linux-stable/fs/xfs/xfs_rmap_item.c

748 lines
20 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0+
/*
* Copyright (C) 2016 Oracle. All Rights Reserved.
* Author: Darrick J. Wong <darrick.wong@oracle.com>
*/
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_bit.h"
#include "xfs_shared.h"
#include "xfs_mount.h"
#include "xfs_defer.h"
#include "xfs_trans.h"
#include "xfs_trans_priv.h"
#include "xfs_rmap_item.h"
#include "xfs_log.h"
#include "xfs_rmap.h"
#include "xfs_error.h"
#include "xfs_log_priv.h"
#include "xfs_log_recover.h"
#include "xfs_ag.h"
struct kmem_cache *xfs_rui_cache;
struct kmem_cache *xfs_rud_cache;
static const struct xfs_item_ops xfs_rui_item_ops;
static inline struct xfs_rui_log_item *RUI_ITEM(struct xfs_log_item *lip)
{
return container_of(lip, struct xfs_rui_log_item, rui_item);
}
STATIC void
xfs_rui_item_free(
struct xfs_rui_log_item *ruip)
{
kmem_free(ruip->rui_item.li_lv_shadow);
if (ruip->rui_format.rui_nextents > XFS_RUI_MAX_FAST_EXTENTS)
kmem_free(ruip);
else
kmem_cache_free(xfs_rui_cache, ruip);
}
/*
* Freeing the RUI requires that we remove it from the AIL if it has already
* been placed there. However, the RUI may not yet have been placed in the AIL
* when called by xfs_rui_release() from RUD processing due to the ordering of
* committed vs unpin operations in bulk insert operations. Hence the reference
* count to ensure only the last caller frees the RUI.
*/
STATIC void
xfs_rui_release(
struct xfs_rui_log_item *ruip)
{
ASSERT(atomic_read(&ruip->rui_refcount) > 0);
if (!atomic_dec_and_test(&ruip->rui_refcount))
return;
xfs_trans_ail_delete(&ruip->rui_item, 0);
xfs_rui_item_free(ruip);
}
STATIC void
xfs_rui_item_size(
struct xfs_log_item *lip,
int *nvecs,
int *nbytes)
{
struct xfs_rui_log_item *ruip = RUI_ITEM(lip);
*nvecs += 1;
*nbytes += xfs_rui_log_format_sizeof(ruip->rui_format.rui_nextents);
}
/*
* This is called to fill in the vector of log iovecs for the
* given rui log item. We use only 1 iovec, and we point that
* at the rui_log_format structure embedded in the rui item.
* It is at this point that we assert that all of the extent
* slots in the rui item have been filled.
*/
STATIC void
xfs_rui_item_format(
struct xfs_log_item *lip,
struct xfs_log_vec *lv)
{
struct xfs_rui_log_item *ruip = RUI_ITEM(lip);
struct xfs_log_iovec *vecp = NULL;
ASSERT(atomic_read(&ruip->rui_next_extent) ==
ruip->rui_format.rui_nextents);
ruip->rui_format.rui_type = XFS_LI_RUI;
ruip->rui_format.rui_size = 1;
xlog_copy_iovec(lv, &vecp, XLOG_REG_TYPE_RUI_FORMAT, &ruip->rui_format,
xfs_rui_log_format_sizeof(ruip->rui_format.rui_nextents));
}
/*
* The unpin operation is the last place an RUI is manipulated in the log. It is
* either inserted in the AIL or aborted in the event of a log I/O error. In
* either case, the RUI transaction has been successfully committed to make it
* this far. Therefore, we expect whoever committed the RUI to either construct
* and commit the RUD or drop the RUD's reference in the event of error. Simply
* drop the log's RUI reference now that the log is done with it.
*/
STATIC void
xfs_rui_item_unpin(
struct xfs_log_item *lip,
int remove)
{
struct xfs_rui_log_item *ruip = RUI_ITEM(lip);
xfs_rui_release(ruip);
}
/*
* The RUI has been either committed or aborted if the transaction has been
* cancelled. If the transaction was cancelled, an RUD isn't going to be
* constructed and thus we free the RUI here directly.
*/
STATIC void
xfs: split iop_unlock The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-06-29 02:27:32 +00:00
xfs_rui_item_release(
struct xfs_log_item *lip)
{
xfs: split iop_unlock The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-06-29 02:27:32 +00:00
xfs_rui_release(RUI_ITEM(lip));
}
/*
* Allocate and initialize an rui item with the given number of extents.
