linux-stable/fs/xfs/xfs_alloc_btree.c
Dave Chinner ed3b4d6cdc 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>
2010-05-24 10:34:00 -05:00

504 lines
12 KiB
C

/*
* Copyright (c) 2000-2001,2005 Silicon Graphics, Inc.
* 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 as
* published by the Free Software Foundation.
*
* 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.
*
* 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
*/
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_types.h"
#include "xfs_bit.h"
#include "xfs_log.h"
#include "xfs_inum.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_bmap_btree.h"
#include "xfs_alloc_btree.h"
#include "xfs_ialloc_btree.h"
#include "xfs_dir2_sf.h"
#include "xfs_attr_sf.h"
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_btree.h"
#include "xfs_btree_trace.h"
#include "xfs_ialloc.h"
#include "xfs_alloc.h"
#include "xfs_error.h"
#include "xfs_trace.h"
STATIC struct xfs_btree_cur *
xfs_allocbt_dup_cursor(
struct xfs_btree_cur *cur)
{
return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp,
cur->bc_private.a.agbp, cur->bc_private.a.agno,
cur->bc_btnum);
}
STATIC void
xfs_allocbt_set_root(
struct xfs_btree_cur *cur,
union xfs_btree_ptr *ptr,
int inc)
{
struct xfs_buf *agbp = cur->bc_private.a.agbp;
struct xfs_agf *agf = XFS_BUF_TO_AGF(agbp);
xfs_agnumber_t seqno = be32_to_cpu(agf->agf_seqno);
int btnum = cur->bc_btnum;
struct xfs_perag *pag = xfs_perag_get(cur->bc_mp, seqno);
ASSERT(ptr->s != 0);
agf->agf_roots[btnum] = ptr->s;
be32_add_cpu(&agf->agf_levels[btnum], inc);
pag->pagf_levels[btnum] += inc;
xfs_perag_put(pag);
xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
}
STATIC int
xfs_allocbt_alloc_block(
struct xfs_btree_cur *cur,
union xfs_btree_ptr *start,
union xfs_btree_ptr *new,
int length,
int *stat)
{
int error;
xfs_agblock_t bno;
XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
/* Allocate the new block from the freelist. If we can't, give up. */
error = xfs_alloc_get_freelist(cur->bc_tp, cur->bc_private.a.agbp,
&bno, 1);
if (error) {
XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
return error;
}
if (bno == NULLAGBLOCK) {
XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
*stat = 0;
return 0;
}
xfs_trans_agbtree_delta(cur->bc_tp, 1);
new->s = cpu_to_be32(bno);
XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
*stat = 1;
return 0;
}
STATIC int
xfs_allocbt_free_block(
struct xfs_btree_cur *cur,
struct xfs_buf *bp)
{
struct xfs_buf *agbp = cur->bc_private.a.agbp;
struct xfs_agf *agf = XFS_BUF_TO_AGF(agbp);
xfs_agblock_t bno;
int error;
bno = xfs_daddr_to_agbno(cur->bc_mp, XFS_BUF_ADDR(bp));
error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1);
if (error)
return error;
/*
* Since blocks move to the free list without the coordination used in
* xfs_bmap_finish, we can't allow block to be available for
* reallocation and non-transaction writing (user data) until we know
* that the transaction that moved it to the free list is permanently
* on disk. We track the blocks by declaring these blocks as "busy";
* the busy list is maintained on a per-ag basis and each transaction
* records which entries should be removed when the iclog commits to
* disk. If a busy block is allocated, the iclog is pushed up to the
* LSN that freed the block.
*/
xfs_alloc_busy_insert(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1);
xfs_trans_agbtree_delta(cur->bc_tp, -1);
return 0;
}
/*
* Update the longest extent in the AGF
*/
STATIC void
xfs_allocbt_update_lastrec(
struct xfs_btree_cur *cur,
struct xfs_btree_block *block,
union xfs_btree_rec *rec,
int ptr,
int reason)
{
struct xfs_agf *agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
xfs_agnumber_t seqno = be32_to_cpu(agf->agf_seqno);
struct xfs_perag *pag;
__be32 len;
int numrecs;
ASSERT(cur->bc_btnum == XFS_BTNUM_CNT);
switch (reason) {
case LASTREC_UPDATE:
/*
* If this is the last leaf block and it's the last record,
* then update the size of the longest extent in the AG.
