linux-stable/fs/xfs/xfs_attr_inactive.c

400 lines
9.8 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (c) 2000-2005 Silicon Graphics, Inc.
* Copyright (c) 2013 Red Hat, Inc.
* All Rights Reserved.
*/
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_shared.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_bit.h"
#include "xfs_mount.h"
#include "xfs_da_format.h"
#include "xfs_da_btree.h"
#include "xfs_inode.h"
xfs: Add delay ready attr remove routines This patch modifies the attr remove routines to be delay ready. This means they no longer roll or commit transactions, but instead return -EAGAIN to have the calling routine roll and refresh the transaction. In this series, xfs_attr_remove_args is merged with xfs_attr_node_removename become a new function, xfs_attr_remove_iter. This new version uses a sort of state machine like switch to keep track of where it was when EAGAIN was returned. A new version of xfs_attr_remove_args consists of a simple loop to refresh the transaction until the operation is completed. A new XFS_DAC_DEFER_FINISH flag is used to finish the transaction where ever the existing code used to. Calls to xfs_attr_rmtval_remove are replaced with the delay ready version __xfs_attr_rmtval_remove. We will rename __xfs_attr_rmtval_remove back to xfs_attr_rmtval_remove when we are done. xfs_attr_rmtval_remove itself is still in use by the set routines (used during a rename). For reasons of preserving existing function, we modify xfs_attr_rmtval_remove to call xfs_defer_finish when the flag is set. Similar to how xfs_attr_remove_args does here. Once we transition the set routines to be delay ready, xfs_attr_rmtval_remove is no longer used and will be removed. This patch also adds a new struct xfs_delattr_context, which we will use to keep track of the current state of an attribute operation. The new xfs_delattr_state enum is used to track various operations that are in progress so that we know not to repeat them, and resume where we left off before EAGAIN was returned to cycle out the transaction. Other members take the place of local variables that need to retain their values across multiple function calls. See xfs_attr.h for a more detailed diagram of the states. Signed-off-by: Allison Henderson <allison.henderson@oracle.com> Reviewed-by: Chandan Babu R <chandanrlinux@gmail.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org>
2021-04-26 22:00:33 +00:00
#include "xfs_attr.h"
#include "xfs_attr_remote.h"
#include "xfs_trans.h"
#include "xfs_bmap.h"
#include "xfs_attr_leaf.h"
#include "xfs_quota.h"
#include "xfs_dir2.h"
#include "xfs_error.h"
/*
xfs: fix memory corruption during remote attr value buffer invalidation While running generic/103, I observed what looks like memory corruption and (with slub debugging turned on) a slub redzone warning on i386 when inactivating an inode with a 64k remote attr value. On a v5 filesystem, maximally sized remote attr values require one block more than 64k worth of space to hold both the remote attribute value header (64 bytes). On a 4k block filesystem this results in a 68k buffer; on a 64k block filesystem, this would be a 128k buffer. Note that even though we'll never use more than 65,600 bytes of this buffer, XFS_MAX_BLOCKSIZE is 64k. This is a problem because the definition of struct xfs_buf_log_format allows for XFS_MAX_BLOCKSIZE worth of dirty bitmap (64k). On i386 when we invalidate a remote attribute, xfs_trans_binval zeroes all 68k worth of the dirty map, writing right off the end of the log item and corrupting memory. We've gotten away with this on x86_64 for years because the compiler inserts a u32 padding on the end of struct xfs_buf_log_format. Fortunately for us, remote attribute values are written to disk with xfs_bwrite(), which is to say that they are not logged. Fix the problem by removing all places where we could end up creating a buffer log item for a remote attribute value and leave a note explaining why. Next, replace the open-coded buffer invalidation with a call to the helper we created in the previous patch that does better checking for bad metadata before marking the buffer stale. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-01-08 00:11:45 +00:00
* Invalidate any incore buffers associated with this remote attribute value
* extent. We never log remote attribute value buffers, which means that they
* won't be attached to a transaction and are therefore safe to mark stale.
