WSL2-Linux-Kernel/fs/xfs/libxfs/xfs_da_btree.c

2689 строки
71 KiB
C
Исходник Обычный вид История

/*
* Copyright (c) 2000-2005 Silicon Graphics, Inc.
* Copyright (c) 2013 Red Hat, 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_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_da_format.h"
#include "xfs_da_btree.h"
#include "xfs_dir2.h"
#include "xfs_dir2_priv.h"
#include "xfs_inode.h"
#include "xfs_trans.h"
#include "xfs_inode_item.h"
#include "xfs_alloc.h"
#include "xfs_bmap.h"
#include "xfs_attr.h"
#include "xfs_attr_leaf.h"
#include "xfs_error.h"
xfs: event tracing support Convert the old xfs tracing support that could only be used with the out of tree kdb and xfsidbg patches to use the generic event tracer. To use it make sure CONFIG_EVENT_TRACING is enabled and then enable all xfs trace channels by: echo 1 > /sys/kernel/debug/tracing/events/xfs/enable or alternatively enable single events by just doing the same in one event subdirectory, e.g. echo 1 > /sys/kernel/debug/tracing/events/xfs/xfs_ihold/enable or set more complex filters, etc. In Documentation/trace/events.txt all this is desctribed in more detail. To reads the events do a cat /sys/kernel/debug/tracing/trace Compared to the last posting this patch converts the tracing mostly to the one tracepoint per callsite model that other users of the new tracing facility also employ. This allows a very fine-grained control of the tracing, a cleaner output of the traces and also enables the perf tool to use each tracepoint as a virtual performance counter, allowing us to e.g. count how often certain workloads git various spots in XFS. Take a look at http://lwn.net/Articles/346470/ for some examples. Also the btree tracing isn't included at all yet, as it will require additional core tracing features not in mainline yet, I plan to deliver it later. And the really nice thing about this patch is that it actually removes many lines of code while adding this nice functionality: fs/xfs/Makefile | 8 fs/xfs/linux-2.6/xfs_acl.c | 1 fs/xfs/linux-2.6/xfs_aops.c | 52 - fs/xfs/linux-2.6/xfs_aops.h | 2 fs/xfs/linux-2.6/xfs_buf.c | 117 +-- fs/xfs/linux-2.6/xfs_buf.h | 33 fs/xfs/linux-2.6/xfs_fs_subr.c | 3 fs/xfs/linux-2.6/xfs_ioctl.c | 1 fs/xfs/linux-2.6/xfs_ioctl32.c | 1 fs/xfs/linux-2.6/xfs_iops.c | 1 fs/xfs/linux-2.6/xfs_linux.h | 1 fs/xfs/linux-2.6/xfs_lrw.c | 87 -- fs/xfs/linux-2.6/xfs_lrw.h | 45 - fs/xfs/linux-2.6/xfs_super.c | 104 --- fs/xfs/linux-2.6/xfs_super.h | 7 fs/xfs/linux-2.6/xfs_sync.c | 1 fs/xfs/linux-2.6/xfs_trace.c | 75 ++ fs/xfs/linux-2.6/xfs_trace.h | 1369 +++++++++++++++++++++++++++++++++++++++++ fs/xfs/linux-2.6/xfs_vnode.h | 4 fs/xfs/quota/xfs_dquot.c | 110 --- fs/xfs/quota/xfs_dquot.h | 21 fs/xfs/quota/xfs_qm.c | 40 - fs/xfs/quota/xfs_qm_syscalls.c | 4 fs/xfs/support/ktrace.c | 323 --------- fs/xfs/support/ktrace.h | 85 -- fs/xfs/xfs.h | 16 fs/xfs/xfs_ag.h | 14 fs/xfs/xfs_alloc.c | 230 +----- fs/xfs/xfs_alloc.h | 27 fs/xfs/xfs_alloc_btree.c | 1 fs/xfs/xfs_attr.c | 107 --- fs/xfs/xfs_attr.h | 10 fs/xfs/xfs_attr_leaf.c | 14 fs/xfs/xfs_attr_sf.h | 40 - fs/xfs/xfs_bmap.c | 507 +++------------ fs/xfs/xfs_bmap.h | 49 - fs/xfs/xfs_bmap_btree.c | 6 fs/xfs/xfs_btree.c | 5 fs/xfs/xfs_btree_trace.h | 17 fs/xfs/xfs_buf_item.c | 87 -- fs/xfs/xfs_buf_item.h | 20 fs/xfs/xfs_da_btree.c | 3 fs/xfs/xfs_da_btree.h | 7 fs/xfs/xfs_dfrag.c | 2 fs/xfs/xfs_dir2.c | 8 fs/xfs/xfs_dir2_block.c | 20 fs/xfs/xfs_dir2_leaf.c | 21 fs/xfs/xfs_dir2_node.c | 27 fs/xfs/xfs_dir2_sf.c | 26 fs/xfs/xfs_dir2_trace.c | 216 ------ fs/xfs/xfs_dir2_trace.h | 72 -- fs/xfs/xfs_filestream.c | 8 fs/xfs/xfs_fsops.c | 2 fs/xfs/xfs_iget.c | 111 --- fs/xfs/xfs_inode.c | 67 -- fs/xfs/xfs_inode.h | 76 -- fs/xfs/xfs_inode_item.c | 5 fs/xfs/xfs_iomap.c | 85 -- fs/xfs/xfs_iomap.h | 8 fs/xfs/xfs_log.c | 181 +---- fs/xfs/xfs_log_priv.h | 20 fs/xfs/xfs_log_recover.c | 1 fs/xfs/xfs_mount.c | 2 fs/xfs/xfs_quota.h | 8 fs/xfs/xfs_rename.c | 1 fs/xfs/xfs_rtalloc.c | 1 fs/xfs/xfs_rw.c | 3 fs/xfs/xfs_trans.h | 47 + fs/xfs/xfs_trans_buf.c | 62 - fs/xfs/xfs_vnodeops.c | 8 70 files changed, 2151 insertions(+), 2592 deletions(-) Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Alex Elder <aelder@sgi.com>
2009-12-15 02:14:59 +03:00
#include "xfs_trace.h"
#include "xfs_cksum.h"
#include "xfs_buf_item.h"
xfs: validate metadata LSNs against log on v5 superblocks Since the onset of v5 superblocks, the LSN of the last modification has been included in a variety of on-disk data structures. This LSN is used to provide log recovery ordering guarantees (e.g., to ensure an older log recovery item is not replayed over a newer target data structure). While this works correctly from the point a filesystem is formatted and mounted, userspace tools have some problematic behaviors that defeat this mechanism. For example, xfs_repair historically zeroes out the log unconditionally (regardless of whether corruption is detected). If this occurs, the LSN of the filesystem is reset and the log is now in a problematic state with respect to on-disk metadata structures that might have a larger LSN. Until either the log catches up to the highest previously used metadata LSN or each affected data structure is modified and written out without incident (which resets the metadata LSN), log recovery is susceptible to filesystem corruption. This problem is ultimately addressed and repaired in the associated userspace tools. The kernel is still responsible to detect the problem and notify the user that something is wrong. Check the superblock LSN at mount time and fail the mount if it is invalid. From that point on, trigger verifier failure on any metadata I/O where an invalid LSN is detected. This results in a filesystem shutdown and guarantees that we do not log metadata changes with invalid LSNs on disk. Since this is a known issue with a known recovery path, present a warning to instruct the user how to recover. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-10-12 07:59:25 +03:00
#include "xfs_log.h"
/*
* xfs_da_btree.c
*
* Routines to implement directories as Btrees of hashed names.
*/
/*========================================================================
* Function prototypes for the kernel.
*========================================================================*/
/*
* Routines used for growing the Btree.
*/
STATIC int xfs_da3_root_split(xfs_da_state_t *state,
xfs_da_state_blk_t *existing_root,
xfs_da_state_blk_t *new_child);
STATIC int xfs_da3_node_split(xfs_da_state_t *state,
xfs_da_state_blk_t *existing_blk,
xfs_da_state_blk_t *split_blk,
xfs_da_state_blk_t *blk_to_add,
int treelevel,
int *result);
STATIC void xfs_da3_node_rebalance(xfs_da_state_t *state,
xfs_da_state_blk_t *node_blk_1,
xfs_da_state_blk_t *node_blk_2);
STATIC void xfs_da3_node_add(xfs_da_state_t *state,
xfs_da_state_blk_t *old_node_blk,
xfs_da_state_blk_t *new_node_blk);
/*
* Routines used for shrinking the Btree.
*/
STATIC int xfs_da3_root_join(xfs_da_state_t *state,
xfs_da_state_blk_t *root_blk);
STATIC int xfs_da3_node_toosmall(xfs_da_state_t *state, int *retval);
STATIC void xfs_da3_node_remove(xfs_da_state_t *state,
xfs_da_state_blk_t *drop_blk);
STATIC void xfs_da3_node_unbalance(xfs_da_state_t *state,
xfs_da_state_blk_t *src_node_blk,
xfs_da_state_blk_t *dst_node_blk);
/*
* Utility routines.
*/
STATIC int xfs_da3_blk_unlink(xfs_da_state_t *state,
xfs_da_state_blk_t *drop_blk,
xfs_da_state_blk_t *save_blk);
kmem_zone_t *xfs_da_state_zone; /* anchor for state struct zone */
/*
* Allocate a dir-state structure.
* We don't put them on the stack since they're large.
*/
xfs_da_state_t *
xfs_da_state_alloc(void)
{
return kmem_zone_zalloc(xfs_da_state_zone, KM_NOFS);
}
/*
* Kill the altpath contents of a da-state structure.
*/
STATIC void
xfs_da_state_kill_altpath(xfs_da_state_t *state)
{
int i;
for (i = 0; i < state->altpath.active; i++)
state->altpath.blk[i].bp = NULL;
state->altpath.active = 0;
}
/*
* Free a da-state structure.
*/
void
xfs_da_state_free(xfs_da_state_t *state)
{
xfs_da_state_kill_altpath(state);
#ifdef DEBUG
memset((char *)state, 0, sizeof(*state));
#endif /* DEBUG */
kmem_zone_free(xfs_da_state_zone, state);
}
static bool
xfs_da3_node_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
struct xfs_da_intnode *hdr = bp->b_addr;
struct xfs_da3_icnode_hdr ichdr;
const struct xfs_dir_ops *ops;
ops = xfs_dir_get_ops(mp, NULL);
ops->node_hdr_from_disk(&ichdr, hdr);
if (xfs_sb_version_hascrc(&mp->m_sb)) {
struct xfs_da3_node_hdr *hdr3 = bp->b_addr;
if (ichdr.magic != XFS_DA3_NODE_MAGIC)
return false;
if (!uuid_equal(&hdr3->info.uuid, &mp->m_sb.sb_meta_uuid))
return false;
if (be64_to_cpu(hdr3->info.blkno) != bp->b_bn)
return false;
xfs: validate metadata LSNs against log on v5 superblocks Since the onset of v5 superblocks, the LSN of the last modification has been included in a variety of on-disk data structures. This LSN is used to provide log recovery ordering guarantees (e.g., to ensure an older log recovery item is not replayed over a newer target data structure). While this works correctly from the point a filesystem is formatted and mounted, userspace tools have some problematic behaviors that defeat this mechanism. For example, xfs_repair historically zeroes out the log unconditionally (regardless of whether corruption is detected). If this occurs, the LSN of the filesystem is reset and the log is now in a problematic state with respect to on-disk metadata structures that might have a larger LSN. Until either the log catches up to the highest previously used metadata LSN or each affected data structure is modified and written out without incident (which resets the metadata LSN), log recovery is susceptible to filesystem corruption. This problem is ultimately addressed and repaired in the associated userspace tools. The kernel is still responsible to detect the problem and notify the user that something is wrong. Check the superblock LSN at mount time and fail the mount if it is invalid. From that point on, trigger verifier failure on any metadata I/O where an invalid LSN is detected. This results in a filesystem shutdown and guarantees that we do not log metadata changes with invalid LSNs on disk. Since this is a known issue with a known recovery path, present a warning to instruct the user how to recover. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-10-12 07:59:25 +03:00
if (!xfs_log_check_lsn(mp, be64_to_cpu(hdr3->info.lsn)))
return false;
} else {
if (ichdr.magic != XFS_DA_NODE_MAGIC)
return false;
}
if (ichdr.level == 0)
return false;
if (ichdr.level > XFS_DA_NODE_MAXDEPTH)
return false;
if (ichdr.count == 0)
return false;
/*
* we don't know if the node is for and attribute or directory tree,
* so only fail if the count is outside both bounds
*/
if (ichdr.count > mp->m_dir_geo->node_ents &&
ichdr.count > mp->m_attr_geo->node_ents)
return false;
/* XXX: hash order check? */
return true;
}
static void
xfs_da3_node_write_verify(
struct xfs_buf *bp)
{
struct xfs_mount *mp = bp->b_target->bt_mount;
struct xfs_buf_log_item *bip = bp->b_fspriv;
struct xfs_da3_node_hdr *hdr3 = bp->b_addr;
if (!xfs_da3_node_verify(bp)) {
xfs_buf_ioerror(bp, -EFSCORRUPTED);
xfs_verifier_error(bp);
return;
}
if (!xfs_sb_version_hascrc(&mp->m_sb))
return;
if (bip)
hdr3->info.lsn = cpu_to_be64(bip->bli_item.li_lsn);
xfs_buf_update_cksum(bp, XFS_DA3_NODE_CRC_OFF);
}
/*
* leaf/node format detection on trees is sketchy, so a node read can be done on
* leaf level blocks when detection identifies the tree as a node format tree
* incorrectly. In this case, we need to swap the verifier to match the correct
* format of the block being read.
