xfs_log.c

linux 内核源代码

    int		     partial_copy;   /* did we split a region? */
    int		     partial_copy_len;/* # bytes copied if split region */
    int		     need_copy;	     /* # bytes need to memcpy this region */
    int		     copy_len;	     /* # bytes actually memcpy'ing */
    int		     copy_off;	     /* # bytes from entry start */
    int		     contwr;	     /* continued write of in-core log? */
    int		     error;
    int		     record_cnt = 0, data_cnt = 0;

    partial_copy_len = partial_copy = 0;

    /* Calculate potential maximum space.  Each region gets its own
     * xlog_op_header_t and may need to be double word aligned.
     */
    len = 0;
    if (ticket->t_flags & XLOG_TIC_INITED) {    /* acct for start rec of xact */
	len += sizeof(xlog_op_header_t);
	ticket->t_res_num_ophdrs++;
    }

    for (index = 0; index < nentries; index++) {
	len += sizeof(xlog_op_header_t);	    /* each region gets >= 1 */
	ticket->t_res_num_ophdrs++;
	len += reg[index].i_len;
	xlog_tic_add_region(ticket, reg[index].i_len, reg[index].i_type);
    }
    contwr = *start_lsn = 0;

    if (ticket->t_curr_res < len) {
	xlog_print_tic_res(mp, ticket);
#ifdef DEBUG
	xlog_panic(
		"xfs_log_write: reservation ran out. Need to up reservation");
#else
	/* Customer configurable panic */
	xfs_cmn_err(XFS_PTAG_LOGRES, CE_ALERT, mp,
		"xfs_log_write: reservation ran out. Need to up reservation");
	/* If we did not panic, shutdown the filesystem */
	xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
#endif
    } else
	ticket->t_curr_res -= len;

    for (index = 0; index < nentries; ) {
	if ((error = xlog_state_get_iclog_space(log, len, &iclog, ticket,
					       &contwr, &log_offset)))
		return error;

	ASSERT(log_offset <= iclog->ic_size - 1);
	ptr = (__psint_t) ((char *)iclog->ic_datap+log_offset);

	/* start_lsn is the first lsn written to. That's all we need. */
	if (! *start_lsn)
	    *start_lsn = INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT);

	/* This loop writes out as many regions as can fit in the amount
	 * of space which was allocated by xlog_state_get_iclog_space().
	 */
	while (index < nentries) {
	    ASSERT(reg[index].i_len % sizeof(__int32_t) == 0);
	    ASSERT((__psint_t)ptr % sizeof(__int32_t) == 0);
	    start_rec_copy = 0;

	    /* If first write for transaction, insert start record.
	     * We can't be trying to commit if we are inited.  We can't
	     * have any "partial_copy" if we are inited.
	     */
	    if (ticket->t_flags & XLOG_TIC_INITED) {
		logop_head		= (xlog_op_header_t *)ptr;
		INT_SET(logop_head->oh_tid, ARCH_CONVERT, ticket->t_tid);
		logop_head->oh_clientid = ticket->t_clientid;
		logop_head->oh_len	= 0;
		logop_head->oh_flags    = XLOG_START_TRANS;
		logop_head->oh_res2	= 0;
		ticket->t_flags		&= ~XLOG_TIC_INITED;	/* clear bit */
		record_cnt++;

		start_rec_copy = sizeof(xlog_op_header_t);
		xlog_write_adv_cnt(ptr, len, log_offset, start_rec_copy);
	    }

	    /* Copy log operation header directly into data section */
	    logop_head			= (xlog_op_header_t *)ptr;
	    INT_SET(logop_head->oh_tid, ARCH_CONVERT, ticket->t_tid);
	    logop_head->oh_clientid	= ticket->t_clientid;
	    logop_head->oh_res2		= 0;

	    /* header copied directly */
	    xlog_write_adv_cnt(ptr, len, log_offset, sizeof(xlog_op_header_t));

	    /* are we copying a commit or unmount record? */
	    logop_head->oh_flags = flags;

