2018-06-06 09:42:14 +07:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
2005-11-02 10:58:39 +07:00
|
|
|
* Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
|
|
|
|
* All Rights Reserved.
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
|
|
|
#ifndef __XFS_INODE_H__
|
|
|
|
#define __XFS_INODE_H__
|
|
|
|
|
2013-08-12 17:49:35 +07:00
|
|
|
#include "xfs_inode_buf.h"
|
2013-08-12 17:49:33 +07:00
|
|
|
#include "xfs_inode_fork.h"
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-11-28 10:23:41 +07:00
|
|
|
/*
|
2013-08-12 17:49:35 +07:00
|
|
|
* Kernel only inode definitions
|
2008-11-28 10:23:41 +07:00
|
|
|
*/
|
2013-08-12 17:49:35 +07:00
|
|
|
struct xfs_dinode;
|
|
|
|
struct xfs_inode;
|
2008-10-30 13:05:38 +07:00
|
|
|
struct xfs_buf;
|
|
|
|
struct xfs_bmbt_irec;
|
|
|
|
struct xfs_inode_log_item;
|
|
|
|
struct xfs_mount;
|
|
|
|
struct xfs_trans;
|
|
|
|
struct xfs_dquot;
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
typedef struct xfs_inode {
|
|
|
|
/* Inode linking and identification information. */
|
|
|
|
struct xfs_mount *i_mount; /* fs mount struct ptr */
|
|
|
|
struct xfs_dquot *i_udquot; /* user dquot */
|
|
|
|
struct xfs_dquot *i_gdquot; /* group dquot */
|
2013-07-11 12:00:40 +07:00
|
|
|
struct xfs_dquot *i_pdquot; /* project dquot */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* Inode location stuff */
|
|
|
|
xfs_ino_t i_ino; /* inode number (agno/agino)*/
|
2008-11-28 10:23:41 +07:00
|
|
|
struct xfs_imap i_imap; /* location for xfs_imap() */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* Extent information. */
|
2018-07-18 06:51:50 +07:00
|
|
|
struct xfs_ifork *i_afp; /* attribute fork pointer */
|
|
|
|
struct xfs_ifork *i_cowfp; /* copy on write extents */
|
|
|
|
struct xfs_ifork i_df; /* data fork */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
|
|
|
/* Transaction and locking information. */
|
|
|
|
struct xfs_inode_log_item *i_itemp; /* logging information */
|
|
|
|
mrlock_t i_lock; /* inode lock */
|
2015-02-23 17:43:37 +07:00
|
|
|
mrlock_t i_mmaplock; /* inode mmap IO lock */
|
2005-04-17 05:20:36 +07:00
|
|
|
atomic_t i_pincount; /* inode pin count */
|
2019-04-12 21:40:25 +07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Bitsets of inode metadata that have been checked and/or are sick.
|
|
|
|
* Callers must hold i_flags_lock before accessing this field.
|
|
|
|
*/
|
|
|
|
uint16_t i_checked;
|
|
|
|
uint16_t i_sick;
|
|
|
|
|
2006-09-28 08:06:03 +07:00
|
|
|
spinlock_t i_flags_lock; /* inode i_flags lock */
|
2005-04-17 05:20:36 +07:00
|
|
|
/* Miscellaneous state. */
|
2011-12-19 03:00:08 +07:00
|
|
|
unsigned long i_flags; /* see defined flags below */
|
2019-04-18 06:30:24 +07:00
|
|
|
uint64_t i_delayed_blks; /* count of delay alloc blks */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2016-02-09 12:54:58 +07:00
|
|
|
struct xfs_icdinode i_d; /* most of ondisk inode */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2008-10-30 13:36:14 +07:00
|
|
|
/* VFS inode */
|
|
|
|
struct inode i_vnode; /* embedded VFS inode */
|
xfs: implement per-inode writeback completion queues
When scheduling writeback of dirty file data in the page cache, XFS uses
IO completion workqueue items to ensure that filesystem metadata only
updates after the write completes successfully. This is essential for
converting unwritten extents to real extents at the right time and
performing COW remappings.
Unfortunately, XFS queues each IO completion work item to an unbounded
workqueue, which means that the kernel can spawn dozens of threads to
try to handle the items quickly. These threads need to take the ILOCK
to update file metadata, which results in heavy ILOCK contention if a
large number of the work items target a single file, which is
inefficient.
Worse yet, the writeback completion threads get stuck waiting for the
ILOCK while holding transaction reservations, which can use up all
available log reservation space. When that happens, metadata updates to
other parts of the filesystem grind to a halt, even if the filesystem
could otherwise have handled it.
Even worse, if one of the things grinding to a halt happens to be a
thread in the middle of a defer-ops finish holding the same ILOCK and
trying to obtain more log reservation having exhausted the permanent
reservation, we now have an ABBA deadlock - writeback completion has a
transaction reserved and wants the ILOCK, and someone else has the ILOCK
and wants a transaction reservation.
Therefore, we create a per-inode writeback io completion queue + work
item. When writeback finishes, it can add the ioend to the per-inode
queue and let the single worker item process that queue. This
dramatically cuts down on the number of kworkers and ILOCK contention in
the system, and seems to have eliminated an occasional deadlock I was
seeing while running generic/476.
Testing with a program that simulates a heavy random-write workload to a
single file demonstrates that the number of kworkers drops from
approximately 120 threads per file to 1, without dramatically changing
write bandwidth or pagecache access latency.