*/
static void
xfs_da3_node_read_verify(
struct xfs_buf *bp)
{
struct xfs_da_blkinfo *info = bp->b_addr;
switch (be16_to_cpu(info->magic)) {
case XFS_DA3_NODE_MAGIC:
if (!xfs_buf_verify_cksum(bp, XFS_DA3_NODE_CRC_OFF)) {
xfs_buf_ioerror(bp, -EFSBADCRC);
break;
}
/* fall through */
case XFS_DA_NODE_MAGIC:
if (!xfs_da3_node_verify(bp)) {
xfs_buf_ioerror(bp, -EFSCORRUPTED);
break;
}
return;
case XFS_ATTR_LEAF_MAGIC:
case XFS_ATTR3_LEAF_MAGIC:
bp->b_ops = &xfs_attr3_leaf_buf_ops;
bp->b_ops->verify_read(bp);
return;
case XFS_DIR2_LEAFN_MAGIC:
case XFS_DIR3_LEAFN_MAGIC:
bp->b_ops = &xfs_dir3_leafn_buf_ops;
bp->b_ops->verify_read(bp);
return;
default:
xfs_buf_ioerror(bp, -EFSCORRUPTED);
break;
}
/* corrupt block */
xfs_verifier_error(bp);
}
const struct xfs_buf_ops xfs_da3_node_buf_ops = {
.name = "xfs_da3_node",
.verify_read = xfs_da3_node_read_verify,
.verify_write = xfs_da3_node_write_verify,
};
int
xfs_da3_node_read(
struct xfs_trans *tp,
struct xfs_inode *dp,
xfs_dablk_t bno,
xfs_daddr_t mappedbno,
struct xfs_buf **bpp,
int which_fork)
{
int err;
err = xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
which_fork, &xfs_da3_node_buf_ops);
if (!err && tp && *bpp) {
struct xfs_da_blkinfo *info = (*bpp)->b_addr;
int type;
switch (be16_to_cpu(info->magic)) {
case XFS_DA_NODE_MAGIC:
case XFS_DA3_NODE_MAGIC:
type = XFS_BLFT_DA_NODE_BUF;
break;
case XFS_ATTR_LEAF_MAGIC:
case XFS_ATTR3_LEAF_MAGIC:
type = XFS_BLFT_ATTR_LEAF_BUF;
break;
case XFS_DIR2_LEAFN_MAGIC:
case XFS_DIR3_LEAFN_MAGIC:
type = XFS_BLFT_DIR_LEAFN_BUF;
break;
default:
type = 0;
ASSERT(0);
break;
}
xfs_trans_buf_set_type(tp, *bpp, type);
}
return err;
}
/*========================================================================
* Routines used for growing the Btree.
*========================================================================*/
/*
* Create the initial contents of an intermediate node.
*/
int
xfs_da3_node_create(
struct xfs_da_args *args,
xfs_dablk_t blkno,
int level,
struct xfs_buf **bpp,
int whichfork)
{
struct xfs_da_intnode *node;
struct xfs_trans *tp = args->trans;
struct xfs_mount *mp = tp->t_mountp;
struct xfs_da3_icnode_hdr ichdr = {0};
struct xfs_buf *bp;
int error;
struct xfs_inode *dp = args->dp;
trace_xfs_da_node_create(args);
ASSERT(level <= XFS_DA_NODE_MAXDEPTH);
error = xfs_da_get_buf(tp, dp, blkno, -1, &bp, whichfork);
if (error)
return error;
bp->b_ops = &xfs_da3_node_buf_ops;
xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DA_NODE_BUF);
node = bp->b_addr;
if (xfs_sb_version_hascrc(&mp->m_sb)) {
struct xfs_da3_node_hdr *hdr3 = bp->b_addr;
xfs: validate metadata LSNs against log on v5 superblocks Since the onset of v5 superblocks, the LSN of the last modification has been included in a variety of on-disk data structures. This LSN is used to provide log recovery ordering guarantees (e.g., to ensure an older log recovery item is not replayed over a newer target data structure). While this works correctly from the point a filesystem is formatted and mounted, userspace tools have some problematic behaviors that defeat this mechanism. For example, xfs_repair historically zeroes out the log unconditionally (regardless of whether corruption is detected). If this occurs, the LSN of the filesystem is reset and the log is now in a problematic state with respect to on-disk metadata structures that might have a larger LSN. Until either the log catches up to the highest previously used metadata LSN or each affected data structure is modified and written out without incident (which resets the metadata LSN), log recovery is susceptible to filesystem corruption. This problem is ultimately addressed and repaired in the associated userspace tools. The kernel is still responsible to detect the problem and notify the user that something is wrong. Check the superblock LSN at mount time and fail the mount if it is invalid. From that point on, trigger verifier failure on any metadata I/O where an invalid LSN is detected. This results in a filesystem shutdown and guarantees that we do not log metadata changes with invalid LSNs on disk. Since this is a known issue with a known recovery path, present a warning to instruct the user how to recover. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-10-12 07:59:25 +03:00
memset(hdr3, 0, sizeof(struct xfs_da3_node_hdr));
ichdr.magic = XFS_DA3_NODE_MAGIC;
hdr3->info.blkno = cpu_to_be64(bp->b_bn);
hdr3->info.owner = cpu_to_be64(args->dp->i_ino);
uuid_copy(&hdr3->info.uuid, &mp->m_sb.sb_meta_uuid);
} else {
ichdr.magic = XFS_DA_NODE_MAGIC;
}
ichdr.level = level;
dp->d_ops->node_hdr_to_disk(node, &ichdr);
xfs_trans_log_buf(tp, bp,
XFS_DA_LOGRANGE(node, &node->hdr, dp->d_ops->node_hdr_size));
*bpp = bp;
return 0;
}
/*
* Split a leaf node, rebalance, then possibly split
* intermediate nodes, rebalance, etc.
*/
int /* error */
xfs_da3_split(
struct xfs_da_state *state)
{
struct xfs_da_state_blk *oldblk;
struct xfs_da_state_blk *newblk;
struct xfs_da_state_blk *addblk;
struct xfs_da_intnode *node;
int max;
int action = 0;
int error;
int i;
trace_xfs_da_split(state->args);
/*
* Walk back up the tree splitting/inserting/adjusting as necessary.
* If we need to insert and there isn't room, split the node, then
* decide which fragment to insert the new block from below into.
* Note that we may split the root this way, but we need more fixup.
*/
max = state->path.active - 1;
ASSERT((max >= 0) && (max < XFS_DA_NODE_MAXDEPTH));
ASSERT(state->path.blk[max].magic == XFS_ATTR_LEAF_MAGIC ||
state->path.blk[max].magic == XFS_DIR2_LEAFN_MAGIC);
addblk = &state->path.blk[max]; /* initial dummy value */
for (i = max; (i >= 0) && addblk; state->path.active--, i--) {
oldblk = &state->path.blk[i];
newblk = &state->altpath.blk[i];
/*
* If a leaf node then
* Allocate a new leaf node, then rebalance across them.
* else if an intermediate node then
* We split on the last layer, must we split the node?
*/
switch (oldblk->magic) {
case XFS_ATTR_LEAF_MAGIC:
error = xfs_attr3_leaf_split(state, oldblk, newblk);
if ((error != 0) && (error != -ENOSPC)) {
return error; /* GROT: attr is inconsistent */
}
if (!error) {
addblk = newblk;
break;
}
/*
libxfs: directory node splitting does not have an extra block xfsprogs source commit 4280e59dcbc4cd8e01585efe788a68eb378048e8 xfs_da3_split() has to handle all three versions of the directory/attribute btree structure. The attr tree is v1, the dir tre is v2 or v3. The main difference between the v1 and v2/3 trees is the way tree nodes are split - in the v1 tree we can require a double split to occur because the object to be inserted may be larger than the space made by splitting a leaf. In this case we need to do a double split - one to split the full leaf, then another to allocate an empty leaf block in the correct location for the new entry. This does not happen with dir (v2/v3) formats as the objects being inserted are always guaranteed to fit into the new space in the split blocks. Indeed, for directories they *may* be an extra block on this buffer pointer. However, it's guaranteed not to be a leaf block (i.e. a directory data block) - the directory code only ever places hash index or free space blocks in this pointer (as a cursor of sorts), and so to use it as a directory data block will immediately corrupt the directory. The problem is that the code assumes that there may be extra blocks that we need to link into the tree once we've split the root, but this is not true for either dir or attr trees, because the extra attr block is always consumed by the last node split before we split the root. Hence the linking in an extra block is always wrong at the root split level, and this manifests itself in repair as a directory corruption in a repaired directory, leaving the directory rebuild incomplete. This is a dir v2 zero-day bug - it was in the initial dir v2 commit that was made back in February 1998. Fix this by ensuring the linking of the blocks after the root split never tries to make use of the extra blocks that may be held in the cursor. They are held there for other purposes and should never be touched by the root splitting code. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 02:51:05 +03:00
* Entry wouldn't fit, split the leaf again. The new
* extrablk will be consumed by xfs_da3_node_split if
* the node is split.
*/
state->extravalid = 1;
if (state->inleaf) {
state->extraafter = 0; /* before newblk */
trace_xfs_attr_leaf_split_before(state->args);
error = xfs_attr3_leaf_split(state, oldblk,
&state->extrablk);
} else {
state->extraafter = 1; /* after newblk */
trace_xfs_attr_leaf_split_after(state->args);
error = xfs_attr3_leaf_split(state, newblk,
&state->extrablk);
}
if (error)
return error; /* GROT: attr inconsistent */
addblk = newblk;
break;
case XFS_DIR2_LEAFN_MAGIC:
error = xfs_dir2_leafn_split(state, oldblk, newblk);
if (error)
return error;
addblk = newblk;
break;
case XFS_DA_NODE_MAGIC:
error = xfs_da3_node_split(state, oldblk, newblk, addblk,
max - i, &action);
addblk->bp = NULL;
if (error)
return error; /* GROT: dir is inconsistent */
/*
* Record the newly split block for the next time thru?
*/
if (action)
addblk = newblk;
else
addblk = NULL;
break;
}
/*
* Update the btree to show the new hashval for this child.
*/
xfs_da3_fixhashpath(state, &state->path);
}
if (!addblk)
return 0;
libxfs: directory node splitting does not have an extra block xfsprogs source commit 4280e59dcbc4cd8e01585efe788a68eb378048e8 xfs_da3_split() has to handle all three versions of the directory/attribute btree structure. The attr tree is v1, the dir tre is v2 or v3. The main difference between the v1 and v2/3 trees is the way tree nodes are split - in the v1 tree we can require a double split to occur because the object to be inserted may be larger than the space made by splitting a leaf. In this case we need to do a double split - one to split the full leaf, then another to allocate an empty leaf block in the correct location for the new entry. This does not happen with dir (v2/v3) formats as the objects being inserted are always guaranteed to fit into the new space in the split blocks. Indeed, for directories they *may* be an extra block on this buffer pointer. However, it's guaranteed not to be a leaf block (i.e. a directory data block) - the directory code only ever places hash index or free space blocks in this pointer (as a cursor of sorts), and so to use it as a directory data block will immediately corrupt the directory. The problem is that the code assumes that there may be extra blocks that we need to link into the tree once we've split the root, but this is not true for either dir or attr trees, because the extra attr block is always consumed by the last node split before we split the root. Hence the linking in an extra block is always wrong at the root split level, and this manifests itself in repair as a directory corruption in a repaired directory, leaving the directory rebuild incomplete. This is a dir v2 zero-day bug - it was in the initial dir v2 commit that was made back in February 1998. Fix this by ensuring the linking of the blocks after the root split never tries to make use of the extra blocks that may be held in the cursor. They are held there for other purposes and should never be touched by the root splitting code. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 02:51:05 +03:00
/*
* xfs_da3_node_split() should have consumed any extra blocks we added
* during a double leaf split in the attr fork. This is guaranteed as
* we can't be here if the attr fork only has a single leaf block.
*/
ASSERT(state->extravalid == 0 ||
state->path.blk[max].magic == XFS_DIR2_LEAFN_MAGIC);
/*
* Split the root node.
*/
ASSERT(state->path.active == 0);
oldblk = &state->path.blk[0];
error = xfs_da3_root_split(state, oldblk, addblk);
if (error) {
addblk->bp = NULL;
return error; /* GROT: dir is inconsistent */
}
/*
libxfs: directory node splitting does not have an extra block xfsprogs source commit 4280e59dcbc4cd8e01585efe788a68eb378048e8 xfs_da3_split() has to handle all three versions of the directory/attribute btree structure. The attr tree is v1, the dir tre is v2 or v3. The main difference between the v1 and v2/3 trees is the way tree nodes are split - in the v1 tree we can require a double split to occur because the object to be inserted may be larger than the space made by splitting a leaf. In this case we need to do a double split - one to split the full leaf, then another to allocate an empty leaf block in the correct location for the new entry. This does not happen with dir (v2/v3) formats as the objects being inserted are always guaranteed to fit into the new space in the split blocks. Indeed, for directories they *may* be an extra block on this buffer pointer. However, it's guaranteed not to be a leaf block (i.e. a directory data block) - the directory code only ever places hash index or free space blocks in this pointer (as a cursor of sorts), and so to use it as a directory data block will immediately corrupt the directory. The problem is that the code assumes that there may be extra blocks that we need to link into the tree once we've split the root, but this is not true for either dir or attr trees, because the extra attr block is always consumed by the last node split before we split the root. Hence the linking in an extra block is always wrong at the root split level, and this manifests itself in repair as a directory corruption in a repaired directory, leaving the directory rebuild incomplete. This is a dir v2 zero-day bug - it was in the initial dir v2 commit that was made back in February 1998. Fix this by ensuring the linking of the blocks after the root split never tries to make use of the extra blocks that may be held in the cursor. They are held there for other purposes and should never be touched by the root splitting code. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 02:51:05 +03:00
* Update pointers to the node which used to be block 0 and just got
* bumped because of the addition of a new root node. Note that the
* original block 0 could be at any position in the list of blocks in
* the tree.
*
libxfs: directory node splitting does not have an extra block xfsprogs source commit 4280e59dcbc4cd8e01585efe788a68eb378048e8 xfs_da3_split() has to handle all three versions of the directory/attribute btree structure. The attr tree is v1, the dir tre is v2 or v3. The main difference between the v1 and v2/3 trees is the way tree nodes are split - in the v1 tree we can require a double split to occur because the object to be inserted may be larger than the space made by splitting a leaf. In this case we need to do a double split - one to split the full leaf, then another to allocate an empty leaf block in the correct location for the new entry. This does not happen with dir (v2/v3) formats as the objects being inserted are always guaranteed to fit into the new space in the split blocks. Indeed, for directories they *may* be an extra block on this buffer pointer. However, it's guaranteed not to be a leaf block (i.e. a directory data block) - the directory code only ever places hash index or free space blocks in this pointer (as a cursor of sorts), and so to use it as a directory data block will immediately corrupt the directory. The problem is that the code assumes that there may be extra blocks that we need to link into the tree once we've split the root, but this is not true for either dir or attr trees, because the extra attr block is always consumed by the last node split before we split the root. Hence the linking in an extra block is always wrong at the root split level, and this manifests itself in repair as a directory corruption in a repaired directory, leaving the directory rebuild incomplete. This is a dir v2 zero-day bug - it was in the initial dir v2 commit that was made back in February 1998. Fix this by ensuring the linking of the blocks after the root split never tries to make use of the extra blocks that may be held in the cursor. They are held there for other purposes and should never be touched by the root splitting code. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 02:51:05 +03:00
* Note: the magic numbers and sibling pointers are in the same physical
* place for both v2 and v3 headers (by design). Hence it doesn't matter
* which version of the xfs_da_intnode structure we use here as the
* result will be the same using either structure.
