xfs_log.c

linux 内核源代码

	tail_lsn = xfs_trans_tail_ail(mp);
	s = GRANT_LOCK(log);
	if (tail_lsn != 0) {
		log->l_tail_lsn = tail_lsn;
	} else {
		tail_lsn = log->l_tail_lsn = log->l_last_sync_lsn;
	}
	GRANT_UNLOCK(log, s);

	return tail_lsn;
}	/* xlog_assign_tail_lsn */


/*
 * Return the space in the log between the tail and the head.  The head
 * is passed in the cycle/bytes formal parms.  In the special case where
 * the reserve head has wrapped passed the tail, this calculation is no
 * longer valid.  In this case, just return 0 which means there is no space
 * in the log.  This works for all places where this function is called
 * with the reserve head.  Of course, if the write head were to ever
 * wrap the tail, we should blow up.  Rather than catch this case here,
 * we depend on other ASSERTions in other parts of the code.   XXXmiken
 *
 * This code also handles the case where the reservation head is behind
 * the tail.  The details of this case are described below, but the end
 * result is that we return the size of the log as the amount of space left.
 */
int
xlog_space_left(xlog_t *log, int cycle, int bytes)
{
	int free_bytes;
	int tail_bytes;
	int tail_cycle;

	tail_bytes = BBTOB(BLOCK_LSN(log->l_tail_lsn));
	tail_cycle = CYCLE_LSN(log->l_tail_lsn);
	if ((tail_cycle == cycle) && (bytes >= tail_bytes)) {
		free_bytes = log->l_logsize - (bytes - tail_bytes);
	} else if ((tail_cycle + 1) < cycle) {
		return 0;
	} else if (tail_cycle < cycle) {
		ASSERT(tail_cycle == (cycle - 1));
		free_bytes = tail_bytes - bytes;
	} else {
		/*
		 * The reservation head is behind the tail.
		 * In this case we just want to return the size of the
		 * log as the amount of space left.
		 */
		xfs_fs_cmn_err(CE_ALERT, log->l_mp,
			"xlog_space_left: head behind tail\n"
			"  tail_cycle = %d, tail_bytes = %d\n"
			"  GH   cycle = %d, GH   bytes = %d",
			tail_cycle, tail_bytes, cycle, bytes);
		ASSERT(0);
		free_bytes = log->l_logsize;
	}
	return free_bytes;
}	/* xlog_space_left */


/*
 * Log function which is called when an io completes.
 *
 * The log manager needs its own routine, in order to control what
 * happens with the buffer after the write completes.
 */
void
xlog_iodone(xfs_buf_t *bp)
{
	xlog_in_core_t	*iclog;
	xlog_t		*l;
	int		aborted;

	iclog = XFS_BUF_FSPRIVATE(bp, xlog_in_core_t *);
	ASSERT(XFS_BUF_FSPRIVATE2(bp, unsigned long) == (unsigned long) 2);
	XFS_BUF_SET_FSPRIVATE2(bp, (unsigned long)1);
	aborted = 0;

	/*
	 * Some versions of cpp barf on the recursive definition of
	 * ic_log -> hic_fields.ic_log and expand ic_log twice when
	 * it is passed through two macros.  Workaround broken cpp.
	 */
	l = iclog->ic_log;

	/*
	 * If the ordered flag has been removed by a lower
	 * layer, it means the underlyin device no longer supports
	 * barrier I/O. Warn loudly and turn off barriers.
	 */
	if ((l->l_mp->m_flags & XFS_MOUNT_BARRIER) && !XFS_BUF_ORDERED(bp)) {
		l->l_mp->m_flags &= ~XFS_MOUNT_BARRIER;
		xfs_fs_cmn_err(CE_WARN, l->l_mp,
				"xlog_iodone: Barriers are no longer supported"
				" by device. Disabling barriers\n");
		xfs_buftrace("XLOG_IODONE BARRIERS OFF", bp);
	}

	/*
	 * Race to shutdown the filesystem if we see an error.
	 */
	if (XFS_TEST_ERROR((XFS_BUF_GETERROR(bp)), l->l_mp,
			XFS_ERRTAG_IODONE_IOERR, XFS_RANDOM_IODONE_IOERR)) {
		xfs_ioerror_alert("xlog_iodone", l->l_mp, bp, XFS_BUF_ADDR(bp));
		XFS_BUF_STALE(bp);
		xfs_force_shutdown(l->l_mp, SHUTDOWN_LOG_IO_ERROR);
		/*
		 * This flag will be propagated to the trans-committed
		 * callback routines to let them know that the log-commit
		 * didn't succeed.
		 */
		aborted = XFS_LI_ABORTED;
	} else if (iclog->ic_state & XLOG_STATE_IOERROR) {
		aborted = XFS_LI_ABORTED;
	}

