2005-04-17 06:20:36 +08:00
|
|
|
/*
|
2005-11-02 11:58:39 +08:00
|
|
|
* Copyright (c) 2000-2005 Silicon Graphics, Inc.
|
|
|
|
* All Rights Reserved.
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
2005-11-02 11:58:39 +08:00
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public License as
|
2005-04-17 06:20:36 +08:00
|
|
|
* published by the Free Software Foundation.
|
|
|
|
*
|
2005-11-02 11:58:39 +08:00
|
|
|
* This program is distributed in the hope that it would be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
|
* GNU General Public License for more details.
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
2005-11-02 11:58:39 +08:00
|
|
|
* You should have received a copy of the GNU General Public License
|
|
|
|
* along with this program; if not, write the Free Software Foundation,
|
|
|
|
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
|
|
|
#ifndef __XFS_MOUNT_H__
|
|
|
|
#define __XFS_MOUNT_H__
|
|
|
|
|
|
|
|
typedef struct xfs_trans_reservations {
|
|
|
|
uint tr_write; /* extent alloc trans */
|
|
|
|
uint tr_itruncate; /* truncate trans */
|
|
|
|
uint tr_rename; /* rename trans */
|
|
|
|
uint tr_link; /* link trans */
|
|
|
|
uint tr_remove; /* unlink trans */
|
|
|
|
uint tr_symlink; /* symlink trans */
|
|
|
|
uint tr_create; /* create trans */
|
|
|
|
uint tr_mkdir; /* mkdir trans */
|
|
|
|
uint tr_ifree; /* inode free trans */
|
|
|
|
uint tr_ichange; /* inode update trans */
|
|
|
|
uint tr_growdata; /* fs data section grow trans */
|
|
|
|
uint tr_swrite; /* sync write inode trans */
|
|
|
|
uint tr_addafork; /* cvt inode to attributed trans */
|
|
|
|
uint tr_writeid; /* write setuid/setgid file */
|
|
|
|
uint tr_attrinval; /* attr fork buffer invalidation */
|
|
|
|
uint tr_attrset; /* set/create an attribute */
|
|
|
|
uint tr_attrrm; /* remove an attribute */
|
|
|
|
uint tr_clearagi; /* clear bad agi unlinked ino bucket */
|
|
|
|
uint tr_growrtalloc; /* grow realtime allocations */
|
|
|
|
uint tr_growrtzero; /* grow realtime zeroing */
|
|
|
|
uint tr_growrtfree; /* grow realtime freeing */
|
|
|
|
} xfs_trans_reservations_t;
|
|
|
|
|
|
|
|
#ifndef __KERNEL__
|
2008-10-30 14:05:38 +08:00
|
|
|
|
2009-01-15 13:22:07 +08:00
|
|
|
#define xfs_daddr_to_agno(mp,d) \
|
2005-04-17 06:20:36 +08:00
|
|
|
((xfs_agnumber_t)(XFS_BB_TO_FSBT(mp, d) / (mp)->m_sb.sb_agblocks))
|
2009-01-15 13:22:07 +08:00
|
|
|
#define xfs_daddr_to_agbno(mp,d) \
|
2005-04-17 06:20:36 +08:00
|
|
|
((xfs_agblock_t)(XFS_BB_TO_FSBT(mp, d) % (mp)->m_sb.sb_agblocks))
|
2008-10-30 14:05:38 +08:00
|
|
|
|
|
|
|
#else /* __KERNEL__ */
|
|
|
|
|
2008-12-09 17:47:32 +08:00
|
|
|
#include "xfs_sync.h"
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
struct log;
|
|
|
|
struct xfs_mount_args;
|
|
|
|
struct xfs_inode;
|
|
|
|
struct xfs_bmbt_irec;
|
|
|
|
struct xfs_bmap_free;
|
2006-06-09 12:48:12 +08:00
|
|
|
struct xfs_extdelta;
|
|
|
|
struct xfs_swapext;
|
2007-07-11 09:09:12 +08:00
|
|
|
struct xfs_mru_cache;
|
2008-05-21 14:41:01 +08:00
|
|
|
struct xfs_nameops;
|
2008-10-30 14:38:26 +08:00
|
|
|
struct xfs_ail;
|
2009-06-08 21:33:32 +08:00
|
|
|
struct xfs_quotainfo;
|
|
|
|
|
2006-03-14 10:13:09 +08:00
|
|
|
#ifdef HAVE_PERCPU_SB
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Valid per-cpu incore superblock counters. Note that if you add new counters,
|
|
|
|
* you may need to define new counter disabled bit field descriptors as there
|
|
|
|
* are more possible fields in the superblock that can fit in a bitfield on a
|
|
|
|
* 32 bit platform. The XFS_SBS_* values for the current current counters just
|
|
|
|
* fit.