* The actual bunmapi will be taken care of later.
*/
STATIC int
xfs: fix memory corruption during remote attr value buffer invalidation While running generic/103, I observed what looks like memory corruption and (with slub debugging turned on) a slub redzone warning on i386 when inactivating an inode with a 64k remote attr value. On a v5 filesystem, maximally sized remote attr values require one block more than 64k worth of space to hold both the remote attribute value header (64 bytes). On a 4k block filesystem this results in a 68k buffer; on a 64k block filesystem, this would be a 128k buffer. Note that even though we'll never use more than 65,600 bytes of this buffer, XFS_MAX_BLOCKSIZE is 64k. This is a problem because the definition of struct xfs_buf_log_format allows for XFS_MAX_BLOCKSIZE worth of dirty bitmap (64k). On i386 when we invalidate a remote attribute, xfs_trans_binval zeroes all 68k worth of the dirty map, writing right off the end of the log item and corrupting memory. We've gotten away with this on x86_64 for years because the compiler inserts a u32 padding on the end of struct xfs_buf_log_format. Fortunately for us, remote attribute values are written to disk with xfs_bwrite(), which is to say that they are not logged. Fix the problem by removing all places where we could end up creating a buffer log item for a remote attribute value and leave a note explaining why. Next, replace the open-coded buffer invalidation with a call to the helper we created in the previous patch that does better checking for bad metadata before marking the buffer stale. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-01-08 00:11:45 +00:00
xfs_attr3_rmt_stale(
struct xfs_inode *dp,
xfs_dablk_t blkno,
int blkcnt)
{
struct xfs_bmbt_irec map;
int nmap;
int error;
/*
* Roll through the "value", invalidating the attribute value's
* blocks.
*/
while (blkcnt > 0) {
/*
* Try to remember where we decided to put the value.
*/
nmap = 1;
error = xfs_bmapi_read(dp, (xfs_fileoff_t)blkno, blkcnt,
&map, &nmap, XFS_BMAPI_ATTRFORK);
xfs: fix memory corruption during remote attr value buffer invalidation While running generic/103, I observed what looks like memory corruption and (with slub debugging turned on) a slub redzone warning on i386 when inactivating an inode with a 64k remote attr value. On a v5 filesystem, maximally sized remote attr values require one block more than 64k worth of space to hold both the remote attribute value header (64 bytes). On a 4k block filesystem this results in a 68k buffer; on a 64k block filesystem, this would be a 128k buffer. Note that even though we'll never use more than 65,600 bytes of this buffer, XFS_MAX_BLOCKSIZE is 64k. This is a problem because the definition of struct xfs_buf_log_format allows for XFS_MAX_BLOCKSIZE worth of dirty bitmap (64k). On i386 when we invalidate a remote attribute, xfs_trans_binval zeroes all 68k worth of the dirty map, writing right off the end of the log item and corrupting memory. We've gotten away with this on x86_64 for years because the compiler inserts a u32 padding on the end of struct xfs_buf_log_format. Fortunately for us, remote attribute values are written to disk with xfs_bwrite(), which is to say that they are not logged. Fix the problem by removing all places where we could end up creating a buffer log item for a remote attribute value and leave a note explaining why. Next, replace the open-coded buffer invalidation with a call to the helper we created in the previous patch that does better checking for bad metadata before marking the buffer stale. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-01-08 00:11:45 +00:00
if (error)
return error;
xfs: fix memory corruption during remote attr value buffer invalidation While running generic/103, I observed what looks like memory corruption and (with slub debugging turned on) a slub redzone warning on i386 when inactivating an inode with a 64k remote attr value. On a v5 filesystem, maximally sized remote attr values require one block more than 64k worth of space to hold both the remote attribute value header (64 bytes). On a 4k block filesystem this results in a 68k buffer; on a 64k block filesystem, this would be a 128k buffer. Note that even though we'll never use more than 65,600 bytes of this buffer, XFS_MAX_BLOCKSIZE is 64k. This is a problem because the definition of struct xfs_buf_log_format allows for XFS_MAX_BLOCKSIZE worth of dirty bitmap (64k). On i386 when we