freebsd-nq/sys/kern/subr_disklabel.c
Poul-Henning Kamp 29f88f470e Comment out Kirks io-request priority hack until we can do this in a
civilized way which doesn't cause grief.

The problem is that it is not generally safe to cast a "struct bio
*" to a "struct buf *".  Things like ccd, vinum, ata-raid and GEOM
constructs bio's which are not entrails of a struct buf.

Also, curthread may or may not have anything to do with the I/O request
at hand.

The correct solution can either be to tag struct bio's with a
priority derived from the requesting threads nice and have disksort
act on this field, this wouldn't address the "silly-seek syndrome"
where two equal processes bang the diskheads from one edge to the
other of the disk repeatedly.

Alternatively, and probably better: a sleep should be introduced
either at the time the I/O is requested or at the time it is completed
where we can be sure to sleep in the right thread.

The sleep also needs to be in constant timeunits, 1/hz can be practicaly
any sub-second size, at high HZ the current code practically doesn't
do anything.
2002-04-22 06:53:20 +00:00

426 lines
12 KiB
C

/*
* Copyright (c) 1982, 1986, 1988, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/buf.h>
#include <sys/conf.h>
#include <sys/kernel.h>
#include <sys/disklabel.h>
#include <sys/diskslice.h>
#include <sys/syslog.h>
#include <machine/atomic.h>
#ifdef notquite
/*
* Mutex to use when delaying niced I/O bound processes in bioqdisksort().
*/
static struct mtx dksort_mtx;
static void
dksort_init(void)
{
mtx_init(&dksort_mtx, "dksort", NULL, MTX_DEF);
}
SYSINIT(dksort, SI_SUB_DRIVERS, SI_ORDER_MIDDLE, dksort_init, NULL)
#endif
/*
* Seek sort for disks.
*
* The buf_queue keep two queues, sorted in ascending block order. The first
* queue holds those requests which are positioned after the current block
* (in the first request); the second, which starts at queue->switch_point,
* holds requests which came in after their block 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.
*/
void
bioqdisksort(bioq, bp)
struct bio_queue_head *bioq;
struct bio *bp;
{
struct bio *bq;
struct bio *bn;
struct bio *be;
#ifdef notquite
struct thread *td = curthread;
if (td && td->td_ksegrp->kg_nice > 0) {
TAILQ_FOREACH(bn, &bioq->queue, bio_queue)
if (BIOTOBUF(bp)->b_vp != BIOTOBUF(bn)->b_vp)
break;
if (bn != NULL) {
mtx_lock(&dksort_mtx);
msleep((caddr_t)&dksort_mtx, &dksort_mtx,
PPAUSE | PCATCH | PDROP, "ioslow",
td->td_ksegrp->kg_nice);
}
}
#endif
if (!atomic_cmpset_int(&bioq->busy, 0, 1))
panic("Recursing in bioqdisksort()");
be = TAILQ_LAST(&bioq->queue, bio_queue);
/*
* If the queue is empty or we are an
* ordered transaction, then it's easy.
*/
if ((bq = bioq_first(bioq)) == NULL) {
bioq_insert_tail(bioq, bp);
bioq->busy = 0;
return;
} else if (bioq->insert_point != NULL) {
/*
* A certain portion of the list is
* "locked" to preserve ordering, so
* we can only insert after the insert
* point.
*/
bq = bioq->insert_point;
} else {
/*
* If we lie before the last removed (currently active)
* request, and are not inserting ourselves into the
* "locked" portion of the list, then we must add ourselves
* to the second request list.
*/
if (bp->bio_pblkno < bioq->last_pblkno) {
bq = bioq->switch_point;
/*
* If we are starting a new secondary list,
* then it's easy.
*/
if (bq == NULL) {
bioq->switch_point = bp;
bioq_insert_tail(bioq, bp);
bioq->busy = 0;
return;
}
/*
* If we lie ahead of the current switch point,
* insert us before the switch point and move