*/
if (ptr != xfs_btree_get_numrecs(block))
return;
len = rec->alloc.ar_blockcount;
break;
case LASTREC_INSREC:
if (be32_to_cpu(rec->alloc.ar_blockcount) <=
be32_to_cpu(agf->agf_longest))
return;
len = rec->alloc.ar_blockcount;
break;
case LASTREC_DELREC:
numrecs = xfs_btree_get_numrecs(block);
if (ptr <= numrecs)
return;
ASSERT(ptr == numrecs + 1);
if (numrecs) {
xfs_alloc_rec_t *rrp;
rrp = XFS_ALLOC_REC_ADDR(cur->bc_mp, block, numrecs);
len = rrp->ar_blockcount;
} else {
len = 0;
}
break;
default:
ASSERT(0);
return;
}
agf->agf_longest = len;
pag = xfs_perag_get(cur->bc_mp, seqno);
pag->pagf_longest = be32_to_cpu(len);
xfs_perag_put(pag);
xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp, XFS_AGF_LONGEST);
}
STATIC int
xfs_allocbt_get_minrecs(
struct xfs_btree_cur *cur,
int level)
{
return cur->bc_mp->m_alloc_mnr[level != 0];
}
STATIC int
xfs_allocbt_get_maxrecs(
struct xfs_btree_cur *cur,
int level)
{
return cur->bc_mp->m_alloc_mxr[level != 0];
}
STATIC void
xfs_allocbt_init_key_from_rec(
union xfs_btree_key *key,
union xfs_btree_rec *rec)
{
ASSERT(rec->alloc.ar_startblock != 0);
key->alloc.ar_startblock = rec->alloc.ar_startblock;
key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
}
STATIC void
xfs_allocbt_init_rec_from_key(
union xfs_btree_key *key,
union xfs_btree_rec *rec)
{
ASSERT(key->alloc.ar_startblock != 0);
rec->alloc.ar_startblock = key->alloc.ar_startblock;
rec->alloc.ar_blockcount = key->alloc.ar_blockcount;
}
STATIC void
xfs_allocbt_init_rec_from_cur(
struct xfs_btree_cur *cur,
union xfs_btree_rec *rec)
{
ASSERT(cur->bc_rec.a.ar_startblock != 0);
rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
}
STATIC void
xfs_allocbt_init_ptr_from_cur(
struct xfs_btree_cur *cur,
union xfs_btree_ptr *ptr)
{
struct xfs_agf *agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
ASSERT(cur->bc_private.a.agno == be32_to_cpu(agf->agf_seqno));
ASSERT(agf->agf_roots[cur->bc_btnum] != 0);
ptr->s = agf->agf_roots[cur->bc_btnum];
}
STATIC __int64_t
xfs_allocbt_key_diff(
struct xfs_btree_cur *cur,
union xfs_btree_key *key)
{
xfs_alloc_rec_incore_t *rec = &cur->bc_rec.a;
xfs_alloc_key_t *kp = &key->alloc;
__int64_t diff;
if (cur->bc_btnum == XFS_BTNUM_BNO) {
return (__int64_t)be32_to_cpu(kp->ar_startblock) -
rec->ar_startblock;
}
diff = (__int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount;
if (diff)
return diff;
return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
}
STATIC int
xfs_allocbt_kill_root(
struct xfs_btree_cur *cur,
struct xfs_buf *bp,
int level,
union xfs_btree_ptr *newroot)
{
int error;
XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
XFS_BTREE_STATS_INC(cur, killroot);
/*
* Update the root pointer, decreasing the level by 1 and then
* free the old root.
*/
xfs_allocbt_set_root(cur, newroot, -1);
error = xfs_allocbt_free_block(cur, bp);
if (error) {
XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
return error;
}
XFS_BTREE_STATS_INC(cur, free);
xfs_btree_setbuf(cur, level, NULL);
cur->bc_nlevels--;
XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
return 0;
}
#ifdef DEBUG
STATIC int
xfs_allocbt_keys_inorder(
struct xfs_btree_cur *cur,
union xfs_btree_key *k1,
union xfs_btree_key *k2)
{
if (cur->bc_btnum == XFS_BTNUM_BNO) {
return be32_to_cpu(k1->alloc.ar_startblock) <
be32_to_cpu(k2->alloc.ar_startblock);
} else {
return be32_to_cpu(k1->alloc.ar_blockcount) <
be32_to_cpu(k2->alloc.ar_blockcount) ||
(k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
be32_to_cpu(k1->alloc.ar_startblock) <
be32_to_cpu(k2->alloc.ar_startblock));
}
}
STATIC int
xfs_allocbt_recs_inorder(
struct xfs_btree_cur *cur,
union xfs_btree_rec *r1,
union xfs_btree_rec *r2)
{
if (cur->bc_btnum == XFS_BTNUM_BNO) {
return be32_to_cpu(r1->alloc.ar_startblock) +
be32_to_cpu(r1->alloc.ar_blockcount) <=
