diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2022-07-16 13:48:55 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2022-07-16 13:48:55 -0700 |
commit | 972a278fe60c361eb8f37619f562f092e8786d7c (patch) | |
tree | f80ef197b2adcab1bf504b255ddf46398e13f39f /fs/btrfs/extent_io.c | |
parent | c5fe7a97f20c7f3070ac870144515c0fabc6b999 (diff) | |
parent | 088aea3b97e0ae5a2a86f5d142ad10fec8a1b80f (diff) | |
download | linux-972a278fe60c361eb8f37619f562f092e8786d7c.tar.gz |
Merge tag 'for-5.19-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux
Pull btrfs reverts from David Sterba:
"Due to a recent report [1] we need to revert the radix tree to xarray
conversion patches.
There's a problem with sleeping under spinlock, when xa_insert could
allocate memory under pressure. We use GFP_NOFS so this is a real
problem that we unfortunately did not discover during review.
I'm sorry to do such change at rc6 time but the revert is IMO the
safer option, there are patches to use mutex instead of the spin locks
but that would need more testing. The revert branch has been tested on
a few setups, all seem ok.
The conversion to xarray will be revisited in the future"
Link: https://lore.kernel.org/linux-btrfs/cover.1657097693.git.fdmanana@suse.com/ [1]
* tag 'for-5.19-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
Revert "btrfs: turn delayed_nodes_tree into an XArray"
Revert "btrfs: turn name_cache radix tree into XArray in send_ctx"
Revert "btrfs: turn fs_info member buffer_radix into XArray"
Revert "btrfs: turn fs_roots_radix in btrfs_fs_info into an XArray"
Diffstat (limited to 'fs/btrfs/extent_io.c')
-rw-r--r-- | fs/btrfs/extent_io.c | 122 |
1 files changed, 74 insertions, 48 deletions
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 04e36343da3a..f03ab5dbda7a 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -2966,7 +2966,7 @@ static void begin_page_read(struct btrfs_fs_info *fs_info, struct page *page) } /* - * Find extent buffer for a given bytenr. + * Find extent buffer for a givne bytenr. * * This is for end_bio_extent_readpage(), thus we can't do any unsafe locking * in endio context. @@ -2985,9 +2985,11 @@ static struct extent_buffer *find_extent_buffer_readpage( return (struct extent_buffer *)page->private; } - /* For subpage case, we need to lookup extent buffer xarray */ - eb = xa_load(&fs_info->extent_buffers, - bytenr >> fs_info->sectorsize_bits); + /* For subpage case, we need to lookup buffer radix tree */ + rcu_read_lock(); + eb = radix_tree_lookup(&fs_info->buffer_radix, + bytenr >> fs_info->sectorsize_bits); + rcu_read_unlock(); ASSERT(eb); return eb; } @@ -4435,8 +4437,8 @@ static struct extent_buffer *find_extent_buffer_nolock( struct extent_buffer *eb; rcu_read_lock(); - eb = xa_load(&fs_info->extent_buffers, - start >> fs_info->sectorsize_bits); + eb = radix_tree_lookup(&fs_info->buffer_radix, + start >> fs_info->sectorsize_bits); if (eb && atomic_inc_not_zero(&eb->refs)) { rcu_read_unlock(); return eb; @@ -6129,22 +6131,24 @@ struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info, if (!eb) return ERR_PTR(-ENOMEM); eb->fs_info = fs_info; - - do { - ret = xa_insert(&fs_info->extent_buffers, - start >> fs_info->sectorsize_bits, - eb, GFP_NOFS); - if (ret == -ENOMEM) { - exists = ERR_PTR(ret); +again: + ret = radix_tree_preload(GFP_NOFS); + if (ret) { + exists = ERR_PTR(ret); + goto