2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2005-11-02 10:58:39 +07:00
|
|
|
* Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
|
|
|
|
* All Rights Reserved.
|
2005-04-17 05:20:36 +07:00
|
|
|
*
|
2005-11-02 10:58:39 +07:00
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License as
|
2005-04-17 05:20:36 +07:00
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*
|
2005-11-02 10:58:39 +07:00
|
|
|
* This program is distributed in the hope that it would 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.
|
2005-04-17 05:20:36 +07:00
|
|
|
*
|
2005-11-02 10:58:39 +07:00
|
|
|
* You should have received a copy of the GNU General Public License
|
|
|
|
* along with this program; if not, write the Free Software Foundation,
|
|
|
|
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
#ifndef __XFS_AG_H__
|
|
|
|
#define __XFS_AG_H__
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocation group header
|
|
|
|
* This is divided into three structures, placed in sequential 512-byte
|
|
|
|
* buffers after a copy of the superblock (also in a 512-byte buffer).
|
|
|
|
*/
|
|
|
|
|
|
|
|
struct xfs_buf;
|
|
|
|
struct xfs_mount;
|
|
|
|
struct xfs_trans;
|
|
|
|
|
|
|
|
#define XFS_AGF_MAGIC 0x58414746 /* 'XAGF' */
|
|
|
|
#define XFS_AGI_MAGIC 0x58414749 /* 'XAGI' */
|
|
|
|
#define XFS_AGF_VERSION 1
|
|
|
|
#define XFS_AGI_VERSION 1
|
2005-11-02 10:38:42 +07:00
|
|
|
|
|
|
|
#define XFS_AGF_GOOD_VERSION(v) ((v) == XFS_AGF_VERSION)
|
|
|
|
#define XFS_AGI_GOOD_VERSION(v) ((v) == XFS_AGI_VERSION)
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Btree number 0 is bno, 1 is cnt. This value gives the size of the
|
|
|
|
* arrays below.
|
|
|
|
*/
|
|
|
|
#define XFS_BTNUM_AGF ((int)XFS_BTNUM_CNTi + 1)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The second word of agf_levels in the first a.g. overlaps the EFS
|
|
|
|
* superblock's magic number. Since the magic numbers valid for EFS
|
|
|
|
* are > 64k, our value cannot be confused for an EFS superblock's.
|
|
|
|
*/
|
|
|
|
|
2005-11-02 11:11:25 +07:00
|
|
|
typedef struct xfs_agf {
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Common allocation group header information
|
|
|
|
*/
|
2005-11-02 11:11:25 +07:00
|
|
|
__be32 agf_magicnum; /* magic number == XFS_AGF_MAGIC */
|
|
|
|
__be32 agf_versionnum; /* header version == XFS_AGF_VERSION */
|
|
|
|
__be32 agf_seqno; /* sequence # starting from 0 */
|
|
|
|
__be32 agf_length; /* size in blocks of a.g. */
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Freespace information
|
|
|
|
*/
|
2005-11-02 11:11:25 +07:00
|
|
|
__be32 agf_roots[XFS_BTNUM_AGF]; /* root blocks */
|
|
|
|
__be32 agf_spare0; /* spare field */
|
|
|
|
__be32 agf_levels[XFS_BTNUM_AGF]; /* btree levels */
|
|
|
|
__be32 agf_spare1; /* spare field */
|
|
|
|
__be32 agf_flfirst; /* first freelist block's index */
|
|
|
|
__be32 agf_fllast; /* last freelist block's index */
|
|
|
|
__be32 agf_flcount; /* count of blocks in freelist */
|
