Miao Xie
2013-Jul-25 11:22 UTC
[PATCH V2 0/6] reduce the lock contention when reading pages
Before applying this patchset, we had to do extent state operation by the block size when reading pages because we need cache the checksum value of each block into the extent state, it was inefficient and increased the lock contenttion of extent state. Besides that, the acquisition of the extent map during the page read was also inefficient, we searched it every time we dealt with a page. In fact, if we don''t cache the checksum into the extent state, we can batch those extent state operation to reduce the time of the operations and the lock contention. Patch 3-5 do this work. Besides that, the acquisition of the extent map during the page read was also inefficient, we searched it every time we dealt with a page. In fact, it is very likely that the extent map objects are the same one. So we can cache the extent map object to reduce its search time and the lock contention of the extent map tree. Patch 6 does this work. I tested this patchset by sysbench on my SSD, it makes the performance of the random read be faster by ~7%. The lock contention is reduced by ~97%. Miao Xie (6): Btrfs: remove unnecessary argument of bio_readpage_error() Btrfs: add branch prediction hints in the read page end IO function Btrfs: don''t cache the csum value into the extent state tree Btrfs: batch the extent state operation in the end io handle of the read page Btrfs: batch the extent state operation when reading pages Btrfs: cache the extent map struct when reading several pages fs/btrfs/btrfs_inode.h | 21 +++ fs/btrfs/ctree.h | 4 +- fs/btrfs/disk-io.c | 5 +- fs/btrfs/extent_io.c | 392 ++++++++++++++++++++++++++++++------------------- fs/btrfs/extent_io.h | 10 +- fs/btrfs/file-item.c | 81 ++++++---- fs/btrfs/inode.c | 85 +++++------ fs/btrfs/volumes.h | 7 + 8 files changed, 370 insertions(+), 235 deletions(-) -- 1.8.1.4 -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Miao Xie
2013-Jul-25 11:22 UTC
[PATCH V2 1/6] Btrfs: remove unnecessary argument of bio_readpage_error()
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
Changelog v1 -> v2:
- None
---
fs/btrfs/extent_io.c | 26 +++++++++++---------------
1 file changed, 11 insertions(+), 15 deletions(-)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index f8586a9..a722235 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2202,8 +2202,7 @@ out:
*/
static int bio_readpage_error(struct bio *failed_bio, struct page *page,
- u64 start, u64 end, int failed_mirror,
- struct extent_state *state)
+ u64 start, u64 end, int failed_mirror)
{
struct io_failure_record *failrec = NULL;
u64 private;
@@ -2212,6 +2211,7 @@ static int bio_readpage_error(struct bio *failed_bio,
struct page *page,
struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+ struct extent_state *state;
struct bio *bio;
int num_copies;
int ret;
@@ -2296,22 +2296,18 @@ static int bio_readpage_error(struct bio *failed_bio,
struct page *page,
* all the retry and error correction code that follows. no
* matter what the error is, it is very likely to persist.
*/
- pr_debug("bio_readpage_error: cannot repair, num_copies == 1. "
- "state=%p, num_copies=%d, next_mirror %d, "
- "failed_mirror %d\n", state, num_copies,
- failrec->this_mirror, failed_mirror);
+ pr_debug("bio_readpage_error: cannot repair, num_copies=%d, next_mirror
%d, failed_mirror %d\n",
+ num_copies, failrec->this_mirror, failed_mirror);
free_io_failure(inode, failrec, 0);
return -EIO;
}
- if (!state) {
- spin_lock(&tree->lock);
- state = find_first_extent_bit_state(tree, failrec->start,
- EXTENT_LOCKED);
- if (state && state->start != failrec->start)
- state = NULL;
- spin_unlock(&tree->lock);
- }
+ spin_lock(&tree->lock);
+ state = find_first_extent_bit_state(tree, failrec->start,
+ EXTENT_LOCKED);
+ if (state && state->start != failrec->start)
+ state = NULL;
+ spin_unlock(&tree->lock);
/*
* there are two premises:
@@ -2541,7 +2537,7 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
* can''t handle the error it will return -EIO and we
* remain responsible for that page.
*/
- ret = bio_readpage_error(bio, page, start, end, mirror, NULL);
+ ret = bio_readpage_error(bio, page, start, end, mirror);
if (ret == 0) {
uptodate test_bit(BIO_UPTODATE, &bio->bi_flags);
--
1.8.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Miao Xie
2013-Jul-25 11:22 UTC
[PATCH V2 2/6] Btrfs: add branch prediction hints in the read page end IO function
This patch add some branch prediction hints into the end IO function
of the read page, it reduced the percentage of the branch misses from
5.5% to 4.9%.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
Changelog v1 -> v2:
- Just change the description in the changelog
---
fs/btrfs/extent_io.c | 14 +++++++++-----
1 file changed, 9 insertions(+), 5 deletions(-)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index a722235..6b6ba1a 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2502,7 +2502,7 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
spin_lock(&tree->lock);
state = find_first_extent_bit_state(tree, start, EXTENT_LOCKED);
- if (state && state->start == start) {
+ if (likely(state && state->start == start)) {
/*
* take a reference on the state, unlock will drop
* the ref
@@ -2512,7 +2512,8 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
spin_unlock(&tree->lock);
mirror = io_bio->mirror_num;
- if (uptodate && tree->ops &&
tree->ops->readpage_end_io_hook) {
+ if (likely(uptodate && tree->ops &&
+ tree->ops->readpage_end_io_hook)) {
ret = tree->ops->readpage_end_io_hook(page, start, end,
state, mirror);
if (ret)
@@ -2521,12 +2522,15 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
clean_io_failure(start, page);
}
- if (!uptodate && tree->ops &&
tree->ops->readpage_io_failed_hook) {
+ 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 && !err &&
test_bit(BIO_UPTODATE, &bio->bi_flags))
uptodate = 1;
- } else if (!uptodate) {
+ } else {
/*
* The generic bio_readpage_error handles errors the
* following way: If possible, new read requests are
@@ -2547,7 +2551,7 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
continue;
}
}
-
+readpage_ok:
if (uptodate && tree->track_uptodate) {
set_extent_uptodate(tree, start, end, &cached,
GFP_ATOMIC);
--
1.8.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Miao Xie
2013-Jul-25 11:22 UTC
[PATCH V2 3/6] Btrfs: don''t cache the csum value into the extent state tree
Before applying this patch, we cached the csum value into the extent state
tree when reading some data from the disk, this operation increased the lock
contention of the state tree.
