Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
/*
|
|
|
|
* Copyright (C) 2008 Red Hat. All rights reserved.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public
|
|
|
|
* License v2 as published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU General Public
|
|
|
|
* License along with this program; if not, write to the
|
|
|
|
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
|
|
|
|
* Boston, MA 021110-1307, USA.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include "ctree.h"
|
|
|
|
|
|
|
|
static int tree_insert_offset(struct rb_root *root, u64 offset,
|
|
|
|
struct rb_node *node)
|
|
|
|
{
|
|
|
|
struct rb_node **p = &root->rb_node;
|
|
|
|
struct rb_node *parent = NULL;
|
|
|
|
struct btrfs_free_space *info;
|
|
|
|
|
|
|
|
while (*p) {
|
|
|
|
parent = *p;
|
|
|
|
info = rb_entry(parent, struct btrfs_free_space, offset_index);
|
|
|
|
|
|
|
|
if (offset < info->offset)
|
|
|
|
p = &(*p)->rb_left;
|
|
|
|
else if (offset > info->offset)
|
|
|
|
p = &(*p)->rb_right;
|
|
|
|
else
|
|
|
|
return -EEXIST;
|
|
|
|
}
|
|
|
|
|
|
|
|
rb_link_node(node, parent, p);
|
|
|
|
rb_insert_color(node, root);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int tree_insert_bytes(struct rb_root *root, u64 bytes,
|
|
|
|
struct rb_node *node)
|
|
|
|
{
|
|
|
|
struct rb_node **p = &root->rb_node;
|
|
|
|
struct rb_node *parent = NULL;
|
|
|
|
struct btrfs_free_space *info;
|
|
|
|
|
|
|
|
while (*p) {
|
|
|
|
parent = *p;
|
|
|
|
info = rb_entry(parent, struct btrfs_free_space, bytes_index);
|
|
|
|
|
|
|
|
if (bytes < info->bytes)
|
|
|
|
p = &(*p)->rb_left;
|
|
|
|
else
|
|
|
|
p = &(*p)->rb_right;
|
|
|
|
}
|
|
|
|
|
|
|
|
rb_link_node(node, parent, p);
|
|
|
|
rb_insert_color(node, root);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* searches the tree for the given offset. If contains is set we will return
|
|
|
|
* the free space that contains the given offset. If contains is not set we
|
|
|
|
* will return the free space that starts at or after the given offset and is
|
|
|
|
* at least bytes long.
|
|
|
|
*/
|
|
|
|
static struct btrfs_free_space *tree_search_offset(struct rb_root *root,
|
|
|
|
u64 offset, u64 bytes,
|
|
|
|
int contains)
|
|
|
|
{
|
|
|
|
struct rb_node *n = root->rb_node;
|
|
|
|
struct btrfs_free_space *entry, *ret = NULL;
|
|
|
|
|
|
|
|
while (n) {
|
|
|
|
entry = rb_entry(n, struct btrfs_free_space, offset_index);
|
|
|
|
|
|
|
|
if (offset < entry->offset) {
|
|
|
|
if (!contains &&
|
|
|
|
(!ret || entry->offset < ret->offset) &&
|
|
|
|
(bytes <= entry->bytes))
|
|
|
|
ret = entry;
|
|
|
|
n = n->rb_left;
|
|
|
|
} else if (offset > entry->offset) {
|
2008-10-10 18:24:32 +04:00
|
|
|
if ((entry->offset + entry->bytes - 1) >= offset &&
|
|
|
|
bytes <= entry->bytes) {
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
ret = entry;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
n = n->rb_right;
|
|
|
|
} else {
|
|
|
|
if (bytes > entry->bytes) {
|
|
|
|
n = n->rb_right;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
ret = entry;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return a chunk at least bytes size, as close to offset that we can get.
