linux/fs/xfs/libxfs/xfs_bmap_btree.c

657 lines
17 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
* All Rights Reserved.
*/
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_shared.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_bit.h"
#include "xfs_mount.h"
#include "xfs_inode.h"
#include "xfs_trans.h"
#include "xfs_alloc.h"
#include "xfs_btree.h"
#include "xfs_bmap_btree.h"
#include "xfs_bmap.h"
#include "xfs_error.h"
#include "xfs_quota.h"
#include "xfs_trace.h"
xfs: add owner field to extent allocation and freeing For the rmap btree to work, we have to feed the extent owner information to the the allocation and freeing functions. This information is what will end up in the rmap btree that tracks allocated extents. While we technically don't need the owner information when freeing extents, passing it allows us to validate that the extent we are removing from the rmap btree actually belonged to the owner we expected it to belong to. We also define a special set of owner values for internal metadata that would otherwise have no owner. This allows us to tell the difference between metadata owned by different per-ag btrees, as well as static fs metadata (e.g. AG headers) and internal journal blocks. There are also a couple of special cases we need to take care of - during EFI recovery, we don't actually know who the original owner was, so we need to pass a wildcard to indicate that we aren't checking the owner for validity. We also need special handling in growfs, as we "free" the space in the last AG when extending it, but because it's new space it has no actual owner... While touching the xfs_bmap_add_free() function, re-order the parameters to put the struct xfs_mount first. Extend the owner field to include both the owner type and some sort of index within the owner. The index field will be used to support reverse mappings when reflink is enabled. When we're freeing extents from an EFI, we don't have the owner information available (rmap updates have their own redo items). xfs_free_extent therefore doesn't need to do an rmap update. Make sure that the log replay code signals this correctly. This is based upon a patch originally from Dave Chinner. It has been extended to add more owner information with the intent of helping recovery operations when things go wrong (e.g. offset of user data block in a file). [dchinner: de-shout the xfs_rmap_*_owner helpers] [darrick: minor style fixes suggested by Christoph Hellwig] Signed-off-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-08-03 01:33:42 +00:00
#include "xfs_rmap.h"
/*
* Convert on-disk form of btree root to in-memory form.
*/
void
xfs_bmdr_to_bmbt(
struct xfs_inode *ip,
xfs_bmdr_block_t *dblock,
int dblocklen,
struct xfs_btree_block *rblock,
int rblocklen)
{
struct xfs_mount *mp = ip->i_mount;
int dmxr;
xfs_bmbt_key_t *fkp;
__be64 *fpp;
xfs_bmbt_key_t *tkp;
__be64 *tpp;
xfs_btree_init_block_int(mp, rblock, XFS_BUF_DADDR_NULL,
XFS_BTNUM_BMAP, 0, 0, ip->i_ino,
XFS_BTREE_LONG_PTRS);
rblock->bb_level = dblock->bb_level;
ASSERT(be16_to_cpu(rblock->bb_level) > 0);
rblock->bb_numrecs = dblock->bb_numrecs;
dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
fkp = XFS_BMDR_KEY_ADDR(dblock, 1);
tkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
fpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
tpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
dmxr = be16_to_cpu(dblock->bb_numrecs);
memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
}
void
xfs: use a b+tree for the in-core extent list Replace the current linear list and the indirection array for the in-core extent list with a b+tree to avoid the need for larger memory allocations for the indirection array when lots of extents are present. The current extent list implementations leads to heavy pressure on the memory allocator when modifying files with a high extent count, and can lead to high latencies because of that. The replacement is a b+tree with a few quirks. The leaf nodes directly store the extent record in two u64 values. The encoding is a little bit different from the existing in-core extent records so that the start offset and length which are required for lookups can be retreived with simple mask operations. The inner nodes store a 64-bit key containing the start offset in the first half of the node, and the pointers to the next lower level in the second half. In either case we walk the node from the beginninig to the end and do a linear search, as that is more efficient for the low number of cache lines touched during a search (2 for the inner nodes, 4 for the leaf nodes) than a binary search. We store termination markers (zero length for the leaf nodes, an otherwise impossible high bit for the inner nodes) to terminate the key list / records instead of storing a count to use the available cache lines as efficiently as possible. One quirk of the algorithm is that while we normally split a node half and half like usual btree implementations we just spill over entries added at the very end of the list to a new node on its own. This means we get a 100% fill grade for the common cases of bulk insertion when reading an inode into memory, and when only sequentially appending to a file. The downside is a slightly higher chance of splits on the first random insertions. Both insert and removal manually recurse into the lower levels, but the bulk deletion of the whole tree is still implemented as a recursive function call, although one limited by the overall depth and with very little stack usage in every iteration. For the first few extents we dynamically grow the list from a single extent to the next powers of two until we have a first full leaf block and that building the actual tree. The code started out based on the generic lib/btree.c code from Joern Engel based on earlier work from Peter Zijlstra, but has since been rewritten beyond recognition. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2017-11-03 17:34:46 +00:00
xfs_bmbt_disk_get_all(
const struct xfs_bmbt_rec *rec,
xfs: use a b+tree for the in-core extent list Replace the current linear list and the indirection array for the in-core extent list with a b+tree to avoid the need for larger memory allocations for the indirection array when lots of extents are present. The current extent list implementations leads to heavy pressure on the memory allocator when modifying files with a high extent count, and can lead to high latencies because of that. The replacement is a b+tree with a few quirks. The leaf nodes directly store the extent record in two u64 values. The encoding is a little bit different from the existing in-core extent records so that the start offset and length which are required for lookups can be retreived with simple mask operations. The inner nodes store a 64-bit key containing the start offset in the first half of the node, and the pointers to the next lower level in the second half. In either case we walk the node from the beginninig to the end and do a linear search, as that is more efficient for the low number of cache lines touched during a search (2 for the inner nodes, 4 for the leaf nodes) than a binary search. We store termination markers (zero length for the leaf nodes, an otherwise impossible high bit for the inner nodes) to terminate the key list / records instead of storing a count to use the available cache lines as efficiently as possible. One quirk of the algorithm is that while we normally split a node half and half like usual btree implementations we just spill over entries added at the very end of the list to a new node on its own. This means we get a 100% fill grade for the common cases of bulk insertion when reading an inode into memory, and when only sequentially appending to a file. The downside is a slightly higher chance of splits on the first random insertions. Both insert and removal manually recurse into the lower levels, but the bulk deletion of the whole tree is still implemented as a recursive function call, although one limited by the overall depth and with very little stack usage in every iteration. For the first few extents we dynamically grow the list from a single extent to the next powers of two until we have a first full leaf block and that building the actual tree. The code started out based on the generic lib/btree.c code from Joern Engel based on earlier work from Peter Zijlstra, but has since been rewritten beyond recognition. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2017-11-03 17:34:46 +00:00