Now, we just store the csum value into the bio structure or other unshared
structure, so we can reduce the lock contention.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
Changelog v1 -> v2:
- Fix uninitialized function pointer in the btrfs_io_bio
---
fs/btrfs/btrfs_inode.h | 21 +++++++++
fs/btrfs/ctree.h | 4 +-
fs/btrfs/disk-io.c | 5 +-
fs/btrfs/extent_io.c | 125 +++++++++++++++++++------------------------------
fs/btrfs/extent_io.h | 10 ++--
fs/btrfs/file-item.c | 81 +++++++++++++++++++++-----------
fs/btrfs/inode.c | 85 ++++++++++++++-------------------
fs/btrfs/volumes.h | 7 +++
8 files changed, 174 insertions(+), 164 deletions(-)
diff --git a/fs/btrfs/btrfs_inode.h b/fs/btrfs/btrfs_inode.h
index 08b286b..d0ae226 100644
--- a/fs/btrfs/btrfs_inode.h
+++ b/fs/btrfs/btrfs_inode.h
@@ -218,6 +218,27 @@ static inline int btrfs_inode_in_log(struct inode *inode,
u64 generation)
return 0;
}
+struct btrfs_dio_private {
+ struct inode *inode;
+ u64 logical_offset;
+ u64 disk_bytenr;
+ u64 bytes;
+ void *private;
+
+ /* number of bios pending for this dio */
+ atomic_t pending_bios;
+
+ /* IO errors */
+ int errors;
+
+ /* orig_bio is our btrfs_io_bio */
+ struct bio *orig_bio;
+
+ /* dio_bio came from fs/direct-io.c */
+ struct bio *dio_bio;
+ u8 csum[0];
+};
+
/*
* Disable DIO read nolock optimization, so new dio readers will be forced
* to grab i_mutex. It is used to avoid the endless truncate due to
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index e795bf1..80a1b5c 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -3509,12 +3509,14 @@ int btrfs_find_name_in_ext_backref(struct btrfs_path
*path,
struct btrfs_inode_extref **extref_ret);
/* file-item.c */
+struct btrfs_dio_private;
int btrfs_del_csums(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytenr, u64 len);
int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
struct bio *bio, u32 *dst);
int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode,
- struct bio *bio, u64 logical_offset);
+ struct btrfs_dio_private *dip, struct bio *bio,
+ u64 logical_offset);
int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 objectid, u64 pos,
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3c2886c..7f452cd 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -576,8 +576,9 @@ static noinline int check_leaf(struct btrfs_root *root,
return 0;
}
-static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
- struct extent_state *state, int mirror)
+static int btree_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
+ u64 phy_offset, struct page *page,
+ u64 start, u64 end, int mirror)
{
struct extent_io_tree *tree;
u64 found_start;
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 6b6ba1a..8eccbbc 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -1837,64 +1837,6 @@ out:
return ret;
}
-void extent_cache_csums_dio(struct extent_io_tree *tree, u64 start, u32
csums[],
- int count)
-{
- struct rb_node *node;
- struct extent_state *state;
-
- spin_lock(&tree->lock);
- /*
- * this search will find all the extents that end after
- * our range starts.
- */
- node = tree_search(tree, start);
- BUG_ON(!node);
-
- state = rb_entry(node, struct extent_state, rb_node);
- BUG_ON(state->start != start);
-
- while (count) {
- state->private = *csums++;
- count--;
- state = next_state(state);
- }
- spin_unlock(&tree->lock);
-}
-
-static inline u64 __btrfs_get_bio_offset(struct bio *bio, int bio_index)
-{
- struct bio_vec *bvec = bio->bi_io_vec + bio_index;
-
- return page_offset(bvec->bv_page) + bvec->bv_offset;
-}
-
-void extent_cache_csums(struct extent_io_tree *tree, struct bio *bio, int
bio_index,
- u32 csums[], int count)
-{
- struct rb_node *node;
- struct extent_state *state = NULL;
- u64 start;
-
- spin_lock(&tree->lock);
- do {
- start = __btrfs_get_bio_offset(bio, bio_index);
- if (state == NULL || state->start != start) {
- node = tree_search(tree, start);
- BUG_ON(!node);
-
- state = rb_entry(node, struct extent_state, rb_node);
- BUG_ON(state->start != start);
- }
- state->private = *csums++;
- count--;
- bio_index++;
-
- state = next_state(state);
- } while (count);
- spin_unlock(&tree->lock);
-}
-
int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private)
{
struct rb_node *node;
@@ -2201,8 +2143,9 @@ out:
* needed
*/
-static int bio_readpage_error(struct bio *failed_bio, struct page *page,
- u64 start, u64 end, int failed_mirror)
+static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
+ struct page *page, u64 start, u64 end,
+ int failed_mirror)
{
struct io_failure_record *failrec = NULL;
u64 private;
@@ -2211,8 +2154,9 @@ static int bio_readpage_error(struct bio *failed_bio,
struct page *page,
struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
- struct extent_state *state;
struct bio *bio;
+ struct btrfs_io_bio *btrfs_failed_bio;
+ struct btrfs_io_bio *btrfs_bio;
int num_copies;
int ret;
int read_mode;
@@ -2302,13 +2246,6 @@ static int bio_readpage_error(struct bio *failed_bio,
struct page *page,
return -EIO;
}
- spin_lock(&tree->lock);
- state = find_first_extent_bit_state(tree, failrec->start,
- EXTENT_LOCKED);
- if (state && state->start != failrec->start)