Note that we leave the xfs-conv workqueue's max_active alone because we
still want to be able to run ioend processing for as many inodes as the
system can handle.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
2019-04-16 03:13:20 +07:00
|
|
|
|
|
|
|
/* pending io completions */
|
|
|
|
spinlock_t i_ioend_lock;
|
|
|
|
struct work_struct i_ioend_work;
|
|
|
|
struct list_head i_ioend_list;
|
2005-04-17 05:20:36 +07:00
|
|
|
} xfs_inode_t;
|
|
|
|
|
2008-08-13 12:45:15 +07:00
|
|
|
/* Convert from vfs inode to xfs inode */
|
|
|
|
static inline struct xfs_inode *XFS_I(struct inode *inode)
|
|
|
|
{
|
2008-10-30 13:36:14 +07:00
|
|
|
return container_of(inode, struct xfs_inode, i_vnode);
|
2008-08-13 12:45:15 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
/* convert from xfs inode to vfs inode */
|
|
|
|
static inline struct inode *VFS_I(struct xfs_inode *ip)
|
|
|
|
{
|
2008-10-30 13:36:14 +07:00
|
|
|
return &ip->i_vnode;
|
2008-08-13 12:45:15 +07:00
|
|
|
}
|
|
|
|
|
2011-12-19 03:00:11 +07:00
|
|
|
/*
|
|
|
|
* For regular files we only update the on-disk filesize when actually
|
|
|
|
* writing data back to disk. Until then only the copy in the VFS inode
|
|
|
|
* is uptodate.
|
|
|
|
*/
|
|
|
|
static inline xfs_fsize_t XFS_ISIZE(struct xfs_inode *ip)
|
|
|
|
{
|
2016-02-09 12:54:58 +07:00
|
|
|
if (S_ISREG(VFS_I(ip)->i_mode))
|
2011-12-19 03:00:11 +07:00
|
|
|
return i_size_read(VFS_I(ip));
|
|
|
|
return ip->i_d.di_size;
|
|
|
|
}
|
|
|
|
|
2012-02-29 16:53:49 +07:00
|
|
|
/*
|
|
|
|
* If this I/O goes past the on-disk inode size update it unless it would
|
|
|
|
* be past the current in-core inode size.
|
|
|
|
*/
|
|
|
|
static inline xfs_fsize_t
|
|
|
|
xfs_new_eof(struct xfs_inode *ip, xfs_fsize_t new_size)
|
|
|
|
{
|
|
|
|
xfs_fsize_t i_size = i_size_read(VFS_I(ip));
|
|
|
|
|
2014-10-02 06:21:53 +07:00
|
|
|
if (new_size > i_size || new_size < 0)
|
2012-02-29 16:53:49 +07:00
|
|
|
new_size = i_size;
|
|
|
|
return new_size > ip->i_d.di_size ? new_size : 0;
|
|
|
|
}
|
|
|
|
|
2006-11-11 14:04:54 +07:00
|
|
|
/*
|
|
|
|
* i_flags helper functions
|
|
|
|
*/
|
|
|
|
static inline void
|
|
|
|
__xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
|
|
|
|
{
|
|
|
|
ip->i_flags |= flags;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
|
|
|
|
{
|
|
|
|
spin_lock(&ip->i_flags_lock);
|
|
|
|
__xfs_iflags_set(ip, flags);
|
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
xfs_iflags_clear(xfs_inode_t *ip, unsigned short flags)
|
|
|
|
{
|
|
|
|
spin_lock(&ip->i_flags_lock);
|
|
|
|
ip->i_flags &= ~flags;
|
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int
|
|
|
|
__xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
|
|
|
|
{
|
|
|
|
return (ip->i_flags & flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int
|
|
|
|
xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
spin_lock(&ip->i_flags_lock);
|
|
|
|
ret = __xfs_iflags_test(ip, flags);
|
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
return ret;
|
|
|
|
}
|
2007-08-29 08:44:50 +07:00
|
|
|
|
|
|
|
static inline int
|
|
|
|
xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned short flags)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
spin_lock(&ip->i_flags_lock);
|
|
|
|
ret = ip->i_flags & flags;
|
|
|
|
if (ret)
|
|
|
|
ip->i_flags &= ~flags;
|
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
return ret;
|
|
|
|
}
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-12-19 03:00:09 +07:00
|
|
|
static inline int
|
|
|
|
xfs_iflags_test_and_set(xfs_inode_t *ip, unsigned short flags)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
spin_lock(&ip->i_flags_lock);
|
|
|
|
ret = ip->i_flags & flags;
|
|
|
|
if (!ret)
|
|
|
|
ip->i_flags |= flags;
|
|
|
|
spin_unlock(&ip->i_flags_lock);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-12-18 07:22:39 +07:00
|
|
|
static inline prid_t
|
|
|
|
xfs_get_initial_prid(struct xfs_inode *dp)
|
|
|
|
{
|
|
|
|
if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
|
2019-11-12 23:22:54 +07:00
|
|
|
return dp->i_d.di_projid;
|
2013-12-18 07:22:39 +07:00
|
|
|
|
|
|
|
return XFS_PROJID_DEFAULT;
|
|
|
|
}
|
|
|
|
|
2016-10-03 23:11:16 +07:00
|
|
|
static inline bool xfs_is_reflink_inode(struct xfs_inode *ip)
|
|
|
|
{
|
|
|
|
return ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK;
|
|
|
|
}
|
|
|
|
|
2018-07-18 06:51:51 +07:00
|
|
|
/*
|
|
|
|
* Check if an inode has any data in the COW fork. This might be often false
|
|
|
|
* even for inodes with the reflink flag when there is no pending COW operation.
