diff options
Diffstat (limited to 'fs/btrfs')
-rw-r--r-- | fs/btrfs/ioctl.c | 150 |
1 files changed, 140 insertions, 10 deletions
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index 7875a75315d0..3a89cd77f307 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c | |||
@@ -475,6 +475,73 @@ out_unlock: | |||
475 | return error; | 475 | return error; |
476 | } | 476 | } |
477 | 477 | ||
478 | static int should_defrag_range(struct inode *inode, u64 start, u64 len, | ||
479 | u64 *last_len, u64 *skip, u64 *defrag_end) | ||
480 | { | ||
481 | struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; | ||
482 | struct extent_map *em = NULL; | ||
483 | struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; | ||
484 | int ret = 1; | ||
485 | |||
486 | /* | ||
487 | * make sure that once we start defragging and extent, we keep on | ||
488 | * defragging it | ||
489 | */ | ||
490 | if (start < *defrag_end) | ||
491 | return 1; | ||
492 | |||
493 | *skip = 0; | ||
494 | |||
495 | /* | ||
496 | * hopefully we have this extent in the tree already, try without | ||
497 | * the full extent lock | ||
498 | */ | ||
499 | read_lock(&em_tree->lock); | ||
500 | em = lookup_extent_mapping(em_tree, start, len); | ||
501 | read_unlock(&em_tree->lock); | ||
502 | |||
503 | if (!em) { | ||
504 | /* get the big lock and read metadata off disk */ | ||
505 | lock_extent(io_tree, start, start + len - 1, GFP_NOFS); | ||
506 | em = btrfs_get_extent(inode, NULL, 0, start, len, 0); | ||
507 | unlock_extent(io_tree, start, start + len - 1, GFP_NOFS); | ||
508 | |||
509 | if (!em) | ||
510 | return 0; | ||
511 | } | ||
512 | |||
513 | /* this will cover holes, and inline extents */ | ||
514 | if (em->block_start >= EXTENT_MAP_LAST_BYTE) | ||
515 | ret = 0; | ||
516 | |||
517 | /* | ||
518 | * we hit a real extent, if it is big don't bother defragging it again | ||
519 | */ | ||
520 | if ((*last_len == 0 || *last_len >= 256 * 1024) && | ||
521 | em->len >= 256 * 1024) | ||
522 | ret = 0; | ||
523 | |||
524 | /* | ||
525 | * last_len ends up being a counter of how many bytes we've defragged. | ||
526 | * every time we choose not to defrag an extent, we reset *last_len | ||
527 | * so that the next tiny extent will force a defrag. | ||
528 | * | ||
529 | * The end result of this is that tiny extents before a single big | ||
530 | * extent will force at least part of that big extent to be defragged. | ||
531 | */ | ||
532 | if (ret) { | ||
533 | *last_len += len; | ||
534 | *defrag_end = extent_map_end(em); | ||
535 | } else { | ||
536 | *last_len = 0; | ||
537 | *skip = extent_map_end(em); | ||
538 | *defrag_end = 0; | ||
539 | } | ||
540 | |||
541 | free_extent_map(em); | ||
542 | return ret; | ||
543 | } | ||
544 | |||
478 | static int btrfs_defrag_file(struct file *file) | 545 | static int btrfs_defrag_file(struct file *file) |
479 | { | 546 | { |
480 | struct inode *inode = fdentry(file)->d_inode; | 547 | struct inode *inode = fdentry(file)->d_inode; |
@@ -487,37 +554,86 @@ static int btrfs_defrag_file(struct file *file) | |||
487 | unsigned long total_read = 0; | 554 | unsigned long total_read = 0; |
488 | u64 page_start; | 555 | u64 page_start; |
489 | u64 page_end; | 556 | u64 page_end; |
557 | u64 last_len = 0; | ||
558 | u64 skip = 0; | ||
559 | u64 defrag_end = 0; | ||
490 | unsigned long i; | 560 | unsigned long i; |
491 | int ret; | 561 | int ret; |
492 | 562 | ||
493 | ret = btrfs_check_data_free_space(root, inode, inode->i_size); | 563 | if (inode->i_size == 0) |
494 | if (ret) | 564 | return 0; |
495 | return -ENOSPC; | 565 | |
566 | last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT; | ||
567 | i = 0; | ||
568 | while (i <= last_index) { | ||
569 | if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT, | ||
570 | PAGE_CACHE_SIZE, &last_len, &skip, | ||
571 | &defrag_end)) { | ||
572 | unsigned long next; | ||
573 | /* | ||
574 | * the should_defrag function tells us how much to skip | ||
575 | * bump our counter by the suggested amount | ||
576 | */ | ||
577 | next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; | ||
578 | i = max(i + 1, next); | ||
579 | continue; | ||
580 | } | ||
496 | 581 | ||
497 | mutex_lock(&inode->i_mutex); | ||
498 | last_index = inode->i_size >> PAGE_CACHE_SHIFT; | ||