	/* log I/O is always issued ASYNC */
	ASSERT(XFS_BUF_ISASYNC(bp));
	xlog_state_done_syncing(iclog, aborted);
	/*
	 * do not reference the buffer (bp) here as we could race
	 * with it being freed after writing the unmount record to the
	 * log.
	 */

}	/* xlog_iodone */

/*
 * The bdstrat callback function for log bufs. This gives us a central
 * place to trap bufs in case we get hit by a log I/O error and need to
 * shutdown. Actually, in practice, even when we didn't get a log error,
 * we transition the iclogs to IOERROR state *after* flushing all existing
 * iclogs to disk. This is because we don't want anymore new transactions to be
 * started or completed afterwards.
 */
STATIC int
xlog_bdstrat_cb(struct xfs_buf *bp)
{
	xlog_in_core_t *iclog;

	iclog = XFS_BUF_FSPRIVATE(bp, xlog_in_core_t *);

	if ((iclog->ic_state & XLOG_STATE_IOERROR) == 0) {
	  /* note for irix bstrat will need  struct bdevsw passed
	   * Fix the following macro if the code ever is merged
	   */
	    XFS_bdstrat(bp);
		return 0;
	}

	xfs_buftrace("XLOG__BDSTRAT IOERROR", bp);
	XFS_BUF_ERROR(bp, EIO);
	XFS_BUF_STALE(bp);
	xfs_biodone(bp);
	return XFS_ERROR(EIO);


}

/*
 * Return size of each in-core log record buffer.
 *
 * All machines get 8 x 32KB buffers by default, unless tuned otherwise.
 *
 * If the filesystem blocksize is too large, we may need to choose a
 * larger size since the directory code currently logs entire blocks.
 */

STATIC void
xlog_get_iclog_buffer_size(xfs_mount_t	*mp,
			   xlog_t	*log)
{
	int size;
	int xhdrs;

	if (mp->m_logbufs <= 0)
		log->l_iclog_bufs = XLOG_MAX_ICLOGS;
	else
		log->l_iclog_bufs = mp->m_logbufs;

	/*
	 * Buffer size passed in from mount system call.
	 */
	if (mp->m_logbsize > 0) {
		size = log->l_iclog_size = mp->m_logbsize;
		log->l_iclog_size_log = 0;
		while (size != 1) {
			log->l_iclog_size_log++;
			size >>= 1;
		}

		if (XFS_SB_VERSION_HASLOGV2(&mp->m_sb)) {
			/* # headers = size / 32K
			 * one header holds cycles from 32K of data
			 */

			xhdrs = mp->m_logbsize / XLOG_HEADER_CYCLE_SIZE;
			if (mp->m_logbsize % XLOG_HEADER_CYCLE_SIZE)
				xhdrs++;
			log->l_iclog_hsize = xhdrs << BBSHIFT;
			log->l_iclog_heads = xhdrs;
		} else {
			ASSERT(mp->m_logbsize <= XLOG_BIG_RECORD_BSIZE);
			log->l_iclog_hsize = BBSIZE;
			log->l_iclog_heads = 1;
		}
		goto done;
	}

	/* All machines use 32KB buffers by default. */
	log->l_iclog_size = XLOG_BIG_RECORD_BSIZE;
	log->l_iclog_size_log = XLOG_BIG_RECORD_BSHIFT;

	/* the default log size is 16k or 32k which is one header sector */
	log->l_iclog_hsize = BBSIZE;
	log->l_iclog_heads = 1;

	/*
	 * For 16KB, we use 3 32KB buffers.  For 32KB block sizes, we use
	 * 4 32KB buffers.  For 64KB block sizes, we use 8 32KB buffers.
	 */
	if (mp->m_sb.sb_blocksize >= 16*1024) {
		log->l_iclog_size = XLOG_BIG_RECORD_BSIZE;
		log->l_iclog_size_log = XLOG_BIG_RECORD_BSHIFT;
		if (mp->m_logbufs <= 0) {
			switch (mp->m_sb.sb_blocksize) {
			    case 16*1024:			/* 16 KB */
				log->l_iclog_bufs = 3;
				break;
			    case 32*1024:			/* 32 KB */
				log->l_iclog_bufs = 4;
				break;
			    case 64*1024:			/* 64 KB */
				log->l_iclog_bufs = 8;
				break;
			    default:
				xlog_panic("XFS: Invalid blocksize");
				break;
			}
		}
	}

done:	/* are we being asked to make the sizes selected above visible? */
	if (mp->m_logbufs == 0)
		mp->m_logbufs = log->l_iclog_bufs;
	if (mp->m_logbsize == 0)
		mp->m_logbsize = log->l_iclog_size;
}	/* xlog_get_iclog_buffer_size */


