2->address) 4313 break; 4314 4315 le2 = le2->Flink; 4316 } 4317 4318 er = ExAllocatePoolWithTag(PagedPool, sizeof(extent_range), ALLOC_TAG); // FIXME - should be from lookaside? 4319 if (!er) { 4320 ERR("out of memory\n"); 4321 goto end; 4322 } 4323 4324 er->address = ed2->address; 4325 er->length = ed2->size; 4326 er->offset = ext->offset - ed2->offset; 4327 er->changed = FALSE; 4328 er->chunk = NULL; 4329 er->skip_start = ed2->offset; 4330 er->skip_end = ed2->size - ed2->offset - ed2->num_bytes; 4331 4332 if (er->skip_start != 0 || er->skip_end != 0) 4333 truncating = TRUE; 4334 4335 InsertHeadList(le2->Blink, &er->list_entry); 4336 num_extents++; 4337 } 4338 } 4339 4340 cont: 4341 le = le->Flink; 4342 } 4343 4344 if (num_extents == 0 || (num_extents == 1 && !truncating)) 4345 goto end; 4346 4347 le = extent_ranges.Flink; 4348 while (le != &extent_ranges) { 4349 er = CONTAINING_RECORD(le, extent_range, list_entry); 4350 4351 if (!er->chunk) { 4352 LIST_ENTRY* le2; 4353 4354 er->chunk = get_chunk_from_address(fcb->Vcb, er->address); 4355 4356 if (!er->chunk) { 4357 ERR("get_chunk_from_address(%llx) failed\n", er->address); 4358 goto end; 4359 } 4360 4361 le2 = le->Flink; 4362 while (le2 != &extent_ranges) { 4363 extent_range* er2 = CONTAINING_RECORD(le2, extent_range, list_entry); 4364 4365 if (!er2->chunk && er2->address >= er->chunk->offset && er2->address < er->chunk->offset + er->chunk->chunk_item->size) 4366 er2->chunk = er->chunk; 4367 4368 le2 = le2->Flink; 4369 } 4370 } 4371 4372 le = le->Flink; 4373 } 4374 4375 if (truncating) { 4376 // truncate beginning or end of extent if unused 4377 4378 le = extent_ranges.Flink; 4379 while (le != &extent_ranges) { 4380 er = CONTAINING_RECORD(le, extent_range, list_entry); 4381 4382 if (er->skip_start > 0) { 4383 LIST_ENTRY* le2 = fcb->extents.Flink; 4384 while (le2 != &fcb->extents) { 4385 extent* ext = CONTAINING_RECORD(le2, extent, list_entry); 4386 4387 if ((ext->extent_data.type == EXTENT_TYPE_REGULAR || ext->extent_data.type == EXTENT_TYPE_PREALLOC) && ext->extent_data.compression == BTRFS_COMPRESSION_NONE && ext->unique) { 4388 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ext->extent_data.data; 4389 4390 if (ed2->size != 0 && ed2->address == er->address) { 4391 NTSTATUS Status; 4392 4393 Status = update_changed_extent_ref(fcb->Vcb, er->chunk, ed2->address, ed2->size, fcb->subvol->id, fcb->inode, ext->offset - ed2->offset, 4394 -1, fcb->inode_item.flags & BTRFS_INODE_NODATASUM, TRUE, Irp); 4395 if (!NT_SUCCESS(Status)) { 4396 ERR("update_changed_extent_ref returned %08x\n", Status); 4397 goto end; 4398 } 4399 4400 ext->extent_data.decoded_size -= er->skip_start; 4401 ed2->size -= er->skip_start; 4402 ed2->address += er->skip_start; 4403 ed2->offset -= er->skip_start; 4404 4405 add_changed_extent_ref(er->chunk, ed2->address, ed2->size, fcb->subvol->id, fcb->inode, ext->offset - ed2->offset, 4406 1, fcb->inode_item.flags & BTRFS_INODE_NODATASUM); 4407 } 4408 } 4409 4410 le2 = le2->Flink; 4411 } 4412 4413 if (!(fcb->inode_item.flags & BTRFS_INODE_NODATASUM)) 4414 add_checksum_entry(fcb->Vcb, er->address, (ULONG)(er->skip_start / fcb->Vcb->superblock.sector_size), NULL, NULL); 4415 4416 ExAcquireResourceExclusiveLite(&er->chunk->lock, TRUE); 4417 4418 if (!er->chunk->cache_loaded) { 4419 NTSTATUS Status = load_cache_chunk(fcb->Vcb, er->chunk, NULL); 4420 4421 if (!NT_SUCCESS(Status)) { 4422 ERR("load_cache_chunk returned %08x\n", Status); 4423 ExReleaseResourceLite(&er->chunk->lock); 4424 goto end; 4425 } 4426 } 4427 4428 er->chunk->used -= er->skip_start; 4429 4430 space_list_add(er->chunk, er->address, er->skip_start, NULL); 4431 4432 ExReleaseResourceLite(&er->chunk->lock); 4433 4434 er->address += er->skip_start; 4435 er->length -= er->skip_start; 4436 } 4437 4438 if (er->skip_end > 0) { 4439 LIST_ENTRY* le2 = fcb->extents.Flink; 4440 while (le2 != &fcb->extents) { 4441 extent* ext = CONTAINING_RECORD(le2, extent, list_entry); 4442 4443 if ((ext->extent_data.type == EXTENT_TYPE_REGULAR || ext->extent_data.type == EXTENT_TYPE_PREALLOC) && ext->extent_data.compression == BTRFS_COMPRESSION_NONE && ext->unique) { 4444 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ext->extent_data.data; 4445 4446 if (ed2->size != 0 && ed2->address == er->address) { 4447 NTSTATUS Status; 4448 4449 Status = update_changed_extent_ref(fcb->Vcb, er->chunk, ed2->address, ed2->size, fcb->subvol->id, fcb->inode, ext->offset - ed2->offset, 4450 -1, fcb->inode_item.flags & BTRFS_INODE_NODATASUM, TRUE, Irp); 4451 if (!NT_SUCCESS(Status)) { 4452 ERR("update_changed_extent_ref returned %08x\n", Status); 4453 goto end; 4454 } 4455 4456 ext->extent_data.decoded_size -= er->skip_end; 4457 ed2->size -= er->skip_end; 4458 4459 add_changed_extent_ref(er->chunk, ed2->address, ed2->size, fcb->subvol->id, fcb->inode, ext->offset - ed2->offset, 4460 1, fcb->inode_item.flags & BTRFS_INODE_NODATASUM); 4461 } 4462 } 4463 4464 le2 = le2->Flink; 4465 } 4466 4467 if (!(fcb->inode_item.flags & BTRFS_INODE_NODATASUM)) 4468 add_checksum_entry(fcb->Vcb, er->address + er->length - er->skip_end, (ULONG)(er->skip_end / fcb->Vcb->superblock.sector_size), NULL, NULL); 4469 4470 ExAcquireResourceExclusiveLite(&er->chunk->lock, TRUE); 4471 4472 if (!er->chunk->cache_loaded) { 4473 NTSTATUS Status = load_cache_chunk(fcb->Vcb, er->chunk, NULL); 4474 4475 if (!NT_SUCCESS(Status)) { 4476 ERR("load_cache_chunk returned %08x\n", Status); 4477 ExReleaseResourceLite(&er->chunk->lock); 4478 goto end; 4479 } 4480 } 4481 4482 er->chunk->used -= er->skip_end; 4483 4484 space_list_add(er->chunk, er->address + er->length - er->skip_end, er->skip_end, NULL); 4485 4486 ExReleaseResourceLite(&er->chunk->lock); 4487 4488 er->length -= er->skip_end; 4489 } 4490 4491 le = le->Flink; 4492 } 4493 } 4494 4495 if (num_extents < 2) 4496 goto end; 4497 4498 // merge together adjacent extents 4499 le = extent_ranges.Flink; 4500 while (le != &extent_ranges) { 4501 er = CONTAINING_RECORD(le, extent_range, list_entry); 4502 4503 if (le->Flink != &extent_ranges && er->length < MAX_EXTENT_SIZE) { 4504 extent_range* er2 = CONTAINING_RECORD(le->Flink, extent_range, list_entry); 4505 4506 if (er->chunk == er2->chunk) { 4507 if (er2->address == er->address + er->length && er2->offset >= er->offset + er->length) { 4508 if (er->length + er2->length <= MAX_EXTENT_SIZE) { 4509 er->length += er2->length; 4510 er->changed = TRUE; 4511 4512 RemoveEntryList(&er2->list_entry); 4513 ExFreePool(er2); 4514 4515 changed = TRUE; 4516 continue; 4517 } 4518 } 4519 } 4520 } 4521 4522 le = le->Flink; 4523 } 4524 4525 if (!changed) 4526 goto end; 4527 4528 le = fcb->extents.Flink; 4529 while (le != &fcb->extents) { 4530 extent* ext = CONTAINING_RECORD(le, extent, list_entry); 4531 4532 if ((ext->extent_data.type == EXTENT_TYPE_REGULAR || ext->extent_data.type == EXTENT_TYPE_PREALLOC) && ext->extent_data.compression == BTRFS_COMPRESSION_NONE && ext->unique) { 4533 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ext->extent_data.data; 4534 4535 if (ed2->size != 0) { 4536 LIST_ENTRY* le2; 4537 4538 le2 = extent_ranges.Flink; 4539 while (le2 != &extent_ranges) { 4540 extent_range* er2 = CONTAINING_RECORD(le2, extent_range, list_entry); 4541 4542 if (ed2->address >= er2->address && ed2->address + ed2->size <= er2->address + er2->length && er2->changed) { 4543 NTSTATUS Status; 4544 4545 Status = update_changed_extent_ref(fcb->Vcb, er2->chunk, ed2->address, ed2->size, fcb->subvol->id, fcb->inode, ext->offset - ed2->offset, 4546 -1, fcb->inode_item.flags & BTRFS_INODE_NODATASUM, TRUE, Irp); 4547 if (!NT_SUCCESS(Status)) { 4548 ERR("update_changed_extent_ref returned %08x\n", Status); 4549 goto end; 4550 } 4551 4552 ed2->offset += ed2->address - er2->address; 4553 ed2->address = er2->address; 4554 ed2->size = er2->length; 4555 ext->extent_data.decoded_size = ed2->size; 4556 4557 add_changed_extent_ref(er2->chunk, ed2->address, ed2->size, fcb->subvol->id, fcb->inode, ext->offset - ed2->offset, 4558 1, fcb->inode_item.flags & BTRFS_INODE_NODATASUM); 4559 4560 break; 4561 } 4562 4563 le2 = le2->Flink; 4564 } 4565 } 4566 } 4567 4568 le = le->Flink; 4569 } 4570 4571 end: 4572 while (!IsListEmpty(&extent_ranges)) { 4573 le = RemoveHeadList(&extent_ranges); 4574 er = CONTAINING_RECORD(le, extent_range, list_entry); 4575 4576 ExFreePool(er); 4577 } 4578 } 4579 4580 NTSTATUS flush_fcb(fcb* fcb, BOOL cache, LIST_ENTRY* batchlist, PIRP Irp) { 4581 traverse_ptr tp; 4582 KEY searchkey; 4583 NTSTATUS Status; 4584 INODE_ITEM* ii; 4585 UINT64 ii_offset; 4586 #ifdef DEBUG_PARANOID 4587 UINT64 old_size = 0; 4588 BOOL extents_changed; 4589 #endif 4590 4591 if (fcb->ads) { 4592 if (fcb->deleted) { 4593 Status = delete_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, fcb->adsxattr.Buffer, fcb->adsxattr.Length, fcb->adshash); 4594 if (!NT_SUCCESS(Status)) { 4595 ERR("delete_xattr returned %08x\n", Status); 4596 goto end; 4597 } 4598 } else { 4599 Status = set_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, fcb->adsxattr.Buffer, fcb->adsxattr.Length, 4600 fcb->adshash, (UINT8*)fcb->adsdata.Buffer, fcb->adsdata.Length); 4601 if (!NT_SUCCESS(Status)) { 4602 ERR("set_xattr returned %08x\n", Status); 4603 goto end; 4604 } 4605 } 4606 4607 Status = STATUS_SUCCESS; 4608 goto end; 4609 } 4610 4611 if (fcb->deleted) { 4612 Status = insert_tree_item_batch(batchlist, fcb->Vcb, fcb->subvol, fcb->inode, TYPE_INODE_ITEM, 0xffffffffffffffff, NULL, 0, Batch_DeleteInode); 4613 if (!NT_SUCCESS(Status)) { 4614 ERR("insert_tree_item_batch returned %08x\n", Status); 4615 goto end; 4616 } 4617 4618 Status = STATUS_SUCCESS; 4619 goto end; 4620 } 4621 4622 #ifdef DEBUG_PARANOID 4623 extents_changed = fcb->extents_changed; 4624 #endif 4625 4626 if (fcb->extents_changed) { 4627 LIST_ENTRY* le; 4628 BOOL prealloc = FALSE, extents_inline = FALSE; 4629 UINT64 last_end; 4630 4631 // delete ignored extent items 4632 le = fcb->extents.Flink; 4633 while (le != &fcb->extents) { 4634 LIST_ENTRY* le2 = le->Flink; 4635 extent* ext = CONTAINING_RECORD(le, extent, list_entry); 4636 4637 if (ext->ignore) { 4638 RemoveEntryList(&ext->list_entry); 4639 4640 if (ext->csum) 4641 ExFreePool(ext->csum); 4642 4643 ExFreePool(ext); 4644 } 4645 4646 le = le2; 4647 } 4648 4649 le = fcb->extents.Flink; 4650 while (le != &fcb->extents) { 4651 extent* ext = CONTAINING_RECORD(le, extent, list_entry); 4652 4653 if (ext->inserted && ext->csum && ext->extent_data.type == EXTENT_TYPE_REGULAR) { 4654 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ext->extent_data.data; 4655 4656 if (ed2->size > 0) { // not sparse 4657 if (ext->extent_data.compression == BTRFS_COMPRESSION_NONE) 4658 add_checksum_entry(fcb->Vcb, ed2->address + ed2->offset, (ULONG)(ed2->num_bytes / fcb->Vcb->superblock.sector_size), ext->csum, Irp); 4659 else 4660 add_checksum_entry(fcb->Vcb, ed2->address, (ULONG)(ed2->size / fcb->Vcb->superblock.sector_size), ext->csum, Irp); 4661 } 4662 } 4663 4664 le = le->Flink; 4665 } 4666 4667 if (!IsListEmpty(&fcb->extents)) { 4668 rationalize_extents(fcb, Irp); 4669 4670 // merge together adjacent EXTENT_DATAs pointing to same extent 4671 4672 le = fcb->extents.Flink; 4673 while (le != &fcb->extents) { 4674 LIST_ENTRY* le2 = le->Flink; 4675 extent* ext = CONTAINING_RECORD(le, extent, list_entry); 4676 4677 if ((ext->extent_data.type == EXTENT_TYPE_REGULAR || ext->extent_data.type == EXTENT_TYPE_PREALLOC) && le->Flink != &fcb->extents) { 4678 extent* nextext = CONTAINING_RECORD(le->Flink, extent, list_entry); 4679 4680 if (ext->extent_data.type == nextext->extent_data.type) { 4681 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ext->extent_data.data; 4682 EXTENT_DATA2* ned2 = (EXTENT_DATA2*)nextext->extent_data.data; 4683 4684 if (ed2->size != 0 && ed2->address == ned2->address && ed2->size == ned2->size && 4685 nextext->offset == ext->offset + ed2->num_bytes && ned2->offset == ed2->offset + ed2->num_bytes) { 4686 chunk* c; 4687 4688 if (ext->extent_data.compression == BTRFS_COMPRESSION_NONE && ext->csum) { 4689 ULONG len = (ULONG)((ed2->num_bytes + ned2->num_bytes) / fcb->Vcb->superblock.sector_size); 4690 UINT32* csum; 4691 4692 csum = ExAllocatePoolWithTag(NonPagedPool, len * sizeof(UINT32), ALLOC_TAG); 4693 if (!csum) { 4694 ERR("out of memory\n"); 4695 Status = STATUS_INSUFFICIENT_RESOURCES; 4696 goto end; 4697 } 4698 4699 RtlCopyMemory(csum, ext->csum, (ULONG)(ed2->num_bytes * sizeof(UINT32) / fcb->Vcb->superblock.sector_size)); 4700 RtlCopyMemory(&csum[ed2->num_bytes / fcb->Vcb->superblock.sector_size], nextext->csum, 4701 (ULONG)(ned2->num_bytes * sizeof(UINT32) / fcb->Vcb->superblock.sector_size)); 4702 4703 ExFreePool(ext->csum); 4704 ext->csum = csum; 4705 } 4706 4707 ext->extent_data.generation = fcb->Vcb->superblock.generation; 4708 ed2->num_bytes += ned2->num_bytes; 4709 4710 RemoveEntryList(&nextext->list_entry); 4711 4712 if (nextext->csum) 4713 ExFreePool(nextext->csum); 4714 4715 ExFreePool(nextext); 4716 4717 c = get_chunk_from_address(fcb->Vcb, ed2->address); 4718 4719 if (!c) { 4720 ERR("get_chunk_from_address(%llx) failed\n", ed2->address); 4721 } else { 4722 Status = update_changed_extent_ref(fcb->Vcb, c, ed2->address, ed2->size, fcb->subvol->id, fcb->inode, ext->offset - ed2->offset, -1, 4723 fcb->inode_item.flags & BTRFS_INODE_NODATASUM, FALSE, Irp); 4724 if (!NT_SUCCESS(Status)) { 4725 ERR("update_changed_extent_ref returned %08x\n", Status); 4726 goto end; 4727 } 4728 } 4729 4730 le2 = le; 4731 } 4732 } 4733 } 4734 4735 le = le2; 4736 } 4737 } 4738 4739 if (!fcb->created) { 4740 // delete existing EXTENT_DATA items 4741 4742 Status = insert_tree_item_batch(batchlist, fcb->Vcb, fcb->subvol, fcb->inode, TYPE_EXTENT_DATA, 0, NULL, 0, Batch_DeleteExtentData); 4743 if (!NT_SUCCESS(Status)) { 4744 ERR("insert_tree_item_batch returned %08x\n", Status); 4745 goto end; 4746 } 4747 } 4748 4749 // add new EXTENT_DATAs 4750 4751 last_end = 0; 4752 4753 le = fcb->extents.Flink; 4754 while (le != &fcb->extents) { 4755 extent* ext = CONTAINING_RECORD(le, extent, list_entry); 4756 EXTENT_DATA* ed; 4757 4758 ext->inserted = FALSE; 4759 4760 if (!(fcb->Vcb->superblock.incompat_flags & BTRFS_INCOMPAT_FLAGS_NO_HOLES) && ext->offset > last_end) { 4761 Status = insert_sparse_extent(fcb, batchlist, last_end, ext->offset - last_end); 4762 if (!NT_SUCCESS(Status)) { 4763 ERR("insert_sparse_extent returned %08x\n", Status); 4764 goto end; 4765 } 4766 } 4767 4768 ed = ExAllocatePoolWithTag(PagedPool, ext->datalen, ALLOC_TAG); 4769 if (!ed) { 4770 ERR("out of memory\n"); 4771 Status = STATUS_INSUFFICIENT_RESOURCES; 4772 goto end; 4773 } 4774 4775 RtlCopyMemory(ed, &ext->extent_data, ext->datalen); 4776 4777 Status = insert_tree_item_batch(batchlist, fcb->Vcb, fcb->subvol, fcb->inode, TYPE_EXTENT_DATA, ext->offset, 4778 ed, ext->datalen, Batch_Insert); 4779 if (!NT_SUCCESS(Status)) { 4780 ERR("insert_tree_item_batch returned %08x\n", Status); 4781 goto end; 4782 } 4783 4784 if (ed->type == EXTENT_TYPE_PREALLOC) 4785 prealloc = TRUE; 4786 4787 if (ed->type == EXTENT_TYPE_INLINE) 4788 extents_inline = TRUE; 4789 4790 if (!(fcb->Vcb->superblock.incompat_flags & BTRFS_INCOMPAT_FLAGS_NO_HOLES)) { 4791 if (ed->type == EXTENT_TYPE_INLINE) 4792 last_end = ext->offset + ed->decoded_size; 4793 else { 4794 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ed->data; 4795 4796 last_end = ext->offset + ed2->num_bytes; 4797 } 4798 } 4799 4800 le = le->Flink; 4801 } 4802 4803 if (!(fcb->Vcb->superblock.incompat_flags & BTRFS_INCOMPAT_FLAGS_NO_HOLES) && !extents_inline && 4804 sector_align(fcb->inode_item.st_size, fcb->Vcb->superblock.sector_size) > last_end) { 4805 Status = insert_sparse_extent(fcb, batchlist, last_end, sector_align(fcb->inode_item.st_size, fcb->Vcb->superblock.sector_size) - last_end); 4806 if (!NT_SUCCESS(Status)) { 4807 ERR("insert_sparse_extent returned %08x\n", Status); 4808 goto end; 4809 } 4810 } 4811 4812 // update prealloc flag in INODE_ITEM 4813 4814 if (!prealloc) 4815 fcb->inode_item.flags &= ~BTRFS_INODE_PREALLOC; 4816 else 4817 fcb->inode_item.flags |= BTRFS_INODE_PREALLOC; 4818 4819 fcb->inode_item_changed = TRUE; 4820 4821 fcb->extents_changed = FALSE; 4822 } 4823 4824 if ((!fcb->created && fcb->inode_item_changed) || cache) { 4825 searchkey.obj_id = fcb->inode; 4826 searchkey.obj_type = TYPE_INODE_ITEM; 4827 searchkey.offset = 0xffffffffffffffff; 4828 4829 Status = find_item(fcb->Vcb, fcb->subvol, &tp, &searchkey, FALSE, Irp); 4830 if (!NT_SUCCESS(Status)) { 4831 ERR("error - find_item returned %08x\n", Status); 4832 goto end; 4833 } 4834 4835 if (tp.item->key.obj_id != searchkey.obj_id || tp.item->key.obj_type != searchkey.obj_type) { 4836 if (cache) { 4837 ii = ExAllocatePoolWithTag(PagedPool, sizeof(INODE_ITEM), ALLOC_TAG); 4838 if (!ii) { 4839 ERR("out of memory\n"); 4840 Status = STATUS_INSUFFICIENT_RESOURCES; 4841 goto end; 4842 } 4843 4844 RtlCopyMemory(ii, &fcb->inode_item, sizeof(INODE_ITEM)); 4845 4846 Status = insert_tree_item(fcb->Vcb, fcb->subvol, fcb->inode, TYPE_INODE_ITEM, 0, ii, sizeof(INODE_ITEM), NULL, Irp); 4847 if (!NT_SUCCESS(Status)) { 4848 ERR("insert_tree_item returned %08x\n", Status); 4849 goto end; 4850 } 4851 4852 ii_offset = 0; 4853 } else { 4854 ERR("could not find INODE_ITEM for inode %llx in subvol %llx\n", fcb->inode, fcb->subvol->id); 4855 Status = STATUS_INTERNAL_ERROR; 4856 goto end; 4857 } 4858 } else { 4859 #ifdef DEBUG_PARANOID 4860 INODE_ITEM* ii2 = (INODE_ITEM*)tp.item->data; 4861 4862 old_size = ii2->st_size; 4863 #endif 4864 4865 ii_offset = tp.item->key.offset; 4866 } 4867 4868 if (!cache) { 4869 Status = delete_tree_item(fcb->Vcb, &tp); 4870 if (!NT_SUCCESS(Status)) { 4871 ERR("delete_tree_item returned %08x\n", Status); 4872 goto end; 4873 } 4874 } else { 4875 searchkey.obj_id = fcb->inode; 4876 searchkey.obj_type = TYPE_INODE_ITEM; 4877 searchkey.offset = ii_offset; 4878 4879 Status = find_item(fcb->Vcb, fcb->subvol, &tp, &searchkey, FALSE, Irp); 4880 if (!NT_SUCCESS(Status)) { 4881 ERR("error - find_item returned %08x\n", Status); 4882 goto end; 4883 } 4884 4885 if (keycmp(tp.item->key, searchkey)) { 4886 ERR("could not find INODE_ITEM for inode %llx in subvol %llx\n", fcb->inode, fcb->subvol->id); 4887 Status = STATUS_INTERNAL_ERROR; 4888 goto end; 4889 } else 4890 RtlCopyMemory(tp.item->data, &fcb->inode_item, min(tp.item->size, sizeof(INODE_ITEM))); 4891 } 4892 4893 #ifdef DEBUG_PARANOID 4894 if (!extents_changed && fcb->type != BTRFS_TYPE_DIRECTORY && old_size != fcb->inode_item.st_size) { 4895 ERR("error - size has changed but extents not marked as changed\n"); 4896 int3; 4897 } 4898 #endif 4899 } else 4900 ii_offset = 0; 4901 4902 fcb->created = FALSE; 4903 4904 if (!cache && fcb->inode_item_changed) { 4905 ii = ExAllocatePoolWithTag(PagedPool, sizeof(INODE_ITEM), ALLOC_TAG); 4906 if (!ii) { 4907 ERR("out of memory\n"); 4908 Status = STATUS_INSUFFICIENT_RESOURCES; 4909 goto end; 4910 } 4911 4912 RtlCopyMemory(ii, &fcb->inode_item, sizeof(INODE_ITEM)); 4913 4914 Status = insert_tree_item_batch(batchlist, fcb->Vcb, fcb->subvol, fcb->inode, TYPE_INODE_ITEM, ii_offset, ii, sizeof(INODE_ITEM), 4915 Batch_Insert); 4916 if (!NT_SUCCESS(Status)) { 4917 ERR("insert_tree_item_batch returned %08x\n", Status); 4918 goto end; 4919 } 4920 4921 fcb->inode_item_changed = FALSE; 4922 } 4923 4924 if (fcb->sd_dirty) { 4925 if (!fcb->sd_deleted) { 4926 Status = set_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_NTACL, (UINT16)strlen(EA_NTACL), 4927 EA_NTACL_HASH, (UINT8*)fcb->sd, (UINT16)RtlLengthSecurityDescriptor(fcb->sd)); 4928 if (!NT_SUCCESS(Status)) { 4929 ERR("set_xattr returned %08x\n", Status); 4930 goto end; 4931 } 4932 } else { 4933 Status = delete_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_NTACL, (UINT16)strlen(EA_NTACL), EA_NTACL_HASH); 4934 if (!NT_SUCCESS(Status)) { 4935 ERR("delete_xattr returned %08x\n", Status); 4936 goto end; 4937 } 4938 } 4939 4940 fcb->sd_deleted = FALSE; 4941 fcb->sd_dirty = FALSE; 4942 } 4943 4944 if (fcb->atts_changed) { 4945 if (!fcb->atts_deleted) { 4946 UINT8 val[16], *val2; 4947 ULONG atts = fcb->atts; 4948 4949 TRACE("inserting new DOSATTRIB xattr\n"); 4950 4951 if (fcb->inode == SUBVOL_ROOT_INODE) 4952 atts &= ~FILE_ATTRIBUTE_READONLY; 4953 4954 val2 = &val[sizeof(val) - 1]; 4955 4956 do { 4957 UINT8 c = atts % 16; 4958 *val2 = c <= 9 ? (c + '0') : (c - 0xa + 'a'); 4959 4960 val2--; 4961 atts >>= 4; 4962 } while (atts != 0); 4963 4964 *val2 = 'x'; 4965 val2--; 4966 *val2 = '0'; 4967 4968 Status = set_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_DOSATTRIB, (UINT16)strlen(EA_DOSATTRIB), 4969 EA_DOSATTRIB_HASH, val2, (UINT16)(val + sizeof(val) - val2)); 4970 if (!NT_SUCCESS(Status)) { 4971 ERR("set_xattr returned %08x\n", Status); 4972 goto end; 4973 } 4974 } else { 4975 Status = delete_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_DOSATTRIB, (UINT16)strlen(EA_DOSATTRIB), EA_DOSATTRIB_HASH); 4976 if (!NT_SUCCESS(Status)) { 4977 ERR("delete_xattr returned %08x\n", Status); 4978 goto end; 4979 } 4980 } 4981 4982 fcb->atts_changed = FALSE; 4983 fcb->atts_deleted = FALSE; 4984 } 4985 4986 if (fcb->reparse_xattr_changed) { 4987 if (fcb->reparse_xattr.Buffer && fcb->reparse_xattr.Length > 0) { 4988 Status = set_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_REPARSE, (UINT16)strlen(EA_REPARSE), 4989 EA_REPARSE_HASH, (UINT8*)fcb->reparse_xattr.Buffer, (UINT16)fcb->reparse_xattr.Length); 4990 if (!NT_SUCCESS(Status)) { 4991 ERR("set_xattr returned %08x\n", Status); 4992 goto end; 4993 } 4994 } else { 4995 Status = delete_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_REPARSE, (UINT16)strlen(EA_REPARSE), EA_REPARSE_HASH); 4996 if (!NT_SUCCESS(Status)) { 4997 ERR("delete_xattr returned %08x\n", Status); 4998 goto end; 4999 } 5000 } 5001 5002 fcb->reparse_xattr_changed = FALSE; 5003 } 5004 5005 if (fcb->ea_changed) { 5006 if (fcb->ea_xattr.Buffer && fcb->ea_xattr.Length > 0) { 5007 Status = set_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_EA, (UINT16)strlen(EA_EA), 5008 EA_EA_HASH, (UINT8*)fcb->ea_xattr.Buffer, (UINT16)fcb->ea_xattr.Length); 5009 if (!NT_SUCCESS(Status)) { 5010 ERR("set_xattr returned %08x\n", Status); 5011 goto end; 5012 } 5013 } else { 5014 Status = delete_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_EA, (UINT16)strlen(EA_EA), EA_EA_HASH); 5015 if (!NT_SUCCESS(Status)) { 5016 ERR("delete_xattr returned %08x\n", Status); 5017 goto end; 5018 } 5019 } 5020 5021 fcb->ea_changed = FALSE; 5022 } 5023 5024 if (fcb->prop_compression_changed) { 5025 if (fcb->prop_compression == PropCompression_None) { 5026 Status = delete_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_PROP_COMPRESSION, (UINT16)strlen(EA_PROP_COMPRESSION), EA_PROP_COMPRESSION_HASH); 5027 if (!NT_SUCCESS(Status)) { 5028 ERR("delete_xattr returned %08x\n", Status); 5029 goto end; 5030 } 5031 } else if (fcb->prop_compression == PropCompression_Zlib) { 5032 const char zlib[] = "zlib"; 5033 5034 Status = set_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_PROP_COMPRESSION, (UINT16)strlen(EA_PROP_COMPRESSION), 5035 EA_PROP_COMPRESSION_HASH, (UINT8*)zlib, (UINT16)strlen(zlib)); 5036 if (!NT_SUCCESS(Status)) { 5037 ERR("set_xattr returned %08x\n", Status); 5038 goto end; 5039 } 5040 } else if (fcb->prop_compression == PropCompression_LZO) { 5041 const char lzo[] = "lzo"; 5042 5043 Status = set_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, EA_PROP_COMPRESSION, (UINT16)strlen(EA_PROP_COMPRESSION), 5044 EA_PROP_COMPRESSION_HASH, (UINT8*)lzo, (UINT16)strlen(lzo)); 5045 if (!NT_SUCCESS(Status)) { 5046 ERR("set_xattr returned %08x\n", Status); 5047 goto end; 5048 } 5049 } 5050 5051 fcb->prop_compression_changed = FALSE; 5052 } 5053 5054 if (fcb->xattrs_changed) { 5055 LIST_ENTRY* le; 5056 5057 le = fcb->xattrs.Flink; 5058 while (le != &fcb->xattrs) { 5059 xattr* xa = CONTAINING_RECORD(le, xattr, list_entry); 5060 LIST_ENTRY* le2 = le->Flink; 5061 5062 if (xa->dirty) { 5063 UINT32 hash = calc_crc32c(0xfffffffe, (UINT8*)xa->data, xa->namelen); 5064 5065 if (xa->valuelen == 0) { 5066 Status = delete_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, xa->data, xa->namelen, hash); 5067 if (!NT_SUCCESS(Status)) { 5068 ERR("delete_xattr returned %08x\n", Status); 5069 goto end; 5070 } 5071 5072 RemoveEntryList(&xa->list_entry); 5073 ExFreePool(xa); 5074 } else { 5075 Status = set_xattr(fcb->Vcb, batchlist, fcb->subvol, fcb->inode, xa->data, xa->namelen, 5076 hash, (UINT8*)&xa->data[xa->namelen], xa->valuelen); 5077 if (!NT_SUCCESS(Status)) { 5078 ERR("set_xattr returned %08x\n", Status); 5079 goto end; 5080 } 5081 5082 xa->dirty = FALSE; 5083 } 5084 } 5085 5086 le = le2; 5087 } 5088 5089 fcb->xattrs_changed = FALSE; 5090 } 5091 5092 Status = STATUS_SUCCESS; 5093 5094 end: 5095 if (fcb->dirty) { 5096 BOOL lock = FALSE; 5097 5098 fcb->dirty = FALSE; 5099 5100 if (!ExIsResourceAcquiredExclusiveLite(&fcb->Vcb->dirty_fcbs_lock)) { 5101 ExAcquireResourceExclusiveLite(&fcb->Vcb->dirty_fcbs_lock, TRUE); 5102 lock = TRUE; 5103 } 5104 5105 RemoveEntryList(&fcb->list_entry_dirty); 5106 5107 if (lock) 5108 ExReleaseResourceLite(&fcb->Vcb->dirty_fcbs_lock); 5109 } 5110 5111 return Status; 5112 } 5113 5114 void add_trim_entry_avoid_sb(device_extension* Vcb, device* dev, UINT64 address, UINT64 size) { 5115 int i; 5116 ULONG sblen = (ULONG)sector_align(sizeof(superblock), Vcb->superblock.sector_size); 5117 5118 i = 0; 5119 while (superblock_addrs[i] != 0) { 5120 if (superblock_addrs[i] + sblen >= address && superblock_addrs[i] < address + size) { 5121 if (superblock_addrs[i] > address) 5122 add_trim_entry(dev, address, superblock_addrs[i] - address); 5123 5124 if (size <= superblock_addrs[i] + sblen - address) 5125 return; 5126 5127 size -= superblock_addrs[i] + sblen - address; 5128 address = superblock_addrs[i] + sblen; 5129 } else if (superblock_addrs[i] > address + size) 5130 break; 5131 5132 i++; 5133 } 5134 5135 add_trim_entry(dev, address, size); 5136 } 5137 5138 static NTSTATUS drop_chunk(device_extension* Vcb, chunk* c, LIST_ENTRY* batchlist, PIRP Irp, LIST_ENTRY* rollback) { 5139 NTSTATUS Status; 5140 KEY searchkey; 5141 traverse_ptr tp; 5142 UINT64 i, factor; 5143 CHUNK_ITEM_STRIPE* cis = (CHUNK_ITEM_STRIPE*)&c->chunk_item[1];; 5144 5145 TRACE("dropping chunk %llx\n", c->offset); 5146 5147 if (c->chunk_item->type & BLOCK_FLAG_RAID0) 5148 factor = c->chunk_item->num_stripes; 5149 else if (c->chunk_item->type & BLOCK_FLAG_RAID10) 5150 factor = c->chunk_item->num_stripes / c->chunk_item->sub_stripes; 5151 else if (c->chunk_item->type & BLOCK_FLAG_RAID5) 5152 factor = c->chunk_item->num_stripes - 1; 5153 else if (c->chunk_item->type & BLOCK_FLAG_RAID6) 5154 factor = c->chunk_item->num_stripes - 2; 5155 else // SINGLE, DUPLICATE, RAID1 5156 factor = 1; 5157 5158 // do TRIM 5159 if (Vcb->trim && !Vcb->options.no_trim) { 5160 UINT64 len = c->chunk_item->size / factor; 5161 5162 for (i = 0; i < c->chunk_item->num_stripes; i++) { 5163 if (c->devices[i] && c->devices[i]->devobj && !c->devices[i]->readonly && c->devices[i]->trim) 5164 add_trim_entry_avoid_sb(Vcb, c->devices[i], cis[i].offset, len); 5165 } 5166 } 5167 5168 if (!c->cache) { 5169 Status = load_stored_free_space_cache(Vcb, c, TRUE, Irp); 5170 5171 if (!NT_SUCCESS(Status) && Status != STATUS_NOT_FOUND) 5172 WARN("load_stored_free_space_cache returned %08x\n", Status); 5173 } 5174 5175 // remove free space cache 5176 if (c->cache) { 5177 c->cache->deleted = TRUE; 5178 5179 Status = excise_extents(Vcb, c->cache, 0, c->cache->inode_item.st_size, Irp, rollback); 5180 if (!NT_SUCCESS(Status)) { 5181 ERR("excise_extents returned %08x\n", Status); 5182 return Status; 5183 } 5184 5185 Status = flush_fcb(c->cache, TRUE, batchlist, Irp); 5186 5187 free_fcb(Vcb, c->cache); 5188 5189 if (!NT_SUCCESS(Status)) { 5190 ERR("flush_fcb returned %08x\n", Status); 5191 return Status; 5192 } 5193 5194 searchkey.obj_id = FREE_SPACE_CACHE_ID; 5195 searchkey.obj_type = 0; 5196 searchkey.offset = c->offset; 5197 5198 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 5199 if (!NT_SUCCESS(Status)) { 5200 ERR("error - find_item returned %08x\n", Status); 5201 return Status; 5202 } 5203 5204 if (!keycmp(tp.item->key, searchkey)) { 5205 Status = delete_tree_item(Vcb, &tp); 5206 if (!NT_SUCCESS(Status)) { 5207 ERR("delete_tree_item returned %08x\n", Status); 5208 return Status; 5209 } 5210 } 5211 } 5212 5213 if (Vcb->space_root) { 5214 Status = insert_tree_item_batch(batchlist, Vcb, Vcb->space_root, c->offset, TYPE_FREE_SPACE_INFO, c->chunk_item->size, 5215 NULL, 0, Batch_DeleteFreeSpace); 5216 if (!NT_SUCCESS(Status)) { 5217 ERR("insert_tree_item_batch returned %08x\n", Status); 5218 return Status; 5219 } 5220 } 5221 5222 for (i = 0; i < c->chunk_item->num_stripes; i++) { 5223 if (!c->created) { 5224 // remove DEV_EXTENTs from tree 4 5225 searchkey.obj_id = cis[i].dev_id; 5226 searchkey.obj_type = TYPE_DEV_EXTENT; 5227 searchkey.offset = cis[i].offset; 5228 5229 Status = find_item(Vcb, Vcb->dev_root, &tp, &searchkey, FALSE, Irp); 5230 if (!NT_SUCCESS(Status)) { 5231 ERR("error - find_item returned %08x\n", Status); 5232 return Status; 5233 } 5234 5235 if (!keycmp(tp.item->key, searchkey)) { 5236 Status = delete_tree_item(Vcb, &tp); 5237 if (!NT_SUCCESS(Status)) { 5238 ERR("delete_tree_item returned %08x\n", Status); 5239 return Status; 5240 } 5241 5242 if (tp.item->size >= sizeof(DEV_EXTENT)) { 5243 DEV_EXTENT* de = (DEV_EXTENT*)tp.item->data; 5244 5245 c->devices[i]->devitem.bytes_used -= de->length; 5246 5247 if (Vcb->balance.thread && Vcb->balance.shrinking && Vcb->balance.opts[0].devid == c->devices[i]->devitem.dev_id) { 5248 if (cis[i].offset < Vcb->balance.opts[0].drange_start && cis[i].offset + de->length > Vcb->balance.opts[0].drange_start) 5249 space_list_add2(&c->devices[i]->space, NULL, cis[i].offset, Vcb->balance.opts[0].drange_start - cis[i].offset, NULL, rollback); 5250 } else 5251 space_list_add2(&c->devices[i]->space, NULL, cis[i].offset, de->length, NULL, rollback); 5252 } 5253 } else 5254 WARN("could not find (%llx,%x,%llx) in dev tree\n", searchkey.obj_id, searchkey.obj_type, searchkey.offset); 5255 } else { 5256 UINT64 len = c->chunk_item->size / factor; 5257 5258 c->devices[i]->devitem.bytes_used -= len; 5259 5260 if (Vcb->balance.thread && Vcb->balance.shrinking && Vcb->balance.opts[0].devid == c->devices[i]->devitem.dev_id) { 5261 if (cis[i].offset < Vcb->balance.opts[0].drange_start && cis[i].offset + len > Vcb->balance.opts[0].drange_start) 5262 space_list_add2(&c->devices[i]->space, NULL, cis[i].offset, Vcb->balance.opts[0].drange_start - cis[i].offset, NULL, rollback); 5263 } else 5264 space_list_add2(&c->devices[i]->space, NULL, cis[i].offset, len, NULL, rollback); 5265 } 5266 } 5267 5268 // modify DEV_ITEMs in chunk tree 5269 for (i = 0; i < c->chunk_item->num_stripes; i++) { 5270 if (c->devices[i]) { 5271 UINT64 j; 5272 DEV_ITEM* di; 5273 5274 searchkey.obj_id = 1; 5275 searchkey.obj_type = TYPE_DEV_ITEM; 5276 searchkey.offset = c->devices[i]->devitem.dev_id; 5277 5278 Status = find_item(Vcb, Vcb->chunk_root, &tp, &searchkey, FALSE, Irp); 5279 if (!NT_SUCCESS(Status)) { 5280 ERR("error - find_item returned %08x\n", Status); 5281 return Status; 5282 } 5283 5284 if (keycmp(tp.item->key, searchkey)) { 5285 ERR("error - could not find DEV_ITEM for device %llx\n", searchkey.offset); 5286 return STATUS_INTERNAL_ERROR; 5287 } 5288 5289 Status = delete_tree_item(Vcb, &tp); 5290 if (!NT_SUCCESS(Status)) { 5291 ERR("delete_tree_item returned %08x\n", Status); 5292 return Status; 5293 } 5294 5295 di = ExAllocatePoolWithTag(PagedPool, sizeof(DEV_ITEM), ALLOC_TAG); 5296 if (!di) { 5297 ERR("out of memory\n"); 5298 return STATUS_INSUFFICIENT_RESOURCES; 5299 } 5300 5301 RtlCopyMemory(di, &c->devices[i]->devitem, sizeof(DEV_ITEM)); 5302 5303 Status = insert_tree_item(Vcb, Vcb->chunk_root, 1, TYPE_DEV_ITEM, c->devices[i]->devitem.dev_id, di, sizeof(DEV_ITEM), NULL, Irp); 5304 if (!NT_SUCCESS(Status)) { 5305 ERR("insert_tree_item returned %08x\n", Status); 5306 return Status; 5307 } 5308 5309 for (j = i + 1; j < c->chunk_item->num_stripes; j++) { 5310 if (c->devices[j] == c->devices[i]) 5311 c->devices[j] = NULL; 5312 } 5313 } 5314 } 5315 5316 if (!c->created) { 5317 // remove CHUNK_ITEM from chunk tree 5318 searchkey.obj_id = 0x100; 5319 searchkey.obj_type = TYPE_CHUNK_ITEM; 5320 searchkey.offset = c->offset; 5321 5322 Status = find_item(Vcb, Vcb->chunk_root, &tp, &searchkey, FALSE, Irp); 5323 if (!NT_SUCCESS(Status)) { 5324 ERR("error - find_item returned %08x\n", Status); 5325 return Status; 5326 } 5327 5328 if (!keycmp(tp.item->key, searchkey)) { 5329 Status = delete_tree_item(Vcb, &tp); 5330 5331 if (!NT_SUCCESS(Status)) { 5332 ERR("delete_tree_item returned %08x\n", Status); 5333 return Status; 5334 } 5335 } else 5336 WARN("could not find CHUNK_ITEM for chunk %llx\n", c->offset); 5337 5338 // remove BLOCK_GROUP_ITEM from extent tree 5339 searchkey.obj_id = c->offset; 5340 searchkey.obj_type = TYPE_BLOCK_GROUP_ITEM; 5341 searchkey.offset = 0xffffffffffffffff; 5342 5343 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 5344 if (!NT_SUCCESS(Status)) { 5345 ERR("error - find_item returned %08x\n", Status); 5346 return Status; 5347 } 5348 5349 if (tp.item->key.obj_id == searchkey.obj_id && tp.item->key.obj_type == searchkey.obj_type) { 5350 Status = delete_tree_item(Vcb, &tp); 5351 5352 if (!NT_SUCCESS(Status)) { 5353 ERR("delete_tree_item returned %08x\n", Status); 5354 return Status; 5355 } 5356 } else 5357 WARN("could not find BLOCK_GROUP_ITEM for chunk %llx\n", c->offset); 5358 } 5359 5360 if (c->chunk_item->type & BLOCK_FLAG_SYSTEM) 5361 remove_from_bootstrap(Vcb, 0x100, TYPE_CHUNK_ITEM, c->offset); 5362 5363 RemoveEntryList(&c->list_entry); 5364 5365 // clear raid56 incompat flag if dropping last RAID5/6 chunk 5366 5367 if (c->chunk_item->type & BLOCK_FLAG_RAID5 || c->chunk_item->type & BLOCK_FLAG_RAID6) { 5368 LIST_ENTRY* le; 5369 BOOL clear_flag = TRUE; 5370 5371 le = Vcb->chunks.Flink; 5372 while (le != &Vcb->chunks) { 5373 chunk* c2 = CONTAINING_RECORD(le, chunk, list_entry); 5374 5375 if (c2->chunk_item->type & BLOCK_FLAG_RAID5 || c2->chunk_item->type & BLOCK_FLAG_RAID6) { 5376 clear_flag = FALSE; 5377 break; 5378 } 5379 5380 le = le->Flink; 5381 } 5382 5383 if (clear_flag) 5384 Vcb->superblock.incompat_flags &= ~BTRFS_INCOMPAT_FLAGS_RAID56; 5385 } 5386 5387 Vcb->superblock.bytes_used -= c->oldused; 5388 5389 ExFreePool(c->chunk_item); 5390 ExFreePool(c->devices); 5391 5392 while (!IsListEmpty(&c->space)) { 5393 space* s = CONTAINING_RECORD(c->space.Flink, space, list_entry); 5394 5395 RemoveEntryList(&s->list_entry); 5396 ExFreePool(s); 5397 } 5398 5399 while (!IsListEmpty(&c->deleting)) { 5400 space* s = CONTAINING_RECORD(c->deleting.Flink, space, list_entry); 5401 5402 RemoveEntryList(&s->list_entry); 5403 ExFreePool(s); 5404 } 5405 5406 ExDeleteResourceLite(&c->partial_stripes_lock); 5407 ExDeleteResourceLite(&c->range_locks_lock); 5408 ExDeleteResourceLite(&c->lock); 5409 ExDeleteResourceLite(&c->changed_extents_lock); 5410 5411 ExFreePool(c); 5412 5413 return STATUS_SUCCESS; 5414 } 5415 5416 static NTSTATUS partial_stripe_read(device_extension* Vcb, chunk* c, partial_stripe* ps, UINT64 startoff, UINT16 parity, ULONG offset, ULONG len) { 5417 NTSTATUS Status; 5418 ULONG sl = (ULONG)(c->chunk_item->stripe_length / Vcb->superblock.sector_size); 5419 CHUNK_ITEM_STRIPE* cis = (CHUNK_ITEM_STRIPE*)&c->chunk_item[1]; 5420 5421 while (len > 0) { 5422 ULONG readlen = min(offset + len, offset + (sl - (offset % sl))) - offset; 5423 UINT16 stripe; 5424 5425 stripe = (parity + (offset / sl) + 1) % c->chunk_item->num_stripes; 5426 5427 if (c->devices[stripe]->devobj) { 5428 Status = sync_read_phys(c->devices[stripe]->devobj, cis[stripe].offset + startoff + ((offset % sl) * Vcb->superblock.sector_size), 5429 readlen * Vcb->superblock.sector_size, ps->data + (offset * Vcb->superblock.sector_size), FALSE); 5430 if (!NT_SUCCESS(Status)) { 5431 