*/
STATIC struct xfs_rui_log_item *
xfs_rui_init(
struct xfs_mount *mp,
uint nextents)
{
struct xfs_rui_log_item *ruip;
ASSERT(nextents > 0);
if (nextents > XFS_RUI_MAX_FAST_EXTENTS)
ruip = kmem_zalloc(xfs_rui_log_item_sizeof(nextents), 0);
else
ruip = kmem_cache_zalloc(xfs_rui_cache,
GFP_KERNEL | __GFP_NOFAIL);
xfs_log_item_init(mp, &ruip->rui_item, XFS_LI_RUI, &xfs_rui_item_ops);
ruip->rui_format.rui_nextents = nextents;
ruip->rui_format.rui_id = (uintptr_t)(void *)ruip;
atomic_set(&ruip->rui_next_extent, 0);
atomic_set(&ruip->rui_refcount, 2);
return ruip;
}
static inline struct xfs_rud_log_item *RUD_ITEM(struct xfs_log_item *lip)
{
return container_of(lip, struct xfs_rud_log_item, rud_item);
}
STATIC void
xfs_rud_item_size(
struct xfs_log_item *lip,
int *nvecs,
int *nbytes)
{
*nvecs += 1;
*nbytes += sizeof(struct xfs_rud_log_format);
}
/*
* This is called to fill in the vector of log iovecs for the
* given rud log item. We use only 1 iovec, and we point that
* at the rud_log_format structure embedded in the rud item.
* It is at this point that we assert that all of the extent
* slots in the rud item have been filled.
*/
STATIC void
xfs_rud_item_format(
struct xfs_log_item *lip,
struct xfs_log_vec *lv)
{
struct xfs_rud_log_item *rudp = RUD_ITEM(lip);
struct xfs_log_iovec *vecp = NULL;
rudp->rud_format.rud_type = XFS_LI_RUD;
rudp->rud_format.rud_size = 1;
xlog_copy_iovec(lv, &vecp, XLOG_REG_TYPE_RUD_FORMAT, &rudp->rud_format,
sizeof(struct xfs_rud_log_format));
}
/*
* The RUD is either committed or aborted if the transaction is cancelled. If
* the transaction is cancelled, drop our reference to the RUI and free the
* RUD.
*/
STATIC void
xfs: split iop_unlock The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-06-29 02:27:32 +00:00
xfs_rud_item_release(
struct xfs_log_item *lip)
{
struct xfs_rud_log_item *rudp = RUD_ITEM(lip);
xfs: split iop_unlock The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-06-29 02:27:32 +00:00
xfs_rui_release(rudp->rud_ruip);
kmem_free(rudp->rud_item.li_lv_shadow);
kmem_cache_free(xfs_rud_cache, rudp);
}
static struct xfs_log_item *
xfs_rud_item_intent(
struct xfs_log_item *lip)
{
return &RUD_ITEM(lip)->rud_ruip->rui_item;
}
static const struct xfs_item_ops xfs_rud_item_ops = {
.flags = XFS_ITEM_RELEASE_WHEN_COMMITTED |
XFS_ITEM_INTENT_DONE,
.iop_size = xfs_rud_item_size,
.iop_format = xfs_rud_item_format,
xfs: split iop_unlock The iop_unlock method is called when comitting or cancelling a transaction. In the latter case, the transaction may or may not be aborted. While there is no known problem with the current code in practice, this implementation is limited in that any log item implementation that might want to differentiate between a commit and a cancellation must rely on the aborted state. The aborted bit is only set when the cancelled transaction is dirty, however. This means that there is no way to distinguish between a commit and a clean transaction cancellation. For example, intent log items currently rely on this distinction. The log item is either transferred to the CIL on commit or released on transaction cancel. There is currently no possibility for a clean intent log item in a transaction, but if that state is ever introduced a cancel of such a transaction will immediately result in memory leaks of the associated log item(s). This is an interface deficiency and landmine. To clean this up, replace the iop_unlock method with an iop_release method that is specific to transaction cancel. The existing iop_committing method occurs at the same time as iop_unlock in the commit path and there is no need for two separate callbacks here. Overload the iop_committing method with the current commit time iop_unlock implementations to eliminate the need for the latter and further simplify the interface. Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-06-29 02:27:32 +00:00