|
|
|
|
*/
|
|
|
|
typedef struct xfs_icsb_cnts {
|
|
|
|
uint64_t icsb_fdblocks;
|
|
|
|
uint64_t icsb_ifree;
|
|
|
|
uint64_t icsb_icount;
|
2006-03-14 10:29:16 +08:00
|
|
|
unsigned long icsb_flags;
|
2006-03-14 10:13:09 +08:00
|
|
|
} xfs_icsb_cnts_t;
|
|
|
|
|
2006-03-14 10:29:16 +08:00
|
|
|
#define XFS_ICSB_FLAG_LOCK (1 << 0) /* counter lock bit */
|
|
|
|
|
2006-03-14 10:13:09 +08:00
|
|
|
#define XFS_ICSB_LAZY_COUNT (1 << 1) /* accuracy not needed */
|
|
|
|
|
|
|
|
extern int xfs_icsb_init_counters(struct xfs_mount *);
|
2007-02-10 15:36:29 +08:00
|
|
|
extern void xfs_icsb_reinit_counters(struct xfs_mount *);
|
2008-05-20 13:10:52 +08:00
|
|
|
extern void xfs_icsb_destroy_counters(struct xfs_mount *);
|
2008-04-22 15:34:37 +08:00
|
|
|
extern void xfs_icsb_sync_counters(struct xfs_mount *, int);
|
|
|
|
extern void xfs_icsb_sync_counters_locked(struct xfs_mount *, int);
|
2010-09-30 10:25:55 +08:00
|
|
|
extern int xfs_icsb_modify_counters(struct xfs_mount *, xfs_sb_field_t,
|
|
|
|
int64_t, int);
|
2006-03-14 10:13:09 +08:00
|
|
|
|
|
|
|
#else
|
2008-05-20 13:10:52 +08:00
|
|
|
#define xfs_icsb_init_counters(mp) (0)
|
|
|
|
#define xfs_icsb_destroy_counters(mp) do { } while (0)
|
|
|
|
#define xfs_icsb_reinit_counters(mp) do { } while (0)
|
2008-04-22 15:34:37 +08:00
|
|
|
#define xfs_icsb_sync_counters(mp, flags) do { } while (0)
|
2008-04-29 10:53:00 +08:00
|
|
|
#define xfs_icsb_sync_counters_locked(mp, flags) do { } while (0)
|
2010-09-30 10:25:55 +08:00
|
|
|
#define xfs_icsb_modify_counters(mp, field, delta, rsvd) \
|
|
|
|
xfs_mod_incore_sb(mp, field, delta, rsvd)
|
2006-03-14 10:13:09 +08:00
|
|
|
#endif
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2011-01-04 08:35:03 +08:00
|
|
|
/* dynamic preallocation free space thresholds, 5% down to 1% */
|
|
|
|
enum {
|
|
|
|
XFS_LOWSP_1_PCNT = 0,
|
|
|
|
XFS_LOWSP_2_PCNT,
|
|
|
|
XFS_LOWSP_3_PCNT,
|
|
|
|
XFS_LOWSP_4_PCNT,
|
|
|
|
XFS_LOWSP_5_PCNT,
|
|
|
|
XFS_LOWSP_MAX,
|
|
|
|
};
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
typedef struct xfs_mount {
|
2007-08-30 15:21:30 +08:00
|
|
|
struct super_block *m_super;
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_tid_t m_tid; /* next unused tid for fs */
|
2008-10-30 14:38:26 +08:00
|
|
|
struct xfs_ail *m_ail; /* fs active log item list */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_sb_t m_sb; /* copy of fs superblock */
|
2007-10-11 15:42:32 +08:00
|
|
|
spinlock_t m_sb_lock; /* sb counter lock */
|
2005-04-17 06:20:36 +08:00
|
|
|
struct xfs_buf *m_sb_bp; /* buffer for superblock */
|
|
|
|
char *m_fsname; /* filesystem name */
|
|
|
|
int m_fsname_len; /* strlen of fs name */
|
2005-11-02 08:44:33 +08:00
|
|
|
char *m_rtname; /* realtime device name */
|
|
|
|
char *m_logname; /* external log device name */
|
2005-04-17 06:20:36 +08:00
|
|
|
int m_bsize; /* fs logical block size */
|
|
|
|
xfs_agnumber_t m_agfrotor; /* last ag where space found */
|
|
|
|
xfs_agnumber_t m_agirotor; /* last ag dir inode alloced */
|
2007-10-11 15:43:43 +08:00
|
|
|
spinlock_t m_agirotor_lock;/* .. and lock protecting it */
|
2005-04-17 06:20:36 +08:00
|
|
|
xfs_agnumber_t m_maxagi; /* highest inode alloc group */
|
|
|
|
uint m_readio_log; /* min read size log bytes */
|
|
|
|
uint m_readio_blocks; /* min read size blocks */
|
|
|
|