|
|
|
|
*/
|
|
|
|
static struct btrfs_free_space *tree_search_bytes(struct rb_root *root,
|
|
|
|
u64 offset, u64 bytes)
|
|
|
|
{
|
|
|
|
struct rb_node *n = root->rb_node;
|
|
|
|
struct btrfs_free_space *entry, *ret = NULL;
|
|
|
|
|
|
|
|
while (n) {
|
|
|
|
entry = rb_entry(n, struct btrfs_free_space, bytes_index);
|
|
|
|
|
|
|
|
if (bytes < entry->bytes) {
|
|
|
|
/*
|
|
|
|
* We prefer to get a hole size as close to the size we
|
|
|
|
* are asking for so we don't take small slivers out of
|
|
|
|
* huge holes, but we also want to get as close to the
|
|
|
|
* offset as possible so we don't have a whole lot of
|
|
|
|
* fragmentation.
|
|
|
|
*/
|
|
|
|
if (offset <= entry->offset) {
|
|
|
|
if (!ret)
|
|
|
|
ret = entry;
|
|
|
|
else if (entry->bytes < ret->bytes)
|
|
|
|
ret = entry;
|
|
|
|
else if (entry->offset < ret->offset)
|
|
|
|
ret = entry;
|
|
|
|
}
|
|
|
|
n = n->rb_left;
|
|
|
|
} else if (bytes > entry->bytes) {
|
|
|
|
n = n->rb_right;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Ok we may have multiple chunks of the wanted size,
|
|
|
|
* so we don't want to take the first one we find, we
|
|
|
|
* want to take the one closest to our given offset, so
|
|
|
|
* keep searching just in case theres a better match.
|
|
|
|
*/
|
|
|
|
n = n->rb_right;
|
|
|
|
if (offset > entry->offset)
|
|
|
|
continue;
|
|
|
|
else if (!ret || entry->offset < ret->offset)
|
|
|
|
ret = entry;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void unlink_free_space(struct btrfs_block_group_cache *block_group,
|
|
|
|
struct btrfs_free_space *info)
|
|
|
|
{
|
|
|
|
rb_erase(&info->offset_index, &block_group->free_space_offset);
|
|
|
|
rb_erase(&info->bytes_index, &block_group->free_space_bytes);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int link_free_space(struct btrfs_block_group_cache *block_group,
|
|
|
|
struct btrfs_free_space *info)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
|
|
|
|
ret = tree_insert_offset(&block_group->free_space_offset, info->offset,
|
|
|
|
&info->offset_index);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = tree_insert_bytes(&block_group->free_space_bytes, info->bytes,
|
|
|
|
&info->bytes_index);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
static int __btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
|
|
|
|
u64 offset, u64 bytes)
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
{
|
|
|
|
struct btrfs_free_space *right_info;
|
|
|
|
struct btrfs_free_space *left_info;
|
|
|
|
struct btrfs_free_space *info = NULL;
|
|
|
|
struct btrfs_free_space *alloc_info;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
alloc_info = kzalloc(sizeof(struct btrfs_free_space), GFP_NOFS);
|
|
|
|
if (!alloc_info)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* first we want to see if there is free space adjacent to the range we
|
|
|
|
* are adding, if there is remove that struct and add a new one to
|
|
|
|
* cover the entire range
|
|
|
|
*/
|
|
|
|
right_info = tree_search_offset(&block_group->free_space_offset,
|
|
|
|
offset+bytes, 0, 1);
|
|
|
|
left_info = tree_search_offset(&block_group->free_space_offset,
|
|
|
|
offset-1, 0, 1);
|
|
|
|
|
|
|
|
if (right_info && right_info->offset == offset+bytes) {
|
|
|
|
unlink_free_space(block_group, right_info);
|
|
|
|
info = right_info;
|
|
|
|
info->offset = offset;
|
|
|
|
info->bytes += bytes;
|
|
|
|
} else if (right_info && right_info->offset != offset+bytes) {
|
2009-01-06 05:25:51 +03:00
|
|
|
printk(KERN_ERR "btrfs adding space in the middle of an "
|
|
|
|
"existing free space area. existing: "
|
|
|
|
"offset=%llu, bytes=%llu. new: offset=%llu, "
|
|
|
|
"bytes=%llu\n", (unsigned long long)right_info->offset,
|
|
|
|
(unsigned long long)right_info->bytes,
|
|
|
|
(unsigned long long)offset,
|
|
|
|
(unsigned long long)bytes);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (left_info) {
|
|
|
|
unlink_free_space(block_group, left_info);
|
|
|
|
|
|
|
|
if (unlikely((left_info->offset + left_info->bytes) !=
