/*
 * Copyright (c) 1982, 1986, 1988, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * %sccs.include.redist.c%
 *
 *	@(#)ufs_disksubr.c	8.1 (Berkeley) 06/11/93
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/buf.h>
#include <sys/disklabel.h>
#include <sys/syslog.h>

/*
 * Seek sort for disks.  We depend on the driver which calls us using b_resid
 * as the current cylinder number.
 *
 * The argument dp structure holds a b_actf activity chain pointer on which we
 * keep two queues, sorted in ascending cylinder order.  The first queue holds
 * those requests which are positioned after the current cylinder (in the first
 * request); the second holds requests which came in after their cylinder number
 * was passed.  Thus we implement a one way scan, retracting after reaching the
 * end of the drive to the first request on the second queue, at which time it
 * becomes the first queue.
 *
 * A one-way scan is natural because of the way UNIX read-ahead blocks are
 * allocated.
 */

#define	b_cylin	b_resid

void
disksort(dp, bp)
	register struct buf *dp, *bp;
{
	register struct buf *ap;

	/*
	 * If nothing on the activity queue, then
	 * we become the only thing.
	 */
	ap = dp->b_actf;
	if(ap == NULL) {
		dp->b_actf = bp;
		bp->b_actf = NULL;
		return;
	}
	/*
	 * If we lie after the first (currently active)
	 * request, then we must locate the second request list
	 * and add ourselves to it.
	 */
	if (bp->b_cylin < ap->b_cylin) {
		while (ap->b_actf) {
			/*
			 * Check for an ``inversion'' in the
			 * normally ascending cylinder numbers,
			 * indicating the start of the second request list.
			 */
			if (ap->b_actf->b_cylin < ap->b_cylin) {
				/*
				 * Search the second request list
				 * for the first request at a larger
				 * cylinder number.  We go before that;
				 * if there is no such request, we go at end.
				 */
				do {
					if (bp->b_cylin < ap->b_actf->b_cylin)
						goto insert;
					if (bp->b_cylin == ap->b_actf->b_cylin &&
					    bp->b_blkno < ap->b_actf->b_blkno)
						goto insert;
					ap = ap->b_actf;
				} while (ap->b_actf);
				goto insert;		/* after last */
			}
			ap = ap->b_actf;
		}
		/*
		 * No inversions... we will go after the last, and
		 * be the first request in the second request list.
		 */
		goto insert;
	}
	/*
	 * Request is at/after the current request...
	 * sort in the first request list.
	 */
	while (ap->b_actf) {
		/*
		 * We want to go after the current request
		 * if there is an inversion after it (i.e. it is
		 * the end of the first request list), or if
		 * the next request is a larger cylinder than our request.
		 */
		if (ap->b_actf->b_cylin < ap->b_cylin ||
		    bp->b_cylin < ap->b_actf->b_cylin ||
		    (bp->b_cylin == ap->b_actf->b_cylin &&
		    bp->b_blkno < ap->b_actf->b_blkno))
			goto insert;
		ap = ap->b_actf;
	}
	/*
	 * Neither a second list nor a larger
	 * request... we go at the end of the first list,
	 * which is the same as the end of the whole schebang.
	 */
insert:
	bp->b_actf = ap->b_actf;
	ap->b_actf = bp;
}

/*
 * Attempt to read a disk label from a device using the indicated stategy
 * routine.  The label must be partly set up before this: secpercyl and
 * anything required in the strategy routine (e.g., sector size) must be
 * filled in before calling us.  Returns NULL on success and an error
 * string on failure.
 */
char *
readdisklabel(dev, strat, lp)
	dev_t dev;
	int (*strat)();
	register struct disklabel *lp;
{
	register struct buf *bp;
	struct disklabel *dlp;
	char *msg = NULL;

	if (lp->d_secperunit == 0)
		lp->d_secperunit = 0x1fffffff;
	lp->d_npartitions = 1;
	if (lp->d_partitions[0].p_size == 0)
		lp->d_partitions[0].p_size = 0x1fffffff;
	lp->d_partitions[0].p_offset = 0;

	bp = geteblk((int)lp->d_secsize);
	bp->b_dev = dev;
	bp->b_blkno = LABELSECTOR;
	bp->b_bcount = lp->d_secsize;
	bp->b_flags = B_BUSY | B_READ;
	bp->b_cylin = LABELSECTOR / lp->d_secpercyl;
	(*strat)(bp);
	if (biowait(bp)) {
		msg = "I/O error";
	} else for (dlp = (struct disklabel *)bp->b_un.b_addr;
	    dlp <= (struct disklabel *)(bp->b_un.b_addr+DEV_BSIZE-sizeof(*dlp));
	    dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
		if (dlp->d_magic != DISKMAGIC || dlp->d_magic2 != DISKMAGIC) {
			if (msg == NULL)
				msg = "no disk label";
		} else if (dlp->d_npartitions > MAXPARTITIONS ||
			   dkcksum(dlp) != 0)
			msg = "disk label corrupted";
		else {
			*lp = *dlp;
			msg = NULL;
			break;
		}
	}
	bp->b_flags = B_INVAL | B_AGE;
	brelse(bp);
	return (msg);
}