uint m_writeio_log; /* min write size log bytes */
|
|
|
|
uint m_writeio_blocks; /* min write size blocks */
|
|
|
|
struct log *m_log; /* log specific stuff */
|
|
|
|
int m_logbufs; /* number of log buffers */
|
|
|
|
int m_logbsize; /* size of each log buffer */
|
|
|
|
uint m_rsumlevels; /* rt summary levels */
|
|
|
|
uint m_rsumsize; /* size of rt summary, bytes */
|
|
|
|
struct xfs_inode *m_rbmip; /* pointer to bitmap inode */
|
|
|
|
struct xfs_inode *m_rsumip; /* pointer to summary inode */
|
|
|
|
struct xfs_inode *m_rootip; /* pointer to root directory */
|
|
|
|
struct xfs_quotainfo *m_quotainfo; /* disk quota information */
|
|
|
|
xfs_buftarg_t *m_ddev_targp; /* saves taking the address */
|
|
|
|
xfs_buftarg_t *m_logdev_targp;/* ptr to log device */
|
|
|
|
xfs_buftarg_t *m_rtdev_targp; /* ptr to rt device */
|
|
|
|
__uint8_t m_blkbit_log; /* blocklog + NBBY */
|
|
|
|
__uint8_t m_blkbb_log; /* blocklog - BBSHIFT */
|
|
|
|
__uint8_t m_agno_log; /* log #ag's */
|
|
|
|
__uint8_t m_agino_log; /* #bits for agino in inum */
|
|
|
|
__uint16_t m_inode_cluster_size;/* min inode buf size */
|
|
|
|
uint m_blockmask; /* sb_blocksize-1 */
|
|
|
|
uint m_blockwsize; /* sb_blocksize in words */
|
|
|
|
uint m_blockwmask; /* blockwsize-1 */
|
2008-10-30 14:11:19 +08:00
|
|
|
uint m_alloc_mxr[2]; /* max alloc btree records */
|
|
|
|
uint m_alloc_mnr[2]; /* min alloc btree records */
|
|
|
|
uint m_bmap_dmxr[2]; /* max bmap btree records */
|
|
|
|
uint m_bmap_dmnr[2]; /* min bmap btree records */
|
|
|
|
uint m_inobt_mxr[2]; /* max inobt btree records */
|
|
|
|
uint m_inobt_mnr[2]; /* min inobt btree records */
|
2005-04-17 06:20:36 +08:00
|
|
|
uint m_ag_maxlevels; /* XFS_AG_MAXLEVELS */
|
|
|
|
uint m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
|
2009-02-09 15:37:14 +08:00
|
|
|
uint m_in_maxlevels; /* max inobt btree levels. */
|
2010-01-11 19:47:44 +08:00
|
|
|
struct radix_tree_root m_perag_tree; /* per-ag accounting info */
|
|
|
|
spinlock_t m_perag_lock; /* lock for m_perag_tree */
|
2007-08-30 15:21:54 +08:00
|
|
|
struct mutex m_growlock; /* growfs mutex */
|
2005-04-17 06:20:36 +08:00
|
|
|
int m_fixedfsid[2]; /* unchanged for life of FS */
|
|
|
|
uint m_dmevmask; /* DMI events for this FS */
|
2005-11-02 12:09:22 +08:00
|
|
|
__uint64_t m_flags; /* global mount flags */
|
2005-04-17 06:20:36 +08:00
|
|
|
uint m_dir_node_ents; /* #entries in a dir danode */
|
|
|
|
uint m_attr_node_ents; /* #entries in attr danode */
|
|
|
|
int m_ialloc_inos; /* inodes in inode allocation */
|
|
|
|
int m_ialloc_blks; /* blocks in inode allocation */
|
|
|
|
int m_inoalign_mask;/* mask sb_inoalignmt if used */
|
|
|
|
uint m_qflags; /* quota status flags */
|
|
|
|
xfs_trans_reservations_t m_reservations;/* precomputed res values */
|
|
|
|
__uint64_t m_maxicount; /* maximum inode count */
|
|
|
|
__uint64_t m_maxioffset; /* maximum inode offset */
|
|
|
|
__uint64_t m_resblks; /* total reserved blocks */
|
|
|
|
__uint64_t m_resblks_avail;/* available reserved blocks */
|
xfs: don't hold onto reserved blocks on remount,ro
If we hold onto reserved blocks when doing a remount,ro we end
up writing the blocks used count to disk that includes the reserved
blocks. Reserved blocks are not actually used, so this results in
the values in the superblock being incorrect.