|
|
|
|
*/
|
|
|
|
static inline bool xfs_inode_has_cow_data(struct xfs_inode *ip)
|
|
|
|
{
|
|
|
|
return ip->i_cowfp && ip->i_cowfp->if_bytes;
|
|
|
|
}
|
|
|
|
|
2020-08-17 23:59:07 +07:00
|
|
|
static inline bool xfs_inode_has_bigtime(struct xfs_inode *ip)
|
|
|
|
{
|
|
|
|
return ip->i_d.di_flags2 & XFS_DIFLAG2_BIGTIME;
|
|
|
|
}
|
|
|
|
|
2019-10-25 12:25:38 +07:00
|
|
|
/*
|
|
|
|
* Return the buftarg used for data allocations on a given inode.
|
|
|
|
*/
|
|
|
|
#define xfs_inode_buftarg(ip) \
|
|
|
|
(XFS_IS_REALTIME_INODE(ip) ? \
|
|
|
|
(ip)->i_mount->m_rtdev_targp : (ip)->i_mount->m_ddev_targp)
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* In-core inode flags.
|
|
|
|
*/
|
2011-12-19 03:00:09 +07:00
|
|
|
#define XFS_IRECLAIM (1 << 0) /* started reclaiming this inode */
|
|
|
|
#define XFS_ISTALE (1 << 1) /* inode has been staled */
|
|
|
|
#define XFS_IRECLAIMABLE (1 << 2) /* inode can be reclaimed */
|
2017-04-26 22:30:39 +07:00
|
|
|
#define __XFS_INEW_BIT 3 /* inode has just been allocated */
|
|
|
|
#define XFS_INEW (1 << __XFS_INEW_BIT)
|
2011-12-19 03:00:09 +07:00
|
|
|
#define XFS_ITRUNCATED (1 << 5) /* truncated down so flush-on-close */
|
|
|
|
#define XFS_IDIRTY_RELEASE (1 << 6) /* dirty release already seen */
|
2020-08-18 06:41:01 +07:00
|
|
|
#define XFS_IFLUSHING (1 << 7) /* inode is being flushed */
|
2011-12-19 03:00:10 +07:00
|
|
|
#define __XFS_IPINNED_BIT 8 /* wakeup key for zero pin count */
|
|
|
|
#define XFS_IPINNED (1 << __XFS_IPINNED_BIT)
|
2020-04-30 21:41:37 +07:00
|
|
|
#define XFS_IEOFBLOCKS (1 << 9) /* has the preallocblocks tag set */
|
2016-10-03 23:11:29 +07:00
|
|
|
/*
|
|
|
|
* If this unlinked inode is in the middle of recovery, don't let drop_inode
|
|
|
|
* truncate and free the inode. This can happen if we iget the inode during
|
|
|
|
* log recovery to replay a bmap operation on the inode.
|
|
|
|
*/
|
|
|
|
#define XFS_IRECOVERY (1 << 11)
|
2017-12-15 06:42:22 +07:00
|
|
|
#define XFS_ICOWBLOCKS (1 << 12)/* has the cowblocks tag set */
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2011-06-23 08:34:59 +07:00
|
|
|
/*
|
|
|
|
* Per-lifetime flags need to be reset when re-using a reclaimable inode during
|
2012-03-22 12:15:10 +07:00
|
|
|
* inode lookup. This prevents unintended behaviour on the new inode from
|
2011-06-23 08:34:59 +07:00
|
|
|
* ocurring.
|
|
|
|
*/
|
|
|
|
#define XFS_IRECLAIM_RESET_FLAGS \
|
|
|
|
(XFS_IRECLAIMABLE | XFS_IRECLAIM | \
|
2014-04-23 04:11:51 +07:00
|
|
|
XFS_IDIRTY_RELEASE | XFS_ITRUNCATED)
|
2011-06-23 08:34:59 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* Flags for inode locking.
|
2007-05-08 10:50:19 +07:00
|
|
|
* Bit ranges: 1<<1 - 1<<16-1 -- iolock/ilock modes (bitfield)
|
|
|
|
* 1<<16 - 1<<32-1 -- lockdep annotation (integers)
|
2005-04-17 05:20:36 +07:00
|
|
|
*/
|
2007-05-08 10:50:19 +07:00
|
|
|
#define XFS_IOLOCK_EXCL (1<<0)
|
|
|
|
#define XFS_IOLOCK_SHARED (1<<1)
|
|
|
|
#define XFS_ILOCK_EXCL (1<<2)
|
|
|
|
#define XFS_ILOCK_SHARED (1<<3)
|
2015-02-23 17:43:37 +07:00
|
|
|
#define XFS_MMAPLOCK_EXCL (1<<4)
|
|
|
|
#define XFS_MMAPLOCK_SHARED (1<<5)
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2007-05-08 10:50:19 +07:00
|
|
|
#define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
|
2015-02-23 17:43:37 +07:00
|
|
|
| XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \
|
|
|
|
| XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED)
|
2007-05-08 10:50:19 +07:00
|
|
|
|
2009-12-15 06:14:59 +07:00
|
|
|
#define XFS_LOCK_FLAGS \
|
|
|
|
{ XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \
|
|
|
|
{ XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \
|
|
|
|
{ XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \
|
2015-02-23 17:43:37 +07:00
|
|
|
{ XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \
|
|
|
|
{ XFS_MMAPLOCK_EXCL, "MMAPLOCK_EXCL" }, \
|
|
|
|
{ XFS_MMAPLOCK_SHARED, "MMAPLOCK_SHARED" }
|
2009-12-15 06:14:59 +07:00
|
|
|
|
|
|
|
|
2007-05-08 10:50:19 +07:00
|
|
|
/*
|
|
|
|
* Flags for lockdep annotations.