struct xfs_bmbt_irec *irec)
{
xfs: use a b+tree for the in-core extent list Replace the current linear list and the indirection array for the in-core extent list with a b+tree to avoid the need for larger memory allocations for the indirection array when lots of extents are present. The current extent list implementations leads to heavy pressure on the memory allocator when modifying files with a high extent count, and can lead to high latencies because of that. The replacement is a b+tree with a few quirks. The leaf nodes directly store the extent record in two u64 values. The encoding is a little bit different from the existing in-core extent records so that the start offset and length which are required for lookups can be retreived with simple mask operations. The inner nodes store a 64-bit key containing the start offset in the first half of the node, and the pointers to the next lower level in the second half. In either case we walk the node from the beginninig to the end and do a linear search, as that is more efficient for the low number of cache lines touched during a search (2 for the inner nodes, 4 for the leaf nodes) than a binary search. We store termination markers (zero length for the leaf nodes, an otherwise impossible high bit for the inner nodes) to terminate the key list / records instead of storing a count to use the available cache lines as efficiently as possible. One quirk of the algorithm is that while we normally split a node half and half like usual btree implementations we just spill over entries added at the very end of the list to a new node on its own. This means we get a 100% fill grade for the common cases of bulk insertion when reading an inode into memory, and when only sequentially appending to a file. The downside is a slightly higher chance of splits on the first random insertions. Both insert and removal manually recurse into the lower levels, but the bulk deletion of the whole tree is still implemented as a recursive function call, although one limited by the overall depth and with very little stack usage in every iteration. For the first few extents we dynamically grow the list from a single extent to the next powers of two until we have a first full leaf block and that building the actual tree. The code started out based on the generic lib/btree.c code from Joern Engel based on earlier work from Peter Zijlstra, but has since been rewritten beyond recognition. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2017-11-03 17:34:46 +00:00
uint64_t l0 = get_unaligned_be64(&rec->l0);
uint64_t l1 = get_unaligned_be64(&rec->l1);
xfs: use a b+tree for the in-core extent list Replace the current linear list and the indirection array for the in-core extent list with a b+tree to avoid the need for larger memory allocations for the indirection array when lots of extents are present. The current extent list implementations leads to heavy pressure on the memory allocator when modifying files with a high extent count, and can lead to high latencies because of that. The replacement is a b+tree with a few quirks. The leaf nodes directly store the extent record in two u64 values. The encoding is a little bit different from the existing in-core extent records so that the start offset and length which are required for lookups can be retreived with simple mask operations. The inner nodes store a 64-bit key containing the start offset in the first half of the node, and the pointers to the next lower level in the second half. In either case we walk the node from the beginninig to the end and do a linear search, as that is more efficient for the low number of cache lines touched during a search (2 for the inner nodes, 4 for the leaf nodes) than a binary search. We store termination markers (zero length for the leaf nodes, an otherwise impossible high bit for the inner nodes) to terminate the key list / records instead of storing a count to use the available cache lines as efficiently as possible. One quirk of the algorithm is that while we normally split a node half and half like usual btree implementations we just spill over entries added at the very end of the list to a new node on its own. This means we get a 100% fill grade for the common cases of bulk insertion when reading an inode into memory, and when only sequentially appending to a file. The downside is a slightly higher chance of splits on the first random insertions. Both insert and removal manually recurse into the lower levels, but the bulk deletion of the whole tree is still implemented as a recursive function call, although one limited by the overall depth and with very little stack usage in every iteration. For the first few extents we dynamically grow the list from a single extent to the next powers of two until we have a first full leaf block and that building the actual tree. The code started out based on the generic lib/btree.c code from Joern Engel based on earlier work from Peter Zijlstra, but has since been rewritten beyond recognition. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2017-11-03 17:34:46 +00:00
irec->br_startoff = (l0 & xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
irec->br_startblock = ((l0 & xfs_mask64lo(9)) << 43) | (l1 >> 21);
irec->br_blockcount = l1 & xfs_mask64lo(21);
if (l0 >> (64 - BMBT_EXNTFLAG_BITLEN))
irec->br_state = XFS_EXT_UNWRITTEN;
else
irec->br_state = XFS_EXT_NORM;
}
/*
* Extract the blockcount field from an on disk bmap extent record.
*/
xfs_filblks_t
xfs_bmbt_disk_get_blockcount(
const struct xfs_bmbt_rec *r)
{
return (xfs_filblks_t)(be64_to_cpu(r->l1) & xfs_mask64lo(21));
}
/*
* Extract the startoff field from a disk format bmap extent record.
*/
xfs_fileoff_t
xfs_bmbt_disk_get_startoff(
const struct xfs_bmbt_rec *r)
{
return ((xfs_fileoff_t)be64_to_cpu(r->l0) &
xfs_mask64lo(64 - BMBT_EXNTFLAG_BITLEN)) >> 9;
}
/*
* Set all the fields in a bmap extent record from the uncompressed form.
*/
void
xfs_bmbt_disk_set_all(
struct xfs_bmbt_rec *r,
struct xfs_bmbt_irec *s)
{
int extent_flag = (s->br_state != XFS_EXT_NORM);
ASSERT(s->br_state == XFS_EXT_NORM || s->br_state == XFS_EXT_UNWRITTEN);
ASSERT(!(s->br_startoff & xfs_mask64hi(64-BMBT_STARTOFF_BITLEN)));
ASSERT(!(s->br_blockcount & xfs_mask64hi(64-BMBT_BLOCKCOUNT_BITLEN)));
ASSERT(!(s->br_startblock & xfs_mask64hi(64-BMBT_STARTBLOCK_BITLEN)));
put_unaligned_be64(
((xfs_bmbt_rec_base_t)extent_flag << 63) |
((xfs_bmbt_rec_base_t)s->br_startoff << 9) |
((xfs_bmbt_rec_base_t)s->br_startblock >> 43), &r->l0);
put_unaligned_be64(
((xfs_bmbt_rec_base_t)s->br_startblock << 21) |
((xfs_bmbt_rec_base_t)s->br_blockcount &
(xfs_bmbt_rec_base_t)xfs_mask64lo(21)), &r->l1);
}
/*
* Convert in-memory form of btree root to on-disk form.