|
|
|
|
offset)) {
|
2009-01-06 05:25:51 +03:00
|
|
|
printk(KERN_ERR "btrfs free space to the left "
|
|
|
|
"of new free space isn't "
|
|
|
|
"quite right. existing: offset=%llu, "
|
|
|
|
"bytes=%llu. new: offset=%llu, bytes=%llu\n",
|
|
|
|
(unsigned long long)left_info->offset,
|
|
|
|
(unsigned long long)left_info->bytes,
|
|
|
|
(unsigned long long)offset,
|
|
|
|
(unsigned long long)bytes);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (info) {
|
|
|
|
info->offset = left_info->offset;
|
|
|
|
info->bytes += left_info->bytes;
|
|
|
|
kfree(left_info);
|
|
|
|
} else {
|
|
|
|
info = left_info;
|
|
|
|
info->bytes += bytes;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (info) {
|
|
|
|
ret = link_free_space(block_group, info);
|
|
|
|
if (!ret)
|
|
|
|
info = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
info = alloc_info;
|
|
|
|
alloc_info = NULL;
|
|
|
|
info->offset = offset;
|
|
|
|
info->bytes = bytes;
|
|
|
|
|
|
|
|
ret = link_free_space(block_group, info);
|
|
|
|
if (ret)
|
|
|
|
kfree(info);
|
|
|
|
out:
|
|
|
|
if (ret) {
|
|
|
|
printk(KERN_ERR "btrfs: unable to add free space :%d\n", ret);
|
|
|
|
if (ret == -EEXIST)
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
2009-01-06 05:25:51 +03:00
|
|
|
kfree(alloc_info);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
static int
|
|
|
|
__btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
|
|
|
|
u64 offset, u64 bytes)
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
{
|
|
|
|
struct btrfs_free_space *info;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
info = tree_search_offset(&block_group->free_space_offset, offset, 0,
|
|
|
|
1);
|
|
|
|
|
|
|
|
if (info && info->offset == offset) {
|
|
|
|
if (info->bytes < bytes) {
|
2009-01-06 05:25:51 +03:00
|
|
|
printk(KERN_ERR "Found free space at %llu, size %llu,"
|
|
|
|
"trying to use %llu\n",
|
|
|
|
(unsigned long long)info->offset,
|
|
|
|
(unsigned long long)info->bytes,
|
|
|
|
(unsigned long long)bytes);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
WARN_ON(1);
|
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
unlink_free_space(block_group, info);
|
|
|
|
|
|
|
|
if (info->bytes == bytes) {
|
|
|
|
kfree(info);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
info->offset += bytes;
|
|
|
|
info->bytes -= bytes;
|
|
|
|
|
|
|
|
ret = link_free_space(block_group, info);
|
|
|
|
BUG_ON(ret);
|
2008-09-24 19:23:25 +04:00
|
|
|
} else if (info && info->offset < offset &&
|
|
|
|
info->offset + info->bytes >= offset + bytes) {
|
|
|
|
u64 old_start = info->offset;
|
|
|
|
/*
|
|
|
|
* we're freeing space in the middle of the info,
|
|
|
|
* this can happen during tree log replay
|
|
|
|
*
|
|
|
|
* first unlink the old info and then
|
|
|
|
* insert it again after the hole we're creating
|
|
|
|
*/
|
|
|
|
unlink_free_space(block_group, info);
|
|
|
|
if (offset + bytes < info->offset + info->bytes) {
|
|
|
|
u64 old_end = info->offset + info->bytes;
|
|
|
|
|
|
|
|
info->offset = offset + bytes;
|
|
|
|
info->bytes = old_end - info->offset;
|
|
|
|
ret = link_free_space(block_group, info);
|
|
|
|
BUG_ON(ret);
|
|
|
|
} else {
|
|
|
|
/* the hole we're creating ends at the end
|
|
|
|
* of the info struct, just free the info
|
|
|
|
*/
|
|
|
|
kfree(info);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* step two, insert a new info struct to cover anything
|
|
|
|
* before the hole
|
|
|
|
*/
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
ret = __btrfs_add_free_space(block_group, old_start,
|
|
|
|
offset - old_start);
|
2008-09-24 19:23:25 +04:00
|
|
|
BUG_ON(ret);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
} else {
|
|
|
|
WARN_ON(1);
|
|
|
|
}
|
|
|
|
out:
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
|
|
|
|
u64 offset, u64 bytes)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_free_space *sp;
|
|
|
|
|
|
|
|
mutex_lock(&block_group->alloc_mutex);
|
|
|
|
ret = __btrfs_add_free_space(block_group, offset, bytes);