*/
node = oldblk->bp->b_addr;
if (node->hdr.info.forw) {
libxfs: directory node splitting does not have an extra block xfsprogs source commit 4280e59dcbc4cd8e01585efe788a68eb378048e8 xfs_da3_split() has to handle all three versions of the directory/attribute btree structure. The attr tree is v1, the dir tre is v2 or v3. The main difference between the v1 and v2/3 trees is the way tree nodes are split - in the v1 tree we can require a double split to occur because the object to be inserted may be larger than the space made by splitting a leaf. In this case we need to do a double split - one to split the full leaf, then another to allocate an empty leaf block in the correct location for the new entry. This does not happen with dir (v2/v3) formats as the objects being inserted are always guaranteed to fit into the new space in the split blocks. Indeed, for directories they *may* be an extra block on this buffer pointer. However, it's guaranteed not to be a leaf block (i.e. a directory data block) - the directory code only ever places hash index or free space blocks in this pointer (as a cursor of sorts), and so to use it as a directory data block will immediately corrupt the directory. The problem is that the code assumes that there may be extra blocks that we need to link into the tree once we've split the root, but this is not true for either dir or attr trees, because the extra attr block is always consumed by the last node split before we split the root. Hence the linking in an extra block is always wrong at the root split level, and this manifests itself in repair as a directory corruption in a repaired directory, leaving the directory rebuild incomplete. This is a dir v2 zero-day bug - it was in the initial dir v2 commit that was made back in February 1998. Fix this by ensuring the linking of the blocks after the root split never tries to make use of the extra blocks that may be held in the cursor. They are held there for other purposes and should never be touched by the root splitting code. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 02:51:05 +03:00
ASSERT(be32_to_cpu(node->hdr.info.forw) == addblk->blkno);
node = addblk->bp->b_addr;
node->hdr.info.back = cpu_to_be32(oldblk->blkno);
libxfs: directory node splitting does not have an extra block xfsprogs source commit 4280e59dcbc4cd8e01585efe788a68eb378048e8 xfs_da3_split() has to handle all three versions of the directory/attribute btree structure. The attr tree is v1, the dir tre is v2 or v3. The main difference between the v1 and v2/3 trees is the way tree nodes are split - in the v1 tree we can require a double split to occur because the object to be inserted may be larger than the space made by splitting a leaf. In this case we need to do a double split - one to split the full leaf, then another to allocate an empty leaf block in the correct location for the new entry. This does not happen with dir (v2/v3) formats as the objects being inserted are always guaranteed to fit into the new space in the split blocks. Indeed, for directories they *may* be an extra block on this buffer pointer. However, it's guaranteed not to be a leaf block (i.e. a directory data block) - the directory code only ever places hash index or free space blocks in this pointer (as a cursor of sorts), and so to use it as a directory data block will immediately corrupt the directory. The problem is that the code assumes that there may be extra blocks that we need to link into the tree once we've split the root, but this is not true for either dir or attr trees, because the extra attr block is always consumed by the last node split before we split the root. Hence the linking in an extra block is always wrong at the root split level, and this manifests itself in repair as a directory corruption in a repaired directory, leaving the directory rebuild incomplete. This is a dir v2 zero-day bug - it was in the initial dir v2 commit that was made back in February 1998. Fix this by ensuring the linking of the blocks after the root split never tries to make use of the extra blocks that may be held in the cursor. They are held there for other purposes and should never be touched by the root splitting code. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 02:51:05 +03:00
xfs_trans_log_buf(state->args->trans, addblk->bp,
XFS_DA_LOGRANGE(node, &node->hdr.info,
sizeof(node->hdr.info)));
}
node = oldblk->bp->b_addr;
if (node->hdr.info.back) {
libxfs: directory node splitting does not have an extra block xfsprogs source commit 4280e59dcbc4cd8e01585efe788a68eb378048e8 xfs_da3_split() has to handle all three versions of the directory/attribute btree structure. The attr tree is v1, the dir tre is v2 or v3. The main difference between the v1 and v2/3 trees is the way tree nodes are split - in the v1 tree we can require a double split to occur because the object to be inserted may be larger than the space made by splitting a leaf. In this case we need to do a double split - one to split the full leaf, then another to allocate an empty leaf block in the correct location for the new entry. This does not happen with dir (v2/v3) formats as the objects being inserted are always guaranteed to fit into the new space in the split blocks. Indeed, for directories they *may* be an extra block on this buffer pointer. However, it's guaranteed not to be a leaf block (i.e. a directory data block) - the directory code only ever places hash index or free space blocks in this pointer (as a cursor of sorts), and so to use it as a directory data block will immediately corrupt the directory. The problem is that the code assumes that there may be extra blocks that we need to link into the tree once we've split the root, but this is not true for either dir or attr trees, because the extra attr block is always consumed by the last node split before we split the root. Hence the linking in an extra block is always wrong at the root split level, and this manifests itself in repair as a directory corruption in a repaired directory, leaving the directory rebuild incomplete. This is a dir v2 zero-day bug - it was in the initial dir v2 commit that was made back in February 1998. Fix this by ensuring the linking of the blocks after the root split never tries to make use of the extra blocks that may be held in the cursor. They are held there for other purposes and should never be touched by the root splitting code. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 02:51:05 +03:00
ASSERT(be32_to_cpu(node->hdr.info.back) == addblk->blkno);
node = addblk->bp->b_addr;
node->hdr.info.forw = cpu_to_be32(oldblk->blkno);
libxfs: directory node splitting does not have an extra block xfsprogs source commit 4280e59dcbc4cd8e01585efe788a68eb378048e8 xfs_da3_split() has to handle all three versions of the directory/attribute btree structure. The attr tree is v1, the dir tre is v2 or v3. The main difference between the v1 and v2/3 trees is the way tree nodes are split - in the v1 tree we can require a double split to occur because the object to be inserted may be larger than the space made by splitting a leaf. In this case we need to do a double split - one to split the full leaf, then another to allocate an empty leaf block in the correct location for the new entry. This does not happen with dir (v2/v3) formats as the objects being inserted are always guaranteed to fit into the new space in the split blocks. Indeed, for directories they *may* be an extra block on this buffer pointer. However, it's guaranteed not to be a leaf block (i.e. a directory data block) - the directory code only ever places hash index or free space blocks in this pointer (as a cursor of sorts), and so to use it as a directory data block will immediately corrupt the directory. The problem is that the code assumes that there may be extra blocks that we need to link into the tree once we've split the root, but this is not true for either dir or attr trees, because the extra attr block is always consumed by the last node split before we split the root. Hence the linking in an extra block is always wrong at the root split level, and this manifests itself in repair as a directory corruption in a repaired directory, leaving the directory rebuild incomplete. This is a dir v2 zero-day bug - it was in the initial dir v2 commit that was made back in February 1998. Fix this by ensuring the linking of the blocks after the root split never tries to make use of the extra blocks that may be held in the cursor. They are held there for other purposes and should never be touched by the root splitting code. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Dave Chinner <david@fromorbit.com>
2016-07-22 02:51:05 +03:00
xfs_trans_log_buf(state->args->trans, addblk->bp,
XFS_DA_LOGRANGE(node, &node->hdr.info,
sizeof(node->hdr.info)));
}
addblk->bp = NULL;
return 0;
}
/*
* Split the root. We have to create a new root and point to the two
* parts (the split old root) that we just created. Copy block zero to
* the EOF, extending the inode in process.
*/
STATIC int /* error */
xfs_da3_root_split(
struct xfs_da_state *state,
struct xfs_da_state_blk *blk1,
struct xfs_da_state_blk *blk2)
{
struct xfs_da_intnode *node;
struct xfs_da_intnode *oldroot;
struct xfs_da_node_entry *btree;
struct xfs_da3_icnode_hdr nodehdr;
struct xfs_da_args *args;
struct xfs_buf *bp;
struct xfs_inode *dp;
struct xfs_trans *tp;
struct xfs_dir2_leaf *leaf;
xfs_dablk_t blkno;
int level;
int error;
int size;
trace_xfs_da_root_split(state->args);
/*
* Copy the existing (incorrect) block from the root node position
* to a free space somewhere.
*/
args = state->args;
error = xfs_da_grow_inode(args, &blkno);
if (error)
return error;
dp = args->dp;
tp = args->trans;
error = xfs_da_get_buf(tp, dp, blkno, -1, &bp, args->whichfork);
if (error)
return error;
node = bp->b_addr;
oldroot = blk1->bp->b_addr;
if (oldroot->hdr.info.magic == cpu_to_be16(XFS_DA_NODE_MAGIC) ||
oldroot->hdr.info.magic == cpu_to_be16(XFS_DA3_NODE_MAGIC)) {
struct xfs_da3_icnode_hdr icnodehdr;
dp->d_ops->node_hdr_from_disk(&icnodehdr, oldroot);
btree = dp->d_ops->node_tree_p(oldroot);
size = (int)((char *)&btree[icnodehdr.count] - (char *)oldroot);
level = icnodehdr.level;
/*
* we are about to copy oldroot to bp, so set up the type
* of bp while we know exactly what it will be.
*/
xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DA_NODE_BUF);
} else {
struct xfs_dir3_icleaf_hdr leafhdr;
struct xfs_dir2_leaf_entry *ents;
leaf = (xfs_dir2_leaf_t *)oldroot;
dp->d_ops->leaf_hdr_from_disk(&leafhdr, leaf);
ents = dp->d_ops->leaf_ents_p(leaf);
ASSERT(leafhdr.magic == XFS_DIR2_LEAFN_MAGIC ||
leafhdr.magic == XFS_DIR3_LEAFN_MAGIC);
size = (int)((char *)&ents[leafhdr.count] - (char *)leaf);
level = 0;
/*
* we are about to copy oldroot to bp, so set up the type
* of bp while we know exactly what it will be.
*/
xfs_trans_buf_set_type(tp, bp, XFS_BLFT_DIR_LEAFN_BUF);
}
/*
* we can copy most of the information in the node from one block to
* another, but for CRC enabled headers we have to make sure that the
* block specific identifiers are kept intact. We update the buffer
* directly for this.
*/
memcpy(node, oldroot, size);
if (oldroot->hdr.info.magic == cpu_to_be16(XFS_DA3_NODE_MAGIC) ||
oldroot->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC)) {
struct xfs_da3_intnode *node3 = (struct xfs_da3_intnode *)node;
node3->hdr.info.blkno = cpu_to_be64(bp->b_bn);
}
xfs_trans_log_buf(tp, bp, 0, size - 1);
bp->b_ops = blk1->bp->b_ops;
xfs_trans_buf_copy_type(bp, blk1->bp);
blk1->bp = bp;
blk1->blkno = blkno;
/*
* Set up the new root node.
*/
error = xfs_da3_node_create(args,
(args->whichfork == XFS_DATA_FORK) ? args->geo->leafblk : 0,
level + 1, &bp, args->whichfork);
if (error)
return error;
node = bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
btree = dp->d_ops->node_tree_p(node);
btree[0].hashval = cpu_to_be32(blk1->hashval);
btree[0].before = cpu_to_be32(blk1->blkno);
btree[1].hashval = cpu_to_be32(blk2->hashval);
btree[1].before = cpu_to_be32(blk2->blkno);
nodehdr.count = 2;
dp->d_ops->node_hdr_to_disk(node, &nodehdr);
#ifdef DEBUG
if (oldroot->hdr.info.magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
oldroot->hdr.info.magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC)) {
ASSERT(blk1->blkno >= args->geo->leafblk &&
blk1->blkno < args->geo->freeblk);
ASSERT(blk2->blkno >= args->geo->leafblk &&
blk2->blkno < args->geo->freeblk);
}
#endif
/* Header is already logged by xfs_da_node_create */
xfs_trans_log_buf(tp, bp,
XFS_DA_LOGRANGE(node, btree, sizeof(xfs_da_node_entry_t) * 2));
return 0;
}
/*
* Split the node, rebalance, then add the new entry.
*/
STATIC int /* error */
xfs_da3_node_split(
struct xfs_da_state *state,
struct xfs_da_state_blk *oldblk,
struct xfs_da_state_blk *newblk,
struct xfs_da_state_blk *addblk,
int treelevel,
int *result)
{
struct xfs_da_intnode *node;
struct xfs_da3_icnode_hdr nodehdr;
xfs_dablk_t blkno;
int newcount;
int error;
int useextra;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_node_split(state->args);
node = oldblk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
/*
* With V2 dirs the extra block is data or freespace.
*/
useextra = state->extravalid && state->args->whichfork == XFS_ATTR_FORK;
newcount = 1 + useextra;
/*
* Do we have to split the node?
*/
if (nodehdr.count + newcount > state->args->geo->node_ents) {
/*
* Allocate a new node, add to the doubly linked chain of
* nodes, then move some of our excess entries into it.
*/
error = xfs_da_grow_inode(state->args, &blkno);
if (error)
return error; /* GROT: dir is inconsistent */
error = xfs_da3_node_create(state->args, blkno, treelevel,
&newblk->bp, state->args->whichfork);
if (error)
return error; /* GROT: dir is inconsistent */
newblk->blkno = blkno;
newblk->magic = XFS_DA_NODE_MAGIC;
xfs_da3_node_rebalance(state, oldblk, newblk);
error = xfs_da3_blk_link(state, oldblk, newblk);
if (error)
return error;
*result = 1;
} else {
*result = 0;
}
/*
* Insert the new entry(s) into the correct block
* (updating last hashval in the process).
*
* xfs_da3_node_add() inserts BEFORE the given index,
* and as a result of using node_lookup_int() we always
* point to a valid entry (not after one), but a split
* operation always results in a new block whose hashvals
* FOLLOW the current block.
*
* If we had double-split op below us, then add the extra block too.
*/
node = oldblk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
if (oldblk->index <= nodehdr.count) {
oldblk->index++;
xfs_da3_node_add(state, oldblk, addblk);
if (useextra) {
if (state->extraafter)
oldblk->index++;
xfs_da3_node_add(state, oldblk, &state->extrablk);
state->extravalid = 0;
}
} else {
newblk->index++;
xfs_da3_node_add(state, newblk, addblk);
if (useextra) {
if (state->extraafter)
newblk->index++;
xfs_da3_node_add(state, newblk, &state->extrablk);
state->extravalid = 0;
}
}
return 0;
}
/*
* Balance the btree elements between two intermediate nodes,
* usually one full and one empty.
*
* NOTE: if blk2 is empty, then it will get the upper half of blk1.
*/
STATIC void
xfs_da3_node_rebalance(
struct xfs_da_state *state,
struct xfs_da_state_blk *blk1,
struct xfs_da_state_blk *blk2)
{
struct xfs_da_intnode *node1;
struct xfs_da_intnode *node2;
struct xfs_da_intnode *tmpnode;
struct xfs_da_node_entry *btree1;
struct xfs_da_node_entry *btree2;
struct xfs_da_node_entry *btree_s;
struct xfs_da_node_entry *btree_d;
struct xfs_da3_icnode_hdr nodehdr1;
struct xfs_da3_icnode_hdr nodehdr2;
struct xfs_trans *tp;
int count;
int tmp;
int swap = 0;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_node_rebalance(state->args);
node1 = blk1->bp->b_addr;
node2 = blk2->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr1, node1);
dp->d_ops->node_hdr_from_disk(&nodehdr2, node2);
btree1 = dp->d_ops->node_tree_p(node1);
btree2 = dp->d_ops->node_tree_p(node2);
/*
* Figure out how many entries need to move, and in which direction.