Hence if we run xfs_check or xfs_repair -n while the filesystem is
mounted remount,ro we end up with an inconsistent filesystem being
reported. Also, running xfs_copy on the remount,ro filesystem will
result in an inconsistent image being generated.
To fix this, unreserve the blocks when doing the remount,ro, and
reserved them again on remount,rw. This way a remount,ro filesystem
will appear consistent on disk to all utilities.
Signed-off-by: Dave Chinner <david@fromorbit.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
2010-01-26 12:08:49 +08:00
|
|
|
__uint64_t m_resblks_save; /* reserved blks @ remount,ro */
|
2005-04-17 06:20:36 +08:00
|
|
|
int m_dalign; /* stripe unit */
|
|
|
|
int m_swidth; /* stripe width */
|
2006-03-29 06:55:14 +08:00
|
|
|
int m_sinoalign; /* stripe unit inode alignment */
|
2005-04-17 06:20:36 +08:00
|
|
|
int m_attr_magicpct;/* 37% of the blocksize */
|
|
|
|
int m_dir_magicpct; /* 37% of the dir blocksize */
|
|
|
|
__uint8_t m_sectbb_log; /* sectlog - BBSHIFT */
|
2008-05-21 14:41:01 +08:00
|
|
|
const struct xfs_nameops *m_dirnameops; /* vector of dir name ops */
|
2005-04-17 06:20:36 +08:00
|
|
|
int m_dirblksize; /* directory block sz--bytes */
|
|
|
|
int m_dirblkfsbs; /* directory block sz--fsbs */
|
|
|
|
xfs_dablk_t m_dirdatablk; /* blockno of dir data v2 */
|
|
|
|
xfs_dablk_t m_dirleafblk; /* blockno of dir non-data v2 */
|
|
|
|
xfs_dablk_t m_dirfreeblk; /* blockno of dirfreeindex v2 */
|
|
|
|
uint m_chsize; /* size of next field */
|
|
|
|
struct xfs_chash *m_chash; /* fs private inode per-cluster
|
|
|
|
* hash table */
|
|
|
|
atomic_t m_active_trans; /* number trans frozen */
|
2006-03-14 10:13:09 +08:00
|
|
|
#ifdef HAVE_PERCPU_SB
|
2010-02-02 13:39:01 +08:00
|
|
|
xfs_icsb_cnts_t __percpu *m_sb_cnts; /* per-cpu superblock counters */
|
2006-03-14 10:13:09 +08:00
|
|
|
unsigned long m_icsb_counters; /* disabled per-cpu counters */
|
2006-03-14 10:23:52 +08:00
|
|
|
struct notifier_block m_icsb_notifier; /* hotplug cpu notifier */
|
2007-02-10 15:35:09 +08:00
|
|
|
struct mutex m_icsb_mutex; /* balancer sync lock */
|
2006-03-14 10:13:09 +08:00
|
|
|
#endif
|
2007-07-11 09:09:12 +08:00
|
|
|
struct xfs_mru_cache *m_filestream; /* per-mount filestream data */
|
2011-04-08 10:45:07 +08:00
|
|
|
struct delayed_work m_sync_work; /* background sync work */
|
2011-04-08 10:45:07 +08:00
|
|
|
struct delayed_work m_reclaim_work; /* background inode reclaim */
|
2011-04-08 10:45:07 +08:00
|
|
|
struct work_struct m_flush_work; /* background inode flush */
|
2009-01-19 09:04:07 +08:00
|
|
|
__int64_t m_update_flags; /* sb flags we need to update
|
|
|
|
on the next remount,rw */
|
2010-07-20 06:07:02 +08:00
|
|
|
struct shrinker m_inode_shrink; /* inode reclaim shrinker */
|
2011-01-04 08:35:03 +08:00
|
|
|
int64_t m_low_space[XFS_LOWSP_MAX];
|
|
|
|
/* low free space thresholds */
|
2005-04-17 06:20:36 +08:00
|
|
|
} xfs_mount_t;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Flags for m_flags.