- state = NULL;
- spin_unlock(&tree->lock);
-
/*
* there are two premises:
* a) deliver good data to the caller
@@ -2345,9 +2282,8 @@ static int bio_readpage_error(struct bio *failed_bio,
struct page *page,
read_mode = READ_SYNC;
}
- if (!state || failrec->this_mirror > num_copies) {
- pr_debug("bio_readpage_error: (fail) state=%p, num_copies=%d, "
- "next_mirror %d, failed_mirror %d\n", state,
+ if (failrec->this_mirror > num_copies) {
+ pr_debug("bio_readpage_error: (fail) num_copies=%d, next_mirror %d,
failed_mirror %d\n",
num_copies, failrec->this_mirror, failed_mirror);
free_io_failure(inode, failrec, 0);
return -EIO;
@@ -2358,12 +2294,24 @@ static int bio_readpage_error(struct bio *failed_bio,
struct page *page,
free_io_failure(inode, failrec, 0);
return -EIO;
}
- bio->bi_private = state;
bio->bi_end_io = failed_bio->bi_end_io;
bio->bi_sector = failrec->logical >> 9;
bio->bi_bdev =
BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
bio->bi_size = 0;
+ btrfs_failed_bio = btrfs_io_bio(failed_bio);
+ if (btrfs_failed_bio->csum) {
+ struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
+ u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
+
+ btrfs_bio = btrfs_io_bio(bio);
+ btrfs_bio->csum = btrfs_bio->csum_inline;
+ phy_offset >>= inode->i_sb->s_blocksize_bits;
+ phy_offset *= csum_size;
+ memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + phy_offset,
+ csum_size);
+ }
+
bio_add_page(bio, page, failrec->len, start - page_offset(page));
pr_debug("bio_readpage_error: submitting new read[%#x] to "
@@ -2462,9 +2410,12 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
struct bio_vec *bvec = bio->bi_io_vec;
+ struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
struct extent_io_tree *tree;
+ u64 offset = 0;
u64 start;
u64 end;
+ u64 len;
int mirror;
int ret;
@@ -2475,7 +2426,6 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
struct page *page = bvec->bv_page;
struct extent_state *cached = NULL;
struct extent_state *state;
- struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
struct inode *inode = page->mapping->host;
pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
@@ -2496,6 +2446,7 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
start = page_offset(page);
end = start + bvec->bv_offset + bvec->bv_len - 1;
+ len = bvec->bv_len;
if (++bvec <= bvec_end)
prefetchw(&bvec->bv_page->flags);
@@ -2514,8 +2465,9 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
mirror = io_bio->mirror_num;
if (likely(uptodate && tree->ops &&
tree->ops->readpage_end_io_hook)) {
- ret = tree->ops->readpage_end_io_hook(page, start, end,
- state, mirror);
+ ret = tree->ops->readpage_end_io_hook(io_bio, offset,
+ page, start, end,
+ mirror);
if (ret)
uptodate = 0;
else
@@ -2541,7 +2493,8 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
* can''t handle the error it will return -EIO and we
* remain responsible for that page.
*/
- ret = bio_readpage_error(bio, page, start, end, mirror);
+ ret = bio_readpage_error(bio, offset, page, start, end,
+ mirror);
if (ret == 0) {
uptodate test_bit(BIO_UPTODATE, &bio->bi_flags);
@@ -2573,8 +2526,11 @@ readpage_ok:
SetPageError(page);
}
unlock_page(page);
+ offset += len;
} while (bvec <= bvec_end);
+ if (io_bio->end_io)
+ io_bio->end_io(io_bio, err);
bio_put(bio);
}
@@ -2586,6 +2542,7 @@ struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
gfp_t gfp_flags)
{
+ struct btrfs_io_bio *btrfs_bio;
struct bio *bio;
bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
@@ -2601,6 +2558,10 @@ btrfs_bio_alloc(struct block_device *bdev, u64
first_sector, int nr_vecs,
bio->bi_size = 0;
bio->bi_bdev = bdev;
bio->bi_sector = first_sector;
+ btrfs_bio = btrfs_io_bio(bio);
+ btrfs_bio->csum = NULL;
+ btrfs_bio->csum_allocated = NULL;
+ btrfs_bio->end_io = NULL;
}
return bio;
}
@@ -2614,7 +2575,17 @@ struct bio *btrfs_bio_clone(struct bio *bio, gfp_t
gfp_mask)
/* this also allocates from the btrfs_bioset */
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
{
- return bio_alloc_bioset(gfp_mask, nr_iovecs, btrfs_bioset);
+ struct btrfs_io_bio *btrfs_bio;
+ struct bio *bio;
+
+ bio = bio_alloc_bioset(gfp_mask, nr_iovecs, btrfs_bioset);
+ if (bio) {
+ btrfs_bio = btrfs_io_bio(bio);
+ btrfs_bio->csum = NULL;
+ btrfs_bio->csum_allocated = NULL;
+ btrfs_bio->end_io = NULL;
+ }
+ return bio;
}
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index 3b8c4e2..f7544af 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -62,6 +62,7 @@
struct extent_state;
struct btrfs_root;
+struct btrfs_io_bio;
typedef int (extent_submit_bio_hook_t)(struct inode *inode, int rw,
struct bio *bio, int mirror_num,
@@ -77,8 +78,9 @@ struct extent_io_ops {
size_t size, struct bio *bio,
unsigned long bio_flags);
int (*readpage_io_failed_hook)(struct page *page, int failed_mirror);
- int (*readpage_end_io_hook)(struct page *page, u64 start, u64 end,
- struct extent_state *state, int mirror);
+ int (*readpage_end_io_hook)(struct btrfs_io_bio *io_bio, u64 phy_offset,
+ struct page *page, u64 start, u64 end,
+ int mirror);
int (*writepage_end_io_hook)(struct page *page, u64 start, u64 end,
struct extent_state *state, int uptodate);