.iop_release = xfs_rud_item_release,
.iop_intent = xfs_rud_item_intent,
};
static struct xfs_rud_log_item *
xfs_trans_get_rud(
struct xfs_trans *tp,
struct xfs_rui_log_item *ruip)
{
struct xfs_rud_log_item *rudp;
rudp = kmem_cache_zalloc(xfs_rud_cache, GFP_KERNEL | __GFP_NOFAIL);
xfs_log_item_init(tp->t_mountp, &rudp->rud_item, XFS_LI_RUD,
&xfs_rud_item_ops);
rudp->rud_ruip = ruip;
rudp->rud_format.rud_rui_id = ruip->rui_format.rui_id;
xfs_trans_add_item(tp, &rudp->rud_item);
return rudp;
}
/* Set the map extent flags for this reverse mapping. */
static void
xfs_trans_set_rmap_flags(
struct xfs_map_extent *map,
enum xfs_rmap_intent_type type,
int whichfork,
xfs_exntst_t state)
{
map->me_flags = 0;
if (state == XFS_EXT_UNWRITTEN)
map->me_flags |= XFS_RMAP_EXTENT_UNWRITTEN;
if (whichfork == XFS_ATTR_FORK)
map->me_flags |= XFS_RMAP_EXTENT_ATTR_FORK;
switch (type) {
case XFS_RMAP_MAP:
map->me_flags |= XFS_RMAP_EXTENT_MAP;
break;
case XFS_RMAP_MAP_SHARED:
map->me_flags |= XFS_RMAP_EXTENT_MAP_SHARED;
break;
case XFS_RMAP_UNMAP:
map->me_flags |= XFS_RMAP_EXTENT_UNMAP;
break;
case XFS_RMAP_UNMAP_SHARED:
map->me_flags |= XFS_RMAP_EXTENT_UNMAP_SHARED;
break;
case XFS_RMAP_CONVERT:
map->me_flags |= XFS_RMAP_EXTENT_CONVERT;
break;
case XFS_RMAP_CONVERT_SHARED:
map->me_flags |= XFS_RMAP_EXTENT_CONVERT_SHARED;
break;
case XFS_RMAP_ALLOC:
map->me_flags |= XFS_RMAP_EXTENT_ALLOC;
break;
case XFS_RMAP_FREE:
map->me_flags |= XFS_RMAP_EXTENT_FREE;
break;
default:
ASSERT(0);
}
}
/*
* Finish an rmap update and log it to the RUD. Note that the transaction is
* marked dirty regardless of whether the rmap update succeeds or fails to
* support the RUI/RUD lifecycle rules.
*/
static int
xfs_trans_log_finish_rmap_update(
struct xfs_trans *tp,
struct xfs_rud_log_item *rudp,
struct xfs_rmap_intent *ri,
struct xfs_btree_cur **pcur)
{
int error;
error = xfs_rmap_finish_one(tp, ri, pcur);
/*
* Mark the transaction dirty, even on error. This ensures the
* transaction is aborted, which:
*
* 1.) releases the RUI and frees the RUD
* 2.) shuts down the filesystem
*/
tp->t_flags |= XFS_TRANS_DIRTY | XFS_TRANS_HAS_INTENT_DONE;
set_bit(XFS_LI_DIRTY, &rudp->rud_item.li_flags);
return error;
}
/* Sort rmap intents by AG. */
static int
xfs_rmap_update_diff_items(
void *priv,
const struct list_head *a,
const struct list_head *b)
{
struct xfs_rmap_intent *ra;
struct xfs_rmap_intent *rb;
ra = container_of(a, struct xfs_rmap_intent, ri_list);
rb = container_of(b, struct xfs_rmap_intent, ri_list);
return ra->ri_pag->pag_agno - rb->ri_pag->pag_agno;
}
/* Log rmap updates in the intent item. */
STATIC void
xfs_rmap_update_log_item(
struct xfs_trans *tp,
struct xfs_rui_log_item *ruip,
struct xfs_rmap_intent *ri)
{
uint next_extent;
struct xfs_map_extent *map;
tp->t_flags |= XFS_TRANS_DIRTY;
set_bit(XFS_LI_DIRTY, &ruip->rui_item.li_flags);
/*
* atomic_inc_return gives us the value after the increment;
* we want to use it as an array index so we need to subtract 1 from
* it.
