For the subpage-blocksize scenario, a page can contain multiple
blocks. In such cases, this patch handles reading data from files.
To track the status of individual blocks of a page, this patch makes use
of a bitmap pointed to by the newly introduced per-page 'struct
btrfs_page_private'.
The per-page btrfs_page_private->io_lock plays the same role as
BH_Uptodate_Lock (see end_buffer_async_read()) i.e. without the io_lock
we may end up in the following situation,
NOTE: Assume 64k page size and 4k block size. Also assume that the first
12 blocks of the page are contiguous while the next 4 blocks are
contiguous. When reading the page we end up submitting two "logical
address space" bios. So end_bio_extent_readpage function is invoked
twice, once for each bio.
|-------------------------+-------------------------+-------------|
| Task A | Task B | Task C |
|-------------------------+-------------------------+-------------|
| end_bio_extent_readpage | | |
| process block 0 | | |
| - clear BLK_STATE_IO | | |
| - page_read_complete | | |
| process block 1 | | |
| | | |
| | | |
| | end_bio_extent_readpage | |
| | process block 0 | |
| | - clear BLK_STATE_IO | |
| | - page_read_complete | |
| | process block 1 | |
| | | |
| process block 11 | process block 3 | |
| - clear BLK_STATE_IO | - clear BLK_STATE_IO | |
| - page_read_complete | - page_read_complete | |
| - returns true | - returns true | |
| - unlock_page() | | |
| | | lock_page() |
| | - unlock_page() | |
|-------------------------+-------------------------+-------------|
We end up incorrectly unlocking the page twice and "Task C" ends up
working on an unlocked page. So private->io_lock makes sure that only
one of the tasks gets "true" as the return value when page_io_complete()
is invoked. As an optimization the patch gets the io_lock only when the
last block of the bio_vec is being processed.
Signed-off-by: Chandan Rajendra <chan...@linux.vnet.ibm.com>
---
fs/btrfs/extent_io.c | 299 +++++++++++++++++++++++++++++++++++++++++----------
fs/btrfs/extent_io.h | 76 ++++++++++++-
fs/btrfs/inode.c | 13 +--
3 files changed, 320 insertions(+), 68 deletions(-)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 522c943..b3885cc 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -23,6 +23,7 @@
static struct kmem_cache *extent_state_cache;
static struct kmem_cache *extent_buffer_cache;
+static struct kmem_cache *page_private_cache;
static struct bio_set *btrfs_bioset;
static inline bool extent_state_in_tree(const struct extent_state *state)
@@ -163,10 +164,16 @@ int __init extent_io_init(void)
if (!extent_buffer_cache)
goto free_state_cache;
+ page_private_cache = kmem_cache_create("btrfs_page_private",
+ sizeof(struct btrfs_page_private), 0,
+ SLAB_MEM_SPREAD, NULL);
+ if (!page_private_cache)
+ goto free_buffer_cache;
+
btrfs_bioset = bioset_create(BIO_POOL_SIZE,
offsetof(struct btrfs_io_bio, bio));
if (!btrfs_bioset)
- goto free_buffer_cache;
+ goto free_page_private_cache;
if (bioset_integrity_create(btrfs_bioset, BIO_POOL_SIZE))
goto free_bioset;
@@ -177,6 +184,10 @@ free_bioset:
bioset_free(btrfs_bioset);
btrfs_bioset = NULL;
+free_page_private_cache:
+ kmem_cache_destroy(page_private_cache);
+ page_private_cache = NULL;
+
free_buffer_cache:
kmem_cache_destroy(extent_buffer_cache);
extent_buffer_cache = NULL;
@@ -1311,6 +1322,96 @@ int clear_record_extent_bits(struct extent_io_tree
*tree, u64 start, u64 end,
changeset);
}
+static int modify_page_blks_state(struct page *page,
+ unsigned long blk_states,
+ u64 start, u64 end, int set)
+{
+ struct inode *inode = page->mapping->host;
+ unsigned long *bitmap;
+ unsigned long first_state;
+ unsigned long state;
+ u64 nr_blks;
+ u64 blk;
+
+ if (BTRFS_I(inode)->root->sectorsize == PAGE_SIZE)
+ return 0;
+
+ bitmap = ((struct btrfs_page_private *)page->private)->bstate;
+
+ blk = BTRFS_BYTES_TO_BLKS(BTRFS_I(inode)->root->fs_info,
+ start & (PAGE_SIZE - 1));