|
|
|
|
sp = tree_search_offset(&block_group->free_space_offset, offset, 0, 1);
|
|
|
|
BUG_ON(!sp);
|
|
|
|
mutex_unlock(&block_group->alloc_mutex);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_add_free_space_lock(struct btrfs_block_group_cache *block_group,
|
|
|
|
u64 offset, u64 bytes)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_free_space *sp;
|
|
|
|
|
|
|
|
ret = __btrfs_add_free_space(block_group, offset, bytes);
|
|
|
|
sp = tree_search_offset(&block_group->free_space_offset, offset, 0, 1);
|
|
|
|
BUG_ON(!sp);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
|
|
|
|
u64 offset, u64 bytes)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
mutex_lock(&block_group->alloc_mutex);
|
|
|
|
ret = __btrfs_remove_free_space(block_group, offset, bytes);
|
|
|
|
mutex_unlock(&block_group->alloc_mutex);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_remove_free_space_lock(struct btrfs_block_group_cache *block_group,
|
|
|
|
u64 offset, u64 bytes)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = __btrfs_remove_free_space(block_group, offset, bytes);
|
|
|
|
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group,
|
|
|
|
u64 bytes)
|
|
|
|
{
|
|
|
|
struct btrfs_free_space *info;
|
|
|
|
struct rb_node *n;
|
|
|
|
int count = 0;
|
|
|
|
|
|
|
|
for (n = rb_first(&block_group->free_space_offset); n; n = rb_next(n)) {
|
|
|
|
info = rb_entry(n, struct btrfs_free_space, offset_index);
|
|
|
|
if (info->bytes >= bytes)
|
|
|
|
count++;
|
|
|
|
}
|
|
|
|
printk(KERN_INFO "%d blocks of free space at or bigger than bytes is"
|
|
|
|
"\n", count);
|
|
|
|
}
|
|
|
|
|
|
|
|
u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group)
|
|
|
|
{
|
|
|
|
struct btrfs_free_space *info;
|
|
|
|
struct rb_node *n;
|
|
|
|
u64 ret = 0;
|
|
|
|
|
|
|
|
for (n = rb_first(&block_group->free_space_offset); n;
|
|
|
|
n = rb_next(n)) {
|
|
|
|
info = rb_entry(n, struct btrfs_free_space, offset_index);
|
|
|
|
ret += info->bytes;
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
|
|
|
|
{
|
|
|
|
struct btrfs_free_space *info;
|
|
|
|
struct rb_node *node;
|
|
|
|
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
mutex_lock(&block_group->alloc_mutex);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
while ((node = rb_last(&block_group->free_space_bytes)) != NULL) {
|
|
|
|
info = rb_entry(node, struct btrfs_free_space, bytes_index);
|
|
|
|
unlink_free_space(block_group, info);
|
|
|
|
kfree(info);
|
|
|
|
if (need_resched()) {
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
mutex_unlock(&block_group->alloc_mutex);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
cond_resched();
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
mutex_lock(&block_group->alloc_mutex);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
}
|
|
|
|
}
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
mutex_unlock(&block_group->alloc_mutex);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
}
|
|
|
|
|
2008-12-02 17:54:17 +03:00
|
|
|
#if 0
|
|
|
|
static struct btrfs_free_space *btrfs_find_free_space_offset(struct
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
btrfs_block_group_cache
|
|
|
|
*block_group, u64 offset,
|
|
|
|
u64 bytes)
|
|
|
|
{
|
|
|
|
struct btrfs_free_space *ret;
|
|
|
|
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
mutex_lock(&block_group->alloc_mutex);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
ret = tree_search_offset(&block_group->free_space_offset, offset,
|
|
|
|
bytes, 0);
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
mutex_unlock(&block_group->alloc_mutex);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-12-02 17:54:17 +03:00
|
|
|
static struct btrfs_free_space *btrfs_find_free_space_bytes(struct
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
btrfs_block_group_cache
|
|
|
|
*block_group, u64 offset,
|
|
|
|
u64 bytes)
|
|
|
|
{
|
|
|
|
struct btrfs_free_space *ret;
|
|
|
|
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