*/
next_extent = atomic_inc_return(&ruip->rui_next_extent) - 1;
ASSERT(next_extent < ruip->rui_format.rui_nextents);
map = &ruip->rui_format.rui_extents[next_extent];
map->me_owner = ri->ri_owner;
map->me_startblock = ri->ri_bmap.br_startblock;
map->me_startoff = ri->ri_bmap.br_startoff;
map->me_len = ri->ri_bmap.br_blockcount;
xfs_trans_set_rmap_flags(map, ri->ri_type, ri->ri_whichfork,
ri->ri_bmap.br_state);
}
static struct xfs_log_item *
xfs_rmap_update_create_intent(
struct xfs_trans *tp,
struct list_head *items,
unsigned int count,
bool sort)
{
struct xfs_mount *mp = tp->t_mountp;
struct xfs_rui_log_item *ruip = xfs_rui_init(mp, count);
struct xfs_rmap_intent *ri;
ASSERT(count > 0);
xfs_trans_add_item(tp, &ruip->rui_item);
if (sort)
list_sort(mp, items, xfs_rmap_update_diff_items);
list_for_each_entry(ri, items, ri_list)
xfs_rmap_update_log_item(tp, ruip, ri);
return &ruip->rui_item;
}
/* Get an RUD so we can process all the deferred rmap updates. */
static struct xfs_log_item *
xfs_rmap_update_create_done(
struct xfs_trans *tp,
struct xfs_log_item *intent,
unsigned int count)
{
return &xfs_trans_get_rud(tp, RUI_ITEM(intent))->rud_item;
}
/* Take a passive ref to the AG containing the space we're rmapping. */
void
xfs_rmap_update_get_group(
struct xfs_mount *mp,
struct xfs_rmap_intent *ri)
{
xfs_agnumber_t agno;
agno = XFS_FSB_TO_AGNO(mp, ri->ri_bmap.br_startblock);
ri->ri_pag = xfs_perag_get(mp, agno);
}
/* Release a passive AG ref after finishing rmapping work. */
static inline void
xfs_rmap_update_put_group(
struct xfs_rmap_intent *ri)
{
xfs_perag_put(ri->ri_pag);
}
/* Process a deferred rmap update. */
STATIC int
xfs_rmap_update_finish_item(
struct xfs_trans *tp,
struct xfs_log_item *done,
struct list_head *item,
struct xfs_btree_cur **state)
{
struct xfs_rmap_intent *ri;
int error;
ri = container_of(item, struct xfs_rmap_intent, ri_list);
error = xfs_trans_log_finish_rmap_update(tp, RUD_ITEM(done), ri,
state);
xfs_rmap_update_put_group(ri);
kmem_cache_free(xfs_rmap_intent_cache, ri);
return error;
}
/* Abort all pending RUIs. */
STATIC void
xfs_rmap_update_abort_intent(
struct xfs_log_item *intent)
{
xfs_rui_release(RUI_ITEM(intent));
}
/* Cancel a deferred rmap update. */
STATIC void
xfs_rmap_update_cancel_item(
struct list_head *item)
{
struct xfs_rmap_intent *ri;
ri = container_of(item, struct xfs_rmap_intent, ri_list);
xfs_rmap_update_put_group(ri);
kmem_cache_free(xfs_rmap_intent_cache, ri);
}
const struct xfs_defer_op_type xfs_rmap_update_defer_type = {
.max_items = XFS_RUI_MAX_FAST_EXTENTS,
.create_intent = xfs_rmap_update_create_intent,
.abort_intent = xfs_rmap_update_abort_intent,
.create_done = xfs_rmap_update_create_done,
.finish_item = xfs_rmap_update_finish_item,
.finish_cleanup = xfs_rmap_finish_one_cleanup,
.cancel_item = xfs_rmap_update_cancel_item,
};
/* Is this recovered RUI ok? */
static inline bool
xfs_rui_validate_map(
struct xfs_mount *mp,
struct xfs_map_extent *map)
{
if (!xfs_has_rmapbt(mp))
return false;
if (map->me_flags & ~XFS_RMAP_EXTENT_FLAGS)
return false;
switch (map->me_flags & XFS_RMAP_EXTENT_TYPE_MASK) {
case XFS_RMAP_EXTENT_MAP:
case XFS_RMAP_EXTENT_MAP_SHARED:
case XFS_RMAP_EXTENT_UNMAP:
case XFS_RMAP_EXTENT_UNMAP_SHARED:
case XFS_RMAP_EXTENT_CONVERT:
case XFS_RMAP_EXTENT_CONVERT_SHARED:
case XFS_RMAP_EXTENT_ALLOC:
case XFS_RMAP_EXTENT_FREE:
break;
default:
return false;
}
if (!XFS_RMAP_NON_INODE_OWNER(map->me_owner) &&
!xfs_verify_ino(mp, map->me_owner))
return false;
if (!xfs_verify_fileext(mp, map->me_startoff, map->me_len))
return false;
return xfs_verify_fsbext(mp, map->me_startblock, map->me_len);
}
/*
* Process an rmap update intent item that was recovered from the log.