|
|
|
__be32 agf_freeblks; /* total free blocks */
|
|
|
|
__be32 agf_longest; /* longest free space */
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 12:26:31 +07:00
|
|
|
__be32 agf_btreeblks; /* # of blocks held in AGF btrees */
|
2005-04-17 05:20:36 +07:00
|
|
|
} xfs_agf_t;
|
|
|
|
|
|
|
|
#define XFS_AGF_MAGICNUM 0x00000001
|
|
|
|
#define XFS_AGF_VERSIONNUM 0x00000002
|
|
|
|
#define XFS_AGF_SEQNO 0x00000004
|
|
|
|
#define XFS_AGF_LENGTH 0x00000008
|
|
|
|
#define XFS_AGF_ROOTS 0x00000010
|
|
|
|
#define XFS_AGF_LEVELS 0x00000020
|
|
|
|
#define XFS_AGF_FLFIRST 0x00000040
|
|
|
|
#define XFS_AGF_FLLAST 0x00000080
|
|
|
|
#define XFS_AGF_FLCOUNT 0x00000100
|
|
|
|
#define XFS_AGF_FREEBLKS 0x00000200
|
|
|
|
#define XFS_AGF_LONGEST 0x00000400
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 12:26:31 +07:00
|
|
|
#define XFS_AGF_BTREEBLKS 0x00000800
|
|
|
|
#define XFS_AGF_NUM_BITS 12
|
2005-04-17 05:20:36 +07:00
|
|
|
#define XFS_AGF_ALL_BITS ((1 << XFS_AGF_NUM_BITS) - 1)
|
|
|
|
|
2009-12-15 06:14:59 +07:00
|
|
|
#define XFS_AGF_FLAGS \
|
|
|
|
{ XFS_AGF_MAGICNUM, "MAGICNUM" }, \
|
|
|
|
{ XFS_AGF_VERSIONNUM, "VERSIONNUM" }, \
|
|
|
|
{ XFS_AGF_SEQNO, "SEQNO" }, \
|
|
|
|
{ XFS_AGF_LENGTH, "LENGTH" }, \
|
|
|
|
{ XFS_AGF_ROOTS, "ROOTS" }, \
|
|
|
|
{ XFS_AGF_LEVELS, "LEVELS" }, \
|
|
|
|
{ XFS_AGF_FLFIRST, "FLFIRST" }, \
|
|
|
|
{ XFS_AGF_FLLAST, "FLLAST" }, \
|
|
|
|
{ XFS_AGF_FLCOUNT, "FLCOUNT" }, \
|
|
|
|
{ XFS_AGF_FREEBLKS, "FREEBLKS" }, \
|
|
|
|
{ XFS_AGF_LONGEST, "LONGEST" }, \
|
|
|
|
{ XFS_AGF_BTREEBLKS, "BTREEBLKS" }
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/* disk block (xfs_daddr_t) in the AG */
|
|
|
|
#define XFS_AGF_DADDR(mp) ((xfs_daddr_t)(1 << (mp)->m_sectbb_log))
|
2005-11-02 10:38:42 +07:00
|
|
|
#define XFS_AGF_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGF_DADDR(mp))
|
2011-07-23 06:40:15 +07:00
|
|
|
#define XFS_BUF_TO_AGF(bp) ((xfs_agf_t *)((bp)->b_addr))
|
2005-11-02 10:38:42 +07:00
|
|
|
|
2009-07-03 09:35:43 +07:00
|
|
|
extern int xfs_read_agf(struct xfs_mount *mp, struct xfs_trans *tp,
|
|
|
|
xfs_agnumber_t agno, int flags, struct xfs_buf **bpp);
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Size of the unlinked inode hash table in the agi.
|
|
|
|
*/
|
|
|
|
#define XFS_AGI_UNLINKED_BUCKETS 64
|
|
|
|
|
2005-11-02 11:11:25 +07:00
|
|
|
typedef struct xfs_agi {
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Common allocation group header information
|
|
|
|
*/
|
2005-11-02 11:11:25 +07:00
|
|
|
__be32 agi_magicnum; /* magic number == XFS_AGI_MAGIC */
|
|
|
|
__be32 agi_versionnum; /* header version == XFS_AGI_VERSION */
|
|
|
|
__be32 agi_seqno; /* sequence # starting from 0 */
|
|
|
|
__be32 agi_length; /* size in blocks of a.g. */
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Inode information
|
|
|
|
* Inodes are mapped by interpreting the inode number, so no
|
|
|
|
* mapping data is needed here.
