[unix-history] / usr / src / sys / ufs / ffs / ufs_disksubr.c

/*
 * Copyright (c) 1982, 1986, 1988 Regents of the University of California.
 * All rights reserved.
 *
 * %sccs.include.redist.c%
 *
 *	@(#)ufs_disksubr.c	7.14 (Berkeley) %G%
 */

#include "param.h"
#include "systm.h"
#include "buf.h"
#include "disklabel.h"
#include "syslog.h"
#include "user.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

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;
		dp->b_actl = bp;
		bp->av_forw = 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->av_forw) {
			/*
			 * Check for an ``inversion'' in the
			 * normally ascending cylinder numbers,
			 * indicating the start of the second request list.
			 */
			if (ap->av_forw->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->av_forw->b_cylin)
						goto insert;
					if (bp->b_cylin == ap->av_forw->b_cylin &&
					    bp->b_blkno < ap->av_forw->b_blkno)
						goto insert;
					ap = ap->av_forw;
				} while (ap->av_forw);
				goto insert;		/* after last */
			}
			ap = ap->av_forw;
		}
		/*
		 * 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->av_forw) {
		/*
		 * 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->av_forw->b_cylin < ap->b_cylin ||
		    bp->b_cylin < ap->av_forw->b_cylin ||
		    (bp->b_cylin == ap->av_forw->b_cylin &&
		    bp->b_blkno < ap->av_forw->b_blkno))
			goto insert;
		ap = ap->av_forw;
	}
	/*
	 * 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->av_forw = ap->av_forw;
	ap->av_forw = bp;
	if (ap == dp->b_actl)
		dp->b_actl = 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.
 */
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.
 */
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.
 */
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 int (*pr)(), sn;
	char partname = 'a' + part;
	extern printf(), addlog();

	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);
	}
}
Commit	Line	Data
da7c5cc6	1	/*
1ad7f5c8	2	* Copyright (c) 1982, 1986, 1988 Regents of the University of California.
7188ac27	3	* All rights reserved.
da7c5cc6	4	*
b702c21d	5	* %sccs.include.redist.c%
7188ac27	6	*
b702c21d	7	* @(#)ufs_disksubr.c 7.14 (Berkeley) %G%
da7c5cc6	8	*/
200a8f27	9
97934dfa MK	10	#include "param.h"
	11	#include "systm.h"
	12	#include "buf.h"
	13	#include "disklabel.h"
1ad7f5c8	14	#include "syslog.h"
9573c71c MK	15	#include "user.h"
9573c71c MK	16
200a8f27	17	/*
0d67fde4 BJ	18	* Seek sort for disks. We depend on the driver
	19	* which calls us using b_resid as the current cylinder number.
	20	*
	21	* The argument dp structure holds a b_actf activity chain pointer
	22	* on which we keep two queues, sorted in ascending cylinder order.
	23	* The first queue holds those requests which are positioned after
	24	* the current cylinder (in the first request); the second holds
	25	* requests which came in after their cylinder number was passed.
	26	* Thus we implement a one way scan, retracting after reaching the
	27	* end of the drive to the first request on the second queue,
	28	* at which time it becomes the first queue.
	29	*
	30	* A one-way scan is natural because of the way UNIX read-ahead
	31	* blocks are allocated.
200a8f27 BJ	32	*/
200a8f27 BJ	33
200a8f27 BJ	34	#define b_cylin b_resid
	35
	36	disksort(dp, bp)
0d67fde4	37	register struct buf dp, bp;
200a8f27 BJ	38	{
200a8f27 BJ	39	register struct buf *ap;
200a8f27	40
0d67fde4 BJ	41	/*
	42	* If nothing on the activity queue, then
	43	* we become the only thing.
	44	*/
200a8f27 BJ	45	ap = dp->b_actf;
	46	if(ap == NULL) {
	47	dp->b_actf = bp;
	48	dp->b_actl = bp;
	49	bp->av_forw = NULL;
	50	return;
	51	}
0d67fde4 BJ	52	/*
	53	* If we lie after the first (currently active)
	54	* request, then we must locate the second request list
	55	* and add ourselves to it.