void (*set_bit_hook)(struct inode *inode, struct extent_state *state,
@@ -262,10 +264,6 @@ int extent_readpages(struct extent_io_tree *tree,
int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
__u64 start, __u64 len, get_extent_t *get_extent);
int set_state_private(struct extent_io_tree *tree, u64 start, u64 private);
-void extent_cache_csums_dio(struct extent_io_tree *tree, u64 start, u32
csums[],
- int count);
-void extent_cache_csums(struct extent_io_tree *tree, struct bio *bio,
- int bvec_index, u32 csums[], int count);
int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private);
void set_page_extent_mapped(struct page *page);
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index a7bfc95..f87d09a 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -23,6 +23,7 @@
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
+#include "volumes.h"
#include "print-tree.h"
#define __MAX_CSUM_ITEMS(r, size) ((unsigned long)(((BTRFS_LEAF_DATA_SIZE(r) -
\
@@ -152,28 +153,54 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle
*trans,
return ret;
}
+static void btrfs_io_bio_endio_readpage(struct btrfs_io_bio *bio, int err)
+{
+ kfree(bio->csum_allocated);
+}
+
static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
struct inode *inode, struct bio *bio,
u64 logical_offset, u32 *dst, int dio)
{
- u32 sum[16];
- int len;
struct bio_vec *bvec = bio->bi_io_vec;
- int bio_index = 0;
+ struct btrfs_io_bio *btrfs_bio = btrfs_io_bio(bio);
+ struct btrfs_csum_item *item = NULL;
+ struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+ struct btrfs_path *path;
+ u8 *csum;
u64 offset = 0;
u64 item_start_offset = 0;
u64 item_last_offset = 0;
u64 disk_bytenr;
u32 diff;
- u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
+ int nblocks;
+ int bio_index = 0;
int count;
- struct btrfs_path *path;
- struct btrfs_csum_item *item = NULL;
- struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
+ u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
+
+ nblocks = bio->bi_size >> inode->i_sb->s_blocksize_bits;
+ if (!dst) {
+ if (nblocks * csum_size > BTRFS_BIO_INLINE_CSUM_SIZE) {
+ btrfs_bio->csum_allocated = kmalloc(nblocks * csum_size,
+ GFP_NOFS);
+ if (!btrfs_bio->csum_allocated) {
+ btrfs_free_path(path);
+ return -ENOMEM;
+ }
+ btrfs_bio->csum = btrfs_bio->csum_allocated;
+ btrfs_bio->end_io = btrfs_io_bio_endio_readpage;
+ } else {
+ btrfs_bio->csum = btrfs_bio->csum_inline;
+ }
+ csum = btrfs_bio->csum;
+ } else {
+ csum = (u8 *)dst;
+ }
+
if (bio->bi_size > PAGE_CACHE_SIZE * 8)
path->reada = 2;
@@ -194,11 +221,10 @@ static int __btrfs_lookup_bio_sums(struct btrfs_root
*root,
if (dio)
offset = logical_offset;
while (bio_index < bio->bi_vcnt) {
- len = min_t(int, ARRAY_SIZE(sum), bio->bi_vcnt - bio_index);
if (!dio)
offset = page_offset(bvec->bv_page) + bvec->bv_offset;
- count = btrfs_find_ordered_sum(inode, offset, disk_bytenr, sum,
- len);
+ count = btrfs_find_ordered_sum(inode, offset, disk_bytenr,
+ (u32 *)csum, nblocks);
if (count)
goto found;
@@ -213,7 +239,7 @@ static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
path, disk_bytenr, 0);
if (IS_ERR(item)) {
count = 1;
- sum[0] = 0;
+ memset(csum, 0, csum_size);
if (BTRFS_I(inode)->root->root_key.objectid =
BTRFS_DATA_RELOC_TREE_OBJECTID) {
set_extent_bits(io_tree, offset,
@@ -249,23 +275,14 @@ static int __btrfs_lookup_bio_sums(struct btrfs_root
*root,
diff = disk_bytenr - item_start_offset;
diff = diff / root->sectorsize;
diff = diff * csum_size;
- count = min_t(int, len, (item_last_offset - disk_bytenr) >>
- inode->i_sb->s_blocksize_bits);
- read_extent_buffer(path->nodes[0], sum,
+ count = min_t(int, nblocks, (item_last_offset - disk_bytenr) >>
+ inode->i_sb->s_blocksize_bits);
+ read_extent_buffer(path->nodes[0], csum,
((unsigned long)item) + diff,
csum_size * count);
found:
- if (dst) {
- memcpy(dst, sum, count * csum_size);
- dst += count;
- } else {
- if (dio)
- extent_cache_csums_dio(io_tree, offset, sum,
- count);
- else
- extent_cache_csums(io_tree, bio, bio_index, sum,
- count);
- }
+ csum += count * csum_size;
+ nblocks -= count;
while (count--) {
disk_bytenr += bvec->bv_len;
offset += bvec->bv_len;
@@ -284,9 +301,19 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct
inode *inode,
}
int btrfs_lookup_bio_sums_dio(struct btrfs_root *root, struct inode *inode,
- struct bio *bio, u64 offset)
+ struct btrfs_dio_private *dip, struct bio *bio,
+ u64 offset)
{
- return __btrfs_lookup_bio_sums(root, inode, bio, offset, NULL, 1);
+ int len = (bio->bi_sector << 9) - dip->disk_bytenr;
+ u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
+ int ret;
+
+ len >>= inode->i_sb->s_blocksize_bits;
+ len *= csum_size;
+
+ ret = __btrfs_lookup_bio_sums(root, inode, bio, offset,
+ (u32 *)(dip->csum + len), 1);
+ return ret;
}
int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 179f3ef..687fd9a 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -2818,16 +2818,16 @@ static int btrfs_writepage_end_io_hook(struct page
*page, u64 start, u64 end,
* if there''s a match, we allow the bio to finish. If not, the code
in
* extent_io.c will try to find good copies for us.