* Swap the nodes around if that makes it simpler.
*/
if (nodehdr1.count > 0 && nodehdr2.count > 0 &&
((be32_to_cpu(btree2[0].hashval) < be32_to_cpu(btree1[0].hashval)) ||
(be32_to_cpu(btree2[nodehdr2.count - 1].hashval) <
be32_to_cpu(btree1[nodehdr1.count - 1].hashval)))) {
tmpnode = node1;
node1 = node2;
node2 = tmpnode;
dp->d_ops->node_hdr_from_disk(&nodehdr1, node1);
dp->d_ops->node_hdr_from_disk(&nodehdr2, node2);
btree1 = dp->d_ops->node_tree_p(node1);
btree2 = dp->d_ops->node_tree_p(node2);
swap = 1;
}
count = (nodehdr1.count - nodehdr2.count) / 2;
if (count == 0)
return;
tp = state->args->trans;
/*
* Two cases: high-to-low and low-to-high.
*/
if (count > 0) {
/*
* Move elements in node2 up to make a hole.
*/
tmp = nodehdr2.count;
if (tmp > 0) {
tmp *= (uint)sizeof(xfs_da_node_entry_t);
btree_s = &btree2[0];
btree_d = &btree2[count];
memmove(btree_d, btree_s, tmp);
}
/*
* Move the req'd B-tree elements from high in node1 to
* low in node2.
*/
nodehdr2.count += count;
tmp = count * (uint)sizeof(xfs_da_node_entry_t);
btree_s = &btree1[nodehdr1.count - count];
btree_d = &btree2[0];
memcpy(btree_d, btree_s, tmp);
nodehdr1.count -= count;
} else {
/*
* Move the req'd B-tree elements from low in node2 to
* high in node1.
*/
count = -count;
tmp = count * (uint)sizeof(xfs_da_node_entry_t);
btree_s = &btree2[0];
btree_d = &btree1[nodehdr1.count];
memcpy(btree_d, btree_s, tmp);
nodehdr1.count += count;
xfs_trans_log_buf(tp, blk1->bp,
XFS_DA_LOGRANGE(node1, btree_d, tmp));
/*
* Move elements in node2 down to fill the hole.
*/
tmp = nodehdr2.count - count;
tmp *= (uint)sizeof(xfs_da_node_entry_t);
btree_s = &btree2[count];
btree_d = &btree2[0];
memmove(btree_d, btree_s, tmp);
nodehdr2.count -= count;
}
/*
* Log header of node 1 and all current bits of node 2.
*/
dp->d_ops->node_hdr_to_disk(node1, &nodehdr1);
xfs_trans_log_buf(tp, blk1->bp,
XFS_DA_LOGRANGE(node1, &node1->hdr, dp->d_ops->node_hdr_size));
dp->d_ops->node_hdr_to_disk(node2, &nodehdr2);
xfs_trans_log_buf(tp, blk2->bp,
XFS_DA_LOGRANGE(node2, &node2->hdr,
dp->d_ops->node_hdr_size +
(sizeof(btree2[0]) * nodehdr2.count)));
/*
* Record the last hashval from each block for upward propagation.
* (note: don't use the swapped node pointers)
*/
if (swap) {
node1 = blk1->bp->b_addr;
node2 = blk2->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr1, node1);
dp->d_ops->node_hdr_from_disk(&nodehdr2, node2);
btree1 = dp->d_ops->node_tree_p(node1);
btree2 = dp->d_ops->node_tree_p(node2);
}
blk1->hashval = be32_to_cpu(btree1[nodehdr1.count - 1].hashval);
blk2->hashval = be32_to_cpu(btree2[nodehdr2.count - 1].hashval);
/*
* Adjust the expected index for insertion.
*/
if (blk1->index >= nodehdr1.count) {
blk2->index = blk1->index - nodehdr1.count;
blk1->index = nodehdr1.count + 1; /* make it invalid */
}
}
/*
* Add a new entry to an intermediate node.
*/
STATIC void
xfs_da3_node_add(
struct xfs_da_state *state,
struct xfs_da_state_blk *oldblk,
struct xfs_da_state_blk *newblk)
{
struct xfs_da_intnode *node;
struct xfs_da3_icnode_hdr nodehdr;
struct xfs_da_node_entry *btree;
int tmp;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_node_add(state->args);
node = oldblk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
btree = dp->d_ops->node_tree_p(node);
ASSERT(oldblk->index >= 0 && oldblk->index <= nodehdr.count);
ASSERT(newblk->blkno != 0);
if (state->args->whichfork == XFS_DATA_FORK)
ASSERT(newblk->blkno >= state->args->geo->leafblk &&
newblk->blkno < state->args->geo->freeblk);
/*
* We may need to make some room before we insert the new node.
*/
tmp = 0;
if (oldblk->index < nodehdr.count) {
tmp = (nodehdr.count - oldblk->index) * (uint)sizeof(*btree);
memmove(&btree[oldblk->index + 1], &btree[oldblk->index], tmp);
}
btree[oldblk->index].hashval = cpu_to_be32(newblk->hashval);
btree[oldblk->index].before = cpu_to_be32(newblk->blkno);
xfs_trans_log_buf(state->args->trans, oldblk->bp,
XFS_DA_LOGRANGE(node, &btree[oldblk->index],
tmp + sizeof(*btree)));
nodehdr.count += 1;
dp->d_ops->node_hdr_to_disk(node, &nodehdr);
xfs_trans_log_buf(state->args->trans, oldblk->bp,
XFS_DA_LOGRANGE(node, &node->hdr, dp->d_ops->node_hdr_size));
/*
* Copy the last hash value from the oldblk to propagate upwards.
*/
oldblk->hashval = be32_to_cpu(btree[nodehdr.count - 1].hashval);
}
/*========================================================================
* Routines used for shrinking the Btree.
*========================================================================*/
/*
* Deallocate an empty leaf node, remove it from its parent,
* possibly deallocating that block, etc...
*/
int
xfs_da3_join(
struct xfs_da_state *state)
{
struct xfs_da_state_blk *drop_blk;
struct xfs_da_state_blk *save_blk;
int action = 0;
int error;
trace_xfs_da_join(state->args);
drop_blk = &state->path.blk[ state->path.active-1 ];
save_blk = &state->altpath.blk[ state->path.active-1 ];
ASSERT(state->path.blk[0].magic == XFS_DA_NODE_MAGIC);
ASSERT(drop_blk->magic == XFS_ATTR_LEAF_MAGIC ||
drop_blk->magic == XFS_DIR2_LEAFN_MAGIC);
/*
* Walk back up the tree joining/deallocating as necessary.
* When we stop dropping blocks, break out.
*/
for ( ; state->path.active >= 2; drop_blk--, save_blk--,
state->path.active--) {
/*
* See if we can combine the block with a neighbor.
* (action == 0) => no options, just leave
* (action == 1) => coalesce, then unlink
* (action == 2) => block empty, unlink it
*/
switch (drop_blk->magic) {
case XFS_ATTR_LEAF_MAGIC:
error = xfs_attr3_leaf_toosmall(state, &action);
if (error)
return error;
if (action == 0)
return 0;
xfs_attr3_leaf_unbalance(state, drop_blk, save_blk);
break;
case XFS_DIR2_LEAFN_MAGIC:
error = xfs_dir2_leafn_toosmall(state, &action);
if (error)
return error;
if (action == 0)
return 0;
xfs_dir2_leafn_unbalance(state, drop_blk, save_blk);
break;
case XFS_DA_NODE_MAGIC:
/*
* Remove the offending node, fixup hashvals,
* check for a toosmall neighbor.
*/
xfs_da3_node_remove(state, drop_blk);
xfs_da3_fixhashpath(state, &state->path);
error = xfs_da3_node_toosmall(state, &action);
if (error)
return error;
if (action == 0)
return 0;
xfs_da3_node_unbalance(state, drop_blk, save_blk);
break;
}
xfs_da3_fixhashpath(state, &state->altpath);
error = xfs_da3_blk_unlink(state, drop_blk, save_blk);
xfs_da_state_kill_altpath(state);
if (error)
return error;
error = xfs_da_shrink_inode(state->args, drop_blk->blkno,
drop_blk->bp);
drop_blk->bp = NULL;
if (error)
return error;
}
/*
* We joined all the way to the top. If it turns out that
* we only have one entry in the root, make the child block
* the new root.
*/
xfs_da3_node_remove(state, drop_blk);
xfs_da3_fixhashpath(state, &state->path);
error = xfs_da3_root_join(state, &state->path.blk[0]);
return error;
}
#ifdef DEBUG
static void
xfs_da_blkinfo_onlychild_validate(struct xfs_da_blkinfo *blkinfo, __u16 level)
{
__be16 magic = blkinfo->magic;
if (level == 1) {
ASSERT(magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC) ||
magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC) ||
magic == cpu_to_be16(XFS_ATTR3_LEAF_MAGIC));
} else {
ASSERT(magic == cpu_to_be16(XFS_DA_NODE_MAGIC) ||
magic == cpu_to_be16(XFS_DA3_NODE_MAGIC));
}
ASSERT(!blkinfo->forw);
ASSERT(!blkinfo->back);
}
#else /* !DEBUG */
#define xfs_da_blkinfo_onlychild_validate(blkinfo, level)
#endif /* !DEBUG */
/*
* We have only one entry in the root. Copy the only remaining child of
* the old root to block 0 as the new root node.
*/
STATIC int
xfs_da3_root_join(
struct xfs_da_state *state,
struct xfs_da_state_blk *root_blk)
{
struct xfs_da_intnode *oldroot;
struct xfs_da_args *args;
xfs_dablk_t child;
struct xfs_buf *bp;
struct xfs_da3_icnode_hdr oldroothdr;
struct xfs_da_node_entry *btree;
int error;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_root_join(state->args);
ASSERT(root_blk->magic == XFS_DA_NODE_MAGIC);
args = state->args;
oldroot = root_blk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&oldroothdr, oldroot);
ASSERT(oldroothdr.forw == 0);
ASSERT(oldroothdr.back == 0);
/*
* If the root has more than one child, then don't do anything.
*/
if (oldroothdr.count > 1)
return 0;
/*
* Read in the (only) child block, then copy those bytes into
* the root block's buffer and free the original child block.
*/
btree = dp->d_ops->node_tree_p(oldroot);
child = be32_to_cpu(btree[0].before);
ASSERT(child != 0);
error = xfs_da3_node_read(args->trans, dp, child, -1, &bp,
args->whichfork);
if (error)
return error;
xfs_da_blkinfo_onlychild_validate(bp->b_addr, oldroothdr.level);
/*
* This could be copying a leaf back into the root block in the case of
* there only being a single leaf block left in the tree. Hence we have
* to update the b_ops pointer as well to match the buffer type change
* that could occur. For dir3 blocks we also need to update the block
* number in the buffer header.
*/
memcpy(root_blk->bp->b_addr, bp->b_addr, args->geo->blksize);
root_blk->bp->b_ops = bp->b_ops;
xfs_trans_buf_copy_type(root_blk->bp, bp);
if (oldroothdr.magic == XFS_DA3_NODE_MAGIC) {
struct xfs_da3_blkinfo *da3 = root_blk->bp->b_addr;
da3->blkno = cpu_to_be64(root_blk->bp->b_bn);
}
xfs_trans_log_buf(args->trans, root_blk->bp, 0,
args->geo->blksize - 1);
error = xfs_da_shrink_inode(args, child, bp);
return error;
}
/*
* Check a node block and its neighbors to see if the block should be
* collapsed into one or the other neighbor. Always keep the block
* with the smaller block number.
* If the current block is over 50% full, don't try to join it, return 0.
* If the block is empty, fill in the state structure and return 2.
* If it can be collapsed, fill in the state structure and return 1.
* If nothing can be done, return 0.
*/
STATIC int
xfs_da3_node_toosmall(
struct xfs_da_state *state,
int *action)
{
struct xfs_da_intnode *node;
struct xfs_da_state_blk *blk;
struct xfs_da_blkinfo *info;
xfs_dablk_t blkno;
struct xfs_buf *bp;
struct xfs_da3_icnode_hdr nodehdr;
int count;
int forward;
int error;
int retval;
int i;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_node_toosmall(state->args);
/*
* Check for the degenerate case of the block being over 50% full.
* If so, it's not worth even looking to see if we might be able
* to coalesce with a sibling.
*/
blk = &state->path.blk[ state->path.active-1 ];
info = blk->bp->b_addr;
node = (xfs_da_intnode_t *)info;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
if (nodehdr.count > (state->args->geo->node_ents >> 1)) {
*action = 0; /* blk over 50%, don't try to join */
return 0; /* blk over 50%, don't try to join */
}
/*
* Check for the degenerate case of the block being empty.
* If the block is empty, we'll simply delete it, no need to
* coalesce it with a sibling block. We choose (arbitrarily)
* to merge with the forward block unless it is NULL.
*/
if (nodehdr.count == 0) {
/*
* Make altpath point to the block we want to keep and
* path point to the block we want to drop (this one).
*/
forward = (info->forw != 0);
memcpy(&state->altpath, &state->path, sizeof(state->path));
error = xfs_da3_path_shift(state, &state->altpath, forward,
0, &retval);
if (error)
return error;
if (retval) {
*action = 0;
} else {
*action = 2;
}
return 0;
}
/*
* Examine each sibling block to see if we can coalesce with
* at least 25% free space to spare. We need to figure out
* whether to merge with the forward or the backward block.
* We prefer coalescing with the lower numbered sibling so as
* to shrink a directory over time.
*/
count = state->args->geo->node_ents;
count -= state->args->geo->node_ents >> 2;
count -= nodehdr.count;
/* start with smaller blk num */
forward = nodehdr.forw < nodehdr.back;
for (i = 0; i < 2; forward = !forward, i++) {
struct xfs_da3_icnode_hdr thdr;
if (forward)
blkno = nodehdr.forw;
else
blkno = nodehdr.back;
if (blkno == 0)
continue;
error = xfs_da3_node_read(state->args->trans, dp,
blkno, -1, &bp, state->args->whichfork);
if (error)
return error;
node = bp->b_addr;
dp->d_ops->node_hdr_from_disk(&thdr, node);
xfs_trans_brelse(state->args->trans, bp);
if (count - thdr.count >= 0)
break; /* fits with at least 25% to spare */
}
if (i >= 2) {
*action = 0;
return 0;
}
/*
* Make altpath point to the block we want to keep (the lower
* numbered block) and path point to the block we want to drop.