|
|
|
|
*
|
2011-01-25 16:06:21 +07:00
|
|
|
* XFS_LOCK_PARENT - for directory operations that require locking a
|
xfs: clean up inode lockdep annotations
Lockdep annotations are a maintenance nightmare. Locking has to be
modified to suit the limitations of the annotations, and we're
always having to fix the annotations because they are unable to
express the complexity of locking heirarchies correctly.
So, next up, we've got more issues with lockdep annotations for
inode locking w.r.t. XFS_LOCK_PARENT:
- lockdep classes are exclusive and can't be ORed together
to form new classes.
- IOLOCK needs multiple PARENT subclasses to express the
changes needed for the readdir locking rework needed to
stop the endless flow of lockdep false positives involving
readdir calling filldir under the ILOCK.
- there are only 8 unique lockdep subclasses available,
so we can't create a generic solution.
IOWs we need to treat the 3-bit space available to each lock type
differently:
- IOLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 IOLOCK subclasses
- at least 2 IOLOCK_PARENT subclasses
- MMAPLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 MMAPLOCK subclasses
- ILOCK uses xfs_lock_inodes with up to 5 inodes, so needs:
- at least 5 ILOCK subclasses
- one ILOCK_PARENT subclass
- one RTBITMAP subclass
- one RTSUM subclass
For the IOLOCK, split the space into two sets of subclasses.
For the MMAPLOCK, just use half the space for the one subclass to
match the non-parent lock classes of the IOLOCK.
For the ILOCK, use 0-4 as the ILOCK subclasses, 5-7 for the
remaining individual subclasses.
Because they are now all different, modify xfs_lock_inumorder() to
handle the nested subclasses, and to assert fail if passed an
invalid subclass. Further, annotate xfs_lock_inodes() to assert fail
if an invalid combination of lock primitives and inode counts are
passed that would result in a lockdep subclass annotation overflow.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 07:32:49 +07:00
|
|
|
* parent directory inode and a child entry inode. IOLOCK requires nesting,
|
|
|
|
* MMAPLOCK does not support this class, ILOCK requires a single subclass
|
|
|
|
* to differentiate parent from child.
|
2011-01-25 16:06:21 +07:00
|
|
|
*
|
|
|
|
* XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
|
|
|
|
* inodes do not participate in the normal lock order, and thus have their
|
|
|
|
* own subclasses.
|
2007-05-08 10:50:19 +07:00
|
|
|
*
|
2007-06-29 14:26:09 +07:00
|
|
|
* XFS_LOCK_INUMORDER - for locking several inodes at the some time
|
2007-05-08 10:50:19 +07:00
|
|
|
* with xfs_lock_inodes(). This flag is used as the starting subclass
|
|
|
|
* and each subsequent lock acquired will increment the subclass by one.
|
xfs: clean up inode lockdep annotations
Lockdep annotations are a maintenance nightmare. Locking has to be
modified to suit the limitations of the annotations, and we're
always having to fix the annotations because they are unable to
express the complexity of locking heirarchies correctly.
So, next up, we've got more issues with lockdep annotations for
inode locking w.r.t. XFS_LOCK_PARENT:
- lockdep classes are exclusive and can't be ORed together
to form new classes.
- IOLOCK needs multiple PARENT subclasses to express the
changes needed for the readdir locking rework needed to
stop the endless flow of lockdep false positives involving
readdir calling filldir under the ILOCK.
- there are only 8 unique lockdep subclasses available,
so we can't create a generic solution.
IOWs we need to treat the 3-bit space available to each lock type
differently:
- IOLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 IOLOCK subclasses
- at least 2 IOLOCK_PARENT subclasses
- MMAPLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 MMAPLOCK subclasses
- ILOCK uses xfs_lock_inodes with up to 5 inodes, so needs:
- at least 5 ILOCK subclasses
- one ILOCK_PARENT subclass
- one RTBITMAP subclass
- one RTSUM subclass
For the IOLOCK, split the space into two sets of subclasses.
For the MMAPLOCK, just use half the space for the one subclass to
match the non-parent lock classes of the IOLOCK.
For the ILOCK, use 0-4 as the ILOCK subclasses, 5-7 for the
remaining individual subclasses.