ERR("sync_read_phys returned %08x\n", Status); 5432 return Status; 5433 } 5434 } else if (c->chunk_item->type & BLOCK_FLAG_RAID5) { 5435 UINT16 i; 5436 UINT8* scratch; 5437 5438 scratch = ExAllocatePoolWithTag(NonPagedPool, readlen * Vcb->superblock.sector_size, ALLOC_TAG); 5439 if (!scratch) { 5440 ERR("out of memory\n"); 5441 return STATUS_INSUFFICIENT_RESOURCES; 5442 } 5443 5444 for (i = 0; i < c->chunk_item->num_stripes; i++) { 5445 if (i != stripe) { 5446 if (!c->devices[i]->devobj) { 5447 ExFreePool(scratch); 5448 return STATUS_UNEXPECTED_IO_ERROR; 5449 } 5450 5451 if (i == 0 || (stripe == 0 && i == 1)) { 5452 Status = sync_read_phys(c->devices[i]->devobj, cis[i].offset + startoff + ((offset % sl) * Vcb->superblock.sector_size), 5453 readlen * Vcb->superblock.sector_size, ps->data + (offset * Vcb->superblock.sector_size), FALSE); 5454 if (!NT_SUCCESS(Status)) { 5455 ERR("sync_read_phys returned %08x\n", Status); 5456 ExFreePool(scratch); 5457 return Status; 5458 } 5459 } else { 5460 Status = sync_read_phys(c->devices[i]->devobj, cis[i].offset + startoff + ((offset % sl) * Vcb->superblock.sector_size), 5461 readlen * Vcb->superblock.sector_size, scratch, FALSE); 5462 if (!NT_SUCCESS(Status)) { 5463 ERR("sync_read_phys returned %08x\n", Status); 5464 ExFreePool(scratch); 5465 return Status; 5466 } 5467 5468 do_xor(ps->data + (offset * Vcb->superblock.sector_size), scratch, readlen * Vcb->superblock.sector_size); 5469 } 5470 } 5471 } 5472 5473 ExFreePool(scratch); 5474 } else { 5475 UINT8* scratch; 5476 UINT16 k, i, logstripe, error_stripe, num_errors = 0; 5477 5478 scratch = ExAllocatePoolWithTag(NonPagedPool, (c->chunk_item->num_stripes + 2) * readlen * Vcb->superblock.sector_size, ALLOC_TAG); 5479 if (!scratch) { 5480 ERR("out of memory\n"); 5481 return STATUS_INSUFFICIENT_RESOURCES; 5482 } 5483 5484 i = (parity + 1) % c->chunk_item->num_stripes; 5485 for (k = 0; k < c->chunk_item->num_stripes; k++) { 5486 if (i != stripe) { 5487 if (c->devices[i]->devobj) { 5488 Status = sync_read_phys(c->devices[i]->devobj, cis[i].offset + startoff + ((offset % sl) * Vcb->superblock.sector_size), 5489 readlen * Vcb->superblock.sector_size, scratch + (k * readlen * Vcb->superblock.sector_size), FALSE); 5490 if (!NT_SUCCESS(Status)) { 5491 ERR("sync_read_phys returned %08x\n", Status); 5492 num_errors++; 5493 error_stripe = k; 5494 } 5495 } else { 5496 num_errors++; 5497 error_stripe = k; 5498 } 5499 5500 if (num_errors > 1) { 5501 ExFreePool(scratch); 5502 return STATUS_UNEXPECTED_IO_ERROR; 5503 } 5504 } else 5505 logstripe = k; 5506 5507 i = (i + 1) % c->chunk_item->num_stripes; 5508 } 5509 5510 if (num_errors == 0 || error_stripe == c->chunk_item->num_stripes - 1) { 5511 for (k = 0; k < c->chunk_item->num_stripes - 1; k++) { 5512 if (k != logstripe) { 5513 if (k == 0 || (k == 1 && logstripe == 0)) { 5514 RtlCopyMemory(ps->data + (offset * Vcb->superblock.sector_size), scratch + (k * readlen * Vcb->superblock.sector_size), 5515 readlen * Vcb->superblock.sector_size); 5516 } else { 5517 do_xor(ps->data + (offset * Vcb->superblock.sector_size), scratch + (k * readlen * Vcb->superblock.sector_size), 5518 readlen * Vcb->superblock.sector_size); 5519 } 5520 } 5521 } 5522 } else { 5523 raid6_recover2(scratch, c->chunk_item->num_stripes, readlen * Vcb->superblock.sector_size, logstripe, 5524 error_stripe, scratch + (c->chunk_item->num_stripes * readlen * Vcb->superblock.sector_size)); 5525 5526 RtlCopyMemory(ps->data + (offset * Vcb->superblock.sector_size), scratch + (c->chunk_item->num_stripes * readlen * Vcb->superblock.sector_size), 5527 readlen * Vcb->superblock.sector_size); 5528 } 5529 5530 ExFreePool(scratch); 5531 } 5532 5533 offset += readlen; 5534 len -= readlen; 5535 } 5536 5537 return STATUS_SUCCESS; 5538 } 5539 5540 NTSTATUS flush_partial_stripe(device_extension* Vcb, chunk* c, partial_stripe* ps) { 5541 NTSTATUS Status; 5542 UINT16 parity2, stripe, startoffstripe; 5543 UINT8* data; 5544 UINT64 startoff; 5545 ULONG runlength, index, last1; 5546 CHUNK_ITEM_STRIPE* cis = (CHUNK_ITEM_STRIPE*)&c->chunk_item[1]; 5547 LIST_ENTRY* le; 5548 UINT16 k, num_data_stripes = c->chunk_item->num_stripes - (c->chunk_item->type & BLOCK_FLAG_RAID5 ? 1 : 2); 5549 UINT64 ps_length = num_data_stripes * c->chunk_item->stripe_length; 5550 ULONG stripe_length = (ULONG)c->chunk_item->stripe_length; 5551 5552 // FIXME - do writes asynchronously? 5553 5554 get_raid0_offset(ps->address - c->offset, stripe_length, num_data_stripes, &startoff, &startoffstripe); 5555 5556 parity2 = (((ps->address - c->offset) / ps_length) + c->chunk_item->num_stripes - 1) % c->chunk_item->num_stripes; 5557 5558 // read data (or reconstruct if degraded) 5559 5560 runlength = RtlFindFirstRunClear(&ps->bmp, &index); 5561 last1 = 0; 5562 5563 while (runlength != 0) { 5564 if (index > last1) { 5565 Status = partial_stripe_read(Vcb, c, ps, startoff, parity2, last1, index - last1); 5566 if (!NT_SUCCESS(Status)) { 5567 ERR("partial_stripe_read returned %08x\n", Status); 5568 return Status; 5569 } 5570 } 5571 5572 last1 = index + runlength; 5573 5574 runlength = RtlFindNextForwardRunClear(&ps->bmp, index + runlength, &index); 5575 } 5576 5577 if (last1 < ps_length / Vcb->superblock.sector_size) { 5578 Status = partial_stripe_read(Vcb, c, ps, startoff, parity2, last1, (ULONG)((ps_length / Vcb->superblock.sector_size) - last1)); 5579 if (!NT_SUCCESS(Status)) { 5580 ERR("partial_stripe_read returned %08x\n", Status); 5581 return Status; 5582 } 5583 } 5584 5585 // set unallocated data to 0 5586 le = c->space.Flink; 5587 while (le != &c->space) { 5588 space* s = CONTAINING_RECORD(le, space, list_entry); 5589 5590 if (s->address + s->size > ps->address && s->address < ps->address + ps_length) { 5591 UINT64 start = max(ps->address, s->address); 5592 UINT64 end = min(ps->address + ps_length, s->address + s->size); 5593 5594 RtlZeroMemory(ps->data + start - ps->address, (ULONG)(end - start)); 5595 } else if (s->address >= ps->address + ps_length) 5596 break; 5597 5598 le = le->Flink; 5599 } 5600 5601 le = c->deleting.Flink; 5602 while (le != &c->deleting) { 5603 space* s = CONTAINING_RECORD(le, space, list_entry); 5604 5605 if (s->address + s->size > ps->address && s->address < ps->address + ps_length) { 5606 UINT64 start = max(ps->address, s->address); 5607 UINT64 end = min(ps->address + ps_length, s->address + s->size); 5608 5609 RtlZeroMemory(ps->data + start - ps->address, (ULONG)(end - start)); 5610 } else if (s->address >= ps->address + ps_length) 5611 break; 5612 5613 le = le->Flink; 5614 } 5615 5616 stripe = (parity2 + 1) % c->chunk_item->num_stripes; 5617 5618 data = ps->data; 5619 for (k = 0; k < num_data_stripes; k++) { 5620 if (c->devices[stripe]->devobj) { 5621 Status = write_data_phys(c->devices[stripe]->devobj, cis[stripe].offset + startoff, data, stripe_length); 5622 if (!NT_SUCCESS(Status)) { 5623 ERR("write_data_phys returned %08x\n", Status); 5624 return Status; 5625 } 5626 } 5627 5628 data += stripe_length; 5629 stripe = (stripe + 1) % c->chunk_item->num_stripes; 5630 } 5631 5632 // write parity 5633 if (c->chunk_item->type & BLOCK_FLAG_RAID5) { 5634 if (c->devices[parity2]->devobj) { 5635 UINT16 i; 5636 5637 for (i = 1; i < c->chunk_item->num_stripes - 1; i++) { 5638 do_xor(ps->data, ps->data + (i * stripe_length), stripe_length); 5639 } 5640 5641 Status = write_data_phys(c->devices[parity2]->devobj, cis[parity2].offset + startoff, ps->data, stripe_length); 5642 if (!NT_SUCCESS(Status)) { 5643 ERR("write_data_phys returned %08x\n", Status); 5644 return Status; 5645 } 5646 } 5647 } else { 5648 UINT16 parity1 = (parity2 + c->chunk_item->num_stripes - 1) % c->chunk_item->num_stripes; 5649 5650 if (c->devices[parity1]->devobj || c->devices[parity2]->devobj) { 5651 UINT8* scratch; 5652 UINT16 i; 5653 5654 scratch = ExAllocatePoolWithTag(NonPagedPool, stripe_length * 2, ALLOC_TAG); 5655 if (!scratch) { 5656 ERR("out of memory\n"); 5657 return STATUS_INSUFFICIENT_RESOURCES; 5658 } 5659 5660 i = c->chunk_item->num_stripes - 3; 5661 5662 while (TRUE) { 5663 if (i == c->chunk_item->num_stripes - 3) { 5664 RtlCopyMemory(scratch, ps->data + (i * stripe_length), stripe_length); 5665 RtlCopyMemory(scratch + stripe_length, ps->data + (i * stripe_length), stripe_length); 5666 } else { 5667 do_xor(scratch, ps->data + (i * stripe_length), stripe_length); 5668 5669 galois_double(scratch + stripe_length, stripe_length); 5670 do_xor(scratch + stripe_length, ps->data + (i * stripe_length), stripe_length); 5671 } 5672 5673 if (i == 0) 5674 break; 5675 5676 i--; 5677 } 5678 5679 if (c->devices[parity1]->devobj) { 5680 Status = write_data_phys(c->devices[parity1]->devobj, cis[parity1].offset + startoff, scratch, stripe_length); 5681 if (!NT_SUCCESS(Status)) { 5682 ERR("write_data_phys returned %08x\n", Status); 5683 ExFreePool(scratch); 5684 return Status; 5685 } 5686 } 5687 5688 if (c->devices[parity2]->devobj) { 5689 Status = write_data_phys(c->devices[parity2]->devobj, cis[parity2].offset + startoff, scratch + stripe_length, stripe_length); 5690 if (!NT_SUCCESS(Status)) { 5691 ERR("write_data_phys returned %08x\n", Status); 5692 ExFreePool(scratch); 5693 return Status; 5694 } 5695 } 5696 5697 ExFreePool(scratch); 5698 } 5699 } 5700 5701 return STATUS_SUCCESS; 5702 } 5703 5704 static NTSTATUS update_chunks(device_extension* Vcb, LIST_ENTRY* batchlist, PIRP Irp, LIST_ENTRY* rollback) { 5705 LIST_ENTRY *le, *le2; 5706 NTSTATUS Status; 5707 UINT64 used_minus_cache; 5708 5709 ExAcquireResourceExclusiveLite(&Vcb->chunk_lock, TRUE); 5710 5711 // FIXME - do tree chunks before data chunks 5712 5713 le = Vcb->chunks.Flink; 5714 while (le != &Vcb->chunks) { 5715 chunk* c = CONTAINING_RECORD(le, chunk, list_entry); 5716 5717 le2 = le->Flink; 5718 5719 if (c->changed) { 5720 ExAcquireResourceExclusiveLite(&c->lock, TRUE); 5721 5722 // flush partial stripes 5723 if (!Vcb->readonly && (c->chunk_item->type & BLOCK_FLAG_RAID5 || c->chunk_item->type & BLOCK_FLAG_RAID6)) { 5724 ExAcquireResourceExclusiveLite(&c->partial_stripes_lock, TRUE); 5725 5726 while (!IsListEmpty(&c->partial_stripes)) { 5727 partial_stripe* ps = CONTAINING_RECORD(RemoveHeadList(&c->partial_stripes), partial_stripe, list_entry); 5728 5729 Status = flush_partial_stripe(Vcb, c, ps); 5730 5731 if (ps->bmparr) 5732 ExFreePool(ps->bmparr); 5733 5734 ExFreePool(ps); 5735 5736 if (!NT_SUCCESS(Status)) { 5737 ERR("flush_partial_stripe returned %08x\n", Status); 5738 ExReleaseResourceLite(&c->partial_stripes_lock); 5739 ExReleaseResourceLite(&c->lock); 5740 ExReleaseResourceLite(&Vcb->chunk_lock); 5741 return Status; 5742 } 5743 } 5744 5745 ExReleaseResourceLite(&c->partial_stripes_lock); 5746 } 5747 5748 if (c->list_entry_balance.Flink) { 5749 ExReleaseResourceLite(&c->lock); 5750 le = le2; 5751 continue; 5752 } 5753 5754 if (c->space_changed || c->created) { 5755 used_minus_cache = c->used; 5756 5757 // subtract self-hosted cache 5758 if (used_minus_cache > 0 && c->chunk_item->type & BLOCK_FLAG_DATA && c->cache && c->cache->inode_item.st_size == c->used) { 5759 LIST_ENTRY* le3; 5760 5761 le3 = c->cache->extents.Flink; 5762 while (le3 != &c->cache->extents) { 5763 extent* ext = CONTAINING_RECORD(le3, extent, list_entry); 5764 EXTENT_DATA* ed = &ext->extent_data; 5765 5766 if (!ext->ignore) { 5767 if (ed->type == EXTENT_TYPE_REGULAR || ed->type == EXTENT_TYPE_PREALLOC) { 5768 EXTENT_DATA2* ed2 = (EXTENT_DATA2*)ed->data; 5769 5770 if (ed2->size != 0 && ed2->address >= c->offset && ed2->address + ed2->size <= c->offset + c->chunk_item->size) 5771 used_minus_cache -= ed2->size; 5772 } 5773 } 5774 5775 le3 = le3->Flink; 5776 } 5777 } 5778 5779 if (used_minus_cache == 0) { 5780 Status = drop_chunk(Vcb, c, batchlist, Irp, rollback); 5781 if (!NT_SUCCESS(Status)) { 5782 ERR("drop_chunk returned %08x\n", Status); 5783 ExReleaseResourceLite(&c->lock); 5784 ExReleaseResourceLite(&Vcb->chunk_lock); 5785 return Status; 5786 } 5787 } else if (c->created) { 5788 Status = create_chunk(Vcb, c, Irp); 5789 if (!NT_SUCCESS(Status)) { 5790 ERR("create_chunk returned %08x\n", Status); 5791 ExReleaseResourceLite(&c->lock); 5792 ExReleaseResourceLite(&Vcb->chunk_lock); 5793 return Status; 5794 } 5795 } 5796 5797 if (used_minus_cache > 0) 5798 ExReleaseResourceLite(&c->lock); 5799 } 5800 } 5801 5802 le = le2; 5803 } 5804 5805 ExReleaseResourceLite(&Vcb->chunk_lock); 5806 5807 return STATUS_SUCCESS; 5808 } 5809 5810 static NTSTATUS delete_root_ref(device_extension* Vcb, UINT64 subvolid, UINT64 parsubvolid, UINT64 parinode, PANSI_STRING utf8, PIRP Irp) { 5811 KEY searchkey; 5812 traverse_ptr tp; 5813 NTSTATUS Status; 5814 5815 searchkey.obj_id = parsubvolid; 5816 searchkey.obj_type = TYPE_ROOT_REF; 5817 searchkey.offset = subvolid; 5818 5819 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 5820 if (!NT_SUCCESS(Status)) { 5821 ERR("error - find_item returned %08x\n", Status); 5822 return Status; 5823 } 5824 5825 if (!keycmp(searchkey, tp.item->key)) { 5826 if (tp.item->size < sizeof(ROOT_REF)) { 5827 ERR("(%llx,%x,%llx) was %u bytes, expected at least %u\n", tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, tp.item->size, sizeof(ROOT_REF)); 5828 return STATUS_INTERNAL_ERROR; 5829 } else { 5830 ROOT_REF* rr; 5831 ULONG len; 5832 5833 rr = (ROOT_REF*)tp.item->data; 5834 len = tp.item->size; 5835 5836 do { 5837 UINT16 itemlen; 5838 5839 if (len < sizeof(ROOT_REF) || len < offsetof(ROOT_REF, name[0]) + rr->n) { 5840 ERR("(%llx,%x,%llx) was truncated\n", tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset); 5841 break; 5842 } 5843 5844 itemlen = (UINT16)offsetof(ROOT_REF, name[0]) + rr->n; 5845 5846 if (rr->dir == parinode && rr->n == utf8->Length && RtlCompareMemory(rr->name, utf8->Buffer, rr->n) == rr->n) { 5847 UINT16 newlen = tp.item->size - itemlen; 5848 5849 Status = delete_tree_item(Vcb, &tp); 5850 if (!NT_SUCCESS(Status)) { 5851 ERR("delete_tree_item returned %08x\n", Status); 5852 return Status; 5853 } 5854 5855 if (newlen == 0) { 5856 TRACE("deleting (%llx,%x,%llx)\n", tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset); 5857 } else { 5858 UINT8 *newrr = ExAllocatePoolWithTag(PagedPool, newlen, ALLOC_TAG), *rroff; 5859 5860 if (!newrr) { 5861 ERR("out of memory\n"); 5862 return STATUS_INSUFFICIENT_RESOURCES; 5863 } 5864 5865 TRACE("modifying (%llx,%x,%llx)\n", tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset); 5866 5867 if ((UINT8*)rr > tp.item->data) { 5868 RtlCopyMemory(newrr, tp.item->data, (UINT8*)rr - tp.item->data); 5869 rroff = newrr + ((UINT8*)rr - tp.item->data); 5870 } else { 5871 rroff = newrr; 5872 } 5873 5874 if ((UINT8*)&rr->name[rr->n] < tp.item->data + tp.item->size) 5875 RtlCopyMemory(rroff, &rr->name[rr->n], tp.item->size - ((UINT8*)&rr->name[rr->n] - tp.item->data)); 5876 5877 Status = insert_tree_item(Vcb, Vcb->root_root, tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, newrr, newlen, NULL, Irp); 5878 if (!NT_SUCCESS(Status)) { 5879 ERR("insert_tree_item returned %08x\n", Status); 5880 ExFreePool(newrr); 5881 return Status; 5882 } 5883 } 5884 5885 break; 5886 } 5887 5888 if (len > itemlen) { 5889 len -= itemlen; 5890 rr = (ROOT_REF*)&rr->name[rr->n]; 5891 } else 5892 break; 5893 } while (len > 0); 5894 } 5895 } else { 5896 WARN("could not find ROOT_REF entry for subvol %llx in %llx\n", searchkey.offset, searchkey.obj_id); 5897 return STATUS_NOT_FOUND; 5898 } 5899 5900 return STATUS_SUCCESS; 5901 } 5902 5903 #ifdef _MSC_VER 5904 #pragma warning(push) 5905 #pragma warning(suppress: 28194) 5906 #endif 5907 static NTSTATUS add_root_ref(_In_ device_extension* Vcb, _In_ UINT64 subvolid, _In_ UINT64 parsubvolid, _In_ __drv_aliasesMem ROOT_REF* rr, _In_opt_ PIRP Irp) { 5908 KEY searchkey; 5909 traverse_ptr tp; 5910 NTSTATUS Status; 5911 5912 searchkey.obj_id = parsubvolid; 5913 searchkey.obj_type = TYPE_ROOT_REF; 5914 searchkey.offset = subvolid; 5915 5916 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 5917 if (!NT_SUCCESS(Status)) { 5918 ERR("error - find_item returned %08x\n", Status); 5919 return Status; 5920 } 5921 5922 if (!keycmp(searchkey, tp.item->key)) { 5923 UINT16 rrsize = tp.item->size + (UINT16)offsetof(ROOT_REF, name[0]) + rr->n; 5924 UINT8* rr2; 5925 5926 rr2 = ExAllocatePoolWithTag(PagedPool, rrsize, ALLOC_TAG); 5927 if (!rr2) { 5928 ERR("out of memory\n"); 5929 return STATUS_INSUFFICIENT_RESOURCES; 5930 } 5931 5932 if (tp.item->size > 0) 5933 RtlCopyMemory(rr2, tp.item->data, tp.item->size); 5934 5935 RtlCopyMemory(rr2 + tp.item->size, rr, offsetof(ROOT_REF, name[0]) + rr->n); 5936 ExFreePool(rr); 5937 5938 Status = delete_tree_item(Vcb, &tp); 5939 if (!NT_SUCCESS(Status)) { 5940 ERR("delete_tree_item returned %08x\n", Status); 5941 ExFreePool(rr2); 5942 return Status; 5943 } 5944 5945 Status = insert_tree_item(Vcb, Vcb->root_root, searchkey.obj_id, searchkey.obj_type, searchkey.offset, rr2, rrsize, NULL, Irp); 5946 if (!NT_SUCCESS(Status)) { 5947 ERR("insert_tree_item returned %08x\n", Status); 5948 ExFreePool(rr2); 5949 return Status; 5950 } 5951 } else { 5952 Status = insert_tree_item(Vcb, Vcb->root_root, searchkey.obj_id, searchkey.obj_type, searchkey.offset, rr, (UINT16)offsetof(ROOT_REF, name[0]) + rr->n, NULL, Irp); 5953 if (!NT_SUCCESS(Status)) { 5954 ERR("insert_tree_item returned %08x\n", Status); 5955 ExFreePool(rr); 5956 return Status; 5957 } 5958 } 5959 5960 return STATUS_SUCCESS; 5961 } 5962 #ifdef _MSC_VER 5963 #pragma warning(pop) 5964 #endif 5965 5966 static NTSTATUS update_root_backref(device_extension* Vcb, UINT64 subvolid, UINT64 parsubvolid, PIRP Irp) { 5967 KEY searchkey; 5968 traverse_ptr tp; 5969 UINT8* data; 5970 UINT16 datalen; 5971 NTSTATUS Status; 5972 5973 searchkey.obj_id = parsubvolid; 5974 searchkey.obj_type = TYPE_ROOT_REF; 5975 searchkey.offset = subvolid; 5976 5977 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 5978 if (!NT_SUCCESS(Status)) { 5979 ERR("error - find_item returned %08x\n", Status); 5980 return Status; 5981 } 5982 5983 if (!keycmp(tp.item->key, searchkey) && tp.item->size > 0) { 5984 datalen = tp.item->size; 5985 5986 data = ExAllocatePoolWithTag(PagedPool, datalen, ALLOC_TAG); 5987 if (!data) { 5988 ERR("out of memory\n"); 5989 return STATUS_INSUFFICIENT_RESOURCES; 5990 } 5991 5992 RtlCopyMemory(data, tp.item->data, datalen); 5993 } else { 5994 datalen = 0; 5995 data = NULL; 5996 } 5997 5998 searchkey.obj_id = subvolid; 5999 searchkey.obj_type = TYPE_ROOT_BACKREF; 6000 searchkey.offset = parsubvolid; 6001 6002 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 6003 if (!NT_SUCCESS(Status)) { 6004 ERR("error - find_item returned %08x\n", Status); 6005 6006 if (datalen > 0) 6007 ExFreePool(data); 6008 6009 return Status; 6010 } 6011 6012 if (!keycmp(tp.item->key, searchkey)) { 6013 Status = delete_tree_item(Vcb, &tp); 6014 if (!NT_SUCCESS(Status)) { 6015 ERR("delete_tree_item returned %08x\n", Status); 6016 6017 if (datalen > 0) 6018 ExFreePool(data); 6019 6020 return Status; 6021 } 6022 } 6023 6024 if (datalen > 0) { 6025 Status = insert_tree_item(Vcb, Vcb->root_root, subvolid, TYPE_ROOT_BACKREF, parsubvolid, data, datalen, NULL, Irp); 6026 if (!NT_SUCCESS(Status)) { 6027 ERR("insert_tree_item returned %08x\n", Status); 6028 ExFreePool(data); 6029 return Status; 6030 } 6031 } 6032 6033 return STATUS_SUCCESS; 6034 } 6035 6036 static NTSTATUS add_root_item_to_cache(device_extension* Vcb, UINT64 root, PIRP Irp) { 6037 KEY searchkey; 6038 traverse_ptr tp; 6039 NTSTATUS Status; 6040 6041 searchkey.obj_id = root; 6042 searchkey.obj_type = TYPE_ROOT_ITEM; 6043 searchkey.offset = 0xffffffffffffffff; 6044 6045 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 6046 if (!NT_SUCCESS(Status)) { 6047 ERR("error - find_item returned %08x\n", Status); 6048 return Status; 6049 } 6050 6051 if (tp.item->key.obj_id != searchkey.obj_id || tp.item->key.obj_type != searchkey.obj_type) { 6052 ERR("could not find ROOT_ITEM for tree %llx\n", searchkey.obj_id); 6053 return STATUS_INTERNAL_ERROR; 6054 } 6055 6056 if (tp.item->size < sizeof(ROOT_ITEM)) { // if not full length, create new entry with new bits zeroed 6057 ROOT_ITEM* ri = ExAllocatePoolWithTag(PagedPool, sizeof(ROOT_ITEM), ALLOC_TAG); 6058 if (!ri) { 6059 ERR("out of memory\n"); 6060 return STATUS_INSUFFICIENT_RESOURCES; 6061 } 6062 6063 if (tp.item->size > 0) 6064 RtlCopyMemory(ri, tp.item->data, tp.item->size); 6065 6066 RtlZeroMemory(((UINT8*)ri) + tp.item->size, sizeof(ROOT_ITEM) - tp.item->size); 6067 6068 Status = delete_tree_item(Vcb, &tp); 6069 if (!NT_SUCCESS(Status)) { 6070 ERR("delete_tree_item returned %08x\n", Status); 6071 ExFreePool(ri); 6072 return Status; 6073 } 6074 6075 Status = insert_tree_item(Vcb, Vcb->root_root, searchkey.obj_id, searchkey.obj_type, tp.item->key.offset, ri, sizeof(ROOT_ITEM), NULL, Irp); 6076 if (!NT_SUCCESS(Status)) { 6077 ERR("insert_tree_item returned %08x\n", Status); 6078 ExFreePool(ri); 6079 return Status; 6080 } 6081 } else { 6082 tp.tree->write = TRUE; 6083 } 6084 6085 return STATUS_SUCCESS; 6086 } 6087 6088 static NTSTATUS flush_fileref(file_ref* fileref, LIST_ENTRY* batchlist, PIRP Irp) { 6089 NTSTATUS Status; 6090 6091 // if fileref created and then immediately deleted, do nothing 6092 if (fileref->created && fileref->deleted) { 6093 fileref->dirty = FALSE; 6094 return STATUS_SUCCESS; 6095 } 6096 6097 if (fileref->fcb->ads) { 6098 fileref->dirty = FALSE; 6099 return STATUS_SUCCESS; 6100 } 6101 6102 if (fileref->created) { 6103 UINT16 disize; 6104 DIR_ITEM *di, *di2; 6105 UINT32 crc32; 6106 6107 crc32 = calc_crc32c(0xfffffffe, (UINT8*)fileref->dc->utf8.Buffer, fileref->dc->utf8.Length); 6108 6109 disize = (UINT16)(offsetof(DIR_ITEM, name[0]) + fileref->dc->utf8.Length); 6110 di = ExAllocatePoolWithTag(PagedPool, disize, ALLOC_TAG); 6111 if (!di) { 6112 ERR("out of memory\n"); 6113 return STATUS_INSUFFICIENT_RESOURCES; 6114 } 6115 6116 if (fileref->parent->fcb->subvol == fileref->fcb->subvol) { 6117 di->key.obj_id = fileref->fcb->inode; 6118 di->key.obj_type = TYPE_INODE_ITEM; 6119 di->key.offset = 0; 6120 } else { // subvolume 6121 di->key.obj_id = fileref->fcb->subvol->id; 6122 di->key.obj_type = TYPE_ROOT_ITEM; 6123 di->key.offset = 0xffffffffffffffff; 6124 } 6125 6126 di->transid = fileref->fcb->Vcb->superblock.generation; 6127 di->m = 0; 6128 di->n = (UINT16)fileref->dc->utf8.Length; 6129 di->type = fileref->fcb->type; 6130 RtlCopyMemory(di->name, fileref->dc->utf8.Buffer, fileref->dc->utf8.Length); 6131 6132 di2 = ExAllocatePoolWithTag(PagedPool, disize, ALLOC_TAG); 6133 if (!di2) { 6134 ERR("out of memory\n"); 6135 return STATUS_INSUFFICIENT_RESOURCES; 6136 } 6137 6138 RtlCopyMemory(di2, di, disize); 6139 6140 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->parent->fcb->inode, TYPE_DIR_INDEX, 6141 fileref->dc->index, di, disize, Batch_Insert); 6142 if (!NT_SUCCESS(Status)) { 6143 ERR("insert_tree_item_batch returned %08x\n", Status); 6144 return Status; 6145 } 6146 6147 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->parent->fcb->inode, TYPE_DIR_ITEM, crc32, 6148 di2, disize, Batch_DirItem); 6149 if (!NT_SUCCESS(Status)) { 6150 ERR("insert_tree_item_batch returned %08x\n", Status); 6151 return Status; 6152 } 6153 6154 if (fileref->parent->fcb->subvol == fileref->fcb->subvol) { 6155 INODE_REF* ir; 6156 6157 ir = ExAllocatePoolWithTag(PagedPool, sizeof(INODE_REF) - 1 + fileref->dc->utf8.Length, ALLOC_TAG); 6158 if (!ir) { 6159 ERR("out of memory\n"); 6160 return STATUS_INSUFFICIENT_RESOURCES; 6161 } 6162 6163 ir->index = fileref->dc->index; 6164 ir->n = fileref->dc->utf8.Length; 6165 RtlCopyMemory(ir->name, fileref->dc->utf8.Buffer, ir->n); 6166 6167 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->fcb->subvol, fileref->fcb->inode, TYPE_INODE_REF, fileref->parent->fcb->inode, 6168 ir, sizeof(INODE_REF) - 1 + ir->n, Batch_InodeRef); 6169 if (!NT_SUCCESS(Status)) { 6170 ERR("insert_tree_item_batch returned %08x\n", Status); 6171 return Status; 6172 } 6173 } else if (fileref->fcb != fileref->fcb->Vcb->dummy_fcb) { 6174 ULONG rrlen; 6175 ROOT_REF* rr; 6176 6177 rrlen = sizeof(ROOT_REF) - 1 + fileref->dc->utf8.Length; 6178 6179 rr = ExAllocatePoolWithTag(PagedPool, rrlen, ALLOC_TAG); 6180 if (!rr) { 6181 ERR("out of memory\n"); 6182 return STATUS_INSUFFICIENT_RESOURCES; 6183 } 6184 6185 rr->dir = fileref->parent->fcb->inode; 6186 rr->index = fileref->dc->index; 6187 rr->n = fileref->dc->utf8.Length; 6188 RtlCopyMemory(rr->name, fileref->dc->utf8.Buffer, fileref->dc->utf8.Length); 6189 6190 Status = add_root_ref(fileref->fcb->Vcb, fileref->fcb->subvol->id, fileref->parent->fcb->subvol->id, rr, Irp); 6191 if (!NT_SUCCESS(Status)) { 6192 ERR("add_root_ref returned %08x\n", Status); 6193 return Status; 6194 } 6195 6196 Status = update_root_backref(fileref->fcb->Vcb, fileref->fcb->subvol->id, fileref->parent->fcb->subvol->id, Irp); 6197 if (!NT_SUCCESS(Status)) { 6198 ERR("update_root_backref returned %08x\n", Status); 6199 return Status; 6200 } 6201 } 6202 6203 fileref->created = FALSE; 6204 } else if (fileref->deleted) { 6205 UINT32 crc32; 6206 ANSI_STRING* name; 6207 DIR_ITEM* di; 6208 6209 name = &fileref->oldutf8; 6210 6211 crc32 = calc_crc32c(0xfffffffe, (UINT8*)name->Buffer, name->Length); 6212 6213 TRACE("deleting %.