+ nr_blks = BTRFS_BYTES_TO_BLKS(BTRFS_I(inode)->root->fs_info,
+ (end - start + 1));
+
+ first_state = find_next_bit(&blk_states, BLK_NR_STATE, 0);
+
+ while (nr_blks--) {
+ state = first_state;
+
+ while (state < BLK_NR_STATE) {
+ if (set)
+ set_bit((blk * BLK_NR_STATE) + state, bitmap);
+ else
+ clear_bit((blk * BLK_NR_STATE) + state, bitmap);
+
+ state = find_next_bit(&blk_states, BLK_NR_STATE,
+ state + 1);
+ }
+
+ ++blk;
+ }
+
+ return 0;
+}
+
+int set_page_blks_state(struct page *page, unsigned long blk_states,
+ u64 start, u64 end)
+{
+ return modify_page_blks_state(page, blk_states, start, end, 1);
+}
+
+int clear_page_blks_state(struct page *page, unsigned long blk_states,
+ u64 start, u64 end)
+{
+ return modify_page_blks_state(page, blk_states, start, end, 0);
+}
+
+int test_page_blks_state(struct page *page, enum blk_state blk_state,
+ u64 start, u64 end, int check_all)
+{
+ struct inode *inode = page->mapping->host;
+ unsigned long *bitmap;
+ unsigned long blk;
+ u64 nr_blks;
+ int found = 0;
+
+ ASSERT(BTRFS_I(inode)->root->sectorsize < PAGE_SIZE);
+
+ bitmap = ((struct btrfs_page_private *)page->private)->bstate;
+
+ blk = BTRFS_BYTES_TO_BLKS(BTRFS_I(inode)->root->fs_info,
+ start & (PAGE_SIZE - 1));
+ nr_blks = BTRFS_BYTES_TO_BLKS(BTRFS_I(inode)->root->fs_info,
+ (end - start + 1));
+
+ while (nr_blks--) {
+ if (test_bit((blk * BLK_NR_STATE) + blk_state, bitmap)) {
+ if (!check_all)
+ return 1;
+ found = 1;
+ } else if (check_all) {
+ return 0;
+ }
+
+ ++blk;
+ }
+
+ if (!check_all && !found)
+ return 0;
+
+ return 1;
+}
+
/*
* either insert or lock state struct between start and end use mask to tell
* us if waiting is desired.
@@ -1950,9 +2051,25 @@ int test_range_bit(struct extent_io_tree *tree, u64
start, u64 end,
* helper function to set a given page up to date if all the
* extents in the tree for that page are up to date
*/
-static void check_page_uptodate(struct page *page)
+void check_page_uptodate(struct page *page)
{
- SetPageUptodate(page);
+ struct inode *inode = page->mapping->host;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ u64 start = page_offset(page);
+ u64 end = start + PAGE_SIZE - 1;
+
+ if (root->sectorsize == PAGE_SIZE
+ || test_page_blks_state(page, BLK_STATE_UPTODATE, start,
+ end, 1))
+ SetPageUptodate(page);
+}
+
+static int page_io_complete(struct page *page)
+{
+ u64 start = page_offset(page);
+ u64 end = start + PAGE_SIZE - 1;
+
+ return !test_page_blks_state(page, BLK_STATE_IO, start, end, 0);
}
int free_io_failure(struct extent_io_tree *failure_tree,
@@ -2282,7 +2399,9 @@ int btrfs_check_repairable(struct inode *inode, struct
bio *failed_bio,
* a) deliver good data to the caller
* b) correct the bad sectors on disk
*/
- if (failed_bio->bi_vcnt > 1) {
+ if ((failed_bio->bi_vcnt > 1)
+ || (failed_bio->bi_io_vec->bv_len
+ > BTRFS_I(inode)->root->sectorsize)) {
/*
* to fulfill b), we need to know the exact failing sectors, as
* we don't want to rewrite any more than the failed ones. thus,
@@ -2506,17 +2625,21 @@ static void end_bio_extent_readpage(struct bio *bio)
int uptodate = !bio->bi_error;
struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
struct extent_io_tree *tree, *failure_tree;
+ struct btrfs_page_private *pg_private;
+ unsigned long flags;
u64 offset = 0;
u64 start;
u64 end;
- u64 len;
+ int nr_sectors;
int mirror;
+ int unlock;
int ret;
int i;
bio_for_each_segment_all(bvec, bio, i) {
struct page *page = bvec->bv_page;
struct inode *inode = page->mapping->host;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
"mirror=%u\n", (u64)bio->bi_iter.bi_sector,
@@ -2524,28 +2647,14 @@ static void end_bio_extent_readpage(struct bio *bio)
tree = &BTRFS_I(inode)->io_tree;
failure_tree = &BTRFS_I(inode)->io_failure_tree;
- /* We always issue full-page reads, but if some block
- * in a page fails to read, blk_update_request() will
- * advance bv_offset and adjust bv_len to compensate.