*/
-static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
- struct extent_state *state, int mirror)
+static int btrfs_readpage_end_io_hook(struct btrfs_io_bio *io_bio,
+ u64 phy_offset, struct page *page,
+ u64 start, u64 end, int mirror)
{
size_t offset = start - page_offset(page);
struct inode *inode = page->mapping->host;
struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
char *kaddr;
- u64 private = ~(u32)0;
- int ret;
struct btrfs_root *root = BTRFS_I(inode)->root;
+ u32 csum_expected;
u32 csum = ~(u32)0;
static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
DEFAULT_RATELIMIT_BURST);
@@ -2847,19 +2847,13 @@ static int btrfs_readpage_end_io_hook(struct page *page,
u64 start, u64 end,
return 0;
}
- if (state && state->start == start) {
- private = state->private;
- ret = 0;
- } else {
- ret = get_state_private(io_tree, start, &private);
- }
- kaddr = kmap_atomic(page);
- if (ret)
- goto zeroit;
+ phy_offset >>= inode->i_sb->s_blocksize_bits;
+ csum_expected = *(((u32 *)io_bio->csum) + phy_offset);
+ kaddr = kmap_atomic(page);
csum = btrfs_csum_data(kaddr + offset, csum, end - start + 1);
btrfs_csum_final(csum, (char *)&csum);
- if (csum != private)
+ if (csum != csum_expected)
goto zeroit;
kunmap_atomic(kaddr);
@@ -2868,14 +2862,13 @@ good:
zeroit:
if (__ratelimit(&_rs))
- btrfs_info(root->fs_info, "csum failed ino %llu off %llu csum %u
private %llu",
+ btrfs_info(root->fs_info, "csum failed ino %llu off %llu csum %u
expected csum %u",
(unsigned long long)btrfs_ino(page->mapping->host),
- (unsigned long long)start, csum,
- (unsigned long long)private);
+ (unsigned long long)start, csum, csum_expected);
memset(kaddr + offset, 1, end - start + 1);
flush_dcache_page(page);
kunmap_atomic(kaddr);
- if (private == 0)
+ if (csum_expected == 0)
return 0;
return -EIO;
}
@@ -6832,26 +6825,6 @@ unlock_err:
return ret;
}
-struct btrfs_dio_private {
- struct inode *inode;
- u64 logical_offset;
- u64 disk_bytenr;
- u64 bytes;
- void *private;
-
- /* number of bios pending for this dio */
- atomic_t pending_bios;
-
- /* IO errors */
- int errors;
-
- /* orig_bio is our btrfs_io_bio */
- struct bio *orig_bio;
-
- /* dio_bio came from fs/direct-io.c */
- struct bio *dio_bio;
-};
-
static void btrfs_endio_direct_read(struct bio *bio, int err)
{
struct btrfs_dio_private *dip = bio->bi_private;
@@ -6860,6 +6833,8 @@ static void btrfs_endio_direct_read(struct bio *bio, int
err)
struct inode *inode = dip->inode;
struct btrfs_root *root = BTRFS_I(inode)->root;
struct bio *dio_bio;
+ u32 *csums = (u32 *)dip->csum;
+ int index = 0;
u64 start;
start = dip->logical_offset;
@@ -6868,12 +6843,8 @@ static void btrfs_endio_direct_read(struct bio *bio, int
err)
struct page *page = bvec->bv_page;
char *kaddr;
u32 csum = ~(u32)0;
- u64 private = ~(u32)0;
unsigned long flags;
- if (get_state_private(&BTRFS_I(inode)->io_tree,
- start, &private))
- goto failed;
local_irq_save(flags);
kaddr = kmap_atomic(page);
csum = btrfs_csum_data(kaddr + bvec->bv_offset,
@@ -6883,18 +6854,18 @@ static void btrfs_endio_direct_read(struct bio *bio, int
err)
local_irq_restore(flags);
flush_dcache_page(bvec->bv_page);
- if (csum != private) {
-failed:
- btrfs_err(root->fs_info, "csum failed ino %llu off %llu csum %u
private %u",
- (unsigned long long)btrfs_ino(inode),
- (unsigned long long)start,
- csum, (unsigned)private);
+ if (csum != csums[index]) {
+ btrfs_err(root->fs_info, "csum failed ino %llu off %llu csum %u
expected csum %u",
+ (unsigned long long)btrfs_ino(inode),
+ (unsigned long long)start,
+ csum, csums[index]);
err = -EIO;
}
}
start += bvec->bv_len;
bvec++;
+ index++;
} while (bvec <= bvec_end);
unlock_extent(&BTRFS_I(inode)->io_tree, dip->logical_offset,
@@ -7011,6 +6982,7 @@ static inline int __btrfs_submit_dio_bio(struct bio *bio,
struct inode *inode,
int rw, u64 file_offset, int skip_sum,
int async_submit)
{
+ struct btrfs_dio_private *dip = bio->bi_private;
int write = rw & REQ_WRITE;
struct btrfs_root *root = BTRFS_I(inode)->root;
int ret;
@@ -7045,7 +7017,8 @@ static inline int __btrfs_submit_dio_bio(struct bio *bio,
struct inode *inode,
if (ret)
goto err;
} else if (!skip_sum) {
- ret = btrfs_lookup_bio_sums_dio(root, inode, bio, file_offset);
+ ret = btrfs_lookup_bio_sums_dio(root, inode, dip, bio,
+ file_offset);
if (ret)
goto err;
}
@@ -7080,6 +7053,7 @@ static int btrfs_submit_direct_hook(int rw, struct
btrfs_dio_private *dip,
bio_put(orig_bio);
return -EIO;
}
+
if (map_length >= orig_bio->bi_size) {
bio = orig_bio;
goto submit;
@@ -7175,19 +7149,28 @@ static void btrfs_submit_direct(int rw, struct bio
*dio_bio,
struct btrfs_dio_private *dip;
struct bio *io_bio;
int skip_sum;
+ int sum_len;
int write = rw & REQ_WRITE;
int ret = 0;
+ u16 csum_size;
skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
io_bio = btrfs_bio_clone(dio_bio, GFP_NOFS);
-
if (!io_bio) {
ret = -ENOMEM;
goto free_ordered;
}
- dip = kmalloc(sizeof(*dip), GFP_NOFS);
+ if (!skip_sum && !write) {
+ csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
+ sum_len = dio_bio->bi_size >> inode->i_sb->s_blocksize_bits;
+ sum_len *= csum_size;
+ } else {
+ sum_len = 0;
+ }
+
+ dip = kmalloc(sizeof(*dip) + sum_len, GFP_NOFS);
if (!dip) {
ret = -ENOMEM;
goto free_io_bio;
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 8670558..08c44d9 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -152,6 +152,8 @@ struct btrfs_fs_devices {
int rotating;
};
+#define BTRFS_BIO_INLINE_CSUM_SIZE 64
+
/*
* we need the mirror number and stripe index to be passed around
* the call chain while we are processing end_io (especially errors).