*S\n", file_desc_fileref(fileref)); 6214 6215 di = ExAllocatePoolWithTag(PagedPool, sizeof(DIR_ITEM) - 1 + name->Length, ALLOC_TAG); 6216 if (!di) { 6217 ERR("out of memory\n"); 6218 return STATUS_INSUFFICIENT_RESOURCES; 6219 } 6220 6221 di->m = 0; 6222 di->n = name->Length; 6223 RtlCopyMemory(di->name, name->Buffer, name->Length); 6224 6225 // delete DIR_ITEM (0x54) 6226 6227 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->parent->fcb->inode, TYPE_DIR_ITEM, 6228 crc32, di, sizeof(DIR_ITEM) - 1 + name->Length, Batch_DeleteDirItem); 6229 if (!NT_SUCCESS(Status)) { 6230 ERR("insert_tree_item_batch returned %08x\n", Status); 6231 return Status; 6232 } 6233 6234 if (fileref->parent->fcb->subvol == fileref->fcb->subvol) { 6235 INODE_REF* ir; 6236 6237 // delete INODE_REF (0xc) 6238 6239 ir = ExAllocatePoolWithTag(PagedPool, sizeof(INODE_REF) - 1 + name->Length, ALLOC_TAG); 6240 if (!ir) { 6241 ERR("out of memory\n"); 6242 return STATUS_INSUFFICIENT_RESOURCES; 6243 } 6244 6245 ir->index = fileref->oldindex; 6246 ir->n = name->Length; 6247 RtlCopyMemory(ir->name, name->Buffer, name->Length); 6248 6249 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->fcb->inode, TYPE_INODE_REF, 6250 fileref->parent->fcb->inode, ir, sizeof(INODE_REF) - 1 + name->Length, Batch_DeleteInodeRef); 6251 if (!NT_SUCCESS(Status)) { 6252 ERR("insert_tree_item_batch returned %08x\n", Status); 6253 return Status; 6254 } 6255 } else if (fileref->fcb != fileref->fcb->Vcb->dummy_fcb) { // subvolume 6256 Status = delete_root_ref(fileref->fcb->Vcb, fileref->fcb->subvol->id, fileref->parent->fcb->subvol->id, fileref->parent->fcb->inode, name, Irp); 6257 if (!NT_SUCCESS(Status)) { 6258 ERR("delete_root_ref returned %08x\n", Status); 6259 return Status; 6260 } 6261 6262 Status = update_root_backref(fileref->fcb->Vcb, fileref->fcb->subvol->id, fileref->parent->fcb->subvol->id, Irp); 6263 if (!NT_SUCCESS(Status)) { 6264 ERR("update_root_backref returned %08x\n", Status); 6265 return Status; 6266 } 6267 } 6268 6269 // delete DIR_INDEX (0x60) 6270 6271 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->parent->fcb->inode, TYPE_DIR_INDEX, 6272 fileref->oldindex, NULL, 0, Batch_Delete); 6273 if (!NT_SUCCESS(Status)) { 6274 ERR("insert_tree_item_batch returned %08x\n", Status); 6275 return Status; 6276 } 6277 6278 if (fileref->oldutf8.Buffer) { 6279 ExFreePool(fileref->oldutf8.Buffer); 6280 fileref->oldutf8.Buffer = NULL; 6281 } 6282 } else { // rename or change type 6283 PANSI_STRING oldutf8 = fileref->oldutf8.Buffer ? &fileref->oldutf8 : &fileref->dc->utf8; 6284 UINT32 crc32, oldcrc32; 6285 UINT16 disize; 6286 DIR_ITEM *olddi, *di, *di2; 6287 6288 crc32 = calc_crc32c(0xfffffffe, (UINT8*)fileref->dc->utf8.Buffer, fileref->dc->utf8.Length); 6289 6290 if (!fileref->oldutf8.Buffer) 6291 oldcrc32 = crc32; 6292 else 6293 oldcrc32 = calc_crc32c(0xfffffffe, (UINT8*)fileref->oldutf8.Buffer, fileref->oldutf8.Length); 6294 6295 olddi = ExAllocatePoolWithTag(PagedPool, sizeof(DIR_ITEM) - 1 + oldutf8->Length, ALLOC_TAG); 6296 if (!olddi) { 6297 ERR("out of memory\n"); 6298 return STATUS_INSUFFICIENT_RESOURCES; 6299 } 6300 6301 olddi->m = 0; 6302 olddi->n = (UINT16)oldutf8->Length; 6303 RtlCopyMemory(olddi->name, oldutf8->Buffer, oldutf8->Length); 6304 6305 // delete DIR_ITEM (0x54) 6306 6307 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->parent->fcb->inode, TYPE_DIR_ITEM, 6308 oldcrc32, olddi, sizeof(DIR_ITEM) - 1 + oldutf8->Length, Batch_DeleteDirItem); 6309 if (!NT_SUCCESS(Status)) { 6310 ERR("insert_tree_item_batch returned %08x\n", Status); 6311 ExFreePool(olddi); 6312 return Status; 6313 } 6314 6315 // add DIR_ITEM (0x54) 6316 6317 disize = (UINT16)(offsetof(DIR_ITEM, name[0]) + fileref->dc->utf8.Length); 6318 di = ExAllocatePoolWithTag(PagedPool, disize, ALLOC_TAG); 6319 if (!di) { 6320 ERR("out of memory\n"); 6321 return STATUS_INSUFFICIENT_RESOURCES; 6322 } 6323 6324 di2 = ExAllocatePoolWithTag(PagedPool, disize, ALLOC_TAG); 6325 if (!di2) { 6326 ERR("out of memory\n"); 6327 ExFreePool(di); 6328 return STATUS_INSUFFICIENT_RESOURCES; 6329 } 6330 6331 if (fileref->dc) 6332 di->key = fileref->dc->key; 6333 else if (fileref->parent->fcb->subvol == fileref->fcb->subvol) { 6334 di->key.obj_id = fileref->fcb->inode; 6335 di->key.obj_type = TYPE_INODE_ITEM; 6336 di->key.offset = 0; 6337 } else { // subvolume 6338 di->key.obj_id = fileref->fcb->subvol->id; 6339 di->key.obj_type = TYPE_ROOT_ITEM; 6340 di->key.offset = 0xffffffffffffffff; 6341 } 6342 6343 di->transid = fileref->fcb->Vcb->superblock.generation; 6344 di->m = 0; 6345 di->n = (UINT16)fileref->dc->utf8.Length; 6346 di->type = fileref->fcb->type; 6347 RtlCopyMemory(di->name, fileref->dc->utf8.Buffer, fileref->dc->utf8.Length); 6348 6349 RtlCopyMemory(di2, di, disize); 6350 6351 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->parent->fcb->inode, TYPE_DIR_ITEM, crc32, 6352 di, disize, Batch_DirItem); 6353 if (!NT_SUCCESS(Status)) { 6354 ERR("insert_tree_item_batch returned %08x\n", Status); 6355 ExFreePool(di2); 6356 ExFreePool(di); 6357 return Status; 6358 } 6359 6360 if (fileref->parent->fcb->subvol == fileref->fcb->subvol) { 6361 INODE_REF *ir, *ir2; 6362 6363 // delete INODE_REF (0xc) 6364 6365 ir = ExAllocatePoolWithTag(PagedPool, sizeof(INODE_REF) - 1 + oldutf8->Length, ALLOC_TAG); 6366 if (!ir) { 6367 ERR("out of memory\n"); 6368 ExFreePool(di2); 6369 return STATUS_INSUFFICIENT_RESOURCES; 6370 } 6371 6372 ir->index = fileref->dc->index; 6373 ir->n = oldutf8->Length; 6374 RtlCopyMemory(ir->name, oldutf8->Buffer, ir->n); 6375 6376 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->fcb->subvol, fileref->fcb->inode, TYPE_INODE_REF, fileref->parent->fcb->inode, 6377 ir, sizeof(INODE_REF) - 1 + ir->n, Batch_DeleteInodeRef); 6378 if (!NT_SUCCESS(Status)) { 6379 ERR("insert_tree_item_batch returned %08x\n", Status); 6380 ExFreePool(ir); 6381 ExFreePool(di2); 6382 return Status; 6383 } 6384 6385 // add INODE_REF (0xc) 6386 6387 ir2 = ExAllocatePoolWithTag(PagedPool, sizeof(INODE_REF) - 1 + fileref->dc->utf8.Length, ALLOC_TAG); 6388 if (!ir2) { 6389 ERR("out of memory\n"); 6390 ExFreePool(di2); 6391 return STATUS_INSUFFICIENT_RESOURCES; 6392 } 6393 6394 ir2->index = fileref->dc->index; 6395 ir2->n = fileref->dc->utf8.Length; 6396 RtlCopyMemory(ir2->name, fileref->dc->utf8.Buffer, ir2->n); 6397 6398 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->fcb->subvol, fileref->fcb->inode, TYPE_INODE_REF, fileref->parent->fcb->inode, 6399 ir2, sizeof(INODE_REF) - 1 + ir2->n, Batch_InodeRef); 6400 if (!NT_SUCCESS(Status)) { 6401 ERR("insert_tree_item_batch returned %08x\n", Status); 6402 ExFreePool(ir2); 6403 ExFreePool(di2); 6404 return Status; 6405 } 6406 } else if (fileref->fcb != fileref->fcb->Vcb->dummy_fcb) { // subvolume 6407 ULONG rrlen; 6408 ROOT_REF* rr; 6409 6410 Status = delete_root_ref(fileref->fcb->Vcb, fileref->fcb->subvol->id, fileref->parent->fcb->subvol->id, fileref->parent->fcb->inode, oldutf8, Irp); 6411 if (!NT_SUCCESS(Status)) { 6412 ERR("delete_root_ref returned %08x\n", Status); 6413 ExFreePool(di2); 6414 return Status; 6415 } 6416 6417 rrlen = sizeof(ROOT_REF) - 1 + fileref->dc->utf8.Length; 6418 6419 rr = ExAllocatePoolWithTag(PagedPool, rrlen, ALLOC_TAG); 6420 if (!rr) { 6421 ERR("out of memory\n"); 6422 ExFreePool(di2); 6423 return STATUS_INSUFFICIENT_RESOURCES; 6424 } 6425 6426 rr->dir = fileref->parent->fcb->inode; 6427 rr->index = fileref->dc->index; 6428 rr->n = fileref->dc->utf8.Length; 6429 RtlCopyMemory(rr->name, fileref->dc->utf8.Buffer, fileref->dc->utf8.Length); 6430 6431 Status = add_root_ref(fileref->fcb->Vcb, fileref->fcb->subvol->id, fileref->parent->fcb->subvol->id, rr, Irp); 6432 if (!NT_SUCCESS(Status)) { 6433 ERR("add_root_ref returned %08x\n", Status); 6434 ExFreePool(di2); 6435 return Status; 6436 } 6437 6438 Status = update_root_backref(fileref->fcb->Vcb, fileref->fcb->subvol->id, fileref->parent->fcb->subvol->id, Irp); 6439 if (!NT_SUCCESS(Status)) { 6440 ERR("update_root_backref returned %08x\n", Status); 6441 ExFreePool(di2); 6442 return Status; 6443 } 6444 } 6445 6446 // delete DIR_INDEX (0x60) 6447 6448 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->parent->fcb->inode, TYPE_DIR_INDEX, 6449 fileref->dc->index, NULL, 0, Batch_Delete); 6450 if (!NT_SUCCESS(Status)) { 6451 ERR("insert_tree_item_batch returned %08x\n", Status); 6452 ExFreePool(di2); 6453 return Status; 6454 } 6455 6456 // add DIR_INDEX (0x60) 6457 6458 Status = insert_tree_item_batch(batchlist, fileref->fcb->Vcb, fileref->parent->fcb->subvol, fileref->parent->fcb->inode, TYPE_DIR_INDEX, 6459 fileref->dc->index, di2, disize, Batch_Insert); 6460 if (!NT_SUCCESS(Status)) { 6461 ERR("insert_tree_item_batch returned %08x\n", Status); 6462 ExFreePool(di2); 6463 return Status; 6464 } 6465 6466 if (fileref->oldutf8.Buffer) { 6467 ExFreePool(fileref->oldutf8.Buffer); 6468 fileref->oldutf8.Buffer = NULL; 6469 } 6470 } 6471 6472 fileref->dirty = FALSE; 6473 6474 return STATUS_SUCCESS; 6475 } 6476 6477 static void flush_disk_caches(device_extension* Vcb) { 6478 LIST_ENTRY* le; 6479 ioctl_context context; 6480 ULONG num; 6481 6482 context.left = 0; 6483 6484 le = Vcb->devices.Flink; 6485 6486 while (le != &Vcb->devices) { 6487 device* dev = CONTAINING_RECORD(le, device, list_entry); 6488 6489 if (dev->devobj && !dev->readonly && dev->can_flush) 6490 context.left++; 6491 6492 le = le->Flink; 6493 } 6494 6495 if (context.left == 0) 6496 return; 6497 6498 num = 0; 6499 6500 KeInitializeEvent(&context.Event, NotificationEvent, FALSE); 6501 6502 context.stripes = ExAllocatePoolWithTag(NonPagedPool, sizeof(ioctl_context_stripe) * context.left, ALLOC_TAG); 6503 if (!context.stripes) { 6504 ERR("out of memory\n"); 6505 return; 6506 } 6507 6508 RtlZeroMemory(context.stripes, sizeof(ioctl_context_stripe) * context.left); 6509 6510 le = Vcb->devices.Flink; 6511 6512 while (le != &Vcb->devices) { 6513 device* dev = CONTAINING_RECORD(le, device, list_entry); 6514 6515 if (dev->devobj && !dev->readonly && dev->can_flush) { 6516 PIO_STACK_LOCATION IrpSp; 6517 ioctl_context_stripe* stripe = &context.stripes[num]; 6518 6519 RtlZeroMemory(&stripe->apte, sizeof(ATA_PASS_THROUGH_EX)); 6520 6521 stripe->apte.Length = sizeof(ATA_PASS_THROUGH_EX); 6522 stripe->apte.TimeOutValue = 5; 6523 stripe->apte.CurrentTaskFile[6] = IDE_COMMAND_FLUSH_CACHE; 6524 6525 stripe->Irp = IoAllocateIrp(dev->devobj->StackSize, FALSE); 6526 6527 if (!stripe->Irp) { 6528 ERR("IoAllocateIrp failed\n"); 6529 goto nextdev; 6530 } 6531 6532 IrpSp = IoGetNextIrpStackLocation(stripe->Irp); 6533 IrpSp->MajorFunction = IRP_MJ_DEVICE_CONTROL; 6534 6535 IrpSp->Parameters.DeviceIoControl.IoControlCode = IOCTL_ATA_PASS_THROUGH; 6536 IrpSp->Parameters.DeviceIoControl.InputBufferLength = sizeof(ATA_PASS_THROUGH_EX); 6537 