	    /*
	     * We've seen logs corrupted with bad transaction client
	     * ids.  This makes sure that XFS doesn't generate them on.
	     * Turn this into an EIO and shut down the filesystem.
	     */
	    switch (logop_head->oh_clientid)  {
	    case XFS_TRANSACTION:
	    case XFS_VOLUME:
	    case XFS_LOG:
		break;
	    default:
		xfs_fs_cmn_err(CE_WARN, mp,
		    "Bad XFS transaction clientid 0x%x in ticket 0x%p",
		    logop_head->oh_clientid, tic);
		return XFS_ERROR(EIO);
	    }

	    /* Partial write last time? => (partial_copy != 0)
	     * need_copy is the amount we'd like to copy if everything could
	     * fit in the current memcpy.
	     */
	    need_copy =	reg[index].i_len - partial_copy_len;

	    copy_off = partial_copy_len;
	    if (need_copy <= iclog->ic_size - log_offset) { /*complete write */
		INT_SET(logop_head->oh_len, ARCH_CONVERT, copy_len = need_copy);
		if (partial_copy)
		    logop_head->oh_flags|= (XLOG_END_TRANS|XLOG_WAS_CONT_TRANS);
		partial_copy_len = partial_copy = 0;
	    } else {					    /* partial write */
		copy_len = iclog->ic_size - log_offset;
		INT_SET(logop_head->oh_len, ARCH_CONVERT, copy_len);
		logop_head->oh_flags |= XLOG_CONTINUE_TRANS;
		if (partial_copy)
			logop_head->oh_flags |= XLOG_WAS_CONT_TRANS;
		partial_copy_len += copy_len;
		partial_copy++;
		len += sizeof(xlog_op_header_t); /* from splitting of region */
		/* account for new log op header */
		ticket->t_curr_res -= sizeof(xlog_op_header_t);
		ticket->t_res_num_ophdrs++;
	    }
	    xlog_verify_dest_ptr(log, ptr);

	    /* copy region */
	    ASSERT(copy_len >= 0);
	    memcpy((xfs_caddr_t)ptr, reg[index].i_addr + copy_off, copy_len);
	    xlog_write_adv_cnt(ptr, len, log_offset, copy_len);

	    /* make copy_len total bytes copied, including headers */
	    copy_len += start_rec_copy + sizeof(xlog_op_header_t);
	    record_cnt++;
	    data_cnt += contwr ? copy_len : 0;
	    if (partial_copy) {			/* copied partial region */
		    /* already marked WANT_SYNC by xlog_state_get_iclog_space */
		    xlog_state_finish_copy(log, iclog, record_cnt, data_cnt);
		    record_cnt = data_cnt = 0;
		    if ((error = xlog_state_release_iclog(log, iclog)))
			    return error;
		    break;			/* don't increment index */
	    } else {				/* copied entire region */
		index++;
		partial_copy_len = partial_copy = 0;

		if (iclog->ic_size - log_offset <= sizeof(xlog_op_header_t)) {
		    xlog_state_finish_copy(log, iclog, record_cnt, data_cnt);
		    record_cnt = data_cnt = 0;
		    xlog_state_want_sync(log, iclog);
		    if (commit_iclog) {
			ASSERT(flags & XLOG_COMMIT_TRANS);
			*commit_iclog = iclog;
		    } else if ((error = xlog_state_release_iclog(log, iclog)))
			   return error;
		    if (index == nentries)
			    return 0;		/* we are done */
		    else
			    break;
		}
	    } /* if (partial_copy) */
	} /* while (index < nentries) */
    } /* for (index = 0; index < nentries; ) */
    ASSERT(len == 0);

    xlog_state_finish_copy(log, iclog, record_cnt, data_cnt);
    if (commit_iclog) {
	ASSERT(flags & XLOG_COMMIT_TRANS);
	*commit_iclog = iclog;
	return 0;
    }
    return xlog_state_release_iclog(log, iclog);
}	/* xlog_write */


/*****************************************************************************
 *
 *		State Machine functions
 *
 *****************************************************************************
 */