/*
 * Check new disk label for sensibility before setting it.
 */
int
setdisklabel(olp, nlp, openmask)
	register struct disklabel *olp, *nlp;
	u_long openmask;
{
	register i;
	register struct partition *opp, *npp;

	if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
	    dkcksum(nlp) != 0)
		return (EINVAL);
	while ((i = ffs((long)openmask)) != 0) {
		i--;
		openmask &= ~(1 << i);
		if (nlp->d_npartitions <= i)
			return (EBUSY);
		opp = &olp->d_partitions[i];
		npp = &nlp->d_partitions[i];
		if (npp->p_offset != opp->p_offset || npp->p_size < opp->p_size)
			return (EBUSY);
		/*
		 * Copy internally-set partition information
		 * if new label doesn't include it.		XXX
		 */
		if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
			npp->p_fstype = opp->p_fstype;
			npp->p_fsize = opp->p_fsize;
			npp->p_frag = opp->p_frag;
			npp->p_cpg = opp->p_cpg;
		}
	}
 	nlp->d_checksum = 0;
 	nlp->d_checksum = dkcksum(nlp);
	*olp = *nlp;
	return (0);
}

/* encoding of disk minor numbers, should be elsewhere... */
#define dkunit(dev)		(minor(dev) >> 3)
#define dkpart(dev)		(minor(dev) & 07)
#define dkminor(unit, part)	(((unit) << 3) | (part))

/*
 * Write disk label back to device after modification.
 */
int
writedisklabel(dev, strat, lp)
	dev_t dev;
	int (*strat)();
	register struct disklabel *lp;
{
	struct buf *bp;
	struct disklabel *dlp;
	int labelpart;
	int error = 0;

	labelpart = dkpart(dev);
	if (lp->d_partitions[labelpart].p_offset != 0) {
		if (lp->d_partitions[0].p_offset != 0)
			return (EXDEV);			/* not quite right */
		labelpart = 0;
	}
	bp = geteblk((int)lp->d_secsize);
	bp->b_dev = makedev(major(dev), dkminor(dkunit(dev), labelpart));
	bp->b_blkno = LABELSECTOR;
	bp->b_bcount = lp->d_secsize;
	bp->b_flags = B_READ;
	(*strat)(bp);
	if (error = biowait(bp))
		goto done;
	for (dlp = (struct disklabel *)bp->b_un.b_addr;
	    dlp <= (struct disklabel *)
	      (bp->b_un.b_addr + lp->d_secsize - sizeof(*dlp));
	    dlp = (struct disklabel *)((char *)dlp + sizeof(long))) {
		if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
		    dkcksum(dlp) == 0) {
			*dlp = *lp;
			bp->b_flags = B_WRITE;
			(*strat)(bp);
			error = biowait(bp);
			goto done;
		}
	}
	error = ESRCH;
done:
	brelse(bp);
	return (error);
}

/*
 * Compute checksum for disk label.
 */
dkcksum(lp)
	register struct disklabel *lp;
{
	register u_short *start, *end;
	register u_short sum = 0;

	start = (u_short *)lp;
	end = (u_short *)&lp->d_partitions[lp->d_npartitions];
	while (start < end)
		sum ^= *start++;
	return (sum);
}

/*
 * Disk error is the preface to plaintive error messages
 * about failing disk transfers.  It prints messages of the form

hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)

 * if the offset of the error in the transfer and a disk label
 * are both available.  blkdone should be -1 if the position of the error
 * is unknown; the disklabel pointer may be null from drivers that have not
 * been converted to use them.  The message is printed with printf
 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
 * The message should be completed (with at least a newline) with printf
 * or addlog, respectively.  There is no trailing space.
 */
void
diskerr(bp, dname, what, pri, blkdone, lp)
	register struct buf *bp;
	char *dname, *what;
	int pri, blkdone;
	register struct disklabel *lp;
{
	int unit = dkunit(bp->b_dev), part = dkpart(bp->b_dev);
	register void (*pr) __P((const char *, ...));
	char partname = 'a' + part;
	int sn;

	if (pri != LOG_PRINTF) {
		log(pri, "");
		pr = addlog;
	} else
		pr = printf;
	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
	    bp->b_flags & B_READ ? "read" : "writ");
	sn = bp->b_blkno;
	if (bp->b_bcount <= DEV_BSIZE)
		(*pr)("%d", sn);
	else {
		if (blkdone >= 0) {
			sn += blkdone;
			(*pr)("%d of ", sn);
		}
		(*pr)("%d-%d", bp->b_blkno,
		    bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
	}
	if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) {
#ifdef tahoe
		sn *= DEV_BSIZE / lp->d_secsize;		/* XXX */
#endif
		sn += lp->d_partitions[part].p_offset;
		(*pr)(" (%s%d bn %d; cn %d", dname, unit, sn,
		    sn / lp->d_secpercyl);
		sn %= lp->d_secpercyl;
		(*pr)(" tn %d sn %d)", sn / lp->d_nsectors, sn % lp->d_nsectors);
	}
}