* We need to update the rmapbt.
*/
STATIC int
xfs_rui_item_recover(
struct xfs_log_item *lip,
xfs: proper replay of deferred ops queued during log recovery When we replay unfinished intent items that have been recovered from the log, it's possible that the replay will cause the creation of more deferred work items. As outlined in commit 509955823cc9c ("xfs: log recovery should replay deferred ops in order"), later work items have an implicit ordering dependency on earlier work items. Therefore, recovery must replay the items (both recovered and created) in the same order that they would have been during normal operation. For log recovery, we enforce this ordering by using an empty transaction to collect deferred ops that get created in the process of recovering a log intent item to prevent them from being committed before the rest of the recovered intent items. After we finish committing all the recovered log items, we allocate a transaction with an enormous block reservation, splice our huge list of created deferred ops into that transaction, and commit it, thereby finishing all those ops. This is /really/ hokey -- it's the one place in XFS where we allow nested transactions; the splicing of the defer ops list is is inelegant and has to be done twice per recovery function; and the broken way we handle inode pointers and block reservations cause subtle use-after-free and allocator problems that will be fixed by this patch and the two patches after it. Therefore, replace the hokey empty transaction with a structure designed to capture each chain of deferred ops that are created as part of recovering a single unfinished log intent. Finally, refactor the loop that replays those chains to do so using one transaction per chain. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2020-09-26 00:39:37 +00:00
struct list_head *capture_list)
{
struct xfs_rui_log_item *ruip = RUI_ITEM(lip);
struct xfs_rud_log_item *rudp;
struct xfs_trans *tp;
struct xfs_btree_cur *rcur = NULL;
struct xfs_mount *mp = lip->li_log->l_mp;
int i;
int error = 0;
/*
* First check the validity of the extents described by the
* RUI. If any are bad, then assume that all are bad and
* just toss the RUI.
*/
for (i = 0; i < ruip->rui_format.rui_nextents; i++) {
if (!xfs_rui_validate_map(mp,
&ruip->rui_format.rui_extents[i])) {
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
&ruip->rui_format,
sizeof(ruip->rui_format));
return -EFSCORRUPTED;
}
}
error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate,
mp->m_rmap_maxlevels, 0, XFS_TRANS_RESERVE, &tp);
if (error)
return error;
rudp = xfs_trans_get_rud(tp, ruip);
for (i = 0; i < ruip->rui_format.rui_nextents; i++) {
struct xfs_rmap_intent fake = { };
struct xfs_map_extent *map;
map = &ruip->rui_format.rui_extents[i];
switch (map->me_flags & XFS_RMAP_EXTENT_TYPE_MASK) {
case XFS_RMAP_EXTENT_MAP:
fake.ri_type = XFS_RMAP_MAP;
break;
case XFS_RMAP_EXTENT_MAP_SHARED:
fake.ri_type = XFS_RMAP_MAP_SHARED;
break;
case XFS_RMAP_EXTENT_UNMAP:
fake.ri_type = XFS_RMAP_UNMAP;
break;
case XFS_RMAP_EXTENT_UNMAP_SHARED:
fake.ri_type = XFS_RMAP_UNMAP_SHARED;
break;
case XFS_RMAP_EXTENT_CONVERT:
fake.ri_type = XFS_RMAP_CONVERT;
break;
case XFS_RMAP_EXTENT_CONVERT_SHARED:
fake.ri_type = XFS_RMAP_CONVERT_SHARED;
break;
case XFS_RMAP_EXTENT_ALLOC:
fake.ri_type = XFS_RMAP_ALLOC;
break;
case XFS_RMAP_EXTENT_FREE:
fake.ri_type = XFS_RMAP_FREE;
break;
default:
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
&ruip->rui_format,
sizeof(ruip->rui_format));
error = -EFSCORRUPTED;
goto abort_error;
}
fake.ri_owner = map->me_owner;
fake.ri_whichfork = (map->me_flags & XFS_RMAP_EXTENT_ATTR_FORK) ?
XFS_ATTR_FORK : XFS_DATA_FORK;
fake.ri_bmap.br_startblock = map->me_startblock;
fake.ri_bmap.br_startoff = map->me_startoff;
fake.ri_bmap.br_blockcount = map->me_len;
fake.ri_bmap.br_state = (map->me_flags & XFS_RMAP_EXTENT_UNWRITTEN) ?