- * Print a warning for nonzero offsets, and an error
- * if they don't add up to a full page. */
- if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
- if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
-
btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
- "partial page read in btrfs with offset %u
and length %u",
- bvec->bv_offset, bvec->bv_len);
- else
-
btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
- "incomplete page read in btrfs with offset
%u and "
- "length %u",
- bvec->bv_offset, bvec->bv_len);
- }
+ start = page_offset(page) + bvec->bv_offset;
+ nr_sectors = BTRFS_BYTES_TO_BLKS(root->fs_info,
+ bvec->bv_len);
+ mirror = io_bio->mirror_num;
- start = page_offset(page);
- end = start + bvec->bv_offset + bvec->bv_len - 1;
- len = bvec->bv_len;
+next_block:
+ end = start + root->sectorsize - 1;
- mirror = io_bio->mirror_num;
if (likely(uptodate && tree->ops &&
tree->ops->readpage_end_io_hook)) {
ret = tree->ops->readpage_end_io_hook(io_bio, offset,
@@ -2556,17 +2665,11 @@ static void end_bio_extent_readpage(struct bio *bio)
else
clean_io_failure(BTRFS_I(inode)->root->fs_info,
failure_tree, tree, start,
- page, btrfs_ino(inode), 0);
+ page, btrfs_ino(inode),
+ start - page_offset(page));
}
- if (likely(uptodate))
- goto readpage_ok;
-
- if (tree->ops && tree->ops->readpage_io_failed_hook) {
- ret = tree->ops->readpage_io_failed_hook(page, mirror);
- if (!ret && !bio->bi_error)
- uptodate = 1;
- } else {
+ if (!uptodate) {
/*
* The generic bio_readpage_error handles errors the
* following way: If possible, new read requests are
@@ -2581,30 +2684,58 @@ static void end_bio_extent_readpage(struct bio *bio)
mirror);
if (ret == 0) {
uptodate = !bio->bi_error;
- offset += len;
- continue;
+ offset += root->sectorsize;
+ if (--nr_sectors) {
+ start = end + 1;
+ goto next_block;
+ } else {
+ continue;
+ }
}
}
-readpage_ok:
- if (likely(uptodate)) {
- loff_t i_size = i_size_read(inode);
- pgoff_t end_index = i_size >> PAGE_SHIFT;
- unsigned off;
-
- /* Zero out the end if this page straddles i_size */
- off = i_size & (PAGE_SIZE-1);
- if (page->index == end_index && off)
- zero_user_segment(page, off, PAGE_SIZE);
+
+ if (uptodate) {
+ set_page_blks_state(page, 1 << BLK_STATE_UPTODATE,
+ start, end);
check_page_uptodate(page);
} else {
ClearPageUptodate(page);
SetPageError(page);
}
- unlock_page(page);
- offset += len;
- unlock_extent_cached(tree, start, end, NULL, GFP_ATOMIC);
+ offset += root->sectorsize;
+
+ if (--nr_sectors) {
+ clear_page_blks_state(page, 1 << BLK_STATE_IO,
+ start, end);
+ clear_extent_bit(tree, start, end,
+ EXTENT_LOCKED, 1, 0, NULL, GFP_ATOMIC);
+ start = end + 1;
+ goto next_block;
+ }
+
+ WARN_ON(!PagePrivate(page));
+
+ unlock = 1;
+
+ if (root->sectorsize < PAGE_SIZE) {
+ pg_private = (struct btrfs_page_private *)page->private;
+
+ spin_lock_irqsave(&pg_private->io_lock, flags);
+
+ clear_page_blks_state(page, 1 << BLK_STATE_IO,
+ start, end);
+
+ unlock = page_io_complete(page);
+
+ spin_unlock_irqrestore(&pg_private->io_lock, flags);
+ }
+
+ clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
+ GFP_ATOMIC);
+ if (unlock)
+ unlock_page(page);
}
if (io_bio->end_io)
@@ -2794,13 +2925,51 @@ static void attach_extent_buffer_page(struct
extent_buffer *eb,
}
}
-void set_page_extent_mapped(struct page *page)
+int set_page_extent_mapped(struct page *page)
{
+ struct inode *inode = page->mapping->host;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_page_private *pg_private;
+ unsigned long private = EXTENT_PAGE_PRIVATE;
+
if (!PagePrivate(page)) {
+ if (root->sectorsize < PAGE_SIZE) {