|
|
|
|
*/
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
#define XFS_MOUNT_WSYNC (1ULL << 0) /* for nfs - all metadata ops
|
2005-04-17 06:20:36 +08:00
|
|
|
must be synchronous except
|
|
|
|
for space allocations */
|
xfs: Introduce delayed logging core code
The delayed logging code only changes in-memory structures and as
such can be enabled and disabled with a mount option. Add the mount
option and emit a warning that this is an experimental feature that
should not be used in production yet.
We also need infrastructure to track committed items that have not
yet been written to the log. This is what the Committed Item List
(CIL) is for.
The log item also needs to be extended to track the current log
vector, the associated memory buffer and it's location in the Commit
Item List. Extend the log item and log vector structures to enable
this tracking.
To maintain the current log format for transactions with delayed
logging, we need to introduce a checkpoint transaction and a context
for tracking each checkpoint from initiation to transaction
completion. This includes adding a log ticket for tracking space
log required/used by the context checkpoint.
To track all the changes we need an io vector array per log item,
rather than a single array for the entire transaction. Using the new
log vector structure for this requires two passes - the first to
allocate the log vector structures and chain them together, and the
second to fill them out. This log vector chain can then be passed
to the CIL for formatting, pinning and insertion into the CIL.
Formatting of the log vector chain is relatively simple - it's just
a loop over the iovecs on each log vector, but it is made slightly
more complex because we re-write the iovec after the copy to point
back at the memory buffer we just copied into.
This code also needs to pin log items. If the log item is not
already tracked in this checkpoint context, then it needs to be
pinned. Otherwise it is already pinned and we don't need to pin it
again.
The only other complexity is calculating the amount of new log space
the formatting has consumed. This needs to be accounted to the
transaction in progress, and the accounting is made more complex
becase we need also to steal space from it for log metadata in the
checkpoint transaction. Calculate all this at insert time and update
all the tickets, counters, etc correctly.
Once we've formatted all the log items in the transaction, attach
the busy extents to the checkpoint context so the busy extents live
until checkpoint completion and can be processed at that point in
time. Transactions can then be freed at this point in time.
Now we need to issue checkpoints - we are tracking the amount of log space
used by the items in the CIL, so we can trigger background checkpoints when the
space usage gets to a certain threshold. Otherwise, checkpoints need ot be
triggered when a log synchronisation point is reached - a log force event.
Because the log write code already handles chained log vectors, writing the
transaction is trivial, too. Construct a transaction header, add it
to the head of the chain and write it into the log, then issue a
commit record write. Then we can release the checkpoint log ticket
and attach the context to the log buffer so it can be called during
Io completion to complete the checkpoint.
We also need to allow for synchronising multiple in-flight
checkpoints. This is needed for two things - the first is to ensure
that checkpoint commit records appear in the log in the correct
sequence order (so they are replayed in the correct order). The
second is so that xfs_log_force_lsn() operates correctly and only
flushes and/or waits for the specific sequence it was provided with.