	56	*/
	57	if (bp->b_cylin < ap->b_cylin) {
	58	while (ap->av_forw) {
	59	/*
	60	* Check for an ``inversion'' in the
	61	* normally ascending cylinder numbers,
	62	* indicating the start of the second request list.
	63	*/
	64	if (ap->av_forw->b_cylin < ap->b_cylin) {
	65	/*
	66	* Search the second request list
	67	* for the first request at a larger
	68	* cylinder number. We go before that;
	69	* if there is no such request, we go at end.
	70	*/
	71	do {
	72	if (bp->b_cylin < ap->av_forw->b_cylin)
	73	goto insert;
667e5665 MK	74	if (bp->b_cylin == ap->av_forw->b_cylin &&
	75	bp->b_blkno < ap->av_forw->b_blkno)
	76	goto insert;
0d67fde4 BJ	77	ap = ap->av_forw;
	78	} while (ap->av_forw);
	79	goto insert; /* after last */
	80	}
	81	ap = ap->av_forw;
200a8f27	82	}
0d67fde4 BJ	83	/*
	84	* No inversions... we will go after the last, and
	85	* be the first request in the second request list.
	86	*/
	87	goto insert;
200a8f27	88	}
0d67fde4 BJ	89	/*
	90	* Request is at/after the current request...
	91	* sort in the first request list.
	92	*/
	93	while (ap->av_forw) {
	94	/*
	95	* We want to go after the current request
	96	* if there is an inversion after it (i.e. it is
	97	* the end of the first request list), or if
	98	* the next request is a larger cylinder than our request.
	99	*/
	100	if (ap->av_forw->b_cylin < ap->b_cylin \|\|
667e5665 MK	101	bp->b_cylin < ap->av_forw->b_cylin \|\|
	102	(bp->b_cylin == ap->av_forw->b_cylin &&
	103	bp->b_blkno < ap->av_forw->b_blkno))
0d67fde4 BJ	104	goto insert;
	105	ap = ap->av_forw;
	106	}
	107	/*
	108	* Neither a second list nor a larger
	109	* request... we go at the end of the first list,
	110	* which is the same as the end of the whole schebang.
	111	*/
	112	insert:
	113	bp->av_forw = ap->av_forw;
	114	ap->av_forw = bp;
	115	if (ap == dp->b_actl)
200a8f27 BJ	116	dp->b_actl = bp;
200a8f27 BJ	117	}
97934dfa	118
9573c71c MK	119	/*
	120	* Attempt to read a disk label from a device
	121	* using the indicated stategy routine.
	122	* The label must be partly set up before this:
	123	* secpercyl and anything required in the strategy routine
	124	* (e.g., sector size) must be filled in before calling us.
	125	* Returns null on success and an error string on failure.