+ pg_private = kmem_cache_zalloc(page_private_cache,
+ GFP_NOFS);
+ if (!pg_private)
+ return -ENOMEM;
+
+ spin_lock_init(&pg_private->io_lock);
+
+ private = (unsigned long)pg_private;
+ }
+
SetPagePrivate(page);
get_page(page);
- set_page_private(page, EXTENT_PAGE_PRIVATE);
+ set_page_private(page, private);
}
+
+ return 0;
+}
+
+int clear_page_extent_mapped(struct page *page)
+{
+ struct inode *inode = page->mapping->host;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
+ struct btrfs_page_private *pg_private;
+
+ if (PagePrivate(page)) {
+ if (root->sectorsize < PAGE_SIZE) {
+ pg_private = (struct btrfs_page_private
*)(page->private);
+ kmem_cache_free(page_private_cache, pg_private);
+ }
+
+ ClearPagePrivate(page);
+ set_page_private(page, 0);
+ put_page(page);
+ }
+
+ return 0;
}
static struct extent_map *
@@ -2846,6 +3015,7 @@ static int __do_readpage(struct extent_io_tree *tree,
u64 *prev_em_start)
{
struct inode *inode = page->mapping->host;
+ struct btrfs_root *root = BTRFS_I(inode)->root;
u64 start = page_offset(page);
u64 page_end = start + PAGE_SIZE - 1;
u64 end;
@@ -2868,12 +3038,10 @@ static int __do_readpage(struct extent_io_tree *tree,
set_page_extent_mapped(page);
end = page_end;
- if (!PageUptodate(page)) {
- if (cleancache_get_page(page) == 0) {
- BUG_ON(blocksize != PAGE_SIZE);
- unlock_extent(tree, start, end);
- goto out;
- }
+ if ((blocksize == PAGE_SIZE) && !PageUptodate(page) &&
+ (cleancache_get_page(page) == 0)) {
+ unlock_extent(tree, start, end);
+ goto out;
}
if (page->index == last_byte >> PAGE_SHIFT) {
@@ -2900,6 +3068,8 @@ static int __do_readpage(struct extent_io_tree *tree,
memset(userpage + pg_offset, 0, iosize);
flush_dcache_page(page);
kunmap_atomic(userpage);
+ set_page_blks_state(page, 1 << BLK_STATE_UPTODATE, cur,
+ cur + iosize - 1);
unlock_extent_cached(tree, cur,
cur + iosize - 1,
NULL, GFP_NOFS);
@@ -2992,6 +3162,9 @@ static int __do_readpage(struct extent_io_tree *tree,
flush_dcache_page(page);
kunmap_atomic(userpage);
+ set_page_blks_state(page, 1 << BLK_STATE_UPTODATE, cur,
+ cur + iosize - 1);
+
unlock_extent_cached(tree, cur,
cur + iosize - 1,
&cached, GFP_NOFS);
@@ -3000,7 +3173,12 @@ static int __do_readpage(struct extent_io_tree *tree,
continue;
}
/* the get_extent function already copied into the page */
- if (PageUptodate(page)) {
+ if ((root->sectorsize == PAGE_SIZE
+ && PageUptodate(page))
+ || (root->sectorsize < PAGE_SIZE
+ && test_page_blks_state(page,
+ BLK_STATE_UPTODATE, cur,
+ cur_end, 1))) {
check_page_uptodate(page);
unlock_extent(tree, cur, cur + iosize - 1);
cur = cur + iosize;
@@ -3019,6 +3197,9 @@ static int __do_readpage(struct extent_io_tree *tree,
}
pnr -= page->index;
+
+ set_page_blks_state(page, 1 << BLK_STATE_IO, cur,
+ cur + iosize - 1);
ret = submit_extent_page(REQ_OP_READ, read_flags, tree, NULL,
page, sector, disk_io_size, pg_offset,
bdev, bio, pnr,
@@ -3031,6 +3212,8 @@ static int __do_readpage(struct extent_io_tree *tree,
*bio_flags = this_bio_flag;
} else {
SetPageError(page);
+ clear_page_blks_state(page, 1 << BLK_STATE_IO,
+ cur, cur + iosize - 1);
unlock_extent(tree, cur, cur + iosize - 1);
goto out;
}
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index 9aa22f9..e7a0462 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -55,11 +55,72 @@
#define PAGE_SET_ERROR (1 << 5)
/*
- * page->private values. Every page that is controlled by the extent
- * map has page->private set to one.
+ * page->private values for "sector size" == "page size" case. Every
+ * page that is controlled by the extent map has page->private set to
+ * one.