To do this we need a wait variable and a list tracking the
checkpoint commits in progress. We can walk this list and wait for
the checkpoints to change state or complete easily, an this provides
the necessary synchronisation for correct operation in both cases.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
2010-05-21 12:37:18 +08:00
|
|
|
#define XFS_MOUNT_DELAYLOG (1ULL << 1) /* delayed logging is enabled */
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
#define XFS_MOUNT_WAS_CLEAN (1ULL << 3)
|
2005-11-02 12:09:22 +08:00
|
|
|
#define XFS_MOUNT_FS_SHUTDOWN (1ULL << 4) /* atomic stop of all filesystem
|
2005-04-17 06:20:36 +08:00
|
|
|
operations, typically for
|
|
|
|
disk errors in metadata */
|
2011-05-20 21:45:32 +08:00
|
|
|
#define XFS_MOUNT_DISCARD (1ULL << 5) /* discard unused blocks */
|
2005-11-02 12:09:22 +08:00
|
|
|
#define XFS_MOUNT_RETERR (1ULL << 6) /* return alignment errors to
|
2005-04-17 06:20:36 +08:00
|
|
|
user */
|
2005-11-02 12:09:22 +08:00
|
|
|
#define XFS_MOUNT_NOALIGN (1ULL << 7) /* turn off stripe alignment
|
2005-04-17 06:20:36 +08:00
|
|
|
allocations */
|
2006-01-11 12:32:01 +08:00
|
|
|
#define XFS_MOUNT_ATTR2 (1ULL << 8) /* allow use of attr2 format */
|
2007-08-30 15:21:12 +08:00
|
|
|
#define XFS_MOUNT_GRPID (1ULL << 9) /* group-ID assigned from directory */
|
2005-11-02 12:09:22 +08:00
|
|
|
#define XFS_MOUNT_NORECOVERY (1ULL << 10) /* no recovery - dirty fs */
|
|
|
|
#define XFS_MOUNT_DFLT_IOSIZE (1ULL << 12) /* set default i/o size */
|
|
|
|
#define XFS_MOUNT_32BITINODES (1ULL << 14) /* do not create inodes above
|
2005-04-17 06:20:36 +08:00
|
|
|
* 32 bits in size */
|
2007-08-30 15:21:12 +08:00
|
|
|
#define XFS_MOUNT_SMALL_INUMS (1ULL << 15) /* users wants 32bit inodes */
|
2005-11-02 12:09:22 +08:00
|
|
|
#define XFS_MOUNT_NOUUID (1ULL << 16) /* ignore uuid during mount */
|
|
|
|
#define XFS_MOUNT_BARRIER (1ULL << 17)
|
2008-02-29 10:58:40 +08:00
|
|
|
#define XFS_MOUNT_IKEEP (1ULL << 18) /* keep empty inode clusters*/
|
2005-11-02 12:09:22 +08:00
|
|
|
#define XFS_MOUNT_SWALLOC (1ULL << 19) /* turn on stripe width
|
2005-04-17 06:20:36 +08:00
|
|
|
* allocation */
|
2007-08-30 15:21:12 +08:00
|
|
|
#define XFS_MOUNT_RDONLY (1ULL << 20) /* read-only fs */
|
2005-11-02 12:09:22 +08:00
|
|
|
#define XFS_MOUNT_DIRSYNC (1ULL << 21) /* synchronous directory ops */
|
|
|
|
#define XFS_MOUNT_COMPAT_IOSIZE (1ULL << 22) /* don't report large preferred
|
2005-11-02 07:33:05 +08:00
|
|
|
* I/O size in stat() */
|
2007-07-11 09:09:12 +08:00
|
|
|
#define XFS_MOUNT_FILESTREAMS (1ULL << 24) /* enable the filestreams
|
|
|
|
allocator */
|
2008-04-30 16:15:28 +08:00
|
|
|
#define XFS_MOUNT_NOATTR2 (1ULL << 25) /* disable use of attr2 format */
|
2005-11-02 07:33:05 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Default minimum read and write sizes.
|
|
|
|
*/
|
|
|
|
#define XFS_READIO_LOG_LARGE 16
|
|
|
|
#define XFS_WRITEIO_LOG_LARGE 16
|
|
|
|
|
|
|
|
/*
|
2005-05-06 04:28:29 +08:00
|
|
|
* Max and min values for mount-option defined I/O
|
|
|
|
* preallocation sizes.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2005-05-06 04:28:29 +08:00
|
|
|
#define XFS_MAX_IO_LOG 30 /* 1G */
|
2005-04-17 06:20:36 +08:00
|
|
|
#define XFS_MIN_IO_LOG PAGE_SHIFT
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Synchronous read and write sizes. This should be
|
|
|
|
* better for NFSv2 wsync filesystems.
|
|
|
|
*/
|
2009-03-29 15:55:42 +08:00
|
|
|
#define XFS_WSYNC_READIO_LOG 15 /* 32k */
|
|
|
|
#define XFS_WSYNC_WRITEIO_LOG 14 /* 16k */
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-11-02 07:33:05 +08:00
|
|
|
/*
|
|
|
|
* Allow large block sizes to be reported to userspace programs if the
|
2008-04-10 10:22:07 +08:00
|
|
|
* "largeio" mount option is used.
|
2005-11-02 07:33:05 +08:00
|
|
|
*
|
|
|
|
* If compatibility mode is specified, simply return the basic unit of caching
|
|
|
|
* so that we don't get inefficient read/modify/write I/O from user apps.
|
|
|
|
* Otherwise....
|
|
|
|
*
|
|
|
|
* If the underlying volume is a stripe, then return the stripe width in bytes
|
|
|
|
* as the recommended I/O size. It is not a stripe and we've set a default
|
|
|
|
* buffered I/O size, return that, otherwise return the compat default.