	126	*/
	127	char *
	128	readdisklabel(dev, strat, lp)
	129	dev_t dev;
	130	int (*strat)();
	131	register struct disklabel *lp;
	132	{
	133	register struct buf *bp;
	134	struct disklabel *dlp;
	135	char *msg = NULL;
	136
	137	if (lp->d_secperunit == 0)
	138	lp->d_secperunit = 0x1fffffff;
	139	lp->d_npartitions = 1;
	140	if (lp->d_partitions[0].p_size == 0)
	141	lp->d_partitions[0].p_size = 0x1fffffff;
	142	lp->d_partitions[0].p_offset = 0;
	143
d3633d66	144	bp = geteblk((int)lp->d_secsize);
9573c71c MK	145	bp->b_dev = dev;
9573c71c MK	146	bp->b_blkno = LABELSECTOR;
cd6cd535	147	bp->b_bcount = lp->d_secsize;
9573c71c MK	148	bp->b_flags = B_BUSY \| B_READ;
	149	bp->b_cylin = LABELSECTOR / lp->d_secpercyl;
	150	(*strat)(bp);
7188ac27	151	if (biowait(bp)) {
9573c71c	152	msg = "I/O error";
cd6cd535 MK	153	} else for (dlp = (struct disklabel *)bp->b_un.b_addr;
	154	dlp <= (struct disklabel )(bp->b_un.b_addr+DEV_BSIZE-sizeof(dlp));
	155	dlp = (struct disklabel )((char )dlp + sizeof(long))) {
	156	if (dlp->d_magic != DISKMAGIC \|\| dlp->d_magic2 != DISKMAGIC) {
	157	if (msg == NULL)
	158	msg = "no disk label";
0328954d KM	159	} else if (dlp->d_npartitions > MAXPARTITIONS \|\|
0328954d KM	160	dkcksum(dlp) != 0)
9573c71c	161	msg = "disk label corrupted";
cd6cd535	162	else {
9573c71c	163	lp = dlp;
cd6cd535 MK	164	msg = NULL;
	165	break;
	166	}
9573c71c MK	167	}
	168	bp->b_flags = B_INVAL \| B_AGE;
	169	brelse(bp);
	170	return (msg);
	171	}
	172
667e5665 MK	173	/*
	174	* Check new disk label for sensibility
	175	* before setting it.
	176	*/
	177	setdisklabel(olp, nlp, openmask)
	178	register struct disklabel olp, nlp;
	179	u_long openmask;
	180	{
	181	register i;
	182	register struct partition opp, npp;
	183
	184	if (nlp->d_magic != DISKMAGIC \|\| nlp->d_magic2 != DISKMAGIC \|\|
	185	dkcksum(nlp) != 0)
	186	return (EINVAL);
5adcb337	187	while ((i = ffs((long)openmask)) != 0) {
667e5665 MK	188	i--;
	189	openmask &= ~(1 << i);
	190	if (nlp->d_npartitions <= i)
	191	return (EBUSY);
	192	opp = &olp->d_partitions[i];
	193	npp = &nlp->d_partitions[i];
	194	if (npp->p_offset != opp->p_offset \|\| npp->p_size < opp->p_size)
	195	return (EBUSY);
	196	/*
	197	* Copy internally-set partition information
	198	* if new label doesn't include it. XXX
	199	*/
	200	if (npp->p_fstype == FS_UNUSED && opp->p_fstype != FS_UNUSED) {
	201	npp->p_fstype = opp->p_fstype;
	202	npp->p_fsize = opp->p_fsize;
	203	npp->p_frag = opp->p_frag;
	204	npp->p_cpg = opp->p_cpg;
	205	}
	206	}
d3633d66 MK	207	nlp->d_checksum = 0;
d3633d66 MK	208	nlp->d_checksum = dkcksum(nlp);
667e5665 MK	209	olp = nlp;
	210	return (0);
	211	}
	212
	213	/* encoding of disk minor numbers, should be elsewhere... */
	214	#define dkunit(dev) (minor(dev) >> 3)
	215	#define dkpart(dev) (minor(dev) & 07)
	216	#define dkminor(unit, part) (((unit) << 3) \| (part))
	217
	218	/*
	219	* Write disk label back to device after modification.