/* Clean iclogs starting from the head.  This ordering must be
 * maintained, so an iclog doesn't become ACTIVE beyond one that
 * is SYNCING.  This is also required to maintain the notion that we use
 * a counting semaphore to hold off would be writers to the log when every
 * iclog is trying to sync to disk.
 *
 * State Change: DIRTY -> ACTIVE
 */
STATIC void
xlog_state_clean_log(xlog_t *log)
{
	xlog_in_core_t	*iclog;
	int changed = 0;

	iclog = log->l_iclog;
	do {
		if (iclog->ic_state == XLOG_STATE_DIRTY) {
			iclog->ic_state	= XLOG_STATE_ACTIVE;
			iclog->ic_offset       = 0;
			iclog->ic_callback	= NULL;   /* don't need to free */
			/*
			 * If the number of ops in this iclog indicate it just
			 * contains the dummy transaction, we can
			 * change state into IDLE (the second time around).
			 * Otherwise we should change the state into
			 * NEED a dummy.
			 * We don't need to cover the dummy.
			 */
			if (!changed &&
			   (INT_GET(iclog->ic_header.h_num_logops, ARCH_CONVERT) == XLOG_COVER_OPS)) {
				changed = 1;
			} else {
				/*
				 * We have two dirty iclogs so start over
				 * This could also be num of ops indicates
				 * this is not the dummy going out.
				 */
				changed = 2;
			}
			iclog->ic_header.h_num_logops = 0;
			memset(iclog->ic_header.h_cycle_data, 0,
			      sizeof(iclog->ic_header.h_cycle_data));
			iclog->ic_header.h_lsn = 0;
		} else if (iclog->ic_state == XLOG_STATE_ACTIVE)
			/* do nothing */;
		else
			break;	/* stop cleaning */
		iclog = iclog->ic_next;
	} while (iclog != log->l_iclog);

	/* log is locked when we are called */
	/*
	 * Change state for the dummy log recording.
	 * We usually go to NEED. But we go to NEED2 if the changed indicates
	 * we are done writing the dummy record.
	 * If we are done with the second dummy recored (DONE2), then
	 * we go to IDLE.
	 */
	if (changed) {
		switch (log->l_covered_state) {
		case XLOG_STATE_COVER_IDLE:
		case XLOG_STATE_COVER_NEED:
		case XLOG_STATE_COVER_NEED2:
			log->l_covered_state = XLOG_STATE_COVER_NEED;
			break;

		case XLOG_STATE_COVER_DONE:
			if (changed == 1)
				log->l_covered_state = XLOG_STATE_COVER_NEED2;
			else
				log->l_covered_state = XLOG_STATE_COVER_NEED;
			break;

		case XLOG_STATE_COVER_DONE2:
			if (changed == 1)
				log->l_covered_state = XLOG_STATE_COVER_IDLE;
			else
				log->l_covered_state = XLOG_STATE_COVER_NEED;
			break;

		default:
			ASSERT(0);
		}
	}
}	/* xlog_state_clean_log */

STATIC xfs_lsn_t
xlog_get_lowest_lsn(
	xlog_t		*log)
{
	xlog_in_core_t  *lsn_log;
	xfs_lsn_t	lowest_lsn, lsn;

	lsn_log = log->l_iclog;
	lowest_lsn = 0;
	do {
	    if (!(lsn_log->ic_state & (XLOG_STATE_ACTIVE|XLOG_STATE_DIRTY))) {
		lsn = INT_GET(lsn_log->ic_header.h_lsn, ARCH_CONVERT);
		if ((lsn && !lowest_lsn) ||
		    (XFS_LSN_CMP(lsn, lowest_lsn) < 0)) {
			lowest_lsn = lsn;
		}
	    }
	    lsn_log = lsn_log->ic_next;
	} while (lsn_log != log->l_iclog);
	return lowest_lsn;
}