|
|
|
|
*/
|
2005-11-02 11:11:25 +07:00
|
|
|
__be32 agi_count; /* count of allocated inodes */
|
|
|
|
__be32 agi_root; /* root of inode btree */
|
|
|
|
__be32 agi_level; /* levels in inode btree */
|
|
|
|
__be32 agi_freecount; /* number of free inodes */
|
|
|
|
__be32 agi_newino; /* new inode just allocated */
|
|
|
|
__be32 agi_dirino; /* last directory inode chunk */
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Hash table of inodes which have been unlinked but are
|
|
|
|
* still being referenced.
|
|
|
|
*/
|
2005-11-02 11:11:25 +07:00
|
|
|
__be32 agi_unlinked[XFS_AGI_UNLINKED_BUCKETS];
|
2005-04-17 05:20:36 +07:00
|
|
|
} xfs_agi_t;
|
|
|
|
|
|
|
|
#define XFS_AGI_MAGICNUM 0x00000001
|
|
|
|
#define XFS_AGI_VERSIONNUM 0x00000002
|
|
|
|
#define XFS_AGI_SEQNO 0x00000004
|
|
|
|
#define XFS_AGI_LENGTH 0x00000008
|
|
|
|
#define XFS_AGI_COUNT 0x00000010
|
|
|
|
#define XFS_AGI_ROOT 0x00000020
|
|
|
|
#define XFS_AGI_LEVEL 0x00000040
|
|
|
|
#define XFS_AGI_FREECOUNT 0x00000080
|
|
|
|
#define XFS_AGI_NEWINO 0x00000100
|
|
|
|
#define XFS_AGI_DIRINO 0x00000200
|
|
|
|
#define XFS_AGI_UNLINKED 0x00000400
|
|
|
|
#define XFS_AGI_NUM_BITS 11
|
|
|
|
#define XFS_AGI_ALL_BITS ((1 << XFS_AGI_NUM_BITS) - 1)
|
|
|
|
|
|
|
|
/* disk block (xfs_daddr_t) in the AG */
|
|
|
|
#define XFS_AGI_DADDR(mp) ((xfs_daddr_t)(2 << (mp)->m_sectbb_log))
|
2005-11-02 10:38:42 +07:00
|
|
|
#define XFS_AGI_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGI_DADDR(mp))
|
2011-07-23 06:40:15 +07:00
|
|
|
#define XFS_BUF_TO_AGI(bp) ((xfs_agi_t *)((bp)->b_addr))
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-11-28 10:23:37 +07:00
|
|
|
extern int xfs_read_agi(struct xfs_mount *mp, struct xfs_trans *tp,
|
|
|
|
xfs_agnumber_t agno, struct xfs_buf **bpp);
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* The third a.g. block contains the a.g. freelist, an array
|
|
|
|
* of block pointers to blocks owned by the allocation btree code.
|
|
|
|
*/
|
|
|
|
#define XFS_AGFL_DADDR(mp) ((xfs_daddr_t)(3 << (mp)->m_sectbb_log))
|
2005-11-02 10:38:42 +07:00
|
|
|
#define XFS_AGFL_BLOCK(mp) XFS_HDR_BLOCK(mp, XFS_AGFL_DADDR(mp))
|
2005-04-17 05:20:36 +07:00
|
|
|
#define XFS_AGFL_SIZE(mp) ((mp)->m_sb.sb_sectsize / sizeof(xfs_agblock_t))
|
2011-07-23 06:40:15 +07:00
|
|
|
#define XFS_BUF_TO_AGFL(bp) ((xfs_agfl_t *)((bp)->b_addr))
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
typedef struct xfs_agfl {
|
2006-09-28 07:56:51 +07:00
|
|
|
__be32 agfl_bno[1]; /* actually XFS_AGFL_SIZE(mp) */
|
2005-04-17 05:20:36 +07:00
|
|
|
} xfs_agfl_t;
|
|
|
|
|
|
|
|
/*
|
xfs: Improve scalability of busy extent tracking
When we free a metadata extent, we record it in the per-AG busy
extent array so that it is not re-used before the freeing
transaction hits the disk. This array is fixed size, so when it
overflows we make further allocation transactions synchronous
because we cannot track more freed extents until those transactions
hit the disk and are completed. Under heavy mixed allocation and
freeing workloads with large log buffers, we can overflow this array
quite easily.