*/
void
xfs_bmbt_to_bmdr(
struct xfs_mount *mp,
struct xfs_btree_block *rblock,
int rblocklen,
xfs_bmdr_block_t *dblock,
int dblocklen)
{
int dmxr;
xfs_bmbt_key_t *fkp;
__be64 *fpp;
xfs_bmbt_key_t *tkp;
__be64 *tpp;
if (xfs_sb_version_hascrc(&mp->m_sb)) {
ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_CRC_MAGIC));
ASSERT(uuid_equal(&rblock->bb_u.l.bb_uuid,
&mp->m_sb.sb_meta_uuid));
ASSERT(rblock->bb_u.l.bb_blkno ==
cpu_to_be64(XFS_BUF_DADDR_NULL));
} else
ASSERT(rblock->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC));
ASSERT(rblock->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK));
ASSERT(rblock->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK));
ASSERT(rblock->bb_level != 0);
dblock->bb_level = rblock->bb_level;
dblock->bb_numrecs = rblock->bb_numrecs;
dmxr = xfs_bmdr_maxrecs(dblocklen, 0);
fkp = XFS_BMBT_KEY_ADDR(mp, rblock, 1);
tkp = XFS_BMDR_KEY_ADDR(dblock, 1);
fpp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, rblocklen);
tpp = XFS_BMDR_PTR_ADDR(dblock, 1, dmxr);
dmxr = be16_to_cpu(dblock->bb_numrecs);
memcpy(tkp, fkp, sizeof(*fkp) * dmxr);
memcpy(tpp, fpp, sizeof(*fpp) * dmxr);
}
STATIC struct xfs_btree_cur *
xfs_bmbt_dup_cursor(
struct xfs_btree_cur *cur)
{
struct xfs_btree_cur *new;
new = xfs_bmbt_init_cursor(cur->bc_mp, cur->bc_tp,
cur->bc_ino.ip, cur->bc_ino.whichfork);
/*
* Copy the firstblock, dfops, and flags values,
* since init cursor doesn't get them.
*/
new->bc_ino.flags = cur->bc_ino.flags;
return new;
}
STATIC void
xfs_bmbt_update_cursor(
struct xfs_btree_cur *src,
struct xfs_btree_cur *dst)
{
ASSERT((dst->bc_tp->t_firstblock != NULLFSBLOCK) ||
(dst->bc_ino.ip->i_diflags & XFS_DIFLAG_REALTIME));
dst->bc_ino.allocated += src->bc_ino.allocated;
dst->bc_tp->t_firstblock = src->bc_tp->t_firstblock;
src->bc_ino.allocated = 0;
}
STATIC int
xfs_bmbt_alloc_block(
struct xfs_btree_cur *cur,
union xfs_btree_ptr *start,
union xfs_btree_ptr *new,
int *stat)
{
xfs_alloc_arg_t args; /* block allocation args */
int error; /* error return value */
memset(&args, 0, sizeof(args));
args.tp = cur->bc_tp;
args.mp = cur->bc_mp;
args.fsbno = cur->bc_tp->t_firstblock;
xfs_rmap_ino_bmbt_owner(&args.oinfo, cur->bc_ino.ip->i_ino,
cur->bc_ino.whichfork);
if (args.fsbno == NULLFSBLOCK) {
args.fsbno = be64_to_cpu(start->l);
args.type = XFS_ALLOCTYPE_START_BNO;
/*
* Make sure there is sufficient room left in the AG to
* complete a full tree split for an extent insert. If
* we are converting the middle part of an extent then
* we may need space for two tree splits.
*
* We are relying on the caller to make the correct block
* reservation for this operation to succeed. If the
* reservation amount is insufficient then we may fail a
* block allocation here and corrupt the filesystem.
*/
args.minleft = args.tp->t_blk_res;
} else if (cur->bc_tp->t_flags & XFS_TRANS_LOWMODE) {
args.type = XFS_ALLOCTYPE_START_BNO;
} else {
args.type = XFS_ALLOCTYPE_NEAR_BNO;
}
args.minlen = args.maxlen = args.prod = 1;
args.wasdel = cur->bc_ino.flags & XFS_BTCUR_BMBT_WASDEL;
if (!args.wasdel && args.tp->t_blk_res == 0) {
error = -ENOSPC;
goto error0;
}
error = xfs_alloc_vextent(&args);
if (error)
goto error0;
if (args.fsbno == NULLFSBLOCK && args.minleft) {
/*
* Could not find an AG with enough free space to satisfy
* a full btree split. Try again and if
* successful activate the lowspace algorithm.
*/
args.fsbno = 0;
args.type = XFS_ALLOCTYPE_FIRST_AG;
error = xfs_alloc_vextent(&args);
if (error)
goto error0;
cur->bc_tp->t_flags |= XFS_TRANS_LOWMODE;
}
if (WARN_ON_ONCE(args.fsbno == NULLFSBLOCK)) {
*stat = 0;
return 0;
}
ASSERT(args.len == 1);
cur->bc_tp->t_firstblock = args.fsbno;
cur->bc_ino.allocated++;
cur->bc_ino.ip->i_nblocks++;
xfs_trans_log_inode(args.tp, cur->bc_ino.ip, XFS_ILOG_CORE);
xfs_trans_mod_dquot_byino(args.tp, cur->bc_ino.ip,
XFS_TRANS_DQ_BCOUNT, 1L);
new->l = cpu_to_be64(args.fsbno);
*stat = 1;
return 0;
error0:
return error;
}
STATIC int
xfs_bmbt_free_block(
struct xfs_btree_cur *cur,
struct xfs_buf *bp)
{
struct xfs_mount *mp = cur->bc_mp;
struct xfs_inode *ip = cur->bc_ino.ip;
struct xfs_trans *tp = cur->bc_tp;
xfs_fsblock_t fsbno = XFS_DADDR_TO_FSB(mp, XFS_BUF_ADDR(bp));
xfs: add owner field to extent allocation and freeing For the rmap btree to work, we have to feed the extent owner information to the the allocation and freeing functions. This information is what will end up in the rmap btree that tracks allocated extents. While we technically don't need the owner information when freeing extents, passing it allows us to validate that the extent we are removing from the rmap btree actually belonged to the owner we expected it to belong to. We also define a special set of owner values for internal metadata that would otherwise have no owner. This allows us to tell the difference between metadata owned by different per-ag btrees, as well as static fs metadata (e.g. AG headers) and internal journal blocks. There are also a couple of special cases we need to take care of - during EFI recovery, we don't actually know who the original owner was, so we need to pass a wildcard to indicate that we aren't checking the owner for validity. We also need special handling in growfs, as we "free" the space in the last AG when extending it, but because it's new space it has no actual owner... While touching the xfs_bmap_add_free() function, re-order the parameters to put the struct xfs_mount first. Extend the owner field to include both the owner type and some sort of index within the owner. The index field will be used to support reverse mappings when reflink is enabled. When we're freeing extents from an EFI, we don't have the owner information available (rmap updates have their own redo items). xfs_free_extent therefore doesn't need to do an rmap update. Make sure that the log replay code signals this correctly. This is based upon a patch originally from Dave Chinner. It has been extended to add more owner information with the intent of helping recovery operations when things go wrong (e.g. offset of user data block in a file). [dchinner: de-shout the xfs_rmap_*_owner helpers] [darrick: minor style fixes suggested by Christoph Hellwig] Signed-off-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-08-03 01:33:42 +00:00
struct xfs_owner_info oinfo;
xfs_rmap_ino_bmbt_owner(&oinfo, ip->i_ino, cur->bc_ino.whichfork);
xfs_bmap_add_free(cur->bc_tp, fsbno, 1, &oinfo);
ip->i_nblocks--;
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, -1L);
return 0;
}
STATIC int
xfs_bmbt_get_minrecs(
struct xfs_btree_cur *cur,
int level)
{
if (level == cur->bc_nlevels - 1) {
struct xfs_ifork *ifp;
ifp = XFS_IFORK_PTR(cur->bc_ino.ip,
cur->bc_ino.whichfork);
return xfs_bmbt_maxrecs(cur->bc_mp,
ifp->if_broot_bytes, level == 0) / 2;
}
return cur->bc_mp->m_bmap_dmnr[level != 0];
}
int
xfs_bmbt_get_maxrecs(
struct xfs_btree_cur *cur,
int level)
{
if (level == cur->bc_nlevels - 1) {
struct xfs_ifork *ifp;
ifp = XFS_IFORK_PTR(cur->bc_ino.ip,
cur->bc_ino.whichfork);
return xfs_bmbt_maxrecs(cur->bc_mp,
ifp->if_broot_bytes, level == 0);
}
return cur->bc_mp->m_bmap_dmxr[level != 0];
}
/*
* Get the maximum records we could store in the on-disk format.