XFS_EXT_UNWRITTEN : XFS_EXT_NORM;
xfs_rmap_update_get_group(mp, &fake);
error = xfs_trans_log_finish_rmap_update(tp, rudp, &fake,
&rcur);
if (error == -EFSCORRUPTED)
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
map, sizeof(*map));
xfs_rmap_update_put_group(&fake);
if (error)
goto abort_error;
}
xfs_rmap_finish_one_cleanup(tp, rcur, error);
return xfs_defer_ops_capture_and_commit(tp, capture_list);
abort_error:
xfs_rmap_finish_one_cleanup(tp, rcur, error);
xfs_trans_cancel(tp);
return error;
}
STATIC bool
xfs_rui_item_match(
struct xfs_log_item *lip,
uint64_t intent_id)
{
return RUI_ITEM(lip)->rui_format.rui_id == intent_id;
}
xfs: periodically relog deferred intent items There's a subtle design flaw in the deferred log item code that can lead to pinning the log tail. Taking up the defer ops chain examples from the previous commit, we can get trapped in sequences like this: Caller hands us a transaction t0 with D0-D3 attached. The defer ops chain will look like the following if the transaction rolls succeed: t1: D0(t0), D1(t0), D2(t0), D3(t0) t2: d4(t1), d5(t1), D1(t0), D2(t0), D3(t0) t3: d5(t1), D1(t0), D2(t0), D3(t0) ... t9: d9(t7), D3(t0) t10: D3(t0) t11: d10(t10), d11(t10) t12: d11(t10) In transaction 9, we finish d9 and try to roll to t10 while holding onto an intent item for D3 that we logged in t0. The previous commit changed the order in which we place new defer ops in the defer ops processing chain to reduce the maximum chain length. Now make xfs_defer_finish_noroll capable of relogging the entire chain periodically so that we can always move the log tail forward. Most chains will never get relogged, except for operations that generate very long chains (large extents containing many blocks with different sharing levels) or are on filesystems with small logs and a lot of ongoing metadata updates. Callers are now required to ensure that the transaction reservation is large enough to handle logging done items and new intent items for the maximum possible chain length. Most callers are careful to keep the chain lengths low, so the overhead should be minimal. The decision to relog an intent item is made based on whether the intent was logged in a previous checkpoint, since there's no point in relogging an intent into the same checkpoint. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-09-27 23:18:13 +00:00
/* Relog an intent item to push the log tail forward. */
static struct xfs_log_item *
xfs_rui_item_relog(
struct xfs_log_item *intent,
struct xfs_trans *tp)
{
struct xfs_rud_log_item *rudp;
struct xfs_rui_log_item *ruip;
struct xfs_map_extent *map;
xfs: periodically relog deferred intent items There's a subtle design flaw in the deferred log item code that can lead to pinning the log tail. Taking up the defer ops chain examples from the previous commit, we can get trapped in sequences like this: Caller hands us a transaction t0 with D0-D3 attached. The defer ops chain will look like the following if the transaction rolls succeed: t1: D0(t0), D1(t0), D2(t0), D3(t0) t2: d4(t1), d5(t1), D1(t0), D2(t0), D3(t0) t3: d5(t1), D1(t0), D2(t0), D3(t0) ... t9: d9(t7), D3(t0) t10: D3(t0) t11: d10(t10), d11(t10) t12: d11(t10) In transaction 9, we finish d9 and try to roll to t10 while holding onto an intent item for D3 that we logged in t0. The previous commit changed the order in which we place new defer ops in the defer ops processing chain to reduce the maximum chain length. Now make xfs_defer_finish_noroll capable of relogging the entire chain periodically so that we can always move the log tail forward. Most chains will never get relogged, except for operations that generate very long chains (large extents containing many blocks with different sharing levels) or are on filesystems with small logs and a lot of ongoing metadata updates. Callers are now required to ensure that the transaction reservation is large enough to handle logging done items and new intent items for the maximum possible chain length. Most callers are careful to keep the chain lengths low, so the overhead should be minimal. The decision to relog an intent item is made based on whether the intent was logged in a previous checkpoint, since there's no point in relogging an intent into the same checkpoint. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-09-27 23:18:13 +00:00
unsigned int count;
count = RUI_ITEM(intent)->rui_format.rui_nextents;