Further, the array is sparsely populated, which means that inserts
need to search for a free slot, and array searches often have to
search many more slots that are actually used to check all the
busy extents. Quite inefficient, really.
To enable this aspect of extent freeing to scale better, we need
a structure that can grow dynamically. While in other areas of
XFS we have used radix trees, the extents being freed are at random
locations on disk so are better suited to being indexed by an rbtree.
So, use a per-AG rbtree indexed by block number to track busy
extents. This incures a memory allocation when marking an extent
busy, but should not occur too often in low memory situations. This
should scale to an arbitrary number of extents so should not be a
limitation for features such as in-memory aggregation of
transactions.
However, there are still situations where we can't avoid allocating
busy extents (such as allocation from the AGFL). To minimise the
overhead of such occurences, we need to avoid doing a synchronous
log force while holding the AGF locked to ensure that the previous
transactions are safely on disk before we use the extent. We can do
this by marking the transaction doing the allocation as synchronous
rather issuing a log force.
Because of the locking involved and the ordering of transactions,
the synchronous transaction provides the same guarantees as a
synchronous log force because it ensures that all the prior
transactions are already on disk when the synchronous transaction
hits the disk. i.e. it preserves the free->allocate order of the
extent correctly in recovery.
By doing this, we avoid holding the AGF locked while log writes are
in progress, hence reducing the length of time the lock is held and
therefore we increase the rate at which we can allocate and free
from the allocation group, thereby increasing overall throughput.
The only problem with this approach is that when a metadata buffer is
marked stale (e.g. a directory block is removed), then buffer remains
pinned and locked until the log goes to disk. The issue here is that
if that stale buffer is reallocated in a subsequent transaction, the
attempt to lock that buffer in the transaction will hang waiting
the log to go to disk to unlock and unpin the buffer. Hence if
someone tries to lock a pinned, stale, locked buffer we need to
push on the log to get it unlocked ASAP. Effectively we are trading
off a guaranteed log force for a much less common trigger for log
force to occur.
Ideally we should not reallocate busy extents. That is a much more
complex fix to the problem as it involves direct intervention in the
allocation btree searches in many places. This is left to a future
set of modifications.
Finally, now that we track busy extents in allocated memory, we
don't need the descriptors in the transaction structure to point to
them. We can replace the complex busy chunk infrastructure with a
simple linked list of busy extents. This allows us to remove a large
chunk of code, making the overall change a net reduction in code
size.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-21 09:07:08 +07:00
|
|
|
* Busy block/extent entry. Indexed by a rbtree in perag to mark blocks that
|
|
|
|
* have been freed but whose transactions aren't committed to disk yet.
|
|
|
|
*
|
|
|
|
* Note that we use the transaction ID to record the transaction, not the
|
|
|
|
* transaction structure itself. See xfs_alloc_busy_insert() for details.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
xfs: Improve scalability of busy extent tracking
When we free a metadata extent, we record it in the per-AG busy
extent array so that it is not re-used before the freeing
transaction hits the disk. This array is fixed size, so when it
overflows we make further allocation transactions synchronous
because we cannot track more freed extents until those transactions
hit the disk and are completed. Under heavy mixed allocation and
freeing workloads with large log buffers, we can overflow this array
quite easily.
Further, the array is sparsely populated, which means that inserts
need to search for a free slot, and array searches often have to
search many more slots that are actually used to check all the
busy extents. Quite inefficient, really.
To enable this aspect of extent freeing to scale better, we need
a structure that can grow dynamically. While in other areas of
XFS we have used radix trees, the extents being freed are at random
locations on disk so are better suited to being indexed by an rbtree.
So, use a per-AG rbtree indexed by block number to track busy
extents. This incures a memory allocation when marking an extent
busy, but should not occur too often in low memory situations. This
should scale to an arbitrary number of extents so should not be a
limitation for features such as in-memory aggregation of
transactions.
However, there are still situations where we can't avoid allocating
busy extents (such as allocation from the AGFL). To minimise the
overhead of such occurences, we need to avoid doing a synchronous
log force while holding the AGF locked to ensure that the previous
transactions are safely on disk before we use the extent. We can do
this by marking the transaction doing the allocation as synchronous
rather issuing a log force.