|
|
|
|
*/
|
|
|
|
static inline unsigned long
|
|
|
|
xfs_preferred_iosize(xfs_mount_t *mp)
|
|
|
|
{
|
|
|
|
if (mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE)
|
|
|
|
return PAGE_CACHE_SIZE;
|
|
|
|
return (mp->m_swidth ?
|
|
|
|
(mp->m_swidth << mp->m_sb.sb_blocklog) :
|
|
|
|
((mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) ?
|
|
|
|
(1 << (int)MAX(mp->m_readio_log, mp->m_writeio_log)) :
|
|
|
|
PAGE_CACHE_SIZE));
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
#define XFS_MAXIOFFSET(mp) ((mp)->m_maxioffset)
|
|
|
|
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
#define XFS_LAST_UNMOUNT_WAS_CLEAN(mp) \
|
|
|
|
((mp)->m_flags & XFS_MOUNT_WAS_CLEAN)
|
2005-04-17 06:20:36 +08:00
|
|
|
#define XFS_FORCED_SHUTDOWN(mp) ((mp)->m_flags & XFS_MOUNT_FS_SHUTDOWN)
|
2007-08-30 15:20:39 +08:00
|
|
|
void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
|
|
|
|
int lnnum);
|
2005-04-17 06:20:36 +08:00
|
|
|
#define xfs_force_shutdown(m,f) \
|
2007-08-30 15:20:39 +08:00
|
|
|
xfs_do_force_shutdown(m, f, __FILE__, __LINE__)
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-11-28 11:23:36 +08:00
|
|
|
#define SHUTDOWN_META_IO_ERROR 0x0001 /* write attempt to metadata failed */
|
|
|
|
#define SHUTDOWN_LOG_IO_ERROR 0x0002 /* write attempt to the log failed */
|
|
|
|
#define SHUTDOWN_FORCE_UMOUNT 0x0004 /* shutdown from a forced unmount */
|
|
|
|
#define SHUTDOWN_CORRUPT_INCORE 0x0008 /* corrupt in-memory data structures */
|
|
|
|
#define SHUTDOWN_REMOTE_REQ 0x0010 /* shutdown came from remote cell */
|
|
|
|
#define SHUTDOWN_DEVICE_REQ 0x0020 /* failed all paths to the device */
|
|
|
|
|
|
|
|
#define xfs_test_for_freeze(mp) ((mp)->m_super->s_frozen)
|
|
|
|
#define xfs_wait_for_freeze(mp,l) vfs_check_frozen((mp)->m_super, (l))
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* Flags for xfs_mountfs
|
|
|
|
*/
|
2006-03-31 11:04:17 +08:00
|
|
|
#define XFS_MFSI_QUIET 0x40 /* Be silent if mount errors found */
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2005-11-02 11:38:42 +08:00
|
|
|
static inline xfs_agnumber_t
|
|
|
|
xfs_daddr_to_agno(struct xfs_mount *mp, xfs_daddr_t d)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2005-11-02 11:38:42 +08:00
|
|
|
xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
|
|
|
|
do_div(ld, mp->m_sb.sb_agblocks);
|
|
|
|
return (xfs_agnumber_t) ld;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2005-11-02 11:38:42 +08:00
|
|
|
static inline xfs_agblock_t
|
|
|
|
xfs_daddr_to_agbno(struct xfs_mount *mp, xfs_daddr_t d)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2005-11-02 11:38:42 +08:00
|
|
|
xfs_daddr_t ld = XFS_BB_TO_FSBT(mp, d);
|
|
|
|
return (xfs_agblock_t) do_div(ld, mp->m_sb.sb_agblocks);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2007-08-28 12:00:13 +08:00
|
|
|
/*
|
2010-01-11 19:47:45 +08:00
|
|
|
* perag get/put wrappers for ref counting
|
2007-08-28 12:00:13 +08:00
|
|
|
*/
|
2010-01-11 19:47:46 +08:00
|
|
|
struct xfs_perag *xfs_perag_get(struct xfs_mount *mp, xfs_agnumber_t agno);
|
2010-09-24 16:40:15 +08:00
|
|
|
struct xfs_perag *xfs_perag_get_tag(struct xfs_mount *mp, xfs_agnumber_t agno,
|
|
|
|
int tag);
|
2010-01-11 19:47:46 +08:00
|
|
|
void xfs_perag_put(struct xfs_perag *pag);
|
2007-08-28 12:00:13 +08:00
|
|
|
|
2007-02-10 15:35:15 +08:00
|
|
|
/*
|
|
|
|
* Per-cpu superblock locking functions
|
|
|
|
*/
|
|
|
|
#ifdef HAVE_PERCPU_SB
|
2009-11-15 00:17:22 +08:00
|
|
|
static inline void
|
2007-02-10 15:35:15 +08:00
|
|
|
xfs_icsb_lock(xfs_mount_t *mp)
|
|
|
|
{
|
|
|
|
mutex_lock(&mp->m_icsb_mutex);
|
|
|
|
}
|
|
|
|
|
2009-11-15 00:17:22 +08:00
|
|
|
static inline void
|
2007-02-10 15:35:15 +08:00
|
|
|
xfs_icsb_unlock(xfs_mount_t *mp)
|
|
|
|
{
|
|
|
|
mutex_unlock(&mp->m_icsb_mutex);
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
#define xfs_icsb_lock(mp)
|
|
|
|
#define xfs_icsb_unlock(mp)
|
|
|
|
#endif
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/*
|
|
|
|
* This structure is for use by the xfs_mod_incore_sb_batch() routine.