@@ -161,9 +163,14 @@ struct btrfs_fs_devices {
* we allocate are actually btrfs_io_bios. We''ll cram as much of
* struct btrfs_bio as we can into this over time.
*/
+typedef void (btrfs_io_bio_end_io_t) (struct btrfs_io_bio *bio, int err);
struct btrfs_io_bio {
unsigned long mirror_num;
unsigned long stripe_index;
+ u8 *csum;
+ u8 csum_inline[BTRFS_BIO_INLINE_CSUM_SIZE];
+ u8 *csum_allocated;
+ btrfs_io_bio_end_io_t *end_io;
struct bio bio;
};
--
1.8.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Miao Xie
2013-Jul-25 11:22 UTC
[PATCH V2 4/6] Btrfs: batch the extent state operation in the end io handle of the read page
Before applying this patch, we set the uptodate flag and unlock the extent
by the page size, it is unnecessary, we can do it in batches, it can reduce
the lock contention of the extent state tree.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
Changelog v1 -> v2:
- Fix wrong check in clean_io_failure()
---
fs/btrfs/extent_io.c | 73 ++++++++++++++++++++++++++++++----------------------
1 file changed, 42 insertions(+), 31 deletions(-)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 8eccbbc..38fb5dd 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -762,15 +762,6 @@ static void cache_state(struct extent_state *state,
}
}
-static void uncache_state(struct extent_state **cached_ptr)
-{
- if (cached_ptr && (*cached_ptr)) {
- struct extent_state *state = *cached_ptr;
- *cached_ptr = NULL;
- free_extent_state(state);
- }
-}
-
/*
* set some bits on a range in the tree. This may require allocations or
* sleeping, so the gfp mask is used to indicate what is allowed.
@@ -2115,7 +2106,8 @@ static int clean_io_failure(u64 start, struct page *page)
EXTENT_LOCKED);
spin_unlock(&BTRFS_I(inode)->io_tree.lock);
- if (state && state->start == failrec->start) {
+ if (state && state->start <= failrec->start &&
+ state->end >= failrec->start + failrec->len - 1) {
fs_info = BTRFS_I(inode)->root->fs_info;
num_copies = btrfs_num_copies(fs_info, failrec->logical,
failrec->len);
@@ -2394,6 +2386,18 @@ static void end_bio_extent_writepage(struct bio *bio, int
err)
bio_put(bio);
}
+static void
+endio_readpage_release_extent(struct extent_io_tree *tree, u64 start, u64 len,
+ int uptodate)
+{
+ struct extent_state *cached = NULL;
+ u64 end = start + len - 1;
+
+ if (uptodate && tree->track_uptodate)
+ set_extent_uptodate(tree, start, end, &cached, GFP_ATOMIC);
+ unlock_extent_cached(tree, start, end, &cached, GFP_ATOMIC);
+}
+
/*
* after a readpage IO is done, we need to:
* clear the uptodate bits on error
@@ -2416,6 +2420,8 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
u64 start;
u64 end;
u64 len;
+ u64 extent_start = 0;
+ u64 extent_len = 0;
int mirror;
int ret;
@@ -2424,8 +2430,6 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
do {
struct page *page = bvec->bv_page;
- struct extent_state *cached = NULL;
- struct extent_state *state;
struct inode *inode = page->mapping->host;
pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
@@ -2451,17 +2455,6 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
if (++bvec <= bvec_end)
prefetchw(&bvec->bv_page->flags);
- spin_lock(&tree->lock);
- state = find_first_extent_bit_state(tree, start, EXTENT_LOCKED);
- if (likely(state && state->start == start)) {
- /*
- * take a reference on the state, unlock will drop
- * the ref
- */
- cache_state(state, &cached);
- }
- spin_unlock(&tree->lock);
-
mirror = io_bio->mirror_num;
if (likely(uptodate && tree->ops &&
tree->ops->readpage_end_io_hook)) {
@@ -2500,18 +2493,11 @@ static void end_bio_extent_readpage(struct bio *bio, int
err)
test_bit(BIO_UPTODATE, &bio->bi_flags);
if (err)
uptodate = 0;
- uncache_state(&cached);
continue;
}
}
readpage_ok:
- if (uptodate && tree->track_uptodate) {
- set_extent_uptodate(tree, start, end, &cached,
- GFP_ATOMIC);
- }
- unlock_extent_cached(tree, start, end, &cached, GFP_ATOMIC);
-
- if (uptodate) {
+ if (likely(uptodate)) {
loff_t i_size = i_size_read(inode);
pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
unsigned offset;
@@ -2527,8 +2513,33 @@ readpage_ok:
}
unlock_page(page);
offset += len;
+
+ if (unlikely(!uptodate)) {
+ if (extent_len) {
+ endio_readpage_release_extent(tree,
+ extent_start,
+ extent_len, 1);
+ extent_start = 0;
+ extent_len = 0;
+ }
+ endio_readpage_release_extent(tree, start,
+ end - start + 1, 0);
+ } else if (!extent_len) {
+ extent_start = start;
+ extent_len = end + 1 - start;
+ } else if (extent_start + extent_len == start) {
+ extent_len += end + 1 - start;
+ } else {
+ endio_readpage_release_extent(tree, extent_start,
+ extent_len, uptodate);
+ extent_start = start;
+ extent_len = end + 1 - start;
+ }
} while (bvec <= bvec_end);
+ if (extent_len)
+ endio_readpage_release_extent(tree, extent_start, extent_len,
+ uptodate);
if (io_bio->end_io)
io_bio->end_io(io_bio, err);
bio_put(bio);
--
1.8.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Miao Xie
2013-Jul-25 11:22 UTC
[PATCH V2 5/6] Btrfs: batch the extent state operation when reading pages
In the past, we cached the checksum value in the extent state object, so we
had to split the extent state object by the block size, or we had no space
to keep this checksum value. But it increased the lock contention of the
extent state tree.