Because of the locking involved and the ordering of transactions,
the synchronous transaction provides the same guarantees as a
synchronous log force because it ensures that all the prior
transactions are already on disk when the synchronous transaction
hits the disk. i.e. it preserves the free->allocate order of the
extent correctly in recovery.
By doing this, we avoid holding the AGF locked while log writes are
in progress, hence reducing the length of time the lock is held and
therefore we increase the rate at which we can allocate and free
from the allocation group, thereby increasing overall throughput.
The only problem with this approach is that when a metadata buffer is
marked stale (e.g. a directory block is removed), then buffer remains
pinned and locked until the log goes to disk. The issue here is that
if that stale buffer is reallocated in a subsequent transaction, the
attempt to lock that buffer in the transaction will hang waiting
the log to go to disk to unlock and unpin the buffer. Hence if
someone tries to lock a pinned, stale, locked buffer we need to
push on the log to get it unlocked ASAP. Effectively we are trading
off a guaranteed log force for a much less common trigger for log
force to occur.
Ideally we should not reallocate busy extents. That is a much more
complex fix to the problem as it involves direct intervention in the
allocation btree searches in many places. This is left to a future
set of modifications.
Finally, now that we track busy extents in allocated memory, we
don't need the descriptors in the transaction structure to point to
them. We can replace the complex busy chunk infrastructure with a
simple linked list of busy extents. This allows us to remove a large
chunk of code, making the overall change a net reduction in code
size.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-21 09:07:08 +07:00
|
|
|
struct xfs_busy_extent {
|
|
|
|
struct rb_node rb_node; /* ag by-bno indexed search tree */
|
|
|
|
struct list_head list; /* transaction busy extent list */
|
|
|
|
xfs_agnumber_t agno;
|
|
|
|
xfs_agblock_t bno;
|
|
|
|
xfs_extlen_t length;
|
2011-05-20 20:45:32 +07:00
|
|
|
unsigned int flags;
|
|
|
|
#define XFS_ALLOC_BUSY_DISCARDED 0x01 /* undergoing a discard op. */
|
2011-05-05 01:55:15 +07:00
|
|
|
#define XFS_ALLOC_BUSY_SKIP_DISCARD 0x02 /* do not discard */
|
xfs: Improve scalability of busy extent tracking
When we free a metadata extent, we record it in the per-AG busy
extent array so that it is not re-used before the freeing
transaction hits the disk. This array is fixed size, so when it
overflows we make further allocation transactions synchronous
because we cannot track more freed extents until those transactions
hit the disk and are completed. Under heavy mixed allocation and
freeing workloads with large log buffers, we can overflow this array
quite easily.
Further, the array is sparsely populated, which means that inserts
need to search for a free slot, and array searches often have to
search many more slots that are actually used to check all the
busy extents. Quite inefficient, really.
To enable this aspect of extent freeing to scale better, we need
a structure that can grow dynamically. While in other areas of
XFS we have used radix trees, the extents being freed are at random
locations on disk so are better suited to being indexed by an rbtree.
So, use a per-AG rbtree indexed by block number to track busy
extents. This incures a memory allocation when marking an extent
busy, but should not occur too often in low memory situations. This
should scale to an arbitrary number of extents so should not be a
limitation for features such as in-memory aggregation of
transactions.
However, there are still situations where we can't avoid allocating
busy extents (such as allocation from the AGFL). To minimise the
overhead of such occurences, we need to avoid doing a synchronous
log force while holding the AGF locked to ensure that the previous
transactions are safely on disk before we use the extent. We can do
this by marking the transaction doing the allocation as synchronous
rather issuing a log force.
Because of the locking involved and the ordering of transactions,
the synchronous transaction provides the same guarantees as a
synchronous log force because it ensures that all the prior
transactions are already on disk when the synchronous transaction
hits the disk. i.e. it preserves the free->allocate order of the
extent correctly in recovery.
By doing this, we avoid holding the AGF locked while log writes are
in progress, hence reducing the length of time the lock is held and
therefore we increase the rate at which we can allocate and free
from the allocation group, thereby increasing overall throughput.