IrpSp->Parameters.DeviceIoControl.OutputBufferLength = sizeof(ATA_PASS_THROUGH_EX); 6538 6539 stripe->Irp->AssociatedIrp.SystemBuffer = &stripe->apte; 6540 stripe->Irp->Flags |= IRP_BUFFERED_IO | IRP_INPUT_OPERATION; 6541 stripe->Irp->UserBuffer = &stripe->apte; 6542 stripe->Irp->UserIosb = &stripe->iosb; 6543 6544 IoSetCompletionRoutine(stripe->Irp, ioctl_completion, &context, TRUE, TRUE, TRUE); 6545 6546 IoCallDriver(dev->devobj, stripe->Irp); 6547 6548 nextdev: 6549 num++; 6550 } 6551 6552 le = le->Flink; 6553 } 6554 6555 KeWaitForSingleObject(&context.Event, Executive, KernelMode, FALSE, NULL); 6556 6557 ExFreePool(context.stripes); 6558 } 6559 6560 static NTSTATUS flush_changed_dev_stats(device_extension* Vcb, device* dev, PIRP Irp) { 6561 NTSTATUS Status; 6562 KEY searchkey; 6563 traverse_ptr tp; 6564 UINT16 statslen; 6565 UINT64* stats; 6566 6567 searchkey.obj_id = 0; 6568 searchkey.obj_type = TYPE_DEV_STATS; 6569 searchkey.offset = dev->devitem.dev_id; 6570 6571 Status = find_item(Vcb, Vcb->dev_root, &tp, &searchkey, FALSE, Irp); 6572 if (!NT_SUCCESS(Status)) { 6573 ERR("find_item returned %08x\n", Status); 6574 return Status; 6575 } 6576 6577 if (!keycmp(tp.item->key, searchkey)) { 6578 Status = delete_tree_item(Vcb, &tp); 6579 if (!NT_SUCCESS(Status)) { 6580 ERR("delete_tree_item returned %08x\n", Status); 6581 return Status; 6582 } 6583 } 6584 6585 statslen = sizeof(UINT64) * 5; 6586 stats = ExAllocatePoolWithTag(PagedPool, statslen, ALLOC_TAG); 6587 if (!stats) { 6588 ERR("out of memory\n"); 6589 return STATUS_INSUFFICIENT_RESOURCES; 6590 } 6591 6592 RtlCopyMemory(stats, dev->stats, statslen); 6593 6594 Status = insert_tree_item(Vcb, Vcb->dev_root, 0, TYPE_DEV_STATS, dev->devitem.dev_id, stats, statslen, NULL, Irp); 6595 if (!NT_SUCCESS(Status)) { 6596 ERR("insert_tree_item returned %08x\n", Status); 6597 ExFreePool(stats); 6598 return Status; 6599 } 6600 6601 return STATUS_SUCCESS; 6602 } 6603 6604 static NTSTATUS flush_subvol(device_extension* Vcb, root* r, PIRP Irp) { 6605 NTSTATUS Status; 6606 6607 if (r != Vcb->root_root && r != Vcb->chunk_root) { 6608 KEY searchkey; 6609 traverse_ptr tp; 6610 ROOT_ITEM* ri; 6611 6612 searchkey.obj_id = r->id; 6613 searchkey.obj_type = TYPE_ROOT_ITEM; 6614 searchkey.offset = 0xffffffffffffffff; 6615 6616 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 6617 if (!NT_SUCCESS(Status)) { 6618 ERR("error - find_item returned %08x\n", Status); 6619 return Status; 6620 } 6621 6622 if (tp.item->key.obj_id != searchkey.obj_id || tp.item->key.obj_type != searchkey.obj_type) { 6623 ERR("could not find ROOT_ITEM for tree %llx\n", searchkey.obj_id); 6624 return STATUS_INTERNAL_ERROR; 6625 } 6626 6627 ri = ExAllocatePoolWithTag(PagedPool, sizeof(ROOT_ITEM), ALLOC_TAG); 6628 if (!ri) { 6629 ERR("out of memory\n"); 6630 return STATUS_INSUFFICIENT_RESOURCES; 6631 } 6632 6633 RtlCopyMemory(ri, &r->root_item, sizeof(ROOT_ITEM)); 6634 6635 Status = delete_tree_item(Vcb, &tp); 6636 if (!NT_SUCCESS(Status)) { 6637 ERR("delete_tree_item returned %08x\n", Status); 6638 return Status; 6639 } 6640 6641 Status = insert_tree_item(Vcb, Vcb->root_root, tp.item->key.obj_id, tp.item->key.obj_type, tp.item->key.offset, ri, sizeof(ROOT_ITEM), NULL, Irp); 6642 if (!NT_SUCCESS(Status)) { 6643 ERR("insert_tree_item returned %08x\n", Status); 6644 return Status; 6645 } 6646 } 6647 6648 if (r->received) { 6649 KEY searchkey; 6650 traverse_ptr tp; 6651 6652 if (!Vcb->uuid_root) { 6653 root* uuid_root; 6654 6655 TRACE("uuid root doesn't exist, creating it\n"); 6656 6657 Status = create_root(Vcb, BTRFS_ROOT_UUID, &uuid_root, FALSE, 0, Irp); 6658 6659 if (!NT_SUCCESS(Status)) { 6660 ERR("create_root returned %08x\n", Status); 6661 return Status; 6662 } 6663 6664 Vcb->uuid_root = uuid_root; 6665 } 6666 6667 RtlCopyMemory(&searchkey.obj_id, &r->root_item.received_uuid, sizeof(UINT64)); 6668 searchkey.obj_type = TYPE_SUBVOL_REC_UUID; 6669 RtlCopyMemory(&searchkey.offset, &r->root_item.received_uuid.uuid[sizeof(UINT64)], sizeof(UINT64)); 6670 6671 Status = find_item(Vcb, Vcb->uuid_root, &tp, &searchkey, FALSE, Irp); 6672 if (!NT_SUCCESS(Status)) { 6673 ERR("find_item returned %08x\n", Status); 6674 return Status; 6675 } 6676 6677 if (!keycmp(tp.item->key, searchkey)) { 6678 if (tp.item->size + sizeof(UINT64) <= Vcb->superblock.node_size - sizeof(tree_header) - sizeof(leaf_node)) { 6679 UINT64* ids; 6680 6681 ids = ExAllocatePoolWithTag(PagedPool, tp.item->size + sizeof(UINT64), ALLOC_TAG); 6682 if (!ids) { 6683 ERR("out of memory\n"); 6684 return STATUS_INSUFFICIENT_RESOURCES; 6685 } 6686 6687 RtlCopyMemory(ids, tp.item->data, tp.item->size); 6688 RtlCopyMemory((UINT8*)ids + tp.item->size, &r->id, sizeof(UINT64)); 6689 6690 Status = delete_tree_item(Vcb, &tp); 6691 if (!NT_SUCCESS(Status)) { 6692 ERR("delete_tree_item returned %08x\n", Status); 6693 ExFreePool(ids); 6694 return Status; 6695 } 6696 6697 Status = insert_tree_item(Vcb, Vcb->uuid_root, searchkey.obj_id, searchkey.obj_type, searchkey.offset, ids, tp.item->size + sizeof(UINT64), NULL, Irp); 6698 if (!NT_SUCCESS(Status)) { 6699 ERR("insert_tree_item returned %08x\n", Status); 6700 ExFreePool(ids); 6701 return Status; 6702 } 6703 } 6704 } else { 6705 UINT64* root_num; 6706 6707 root_num = ExAllocatePoolWithTag(PagedPool, sizeof(UINT64), ALLOC_TAG); 6708 if (!root_num) { 6709 ERR("out of memory\n"); 6710 return STATUS_INSUFFICIENT_RESOURCES; 6711 } 6712 6713 *root_num = r->id; 6714 6715 Status = insert_tree_item(Vcb, Vcb->uuid_root, searchkey.obj_id, searchkey.obj_type, searchkey.offset, root_num, sizeof(UINT64), NULL, Irp); 6716 if (!NT_SUCCESS(Status)) { 6717 ERR("insert_tree_item returned %08x\n", Status); 6718 ExFreePool(root_num); 6719 return Status; 6720 } 6721 } 6722 6723 r->received = FALSE; 6724 } 6725 6726 r->dirty = FALSE; 6727 6728 return STATUS_SUCCESS; 6729 } 6730 6731 static NTSTATUS test_not_full(device_extension* Vcb) { 6732 UINT64 reserve, could_alloc, free_space; 6733 LIST_ENTRY* le; 6734 6735 // This function ensures we drop into readonly mode if we're about to leave very little 6736 // space for metadata - this is similar to the "global reserve" of the Linux driver. 6737 // Otherwise we might completely fill our space, at which point due to COW we can't 6738 // delete anything in order to fix this. 6739 6740 reserve = Vcb->extent_root->root_item.bytes_used; 6741 reserve += Vcb->root_root->root_item.bytes_used; 6742 if (Vcb->checksum_root) reserve += Vcb->checksum_root->root_item.bytes_used; 6743 6744 reserve = max(reserve, 0x1000000); // 16 M 6745 reserve = min(reserve, 0x20000000); // 512 M 6746 6747 // Find out how much space would be available for new metadata chunks 6748 6749 could_alloc = 0; 6750 6751 if (Vcb->metadata_flags & BLOCK_FLAG_RAID5) { 6752 UINT64 s1 = 0, s2 = 0, s3 = 0; 6753 6754 le = Vcb->devices.Flink; 6755 while (le != &Vcb->devices) { 6756 device* dev = CONTAINING_RECORD(le, device, list_entry); 6757 6758 if (!dev->readonly) { 6759 UINT64 space = dev->devitem.num_bytes - dev->devitem.bytes_used; 6760 6761 if (space >= s1) { 6762 s3 = s2; 6763 s2 = s1; 6764 s1 = space; 6765 } else if (space >= s2) { 6766 s3 = s2; 6767 s2 = space; 6768 } else if (space >= s3) 6769 s3 = space; 6770 } 6771 6772 le = le->Flink; 6773 } 6774 6775 could_alloc = s3 * 2; 6776 } else if (Vcb->metadata_flags & (BLOCK_FLAG_RAID10 | BLOCK_FLAG_RAID6)) { 6777 UINT64 s1 = 0, s2 = 0, s3 = 0, s4 = 0; 6778 6779 le = Vcb->devices.Flink; 6780 while (le != &Vcb->devices) { 6781 device* dev = CONTAINING_RECORD(le, device, list_entry); 6782 6783 if (!dev->readonly) { 6784 UINT64 space = dev->devitem.num_bytes - dev->devitem.bytes_used; 6785 6786 if (space >= s1) { 6787 s4 = s3; 6788 s3 = s2; 6789 s2 = s1; 6790 s1 = space; 6791 } else if (space >= s2) { 6792 s4 = s3; 6793 s3 = s2; 6794 s2 = space; 6795 } else if (space >= s3) { 6796 s4 = s3; 6797 s3 = space; 6798 } else if (space >= s4) 6799 s4 = space; 6800 } 6801 6802 le = le->Flink; 6803 } 6804 6805 could_alloc = s4 * 2; 6806 } else if (Vcb->metadata_flags & (BLOCK_FLAG_RAID0 | BLOCK_FLAG_RAID1)) { 6807 UINT64 s1 = 0, s2 = 0; 6808 6809 le = Vcb->devices.Flink; 6810 while (le != &Vcb->devices) { 6811 device* dev = CONTAINING_RECORD(le, device, list_entry); 6812 6813 if (!dev->readonly) { 6814 UINT64 space = dev->devitem.num_bytes - dev->devitem.bytes_used; 6815 6816 if (space >= s1) { 6817 s2 = s1; 6818 s1 = space; 6819 } else if (space >= s2) 6820 s2 = space; 6821 } 6822 6823 le = le->Flink; 6824 } 6825 6826 if (Vcb->metadata_flags & BLOCK_FLAG_RAID1) 6827 could_alloc = s2; 6828 else // RAID0 6829 could_alloc = s2 * 2; 6830 } else if (Vcb->metadata_flags & BLOCK_FLAG_DUPLICATE) { 6831 le = Vcb->devices.Flink; 6832 while (le != &Vcb->devices) { 6833 device* dev = CONTAINING_RECORD(le, device, list_entry); 6834 6835 if (!dev->readonly) { 6836 UINT64 space = (dev->devitem.num_bytes - dev->devitem.bytes_used) / 2; 6837 6838 could_alloc = max(could_alloc, space); 6839 } 6840 6841 le = le->Flink; 6842 } 6843 } else { // SINGLE 6844 le = Vcb->devices.Flink; 6845 while (le != &Vcb->devices) { 6846 device* dev = CONTAINING_RECORD(le, device, list_entry); 6847 6848 if (!dev->readonly) { 6849 UINT64 space = dev->devitem.num_bytes - dev->devitem.bytes_used; 6850 6851 could_alloc = max(could_alloc, space); 6852 } 6853 6854 le = le->Flink; 6855 } 6856 } 6857 6858 if (could_alloc >= reserve) 6859 return STATUS_SUCCESS; 6860 6861 free_space = 0; 6862 6863 le = Vcb->chunks.Flink; 6864 while (le != &Vcb->chunks) { 6865 chunk* c = CONTAINING_RECORD(le, chunk, list_entry); 6866 6867 if (!c->reloc && !c->readonly && c->chunk_item->type & BLOCK_FLAG_METADATA) { 6868 free_space += c->chunk_item->size - c->used; 6869 6870 if (free_space + could_alloc >= reserve) 6871 return STATUS_SUCCESS; 6872 } 6873 6874 le = le->Flink; 6875 } 6876 6877 return STATUS_DISK_FULL; 6878 } 6879 6880 static NTSTATUS do_write2(device_extension* Vcb, PIRP Irp, LIST_ENTRY* rollback) { 6881 NTSTATUS Status; 6882 LIST_ENTRY *le, batchlist; 6883 BOOL cache_changed = FALSE; 6884 volume_device_extension* vde; 6885 BOOL no_cache = FALSE; 6886 #ifdef DEBUG_FLUSH_TIMES 6887 UINT64 filerefs = 0, fcbs = 0; 6888 LARGE_INTEGER freq, time1, time2; 6889 #endif 6890 #ifdef DEBUG_WRITE_LOOPS 6891 UINT loops = 0; 6892 #endif 6893 6894 TRACE("(%p)\n", Vcb); 6895 6896 InitializeListHead(&batchlist); 6897 6898 #ifdef DEBUG_FLUSH_TIMES 6899 time1 = KeQueryPerformanceCounter(&freq); 6900 #endif 6901 6902 ExAcquireResourceExclusiveLite(&Vcb->dirty_filerefs_lock, TRUE); 6903 6904 while (!IsListEmpty(&Vcb->dirty_filerefs)) { 6905 file_ref* fr = CONTAINING_RECORD(RemoveHeadList(&Vcb->dirty_filerefs), file_ref, list_entry_dirty); 6906 6907 flush_fileref(fr, &batchlist, Irp); 6908 free_fileref(Vcb, fr); 6909 6910 #ifdef DEBUG_FLUSH_TIMES 6911 filerefs++; 6912 #endif 6913 } 6914 6915 ExReleaseResourceLite(&Vcb->dirty_filerefs_lock); 6916 6917 Status = commit_batch_list(Vcb, &batchlist, Irp); 6918 if (!NT_SUCCESS(Status)) { 6919 ERR("commit_batch_list returned %08x\n", Status); 6920 return Status; 6921 } 6922 6923 #ifdef DEBUG_FLUSH_TIMES 6924 time2 = KeQueryPerformanceCounter(NULL); 6925 6926 ERR("flushed %llu filerefs in %llu (freq = %llu)\n", filerefs, time2.QuadPart - time1.QuadPart, freq.QuadPart); 6927 6928 time1 = KeQueryPerformanceCounter(&freq); 6929 #endif 6930 6931 // We process deleted streams first, so we don't run over our xattr 6932 // limit unless we absolutely have to. 6933 // We also process deleted normal files, to avoid any problems 6934 // caused by inode collisions. 