*
* For non-root nodes this is equivalent to xfs_bmbt_get_maxrecs, but
* for the root node this checks the available space in the dinode fork
* so that we can resize the in-memory buffer to match it. After a
* resize to the maximum size this function returns the same value
* as xfs_bmbt_get_maxrecs for the root node, too.
*/
STATIC int
xfs_bmbt_get_dmaxrecs(
struct xfs_btree_cur *cur,
int level)
{
if (level != cur->bc_nlevels - 1)
return cur->bc_mp->m_bmap_dmxr[level != 0];
return xfs_bmdr_maxrecs(cur->bc_ino.forksize, level == 0);
}
STATIC void
xfs_bmbt_init_key_from_rec(
union xfs_btree_key *key,
const union xfs_btree_rec *rec)
{
key->bmbt.br_startoff =
cpu_to_be64(xfs_bmbt_disk_get_startoff(&rec->bmbt));
}
STATIC void
xfs_bmbt_init_high_key_from_rec(
union xfs_btree_key *key,
const union xfs_btree_rec *rec)
{
key->bmbt.br_startoff = cpu_to_be64(
xfs_bmbt_disk_get_startoff(&rec->bmbt) +
xfs_bmbt_disk_get_blockcount(&rec->bmbt) - 1);
}
STATIC void
xfs_bmbt_init_rec_from_cur(
struct xfs_btree_cur *cur,
union xfs_btree_rec *rec)
{
xfs_bmbt_disk_set_all(&rec->bmbt, &cur->bc_rec.b);
}
STATIC void
xfs_bmbt_init_ptr_from_cur(
struct xfs_btree_cur *cur,
union xfs_btree_ptr *ptr)
{
ptr->l = 0;
}
STATIC int64_t
xfs_bmbt_key_diff(
struct xfs_btree_cur *cur,
const union xfs_btree_key *key)
{
return (int64_t)be64_to_cpu(key->bmbt.br_startoff) -
cur->bc_rec.b.br_startoff;
}
STATIC int64_t
xfs_bmbt_diff_two_keys(
struct xfs_btree_cur *cur,
const union xfs_btree_key *k1,
const union xfs_btree_key *k2)
{
uint64_t a = be64_to_cpu(k1->bmbt.br_startoff);
uint64_t b = be64_to_cpu(k2->bmbt.br_startoff);
/*
* Note: This routine previously casted a and b to int64 and subtracted
* them to generate a result. This lead to problems if b was the
* "maximum" key value (all ones) being signed incorrectly, hence this
* somewhat less efficient version.
*/
if (a > b)
return 1;
if (b > a)
return -1;
return 0;
}
static xfs_failaddr_t
xfs_bmbt_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_mount;
struct xfs_btree_block *block = XFS_BUF_TO_BLOCK(bp);
xfs_failaddr_t fa;
unsigned int level;
if (!xfs_verify_magic(bp, block->bb_magic))
return __this_address;
if (xfs_sb_version_hascrc(&mp->m_sb)) {
/*
* XXX: need a better way of verifying the owner here. Right now
* just make sure there has been one set.
*/
fa = xfs_btree_lblock_v5hdr_verify(bp, XFS_RMAP_OWN_UNKNOWN);
if (fa)
return fa;
}
/*
* numrecs and level verification.
*
* We don't know what fork we belong to, so just verify that the level
* is less than the maximum of the two. Later checks will be more
* precise.