*/
memcpy(&state->altpath, &state->path, sizeof(state->path));
if (blkno < blk->blkno) {
error = xfs_da3_path_shift(state, &state->altpath, forward,
0, &retval);
} else {
error = xfs_da3_path_shift(state, &state->path, forward,
0, &retval);
}
if (error)
return error;
if (retval) {
*action = 0;
return 0;
}
*action = 1;
return 0;
}
/*
* Pick up the last hashvalue from an intermediate node.
*/
STATIC uint
xfs_da3_node_lasthash(
struct xfs_inode *dp,
struct xfs_buf *bp,
int *count)
{
struct xfs_da_intnode *node;
struct xfs_da_node_entry *btree;
struct xfs_da3_icnode_hdr nodehdr;
node = bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
if (count)
*count = nodehdr.count;
if (!nodehdr.count)
return 0;
btree = dp->d_ops->node_tree_p(node);
return be32_to_cpu(btree[nodehdr.count - 1].hashval);
}
/*
* Walk back up the tree adjusting hash values as necessary,
* when we stop making changes, return.
*/
void
xfs_da3_fixhashpath(
struct xfs_da_state *state,
struct xfs_da_state_path *path)
{
struct xfs_da_state_blk *blk;
struct xfs_da_intnode *node;
struct xfs_da_node_entry *btree;
xfs_dahash_t lasthash=0;
int level;
int count;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_fixhashpath(state->args);
level = path->active-1;
blk = &path->blk[ level ];
switch (blk->magic) {
case XFS_ATTR_LEAF_MAGIC:
lasthash = xfs_attr_leaf_lasthash(blk->bp, &count);
if (count == 0)
return;
break;
case XFS_DIR2_LEAFN_MAGIC:
lasthash = xfs_dir2_leaf_lasthash(dp, blk->bp, &count);
if (count == 0)
return;
break;
case XFS_DA_NODE_MAGIC:
lasthash = xfs_da3_node_lasthash(dp, blk->bp, &count);
if (count == 0)
return;
break;
}
for (blk--, level--; level >= 0; blk--, level--) {
struct xfs_da3_icnode_hdr nodehdr;
node = blk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
btree = dp->d_ops->node_tree_p(node);
xfs: fix directory hash ordering bug Commit f5ea1100 ("xfs: add CRCs to dir2/da node blocks") introduced in 3.10 incorrectly converted the btree hash index array pointer in xfs_da3_fixhashpath(). It resulted in the the current hash always being compared against the first entry in the btree rather than the current block index into the btree block's hash entry array. As a result, it was comparing the wrong hashes, and so could misorder the entries in the btree. For most cases, this doesn't cause any problems as it requires hash collisions to expose the ordering problem. However, when there are hash collisions within a directory there is a very good probability that the entries will be ordered incorrectly and that actually matters when duplicate hashes are placed into or removed from the btree block hash entry array. This bug results in an on-disk directory corruption and that results in directory verifier functions throwing corruption warnings into the logs. While no data or directory entries are lost, access to them may be compromised, and attempts to remove entries from a directory that has suffered from this corruption may result in a filesystem shutdown. xfs_repair will fix the directory hash ordering without data loss occuring. [dchinner: wrote useful a commit message] cc: <stable@vger.kernel.org> Reported-by: Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by: Mark Tinguely <tinguely@sgi.com> Reviewed-by: Ben Myers <bpm@sgi.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2014-04-04 00:10:49 +04:00
if (be32_to_cpu(btree[blk->index].hashval) == lasthash)
break;
blk->hashval = lasthash;
btree[blk->index].hashval = cpu_to_be32(lasthash);
xfs_trans_log_buf(state->args->trans, blk->bp,
XFS_DA_LOGRANGE(node, &btree[blk->index],
sizeof(*btree)));
lasthash = be32_to_cpu(btree[nodehdr.count - 1].hashval);
}
}
/*
* Remove an entry from an intermediate node.
*/
STATIC void
xfs_da3_node_remove(
struct xfs_da_state *state,
struct xfs_da_state_blk *drop_blk)
{
struct xfs_da_intnode *node;
struct xfs_da3_icnode_hdr nodehdr;
struct xfs_da_node_entry *btree;
int index;
int tmp;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_node_remove(state->args);
node = drop_blk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
ASSERT(drop_blk->index < nodehdr.count);
ASSERT(drop_blk->index >= 0);
/*
* Copy over the offending entry, or just zero it out.
*/
index = drop_blk->index;
btree = dp->d_ops->node_tree_p(node);
if (index < nodehdr.count - 1) {
tmp = nodehdr.count - index - 1;
tmp *= (uint)sizeof(xfs_da_node_entry_t);
memmove(&btree[index], &btree[index + 1], tmp);
xfs_trans_log_buf(state->args->trans, drop_blk->bp,
XFS_DA_LOGRANGE(node, &btree[index], tmp));
index = nodehdr.count - 1;
}
memset(&btree[index], 0, sizeof(xfs_da_node_entry_t));
xfs_trans_log_buf(state->args->trans, drop_blk->bp,
XFS_DA_LOGRANGE(node, &btree[index], sizeof(btree[index])));
nodehdr.count -= 1;
dp->d_ops->node_hdr_to_disk(node, &nodehdr);
xfs_trans_log_buf(state->args->trans, drop_blk->bp,
XFS_DA_LOGRANGE(node, &node->hdr, dp->d_ops->node_hdr_size));
/*
* Copy the last hash value from the block to propagate upwards.
*/
drop_blk->hashval = be32_to_cpu(btree[index - 1].hashval);
}
/*
* Unbalance the elements between two intermediate nodes,
* move all Btree elements from one node into another.
*/
STATIC void
xfs_da3_node_unbalance(
struct xfs_da_state *state,
struct xfs_da_state_blk *drop_blk,
struct xfs_da_state_blk *save_blk)
{
struct xfs_da_intnode *drop_node;
struct xfs_da_intnode *save_node;
struct xfs_da_node_entry *drop_btree;
struct xfs_da_node_entry *save_btree;
struct xfs_da3_icnode_hdr drop_hdr;
struct xfs_da3_icnode_hdr save_hdr;
struct xfs_trans *tp;
int sindex;
int tmp;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_node_unbalance(state->args);
drop_node = drop_blk->bp->b_addr;
save_node = save_blk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&drop_hdr, drop_node);
dp->d_ops->node_hdr_from_disk(&save_hdr, save_node);
drop_btree = dp->d_ops->node_tree_p(drop_node);
save_btree = dp->d_ops->node_tree_p(save_node);
tp = state->args->trans;
/*
* If the dying block has lower hashvals, then move all the
* elements in the remaining block up to make a hole.
*/
if ((be32_to_cpu(drop_btree[0].hashval) <
be32_to_cpu(save_btree[0].hashval)) ||
(be32_to_cpu(drop_btree[drop_hdr.count - 1].hashval) <
be32_to_cpu(save_btree[save_hdr.count - 1].hashval))) {
/* XXX: check this - is memmove dst correct? */
tmp = save_hdr.count * sizeof(xfs_da_node_entry_t);
memmove(&save_btree[drop_hdr.count], &save_btree[0], tmp);
sindex = 0;
xfs_trans_log_buf(tp, save_blk->bp,
XFS_DA_LOGRANGE(save_node, &save_btree[0],
(save_hdr.count + drop_hdr.count) *
sizeof(xfs_da_node_entry_t)));
} else {
sindex = save_hdr.count;
xfs_trans_log_buf(tp, save_blk->bp,
XFS_DA_LOGRANGE(save_node, &save_btree[sindex],
drop_hdr.count * sizeof(xfs_da_node_entry_t)));
}
/*
* Move all the B-tree elements from drop_blk to save_blk.
*/
tmp = drop_hdr.count * (uint)sizeof(xfs_da_node_entry_t);
memcpy(&save_btree[sindex], &drop_btree[0], tmp);
save_hdr.count += drop_hdr.count;
dp->d_ops->node_hdr_to_disk(save_node, &save_hdr);
xfs_trans_log_buf(tp, save_blk->bp,
XFS_DA_LOGRANGE(save_node, &save_node->hdr,
dp->d_ops->node_hdr_size));
/*
* Save the last hashval in the remaining block for upward propagation.
*/
save_blk->hashval = be32_to_cpu(save_btree[save_hdr.count - 1].hashval);
}
/*========================================================================
* Routines used for finding things in the Btree.
*========================================================================*/
/*
* Walk down the Btree looking for a particular filename, filling
* in the state structure as we go.
*
* We will set the state structure to point to each of the elements
* in each of the nodes where either the hashval is or should be.
*
* We support duplicate hashval's so for each entry in the current
* node that could contain the desired hashval, descend. This is a
* pruned depth-first tree search.
*/
int /* error */
xfs_da3_node_lookup_int(
struct xfs_da_state *state,
int *result)
{
struct xfs_da_state_blk *blk;
struct xfs_da_blkinfo *curr;
struct xfs_da_intnode *node;
struct xfs_da_node_entry *btree;
struct xfs_da3_icnode_hdr nodehdr;
struct xfs_da_args *args;
xfs_dablk_t blkno;
xfs_dahash_t hashval;
xfs_dahash_t btreehashval;
int probe;
int span;
int max;
int error;
int retval;
unsigned int expected_level = 0;
struct xfs_inode *dp = state->args->dp;
args = state->args;
/*
* Descend thru the B-tree searching each level for the right
* node to use, until the right hashval is found.
*/
blkno = args->geo->leafblk;
for (blk = &state->path.blk[0], state->path.active = 1;
state->path.active <= XFS_DA_NODE_MAXDEPTH;
blk++, state->path.active++) {
/*
* Read the next node down in the tree.
*/
blk->blkno = blkno;
error = xfs_da3_node_read(args->trans, args->dp, blkno,
-1, &blk->bp, args->whichfork);
if (error) {
blk->blkno = 0;
state->path.active--;
return error;
}
curr = blk->bp->b_addr;
blk->magic = be16_to_cpu(curr->magic);
if (blk->magic == XFS_ATTR_LEAF_MAGIC ||
blk->magic == XFS_ATTR3_LEAF_MAGIC) {
blk->magic = XFS_ATTR_LEAF_MAGIC;
blk->hashval = xfs_attr_leaf_lasthash(blk->bp, NULL);
break;
}
if (blk->magic == XFS_DIR2_LEAFN_MAGIC ||
blk->magic == XFS_DIR3_LEAFN_MAGIC) {
blk->magic = XFS_DIR2_LEAFN_MAGIC;
blk->hashval = xfs_dir2_leaf_lasthash(args->dp,
blk->bp, NULL);
break;
}
blk->magic = XFS_DA_NODE_MAGIC;
/*
* Search an intermediate node for a match.
*/
node = blk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
btree = dp->d_ops->node_tree_p(node);
/* Tree taller than we can handle; bail out! */
if (nodehdr.level >= XFS_DA_NODE_MAXDEPTH)
return -EFSCORRUPTED;
/* Check the level from the root. */
if (blkno == args->geo->leafblk)
expected_level = nodehdr.level - 1;
else if (expected_level != nodehdr.level)
return -EFSCORRUPTED;
else
expected_level--;
max = nodehdr.count;
blk->hashval = be32_to_cpu(btree[max - 1].hashval);
/*
* Binary search. (note: small blocks will skip loop)
*/
probe = span = max / 2;
hashval = args->hashval;
while (span > 4) {
span /= 2;
btreehashval = be32_to_cpu(btree[probe].hashval);
if (btreehashval < hashval)
probe += span;
else if (btreehashval > hashval)
probe -= span;
else
break;
}
ASSERT((probe >= 0) && (probe < max));
ASSERT((span <= 4) ||
(be32_to_cpu(btree[probe].hashval) == hashval));
/*
* Since we may have duplicate hashval's, find the first
* matching hashval in the node.
*/
while (probe > 0 &&
be32_to_cpu(btree[probe].hashval) >= hashval) {
probe--;
}
while (probe < max &&
be32_to_cpu(btree[probe].hashval) < hashval) {
probe++;
}
/*
* Pick the right block to descend on.
*/
if (probe == max) {
blk->index = max - 1;
blkno = be32_to_cpu(btree[max - 1].before);
} else {
blk->index = probe;
blkno = be32_to_cpu(btree[probe].before);
}
/* We can't point back to the root. */
if (blkno == args->geo->leafblk)
return -EFSCORRUPTED;
}
if (expected_level != 0)
return -EFSCORRUPTED;
/*
* A leaf block that ends in the hashval that we are interested in
* (final hashval == search hashval) means that the next block may
* contain more entries with the same hashval, shift upward to the
* next leaf and keep searching.
*/
for (;;) {
if (blk->magic == XFS_DIR2_LEAFN_MAGIC) {
retval = xfs_dir2_leafn_lookup_int(blk->bp, args,
&blk->index, state);
} else if (blk->magic == XFS_ATTR_LEAF_MAGIC) {
retval = xfs_attr3_leaf_lookup_int(blk->bp, args);
blk->index = args->index;
args->blkno = blk->blkno;
} else {
ASSERT(0);
return -EFSCORRUPTED;
}
if (((retval == -ENOENT) || (retval == -ENOATTR)) &&
(blk->hashval == args->hashval)) {
error = xfs_da3_path_shift(state, &state->path, 1, 1,
&retval);
if (error)
return error;
if (retval == 0) {
continue;
} else if (blk->magic == XFS_ATTR_LEAF_MAGIC) {
/* path_shift() gives ENOENT */
retval = -ENOATTR;
}
}
break;
}
*result = retval;
return 0;
}
/*========================================================================
* Utility routines.
*========================================================================*/
/*
* Compare two intermediate nodes for "order".
*/
STATIC int
xfs_da3_node_order(
struct xfs_inode *dp,
struct xfs_buf *node1_bp,
struct xfs_buf *node2_bp)
{
struct xfs_da_intnode *node1;
struct xfs_da_intnode *node2;
struct xfs_da_node_entry *btree1;
struct xfs_da_node_entry *btree2;
struct xfs_da3_icnode_hdr node1hdr;
struct xfs_da3_icnode_hdr node2hdr;
node1 = node1_bp->b_addr;
node2 = node2_bp->b_addr;
dp->d_ops->node_hdr_from_disk(&node1hdr, node1);
dp->d_ops->node_hdr_from_disk(&node2hdr, node2);
btree1 = dp->d_ops->node_tree_p(node1);
btree2 = dp->d_ops->node_tree_p(node2);
if (node1hdr.count > 0 && node2hdr.count > 0 &&
((be32_to_cpu(btree2[0].hashval) < be32_to_cpu(btree1[0].hashval)) ||
(be32_to_cpu(btree2[node2hdr.count - 1].hashval) <
be32_to_cpu(btree1[node1hdr.count - 1].hashval)))) {
return 1;
}
return 0;
}
/*
* Link a new block into a doubly linked list of blocks (of whatever type).