invalidate a remote attribute, xfs_trans_binval zeroes all 68k worth of the dirty map, writing right off the end of the log item and corrupting memory. We've gotten away with this on x86_64 for years because the compiler inserts a u32 padding on the end of struct xfs_buf_log_format. Fortunately for us, remote attribute values are written to disk with xfs_bwrite(), which is to say that they are not logged. Fix the problem by removing all places where we could end up creating a buffer log item for a remote attribute value and leave a note explaining why. Next, replace the open-coded buffer invalidation with a call to the helper we created in the previous patch that does better checking for bad metadata before marking the buffer stale. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-01-08 00:11:45 +00:00
if (XFS_IS_CORRUPT(dp->i_mount, nmap != 1))
return -EFSCORRUPTED;
/*
xfs: fix memory corruption during remote attr value buffer invalidation While running generic/103, I observed what looks like memory corruption and (with slub debugging turned on) a slub redzone warning on i386 when inactivating an inode with a 64k remote attr value. On a v5 filesystem, maximally sized remote attr values require one block more than 64k worth of space to hold both the remote attribute value header (64 bytes). On a 4k block filesystem this results in a 68k buffer; on a 64k block filesystem, this would be a 128k buffer. Note that even though we'll never use more than 65,600 bytes of this buffer, XFS_MAX_BLOCKSIZE is 64k. This is a problem because the definition of struct xfs_buf_log_format allows for XFS_MAX_BLOCKSIZE worth of dirty bitmap (64k). On i386 when we invalidate a remote attribute, xfs_trans_binval zeroes all 68k worth of the dirty map, writing right off the end of the log item and corrupting memory. We've gotten away with this on x86_64 for years because the compiler inserts a u32 padding on the end of struct xfs_buf_log_format. Fortunately for us, remote attribute values are written to disk with xfs_bwrite(), which is to say that they are not logged. Fix the problem by removing all places where we could end up creating a buffer log item for a remote attribute value and leave a note explaining why. Next, replace the open-coded buffer invalidation with a call to the helper we created in the previous patch that does better checking for bad metadata before marking the buffer stale. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-01-08 00:11:45 +00:00
* Mark any incore buffers for the remote value as stale. We
* never log remote attr value buffers, so the buffer should be
* easy to kill.
*/
xfs: fix memory corruption during remote attr value buffer invalidation While running generic/103, I observed what looks like memory corruption and (with slub debugging turned on) a slub redzone warning on i386 when inactivating an inode with a 64k remote attr value. On a v5 filesystem, maximally sized remote attr values require one block more than 64k worth of space to hold both the remote attribute value header (64 bytes). On a 4k block filesystem this results in a 68k buffer; on a 64k block filesystem, this would be a 128k buffer. Note that even though we'll never use more than 65,600 bytes of this buffer, XFS_MAX_BLOCKSIZE is 64k. This is a problem because the definition of struct xfs_buf_log_format allows for XFS_MAX_BLOCKSIZE worth of dirty bitmap (64k). On i386 when we invalidate a remote attribute, xfs_trans_binval zeroes all 68k worth of the dirty map, writing right off the end of the log item and corrupting memory. We've gotten away with this on x86_64 for years because the compiler inserts a u32 padding on the end of struct xfs_buf_log_format. Fortunately for us, remote attribute values are written to disk with xfs_bwrite(), which is to say that they are not logged. Fix the problem by removing all places where we could end up creating a buffer log item for a remote attribute value and leave a note explaining why. Next, replace the open-coded buffer invalidation with a call to the helper we created in the previous patch that does better checking for bad metadata before marking the buffer stale. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-01-08 00:11:45 +00:00
error = xfs_attr_rmtval_stale(dp, &map, 0);
if (error)
return error;
blkno += map.br_blockcount;
blkcnt -= map.br_blockcount;
}
return 0;
}
/*
* Invalidate all of the "remote" value regions pointed to by a particular
* leaf block.