free_eb; + } + spin_lock(&fs_info->buffer_lock); + ret = radix_tree_insert(&fs_info->buffer_radix, + start >> fs_info->sectorsize_bits, eb); + spin_unlock(&fs_info->buffer_lock); + radix_tree_preload_end(); + if (ret == -EEXIST) { + exists = find_extent_buffer(fs_info, start); + if (exists) goto free_eb; - } - if (ret == -EBUSY) { - exists = find_extent_buffer(fs_info, start); - if (exists) - goto free_eb; - } - } while (ret); - + else + goto again; + } check_buffer_tree_ref(eb); set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags); @@ -6319,22 +6323,25 @@ struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info, } if (uptodate) set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags); - - do { - ret = xa_insert(&fs_info->extent_buffers, - start >> fs_info->sectorsize_bits, - eb, GFP_NOFS); - if (ret == -ENOMEM) { - exists = ERR_PTR(ret); +again: + ret = radix_tree_preload(GFP_NOFS); + if (ret) { + exists = ERR_PTR(ret); + goto free_eb; + } + + spin_lock(&fs_info->buffer_lock); + ret = radix_tree_insert(&fs_info->buffer_radix, + start >> fs_info->sectorsize_bits, eb); + spin_unlock(&fs_info->buffer_lock); + radix_tree_preload_end(); + if (ret == -EEXIST) { + exists = find_extent_buffer(fs_info, start); + if (exists) goto free_eb; - } - if (ret == -EBUSY) { - exists = find_extent_buffer(fs_info, start); - if (exists) - goto free_eb; - } - } while (ret); - + else + goto again; + } /* add one reference for the tree */ check_buffer_tree_ref(eb); set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags); @@ -6379,8 +6386,10 @@ static int release_extent_buffer(struct extent_buffer *eb) spin_unlock(&eb->refs_lock); - xa_erase(&fs_info->extent_buffers, - eb->start >> fs_info->sectorsize_bits); + spin_lock(&fs_info->buffer_lock); + radix_tree_delete(&fs_info->buffer_radix, + eb->start >> fs_info->sectorsize_bits); + spin_unlock(&fs_info->buffer_lock); } else { spin_unlock(&eb->refs_lock); } @@ -7325,25 +7334,42 @@ void memmove_extent_buffer(const struct extent_buffer *dst, } } +#define GANG_LOOKUP_SIZE 16 static struct extent_buffer *get_next_extent_buffer( struct btrfs_fs_info *fs_info, struct page *page, u64 bytenr) { - struct extent_buffer *eb; - unsigned long index; + struct extent_buffer *gang[GANG_LOOKUP_SIZE]; + struct extent_buffer *found = NULL; u64 page_start = page_offset(page); + u64 cur = page_start; ASSERT(in_range(bytenr, page_start, PAGE_SIZE)); lockdep_assert_held(&fs_info->buffer_lock); - xa_for_each_start(&fs_info->extent_buffers, index, eb, - page_start >> fs_info->sectorsize_bits) { - if (in_range(eb->start, page_start, PAGE_SIZE)) - return eb; - else if (eb->start >= page_start + PAGE_SIZE) - /* Already beyond page end */ - return NULL; + while (cur < page_start + PAGE_SIZE) { + int ret; + int i; + + ret = radix_tree_gang_lookup(&fs_info->buffer_radix, + (void **)gang, cur >> fs_info->sectorsize_bits, + min_t(unsigned int, GANG_LOOKUP_SIZE, + PAGE_SIZE / fs_info->nodesize)); + if (ret == 0) + goto out; + for (i = 0; i < ret; i++) { + /* Already beyond page end */ + if (gang[i]->start >= page_start + PAGE_SIZE) + goto out; + /* Found one */ + if (gang[i]->start >= bytenr) { + found = gang[i]; + goto out; + } + } + cur = gang[ret - 1]->start + gang[ret - 1]->len; } - return NULL; +out: + return found; } static int try_release_subpage_extent_buffer(struct page *page) |