Because they are now all different, modify xfs_lock_inumorder() to
handle the nested subclasses, and to assert fail if passed an
invalid subclass. Further, annotate xfs_lock_inodes() to assert fail
if an invalid combination of lock primitives and inode counts are
passed that would result in a lockdep subclass annotation overflow.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 07:32:49 +07:00
|
|
|
* However, MAX_LOCKDEP_SUBCLASSES == 8, which means we are greatly
|
|
|
|
* limited to the subclasses we can represent via nesting. We need at least
|
|
|
|
* 5 inodes nest depth for the ILOCK through rename, and we also have to support
|
|
|
|
* XFS_ILOCK_PARENT, which gives 6 subclasses. Then we have XFS_ILOCK_RTBITMAP
|
|
|
|
* and XFS_ILOCK_RTSUM, which are another 2 unique subclasses, so that's all
|
|
|
|
* 8 subclasses supported by lockdep.
|
|
|
|
*
|
|
|
|
* This also means we have to number the sub-classes in the lowest bits of
|
|
|
|
* the mask we keep, and we have to ensure we never exceed 3 bits of lockdep
|
|
|
|
* mask and we can't use bit-masking to build the subclasses. What a mess.
|
|
|
|
*
|
|
|
|
* Bit layout:
|
|
|
|
*
|
|
|
|
* Bit Lock Region
|
|
|
|
* 16-19 XFS_IOLOCK_SHIFT dependencies
|
|
|
|
* 20-23 XFS_MMAPLOCK_SHIFT dependencies
|
|
|
|
* 24-31 XFS_ILOCK_SHIFT dependencies
|
|
|
|
*
|
|
|
|
* IOLOCK values
|
|
|
|
*
|
|
|
|
* 0-3 subclass value
|
2016-11-30 10:33:25 +07:00
|
|
|
* 4-7 unused
|
xfs: clean up inode lockdep annotations
Lockdep annotations are a maintenance nightmare. Locking has to be
modified to suit the limitations of the annotations, and we're
always having to fix the annotations because they are unable to
express the complexity of locking heirarchies correctly.
So, next up, we've got more issues with lockdep annotations for
inode locking w.r.t. XFS_LOCK_PARENT:
- lockdep classes are exclusive and can't be ORed together
to form new classes.
- IOLOCK needs multiple PARENT subclasses to express the
changes needed for the readdir locking rework needed to
stop the endless flow of lockdep false positives involving
readdir calling filldir under the ILOCK.
- there are only 8 unique lockdep subclasses available,
so we can't create a generic solution.
IOWs we need to treat the 3-bit space available to each lock type
differently:
- IOLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 IOLOCK subclasses
- at least 2 IOLOCK_PARENT subclasses
- MMAPLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 MMAPLOCK subclasses
- ILOCK uses xfs_lock_inodes with up to 5 inodes, so needs:
- at least 5 ILOCK subclasses
- one ILOCK_PARENT subclass
- one RTBITMAP subclass
- one RTSUM subclass
For the IOLOCK, split the space into two sets of subclasses.
For the MMAPLOCK, just use half the space for the one subclass to
match the non-parent lock classes of the IOLOCK.
For the ILOCK, use 0-4 as the ILOCK subclasses, 5-7 for the
remaining individual subclasses.
Because they are now all different, modify xfs_lock_inumorder() to
handle the nested subclasses, and to assert fail if passed an
invalid subclass. Further, annotate xfs_lock_inodes() to assert fail
if an invalid combination of lock primitives and inode counts are
passed that would result in a lockdep subclass annotation overflow.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 07:32:49 +07:00
|
|
|
*
|
|
|
|
* MMAPLOCK values
|
|
|
|
*
|
|
|
|
* 0-3 subclass value
|
|
|
|
* 4-7 unused
|
|
|
|
*
|
|
|
|
* ILOCK values
|
|
|
|
* 0-4 subclass values
|
|
|
|
* 5 PARENT subclass (not nestable)
|
|
|
|
* 6 RTBITMAP subclass (not nestable)
|
|
|
|
* 7 RTSUM subclass (not nestable)
|
|
|
|
*
|
2007-05-08 10:50:19 +07:00
|
|
|
*/
|
xfs: clean up inode lockdep annotations
Lockdep annotations are a maintenance nightmare. Locking has to be
modified to suit the limitations of the annotations, and we're
always having to fix the annotations because they are unable to
express the complexity of locking heirarchies correctly.
So, next up, we've got more issues with lockdep annotations for
inode locking w.r.t. XFS_LOCK_PARENT:
- lockdep classes are exclusive and can't be ORed together
to form new classes.
- IOLOCK needs multiple PARENT subclasses to express the
changes needed for the readdir locking rework needed to
stop the endless flow of lockdep false positives involving
readdir calling filldir under the ILOCK.
- there are only 8 unique lockdep subclasses available,
so we can't create a generic solution.
IOWs we need to treat the 3-bit space available to each lock type
differently:
- IOLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 IOLOCK subclasses
- at least 2 IOLOCK_PARENT subclasses
- MMAPLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 MMAPLOCK subclasses
- ILOCK uses xfs_lock_inodes with up to 5 inodes, so needs:
- at least 5 ILOCK subclasses
- one ILOCK_PARENT subclass
- one RTBITMAP subclass
- one RTSUM subclass
For the IOLOCK, split the space into two sets of subclasses.
For the MMAPLOCK, just use half the space for the one subclass to
match the non-parent lock classes of the IOLOCK.
For the ILOCK, use 0-4 as the ILOCK subclasses, 5-7 for the
remaining individual subclasses.
Because they are now all different, modify xfs_lock_inumorder() to
handle the nested subclasses, and to assert fail if passed an
invalid subclass. Further, annotate xfs_lock_inodes() to assert fail
if an invalid combination of lock primitives and inode counts are
passed that would result in a lockdep subclass annotation overflow.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 07:32:49 +07:00
|
|
|
#define XFS_IOLOCK_SHIFT 16
|
2016-11-30 10:33:25 +07:00
|
|
|
#define XFS_IOLOCK_MAX_SUBCLASS 3
|
xfs: clean up inode lockdep annotations
Lockdep annotations are a maintenance nightmare. Locking has to be
modified to suit the limitations of the annotations, and we're
always having to fix the annotations because they are unable to
express the complexity of locking heirarchies correctly.