The only problem with this approach is that when a metadata buffer is
marked stale (e.g. a directory block is removed), then buffer remains
pinned and locked until the log goes to disk. The issue here is that
if that stale buffer is reallocated in a subsequent transaction, the
attempt to lock that buffer in the transaction will hang waiting
the log to go to disk to unlock and unpin the buffer. Hence if
someone tries to lock a pinned, stale, locked buffer we need to
push on the log to get it unlocked ASAP. Effectively we are trading
off a guaranteed log force for a much less common trigger for log
force to occur.
Ideally we should not reallocate busy extents. That is a much more
complex fix to the problem as it involves direct intervention in the
allocation btree searches in many places. This is left to a future
set of modifications.
Finally, now that we track busy extents in allocated memory, we
don't need the descriptors in the transaction structure to point to
them. We can replace the complex busy chunk infrastructure with a
simple linked list of busy extents. This allows us to remove a large
chunk of code, making the overall change a net reduction in code
size.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-21 09:07:08 +07:00
|
|
|
};
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Per-ag incore structure, copies of information in agf and agi,
|
|
|
|
* to improve the performance of allocation group selection.
|
|
|
|
*/
|
|
|
|
#define XFS_PAGB_NUM_SLOTS 128
|
|
|
|
|
2010-01-11 18:47:45 +07:00
|
|
|
typedef struct xfs_perag {
|
2010-01-11 18:47:46 +07:00
|
|
|
struct xfs_mount *pag_mount; /* owner filesystem */
|
|
|
|
xfs_agnumber_t pag_agno; /* AG this structure belongs to */
|
2010-01-11 18:47:45 +07:00
|
|
|
atomic_t pag_ref; /* perag reference count */
|
2005-04-17 05:20:36 +07:00
|
|
|
char pagf_init; /* this agf's entry is initialized */
|
|
|
|
char pagi_init; /* this agi's entry is initialized */
|
2006-03-29 05:55:14 +07:00
|
|
|
char pagf_metadata; /* the agf is preferred to be metadata */
|
2005-04-17 05:20:36 +07:00
|
|
|
char pagi_inodeok; /* The agi is ok for inodes */
|
|
|
|
__uint8_t pagf_levels[XFS_BTNUM_AGF];
|
|
|
|
/* # of levels in bno & cnt btree */
|
|
|
|
__uint32_t pagf_flcount; /* count of blocks in freelist */
|
|
|
|
xfs_extlen_t pagf_freeblks; /* total free blocks */
|
|
|
|
xfs_extlen_t pagf_longest; /* longest free space */
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 12:26:31 +07:00
|
|
|
__uint32_t pagf_btreeblks; /* # of blocks held in AGF btrees */
|
2005-04-17 05:20:36 +07:00
|
|
|
xfs_agino_t pagi_freecount; /* number of free inodes */
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 12:26:31 +07:00
|
|
|
xfs_agino_t pagi_count; /* number of allocated inodes */
|
2009-09-01 06:58:28 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Inode allocation search lookup optimisation.
|
|
|
|
* If the pagino matches, the search for new inodes
|
|
|
|
* doesn't need to search the near ones again straight away
|
|
|
|
*/
|
|
|
|
xfs_agino_t pagl_pagino;
|
|
|
|
xfs_agino_t pagl_leftrec;
|
|
|
|
xfs_agino_t pagl_rightrec;
|
2005-04-17 05:20:36 +07:00
|
|
|
#ifdef __KERNEL__
|
xfs: Improve scalability of busy extent tracking
When we free a metadata extent, we record it in the per-AG busy
extent array so that it is not re-used before the freeing
transaction hits the disk. This array is fixed size, so when it
overflows we make further allocation transactions synchronous
because we cannot track more freed extents until those transactions
hit the disk and are completed. Under heavy mixed allocation and
freeing workloads with large log buffers, we can overflow this array
quite easily.
Further, the array is sparsely populated, which means that inserts
need to search for a free slot, and array searches often have to
search many more slots that are actually used to check all the
busy extents. Quite inefficient, really.
