xfs_log_priv.h

linux 内核源代码

/*
 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
 * All Rights Reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */
#ifndef	__XFS_LOG_PRIV_H__
#define __XFS_LOG_PRIV_H__

struct xfs_buf;
struct ktrace;
struct log;
struct xlog_ticket;
struct xfs_buf_cancel;
struct xfs_mount;

/*
 * Macros, structures, prototypes for internal log manager use.
 */

#define XLOG_MIN_ICLOGS		2
#define XLOG_MAX_ICLOGS		8
#define XLOG_HEADER_MAGIC_NUM	0xFEEDbabe	/* Invalid cycle number */
#define XLOG_VERSION_1		1
#define XLOG_VERSION_2		2		/* Large IClogs, Log sunit */
#define XLOG_VERSION_OKBITS	(XLOG_VERSION_1 | XLOG_VERSION_2)
#define XLOG_MIN_RECORD_BSIZE	(16*1024)	/* eventually 32k */
#define XLOG_BIG_RECORD_BSIZE	(32*1024)	/* 32k buffers */
#define XLOG_MAX_RECORD_BSIZE	(256*1024)
#define XLOG_HEADER_CYCLE_SIZE	(32*1024)	/* cycle data in header */
#define XLOG_MIN_RECORD_BSHIFT	14		/* 16384 == 1 << 14 */
#define XLOG_BIG_RECORD_BSHIFT	15		/* 32k == 1 << 15 */
#define XLOG_MAX_RECORD_BSHIFT	18		/* 256k == 1 << 18 */
#define XLOG_BTOLSUNIT(log, b)  (((b)+(log)->l_mp->m_sb.sb_logsunit-1) / \
                                 (log)->l_mp->m_sb.sb_logsunit)
#define XLOG_LSUNITTOB(log, su) ((su) * (log)->l_mp->m_sb.sb_logsunit)

#define XLOG_HEADER_SIZE	512

#define XLOG_REC_SHIFT(log) \
	BTOBB(1 << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \
	 XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))
#define XLOG_TOTAL_REC_SHIFT(log) \
	BTOBB(XLOG_MAX_ICLOGS << (XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? \
	 XLOG_MAX_RECORD_BSHIFT : XLOG_BIG_RECORD_BSHIFT))

/*
 *  set lsns
 */

#define ASSIGN_ANY_LSN_HOST(lsn,cycle,block)  \
    { \
	(lsn) = ((xfs_lsn_t)(cycle)<<32)|(block); \
    }
#define ASSIGN_ANY_LSN_DISK(lsn,cycle,block)  \
    { \
	INT_SET(((uint *)&(lsn))[0], ARCH_CONVERT, (cycle)); \
	INT_SET(((uint *)&(lsn))[1], ARCH_CONVERT, (block)); \
    }
#define ASSIGN_LSN(lsn,log) \
    ASSIGN_ANY_LSN_DISK(lsn,(log)->l_curr_cycle,(log)->l_curr_block);

#define XLOG_SET(f,b)		(((f) & (b)) == (b))

#define GET_CYCLE(ptr, arch) \
    (INT_GET(*(uint *)(ptr), arch) == XLOG_HEADER_MAGIC_NUM ? \
	 INT_GET(*((uint *)(ptr)+1), arch) : \
	 INT_GET(*(uint *)(ptr), arch) \
    )

#define BLK_AVG(blk1, blk2)	((blk1+blk2) >> 1)


#ifdef __KERNEL__

/*
 * get client id from packed copy.
 *
 * this hack is here because the xlog_pack code copies four bytes
 * of xlog_op_header containing the fields oh_clientid, oh_flags
 * and oh_res2 into the packed copy.
 *
 * later on this four byte chunk is treated as an int and the
 * client id is pulled out.
 *
 * this has endian issues, of course.
 */

#ifndef XFS_NATIVE_HOST
#define GET_CLIENT_ID(i,arch) \
    ((i) & 0xff)
#else
#define GET_CLIENT_ID(i,arch) \
    ((i) >> 24)
#endif

#define GRANT_LOCK(log)		mutex_spinlock(&(log)->l_grant_lock)
#define GRANT_UNLOCK(log, s)	mutex_spinunlock(&(log)->l_grant_lock, s)
#define LOG_LOCK(log)		mutex_spinlock(&(log)->l_icloglock)
#define LOG_UNLOCK(log, s)	mutex_spinunlock(&(log)->l_icloglock, s)

#define xlog_panic(args...)	cmn_err(CE_PANIC, ## args)
#define xlog_exit(args...)	cmn_err(CE_PANIC, ## args)
#define xlog_warn(args...)	cmn_err(CE_WARN, ## args)

/*
 * In core log state
 */
#define XLOG_STATE_ACTIVE    0x0001 /* Current IC log being written to */
#define XLOG_STATE_WANT_SYNC 0x0002 /* Want to sync this iclog; no more writes */
#define XLOG_STATE_SYNCING   0x0004 /* This IC log is syncing */
#define XLOG_STATE_DONE_SYNC 0x0008 /* Done syncing to disk */
#define XLOG_STATE_DO_CALLBACK \
			     0x0010 /* Process callback functions */
#define XLOG_STATE_CALLBACK  0x0020 /* Callback functions now */
#define XLOG_STATE_DIRTY     0x0040 /* Dirty IC log, not ready for ACTIVE status*/
#define XLOG_STATE_IOERROR   0x0080 /* IO error happened in sync'ing log */
#define XLOG_STATE_ALL	     0x7FFF /* All possible valid flags */
#define XLOG_STATE_NOTUSED   0x8000 /* This IC log not being used */
#endif	/* __KERNEL__ */