*/
level = be16_to_cpu(block->bb_level);
if (level > max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]))
return __this_address;
return xfs_btree_lblock_verify(bp, mp->m_bmap_dmxr[level != 0]);
}
static void
xfs_bmbt_read_verify(
struct xfs_buf *bp)
{
xfs_failaddr_t fa;
if (!xfs_btree_lblock_verify_crc(bp))
xfs_verifier_error(bp, -EFSBADCRC, __this_address);
else {
fa = xfs_bmbt_verify(bp);
if (fa)
xfs_verifier_error(bp, -EFSCORRUPTED, fa);
}
if (bp->b_error)
trace_xfs_btree_corrupt(bp, _RET_IP_);
}
static void
xfs_bmbt_write_verify(
struct xfs_buf *bp)
{
xfs_failaddr_t fa;
fa = xfs_bmbt_verify(bp);
if (fa) {
trace_xfs_btree_corrupt(bp, _RET_IP_);
xfs_verifier_error(bp, -EFSCORRUPTED, fa);
return;
}
xfs_btree_lblock_calc_crc(bp);
}
const struct xfs_buf_ops xfs_bmbt_buf_ops = {
.name = "xfs_bmbt",
.magic = { cpu_to_be32(XFS_BMAP_MAGIC),
cpu_to_be32(XFS_BMAP_CRC_MAGIC) },
.verify_read = xfs_bmbt_read_verify,
.verify_write = xfs_bmbt_write_verify,
.verify_struct = xfs_bmbt_verify,
};
STATIC int
xfs_bmbt_keys_inorder(
struct xfs_btree_cur *cur,
const union xfs_btree_key *k1,
const union xfs_btree_key *k2)
{
return be64_to_cpu(k1->bmbt.br_startoff) <
be64_to_cpu(k2->bmbt.br_startoff);
}
STATIC int
xfs_bmbt_recs_inorder(
struct xfs_btree_cur *cur,
const union xfs_btree_rec *r1,
const union xfs_btree_rec *r2)
{
return xfs_bmbt_disk_get_startoff(&r1->bmbt) +
xfs_bmbt_disk_get_blockcount(&r1->bmbt) <=
xfs_bmbt_disk_get_startoff(&r2->bmbt);
}
static const struct xfs_btree_ops xfs_bmbt_ops = {
.rec_len = sizeof(xfs_bmbt_rec_t),
.key_len = sizeof(xfs_bmbt_key_t),
.dup_cursor = xfs_bmbt_dup_cursor,
.update_cursor = xfs_bmbt_update_cursor,
.alloc_block = xfs_bmbt_alloc_block,
.free_block = xfs_bmbt_free_block,
.get_maxrecs = xfs_bmbt_get_maxrecs,
.get_minrecs = xfs_bmbt_get_minrecs,
.get_dmaxrecs = xfs_bmbt_get_dmaxrecs,
.init_key_from_rec = xfs_bmbt_init_key_from_rec,
.init_high_key_from_rec = xfs_bmbt_init_high_key_from_rec,
.init_rec_from_cur = xfs_bmbt_init_rec_from_cur,
.init_ptr_from_cur = xfs_bmbt_init_ptr_from_cur,
.key_diff = xfs_bmbt_key_diff,
.diff_two_keys = xfs_bmbt_diff_two_keys,
.buf_ops = &xfs_bmbt_buf_ops,
.keys_inorder = xfs_bmbt_keys_inorder,
.recs_inorder = xfs_bmbt_recs_inorder,
};
/*
* Allocate a new bmap btree cursor.
*/
struct xfs_btree_cur * /* new bmap btree cursor */
xfs_bmbt_init_cursor(
struct xfs_mount *mp, /* file system mount point */
struct xfs_trans *tp, /* transaction pointer */
struct xfs_inode *ip, /* inode owning the btree */
int whichfork) /* data or attr fork */
{
struct xfs_ifork *ifp = XFS_IFORK_PTR(ip, whichfork);
struct xfs_btree_cur *cur;
ASSERT(whichfork != XFS_COW_FORK);
cur = kmem_cache_zalloc(xfs_btree_cur_zone, GFP_NOFS | __GFP_NOFAIL);
cur->bc_tp = tp;
cur->bc_mp = mp;
cur->bc_nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
cur->bc_btnum = XFS_BTNUM_BMAP;
cur->bc_blocklog = mp->m_sb.sb_blocklog;
cur->bc_statoff = XFS_STATS_CALC_INDEX(xs_bmbt_2);
cur->bc_ops = &xfs_bmbt_ops;
cur->bc_flags = XFS_BTREE_LONG_PTRS | XFS_BTREE_ROOT_IN_INODE;
if (xfs_sb_version_hascrc(&mp->m_sb))
cur->bc_flags |= XFS_BTREE_CRC_BLOCKS;
cur->bc_ino.forksize = XFS_IFORK_SIZE(ip, whichfork);
cur->bc_ino.ip = ip;
cur->bc_ino.allocated = 0;
cur->bc_ino.flags = 0;
cur->bc_ino.whichfork = whichfork;
return cur;
}
/*
* Calculate number of records in a bmap btree block.
*/
int
xfs_bmbt_maxrecs(
struct xfs_mount *mp,
int blocklen,
int leaf)
{
blocklen -= XFS_BMBT_BLOCK_LEN(mp);
if (leaf)
return blocklen / sizeof(xfs_bmbt_rec_t);
return blocklen / (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t));
}
/*
* Calculate number of records in a bmap btree inode root.
*/
int
xfs_bmdr_maxrecs(
int blocklen,
int leaf)
{
blocklen -= sizeof(xfs_bmdr_block_t);
if (leaf)
return blocklen / sizeof(xfs_bmdr_rec_t);
return blocklen / (sizeof(xfs_bmdr_key_t) + sizeof(xfs_bmdr_ptr_t));
}
xfs: swap extents operations for CRC filesystems For CRC enabled filesystems, we can't just swap inode forks from one inode to another when defragmenting a file - the blocks in the inode fork bmap btree contain pointers back to the owner inode. Hence if we are to swap the inode forks we have to atomically modify every block in the btree during the transaction. We are doing an entire fork swap here, so we could create a new transaction item type that indicates we are changing the owner of a certain structure from one value to another. If we combine this with ordered buffer logging to modify all the buffers in the tree, then we can change the buffers in the tree without needing log space for the operation. However, this then requires log recovery to perform the modification of the owner information of the objects/structures in question. This does introduce some interesting ordering details into recovery: we have to make sure that the owner change replay occurs after the change that moves the objects is made, not before. Hence we can't use a separate log item for this as we have no guarantee of strict ordering between multiple items in the log due to the relogging action of asynchronous transaction commits. Hence there is no "generic" method we can use for changing the ownership of arbitrary metadata structures. For inode forks, however, there is a simple method of communicating that the fork contents need the owner rewritten - we can pass a inode log format flag for the fork for the transaction that does a fork swap. This flag will then follow the inode fork through relogging actions so when the swap actually gets replayed the ownership can be changed immediately by log recovery. So that gives us a simple method of "whole fork" exchange between two inodes. This is relatively simple to implement, so it makes sense to do this as an initial implementation to support xfs_fsr on CRC enabled filesytems in the same manner as we do on existing filesystems. This commit introduces the swapext driven functionality, the recovery functionality will be in a separate patch. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:44 +00:00
/*
* Change the owner of a btree format fork fo the inode passed in. Change it to
* the owner of that is passed in so that we can change owners before or after
* we switch forks between inodes. The operation that the caller is doing will
* determine whether is needs to change owner before or after the switch.