*/
int /* error */
xfs_da3_blk_link(
struct xfs_da_state *state,
struct xfs_da_state_blk *old_blk,
struct xfs_da_state_blk *new_blk)
{
struct xfs_da_blkinfo *old_info;
struct xfs_da_blkinfo *new_info;
struct xfs_da_blkinfo *tmp_info;
struct xfs_da_args *args;
struct xfs_buf *bp;
int before = 0;
int error;
struct xfs_inode *dp = state->args->dp;
/*
* Set up environment.
*/
args = state->args;
ASSERT(args != NULL);
old_info = old_blk->bp->b_addr;
new_info = new_blk->bp->b_addr;
ASSERT(old_blk->magic == XFS_DA_NODE_MAGIC ||
old_blk->magic == XFS_DIR2_LEAFN_MAGIC ||
old_blk->magic == XFS_ATTR_LEAF_MAGIC);
switch (old_blk->magic) {
case XFS_ATTR_LEAF_MAGIC:
before = xfs_attr_leaf_order(old_blk->bp, new_blk->bp);
break;
case XFS_DIR2_LEAFN_MAGIC:
before = xfs_dir2_leafn_order(dp, old_blk->bp, new_blk->bp);
break;
case XFS_DA_NODE_MAGIC:
before = xfs_da3_node_order(dp, old_blk->bp, new_blk->bp);
break;
}
/*
* Link blocks in appropriate order.
*/
if (before) {
/*
* Link new block in before existing block.
*/
trace_xfs_da_link_before(args);
new_info->forw = cpu_to_be32(old_blk->blkno);
new_info->back = old_info->back;
if (old_info->back) {
error = xfs_da3_node_read(args->trans, dp,
be32_to_cpu(old_info->back),
-1, &bp, args->whichfork);
if (error)
return error;
ASSERT(bp != NULL);
tmp_info = bp->b_addr;
ASSERT(tmp_info->magic == old_info->magic);
ASSERT(be32_to_cpu(tmp_info->forw) == old_blk->blkno);
tmp_info->forw = cpu_to_be32(new_blk->blkno);
xfs_trans_log_buf(args->trans, bp, 0, sizeof(*tmp_info)-1);
}
old_info->back = cpu_to_be32(new_blk->blkno);
} else {
/*
* Link new block in after existing block.
*/
trace_xfs_da_link_after(args);
new_info->forw = old_info->forw;
new_info->back = cpu_to_be32(old_blk->blkno);
if (old_info->forw) {
error = xfs_da3_node_read(args->trans, dp,
be32_to_cpu(old_info->forw),
-1, &bp, args->whichfork);
if (error)
return error;
ASSERT(bp != NULL);
tmp_info = bp->b_addr;
ASSERT(tmp_info->magic == old_info->magic);
ASSERT(be32_to_cpu(tmp_info->back) == old_blk->blkno);
tmp_info->back = cpu_to_be32(new_blk->blkno);
xfs_trans_log_buf(args->trans, bp, 0, sizeof(*tmp_info)-1);
}
old_info->forw = cpu_to_be32(new_blk->blkno);
}
xfs_trans_log_buf(args->trans, old_blk->bp, 0, sizeof(*tmp_info) - 1);
xfs_trans_log_buf(args->trans, new_blk->bp, 0, sizeof(*tmp_info) - 1);
return 0;
}
/*
* Unlink a block from a doubly linked list of blocks.
*/
STATIC int /* error */
xfs_da3_blk_unlink(
struct xfs_da_state *state,
struct xfs_da_state_blk *drop_blk,
struct xfs_da_state_blk *save_blk)
{
struct xfs_da_blkinfo *drop_info;
struct xfs_da_blkinfo *save_info;
struct xfs_da_blkinfo *tmp_info;
struct xfs_da_args *args;
struct xfs_buf *bp;
int error;
/*
* Set up environment.
*/
args = state->args;
ASSERT(args != NULL);
save_info = save_blk->bp->b_addr;
drop_info = drop_blk->bp->b_addr;
ASSERT(save_blk->magic == XFS_DA_NODE_MAGIC ||
save_blk->magic == XFS_DIR2_LEAFN_MAGIC ||
save_blk->magic == XFS_ATTR_LEAF_MAGIC);
ASSERT(save_blk->magic == drop_blk->magic);
ASSERT((be32_to_cpu(save_info->forw) == drop_blk->blkno) ||
(be32_to_cpu(save_info->back) == drop_blk->blkno));
ASSERT((be32_to_cpu(drop_info->forw) == save_blk->blkno) ||
(be32_to_cpu(drop_info->back) == save_blk->blkno));
/*
* Unlink the leaf block from the doubly linked chain of leaves.
*/
if (be32_to_cpu(save_info->back) == drop_blk->blkno) {
trace_xfs_da_unlink_back(args);
save_info->back = drop_info->back;
if (drop_info->back) {
error = xfs_da3_node_read(args->trans, args->dp,
be32_to_cpu(drop_info->back),
-1, &bp, args->whichfork);
if (error)
return error;
ASSERT(bp != NULL);
tmp_info = bp->b_addr;
ASSERT(tmp_info->magic == save_info->magic);
ASSERT(be32_to_cpu(tmp_info->forw) == drop_blk->blkno);
tmp_info->forw = cpu_to_be32(save_blk->blkno);
xfs_trans_log_buf(args->trans, bp, 0,
sizeof(*tmp_info) - 1);
}
} else {
trace_xfs_da_unlink_forward(args);
save_info->forw = drop_info->forw;
if (drop_info->forw) {
error = xfs_da3_node_read(args->trans, args->dp,
be32_to_cpu(drop_info->forw),
-1, &bp, args->whichfork);
if (error)
return error;
ASSERT(bp != NULL);
tmp_info = bp->b_addr;
ASSERT(tmp_info->magic == save_info->magic);
ASSERT(be32_to_cpu(tmp_info->back) == drop_blk->blkno);
tmp_info->back = cpu_to_be32(save_blk->blkno);
xfs_trans_log_buf(args->trans, bp, 0,
sizeof(*tmp_info) - 1);
}
}
xfs_trans_log_buf(args->trans, save_blk->bp, 0, sizeof(*save_info) - 1);
return 0;
}
/*
* Move a path "forward" or "!forward" one block at the current level.
*
* This routine will adjust a "path" to point to the next block
* "forward" (higher hashvalues) or "!forward" (lower hashvals) in the
* Btree, including updating pointers to the intermediate nodes between
* the new bottom and the root.
*/
int /* error */
xfs_da3_path_shift(
struct xfs_da_state *state,
struct xfs_da_state_path *path,
int forward,
int release,
int *result)
{
struct xfs_da_state_blk *blk;
struct xfs_da_blkinfo *info;
struct xfs_da_intnode *node;
struct xfs_da_args *args;
struct xfs_da_node_entry *btree;
struct xfs_da3_icnode_hdr nodehdr;
xfs: swap leaf buffer into path struct atomically during path shift The node directory lookup code uses a state structure that tracks the path of buffers used to search for the hash of a filename through the leaf blocks. When the lookup encounters a block that ends with the requested hash, but the entry has not yet been found, it must shift over to the next block and continue looking for the entry (i.e., duplicate hashes could continue over into the next block). This shift mechanism involves walking back up and down the state structure, replacing buffers at the appropriate btree levels as necessary. When a buffer is replaced, the old buffer is released and the new buffer read into the active slot in the path structure. Because the buffer is read directly into the path slot, a buffer read failure can result in setting a NULL buffer pointer in an active slot. This throws off the state cleanup code in xfs_dir2_node_lookup(), which expects to release a buffer from each active slot. Instead, a BUG occurs due to a NULL pointer dereference: BUG: unable to handle kernel NULL pointer dereference at 00000000000001e8 IP: [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... RIP: 0010:[<ffffffffa0585063>] [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... Call Trace: [<ffffffffa05250c6>] xfs_dir2_node_lookup+0xa6/0x2c0 [xfs] [<ffffffffa0519f7c>] xfs_dir_lookup+0x1ac/0x1c0 [xfs] [<ffffffffa055d0e1>] xfs_lookup+0x91/0x290 [xfs] [<ffffffffa05580b3>] xfs_vn_lookup+0x73/0xb0 [xfs] [<ffffffff8122de8d>] lookup_real+0x1d/0x50 [<ffffffff8123330e>] path_openat+0x91e/0x1490 [<ffffffff81235079>] do_filp_open+0x89/0x100 ... This has been reproduced via a parallel fsstress and filesystem shutdown workload in a loop. The shutdown triggers the read error in the aforementioned codepath and causes the BUG in xfs_dir2_node_lookup(). Update xfs_da3_path_shift() to update the active path slot atomically with respect to the caller when a buffer is replaced. This ensures that the caller always sees the old or new buffer in the slot and prevents the NULL pointer dereference. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 03:32:33 +03:00
struct xfs_buf *bp;
xfs_dablk_t blkno = 0;
int level;
int error;
struct xfs_inode *dp = state->args->dp;
trace_xfs_da_path_shift(state->args);
/*
* Roll up the Btree looking for the first block where our
* current index is not at the edge of the block. Note that
* we skip the bottom layer because we want the sibling block.
*/
args = state->args;
ASSERT(args != NULL);
ASSERT(path != NULL);
ASSERT((path->active > 0) && (path->active < XFS_DA_NODE_MAXDEPTH));
level = (path->active-1) - 1; /* skip bottom layer in path */
for (blk = &path->blk[level]; level >= 0; blk--, level--) {
node = blk->bp->b_addr;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
btree = dp->d_ops->node_tree_p(node);
if (forward && (blk->index < nodehdr.count - 1)) {
blk->index++;
blkno = be32_to_cpu(btree[blk->index].before);
break;
} else if (!forward && (blk->index > 0)) {
blk->index--;
blkno = be32_to_cpu(btree[blk->index].before);
break;
}
}
if (level < 0) {
*result = -ENOENT; /* we're out of our tree */
ASSERT(args->op_flags & XFS_DA_OP_OKNOENT);
return 0;
}
/*
* Roll down the edge of the subtree until we reach the
* same depth we were at originally.
*/
for (blk++, level++; level < path->active; blk++, level++) {
/*
xfs: swap leaf buffer into path struct atomically during path shift The node directory lookup code uses a state structure that tracks the path of buffers used to search for the hash of a filename through the leaf blocks. When the lookup encounters a block that ends with the requested hash, but the entry has not yet been found, it must shift over to the next block and continue looking for the entry (i.e., duplicate hashes could continue over into the next block). This shift mechanism involves walking back up and down the state structure, replacing buffers at the appropriate btree levels as necessary. When a buffer is replaced, the old buffer is released and the new buffer read into the active slot in the path structure. Because the buffer is read directly into the path slot, a buffer read failure can result in setting a NULL buffer pointer in an active slot. This throws off the state cleanup code in xfs_dir2_node_lookup(), which expects to release a buffer from each active slot. Instead, a BUG occurs due to a NULL pointer dereference: BUG: unable to handle kernel NULL pointer dereference at 00000000000001e8 IP: [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... RIP: 0010:[<ffffffffa0585063>] [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... Call Trace: [<ffffffffa05250c6>] xfs_dir2_node_lookup+0xa6/0x2c0 [xfs] [<ffffffffa0519f7c>] xfs_dir_lookup+0x1ac/0x1c0 [xfs] [<ffffffffa055d0e1>] xfs_lookup+0x91/0x290 [xfs] [<ffffffffa05580b3>] xfs_vn_lookup+0x73/0xb0 [xfs] [<ffffffff8122de8d>] lookup_real+0x1d/0x50 [<ffffffff8123330e>] path_openat+0x91e/0x1490 [<ffffffff81235079>] do_filp_open+0x89/0x100 ... This has been reproduced via a parallel fsstress and filesystem shutdown workload in a loop. The shutdown triggers the read error in the aforementioned codepath and causes the BUG in xfs_dir2_node_lookup(). Update xfs_da3_path_shift() to update the active path slot atomically with respect to the caller when a buffer is replaced. This ensures that the caller always sees the old or new buffer in the slot and prevents the NULL pointer dereference. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 03:32:33 +03:00
* Read the next child block into a local buffer.
*/
xfs: swap leaf buffer into path struct atomically during path shift The node directory lookup code uses a state structure that tracks the path of buffers used to search for the hash of a filename through the leaf blocks. When the lookup encounters a block that ends with the requested hash, but the entry has not yet been found, it must shift over to the next block and continue looking for the entry (i.e., duplicate hashes could continue over into the next block). This shift mechanism involves walking back up and down the state structure, replacing buffers at the appropriate btree levels as necessary. When a buffer is replaced, the old buffer is released and the new buffer read into the active slot in the path structure. Because the buffer is read directly into the path slot, a buffer read failure can result in setting a NULL buffer pointer in an active slot. This throws off the state cleanup code in xfs_dir2_node_lookup(), which expects to release a buffer from each active slot. Instead, a BUG occurs due to a NULL pointer dereference: BUG: unable to handle kernel NULL pointer dereference at 00000000000001e8 IP: [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... RIP: 0010:[<ffffffffa0585063>] [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... Call Trace: [<ffffffffa05250c6>] xfs_dir2_node_lookup+0xa6/0x2c0 [xfs] [<ffffffffa0519f7c>] xfs_dir_lookup+0x1ac/0x1c0 [xfs] [<ffffffffa055d0e1>] xfs_lookup+0x91/0x290 [xfs] [<ffffffffa05580b3>] xfs_vn_lookup+0x73/0xb0 [xfs] [<ffffffff8122de8d>] lookup_real+0x1d/0x50 [<ffffffff8123330e>] path_openat+0x91e/0x1490 [<ffffffff81235079>] do_filp_open+0x89/0x100 ... This has been reproduced via a parallel fsstress and filesystem shutdown workload in a loop. The shutdown triggers the read error in the aforementioned codepath and causes the BUG in xfs_dir2_node_lookup(). Update xfs_da3_path_shift() to update the active path slot atomically with respect to the caller when a buffer is replaced. This ensures that the caller always sees the old or new buffer in the slot and prevents the NULL pointer dereference. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 03:32:33 +03:00
error = xfs_da3_node_read(args->trans, dp, blkno, -1, &bp,
args->whichfork);
if (error)
return error;
/*
xfs: swap leaf buffer into path struct atomically during path shift The node directory lookup code uses a state structure that tracks the path of buffers used to search for the hash of a filename through the leaf blocks. When the lookup encounters a block that ends with the requested hash, but the entry has not yet been found, it must shift over to the next block and continue looking for the entry (i.e., duplicate hashes could continue over into the next block). This shift mechanism involves walking back up and down the state structure, replacing buffers at the appropriate btree levels as necessary. When a buffer is replaced, the old buffer is released and the new buffer read into the active slot in the path structure. Because the buffer is read directly into the path slot, a buffer read failure can result in setting a NULL buffer pointer in an active slot. This throws off the state cleanup code in xfs_dir2_node_lookup(), which expects to release a buffer from each active slot. Instead, a BUG occurs due to a NULL pointer dereference: BUG: unable to handle kernel NULL pointer dereference at 00000000000001e8 IP: [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... RIP: 0010:[<ffffffffa0585063>] [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... Call Trace: [<ffffffffa05250c6>] xfs_dir2_node_lookup+0xa6/0x2c0 [xfs] [<ffffffffa0519f7c>] xfs_dir_lookup+0x1ac/0x1c0 [xfs] [<ffffffffa055d0e1>] xfs_lookup+0x91/0x290 [xfs] [<ffffffffa05580b3>] xfs_vn_lookup+0x73/0xb0 [xfs] [<ffffffff8122de8d>] lookup_real+0x1d/0x50 [<ffffffff8123330e>] path_openat+0x91e/0x1490 [<ffffffff81235079>] do_filp_open+0x89/0x100 ... This has been reproduced via a parallel fsstress and filesystem shutdown workload in a loop. The shutdown triggers the read error in the aforementioned codepath and causes the BUG in xfs_dir2_node_lookup(). Update xfs_da3_path_shift() to update the active path slot atomically with respect to the caller when a buffer is replaced. This ensures that the caller always sees the old or new buffer in the slot and prevents the NULL pointer dereference. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 03:32:33 +03:00
* Release the old block (if it's dirty, the trans doesn't
* actually let go) and swap the local buffer into the path
* structure. This ensures failure of the above read doesn't set
* a NULL buffer in an active slot in the path.