	220	*/
	221	writedisklabel(dev, strat, lp)
	222	dev_t dev;
	223	int (*strat)();
	224	register struct disklabel *lp;
	225	{
	226	struct buf *bp;
	227	struct disklabel *dlp;
	228	int labelpart;
	229	int error = 0;
	230
	231	labelpart = dkpart(dev);
	232	if (lp->d_partitions[labelpart].p_offset != 0) {
	233	if (lp->d_partitions[0].p_offset != 0)
	234	return (EXDEV); /* not quite right */
	235	labelpart = 0;
	236	}
d3633d66	237	bp = geteblk((int)lp->d_secsize);
667e5665 MK	238	bp->b_dev = makedev(major(dev), dkminor(dkunit(dev), labelpart));
	239	bp->b_blkno = LABELSECTOR;
	240	bp->b_bcount = lp->d_secsize;
	241	bp->b_flags = B_READ;
	242	(*strat)(bp);
7188ac27	243	if (error = biowait(bp))
d3633d66	244	goto done;
d3633d66 MK	245	for (dlp = (struct disklabel *)bp->b_un.b_addr;
	246	dlp <= (struct disklabel *)
	247	(bp->b_un.b_addr + lp->d_secsize - sizeof(*dlp));
	248	dlp = (struct disklabel )((char )dlp + sizeof(long))) {
	249	if (dlp->d_magic == DISKMAGIC && dlp->d_magic2 == DISKMAGIC &&
	250	dkcksum(dlp) == 0) {
	251	dlp = lp;
	252	bp->b_flags = B_WRITE;
	253	(*strat)(bp);
7188ac27	254	error = biowait(bp);
d3633d66 MK	255	goto done;
d3633d66 MK	256	}
667e5665	257	}
d3633d66 MK	258	error = ESRCH;
d3633d66 MK	259	done:
667e5665 MK	260	brelse(bp);
	261	return (error);
	262	}
	263
97934dfa MK	264	/*
	265	* Compute checksum for disk label.
	266	*/
	267	dkcksum(lp)
	268	register struct disklabel *lp;
	269	{
	270	register u_short start, end;
	271	register u_short sum = 0;
	272
	273	start = (u_short *)lp;
	274	end = (u_short *)&lp->d_partitions[lp->d_npartitions];
	275	while (start < end)
	276	sum ^= *start++;
	277	return (sum);
	278	}
1ad7f5c8 MK	279
	280	/*
	281	* Disk error is the preface to plaintive error messages
	282	* about failing disk transfers. It prints messages of the form
05de9101 MK	283
	284	hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d)
	285
1ad7f5c8 MK	286	* if the offset of the error in the transfer and a disk label
	287	* are both available. blkdone should be -1 if the position of the error
	288	* is unknown; the disklabel pointer may be null from drivers that have not
	289	* been converted to use them. The message is printed with printf
	290	* if pri is LOG_PRINTF, otherwise it uses log at the specified priority.
	291	* The message should be completed (with at least a newline) with printf
	292	* or addlog, respectively. There is no trailing space.
	293	*/
	294	diskerr(bp, dname, what, pri, blkdone, lp)
	295	register struct buf *bp;
	296	char dname, what;
	297	int pri, blkdone;
	298	register struct disklabel *lp;
	299	{
	300	int unit = dkunit(bp->b_dev), part = dkpart(bp->b_dev);
	301	register int (*pr)(), sn;
	302	char partname = 'a' + part;
	303	extern printf(), addlog();
	304
	305	if (pri != LOG_PRINTF) {
	306	log(pri, "");
	307	pr = addlog;
	308	} else
	309	pr = printf;
	310	(*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what,
	311	bp->b_flags & B_READ ? "read" : "writ");
	312	sn = bp->b_blkno;
	313	if (bp->b_bcount <= DEV_BSIZE)
	314	(*pr)("%d", sn);
	315	else {
	316	if (blkdone >= 0) {
	317	sn += blkdone;
	318	(*pr)("%d of ", sn);
	319	}
	320	(*pr)("%d-%d", bp->b_blkno,
	321	bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE);
	322	}
f2a9617b MK	323	if (lp && (blkdone >= 0 \|\| bp->b_bcount <= lp->d_secsize)) {
	324	#ifdef tahoe
	325	sn = DEV_BSIZE / lp->d_secsize; / XXX */
	326	#endif
1ad7f5c8	327	sn += lp->d_partitions[part].p_offset;
05de9101 MK	328	(*pr)(" (%s%d bn %d; cn %d", dname, unit, sn,
	329	sn / lp->d_secpercyl);
	330	sn %= lp->d_secpercyl;
9f79ca31	331	(*pr)(" tn %d sn %d)", sn / lp->d_nsectors, sn % lp->d_nsectors);
1ad7f5c8 MK	332	}
1ad7f5c8 MK	333	}