*
xfs: recovery of swap extents operations for CRC filesystems This is the recovery side of the btree block owner change operation performed by swapext on CRC enabled filesystems. We detect that an owner change is needed by the flag that has been placed on the inode log format flag field. Because the inode recovery is being replayed after the buffers that make up the BMBT in the given checkpoint, we can walk all the buffers and directly modify them when we see the flag set on an inode. Because the inode can be relogged and hence present in multiple chekpoints with the "change owner" flag set, we could do multiple passes across the inode to do this change. While this isn't optimal, we can't directly ignore the flag as there may be multiple independent swap extent operations being replayed on the same inode in different checkpoints so we can't ignore them. Further, because the owner change operation uses ordered buffers, we might have buffers that are newer on disk than the current checkpoint and so already have the owner changed in them. Hence we cannot just peek at a buffer in the tree and check that it has the correct owner and assume that the change was completed. So, for the moment just brute force the owner change every time we see an inode with the flag set. Note that we have to be careful here because the owner of the buffers may point to either the old owner or the new owner. Currently the verifier can't verify the owner directly, so there is no failure case here right now. If we verify the owner exactly in future, then we'll have to take this into account. This was tested in terms of normal operation via xfstests - all of the fsr tests now pass without failure. however, we really need to modify xfs/227 to stress v3 inodes correctly to ensure we fully cover this case for v5 filesystems. In terms of recovery testing, I used a hacked version of xfs_fsr that held the temp inode open for a few seconds before exiting so that the filesystem could be shut down with an open owner change recovery flags set on at least the temp inode. fsr leaves the temp inode unlinked and in btree format, so this was necessary for the owner change to be reliably replayed. logprint confirmed the tmp inode in the log had the correct flag set: INO: cnt:3 total:3 a:0x69e9e0 len:56 a:0x69ea20 len:176 a:0x69eae0 len:88 INODE: #regs:3 ino:0x44 flags:0x209 dsize:88 ^^^^^ 0x200 is set, indicating a data fork owner change needed to be replayed on inode 0x44. A printk in the revoery code confirmed that the inode change was recovered: XFS (vdc): Mounting Filesystem XFS (vdc): Starting recovery (logdev: internal) recovering owner change ino 0x44 XFS (vdc): Version 5 superblock detected. This kernel L support enabled! Use of these features in this kernel is at your own risk! XFS (vdc): Ending recovery (logdev: internal) The script used to test this was: $ cat ./recovery-fsr.sh #!/bin/bash dev=/dev/vdc mntpt=/mnt/scratch testfile=$mntpt/testfile umount $mntpt mkfs.xfs -f -m crc=1 $dev mount $dev $mntpt chmod 777 $mntpt for i in `seq 10000 -1 0`; do xfs_io -f -d -c "pwrite $(($i * 4096)) 4096" $testfile > /dev/null 2>&1 done xfs_bmap -vp $testfile |head -20 xfs_fsr -d -v $testfile & sleep 10 /home/dave/src/xfstests-dev/src/godown -f $mntpt wait umount $mntpt xfs_logprint -t $dev |tail -20 time mount $dev $mntpt xfs_bmap -vp $testfile umount $mntpt $ Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:45 +00:00
* For demand paged transactional modification, the fork switch should be done
* after reading in all the blocks, modifying them and pinning them in the
* transaction. For modification when the buffers are already pinned in memory,
* the fork switch can be done before changing the owner as we won't need to
* validate the owner until the btree buffers are unpinned and writes can occur
* again.
*
* For recovery based ownership change, there is no transactional context and
* so a buffer list must be supplied so that we can record the buffers that we
* modified for the caller to issue IO on.
xfs: swap extents operations for CRC filesystems For CRC enabled filesystems, we can't just swap inode forks from one inode to another when defragmenting a file - the blocks in the inode fork bmap btree contain pointers back to the owner inode. Hence if we are to swap the inode forks we have to atomically modify every block in the btree during the transaction. We are doing an entire fork swap here, so we could create a new transaction item type that indicates we are changing the owner of a certain structure from one value to another. If we combine this with ordered buffer logging to modify all the buffers in the tree, then we can change the buffers in the tree without needing log space for the operation. However, this then requires log recovery to perform the modification of the owner information of the objects/structures in question. This does introduce some interesting ordering details into recovery: we have to make sure that the owner change replay occurs after the change that moves the objects is made, not before. Hence we can't use a separate log item for this as we have no guarantee of strict ordering between multiple items in the log due to the relogging action of asynchronous transaction commits. Hence there is no "generic" method we can use for changing the ownership of arbitrary metadata structures. For inode forks, however, there is a simple method of communicating that the fork contents need the owner rewritten - we can pass a inode log format flag for the fork for the transaction that does a fork swap. This flag will then follow the inode fork through relogging actions so when the swap actually gets replayed the ownership can be changed immediately by log recovery. So that gives us a simple method of "whole fork" exchange between two inodes. This is relatively simple to implement, so it makes sense to do this as an initial implementation to support xfs_fsr on CRC enabled filesytems in the same manner as we do on existing filesystems. This commit introduces the swapext driven functionality, the recovery functionality will be in a separate patch. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:44 +00:00
*/
int
xfs_bmbt_change_owner(
struct xfs_trans *tp,
struct xfs_inode *ip,
int whichfork,
xfs: recovery of swap extents operations for CRC filesystems This is the recovery side of the btree block owner change operation performed by swapext on CRC enabled filesystems. We detect that an owner change is needed by the flag that has been placed on the inode log format flag field. Because the inode recovery is being replayed after the buffers that make up the BMBT in the given checkpoint, we can walk all the buffers and directly modify them when we see the flag set on an inode. Because the inode can be relogged and hence present in multiple chekpoints with the "change owner" flag set, we could do multiple passes across the inode to do this change. While this isn't optimal, we can't directly ignore the flag as there may be multiple independent swap extent operations being replayed on the same inode in different checkpoints so we can't ignore them. Further, because the owner change operation uses ordered buffers, we might have buffers that are newer on disk than the current checkpoint and so already have the owner changed in them. Hence we cannot just peek at a buffer in the tree and check that it has the correct owner and assume that the change was completed. So, for the moment just brute force the owner change every time we see an inode with the flag set. Note that we have to be careful here because the owner of the buffers may point to either the old owner or the new owner. Currently the verifier can't verify the owner directly, so there is no failure case here right now. If we verify the owner exactly in future, then we'll have to take this into account. This was tested in terms of normal operation via xfstests - all of the fsr tests now pass without failure. however, we really need to modify xfs/227 to stress v3 inodes correctly to ensure we fully cover this case for v5 filesystems. In terms of recovery testing, I used a hacked version of xfs_fsr that held the temp inode open for a few seconds before exiting so that the filesystem could be shut down with an open owner change recovery flags set on at least the temp inode. fsr leaves the temp inode unlinked and in btree format, so this was necessary for the owner change to be reliably replayed. logprint confirmed the tmp inode in the log had the correct flag set: INO: cnt:3 total:3 a:0x69e9e0 len:56 a:0x69ea20 len:176 a:0x69eae0 len:88 INODE: #regs:3 ino:0x44 flags:0x209 dsize:88 ^^^^^ 0x200 is set, indicating a data fork owner change needed to be replayed on inode 0x44. A printk in the revoery code confirmed that the inode change was recovered: XFS (vdc): Mounting Filesystem XFS (vdc): Starting recovery (logdev: internal) recovering owner change ino 0x44 XFS (vdc): Version 5 superblock detected. This kernel L support enabled! Use of these features in this kernel is at your own risk! XFS (vdc): Ending recovery (logdev: internal) The script used to test this was: $ cat ./recovery-fsr.sh #!/bin/bash dev=/dev/vdc mntpt=/mnt/scratch testfile=$mntpt/testfile umount $mntpt mkfs.xfs -f -m crc=1 $dev mount $dev $mntpt chmod 777 $mntpt for i in `seq 10000 -1 0`; do xfs_io -f -d -c "pwrite $(($i * 4096)) 4096" $testfile > /dev/null 2>&1 done xfs_bmap -vp $testfile |head -20 xfs_fsr -d -v $testfile & sleep 10 /home/dave/src/xfstests-dev/src/godown -f $mntpt wait umount $mntpt xfs_logprint -t $dev |tail -20 time mount $dev $mntpt xfs_bmap -vp $testfile umount $mntpt $ Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:45 +00:00
xfs_ino_t new_owner,
struct list_head *buffer_list)
xfs: swap extents operations for CRC filesystems For CRC enabled filesystems, we can't just swap inode forks from one inode to another when defragmenting a file - the blocks in the inode fork bmap btree contain pointers back to the owner inode. Hence if we are to swap the inode forks we have to atomically modify every block in the btree during the transaction. We are doing an entire fork swap here, so we could create a new transaction item type that indicates we are changing the owner of a certain structure from one value to another. If we combine this with ordered buffer logging to modify all the buffers in the tree, then we can change the buffers in the tree without needing log space for the operation. However, this then requires log recovery to perform the modification of the owner information of the objects/structures in question. This does introduce some interesting ordering details into recovery: we have to make sure that the owner change replay occurs after the change that moves the objects is made, not before. Hence we can't use a separate log item for this as we have no guarantee of strict ordering between multiple items in the log due to the relogging action of asynchronous transaction commits. Hence there is no "generic" method we can use for changing the ownership of arbitrary metadata structures. For inode forks, however, there is a simple method of communicating that the fork contents need the owner rewritten - we can pass a inode log format flag for the fork for the transaction that does a fork swap. This flag will then follow the inode fork through relogging actions so when the swap actually gets replayed the ownership can be changed immediately by log recovery. So that gives us a simple method of "whole fork" exchange between two inodes. This is relatively simple to implement, so it makes sense to do this as an initial implementation to support xfs_fsr on CRC enabled filesytems in the same manner as we do on existing filesystems. This commit introduces the swapext driven functionality, the recovery functionality will be in a separate patch. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:44 +00:00
{
struct xfs_btree_cur *cur;
int error;
xfs: recovery of swap extents operations for CRC filesystems This is the recovery side of the btree block owner change operation performed by swapext on CRC enabled filesystems. We detect that an owner change is needed by the flag that has been placed on the inode log format flag field. Because the inode recovery is being replayed after the buffers that make up the BMBT in the given checkpoint, we can walk all the buffers and directly modify them when we see the flag set on an inode. Because the inode can be relogged and hence present in multiple chekpoints with the "change owner" flag set, we could do multiple passes across the inode to do this change. While this isn't optimal, we can't directly ignore the flag as there may be multiple independent swap extent operations being replayed on the same inode in different checkpoints so we can't ignore them. Further, because the owner change operation uses ordered buffers, we might have buffers that are newer on disk than the current checkpoint and so already have the owner changed in them. Hence we cannot just peek at a buffer in the tree and check that it has the correct owner and assume that the change was completed. So, for the moment just brute force the owner change every time we see an inode with the flag set. Note that we have to be careful here because the owner of the buffers may point to either the old owner or the new owner. Currently the verifier can't verify the owner directly, so there is no failure case here right now. If we verify the owner exactly in future, then we'll have to take this into account. This was tested in terms of normal operation via xfstests - all of the fsr tests now pass without failure. however, we really need to modify xfs/227 to stress v3 inodes correctly to ensure we fully cover this case for v5 filesystems. In terms of recovery testing, I used a hacked version of xfs_fsr that held the temp inode open for a few seconds before exiting so that the filesystem could be shut down with an open owner change recovery flags set on at least the temp inode. fsr leaves the temp inode unlinked and in btree format, so this was necessary for the owner change to be reliably replayed. logprint confirmed the tmp inode in the log had the correct flag set: INO: cnt:3 total:3 a:0x69e9e0 len:56 a:0x69ea20 len:176 a:0x69eae0 len:88 INODE: #regs:3 ino:0x44 flags:0x209 dsize:88 ^^^^^ 0x200 is set, indicating a data fork owner change needed to be replayed on inode 0x44. A printk in the revoery code confirmed that the inode change was recovered: XFS (vdc): Mounting Filesystem XFS (vdc): Starting recovery (logdev: internal) recovering owner change ino 0x44 XFS (vdc): Version 5 superblock detected. This kernel L support enabled! Use of these features in this kernel is at your own risk! XFS (vdc): Ending recovery (logdev: internal) The script used to test this was: $ cat ./recovery-fsr.sh #!/bin/bash dev=/dev/vdc mntpt=/mnt/scratch testfile=$mntpt/testfile umount $mntpt mkfs.xfs -f -m crc=1 $dev mount $dev $mntpt chmod 777 $mntpt for i in `seq 10000 -1 0`; do xfs_io -f -d -c "pwrite $(($i * 4096)) 4096" $testfile > /dev/null 2>&1 done xfs_bmap -vp $testfile |head -20 xfs_fsr -d -v $testfile & sleep 10 /home/dave/src/xfstests-dev/src/godown -f $mntpt wait umount $mntpt xfs_logprint -t $dev |tail -20 time mount $dev $mntpt xfs_bmap -vp $testfile umount $mntpt $ Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:45 +00:00
ASSERT(tp || buffer_list);
ASSERT(!(tp && buffer_list));
ASSERT(XFS_IFORK_PTR(ip, whichfork)->if_format == XFS_DINODE_FMT_BTREE);