/*
 * Flags to log operation header
 *
 * The first write of a new transaction will be preceded with a start
 * record, XLOG_START_TRANS.  Once a transaction is committed, a commit
 * record is written, XLOG_COMMIT_TRANS.  If a single region can not fit into
 * the remainder of the current active in-core log, it is split up into
 * multiple regions.  Each partial region will be marked with a
 * XLOG_CONTINUE_TRANS until the last one, which gets marked with XLOG_END_TRANS.
 *
 */
#define XLOG_START_TRANS	0x01	/* Start a new transaction */
#define XLOG_COMMIT_TRANS	0x02	/* Commit this transaction */
#define XLOG_CONTINUE_TRANS	0x04	/* Cont this trans into new region */
#define XLOG_WAS_CONT_TRANS	0x08	/* Cont this trans into new region */
#define XLOG_END_TRANS		0x10	/* End a continued transaction */
#define XLOG_UNMOUNT_TRANS	0x20	/* Unmount a filesystem transaction */

#ifdef __KERNEL__
/*
 * Flags to log ticket
 */
#define XLOG_TIC_INITED		0x1	/* has been initialized */
#define XLOG_TIC_PERM_RESERV	0x2	/* permanent reservation */
#define XLOG_TIC_IN_Q		0x4
#endif	/* __KERNEL__ */

#define XLOG_UNMOUNT_TYPE	0x556e	/* Un for Unmount */

/*
 * Flags for log structure
 */
#define XLOG_CHKSUM_MISMATCH	0x1	/* used only during recovery */
#define XLOG_ACTIVE_RECOVERY	0x2	/* in the middle of recovery */
#define	XLOG_RECOVERY_NEEDED	0x4	/* log was recovered */
#define XLOG_IO_ERROR		0x8	/* log hit an I/O error, and being
					   shutdown */
typedef __uint32_t xlog_tid_t;


#ifdef __KERNEL__
/*
 * Below are states for covering allocation transactions.
 * By covering, we mean changing the h_tail_lsn in the last on-disk
 * log write such that no allocation transactions will be re-done during
 * recovery after a system crash. Recovery starts at the last on-disk
 * log write.
 *
 * These states are used to insert dummy log entries to cover
 * space allocation transactions which can undo non-transactional changes
 * after a crash. Writes to a file with space
 * already allocated do not result in any transactions. Allocations
 * might include space beyond the EOF. So if we just push the EOF a
 * little, the last transaction for the file could contain the wrong
 * size. If there is no file system activity, after an allocation
 * transaction, and the system crashes, the allocation transaction
 * will get replayed and the file will be truncated. This could
 * be hours/days/... after the allocation occurred.
 *
 * The fix for this is to do two dummy transactions when the
 * system is idle. We need two dummy transaction because the h_tail_lsn
 * in the log record header needs to point beyond the last possible
 * non-dummy transaction. The first dummy changes the h_tail_lsn to
 * the first transaction before the dummy. The second dummy causes
 * h_tail_lsn to point to the first dummy. Recovery starts at h_tail_lsn.
 *
 * These dummy transactions get committed when everything
 * is idle (after there has been some activity).
 *
 * There are 5 states used to control this.
 *
 *  IDLE -- no logging has been done on the file system or
 *		we are done covering previous transactions.
 *  NEED -- logging has occurred and we need a dummy transaction
 *		when the log becomes idle.
 *  DONE -- we were in the NEED state and have committed a dummy
 *		transaction.
 *  NEED2 -- we detected that a dummy transaction has gone to the
 *		on disk log with no other transactions.
 *  DONE2 -- we committed a dummy transaction when in the NEED2 state.
 *
 * There are two places where we switch states:
 *
 * 1.) In xfs_sync, when we detect an idle log and are in NEED or NEED2.
 *	We commit the dummy transaction and switch to DONE or DONE2,
 *	respectively. In all other states, we don't do anything.
 *
 * 2.) When we finish writing the on-disk log (xlog_state_clean_log).
 *
 *	No matter what state we are in, if this isn't the dummy
 *	transaction going out, the next state is NEED.
 *	So, if we aren't in the DONE or DONE2 states, the next state
 *	is NEED. We can't be finishing a write of the dummy record
 *	unless it was committed and the state switched to DONE or DONE2.
 *
 *	If we are in the DONE state and this was a write of the
 *		dummy transaction, we move to NEED2.
 *
 *	If we are in the DONE2 state and this was a write of the
 *		dummy transaction, we move to IDLE.
 *
 *
 * Writing only one dummy transaction can get appended to
 * one file space allocation. When this happens, the log recovery
 * code replays the space allocation and a file could be truncated.
 * This is why we have the NEED2 and DONE2 states before going idle.
 */

#define XLOG_STATE_COVER_IDLE	0
#define XLOG_STATE_COVER_NEED	1
#define XLOG_STATE_COVER_DONE	2
#define XLOG_STATE_COVER_NEED2	3
#define XLOG_STATE_COVER_DONE2	4

#define XLOG_COVER_OPS		5