* Note that we must release the lock on the buffer so that we are not
* caught holding something that the logging code wants to flush to disk.
*/
STATIC int
xfs_attr3_leaf_inactive(
struct xfs_trans **trans,
struct xfs_inode *dp,
struct xfs_buf *bp)
{
struct xfs_attr3_icleaf_hdr ichdr;
struct xfs_mount *mp = bp->b_mount;
struct xfs_attr_leafblock *leaf = bp->b_addr;
struct xfs_attr_leaf_entry *entry;
struct xfs_attr_leaf_name_remote *name_rmt;
int error = 0;
int i;
xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf);
/*
* Find the remote value extents for this leaf and invalidate their
* incore buffers.
*/
entry = xfs_attr3_leaf_entryp(leaf);
for (i = 0; i < ichdr.count; entry++, i++) {
int blkcnt;
if (!entry->nameidx || (entry->flags & XFS_ATTR_LOCAL))
continue;
name_rmt = xfs_attr3_leaf_name_remote(leaf, i);
if (!name_rmt->valueblk)
continue;
blkcnt = xfs_attr3_rmt_blocks(dp->i_mount,
be32_to_cpu(name_rmt->valuelen));
error = xfs_attr3_rmt_stale(dp,
be32_to_cpu(name_rmt->valueblk), blkcnt);
if (error)
goto err;
}
xfs_trans_brelse(*trans, bp);
err:
return error;
}
/*
* Recurse (gasp!) through the attribute nodes until we find leaves.
* We're doing a depth-first traversal in order to invalidate everything.
*/
STATIC int
xfs_attr3_node_inactive(
struct xfs_trans **trans,
struct xfs_inode *dp,
struct xfs_buf *bp,
int level)
{
struct xfs_mount *mp = dp->i_mount;
struct xfs_da_blkinfo *info;
xfs_dablk_t child_fsb;
xfs_daddr_t parent_blkno, child_blkno;
struct xfs_buf *child_bp;
struct xfs_da3_icnode_hdr ichdr;
int error, i;
/*
* Since this code is recursive (gasp!) we must protect ourselves.
*/
if (level > XFS_DA_NODE_MAXDEPTH) {
xfs_buf_mark_corrupt(bp);
xfs_trans_brelse(*trans, bp); /* no locks for later trans */
return -EFSCORRUPTED;
}
xfs_da3_node_hdr_from_disk(dp->i_mount, &ichdr, bp->b_addr);
parent_blkno = xfs_buf_daddr(bp);
if (!ichdr.count) {
xfs_trans_brelse(*trans, bp);
return 0;
}
child_fsb = be32_to_cpu(ichdr.btree[0].before);
xfs_trans_brelse(*trans, bp); /* no locks for later trans */
/*
* If this is the node level just above the leaves, simply loop
* over the leaves removing all of them. If this is higher up
* in the tree, recurse downward.
*/
for (i = 0; i < ichdr.count; i++) {
/*
* Read the subsidiary block to see what we have to work with.
* Don't do this in a transaction. This is a depth-first
* traversal of the tree so we may deal with many blocks
* before we come back to this one.
*/
error = xfs_da3_node_read(*trans, dp, child_fsb, &child_bp,
XFS_ATTR_FORK);
if (error)
return error;
/* save for re-read later */
child_blkno = xfs_buf_daddr(child_bp);
/*
* Invalidate the subtree, however we have to.