*/
xfs: swap leaf buffer into path struct atomically during path shift The node directory lookup code uses a state structure that tracks the path of buffers used to search for the hash of a filename through the leaf blocks. When the lookup encounters a block that ends with the requested hash, but the entry has not yet been found, it must shift over to the next block and continue looking for the entry (i.e., duplicate hashes could continue over into the next block). This shift mechanism involves walking back up and down the state structure, replacing buffers at the appropriate btree levels as necessary. When a buffer is replaced, the old buffer is released and the new buffer read into the active slot in the path structure. Because the buffer is read directly into the path slot, a buffer read failure can result in setting a NULL buffer pointer in an active slot. This throws off the state cleanup code in xfs_dir2_node_lookup(), which expects to release a buffer from each active slot. Instead, a BUG occurs due to a NULL pointer dereference: BUG: unable to handle kernel NULL pointer dereference at 00000000000001e8 IP: [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... RIP: 0010:[<ffffffffa0585063>] [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... Call Trace: [<ffffffffa05250c6>] xfs_dir2_node_lookup+0xa6/0x2c0 [xfs] [<ffffffffa0519f7c>] xfs_dir_lookup+0x1ac/0x1c0 [xfs] [<ffffffffa055d0e1>] xfs_lookup+0x91/0x290 [xfs] [<ffffffffa05580b3>] xfs_vn_lookup+0x73/0xb0 [xfs] [<ffffffff8122de8d>] lookup_real+0x1d/0x50 [<ffffffff8123330e>] path_openat+0x91e/0x1490 [<ffffffff81235079>] do_filp_open+0x89/0x100 ... This has been reproduced via a parallel fsstress and filesystem shutdown workload in a loop. The shutdown triggers the read error in the aforementioned codepath and causes the BUG in xfs_dir2_node_lookup(). Update xfs_da3_path_shift() to update the active path slot atomically with respect to the caller when a buffer is replaced. This ensures that the caller always sees the old or new buffer in the slot and prevents the NULL pointer dereference. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 03:32:33 +03:00
if (release)
xfs_trans_brelse(args->trans, blk->bp);
blk->blkno = blkno;
xfs: swap leaf buffer into path struct atomically during path shift The node directory lookup code uses a state structure that tracks the path of buffers used to search for the hash of a filename through the leaf blocks. When the lookup encounters a block that ends with the requested hash, but the entry has not yet been found, it must shift over to the next block and continue looking for the entry (i.e., duplicate hashes could continue over into the next block). This shift mechanism involves walking back up and down the state structure, replacing buffers at the appropriate btree levels as necessary. When a buffer is replaced, the old buffer is released and the new buffer read into the active slot in the path structure. Because the buffer is read directly into the path slot, a buffer read failure can result in setting a NULL buffer pointer in an active slot. This throws off the state cleanup code in xfs_dir2_node_lookup(), which expects to release a buffer from each active slot. Instead, a BUG occurs due to a NULL pointer dereference: BUG: unable to handle kernel NULL pointer dereference at 00000000000001e8 IP: [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... RIP: 0010:[<ffffffffa0585063>] [<ffffffffa0585063>] xfs_trans_brelse+0x2a3/0x3c0 [xfs] ... Call Trace: [<ffffffffa05250c6>] xfs_dir2_node_lookup+0xa6/0x2c0 [xfs] [<ffffffffa0519f7c>] xfs_dir_lookup+0x1ac/0x1c0 [xfs] [<ffffffffa055d0e1>] xfs_lookup+0x91/0x290 [xfs] [<ffffffffa05580b3>] xfs_vn_lookup+0x73/0xb0 [xfs] [<ffffffff8122de8d>] lookup_real+0x1d/0x50 [<ffffffff8123330e>] path_openat+0x91e/0x1490 [<ffffffff81235079>] do_filp_open+0x89/0x100 ... This has been reproduced via a parallel fsstress and filesystem shutdown workload in a loop. The shutdown triggers the read error in the aforementioned codepath and causes the BUG in xfs_dir2_node_lookup(). Update xfs_da3_path_shift() to update the active path slot atomically with respect to the caller when a buffer is replaced. This ensures that the caller always sees the old or new buffer in the slot and prevents the NULL pointer dereference. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 03:32:33 +03:00
blk->bp = bp;
info = blk->bp->b_addr;
ASSERT(info->magic == cpu_to_be16(XFS_DA_NODE_MAGIC) ||
info->magic == cpu_to_be16(XFS_DA3_NODE_MAGIC) ||
info->magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
info->magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC) ||
info->magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC) ||
info->magic == cpu_to_be16(XFS_ATTR3_LEAF_MAGIC));
/*
* Note: we flatten the magic number to a single type so we
* don't have to compare against crc/non-crc types elsewhere.
*/
switch (be16_to_cpu(info->magic)) {
case XFS_DA_NODE_MAGIC:
case XFS_DA3_NODE_MAGIC:
blk->magic = XFS_DA_NODE_MAGIC;
node = (xfs_da_intnode_t *)info;
dp->d_ops->node_hdr_from_disk(&nodehdr, node);
btree = dp->d_ops->node_tree_p(node);
blk->hashval = be32_to_cpu(btree[nodehdr.count - 1].hashval);
if (forward)
blk->index = 0;
else
blk->index = nodehdr.count - 1;
blkno = be32_to_cpu(btree[blk->index].before);
break;
case XFS_ATTR_LEAF_MAGIC:
case XFS_ATTR3_LEAF_MAGIC:
blk->magic = XFS_ATTR_LEAF_MAGIC;
ASSERT(level == path->active-1);
blk->index = 0;
blk->hashval = xfs_attr_leaf_lasthash(blk->bp, NULL);
break;
case XFS_DIR2_LEAFN_MAGIC:
case XFS_DIR3_LEAFN_MAGIC:
blk->magic = XFS_DIR2_LEAFN_MAGIC;
ASSERT(level == path->active-1);
blk->index = 0;
blk->hashval = xfs_dir2_leaf_lasthash(args->dp,
blk->bp, NULL);
break;
default:
ASSERT(0);
break;
}
}
*result = 0;
return 0;
}
/*========================================================================
* Utility routines.
*========================================================================*/
/*
* Implement a simple hash on a character string.
* Rotate the hash value by 7 bits, then XOR each character in.
* This is implemented with some source-level loop unrolling.
*/
xfs_dahash_t
xfs_da_hashname(const uint8_t *name, int namelen)
{
xfs_dahash_t hash;
/*
* Do four characters at a time as long as we can.
*/
for (hash = 0; namelen >= 4; namelen -= 4, name += 4)
hash = (name[0] << 21) ^ (name[1] << 14) ^ (name[2] << 7) ^
(name[3] << 0) ^ rol32(hash, 7 * 4);
/*
* Now do the rest of the characters.
*/
switch (namelen) {
case 3:
return (name[0] << 14) ^ (name[1] << 7) ^ (name[2] << 0) ^
rol32(hash, 7 * 3);
case 2:
return (name[0] << 7) ^ (name[1] << 0) ^ rol32(hash, 7 * 2);
case 1:
return (name[0] << 0) ^ rol32(hash, 7 * 1);
default: /* case 0: */
return hash;
}
}
enum xfs_dacmp
xfs_da_compname(
struct xfs_da_args *args,
const unsigned char *name,
int len)
{
return (args->namelen == len && memcmp(args->name, name, len) == 0) ?
XFS_CMP_EXACT : XFS_CMP_DIFFERENT;
}
static xfs_dahash_t
xfs_default_hashname(
struct xfs_name *name)
{
return xfs_da_hashname(name->name, name->len);
}
const struct xfs_nameops xfs_default_nameops = {
.hashname = xfs_default_hashname,
.compname = xfs_da_compname
};
int
xfs_da_grow_inode_int(
struct xfs_da_args *args,
xfs_fileoff_t *bno,
int count)
{
struct xfs_trans *tp = args->trans;
struct xfs_inode *dp = args->dp;
int w = args->whichfork;
xfs_rfsblock_t nblks = dp->i_d.di_nblocks;
struct xfs_bmbt_irec map, *mapp;
int nmap, error, got, i, mapi;
/*
* Find a spot in the file space to put the new block.
*/
error = xfs_bmap_first_unused(tp, dp, count, bno, w);
if (error)
return error;
/*
* Try mapping it in one filesystem block.
*/
nmap = 1;
ASSERT(args->firstblock != NULL);
error = xfs_bmapi_write(tp, dp, *bno, count,
xfs_bmapi_aflag(w)|XFS_BMAPI_METADATA|XFS_BMAPI_CONTIG,
args->firstblock, args->total, &map, &nmap,
args->dfops);
if (error)
return error;
ASSERT(nmap <= 1);
if (nmap == 1) {
mapp = &map;
mapi = 1;
} else if (nmap == 0 && count > 1) {
xfs_fileoff_t b;
int c;
/*
* If we didn't get it and the block might work if fragmented,
* try without the CONTIG flag. Loop until we get it all.
*/
mapp = kmem_alloc(sizeof(*mapp) * count, KM_SLEEP);
for (b = *bno, mapi = 0; b < *bno + count; ) {
nmap = MIN(XFS_BMAP_MAX_NMAP, count);
c = (int)(*bno + count - b);
error = xfs_bmapi_write(tp, dp, b, c,
xfs_bmapi_aflag(w)|XFS_BMAPI_METADATA,
args->firstblock, args->total,
&mapp[mapi], &nmap, args->dfops);
if (error)
goto out_free_map;
if (nmap < 1)
break;
mapi += nmap;
b = mapp[mapi - 1].br_startoff +
mapp[mapi - 1].br_blockcount;
}
} else {
mapi = 0;
mapp = NULL;
}
/*
* Count the blocks we got, make sure it matches the total.
*/
for (i = 0, got = 0; i < mapi; i++)
got += mapp[i].br_blockcount;
if (got != count || mapp[0].br_startoff != *bno ||
mapp[mapi - 1].br_startoff + mapp[mapi - 1].br_blockcount !=
*bno + count) {
error = -ENOSPC;
goto out_free_map;
}
2008-10-30 09:38:12 +03:00
/* account for newly allocated blocks in reserved blocks total */
args->total -= dp->i_d.di_nblocks - nblks;
out_free_map:
if (mapp != &map)
kmem_free(mapp);
return error;
}
/*
* Add a block to the btree ahead of the file.
* Return the new block number to the caller.
*/
int
xfs_da_grow_inode(
struct xfs_da_args *args,
xfs_dablk_t *new_blkno)
{
xfs_fileoff_t bno;
int error;
trace_xfs_da_grow_inode(args);
bno = args->geo->leafblk;
error = xfs_da_grow_inode_int(args, &bno, args->geo->fsbcount);
if (!error)
*new_blkno = (xfs_dablk_t)bno;
return error;
}
/*
* Ick. We need to always be able to remove a btree block, even
* if there's no space reservation because the filesystem is full.
* This is called if xfs_bunmapi on a btree block fails due to ENOSPC.
* It swaps the target block with the last block in the file. The
* last block in the file can always be removed since it can't cause
* a bmap btree split to do that.
*/
STATIC int
xfs_da3_swap_lastblock(
struct xfs_da_args *args,
xfs_dablk_t *dead_blknop,
struct xfs_buf **dead_bufp)
{
struct xfs_da_blkinfo *dead_info;
struct xfs_da_blkinfo *sib_info;
struct xfs_da_intnode *par_node;
struct xfs_da_intnode *dead_node;
struct xfs_dir2_leaf *dead_leaf2;
struct xfs_da_node_entry *btree;
struct xfs_da3_icnode_hdr par_hdr;
struct xfs_inode *dp;
struct xfs_trans *tp;
struct xfs_mount *mp;
struct xfs_buf *dead_buf;
struct xfs_buf *last_buf;
struct xfs_buf *sib_buf;
struct xfs_buf *par_buf;
xfs_dahash_t dead_hash;
xfs_fileoff_t lastoff;
xfs_dablk_t dead_blkno;
xfs_dablk_t last_blkno;
xfs_dablk_t sib_blkno;
xfs_dablk_t par_blkno;
int error;
int w;
int entno;
int level;
int dead_level;
trace_xfs_da_swap_lastblock(args);
dead_buf = *dead_bufp;
dead_blkno = *dead_blknop;
tp = args->trans;
dp = args->dp;
w = args->whichfork;
ASSERT(w == XFS_DATA_FORK);
mp = dp->i_mount;
lastoff = args->geo->freeblk;
error = xfs_bmap_last_before(tp, dp, &lastoff, w);
if (error)
return error;
if (unlikely(lastoff == 0)) {
XFS_ERROR_REPORT("xfs_da_swap_lastblock(1)", XFS_ERRLEVEL_LOW,
mp);
return -EFSCORRUPTED;
}
/*
* Read the last block in the btree space.
*/
last_blkno = (xfs_dablk_t)lastoff - args->geo->fsbcount;
error = xfs_da3_node_read(tp, dp, last_blkno, -1, &last_buf, w);
if (error)
return error;
/*
* Copy the last block into the dead buffer and log it.