|
2007-02-10 15:36:10 +08:00
|
|
|
* xfs_growfs can specify a few fields which are more than int limit
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
|
|
|
typedef struct xfs_mod_sb {
|
|
|
|
xfs_sb_field_t msb_field; /* Field to modify, see below */
|
2007-02-10 15:36:10 +08:00
|
|
|
int64_t msb_delta; /* Change to make to specified field */
|
2005-04-17 06:20:36 +08:00
|
|
|
} xfs_mod_sb_t;
|
|
|
|
|
2011-06-30 06:10:14 +08:00
|
|
|
extern int xfs_log_sbcount(xfs_mount_t *);
|
2010-02-06 06:59:53 +08:00
|
|
|
extern __uint64_t xfs_default_resblks(xfs_mount_t *mp);
|
2008-08-13 14:49:32 +08:00
|
|
|
extern int xfs_mountfs(xfs_mount_t *mp);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2008-08-13 14:49:57 +08:00
|
|
|
extern void xfs_unmountfs(xfs_mount_t *);
|
2005-04-17 06:20:36 +08:00
|
|
|
extern int xfs_unmountfs_writesb(xfs_mount_t *);
|
2007-02-10 15:36:10 +08:00
|
|
|
extern int xfs_mod_incore_sb(xfs_mount_t *, xfs_sb_field_t, int64_t, int);
|
2005-04-17 06:20:36 +08:00
|
|
|
extern int xfs_mod_incore_sb_batch(xfs_mount_t *, xfs_mod_sb_t *,
|
|
|
|
uint, int);
|
2009-01-19 09:04:07 +08:00
|
|
|
extern int xfs_mount_log_sb(xfs_mount_t *, __int64_t);
|
2005-04-17 06:20:36 +08:00
|
|
|
extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
|
2006-03-31 11:04:17 +08:00
|
|
|
extern int xfs_readsb(xfs_mount_t *, int);
|
2005-04-17 06:20:36 +08:00
|
|
|
extern void xfs_freesb(xfs_mount_t *);
|
[XFS] Lazy Superblock Counters
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
2007-05-24 13:26:31 +08:00
|
|
|
extern int xfs_fs_writable(xfs_mount_t *);
|
2007-05-14 16:24:02 +08:00
|
|
|
extern int xfs_sb_validate_fsb_count(struct xfs_sb *, __uint64_t);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2010-02-18 03:36:13 +08:00
|
|
|
extern int xfs_dev_is_read_only(struct xfs_mount *, char *);
|
|
|
|
|
2011-01-04 08:35:03 +08:00
|
|
|
extern void xfs_set_low_space_thresholds(struct xfs_mount *);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
#endif /* __KERNEL__ */
|
|
|
|
|
2008-10-30 14:05:38 +08:00
|
|
|
extern void xfs_mod_sb(struct xfs_trans *, __int64_t);
|
2010-01-11 19:47:44 +08:00
|
|
|
extern int xfs_initialize_perag(struct xfs_mount *, xfs_agnumber_t,
|
|
|
|
xfs_agnumber_t *);
|
2008-10-30 14:05:38 +08:00
|
|
|
extern void xfs_sb_from_disk(struct xfs_sb *, struct xfs_dsb *);
|
|
|
|
extern void xfs_sb_to_disk(struct xfs_dsb *, struct xfs_sb *, __int64_t);
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
#endif /* __XFS_MOUNT_H__ */
|