STATIC void
xlog_state_do_callback(
	xlog_t		*log,
	int		aborted,
	xlog_in_core_t	*ciclog)
{
	xlog_in_core_t	   *iclog;
	xlog_in_core_t	   *first_iclog;	/* used to know when we've
						 * processed all iclogs once */
	xfs_log_callback_t *cb, *cb_next;
	int		   flushcnt = 0;
	xfs_lsn_t	   lowest_lsn;
	int		   ioerrors;	/* counter: iclogs with errors */
	int		   loopdidcallbacks; /* flag: inner loop did callbacks*/
	int		   funcdidcallbacks; /* flag: function did callbacks */
	int		   repeats;	/* for issuing console warnings if
					 * looping too many times */
	SPLDECL(s);

	s = LOG_LOCK(log);
	first_iclog = iclog = log->l_iclog;
	ioerrors = 0;
	funcdidcallbacks = 0;
	repeats = 0;

	do {
		/*
		 * Scan all iclogs starting with the one pointed to by the
		 * log.  Reset this starting point each time the log is
		 * unlocked (during callbacks).
		 *
		 * Keep looping through iclogs until one full pass is made
		 * without running any callbacks.
		 */
		first_iclog = log->l_iclog;
		iclog = log->l_iclog;
		loopdidcallbacks = 0;
		repeats++;

		do {

			/* skip all iclogs in the ACTIVE & DIRTY states */
			if (iclog->ic_state &
			    (XLOG_STATE_ACTIVE|XLOG_STATE_DIRTY)) {
				iclog = iclog->ic_next;
				continue;
			}

			/*
			 * Between marking a filesystem SHUTDOWN and stopping
			 * the log, we do flush all iclogs to disk (if there
			 * wasn't a log I/O error). So, we do want things to
			 * go smoothly in case of just a SHUTDOWN  w/o a
			 * LOG_IO_ERROR.
			 */
			if (!(iclog->ic_state & XLOG_STATE_IOERROR)) {
				/*
				 * Can only perform callbacks in order.  Since
				 * this iclog is not in the DONE_SYNC/
				 * DO_CALLBACK state, we skip the rest and
				 * just try to clean up.  If we set our iclog
				 * to DO_CALLBACK, we will not process it when
				 * we retry since a previous iclog is in the
				 * CALLBACK and the state cannot change since
				 * we are holding the LOG_LOCK.
				 */
				if (!(iclog->ic_state &
					(XLOG_STATE_DONE_SYNC |
						 XLOG_STATE_DO_CALLBACK))) {
					if (ciclog && (ciclog->ic_state ==
							XLOG_STATE_DONE_SYNC)) {
						ciclog->ic_state = XLOG_STATE_DO_CALLBACK;
					}
					break;
				}
				/*
				 * We now have an iclog that is in either the
				 * DO_CALLBACK or DONE_SYNC states. The other
				 * states (WANT_SYNC, SYNCING, or CALLBACK were
				 * caught by the above if and are going to
				 * clean (i.e. we aren't doing their callbacks)
				 * see the above if.
				 */

				/*
				 * We will do one more check here to see if we
				 * have chased our tail around.
				 */

				lowest_lsn = xlog_get_lowest_lsn(log);
				if (lowest_lsn && (
					XFS_LSN_CMP(
						lowest_lsn,
						INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT)
					)<0)) {
					iclog = iclog->ic_next;
					continue; /* Leave this iclog for
						   * another thread */
				}

				iclog->ic_state = XLOG_STATE_CALLBACK;

				LOG_UNLOCK(log, s);

				/* l_last_sync_lsn field protected by
				 * GRANT_LOCK. Don't worry about iclog's lsn.
				 * No one else can be here except us.
				 */
				s = GRANT_LOCK(log);
				ASSERT(XFS_LSN_CMP(
						log->l_last_sync_lsn,
						INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT)
					)<=0);
				log->l_last_sync_lsn = INT_GET(iclog->ic_header.h_lsn, ARCH_CONVERT);
				GRANT_UNLOCK(log, s);

				/*
				 * Keep processing entries in the callback list
				 * until we come around and it is empty.  We
				 * need to atomically see that the list is
				 * empty and change the state to DIRTY so that
				 * we don't miss any more callbacks being added.
				 */
				s = LOG_LOCK(log);
			} else {
				ioerrors++;
			}
			cb = iclog->ic_callback;

			while (cb) {
				iclog->ic_callback_tail = &(iclog->ic_callback);
				iclog->ic_callback = NULL;
				LOG_UNLOCK(log, s);

				/* perform callbacks in the order given */
				for (; cb; cb = cb_next) {
					cb_next = cb->cb_next;
					cb->cb_func(cb->cb_arg, aborte