*/
info = child_bp->b_addr;
switch (info->magic) {
case cpu_to_be16(XFS_DA_NODE_MAGIC):
case cpu_to_be16(XFS_DA3_NODE_MAGIC):
error = xfs_attr3_node_inactive(trans, dp, child_bp,
level + 1);
break;
case cpu_to_be16(XFS_ATTR_LEAF_MAGIC):
case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC):
error = xfs_attr3_leaf_inactive(trans, dp, child_bp);
break;
default:
xfs_buf_mark_corrupt(child_bp);
xfs_trans_brelse(*trans, child_bp);
error = -EFSCORRUPTED;
break;
}
if (error)
return error;
/*
* Remove the subsidiary block from the cache and from the log.
*/
error = xfs_trans_get_buf(*trans, mp->m_ddev_targp,
child_blkno,
XFS_FSB_TO_BB(mp, mp->m_attr_geo->fsbcount), 0,
&child_bp);
if (error)
return error;
error = bp->b_error;
if (error) {
xfs_trans_brelse(*trans, child_bp);
return error;
}
xfs_trans_binval(*trans, child_bp);
/*
* If we're not done, re-read the parent to get the next
* child block number.
*/
if (i + 1 < ichdr.count) {
struct xfs_da3_icnode_hdr phdr;
error = xfs_da3_node_read_mapped(*trans, dp,
parent_blkno, &bp, XFS_ATTR_FORK);
if (error)
return error;
xfs_da3_node_hdr_from_disk(dp->i_mount, &phdr,
bp->b_addr);
child_fsb = be32_to_cpu(phdr.btree[i + 1].before);
xfs_trans_brelse(*trans, bp);
}
/*
* Atomically commit the whole invalidate stuff.
*/
error = xfs_trans_roll_inode(trans, dp);
if (error)
return error;
}
return 0;
}
/*
* Indiscriminately delete the entire attribute fork
*
* Recurse (gasp!) through the attribute nodes until we find leaves.
* We're doing a depth-first traversal in order to invalidate everything.
*/
static int
xfs_attr3_root_inactive(
struct xfs_trans **trans,
struct xfs_inode *dp)
{
struct xfs_mount *mp = dp->i_mount;
struct xfs_da_blkinfo *info;
struct xfs_buf *bp;
xfs_daddr_t blkno;
int error;
/*
* Read block 0 to see what we have to work with.
* We only get here if we have extents, since we remove
* the extents in reverse order the extent containing
* block 0 must still be there.
*/
error = xfs_da3_node_read(*trans, dp, 0, &bp, XFS_ATTR_FORK);
if (error)
return error;
blkno = xfs_buf_daddr(bp);
/*
* Invalidate the tree, even if the "tree" is only a single leaf block.
* This is a depth-first traversal!
*/
info = bp->b_addr;
switch (info->magic) {
case cpu_to_be16(XFS_DA_NODE_MAGIC):
case cpu_to_be16(XFS_DA3_NODE_MAGIC):
error = xfs_attr3_node_inactive(trans, dp, bp, 1);
break;
case cpu_to_be16(XFS_ATTR_LEAF_MAGIC):
case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC):
error = xfs_attr3_leaf_inactive(trans, dp, bp);
break;
default:
error = -EFSCORRUPTED;
xfs_buf_mark_corrupt(bp);
xfs_trans_brelse(*trans, bp);
break;
}
if (error)
return error;
/*
* Invalidate the incore copy of the root block.
*/
error = xfs_trans_get_buf(*trans, mp->m_ddev_targp, blkno,
XFS_FSB_TO_BB(mp, mp->m_attr_geo->fsbcount), 0, &bp);
if (error)
return error;
error = bp->b_error;
if (error) {
xfs_trans_brelse(*trans, bp);
return error;
}
xfs_trans_binval(*trans, bp); /* remove from cache */
/*
* Commit the invalidate and start the next transaction.