map = RUI_ITEM(intent)->rui_format.rui_extents;
xfs: periodically relog deferred intent items There's a subtle design flaw in the deferred log item code that can lead to pinning the log tail. Taking up the defer ops chain examples from the previous commit, we can get trapped in sequences like this: Caller hands us a transaction t0 with D0-D3 attached. The defer ops chain will look like the following if the transaction rolls succeed: t1: D0(t0), D1(t0), D2(t0), D3(t0) t2: d4(t1), d5(t1), D1(t0), D2(t0), D3(t0) t3: d5(t1), D1(t0), D2(t0), D3(t0) ... t9: d9(t7), D3(t0) t10: D3(t0) t11: d10(t10), d11(t10) t12: d11(t10) In transaction 9, we finish d9 and try to roll to t10 while holding onto an intent item for D3 that we logged in t0. The previous commit changed the order in which we place new defer ops in the defer ops processing chain to reduce the maximum chain length. Now make xfs_defer_finish_noroll capable of relogging the entire chain periodically so that we can always move the log tail forward. Most chains will never get relogged, except for operations that generate very long chains (large extents containing many blocks with different sharing levels) or are on filesystems with small logs and a lot of ongoing metadata updates. Callers are now required to ensure that the transaction reservation is large enough to handle logging done items and new intent items for the maximum possible chain length. Most callers are careful to keep the chain lengths low, so the overhead should be minimal. The decision to relog an intent item is made based on whether the intent was logged in a previous checkpoint, since there's no point in relogging an intent into the same checkpoint. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-09-27 23:18:13 +00:00
tp->t_flags |= XFS_TRANS_DIRTY;
rudp = xfs_trans_get_rud(tp, RUI_ITEM(intent));
set_bit(XFS_LI_DIRTY, &rudp->rud_item.li_flags);
ruip = xfs_rui_init(tp->t_mountp, count);
memcpy(ruip->rui_format.rui_extents, map, count * sizeof(*map));
xfs: periodically relog deferred intent items There's a subtle design flaw in the deferred log item code that can lead to pinning the log tail. Taking up the defer ops chain examples from the previous commit, we can get trapped in sequences like this: Caller hands us a transaction t0 with D0-D3 attached. The defer ops chain will look like the following if the transaction rolls succeed: t1: D0(t0), D1(t0), D2(t0), D3(t0) t2: d4(t1), d5(t1), D1(t0), D2(t0), D3(t0) t3: d5(t1), D1(t0), D2(t0), D3(t0) ... t9: d9(t7), D3(t0) t10: D3(t0) t11: d10(t10), d11(t10) t12: d11(t10) In transaction 9, we finish d9 and try to roll to t10 while holding onto an intent item for D3 that we logged in t0. The previous commit changed the order in which we place new defer ops in the defer ops processing chain to reduce the maximum chain length. Now make xfs_defer_finish_noroll capable of relogging the entire chain periodically so that we can always move the log tail forward. Most chains will never get relogged, except for operations that generate very long chains (large extents containing many blocks with different sharing levels) or are on filesystems with small logs and a lot of ongoing metadata updates. Callers are now required to ensure that the transaction reservation is large enough to handle logging done items and new intent items for the maximum possible chain length. Most callers are careful to keep the chain lengths low, so the overhead should be minimal. The decision to relog an intent item is made based on whether the intent was logged in a previous checkpoint, since there's no point in relogging an intent into the same checkpoint. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-09-27 23:18:13 +00:00
atomic_set(&ruip->rui_next_extent, count);
xfs_trans_add_item(tp, &ruip->rui_item);
set_bit(XFS_LI_DIRTY, &ruip->rui_item.li_flags);
return &ruip->rui_item;
}
static const struct xfs_item_ops xfs_rui_item_ops = {
.flags = XFS_ITEM_INTENT,
.iop_size = xfs_rui_item_size,
.iop_format = xfs_rui_item_format,
.iop_unpin = xfs_rui_item_unpin,
.iop_release = xfs_rui_item_release,
.iop_recover = xfs_rui_item_recover,
.iop_match = xfs_rui_item_match,