*/
#define EXTENT_PAGE_PRIVATE 1
+enum blk_state {
+ BLK_STATE_UPTODATE,
+ BLK_STATE_DIRTY,
+ BLK_STATE_IO,
+ BLK_NR_STATE,
+};
+
+/*
+ * The maximum number of blocks per page (i.e. 32) occurs when using 2k
+ * as the block size and having 64k as the page size.
+ */
+#define BLK_STATE_NR_LONGS DIV_ROUND_UP(BLK_NR_STATE * 32, BITS_PER_LONG)
+
+
+/*
+ * btrfs_page_private->io_lock plays the same role as BH_Uptodate_Lock
+ * (see end_buffer_async_read()) i.e. without the io_lock we may end up
+ * in the following situation,
+ *
+ * NOTE: Assume 64k page size and 4k block size. Also assume that the first 12
+ * blocks of the page are contiguous while the next 4 blocks are contiguous.
When
+ * reading the page we end up submitting two "logical address space" bios. So
+ * end_bio_extent_readpage function is invoked twice, once for each bio.
+ *
+ * |-------------------------+-------------------------+-------------|
+ * | Task A | Task B | Task C |
+ * |-------------------------+-------------------------+-------------|
+ * | end_bio_extent_readpage | | |
+ * | process block 0 | | |
+ * | - clear BLK_STATE_IO | | |
+ * | - page_read_complete | | |
+ * | process block 1 | | |
+ * | | | |
+ * | | | |
+ * | | end_bio_extent_readpage | |
+ * | | process block 0 | |
+ * | | - clear BLK_STATE_IO | |
+ * | | - page_read_complete | |
+ * | | process block 1 | |
+ * | | | |
+ * | process block 11 | process block 3 | |
+ * | - clear BLK_STATE_IO | - clear BLK_STATE_IO | |
+ * | - page_read_complete | - page_read_complete | |
+ * | - returns true | - returns true | |
+ * | - unlock_page() | | |
+ * | | | lock_page() |
+ * | | - unlock_page() | |
+ * |-------------------------+-------------------------+-------------|
+ *
+ * We end up incorrectly unlocking the page twice and "Task C" ends up
+ * working on an unlocked page. So private->io_lock makes sure that
+ * only one of the tasks gets "true" as the return value when
+ * page_io_complete() is invoked. As an optimization the patch gets the
+ * io_lock only when the last block of the bio_vec is being processed.
+ */
+struct btrfs_page_private {
+ spinlock_t io_lock;
+ unsigned long bstate[BLK_STATE_NR_LONGS];
+};
+
struct extent_state;
struct btrfs_root;
struct btrfs_io_bio;
@@ -356,8 +417,14 @@ int extent_readpages(struct extent_io_tree *tree,
get_extent_t get_extent);
int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
__u64 start, __u64 len, get_extent_t *get_extent);
-void set_page_extent_mapped(struct page *page);
-
+int set_page_extent_mapped(struct page *page);
+int clear_page_extent_mapped(struct page *page);
+int set_page_blks_state(struct page *page, unsigned long blk_states,
+ u64 start, u64 end);
+int clear_page_blks_state(struct page *page, unsigned long blk_states,
+ u64 start, u64 end);
+int test_page_blks_state(struct page *page, enum blk_state blk_state,
+ u64 start, u64 end, int check_all);
struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
u64 start);
struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_eb_info *eb_info,
@@ -477,6 +544,7 @@ struct io_failure_record {
int in_validation;
};
+void check_page_uptodate(struct page *page);
void btrfs_free_io_failure_record(struct inode *inode, u64 start, u64 end);
int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
struct io_failure_record **failrec_ret);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index ac4a7c0..10dcb44 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -6983,7 +6983,10 @@ next:
btrfs_mark_buffer_dirty(leaf);
}
- SetPageUptodate(page);
+ set_page_blks_state(page, 1 << BLK_STATE_UPTODATE,
+ em->start, extent_map_end(em) - 1);
+ check_page_uptodate(page);
+
goto insert;
}
not_found:
@@ -8802,11 +8805,9 @@ static int __btrfs_releasepage(struct page *page, gfp_t
gfp_flags)
tree = &BTRFS_I(page->mapping->host)->io_tree;
map = &BTRFS_I(page->mapping->host)->extent_tree;
ret = try_release_extent_mapping(map, tree, page, gfp_flags);
- if (ret == 1) {
- ClearPagePrivate(page);
- set_page_private(page, 0);
- put_page(page);
- }
+ if (ret == 1)
+ clear_page_extent_mapped(page);
+
return ret;
}
--
2.5.5
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