be32_to_cpu(r2->alloc.ar_startblock);
} else {
return be32_to_cpu(r1->alloc.ar_blockcount) <
be32_to_cpu(r2->alloc.ar_blockcount) ||
(r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
be32_to_cpu(r1->alloc.ar_startblock) <
be32_to_cpu(r2->alloc.ar_startblock));
}
}
#endif /* DEBUG */
#ifdef XFS_BTREE_TRACE
ktrace_t *xfs_allocbt_trace_buf;
STATIC void
xfs_allocbt_trace_enter(
struct xfs_btree_cur *cur,
const char *func,
char *s,
int type,
int line,
__psunsigned_t a0,
__psunsigned_t a1,
__psunsigned_t a2,
__psunsigned_t a3,
__psunsigned_t a4,
__psunsigned_t a5,
__psunsigned_t a6,
__psunsigned_t a7,
__psunsigned_t a8,
__psunsigned_t a9,
__psunsigned_t a10)
{
ktrace_enter(xfs_allocbt_trace_buf, (void *)(__psint_t)type,
(void *)func, (void *)s, NULL, (void *)cur,
(void *)a0, (void *)a1, (void *)a2, (void *)a3,
(void *)a4, (void *)a5, (void *)a6, (void *)a7,
(void *)a8, (void *)a9, (void *)a10);
}
STATIC void
xfs_allocbt_trace_cursor(
struct xfs_btree_cur *cur,
__uint32_t *s0,
__uint64_t *l0,
__uint64_t *l1)
{
*s0 = cur->bc_private.a.agno;
*l0 = cur->bc_rec.a.ar_startblock;
*l1 = cur->bc_rec.a.ar_blockcount;
}
STATIC void
xfs_allocbt_trace_key(
struct xfs_btree_cur *cur,
union xfs_btree_key *key,
__uint64_t *l0,
__uint64_t *l1)
{
*l0 = be32_to_cpu(key->alloc.ar_startblock);
*l1 = be32_to_cpu(key->alloc.ar_blockcount);
}
STATIC void
xfs_allocbt_trace_record(
struct xfs_btree_cur *cur,
union xfs_btree_rec *rec,
__uint64_t *l0,
__uint64_t *l1,
__uint64_t *l2)
{
*l0 = be32_to_cpu(rec->alloc.ar_startblock);
*l1 = be32_to_cpu(rec->alloc.ar_blockcount);
*l2 = 0;
}
#endif /* XFS_BTREE_TRACE */
static const struct xfs_btree_ops xfs_allocbt_ops = {
.rec_len = sizeof(xfs_alloc_rec_t),
.key_len = sizeof(xfs_alloc_key_t),
.dup_cursor = xfs_allocbt_dup_cursor,
.set_root = xfs_allocbt_set_root,
.kill_root = xfs_allocbt_kill_root,
.alloc_block = xfs_allocbt_alloc_block,
.free_block = xfs_allocbt_free_block,
.update_lastrec = xfs_allocbt_update_lastrec,
.get_minrecs = xfs_allocbt_get_minrecs,
.get_maxrecs = xfs_allocbt_get_maxrecs,
.init_key_from_rec = xfs_allocbt_init_key_from_rec,
.init_rec_from_key = xfs_allocbt_init_rec_from_key,
.init_rec_from_cur = xfs_allocbt_init_rec_from_cur,
.init_ptr_from_cur = xfs_allocbt_init_ptr_from_cur,
.key_diff = xfs_allocbt_key_diff,
#ifdef DEBUG
.keys_inorder = xfs_allocbt_keys_inorder,
.recs_inorder = xfs_allocbt_recs_inorder,
#endif
#ifdef XFS_BTREE_TRACE
.trace_enter = xfs_allocbt_trace_enter,
.trace_cursor = xfs_allocbt_trace_cursor,
.trace_key = xfs_allocbt_trace_key,
.trace_record = xfs_allocbt_trace_record,
#endif
};
/*
* Allocate a new allocation btree cursor.
*/
struct xfs_btree_cur * /* new alloc btree cursor */
xfs_allocbt_init_cursor(
struct xfs_mount *mp, /* file system mount point */
struct xfs_trans *tp, /* transaction pointer */
struct xfs_buf *agbp, /* buffer for agf structure */
xfs_agnumber_t agno, /* allocation group number */
xfs_btnum_t btnum) /* btree identifier */
{
struct xfs_agf *agf = XFS_BUF_TO_AGF(agbp);
struct xfs_btree_cur *cur;
ASSERT(btnum == XFS_BTNUM_BNO || btnum == XFS_BTNUM_CNT);
cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
cur->bc_tp = tp;
cur->bc_mp = mp;
cur->bc_nlevels = be32_to_cpu(agf->agf_levels[btnum]);
cur->bc_btnum = btnum;
cur->bc_blocklog = mp->m_sb.sb_blocklog;
cur->bc_ops = &xfs_allocbt_ops;
if (btnum == XFS_BTNUM_CNT)
cur->bc_flags = XFS_BTREE_LASTREC_UPDATE;
cur->bc_private.a.agbp = agbp;
cur->bc_private.a.agno = agno;
return cur;
}
/*
* Calculate number of records in an alloc btree block.
*/
int
xfs_allocbt_maxrecs(
struct xfs_mount *mp,
int blocklen,
int leaf)
{
blocklen -= XFS_ALLOC_BLOCK_LEN(mp);
if (leaf)
return blocklen / sizeof(xfs_alloc_rec_t);
return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t));
}