xfs: swap extents operations for CRC filesystems For CRC enabled filesystems, we can't just swap inode forks from one inode to another when defragmenting a file - the blocks in the inode fork bmap btree contain pointers back to the owner inode. Hence if we are to swap the inode forks we have to atomically modify every block in the btree during the transaction. We are doing an entire fork swap here, so we could create a new transaction item type that indicates we are changing the owner of a certain structure from one value to another. If we combine this with ordered buffer logging to modify all the buffers in the tree, then we can change the buffers in the tree without needing log space for the operation. However, this then requires log recovery to perform the modification of the owner information of the objects/structures in question. This does introduce some interesting ordering details into recovery: we have to make sure that the owner change replay occurs after the change that moves the objects is made, not before. Hence we can't use a separate log item for this as we have no guarantee of strict ordering between multiple items in the log due to the relogging action of asynchronous transaction commits. Hence there is no "generic" method we can use for changing the ownership of arbitrary metadata structures. For inode forks, however, there is a simple method of communicating that the fork contents need the owner rewritten - we can pass a inode log format flag for the fork for the transaction that does a fork swap. This flag will then follow the inode fork through relogging actions so when the swap actually gets replayed the ownership can be changed immediately by log recovery. So that gives us a simple method of "whole fork" exchange between two inodes. This is relatively simple to implement, so it makes sense to do this as an initial implementation to support xfs_fsr on CRC enabled filesytems in the same manner as we do on existing filesystems. This commit introduces the swapext driven functionality, the recovery functionality will be in a separate patch. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:44 +00:00
cur = xfs_bmbt_init_cursor(ip->i_mount, tp, ip, whichfork);
cur->bc_ino.flags |= XFS_BTCUR_BMBT_INVALID_OWNER;
xfs: recovery of swap extents operations for CRC filesystems This is the recovery side of the btree block owner change operation performed by swapext on CRC enabled filesystems. We detect that an owner change is needed by the flag that has been placed on the inode log format flag field. Because the inode recovery is being replayed after the buffers that make up the BMBT in the given checkpoint, we can walk all the buffers and directly modify them when we see the flag set on an inode. Because the inode can be relogged and hence present in multiple chekpoints with the "change owner" flag set, we could do multiple passes across the inode to do this change. While this isn't optimal, we can't directly ignore the flag as there may be multiple independent swap extent operations being replayed on the same inode in different checkpoints so we can't ignore them. Further, because the owner change operation uses ordered buffers, we might have buffers that are newer on disk than the current checkpoint and so already have the owner changed in them. Hence we cannot just peek at a buffer in the tree and check that it has the correct owner and assume that the change was completed. So, for the moment just brute force the owner change every time we see an inode with the flag set. Note that we have to be careful here because the owner of the buffers may point to either the old owner or the new owner. Currently the verifier can't verify the owner directly, so there is no failure case here right now. If we verify the owner exactly in future, then we'll have to take this into account. This was tested in terms of normal operation via xfstests - all of the fsr tests now pass without failure. however, we really need to modify xfs/227 to stress v3 inodes correctly to ensure we fully cover this case for v5 filesystems. In terms of recovery testing, I used a hacked version of xfs_fsr that held the temp inode open for a few seconds before exiting so that the filesystem could be shut down with an open owner change recovery flags set on at least the temp inode. fsr leaves the temp inode unlinked and in btree format, so this was necessary for the owner change to be reliably replayed. logprint confirmed the tmp inode in the log had the correct flag set: INO: cnt:3 total:3 a:0x69e9e0 len:56 a:0x69ea20 len:176 a:0x69eae0 len:88 INODE: #regs:3 ino:0x44 flags:0x209 dsize:88 ^^^^^ 0x200 is set, indicating a data fork owner change needed to be replayed on inode 0x44. A printk in the revoery code confirmed that the inode change was recovered: XFS (vdc): Mounting Filesystem XFS (vdc): Starting recovery (logdev: internal) recovering owner change ino 0x44 XFS (vdc): Version 5 superblock detected. This kernel L support enabled! Use of these features in this kernel is at your own risk! XFS (vdc): Ending recovery (logdev: internal) The script used to test this was: $ cat ./recovery-fsr.sh #!/bin/bash dev=/dev/vdc mntpt=/mnt/scratch testfile=$mntpt/testfile umount $mntpt mkfs.xfs -f -m crc=1 $dev mount $dev $mntpt chmod 777 $mntpt for i in `seq 10000 -1 0`; do xfs_io -f -d -c "pwrite $(($i * 4096)) 4096" $testfile > /dev/null 2>&1 done xfs_bmap -vp $testfile |head -20 xfs_fsr -d -v $testfile & sleep 10 /home/dave/src/xfstests-dev/src/godown -f $mntpt wait umount $mntpt xfs_logprint -t $dev |tail -20 time mount $dev $mntpt xfs_bmap -vp $testfile umount $mntpt $ Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:45 +00:00
error = xfs_btree_change_owner(cur, new_owner, buffer_list);
xfs_btree_del_cursor(cur, error);
xfs: swap extents operations for CRC filesystems For CRC enabled filesystems, we can't just swap inode forks from one inode to another when defragmenting a file - the blocks in the inode fork bmap btree contain pointers back to the owner inode. Hence if we are to swap the inode forks we have to atomically modify every block in the btree during the transaction. We are doing an entire fork swap here, so we could create a new transaction item type that indicates we are changing the owner of a certain structure from one value to another. If we combine this with ordered buffer logging to modify all the buffers in the tree, then we can change the buffers in the tree without needing log space for the operation. However, this then requires log recovery to perform the modification of the owner information of the objects/structures in question. This does introduce some interesting ordering details into recovery: we have to make sure that the owner change replay occurs after the change that moves the objects is made, not before. Hence we can't use a separate log item for this as we have no guarantee of strict ordering between multiple items in the log due to the relogging action of asynchronous transaction commits. Hence there is no "generic" method we can use for changing the ownership of arbitrary metadata structures. For inode forks, however, there is a simple method of communicating that the fork contents need the owner rewritten - we can pass a inode log format flag for the fork for the transaction that does a fork swap. This flag will then follow the inode fork through relogging actions so when the swap actually gets replayed the ownership can be changed immediately by log recovery. So that gives us a simple method of "whole fork" exchange between two inodes. This is relatively simple to implement, so it makes sense to do this as an initial implementation to support xfs_fsr on CRC enabled filesytems in the same manner as we do on existing filesystems. This commit introduces the swapext driven functionality, the recovery functionality will be in a separate patch. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Mark Tinguely <tinguely@sgi.com> Signed-off-by: Ben Myers <bpm@sgi.com>
2013-08-30 00:23:44 +00:00
return error;
}
/* Calculate the bmap btree size for some records. */
unsigned long long
xfs_bmbt_calc_size(
struct xfs_mount *mp,
unsigned long long len)
{
return xfs_btree_calc_size(mp->m_bmap_dmnr, len);
}