*/
error = xfs_trans_roll_inode(trans, dp);
return error;
}
/*
* xfs_attr_inactive kills all traces of an attribute fork on an inode. It
* removes both the on-disk and in-memory inode fork. Note that this also has to
* handle the condition of inodes without attributes but with an attribute fork
* configured, so we can't use xfs_inode_hasattr() here.
*
* The in-memory attribute fork is removed even on error.
*/
int
xfs_attr_inactive(
struct xfs_inode *dp)
{
struct xfs_trans *trans;
struct xfs_mount *mp;
int lock_mode = XFS_ILOCK_SHARED;
int error = 0;
mp = dp->i_mount;
ASSERT(! XFS_NOT_DQATTACHED(mp, dp));
xfs_ilock(dp, lock_mode);
if (!XFS_IFORK_Q(dp))
goto out_destroy_fork;
xfs_iunlock(dp, lock_mode);
lock_mode = 0;
error = xfs_trans_alloc(mp, &M_RES(mp)->tr_attrinval, 0, 0, 0, &trans);
if (error)
goto out_destroy_fork;
lock_mode = XFS_ILOCK_EXCL;
xfs_ilock(dp, lock_mode);
if (!XFS_IFORK_Q(dp))
goto out_cancel;
/*
* No need to make quota reservations here. We expect to release some
* blocks, not allocate, in the common case.
*/
xfs_trans_ijoin(trans, dp, 0);
xfs: don't truncate attribute extents if no extents exist The xfs_attr3_root_inactive() call from xfs_attr_inactive() assumes that attribute blocks exist to invalidate. It is possible to have an attribute fork without extents, however. Consider the case where the attribute fork is created towards the beginning of xfs_attr_set() but some part of the subsequent attribute set fails. If an inode in such a state hits xfs_attr_inactive(), it eventually calls xfs_dabuf_map() and possibly xfs_bmapi_read(). The former emits a filesystem corruption warning, returns an error that bubbles back up to xfs_attr_inactive(), and leads to destruction of the in-core attribute fork without an on-disk reset. If the inode happens to make it back through xfs_inactive() in this state (e.g., via a concurrent bulkstat that cycles the inode from the reclaim state and releases it), i_afp might not exist when xfs_bmapi_read() is called and causes a NULL dereference panic. A '-p 2' fsstress run to ENOSPC on a relatively small fs (1GB) reproduces these problems. The behavior is a regression caused by: 6dfe5a0 xfs: xfs_attr_inactive leaves inconsistent attr fork state behind ... which removed logic that avoided the attribute extent truncate when no extents exist. Restore this logic to ensure the attribute fork is destroyed and reset correctly if it exists without any allocated extents. cc: stable@vger.kernel.org # 3.12 to 4.0.x Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-06-22 22:47:20 +00:00
/*
* Invalidate and truncate the attribute fork extents. Make sure the
* fork actually has attributes as otherwise the invalidation has no
* blocks to read and returns an error. In this case, just do the fork
* removal below.
*/
if (xfs_inode_hasattr(dp) &&
dp->i_afp->if_format != XFS_DINODE_FMT_LOCAL) {
error = xfs_attr3_root_inactive(&trans, dp);
if (error)
goto out_cancel;
error = xfs_itruncate_extents(&trans, dp, XFS_ATTR_FORK, 0);
if (error)
goto out_cancel;
}
/* Reset the attribute fork - this also destroys the in-core fork */
xfs_attr_fork_remove(dp, trans);
error = xfs_trans_commit(trans);
xfs_iunlock(dp, lock_mode);
return error;
out_cancel:
xfs_trans_cancel(trans);
out_destroy_fork:
/* kill the in-core attr fork before we drop the inode lock */
if (dp->i_afp) {
xfs_idestroy_fork(dp->i_afp);
kmem_cache_free(xfs_ifork_cache, dp->i_afp);
dp->i_afp = NULL;
}
if (lock_mode)
xfs_iunlock(dp, lock_mode);
return error;
}