Miao Xie
2013-Jul-11 05:25 UTC
[PATCH 1/5] Btrfs: remove unnecessary argument of bio_readpage_error()
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
fs/btrfs/extent_io.c | 25 +++++++++++--------------
1 file changed, 11 insertions(+), 14 deletions(-)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index f8586a9..4bfbcc5 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;
@@ -2297,21 +2297,18 @@ static int bio_readpage_error(struct bio *failed_bio,
struct page *page,
* 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);
+ "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 +2538,7 @@ 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-11 05:25 UTC
[PATCH 2/5] Btrfs: add branch prediction hints in the read page end IO function
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
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 4bfbcc5..c9b28cf 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -2503,7 +2503,7 @@ 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
@@ -2513,7 +2513,8 @@ 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)
@@ -2522,12 +2523,15 @@ 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
@@ -2548,7 +2552,7 @@ 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-11 05:25 UTC
[PATCH 3/5] 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>
---
fs/btrfs/btrfs_inode.h | 21 +++++++++
fs/btrfs/ctree.h | 4 +-
fs/btrfs/disk-io.c | 5 ++-
fs/btrfs/extent_io.c | 113 ++++++++++++++++---------------------------------
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, 163 insertions(+), 163 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 f5b4b72..d52ec5d 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -3556,12 +3556,14 @@ 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 dfe6864..290b83f 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 c9b28cf..9f4dedf 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;
@@ -2303,13 +2247,6 @@ 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
@@ -2346,9 +2283,8 @@ 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;
@@ -2359,12 +2295,24 @@ 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 "
@@ -2463,9 +2411,12 @@ 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;
@@ -2476,7 +2427,6 @@ 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, "
@@ -2497,6 +2447,7 @@ 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);
@@ -2515,8 +2466,9 @@ 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
@@ -2542,7 +2494,8 @@ 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);
@@ -2574,8 +2527,11 @@ 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);
}
@@ -2587,6 +2543,7 @@ 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);
@@ -2602,6 +2559,10 @@ 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;
}
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 3332424..2b79026 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -2826,16 +2826,16 @@ 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);
@@ -2855,19 +2855,13 @@ 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);
@@ -2876,14 +2870,13 @@ 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;
}
@@ -6837,26 +6830,6 @@ 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;
@@ -6865,6 +6838,8 @@ 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;
@@ -6873,12 +6848,8 @@ 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,
@@ -6888,18 +6859,18 @@ 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,
@@ -7016,6 +6987,7 @@ 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;
@@ -7050,7 +7022,8 @@ 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;
}
@@ -7085,6 +7058,7 @@ 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;
@@ -7180,19 +7154,28 @@ 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-11 05:25 UTC
[PATCH 4/5] Btrfs: batch the extent state operation in the end io handle of the read page
It is unnecessary to unlock the extent by the page size, we can do it
in batches, it makes the random read be faster by ~6%.
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
---
fs/btrfs/extent_io.c | 70 ++++++++++++++++++++++++++++++----------------------
1 file changed, 40 insertions(+), 30 deletions(-)
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 9f4dedf..8f95418 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.
@@ -2395,6 +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
@@ -2417,6 +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;
@@ -2425,8 +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, "
@@ -2452,17 +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)) {
@@ -2501,18 +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;
@@ -2528,8 +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-11 08:05 UTC
Re: [PATCH 1/5] Btrfs: remove unnecessary argument of bio_readpage_error()
There are only 4 patches in this patchset, not 5. Sorry for my mistake. Miao On thu, 11 Jul 2013 13:25:36 +0800, Miao Xie wrote:> Signed-off-by: Miao Xie <miaox@cn.fujitsu.com> > --- > fs/btrfs/extent_io.c | 25 +++++++++++-------------- > 1 file changed, 11 insertions(+), 14 deletions(-) > > diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c > index f8586a9..4bfbcc5 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; > @@ -2297,21 +2297,18 @@ static int bio_readpage_error(struct bio *failed_bio, struct page *page, > * 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); > + "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 +2538,7 @@ 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); >-- 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
Chris Mason
2013-Jul-11 14:29 UTC
Re: [PATCH 3/5] Btrfs: don''t cache the csum value into the extent state tree
Quoting Miao Xie (2013-07-11 01:25:38)> 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.Perfect, this is a great way to use the extra bio struct. -chris -- 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
Chris Mason
2013-Jul-11 14:31 UTC
Re: [PATCH 2/5] Btrfs: add branch prediction hints in the read page end IO function
Do you have benchmark numbers for how much these help? I hesitate to bring in the likely/unlikely unless we can see it on the benchmarks. (The patch does look fine though) -chris -- 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
Josef Bacik
2013-Jul-11 18:56 UTC
Re: [PATCH 4/5] Btrfs: batch the extent state operation in the end io handle of the read page
On Thu, Jul 11, 2013 at 01:25:39PM +0800, Miao Xie wrote:> It is unnecessary to unlock the extent by the page size, we can do it > in batches, it makes the random read be faster by ~6%. > > Signed-off-by: Miao Xie <miaox@cn.fujitsu.com> > --- > fs/btrfs/extent_io.c | 70 ++++++++++++++++++++++++++++++---------------------- > 1 file changed, 40 insertions(+), 30 deletions(-) > > diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c > index 9f4dedf..8f95418 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. > @@ -2395,6 +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 > @@ -2417,6 +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; > > @@ -2425,8 +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, " > @@ -2452,17 +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)) { > @@ -2501,18 +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; > @@ -2528,8 +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);This patch is causing xfstest btrfs/265 to blow up, I''m kicking this series out until you fix it. Thanks, Josef -- 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
David Sterba
2013-Jul-12 22:19 UTC
Re: [PATCH 2/5] Btrfs: add branch prediction hints in the read page end IO function
On Thu, Jul 11, 2013 at 10:31:22AM -0400, Chris Mason wrote:> Do you have benchmark numbers for how much these help? I hesitate to > bring in the likely/unlikely unless we can see it on the benchmarks.Seconded, I doubt that this particular function gets any measurable speed improvement with the prediction hints. They''re eg suitable for branches that almost never happen like error conditions or shortcuts to the exit. There''s a config option CONFIG_PROFILE_ANNOTATED_BRANCHES to profile all likely/unlikely, so if you want to see the hint effects yourself feel free to experiment with that. Theres'' ''perf branch'' patchset, able to gather information from the Branch Trace Store (BTS), http://lwn.net/Articles/444885/ . There probably are functions in btrfs code where even a small improvement could bring some performance, first guess is btrfs_search_slot + the callees, but I''ve never seen it up in the perf top profile. david -- 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