499 | for (i = 0; i <= last_index; i++) { | ||
500 | if (total_read % ra_pages == 0) { | 582 | if (total_read % ra_pages == 0) { |
501 | btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, | 583 | btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i, |
502 | min(last_index, i + ra_pages - 1)); | 584 | min(last_index, i + ra_pages - 1)); |
503 | } | 585 | } |
504 | total_read++; | 586 | total_read++; |
587 | mutex_lock(&inode->i_mutex); | ||
588 | |||
589 | ret = btrfs_check_data_free_space(root, inode, PAGE_CACHE_SIZE); | ||
590 | if (ret) { | ||
591 | ret = -ENOSPC; | ||
592 | break; | ||
593 | } | ||
594 | |||
595 | ret = btrfs_reserve_metadata_for_delalloc(root, inode, 1); | ||
596 | if (ret) { | ||
597 | btrfs_free_reserved_data_space(root, inode, | ||
598 | PAGE_CACHE_SIZE); | ||
599 | ret = -ENOSPC; | ||
600 | break; | ||
601 | } | ||
505 | again: | 602 | again: |
603 | if (inode->i_size == 0 || | ||
604 | i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) { | ||
605 | ret = 0; | ||
606 | goto err_reservations; | ||
607 | } | ||
608 | |||
506 | page = grab_cache_page(inode->i_mapping, i); | 609 | page = grab_cache_page(inode->i_mapping, i); |
507 | if (!page) | 610 | if (!page) |
508 | goto out_unlock; | 611 | goto err_reservations; |
612 | |||
509 | if (!PageUptodate(page)) { | 613 | if (!PageUptodate(page)) { |
510 | btrfs_readpage(NULL, page); | 614 | btrfs_readpage(NULL, page); |
511 | lock_page(page); | 615 | lock_page(page); |
512 | if (!PageUptodate(page)) { | 616 | if (!PageUptodate(page)) { |
513 | unlock_page(page); | 617 | unlock_page(page); |
514 | page_cache_release(page); | 618 | page_cache_release(page); |
515 | goto out_unlock; | 619 | goto err_reservations; |
516 | } | 620 | } |
517 | } | 621 | } |
518 | 622 | ||
623 | if (page->mapping != inode->i_mapping) { | ||
624 | unlock_page(page); | ||
625 | page_cache_release(page); | ||
626 | goto again; | ||
627 | } | ||
628 | |||
519 | wait_on_page_writeback(page); | 629 | wait_on_page_writeback(page); |
520 | 630 | ||
631 | if (PageDirty(page)) { | ||
632 | btrfs_free_reserved_data_space(root, inode, | ||
633 | PAGE_CACHE_SIZE); | ||
634 | goto loop_unlock; | ||
635 | } | ||
636 | |||
521 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; | 637 | page_start = (u64)page->index << PAGE_CACHE_SHIFT; |
522 | page_end = page_start + PAGE_CACHE_SIZE - 1; | 638 | page_end = page_start + PAGE_CACHE_SIZE - 1; |
523 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); | 639 | lock_extent(io_tree, page_start, page_end, GFP_NOFS); |
@@ -538,18 +654,32 @@ again: | |||
538 | * page if it is dirtied again later | 654 | * page if it is dirtied again later |
539 | */ | 655 | */ |
540 | clear_page_dirty_for_io(page); | 656 | clear_page_dirty_for_io(page); |
657 | clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start, | ||
658 | page_end, EXTENT_DIRTY | EXTENT_DELALLOC | | ||
659 | EXTENT_DO_ACCOUNTING, GFP_NOFS); | ||
541 | 660 | ||
542 | btrfs_set_extent_delalloc(inode, page_start, page_end); | 661 | btrfs_set_extent_delalloc(inode, page_start, page_end); |
662 | ClearPageChecked(page); | ||
543 | set_page_dirty(page); | 663 | set_page_dirty(page); |
544 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); | 664 | unlock_extent(io_tree, page_start, page_end, GFP_NOFS); |
665 | |||
666 | loop_unlock: | ||
545 | unlock_page(page); | 667 | unlock_page(page); |
546 | page_cache_release(page); | 668 | page_cache_release(page); |
669 | mutex_unlock(&inode->i_mutex); | ||
670 | |||
671 | btrfs_unreserve_metadata_for_delalloc(root, inode, 1); | ||
547 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); | 672 | balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1); |
673 | i++; | ||
548 | } | 674 | } |
549 | 675 | ||
550 | out_unlock: | ||
551 | mutex_unlock(&inode->i_mutex); | ||
552 | return 0; | 676 | return 0; |
677 | |||
678 | err_reservations: | ||
679 | mutex_unlock(&inode->i_mutex); | ||
680 | btrfs_free_reserved_data_space(root, inode, PAGE_CACHE_SIZE); | ||
681 | btrfs_unreserve_metadata_for_delalloc(root, inode, 1); | ||
682 | return ret; | ||
553 | } | 683 | } |
554 | 684 | ||
555 | static noinline int btrfs_ioctl_resize(struct btrfs_root *root, | 685 | static noinline int btrfs_ioctl_resize(struct btrfs_root *root, |