Now we removed this limit by caching the checksum into the bio object, so
it is unnecessary to do the extent state operations by the block size, we
can do it in batches, in this way, we can reduce the lock contention of
the extent state tree.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
Changelog v1 -> v2:
- New patch
---
fs/btrfs/extent_io.c | 135 ++++++++++++++++++++++++++++++++++++++++-----------
1 file changed, 107 insertions(+), 28 deletions(-)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 38fb5dd..53a1828 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2727,11 +2727,11 @@ void set_page_extent_mapped(struct page *page)
* handlers)
* XXX JDM: This needs looking at to ensure proper page locking
*/
-static int __extent_read_full_page(struct extent_io_tree *tree,
- struct page *page,
- get_extent_t *get_extent,
- struct bio **bio, int mirror_num,
- unsigned long *bio_flags, int rw)
+static int __do_readpage(struct extent_io_tree *tree,
+ struct page *page,
+ get_extent_t *get_extent,
+ struct bio **bio, int mirror_num,
+ unsigned long *bio_flags, int rw)
{
struct inode *inode = page->mapping->host;
u64 start = page_offset(page);
@@ -2745,7 +2745,6 @@ static int __extent_read_full_page(struct extent_io_tree
*tree,
sector_t sector;
struct extent_map *em;
struct block_device *bdev;
- struct btrfs_ordered_extent *ordered;
int ret;
int nr = 0;
size_t pg_offset = 0;
@@ -2756,24 +2755,15 @@ static int __extent_read_full_page(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;
}
}
- end = page_end;
- while (1) {
- lock_extent(tree, start, end);
- ordered = btrfs_lookup_ordered_extent(inode, start);
- if (!ordered)
- break;
- unlock_extent(tree, start, end);
- btrfs_start_ordered_extent(inode, ordered, 1);
- btrfs_put_ordered_extent(ordered);
- }
-
if (page->index == last_byte >> PAGE_CACHE_SHIFT) {
char *userpage;
size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1);
@@ -2902,6 +2892,101 @@ out:
return 0;
}
+static inline void __do_contiguous_readpages(struct extent_io_tree *tree,
+ struct page *pages[], int nr_pages,
+ u64 start, u64 end,
+ get_extent_t *get_extent,
+ struct bio **bio, int mirror_num,
+ unsigned long *bio_flags, int rw)
+{
+ struct inode *inode;
+ struct btrfs_ordered_extent *ordered;
+ int index;
+
+ inode = pages[0]->mapping->host;
+ while (1) {
+ lock_extent(tree, start, end);
+ ordered = btrfs_lookup_ordered_range(inode, start,
+ end - start + 1);
+ if (!ordered)
+ break;
+ unlock_extent(tree, start, end);
+ btrfs_start_ordered_extent(inode, ordered, 1);
+ btrfs_put_ordered_extent(ordered);
+ }
+
+ for (index = 0; index < nr_pages; index++) {
+ __do_readpage(tree, pages[index], get_extent, bio, mirror_num,
+ bio_flags, rw);
+ page_cache_release(pages[index]);
+ }
+}
+
+static void __extent_readpages(struct extent_io_tree *tree,
+ struct page *pages[],
+ int nr_pages, get_extent_t *get_extent,
+ struct bio **bio, int mirror_num,
+ unsigned long *bio_flags, int rw)
+{
+ u64 start;
+ u64 end = 0;
+ u64 page_start;
+ int index;
+ int first_index;
+
+ for (index = 0; index < nr_pages; index++) {
+ page_start = page_offset(pages[index]);
+ if (!end) {
+ start = page_start;
+ end = start + PAGE_CACHE_SIZE - 1;
+ first_index = index;
+ } else if (end + 1 == page_start) {
+ end += PAGE_CACHE_SIZE;
+ } else {
+ __do_contiguous_readpages(tree, &pages[first_index],
+ index - first_index, start,
+ end, get_extent, bio,
+ mirror_num, bio_flags, rw);
+ start = page_start;
+ end = start + PAGE_CACHE_SIZE - 1;
+ first_index = index;
+ }
+ }
+
+ if (end)
+ __do_contiguous_readpages(tree, &pages[first_index],
+ index - first_index, start,
+ end, get_extent, bio,
+ mirror_num, bio_flags, rw);
+}
+
+static int __extent_read_full_page(struct extent_io_tree *tree,
+ struct page *page,
+ get_extent_t *get_extent,
+ struct bio **bio, int mirror_num,
+ unsigned long *bio_flags, int rw)
+{
+ struct inode *inode = page->mapping->host;
+ struct btrfs_ordered_extent *ordered;
+ u64 start = page_offset(page);
+ u64 end = start + PAGE_CACHE_SIZE - 1;
+ int ret;
+
+ while (1) {
+ lock_extent(tree, start, end);
+ ordered = btrfs_lookup_ordered_extent(inode, start);
+ if (!ordered)
+ break;
+ unlock_extent(tree, start, end);
+ btrfs_start_ordered_extent(inode, ordered, 1);
+ btrfs_put_ordered_extent(ordered);
+ }
+
+ ret = __do_readpage(tree, page, get_extent, bio, mirror_num, bio_flags,
+ rw);
+ return ret;
+}
+
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
get_extent_t *get_extent, int mirror_num)
{
@@ -3752,7 +3837,6 @@ int extent_readpages(struct extent_io_tree *tree,
unsigned long bio_flags = 0;
struct page *pagepool[16];
struct page *page;
- int i = 0;
int nr = 0;
for (page_idx = 0; page_idx < nr_pages; page_idx++) {
@@ -3769,18 +3853,13 @@ int extent_readpages(struct extent_io_tree *tree,
pagepool[nr++] = page;
if (nr < ARRAY_SIZE(pagepool))
continue;
- for (i = 0; i < nr; i++) {
- __extent_read_full_page(tree, pagepool[i], get_extent,
- &bio, 0, &bio_flags, READ);
- page_cache_release(pagepool[i]);
- }
+ __extent_readpages(tree, pagepool, nr, get_extent,
+ &bio, 0, &bio_flags, READ);
nr = 0;
}
- for (i = 0; i < nr; i++) {
- __extent_read_full_page(tree, pagepool[i], get_extent,
- &bio, 0, &bio_flags, READ);
- page_cache_release(pagepool[i]);
- }
+ if (nr)
+ __extent_readpages(tree, pagepool, nr, get_extent,
+ &bio, 0, &bio_flags, READ);
BUG_ON(!list_empty(pages));
if (bio)
--
1.8.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Miao Xie
2013-Jul-25 11:22 UTC
[PATCH V2 6/6] Btrfs: cache the extent map struct when reading several pages
When we read several pages at once, we needn''t get the extent map
object
every time we deal with a page, and we can cache the extent map object.