/*
 * This routine initializes some of the log structure for a given mount point.
 * Its primary purpose is to fill in enough, so recovery can occur.  However,
 * some other stuff may be filled in too.
 */
STATIC xlog_t *
xlog_alloc_log(xfs_mount_t	*mp,
	       xfs_buftarg_t	*log_target,
	       xfs_daddr_t	blk_offset,
	       int		num_bblks)
{
	xlog_t			*log;
	xlog_rec_header_t	*head;
	xlog_in_core_t		**iclogp;
	xlog_in_core_t		*iclog, *prev_iclog=NULL;
	xfs_buf_t		*bp;
	int			i;
	int			iclogsize;

	log = (xlog_t *)kmem_zalloc(sizeof(xlog_t), KM_SLEEP);

	log->l_mp	   = mp;
	log->l_targ	   = log_target;
	log->l_logsize     = BBTOB(num_bblks);
	log->l_logBBstart  = blk_offset;
	log->l_logBBsize   = num_bblks;
	log->l_covered_state = XLOG_STATE_COVER_IDLE;
	log->l_flags	   |= XLOG_ACTIVE_RECOVERY;

	log->l_prev_block  = -1;
	ASSIGN_ANY_LSN_HOST(log->l_tail_lsn, 1, 0);
	/* log->l_tail_lsn = 0x100000000LL; cycle = 1; current block = 0 */
	log->l_last_sync_lsn = log->l_tail_lsn;
	log->l_curr_cycle  = 1;	    /* 0 is bad since this is initial value */
	log->l_grant_reserve_cycle = 1;
	log->l_grant_write_cycle = 1;

	if (XFS_SB_VERSION_HASSECTOR(&mp->m_sb)) {
		log->l_sectbb_log = mp->m_sb.sb_logsectlog - BBSHIFT;
		ASSERT(log->l_sectbb_log <= mp->m_sectbb_log);
		/* for larger sector sizes, must have v2 or external log */
		ASSERT(log->l_sectbb_log == 0 ||
			log->l_logBBstart == 0 ||
			XFS_SB_VERSION_HASLOGV2(&mp->m_sb));
		ASSERT(mp->m_sb.sb_logsectlog >= BBSHIFT);
	}
	log->l_sectbb_mask = (1 << log->l_sectbb_log) - 1;

	xlog_get_iclog_buffer_size(mp, log);

	bp = xfs_buf_get_empty(log->l_iclog_size, mp->m_logdev_targp);
	XFS_BUF_SET_IODONE_FUNC(bp, xlog_iodone);
	XFS_BUF_SET_BDSTRAT_FUNC(bp, xlog_bdstrat_cb);
	XFS_BUF_SET_FSPRIVATE2(bp, (unsigned long)1);
	ASSERT(XFS_BUF_ISBUSY(bp));
	ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
	log->l_xbuf = bp;

	spinlock_init(&log->l_icloglock, "iclog");
	spinlock_init(&log->l_grant_lock, "grhead_iclog");
	initnsema(&log->l_flushsema, 0, "ic-flush");
	xlog_state_ticket_alloc(log);  /* wait until after icloglock inited */

	/* log record size must be multiple of BBSIZE; see xlog_rec_header_t */
	ASSERT((XFS_BUF_SIZE(bp) & BBMASK) == 0);

	iclogp = &log->l_iclog;
	/*
	 * The amount of memory to allocate for the iclog structure is
	 * rather funky due to the way the structure is defined.  It is
	 * done this way so that we can use different sizes for machines
	 * with different amounts of memory.  See the definition of
	 * xlog_in_core_t in xfs_log_priv.h for details.
	 */
	iclogsize = log->l_iclog_size;
	ASSERT(log->l_iclog_size >= 4096);
	for (i=0; i < log->l_iclog_bufs; i++) {
		*iclogp = (xlog_in_core_t *)
			  kmem_zalloc(sizeof(xlog_in_core_t), KM_SLEEP);
		iclog = *iclogp;
		iclog->ic_prev = prev_iclog;
		prev_iclog = iclog;

		bp = xfs_buf_get_noaddr(log->l_iclog_size, mp->m_logdev_targp);
		if (!XFS_BUF_CPSEMA(bp))
			ASSERT(0);
		XFS_BUF_SET_IODONE_FUNC(bp, xlog_iodone);
		XFS_BUF_SET_BDSTRAT_FUNC(bp, xlog_bdstrat_cb);
		XFS_BUF_SET_FSPRIVATE2(bp, (unsigned long)1);
		iclog->ic_bp = bp;
		iclog->hic_data = bp->b_addr;