So, next up, we've got more issues with lockdep annotations for
inode locking w.r.t. XFS_LOCK_PARENT:
- lockdep classes are exclusive and can't be ORed together
to form new classes.
- IOLOCK needs multiple PARENT subclasses to express the
changes needed for the readdir locking rework needed to
stop the endless flow of lockdep false positives involving
readdir calling filldir under the ILOCK.
- there are only 8 unique lockdep subclasses available,
so we can't create a generic solution.
IOWs we need to treat the 3-bit space available to each lock type
differently:
- IOLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 IOLOCK subclasses
- at least 2 IOLOCK_PARENT subclasses
- MMAPLOCK uses xfs_lock_two_inodes(), so needs:
- at least 2 MMAPLOCK subclasses
- ILOCK uses xfs_lock_inodes with up to 5 inodes, so needs:
- at least 5 ILOCK subclasses
- one ILOCK_PARENT subclass
- one RTBITMAP subclass
- one RTSUM subclass
For the IOLOCK, split the space into two sets of subclasses.
For the MMAPLOCK, just use half the space for the one subclass to
match the non-parent lock classes of the IOLOCK.
For the ILOCK, use 0-4 as the ILOCK subclasses, 5-7 for the
remaining individual subclasses.
Because they are now all different, modify xfs_lock_inumorder() to
handle the nested subclasses, and to assert fail if passed an
invalid subclass. Further, annotate xfs_lock_inodes() to assert fail
if an invalid combination of lock primitives and inode counts are
passed that would result in a lockdep subclass annotation overflow.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
2015-08-19 07:32:49 +07:00
|
|
|
#define XFS_IOLOCK_DEP_MASK 0x000f0000
|
|
|
|
|
|
|
|
#define XFS_MMAPLOCK_SHIFT 20
|
|
|
|
#define XFS_MMAPLOCK_NUMORDER 0
|
|
|
|
#define XFS_MMAPLOCK_MAX_SUBCLASS 3
|
|
|
|
#define XFS_MMAPLOCK_DEP_MASK 0x00f00000
|
|
|
|
|
|
|
|
#define XFS_ILOCK_SHIFT 24
|
|
|
|
#define XFS_ILOCK_PARENT_VAL 5
|
|
|
|
#define XFS_ILOCK_MAX_SUBCLASS (XFS_ILOCK_PARENT_VAL - 1)
|
|
|
|
#define XFS_ILOCK_RTBITMAP_VAL 6
|
|
|
|
#define XFS_ILOCK_RTSUM_VAL 7
|
|
|
|
#define XFS_ILOCK_DEP_MASK 0xff000000
|
|
|
|
#define XFS_ILOCK_PARENT (XFS_ILOCK_PARENT_VAL << XFS_ILOCK_SHIFT)
|
|
|
|
#define XFS_ILOCK_RTBITMAP (XFS_ILOCK_RTBITMAP_VAL << XFS_ILOCK_SHIFT)
|
|
|
|
#define XFS_ILOCK_RTSUM (XFS_ILOCK_RTSUM_VAL << XFS_ILOCK_SHIFT)
|
|
|
|
|
|
|
|
#define XFS_LOCK_SUBCLASS_MASK (XFS_IOLOCK_DEP_MASK | \
|
2015-02-23 17:43:37 +07:00
|
|
|
XFS_MMAPLOCK_DEP_MASK | \
|
|
|
|
XFS_ILOCK_DEP_MASK)
|
|
|
|
|
|
|
|
#define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) \
|
|
|
|
>> XFS_IOLOCK_SHIFT)
|
|
|
|
#define XFS_MMAPLOCK_DEP(flags) (((flags) & XFS_MMAPLOCK_DEP_MASK) \
|
|
|
|
>> XFS_MMAPLOCK_SHIFT)
|
|
|
|
#define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) \
|
|
|
|
>> XFS_ILOCK_SHIFT)
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2018-03-21 04:42:38 +07:00
|
|
|
/*
|
|
|
|
* Layouts are broken in the BREAK_WRITE case to ensure that
|
|
|
|
* layout-holders do not collide with local writes. Additionally,
|
|
|
|
* layouts are broken in the BREAK_UNMAP case to make sure the
|
|
|
|
* layout-holder has a consistent view of the file's extent map. While
|
|
|
|
* BREAK_WRITE breaks can be satisfied by recalling FL_LAYOUT leases,
|
|
|
|
* BREAK_UNMAP breaks additionally require waiting for busy dax-pages to
|
|
|
|
* go idle.
|
|
|
|
*/
|
|
|
|
enum layout_break_reason {
|
|
|
|
BREAK_WRITE,
|
|
|
|
BREAK_UNMAP,
|
|
|
|
};
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
/*
|
|
|
|
* For multiple groups support: if S_ISGID bit is set in the parent
|
|
|
|
* directory, group of new file is set to that of the parent, and
|
|
|
|
* new subdirectory gets S_ISGID bit from parent.