So, we can reduce the search time of the extent map, and besides that, we
also can reduce the lock contention of the extent map tree.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
Changelog v1 -> v2:
- New patch
---
fs/btrfs/extent_io.c | 57 ++++++++++++++++++++++++++++++++++++++++++----------
1 file changed, 46 insertions(+), 11 deletions(-)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 53a1828..2e6dfeb 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2721,6 +2721,33 @@ void set_page_extent_mapped(struct page *page)
}
}
+static struct extent_map *
+__get_extent_map(struct inode *inode, struct page *page, size_t pg_offset,
+ u64 start, u64 len, get_extent_t *get_extent,
+ struct extent_map **em_cached)
+{
+ struct extent_map *em;
+
+ if (em_cached && *em_cached) {
+ em = *em_cached;
+ if (em->in_tree && start >= em->start &&
+ start < extent_map_end(em)) {
+ atomic_inc(&em->refs);
+ return em;
+ }
+
+ free_extent_map(em);
+ *em_cached = NULL;
+ }
+
+ em = get_extent(inode, page, pg_offset, start, len, 0);
+ if (em_cached && !IS_ERR_OR_NULL(em)) {
+ BUG_ON(*em_cached);
+ atomic_inc(&em->refs);
+ *em_cached = em;
+ }
+ return em;
+}
/*
* basic readpage implementation. Locked extent state structs are inserted
* into the tree that are removed when the IO is done (by the end_io
@@ -2730,6 +2757,7 @@ void set_page_extent_mapped(struct page *page)
static int __do_readpage(struct extent_io_tree *tree,
struct page *page,
get_extent_t *get_extent,
+ struct extent_map **em_cached,
struct bio **bio, int mirror_num,
unsigned long *bio_flags, int rw)
{
@@ -2794,8 +2822,8 @@ static int __do_readpage(struct extent_io_tree *tree,
&cached, GFP_NOFS);
break;
}
- em = get_extent(inode, page, pg_offset, cur,
- end - cur + 1, 0);
+ em = __get_extent_map(inode, page, pg_offset, cur,
+ end - cur + 1, get_extent, em_cached);
if (IS_ERR_OR_NULL(em)) {
SetPageError(page);
unlock_extent(tree, cur, end);
@@ -2896,6 +2924,7 @@ static inline void __do_contiguous_readpages(struct
extent_io_tree *tree,
struct page *pages[], int nr_pages,
u64 start, u64 end,
get_extent_t *get_extent,
+ struct extent_map **em_cached,
struct bio **bio, int mirror_num,
unsigned long *bio_flags, int rw)
{
@@ -2916,8 +2945,8 @@ static inline void __do_contiguous_readpages(struct
extent_io_tree *tree,
}
for (index = 0; index < nr_pages; index++) {
- __do_readpage(tree, pages[index], get_extent, bio, mirror_num,
- bio_flags, rw);
+ __do_readpage(tree, pages[index], get_extent, em_cached, bio,
+ mirror_num, bio_flags, rw);
page_cache_release(pages[index]);
}
}
@@ -2925,6 +2954,7 @@ static inline void __do_contiguous_readpages(struct
extent_io_tree *tree,
static void __extent_readpages(struct extent_io_tree *tree,
struct page *pages[],
int nr_pages, get_extent_t *get_extent,
+ struct extent_map **em_cached,
struct bio **bio, int mirror_num,
unsigned long *bio_flags, int rw)
{
@@ -2945,8 +2975,9 @@ static void __extent_readpages(struct extent_io_tree
*tree,
} else {
__do_contiguous_readpages(tree, &pages[first_index],
index - first_index, start,
- end, get_extent, bio,
- mirror_num, bio_flags, rw);
+ end, get_extent, em_cached,
+ bio, mirror_num, bio_flags,
+ rw);
start = page_start;
end = start + PAGE_CACHE_SIZE - 1;
first_index = index;
@@ -2956,7 +2987,7 @@ static void __extent_readpages(struct extent_io_tree
*tree,
if (end)
__do_contiguous_readpages(tree, &pages[first_index],
index - first_index, start,
- end, get_extent, bio,
+ end, get_extent, em_cached, bio,
mirror_num, bio_flags, rw);
}
@@ -2982,8 +3013,8 @@ static int __extent_read_full_page(struct extent_io_tree
*tree,
btrfs_put_ordered_extent(ordered);
}
- ret = __do_readpage(tree, page, get_extent, bio, mirror_num, bio_flags,
- rw);
+ ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
+ bio_flags, rw);
return ret;
}
@@ -3837,6 +3868,7 @@ int extent_readpages(struct extent_io_tree *tree,
unsigned long bio_flags = 0;
struct page *pagepool[16];
struct page *page;
+ struct extent_map *em_cached = NULL;
int nr = 0;
for (page_idx = 0; page_idx < nr_pages; page_idx++) {
@@ -3853,14 +3885,17 @@ int extent_readpages(struct extent_io_tree *tree,
pagepool[nr++] = page;
if (nr < ARRAY_SIZE(pagepool))
continue;
- __extent_readpages(tree, pagepool, nr, get_extent,
+ __extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
&bio, 0, &bio_flags, READ);
nr = 0;
}
if (nr)
- __extent_readpages(tree, pagepool, nr, get_extent,
+ __extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
&bio, 0, &bio_flags, READ);
+ if (em_cached)
+ free_extent_map(em_cached);
+
BUG_ON(!list_empty(pages));
if (bio)
return submit_one_bio(READ, bio, 0, bio_flags);
--
1.8.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs"
in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html