To enable this aspect of extent freeing to scale better, we need
a structure that can grow dynamically. While in other areas of
XFS we have used radix trees, the extents being freed are at random
locations on disk so are better suited to being indexed by an rbtree.
So, use a per-AG rbtree indexed by block number to track busy
extents. This incures a memory allocation when marking an extent
busy, but should not occur too often in low memory situations. This
should scale to an arbitrary number of extents so should not be a
limitation for features such as in-memory aggregation of
transactions.
However, there are still situations where we can't avoid allocating
busy extents (such as allocation from the AGFL). To minimise the
overhead of such occurences, we need to avoid doing a synchronous
log force while holding the AGF locked to ensure that the previous
transactions are safely on disk before we use the extent. We can do
this by marking the transaction doing the allocation as synchronous
rather issuing a log force.
Because of the locking involved and the ordering of transactions,
the synchronous transaction provides the same guarantees as a
synchronous log force because it ensures that all the prior
transactions are already on disk when the synchronous transaction
hits the disk. i.e. it preserves the free->allocate order of the
extent correctly in recovery.
By doing this, we avoid holding the AGF locked while log writes are
in progress, hence reducing the length of time the lock is held and
therefore we increase the rate at which we can allocate and free
from the allocation group, thereby increasing overall throughput.
The only problem with this approach is that when a metadata buffer is
marked stale (e.g. a directory block is removed), then buffer remains
pinned and locked until the log goes to disk. The issue here is that
if that stale buffer is reallocated in a subsequent transaction, the
attempt to lock that buffer in the transaction will hang waiting
the log to go to disk to unlock and unpin the buffer. Hence if
someone tries to lock a pinned, stale, locked buffer we need to
push on the log to get it unlocked ASAP. Effectively we are trading
off a guaranteed log force for a much less common trigger for log
force to occur.
Ideally we should not reallocate busy extents. That is a much more
complex fix to the problem as it involves direct intervention in the
allocation btree searches in many places. This is left to a future
set of modifications.
Finally, now that we track busy extents in allocated memory, we
don't need the descriptors in the transaction structure to point to
them. We can replace the complex busy chunk infrastructure with a
simple linked list of busy extents. This allows us to remove a large
chunk of code, making the overall change a net reduction in code
size.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-21 09:07:08 +07:00
|
|
|
spinlock_t pagb_lock; /* lock for pagb_tree */
|
|
|
|
struct rb_root pagb_tree; /* ordered tree of busy extents */
|
2008-10-30 13:05:38 +07:00
|
|
|
|
2007-07-11 08:09:12 +07:00
|
|
|
atomic_t pagf_fstrms; /* # of filestreams active in this AG */
|
2007-08-28 11:00:13 +07:00
|
|
|
|
2010-12-16 13:08:41 +07:00
|
|
|
spinlock_t pag_ici_lock; /* incore inode cache lock */
|