|
|
|
|
*/
|
2007-08-30 14:21:12 +07:00
|
|
|
#define XFS_INHERIT_GID(pip) \
|
|
|
|
(((pip)->i_mount->m_flags & XFS_MOUNT_GRPID) || \
|
2016-02-09 12:54:58 +07:00
|
|
|
(VFS_I(pip)->i_mode & S_ISGID))
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2013-08-12 17:49:45 +07:00
|
|
|
int xfs_release(struct xfs_inode *ip);
|
2013-09-20 22:06:12 +07:00
|
|
|
void xfs_inactive(struct xfs_inode *ip);
|
2013-08-12 17:49:45 +07:00
|
|
|
int xfs_lookup(struct xfs_inode *dp, struct xfs_name *name,
|
|
|
|
struct xfs_inode **ipp, struct xfs_name *ci_name);
|
|
|
|
int xfs_create(struct xfs_inode *dp, struct xfs_name *name,
|
2017-10-20 01:07:09 +07:00
|
|
|
umode_t mode, dev_t rdev, struct xfs_inode **ipp);
|
2018-04-07 00:09:42 +07:00
|
|
|
int xfs_create_tmpfile(struct xfs_inode *dp, umode_t mode,
|
|
|
|
struct xfs_inode **ipp);
|
2013-08-12 17:49:45 +07:00
|
|
|
int xfs_remove(struct xfs_inode *dp, struct xfs_name *name,
|
|
|
|
struct xfs_inode *ip);
|
|
|
|
int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
|
|
|
|
struct xfs_name *target_name);
|
|
|
|
int xfs_rename(struct xfs_inode *src_dp, struct xfs_name *src_name,
|
|
|
|
struct xfs_inode *src_ip, struct xfs_inode *target_dp,
|
|
|
|
struct xfs_name *target_name,
|
2014-12-24 04:51:42 +07:00
|
|
|
struct xfs_inode *target_ip, unsigned int flags);
|
2013-08-12 17:49:45 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
void xfs_ilock(xfs_inode_t *, uint);
|
|
|
|
int xfs_ilock_nowait(xfs_inode_t *, uint);
|
|
|
|
void xfs_iunlock(xfs_inode_t *, uint);
|
|
|
|
void xfs_ilock_demote(xfs_inode_t *, uint);
|
2008-04-22 14:34:00 +07:00
|
|
|
int xfs_isilocked(xfs_inode_t *, uint);
|
2013-12-07 03:30:09 +07:00
|
|
|
uint xfs_ilock_data_map_shared(struct xfs_inode *);
|
2013-12-18 17:14:39 +07:00
|
|
|
uint xfs_ilock_attr_map_shared(struct xfs_inode *);
|
2007-08-28 10:57:51 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
uint xfs_ip2xflags(struct xfs_inode *);
|
2018-07-12 12:26:07 +07:00
|
|
|
int xfs_ifree(struct xfs_trans *, struct xfs_inode *);
|
2018-05-10 23:35:42 +07:00
|
|
|
int xfs_itruncate_extents_flags(struct xfs_trans **,
|
|
|
|
struct xfs_inode *, int, xfs_fsize_t, int);
|
2005-04-17 05:20:36 +07:00
|
|
|
void xfs_iext_realloc(xfs_inode_t *, int, int);
|
2013-08-12 17:49:35 +07:00
|
|
|
|
2020-04-04 01:45:37 +07:00
|
|
|
int xfs_log_force_inode(struct xfs_inode *ip);
|
2010-02-06 08:37:26 +07:00
|
|
|
void xfs_iunpin_wait(xfs_inode_t *);
|
2013-08-12 17:49:35 +07:00
|
|
|
#define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount))
|
|
|
|
|
2020-06-30 04:49:20 +07:00
|
|
|
int xfs_iflush_cluster(struct xfs_buf *);
|
2018-01-27 06:27:33 +07:00
|
|
|
void xfs_lock_two_inodes(struct xfs_inode *ip0, uint ip0_mode,
|
|
|
|
struct xfs_inode *ip1, uint ip1_mode);
|
2005-04-17 05:20:36 +07:00
|
|
|
|
2012-04-23 12:59:02 +07:00
|
|
|
xfs_extlen_t xfs_get_extsz_hint(struct xfs_inode *ip);
|
2016-10-03 23:11:43 +07:00
|
|
|
xfs_extlen_t xfs_get_cowextsz_hint(struct xfs_inode *ip);
|
2012-04-23 12:59:02 +07:00
|
|
|
|
2013-08-12 17:49:47 +07:00
|
|
|
int xfs_dir_ialloc(struct xfs_trans **, struct xfs_inode *, umode_t,
|
2017-12-05 08:32:55 +07:00
|
|
|
xfs_nlink_t, dev_t, prid_t,
|
2018-04-03 05:47:43 +07:00
|
|
|
struct xfs_inode **);
|
2013-08-12 17:49:47 +07:00
|
|
|
|
2018-05-09 22:45:04 +07:00
|
|
|
static inline int
|
|
|
|
xfs_itruncate_extents(
|
|
|
|
struct xfs_trans **tpp,
|
|
|
|
struct xfs_inode *ip,
|
|
|
|
int whichfork,
|
|
|
|
xfs_fsize_t new_size)
|
|
|
|
{
|
2018-05-10 23:35:42 +07:00
|
|
|
return xfs_itruncate_extents_flags(tpp, ip, whichfork, new_size, 0);
|
2018-05-09 22:45:04 +07:00
|
|
|
}
|
|
|
|
|
2013-08-12 17:49:45 +07:00
|
|
|
/* from xfs_file.c */
|
2015-02-02 05:53:56 +07:00
|
|
|
enum xfs_prealloc_flags {
|
|
|
|
XFS_PREALLOC_SET = (1 << 1),
|
|
|
|
XFS_PREALLOC_CLEAR = (1 << 2),
|
|
|
|
XFS_PREALLOC_SYNC = (1 << 3),
|
|
|
|
XFS_PREALLOC_INVISIBLE = (1 << 4),
|
|
|
|
};
|
|
|
|
|
2015-02-23 18:37:08 +07:00
|
|
|
int xfs_update_prealloc_flags(struct xfs_inode *ip,
|
|
|
|
enum xfs_prealloc_flags flags);
|
2018-03-21 04:42:38 +07:00
|
|
|
int xfs_break_layouts(struct inode *inode, uint *iolock,
|
|
|
|
enum layout_break_reason reason);
|
2013-08-12 17:49:45 +07:00
|
|
|
|
2015-02-23 18:38:08 +07:00
|
|
|
/* from xfs_iops.c */