xfs: periodically relog deferred intent items There's a subtle design flaw in the deferred log item code that can lead to pinning the log tail. Taking up the defer ops chain examples from the previous commit, we can get trapped in sequences like this: Caller hands us a transaction t0 with D0-D3 attached. The defer ops chain will look like the following if the transaction rolls succeed: t1: D0(t0), D1(t0), D2(t0), D3(t0) t2: d4(t1), d5(t1), D1(t0), D2(t0), D3(t0) t3: d5(t1), D1(t0), D2(t0), D3(t0) ... t9: d9(t7), D3(t0) t10: D3(t0) t11: d10(t10), d11(t10) t12: d11(t10) In transaction 9, we finish d9 and try to roll to t10 while holding onto an intent item for D3 that we logged in t0. The previous commit changed the order in which we place new defer ops in the defer ops processing chain to reduce the maximum chain length. Now make xfs_defer_finish_noroll capable of relogging the entire chain periodically so that we can always move the log tail forward. Most chains will never get relogged, except for operations that generate very long chains (large extents containing many blocks with different sharing levels) or are on filesystems with small logs and a lot of ongoing metadata updates. Callers are now required to ensure that the transaction reservation is large enough to handle logging done items and new intent items for the maximum possible chain length. Most callers are careful to keep the chain lengths low, so the overhead should be minimal. The decision to relog an intent item is made based on whether the intent was logged in a previous checkpoint, since there's no point in relogging an intent into the same checkpoint. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2020-09-27 23:18:13 +00:00
.iop_relog = xfs_rui_item_relog,
};
static inline void
xfs_rui_copy_format(
struct xfs_rui_log_format *dst,
const struct xfs_rui_log_format *src)
{
unsigned int i;
memcpy(dst, src, offsetof(struct xfs_rui_log_format, rui_extents));
for (i = 0; i < src->rui_nextents; i++)
memcpy(&dst->rui_extents[i], &src->rui_extents[i],
sizeof(struct xfs_map_extent));
}
/*
* This routine is called to create an in-core extent rmap update
* item from the rui format structure which was logged on disk.
* It allocates an in-core rui, copies the extents from the format
* structure into it, and adds the rui to the AIL with the given
* LSN.
*/
STATIC int
xlog_recover_rui_commit_pass2(
struct xlog *log,
struct list_head *buffer_list,
struct xlog_recover_item *item,
xfs_lsn_t lsn)
{
struct xfs_mount *mp = log->l_mp;
struct xfs_rui_log_item *ruip;
struct xfs_rui_log_format *rui_formatp;
size_t len;
rui_formatp = item->ri_buf[0].i_addr;
if (item->ri_buf[0].i_len < xfs_rui_log_format_sizeof(0)) {
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
item->ri_buf[0].i_addr, item->ri_buf[0].i_len);
return -EFSCORRUPTED;
}
len = xfs_rui_log_format_sizeof(rui_formatp->rui_nextents);
if (item->ri_buf[0].i_len != len) {
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
item->ri_buf[0].i_addr, item->ri_buf[0].i_len);
return -EFSCORRUPTED;
}
ruip = xfs_rui_init(mp, rui_formatp->rui_nextents);
xfs_rui_copy_format(&ruip->rui_format, rui_formatp);
atomic_set(&ruip->rui_next_extent, rui_formatp->rui_nextents);
/*
* Insert the intent into the AIL directly and drop one reference so
* that finishing or canceling the work will drop the other.
*/
xfs_trans_ail_insert(log->l_ailp, &ruip->rui_item, lsn);
xfs_rui_release(ruip);
return 0;
}
const struct xlog_recover_item_ops xlog_rui_item_ops = {
.item_type = XFS_LI_RUI,
.commit_pass2 = xlog_recover_rui_commit_pass2,
};
/*
* This routine is called when an RUD format structure is found in a committed
* transaction in the log. Its purpose is to cancel the corresponding RUI if it
* was still in the log. To do this it searches the AIL for the RUI with an id
* equal to that in the RUD format structure. If we find it we drop the RUD
* reference, which removes the RUI from the AIL and frees it.
*/
STATIC int
xlog_recover_rud_commit_pass2(
struct xlog *log,
struct list_head *buffer_list,
struct xlog_recover_item *item,
xfs_lsn_t lsn)
{
struct xfs_rud_log_format *rud_formatp;
rud_formatp = item->ri_buf[0].i_addr;
if (item->ri_buf[0].i_len != sizeof(struct xfs_rud_log_format)) {
XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, log->l_mp,
rud_formatp, item->ri_buf[0].i_len);
return -EFSCORRUPTED;
}
xlog_recover_release_intent(log, XFS_LI_RUI, rud_formatp->rud_rui_id);
return 0;
}
const struct xlog_recover_item_ops xlog_rud_item_ops = {
.item_type = XFS_LI_RUD,
.commit_pass2 = xlog_recover_rud_commit_pass2,
};