2007-08-28 11:00:13 +07:00
|
|
|
struct radix_tree_root pag_ici_root; /* incore inode cache root */
|
2010-04-29 06:55:50 +07:00
|
|
|
int pag_ici_reclaimable; /* reclaimable inodes */
|
2010-09-27 08:09:51 +07:00
|
|
|
struct mutex pag_ici_reclaim_lock; /* serialisation point */
|
|
|
|
unsigned long pag_ici_reclaim_cursor; /* reclaim restart point */
|
2010-09-22 07:47:20 +07:00
|
|
|
|
2010-09-24 16:59:04 +07:00
|
|
|
/* buffer cache index */
|
|
|
|
spinlock_t pag_buf_lock; /* lock for pag_buf_tree */
|
|
|
|
struct rb_root pag_buf_tree; /* ordered tree of active buffers */
|
|
|
|
|
2010-09-22 07:47:20 +07:00
|
|
|
/* for rcu-safe freeing */
|
|
|
|
struct rcu_head rcu_head;
|
2008-10-30 13:05:38 +07:00
|
|
|
#endif
|
2010-01-11 18:47:49 +07:00
|
|
|
int pagb_count; /* pagb slots in use */
|
2005-04-17 05:20:36 +07:00
|
|
|
} xfs_perag_t;
|
|
|
|
|
2008-10-30 13:37:26 +07:00
|
|
|
/*
|
|
|
|
* tags for inode radix tree
|
|
|
|
*/
|
2009-06-08 20:35:14 +07:00
|
|
|
#define XFS_ICI_NO_TAG (-1) /* special flag for an untagged lookup
|
|
|
|
in xfs_inode_ag_iterator */
|
2008-10-30 13:37:26 +07:00
|
|
|
#define XFS_ICI_RECLAIM_TAG 0 /* inode is to be reclaimed */
|
|
|
|
|
2005-11-02 10:38:42 +07:00
|
|
|
#define XFS_AG_MAXLEVELS(mp) ((mp)->m_ag_maxlevels)
|
|
|
|
#define XFS_MIN_FREELIST_RAW(bl,cl,mp) \
|
|
|
|
(MIN(bl + 1, XFS_AG_MAXLEVELS(mp)) + MIN(cl + 1, XFS_AG_MAXLEVELS(mp)))
|
|
|
|
#define XFS_MIN_FREELIST(a,mp) \
|
|
|
|
(XFS_MIN_FREELIST_RAW( \
|
2005-11-02 11:11:25 +07:00
|
|
|
be32_to_cpu((a)->agf_levels[XFS_BTNUM_BNOi]), \
|
|
|
|
be32_to_cpu((a)->agf_levels[XFS_BTNUM_CNTi]), mp))
|
2005-04-17 05:20:36 +07:00
|
|
|
#define XFS_MIN_FREELIST_PAG(pag,mp) \
|
2005-11-02 10:38:42 +07:00
|
|
|
(XFS_MIN_FREELIST_RAW( \
|
2009-02-09 14:37:39 +07:00
|
|
|
(unsigned int)(pag)->pagf_levels[XFS_BTNUM_BNOi], \
|
|
|
|
(unsigned int)(pag)->pagf_levels[XFS_BTNUM_CNTi], mp))
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2005-11-02 10:38:42 +07:00
|
|
|
#define XFS_AGB_TO_FSB(mp,agno,agbno) \
|
2005-04-17 05:20:36 +07:00
|
|
|
(((xfs_fsblock_t)(agno) << (mp)->m_sb.sb_agblklog) | (agbno))
|
2005-11-02 10:38:42 +07:00
|
|
|
#define XFS_FSB_TO_AGNO(mp,fsbno) \
|
2005-04-17 05:20:36 +07:00
|
|
|
((xfs_agnumber_t)((fsbno) >> (mp)->m_sb.sb_agblklog))
|
2005-11-02 10:38:42 +07:00
|
|
|
#define XFS_FSB_TO_AGBNO(mp,fsbno) \
|
2009-01-09 11:53:54 +07:00
|
|
|
((xfs_agblock_t)((fsbno) & xfs_mask32lo((mp)->m_sb.sb_agblklog)))
|
2005-11-02 10:38:42 +07:00
|
|
|
#define XFS_AGB_TO_DADDR(mp,agno,agbno) \
|
|
|
|
((xfs_daddr_t)XFS_FSB_TO_BB(mp, \
|
|
|
|
(xfs_fsblock_t)(agno) * (mp)->m_sb.sb_agblocks + (agbno)))
|
|
|
|
#define XFS_AG_DADDR(mp,agno,d) (XFS_AGB_TO_DADDR(mp, agno, 0) + (d))
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* For checking for bad ranges of xfs_daddr_t's, covering multiple
|
|
|
|
* allocation groups or a single xfs_daddr_t that's a superblock copy.
|
|
|
|
*/
|
|
|
|
#define XFS_AG_CHECK_DADDR(mp,d,len) \
|
|
|
|
((len) == 1 ? \
|
|
|
|
ASSERT((d) == XFS_SB_DADDR || \
|
2009-01-15 12:22:07 +07:00
|
|
|
xfs_daddr_to_agbno(mp, d) != XFS_SB_DADDR) : \
|
|
|
|
ASSERT(xfs_daddr_to_agno(mp, d) == \
|
|
|
|
xfs_daddr_to_agno(mp, (d) + (len) - 1)))
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
#endif /* __XFS_AG_H__ */
|