|
2016-04-06 04:48:27 +07:00
|
|
|
extern void xfs_setup_inode(struct xfs_inode *ip);
|
|
|
|
extern void xfs_setup_iops(struct xfs_inode *ip);
|
2020-05-04 23:02:43 +07:00
|
|
|
extern void xfs_diflags_to_iflags(struct xfs_inode *ip, bool init);
|
2016-04-06 04:48:27 +07:00
|
|
|
|
2015-02-23 18:38:08 +07:00
|
|
|
/*
|
|
|
|
* When setting up a newly allocated inode, we need to call
|
|
|
|
* xfs_finish_inode_setup() once the inode is fully instantiated at
|
|
|
|
* the VFS level to prevent the rest of the world seeing the inode
|
|
|
|
* before we've completed instantiation. Otherwise we can do it
|
|
|
|
* the moment the inode lookup is complete.
|
|
|
|
*/
|
|
|
|
static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
|
|
|
|
{
|
|
|
|
xfs_iflags_clear(ip, XFS_INEW);
|
|
|
|
barrier();
|
|
|
|
unlock_new_inode(VFS_I(ip));
|
2017-04-26 22:30:39 +07:00
|
|
|
wake_up_bit(&ip->i_flags, __XFS_INEW_BIT);
|
2015-02-23 18:38:08 +07:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void xfs_setup_existing_inode(struct xfs_inode *ip)
|
|
|
|
{
|
|
|
|
xfs_setup_inode(ip);
|
2016-04-06 04:48:27 +07:00
|
|
|
xfs_setup_iops(ip);
|
2015-02-23 18:38:08 +07:00
|
|
|
xfs_finish_inode_setup(ip);
|
|
|
|
}
|
|
|
|
|
2018-07-26 02:52:32 +07:00
|
|
|
void xfs_irele(struct xfs_inode *ip);
|
2008-12-03 18:20:40 +07:00
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
extern struct kmem_zone *xfs_inode_zone;
|
|
|
|
|
2016-10-03 23:11:49 +07:00
|
|
|
/* The default CoW extent size hint. */
|
|
|
|
#define XFS_DEFAULT_COWEXTSZ_HINT 32
|
|
|
|
|
2019-02-08 01:37:16 +07:00
|
|
|
int xfs_iunlink_init(struct xfs_perag *pag);
|
|
|
|
void xfs_iunlink_destroy(struct xfs_perag *pag);
|
|
|
|
|
xfs: implement per-inode writeback completion queues
When scheduling writeback of dirty file data in the page cache, XFS uses
IO completion workqueue items to ensure that filesystem metadata only
updates after the write completes successfully. This is essential for
converting unwritten extents to real extents at the right time and
performing COW remappings.
Unfortunately, XFS queues each IO completion work item to an unbounded
workqueue, which means that the kernel can spawn dozens of threads to
try to handle the items quickly. These threads need to take the ILOCK
to update file metadata, which results in heavy ILOCK contention if a
large number of the work items target a single file, which is
inefficient.
Worse yet, the writeback completion threads get stuck waiting for the
ILOCK while holding transaction reservations, which can use up all
available log reservation space. When that happens, metadata updates to
other parts of the filesystem grind to a halt, even if the filesystem
could otherwise have handled it.
Even worse, if one of the things grinding to a halt happens to be a
thread in the middle of a defer-ops finish holding the same ILOCK and
trying to obtain more log reservation having exhausted the permanent
reservation, we now have an ABBA deadlock - writeback completion has a
transaction reserved and wants the ILOCK, and someone else has the ILOCK
and wants a transaction reservation.
Therefore, we create a per-inode writeback io completion queue + work
item. When writeback finishes, it can add the ioend to the per-inode
queue and let the single worker item process that queue. This
dramatically cuts down on the number of kworkers and ILOCK contention in
the system, and seems to have eliminated an occasional deadlock I was
seeing while running generic/476.
Testing with a program that simulates a heavy random-write workload to a
single file demonstrates that the number of kworkers drops from
approximately 120 threads per file to 1, without dramatically changing
write bandwidth or pagecache access latency.
Note that we leave the xfs-conv workqueue's max_active alone because we
still want to be able to run ioend processing for as many inodes as the
system can handle.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
2019-04-16 03:13:20 +07:00
|
|
|
void xfs_end_io(struct work_struct *work);
|
|
|
|
|
2020-06-30 04:47:20 +07:00
|
|
|
int xfs_ilock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
|
|
|
|
void xfs_iunlock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
|
|
|
|
|
2005-04-17 05:20:36 +07:00
|
|
|
#endif /* __XFS_INODE_H__ */
|