		log->l_iclog_bak[i] = (xfs_caddr_t)&(iclog->ic_header);

		head = &iclog->ic_header;
		memset(head, 0, sizeof(xlog_rec_header_t));
		INT_SET(head->h_magicno, ARCH_CONVERT, XLOG_HEADER_MAGIC_NUM);
		INT_SET(head->h_version, ARCH_CONVERT,
			XFS_SB_VERSION_HASLOGV2(&log->l_mp->m_sb) ? 2 : 1);
		INT_SET(head->h_size, ARCH_CONVERT, log->l_iclog_size);
		/* new fields */
		INT_SET(head->h_fmt, ARCH_CONVERT, XLOG_FMT);
		memcpy(&head->h_fs_uuid, &mp->m_sb.sb_uuid, sizeof(uuid_t));


		iclog->ic_size = XFS_BUF_SIZE(bp) - log->l_iclog_hsize;
		iclog->ic_state = XLOG_STATE_ACTIVE;
		iclog->ic_log = log;
		iclog->ic_callback_tail = &(iclog->ic_callback);
		iclog->ic_datap = (char *)iclog->hic_data + log->l_iclog_hsize;

		ASSERT(XFS_BUF_ISBUSY(iclog->ic_bp));
		ASSERT(XFS_BUF_VALUSEMA(iclog->ic_bp) <= 0);
		sv_init(&iclog->ic_forcesema, SV_DEFAULT, "iclog-force");
		sv_init(&iclog->ic_writesema, SV_DEFAULT, "iclog-write");

		iclogp = &iclog->ic_next;
	}
	*iclogp = log->l_iclog;			/* complete ring */
	log->l_iclog->ic_prev = prev_iclog;	/* re-write 1st prev ptr */

	return log;
}	/* xlog_alloc_log */


/*
 * Write out the commit record of a transaction associated with the given
 * ticket.  Return the lsn of the commit record.
 */
STATIC int
xlog_commit_record(xfs_mount_t  *mp,
		   xlog_ticket_t *ticket,
		   xlog_in_core_t **iclog,
		   xfs_lsn_t	*commitlsnp)
{
	int		error;
	xfs_log_iovec_t	reg[1];

	reg[0].i_addr = NULL;
	reg[0].i_len = 0;
	XLOG_VEC_SET_TYPE(&reg[0], XLOG_REG_TYPE_COMMIT);

	ASSERT_ALWAYS(iclog);
	if ((error = xlog_write(mp, reg, 1, ticket, commitlsnp,
			       iclog, XLOG_COMMIT_TRANS))) {
		xfs_force_shutdown(mp, SHUTDOWN_LOG_IO_ERROR);
	}
	return error;
}	/* xlog_commit_record */


/*
 * Push on the buffer cache code if we ever use more than 75% of the on-disk
 * log space.  This code pushes on the lsn which would supposedly free up
 * the 25% which we want to leave free.  We may need to adopt a policy which
 * pushes on an lsn which is further along in the log once we reach the high
 * water mark.  In this manner, we would be creating a low water mark.
 */
void
xlog_grant_push_ail(xfs_mount_t	*mp,
		    int		need_bytes)
{
    xlog_t	*log = mp->m_log;	/* pointer to the log */
    xfs_lsn_t	tail_lsn;		/* lsn of the log tail */
    xfs_lsn_t	threshold_lsn = 0;	/* lsn we'd like to be at */
    int		free_blocks;		/* free blocks left to write to */
    int		free_bytes;		/* free bytes left to write to */
    int		threshold_block;	/* block in lsn we'd like to be at */
    int		threshold_cycle;	/* lsn cycle we'd like to be at */
    int		free_threshold;
    SPLDECL(s);

    ASSERT(BTOBB(need_bytes) < log->l_logBBsize);

    s = GRANT_LOCK(log);
    free_bytes = xlog_space_left(log,
				 log->l_grant_reserve_cycle,
				 log->l_grant_reserve_bytes);
    tail_lsn = log->l_tail_lsn;
    free_blocks = BTOBBT(free_bytes);

    /*
     * Set the threshold for the minimum number of free blocks in the
     * log to the maximum of what the caller needs, one quarter of the
     * log, and 256 blocks.
     */
    free_threshold = BTOBB(need_bytes);
    free_threshold = MAX(free_threshold, (log->l_logBBsize >> 2));
    free_threshold = MAX(free_threshold, 256);
    if (free_blocks < free_threshold) {
	threshold_block = BLOCK_LSN(tail_lsn) + free_threshold;
	threshold_cycle = CYCLE_LSN(tail_lsn);