*/
memcpy(dead_buf->b_addr, last_buf->b_addr, args->geo->blksize);
xfs_trans_log_buf(tp, dead_buf, 0, args->geo->blksize - 1);
dead_info = dead_buf->b_addr;
/*
* Get values from the moved block.
*/
if (dead_info->magic == cpu_to_be16(XFS_DIR2_LEAFN_MAGIC) ||
dead_info->magic == cpu_to_be16(XFS_DIR3_LEAFN_MAGIC)) {
struct xfs_dir3_icleaf_hdr leafhdr;
struct xfs_dir2_leaf_entry *ents;
dead_leaf2 = (xfs_dir2_leaf_t *)dead_info;
dp->d_ops->leaf_hdr_from_disk(&leafhdr, dead_leaf2);
ents = dp->d_ops->leaf_ents_p(dead_leaf2);
dead_level = 0;
dead_hash = be32_to_cpu(ents[leafhdr.count - 1].hashval);
} else {
struct xfs_da3_icnode_hdr deadhdr;
dead_node = (xfs_da_intnode_t *)dead_info;
dp->d_ops->node_hdr_from_disk(&deadhdr, dead_node);
btree = dp->d_ops->node_tree_p(dead_node);
dead_level = deadhdr.level;
dead_hash = be32_to_cpu(btree[deadhdr.count - 1].hashval);
}
sib_buf = par_buf = NULL;
/*
* If the moved block has a left sibling, fix up the pointers.
*/
if ((sib_blkno = be32_to_cpu(dead_info->back))) {
error = xfs_da3_node_read(tp, dp, sib_blkno, -1, &sib_buf, w);
if (error)
goto done;
sib_info = sib_buf->b_addr;
if (unlikely(
be32_to_cpu(sib_info->forw) != last_blkno ||
sib_info->magic != dead_info->magic)) {
XFS_ERROR_REPORT("xfs_da_swap_lastblock(2)",
XFS_ERRLEVEL_LOW, mp);
error = -EFSCORRUPTED;
goto done;
}
sib_info->forw = cpu_to_be32(dead_blkno);
xfs_trans_log_buf(tp, sib_buf,
XFS_DA_LOGRANGE(sib_info, &sib_info->forw,
sizeof(sib_info->forw)));
sib_buf = NULL;
}
/*
* If the moved block has a right sibling, fix up the pointers.
*/
if ((sib_blkno = be32_to_cpu(dead_info->forw))) {
error = xfs_da3_node_read(tp, dp, sib_blkno, -1, &sib_buf, w);
if (error)
goto done;
sib_info = sib_buf->b_addr;
if (unlikely(
be32_to_cpu(sib_info->back) != last_blkno ||
sib_info->magic != dead_info->magic)) {
XFS_ERROR_REPORT("xfs_da_swap_lastblock(3)",
XFS_ERRLEVEL_LOW, mp);
error = -EFSCORRUPTED;
goto done;
}
sib_info->back = cpu_to_be32(dead_blkno);
xfs_trans_log_buf(tp, sib_buf,
XFS_DA_LOGRANGE(sib_info, &sib_info->back,
sizeof(sib_info->back)));
sib_buf = NULL;
}
par_blkno = args->geo->leafblk;
level = -1;
/*
* Walk down the tree looking for the parent of the moved block.
*/
for (;;) {
error = xfs_da3_node_read(tp, dp, par_blkno, -1, &par_buf, w);
if (error)
goto done;
par_node = par_buf->b_addr;
dp->d_ops->node_hdr_from_disk(&par_hdr, par_node);
if (level >= 0 && level != par_hdr.level + 1) {
XFS_ERROR_REPORT("xfs_da_swap_lastblock(4)",
XFS_ERRLEVEL_LOW, mp);
error = -EFSCORRUPTED;
goto done;
}
level = par_hdr.level;
btree = dp->d_ops->node_tree_p(par_node);
for (entno = 0;
entno < par_hdr.count &&
be32_to_cpu(btree[entno].hashval) < dead_hash;
entno++)
continue;
if (entno == par_hdr.count) {
XFS_ERROR_REPORT("xfs_da_swap_lastblock(5)",
XFS_ERRLEVEL_LOW, mp);
error = -EFSCORRUPTED;
goto done;
}
par_blkno = be32_to_cpu(btree[entno].before);
if (level == dead_level + 1)
break;
xfs_trans_brelse(tp, par_buf);
par_buf = NULL;
}
/*
* We're in the right parent block.
* Look for the right entry.
*/
for (;;) {
for (;
entno < par_hdr.count &&
be32_to_cpu(btree[entno].before) != last_blkno;
entno++)
continue;
if (entno < par_hdr.count)
break;
par_blkno = par_hdr.forw;
xfs_trans_brelse(tp, par_buf);
par_buf = NULL;
if (unlikely(par_blkno == 0)) {
XFS_ERROR_REPORT("xfs_da_swap_lastblock(6)",
XFS_ERRLEVEL_LOW, mp);
error = -EFSCORRUPTED;
goto done;
}
error = xfs_da3_node_read(tp, dp, par_blkno, -1, &par_buf, w);
if (error)
goto done;
par_node = par_buf->b_addr;
dp->d_ops->node_hdr_from_disk(&par_hdr, par_node);
if (par_hdr.level != level) {
XFS_ERROR_REPORT("xfs_da_swap_lastblock(7)",
XFS_ERRLEVEL_LOW, mp);
error = -EFSCORRUPTED;
goto done;
}
btree = dp->d_ops->node_tree_p(par_node);
entno = 0;
}
/*
* Update the parent entry pointing to the moved block.
*/
btree[entno].before = cpu_to_be32(dead_blkno);
xfs_trans_log_buf(tp, par_buf,
XFS_DA_LOGRANGE(par_node, &btree[entno].before,
sizeof(btree[entno].before)));
*dead_blknop = last_blkno;
*dead_bufp = last_buf;
return 0;
done:
if (par_buf)
xfs_trans_brelse(tp, par_buf);
if (sib_buf)
xfs_trans_brelse(tp, sib_buf);
xfs_trans_brelse(tp, last_buf);
return error;
}
/*
* Remove a btree block from a directory or attribute.
*/
int
xfs_da_shrink_inode(
xfs_da_args_t *args,
xfs_dablk_t dead_blkno,
struct xfs_buf *dead_buf)
{
xfs_inode_t *dp;
int done, error, w, count;
xfs_trans_t *tp;
trace_xfs_da_shrink_inode(args);
dp = args->dp;
w = args->whichfork;
tp = args->trans;
count = args->geo->fsbcount;
for (;;) {
/*
* Remove extents. If we get ENOSPC for a dir we have to move
* the last block to the place we want to kill.
*/
error = xfs_bunmapi(tp, dp, dead_blkno, count,
xfs_bmapi_aflag(w), 0, args->firstblock,
args->dfops, &done);
if (error == -ENOSPC) {
if (w != XFS_DATA_FORK)
break;
error = xfs_da3_swap_lastblock(args, &dead_blkno,
&dead_buf);
if (error)
break;
} else {
break;
}
}
xfs_trans_binval(tp, dead_buf);
return error;
}
/*
* See if the mapping(s) for this btree block are valid, i.e.
* don't contain holes, are logically contiguous, and cover the whole range.
*/
STATIC int
xfs_da_map_covers_blocks(
int nmap,
xfs_bmbt_irec_t *mapp,
xfs_dablk_t bno,
int count)
{
int i;
xfs_fileoff_t off;
for (i = 0, off = bno; i < nmap; i++) {
if (mapp[i].br_startblock == HOLESTARTBLOCK ||
mapp[i].br_startblock == DELAYSTARTBLOCK) {
return 0;
}
if (off != mapp[i].br_startoff) {
return 0;
}
off += mapp[i].br_blockcount;
}
return off == bno + count;
}
/*
* Convert a struct xfs_bmbt_irec to a struct xfs_buf_map.
*
* For the single map case, it is assumed that the caller has provided a pointer
* to a valid xfs_buf_map. For the multiple map case, this function will
* allocate the xfs_buf_map to hold all the maps and replace the caller's single
* map pointer with the allocated map.
*/
static int
xfs_buf_map_from_irec(
struct xfs_mount *mp,
struct xfs_buf_map **mapp,
int *nmaps,
struct xfs_bmbt_irec *irecs,
int nirecs)
{
struct xfs_buf_map *map;
int i;
ASSERT(*nmaps == 1);
ASSERT(nirecs >= 1);
if (nirecs > 1) {
map = kmem_zalloc(nirecs * sizeof(struct xfs_buf_map),
KM_SLEEP | KM_NOFS);
if (!map)
return -ENOMEM;
*mapp = map;
}
*nmaps = nirecs;
map = *mapp;
for (i = 0; i < *nmaps; i++) {
ASSERT(irecs[i].br_startblock != DELAYSTARTBLOCK &&
irecs[i].br_startblock != HOLESTARTBLOCK);
map[i].bm_bn = XFS_FSB_TO_DADDR(mp, irecs[i].br_startblock);
map[i].bm_len = XFS_FSB_TO_BB(mp, irecs[i].br_blockcount);
}
return 0;
}
/*
* Map the block we are given ready for reading. There are three possible return
* values:
* -1 - will be returned if we land in a hole and mappedbno == -2 so the
* caller knows not to execute a subsequent read.
* 0 - if we mapped the block successfully
* >0 - positive error number if there was an error.
*/
static int
xfs_dabuf_map(
struct xfs_inode *dp,
xfs_dablk_t bno,
xfs_daddr_t mappedbno,
int whichfork,
struct xfs_buf_map **map,
int *nmaps)
{
struct xfs_mount *mp = dp->i_mount;
int nfsb;
int error = 0;
struct xfs_bmbt_irec irec;
struct xfs_bmbt_irec *irecs = &irec;
int nirecs;
ASSERT(map && *map);
ASSERT(*nmaps == 1);
if (whichfork == XFS_DATA_FORK)
nfsb = mp->m_dir_geo->fsbcount;
else
nfsb = mp->m_attr_geo->fsbcount;
/*
* Caller doesn't have a mapping. -2 means don't complain
* if we land in a hole.
*/
if (mappedbno == -1 || mappedbno == -2) {
/*
* Optimize the one-block case.
*/
if (nfsb != 1)
irecs = kmem_zalloc(sizeof(irec) * nfsb,
KM_SLEEP | KM_NOFS);
nirecs = nfsb;
error = xfs_bmapi_read(dp, (xfs_fileoff_t)bno, nfsb, irecs,
&nirecs, xfs_bmapi_aflag(whichfork));
if (error)
goto out;
} else {
irecs->br_startblock = XFS_DADDR_TO_FSB(mp, mappedbno);
irecs->br_startoff = (xfs_fileoff_t)bno;
irecs->br_blockcount = nfsb;
irecs->br_state = 0;
nirecs = 1;
}
if (!xfs_da_map_covers_blocks(nirecs, irecs, bno, nfsb)) {
error = mappedbno == -2 ? -1 : -EFSCORRUPTED;
if (unlikely(error == -EFSCORRUPTED)) {
if (xfs_error_level >= XFS_ERRLEVEL_LOW) {
int i;
xfs_alert(mp, "%s: bno %lld dir: inode %lld",
__func__, (long long)bno,
(long long)dp->i_ino);
for (i = 0; i < *nmaps; i++) {
xfs_alert(mp,
"[%02d] br_startoff %lld br_startblock %lld br_blockcount %lld br_state %d",
i,
(long long)irecs[i].br_startoff,
(long long)irecs[i].br_startblock,
(long long)irecs[i].br_blockcount,
irecs[i].br_state);
}
}
XFS_ERROR_REPORT("xfs_da_do_buf(1)",
XFS_ERRLEVEL_LOW, mp);
}
goto out;
}
error = xfs_buf_map_from_irec(mp, map, nmaps, irecs, nirecs);
out:
if (irecs != &irec)
kmem_free(irecs);
return error;
}
/*
* Get a buffer for the dir/attr block.
*/
int
xfs_da_get_buf(
struct xfs_trans *trans,
struct xfs_inode *dp,
xfs_dablk_t bno,
xfs_daddr_t mappedbno,
struct xfs_buf **bpp,
int whichfork)
{
struct xfs_buf *bp;
struct xfs_buf_map map;
struct xfs_buf_map *mapp;
int nmap;
int error;
*bpp = NULL;
mapp = &map;
nmap = 1;
error = xfs_dabuf_map(dp, bno, mappedbno, whichfork,
&mapp, &nmap);
if (error) {
/* mapping a hole is not an error, but we don't continue */
if (error == -1)
error = 0;
goto out_free;
}
bp = xfs_trans_get_buf_map(trans, dp->i_mount->m_ddev_targp,
mapp, nmap, 0);
error = bp ? bp->b_error : -EIO;
if (error) {
if (bp)
xfs_trans_brelse(trans, bp);
goto out_free;
}
*bpp = bp;
out_free:
if (mapp != &map)
kmem_free(mapp);
return error;
}
/*
* Get a buffer for the dir/attr block, fill in the contents.
*/
int
xfs_da_read_buf(
struct xfs_trans *trans,
struct xfs_inode *dp,
xfs_dablk_t bno,
xfs_daddr_t mappedbno,
struct xfs_buf **bpp,
int whichfork,
const struct xfs_buf_ops *ops)
{
struct xfs_buf *bp;
struct xfs_buf_map map;
struct xfs_buf_map *mapp;
int nmap;
int error;
*bpp = NULL;
mapp = &map;
nmap = 1;
error = xfs_dabuf_map(dp, bno, mappedbno, whichfork,
&mapp, &nmap);
if (error) {
/* mapping a hole is not an error, but we don't continue */
if (error == -1)
error = 0;
goto out_free;
}
error = xfs_trans_read_buf_map(dp->i_mount, trans,
dp->i_mount->m_ddev_targp,
mapp, nmap, 0, &bp, ops);
if (error)
goto out_free;
if (whichfork == XFS_ATTR_FORK)
xfs_buf_set_ref(bp, XFS_ATTR_BTREE_REF);
else
xfs_buf_set_ref(bp, XFS_DIR_BTREE_REF);
*bpp = bp;
out_free:
if (mapp != &map)
kmem_free(mapp);
return error;
}
/*
* Readahead the dir/attr block.
*/
int
xfs_da_reada_buf(
struct xfs_inode *dp,
xfs_dablk_t bno,
xfs_daddr_t mappedbno,
int whichfork,
const struct xfs_buf_ops *ops)
{
struct xfs_buf_map map;
struct xfs_buf_map *mapp;
int nmap;
int error;
mapp = &map;
nmap = 1;
error = xfs_dabuf_map(dp, bno, mappedbno, whichfork,
&mapp, &nmap);
if (error) {
/* mapping a hole is not an error, but we don't continue */
if (error == -1)
error = 0;
goto out_free;
}
mappedbno = mapp[0].bm_bn;
xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap, ops);
out_free:
if (mapp != &map)
kmem_free(mapp);
return error;
}