mutex_lock(&block_group->alloc_mutex);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
|
|
|
|
ret = tree_search_bytes(&block_group->free_space_bytes, offset, bytes);
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
mutex_unlock(&block_group->alloc_mutex);
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
2008-12-02 17:54:17 +03:00
|
|
|
#endif
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
|
|
|
|
struct btrfs_free_space *btrfs_find_free_space(struct btrfs_block_group_cache
|
|
|
|
*block_group, u64 offset,
|
|
|
|
u64 bytes)
|
|
|
|
{
|
Btrfs: nuke fs wide allocation mutex V2
This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.
There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.
The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same. Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.
Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one. I have tested this heavily and it does
not appear to break anything. This has to be applied on top of my
find_free_extent redo patch.
I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out. Thank you,
Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 21:49:05 +03:00
|
|
|
struct btrfs_free_space *ret = NULL;
|
Btrfs: free space accounting redo
1) replace the per fs_info extent_io_tree that tracked free space with two
rb-trees per block group to track free space areas via offset and size. The
reason to do this is because most allocations come with a hint byte where to
start, so we can usually find a chunk of free space at that hint byte to satisfy
the allocation and get good space packing. If we cannot find free space at or
after the given offset we fall back on looking for a chunk of the given size as
close to that given offset as possible. When we fall back on the size search we
also try to find a slot as close to the size we want as possible, to avoid
breaking small chunks off of huge areas if possible.
2) remove the extent_io_tree that tracked the block group cache from fs_info and
replaced it with an rb-tree thats tracks block group cache via offset. also
added a per space_info list that tracks the block group cache for the particular
space so we can lookup related block groups easily.
3) cleaned up the allocation code to make it a little easier to read and a
little less complicated. Basically there are 3 steps, first look from our
provided hint. If we couldn't find from that given hint, start back at our
original search start and look for space from there. If that fails try to
allocate space if we can and start looking again. If not we're screwed and need
to start over again.
4) small fixes. there were some issues in volumes.c where we wouldn't allocate
the rest of the disk. fixed cow_file_range to actually pass the alloc_hint,
which has helped a good bit in making the fs_mark test I run have semi-normal
results as we run out of space. Generally with data allocations we don't track
where we last allocated from, so everytime we did a data allocation we'd search
through every block group that we have looking for free space. Now searching a
block group with no free space isn't terribly time consuming, it was causing a
slight degradation as we got more data block groups. The alloc_hint has fixed
this slight degredation and made things semi-normal.
There is still one nagging problem I'm working on where we will get ENOSPC when
there is definitely plenty of space. This only happens with metadata
allocations, and only when we are almost full. So you generally hit the 85%
mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm
still tracking it down, but until then this seems to be pretty stable and make a
significant performance gain.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 21:14:11 +04:00
|
|
|
|
|
|
|
ret = tree_search_offset(&block_group->free_space_offset, offset,
|
|
|
|
bytes, 0);
|
|
|
|
if (!ret)
|
|
|
|
ret = tree_search_bytes(&block_group->free_space_bytes,
|
|
|
|
offset, bytes);
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|