6935 6936 ExAcquireResourceExclusiveLite(&Vcb->dirty_fcbs_lock, TRUE); 6937 6938 le = Vcb->dirty_fcbs.Flink; 6939 while (le != &Vcb->dirty_fcbs) { 6940 fcb* fcb = CONTAINING_RECORD(le, struct _fcb, list_entry_dirty); 6941 LIST_ENTRY* le2 = le->Flink; 6942 6943 if (fcb->deleted) { 6944 ExAcquireResourceExclusiveLite(fcb->Header.Resource, TRUE); 6945 Status = flush_fcb(fcb, FALSE, &batchlist, Irp); 6946 ExReleaseResourceLite(fcb->Header.Resource); 6947 6948 free_fcb(Vcb, fcb); 6949 6950 if (!NT_SUCCESS(Status)) { 6951 ERR("flush_fcb returned %08x\n", Status); 6952 clear_batch_list(Vcb, &batchlist); 6953 ExReleaseResourceLite(&Vcb->dirty_fcbs_lock); 6954 return Status; 6955 } 6956 6957 #ifdef DEBUG_FLUSH_TIMES 6958 fcbs++; 6959 #endif 6960 } 6961 6962 le = le2; 6963 } 6964 6965 Status = commit_batch_list(Vcb, &batchlist, Irp); 6966 if (!NT_SUCCESS(Status)) { 6967 ERR("commit_batch_list returned %08x\n", Status); 6968 ExReleaseResourceLite(&Vcb->dirty_fcbs_lock); 6969 return Status; 6970 } 6971 6972 le = Vcb->dirty_fcbs.Flink; 6973 while (le != &Vcb->dirty_fcbs) { 6974 fcb* fcb = CONTAINING_RECORD(le, struct _fcb, list_entry_dirty); 6975 LIST_ENTRY* le2 = le->Flink; 6976 6977 if (fcb->subvol != Vcb->root_root) { 6978 ExAcquireResourceExclusiveLite(fcb->Header.Resource, TRUE); 6979 Status = flush_fcb(fcb, FALSE, &batchlist, Irp); 6980 ExReleaseResourceLite(fcb->Header.Resource); 6981 free_fcb(Vcb, fcb); 6982 6983 if (!NT_SUCCESS(Status)) { 6984 ERR("flush_fcb returned %08x\n", Status); 6985 ExReleaseResourceLite(&Vcb->dirty_fcbs_lock); 6986 return Status; 6987 } 6988 6989 #ifdef DEBUG_FLUSH_TIMES 6990 fcbs++; 6991 #endif 6992 } 6993 6994 le = le2; 6995 } 6996 6997 ExReleaseResourceLite(&Vcb->dirty_fcbs_lock); 6998 6999 Status = commit_batch_list(Vcb, &batchlist, Irp); 7000 if (!NT_SUCCESS(Status)) { 7001 ERR("commit_batch_list returned %08x\n", Status); 7002 return Status; 7003 } 7004 7005 #ifdef DEBUG_FLUSH_TIMES 7006 time2 = KeQueryPerformanceCounter(NULL); 7007 7008 ERR("flushed %llu fcbs in %llu (freq = %llu)\n", filerefs, time2.QuadPart - time1.QuadPart, freq.QuadPart); 7009 #endif 7010 7011 // no need to get dirty_subvols_lock here, as we have tree_lock exclusively 7012 while (!IsListEmpty(&Vcb->dirty_subvols)) { 7013 root* r = CONTAINING_RECORD(RemoveHeadList(&Vcb->dirty_subvols), root, list_entry_dirty); 7014 7015 Status = flush_subvol(Vcb, r, Irp); 7016 if (!NT_SUCCESS(Status)) { 7017 ERR("flush_subvol returned %08x\n", Status); 7018 return Status; 7019 } 7020 } 7021 7022 if (!IsListEmpty(&Vcb->drop_roots)) { 7023 Status = drop_roots(Vcb, Irp, rollback); 7024 7025 if (!NT_SUCCESS(Status)) { 7026 ERR("drop_roots returned %08x\n", Status); 7027 return Status; 7028 } 7029 } 7030 7031 Status = update_chunks(Vcb, &batchlist, Irp, rollback); 7032 7033 if (!NT_SUCCESS(Status)) { 7034 ERR("update_chunks returned %08x\n", Status); 7035 return Status; 7036 } 7037 7038 Status = commit_batch_list(Vcb, &batchlist, Irp); 7039 7040 // If only changing superblock, e.g. changing label, we still need to rewrite 7041 // the root tree so the generations match, otherwise you won't be able to mount on Linux. 7042 if (!Vcb->root_root->treeholder.tree || !Vcb->root_root->treeholder.tree->write) { 7043 KEY searchkey; 7044 7045 traverse_ptr tp; 7046 7047 searchkey.obj_id = 0; 7048 searchkey.obj_type = 0; 7049 searchkey.offset = 0; 7050 7051 Status = find_item(Vcb, Vcb->root_root, &tp, &searchkey, FALSE, Irp); 7052 if (!NT_SUCCESS(Status)) { 7053 ERR("error - find_item returned %08x\n", Status); 7054 return Status; 7055 } 7056 7057 Vcb->root_root->treeholder.tree->write = TRUE; 7058 } 7059 7060 // make sure we always update the extent tree 7061 Status = add_root_item_to_cache(Vcb, BTRFS_ROOT_EXTENT, Irp); 7062 if (!NT_SUCCESS(Status)) { 7063 ERR("add_root_item_to_cache returned %08x\n", Status); 7064 return Status; 7065 } 7066 7067 if (Vcb->stats_changed) { 7068 le = Vcb->devices.Flink; 7069 while (le != &Vcb->devices) { 7070 device* dev = CONTAINING_RECORD(le, device, list_entry); 7071 7072 if (dev->stats_changed) { 7073 Status = flush_changed_dev_stats(Vcb, dev, Irp); 7074 if (!NT_SUCCESS(Status)) { 7075 ERR("flush_changed_dev_stats returned %08x\n", Status); 7076 return Status; 7077 } 7078 dev->stats_changed = FALSE; 7079 } 7080 7081 le = le->Flink; 7082 } 7083 7084 Vcb->stats_changed = FALSE; 7085 } 7086 7087 do { 7088 Status = add_parents(Vcb, Irp); 7089 if (!NT_SUCCESS(Status)) { 7090 ERR("add_parents returned %08x\n", Status); 7091 goto end; 7092 } 7093 7094 Status = allocate_tree_extents(Vcb, Irp, rollback); 7095 if (!NT_SUCCESS(Status)) { 7096 ERR("allocate_tree_extents returned %08x\n", Status); 7097 goto end; 7098 } 7099 7100 Status = do_splits(Vcb, Irp, rollback); 7101 if (!NT_SUCCESS(Status)) { 7102 ERR("do_splits returned %08x\n", Status); 7103 goto end; 7104 } 7105 7106 Status = update_chunk_usage(Vcb, Irp, rollback); 7107 if (!NT_SUCCESS(Status)) { 7108 ERR("update_chunk_usage returned %08x\n", Status); 7109 goto end; 7110 } 7111 7112 if (!(Vcb->superblock.compat_ro_flags & BTRFS_COMPAT_RO_FLAGS_FREE_SPACE_CACHE)) { 7113 if (!no_cache) { 7114 Status = allocate_cache(Vcb, &cache_changed, Irp, rollback); 7115 if (!NT_SUCCESS(Status)) { 7116 WARN("allocate_cache returned %08x\n", Status); 7117 no_cache = TRUE; 7118 cache_changed = FALSE; 7119 } 7120 } 7121 } else { 7122 Status = update_chunk_caches_tree(Vcb, Irp); 7123 if (!NT_SUCCESS(Status)) { 7124 ERR("update_chunk_caches_tree returned %08x\n", Status); 7125 goto end; 7126 } 7127 } 7128 7129 #ifdef DEBUG_WRITE_LOOPS 7130 loops++; 7131 7132 if (cache_changed) 7133 ERR("cache has changed, looping again\n"); 7134 #endif 7135 } while (cache_changed || !trees_consistent(Vcb)); 7136 7137 #ifdef DEBUG_WRITE_LOOPS 7138 ERR("%u loops\n", loops); 7139 #endif 7140 7141 TRACE("trees consistent\n"); 7142 7143 Status = update_root_root(Vcb, no_cache, Irp, rollback); 7144 if (!NT_SUCCESS(Status)) { 7145 ERR("update_root_root returned %08x\n", Status); 7146 goto end; 7147 } 7148 7149 Status = write_trees(Vcb, Irp); 7150 if (!NT_SUCCESS(Status)) { 7151 ERR("write_trees returned %08x\n", Status); 7152 goto end; 7153 } 7154 7155 Status = test_not_full(Vcb); 7156 if (!NT_SUCCESS(Status)) { 7157 ERR("test_not_full returned %08x\n", Status); 7158 goto end; 7159 } 7160 7161 #ifdef DEBUG_PARANOID 7162 le = Vcb->trees.Flink; 7163 while (le != &Vcb->trees) { 7164 tree* t = CONTAINING_RECORD(le, tree, list_entry); 7165 KEY searchkey; 7166 traverse_ptr tp; 7167 7168 searchkey.obj_id = t->header.address; 7169 searchkey.obj_type = TYPE_METADATA_ITEM; 7170 searchkey.offset = 0xffffffffffffffff; 7171 7172 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 7173 if (!NT_SUCCESS(Status)) { 7174 ERR("error - find_item returned %08x\n", Status); 7175 goto end; 7176 } 7177 7178 if (tp.item->key.obj_id != searchkey.obj_id || tp.item->key.obj_type != searchkey.obj_type) { 7179 searchkey.obj_id = t->header.address; 7180 searchkey.obj_type = TYPE_EXTENT_ITEM; 7181 searchkey.offset = 0xffffffffffffffff; 7182 7183 Status = find_item(Vcb, Vcb->extent_root, &tp, &searchkey, FALSE, Irp); 7184 if (!NT_SUCCESS(Status)) { 7185 ERR("error - find_item returned %08x\n", Status); 7186 goto end; 7187 } 7188 7189 if (tp.item->key.obj_id != searchkey.obj_id || tp.item->key.obj_type != searchkey.obj_type) { 7190 ERR("error - could not find entry in extent tree for tree at %llx\n", t->header.address); 7191 Status = STATUS_INTERNAL_ERROR; 7192 goto end; 7193 } 7194 } 7195 7196 le = le->Flink; 7197 } 7198 #endif 7199 7200 Vcb->superblock.cache_generation = Vcb->superblock.generation; 7201 7202 if (!Vcb->options.no_barrier) 7203 flush_disk_caches(Vcb); 7204 7205 Status = write_superblocks(Vcb, Irp); 7206 if (!NT_SUCCESS(Status)) { 7207 ERR("write_superblocks returned %08x\n", Status); 7208 goto end; 7209 } 7210 7211 vde = Vcb->vde; 7212 7213 if (vde) { 7214 pdo_device_extension* pdode = vde->pdode; 7215 7216 ExAcquireResourceSharedLite(&pdode->child_lock, TRUE); 7217 7218 le = pdode->children.Flink; 7219 7220 while (le != &pdode->children) { 7221 volume_child* vc = CONTAINING_RECORD(le, volume_child, list_entry); 7222 7223 vc->generation = Vcb->superblock.generation; 7224 le = le->Flink; 7225 } 7226 7227 ExReleaseResourceLite(&pdode->child_lock); 7228 } 7229 7230 clean_space_cache(Vcb); 7231 7232 le = Vcb->chunks.Flink; 7233 while (le != &Vcb->chunks) { 7234 chunk* c = CONTAINING_RECORD(le, chunk, list_entry); 7235 7236 c->changed = FALSE; 7237 c->space_changed = FALSE; 7238 7239 le = le->Flink; 7240 } 7241 7242 Vcb->superblock.generation++; 7243 7244 Status = STATUS_SUCCESS; 7245 7246 le = Vcb->trees.Flink; 7247 while (le != &Vcb->trees) { 7248 tree* t = CONTAINING_RECORD(le, tree, list_entry); 7249 7250 t->write = FALSE; 7251 7252 le = le->Flink; 7253 } 7254 7255 Vcb->need_write = FALSE; 7256 7257 while (!IsListEmpty(&Vcb->drop_roots)) { 7258 root* r = CONTAINING_RECORD(RemoveHeadList(&Vcb->drop_roots), root, list_entry); 7259 7260 ExDeleteResourceLite(&r->nonpaged->load_tree_lock); 7261 ExFreePool(r->nonpaged); 7262 ExFreePool(r); 7263 } 7264 7265 end: 7266 TRACE("do_write returning %08x\n", Status); 7267 7268 return Status; 7269 } 7270 7271 NTSTATUS do_write(device_extension* Vcb, PIRP Irp) { 7272 LIST_ENTRY rollback; 7273 NTSTATUS Status; 7274 7275 InitializeListHead(&rollback); 7276 7277 Status = do_write2(Vcb, Irp, &rollback); 7278 7279 if (!NT_SUCCESS(Status)) { 7280 ERR("do_write2 returned %08x, dropping into readonly mode\n", Status); 7281 Vcb->readonly = TRUE; 7282 FsRtlNotifyVolumeEvent(Vcb->root_file, FSRTL_VOLUME_FORCED_CLOSED); 7283 do_rollback(Vcb, &rollback); 7284 } else 7285 clear_rollback(&rollback); 7286 7287 return Status; 7288 } 7289 7290 #ifdef DEBUG_STATS 7291 static void print_stats(device_extension* Vcb) { 7292 ERR("READ STATS:\n"); 7293 ERR("number of reads: %llu\n", Vcb->stats.num_reads); 7294 ERR("data read: %llu bytes\n", Vcb->stats.data_read); 7295 ERR("total time taken: %llu\n", Vcb->stats.read_total_time); 7296 ERR("csum time taken: %llu\n", Vcb->stats.read_csum_time); 7297 ERR("disk time taken: %llu\n", Vcb->stats.read_disk_time); 7298 ERR("other time taken: %llu\n", Vcb->stats.read_total_time - Vcb->stats.read_csum_time - Vcb->stats.read_disk_time); 7299 7300 ERR("OPEN STATS:\n"); 7301 ERR("number of opens: %llu\n", Vcb->stats.num_opens); 7302 ERR("total time taken: %llu\n", Vcb->stats.open_total_time); 7303 ERR("number of overwrites: %llu\n", Vcb->stats.num_overwrites); 7304 ERR("total time taken: %llu\n", Vcb->stats.overwrite_total_time); 7305 ERR("number of creates: %llu\n", Vcb->stats.num_creates); 7306 ERR("calls to open_fcb: %llu\n", Vcb->stats.open_fcb_calls); 7307 ERR("time spent in open_fcb: %llu\n", Vcb->stats.open_fcb_time); 7308 ERR("total time taken: %llu\n", Vcb->stats.create_total_time); 7309 7310 RtlZeroMemory(&Vcb->stats, sizeof(debug_stats)); 7311 } 7312 #endif 7313 7314 static void do_flush(device_extension* Vcb) { 7315 NTSTATUS Status; 7316 7317 ExAcquireResourceExclusiveLite(&Vcb->tree_lock, TRUE); 7318 7319 #ifdef DEBUG_STATS 7320 print_stats(Vcb); 7321 #endif 7322 7323 if (Vcb->need_write && !Vcb->readonly) 7324 Status = do_write(Vcb, NULL); 7325 else 7326 Status = STATUS_SUCCESS; 7327 7328 free_trees(Vcb); 7329 7330 if (!NT_SUCCESS(Status)) 7331 ERR("do_write returned %08x\n", Status); 7332 7333 ExReleaseResourceLite(&Vcb->tree_lock); 7334 } 7335 7336 _Function_class_(KSTART_ROUTINE) 7337 #ifdef __REACTOS__ 7338 void NTAPI flush_thread(void* context) { 7339 #else 7340 void flush_thread(void* context) { 7341 #endif 7342 DEVICE_OBJECT* devobj = context; 7343 device_extension* Vcb = devobj->DeviceExtension; 7344 LARGE_INTEGER due_time; 7345 7346 ObReferenceObject(devobj); 7347 7348 KeInitializeTimer(&Vcb->flush_thread_timer); 7349 7350 due_time.QuadPart = (UINT64)Vcb->options.flush_interval * -10000000; 7351 7352 KeSetTimer(&Vcb->flush_thread_timer, due_time, NULL); 7353 7354 while (TRUE) { 7355 KeWaitForSingleObject(&Vcb->flush_thread_timer, Executive, KernelMode, FALSE, NULL); 7356 7357 if (!(devobj->Vpb->Flags & VPB_MOUNTED) || Vcb->removing) 7358 break; 7359 7360 if (!Vcb->locked) 7361 do_flush(Vcb); 7362 7363 KeSetTimer(&Vcb->flush_thread_timer, due_time, NULL); 7364 } 7365 7366 ObDereferenceObject(devobj); 7367 KeCancelTimer(&Vcb->flush_thread_timer); 7368 7369 KeSetEvent(&Vcb->flush_thread_finished, 0, FALSE); 7370 7371 PsTerminateSystemThread(STATUS_SUCCESS); 7372 } 7373