* the switch point.
*/
if (bp->bio_pblkno < bq->bio_pblkno) {
bioq->switch_point = bp;
TAILQ_INSERT_BEFORE(bq, bp, bio_queue);
bioq->busy = 0;
return;
}
} else {
if (bioq->switch_point != NULL)
be = TAILQ_PREV(bioq->switch_point,
bio_queue, bio_queue);
/*
* If we lie between last_pblkno and bq,
* insert before bq.
*/
if (bp->bio_pblkno < bq->bio_pblkno) {
TAILQ_INSERT_BEFORE(bq, bp, bio_queue);
bioq->busy = 0;
return;
}
}
}
/*
* Request is at/after our current position in the list.
* Optimize for sequential I/O by seeing if we go at the tail.
*/
if (bp->bio_pblkno > be->bio_pblkno) {
TAILQ_INSERT_AFTER(&bioq->queue, be, bp, bio_queue);
bioq->busy = 0;
return;
}
/* Otherwise, insertion sort */
while ((bn = TAILQ_NEXT(bq, bio_queue)) != NULL) {
/*
* We want to go after the current request if it is the end
* of the first request list, or if the next request is a
* larger cylinder than our request.
*/
if (bn == bioq->switch_point
|| bp->bio_pblkno < bn->bio_pblkno)
break;
bq = bn;
}
TAILQ_INSERT_AFTER(&bioq->queue, bq, bp, bio_queue);
bioq->busy = 0;
}
/*
* Attempt to read a disk label from a device using the indicated strategy
* routine. The label must be partly set up before this: secpercyl, secsize
* and anything required in the strategy routine (e.g., dummy bounds for the
* partition containing the label) must be filled in before calling us.
* Returns NULL on success and an error string on failure.
*/
char *
readdisklabel(dev, lp)
dev_t dev;
register struct disklabel *lp;
{
register struct buf *bp;
struct disklabel *dlp;
char *msg = NULL;
bp = geteblk((int)lp->d_secsize);
bp->b_dev = dev;
bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE);
bp->b_bcount = lp->d_secsize;
bp->b_flags &= ~B_INVAL;
bp->b_iocmd = BIO_READ;
DEV_STRATEGY(bp, 1);
if (bufwait(bp))
msg = "I/O error";
else if (bp->b_resid != 0)
msg = "disk too small for a label";
else for (dlp = (struct disklabel *)bp->b_data;
dlp <= (struct disklabel *)((char *)bp->b_data +
lp->d_secsize - 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 int i;
register struct partition *opp, *npp;
/*
* Check it is actually a disklabel we are looking at.
*/
if (nlp->d_magic != DISKMAGIC || nlp->d_magic2 != DISKMAGIC ||
dkcksum(nlp) != 0)
return (EINVAL);
/*
* For each partition that we think is open,
*/
while ((i = ffs((long)openmask)) != 0) {
i--;
/*
* Check it is not changing....
*/
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 we are using it then we had better stay the same type)
* This is possibly dubious, as someone else noted (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);
}
/*
* Write disk label back to device after modification.
*/
int
writedisklabel(dev, lp)
dev_t dev;
register struct disklabel *lp;
{
struct buf *bp;
struct disklabel *dlp;
int error = 0;
if (lp->d_partitions[RAW_PART].p_offset != 0)
return (EXDEV); /* not quite right */
bp = geteblk((int)lp->d_secsize);
bp->b_dev = dkmodpart(dev, RAW_PART);
bp->b_blkno = LABELSECTOR * ((int)lp->d_secsize/DEV_BSIZE);
bp->b_bcount = lp->d_secsize;
#if 1
/*
* We read the label first to see if it's there,
* in which case we will put ours at the same offset into the block..
* (I think this is stupid [Julian])
* Note that you can't write a label out over a corrupted label!
* (also stupid.. how do you write the first one? by raw writes?)
*/
bp->b_flags &= ~B_INVAL;
bp->b_iocmd = BIO_READ;
DEV_STRATEGY(bp, 1);
error = bufwait(bp);
if (error)
goto done;
if (bp->b_resid != 0) {
error = ENOSPC;
goto done;
}
for (dlp = (struct disklabel *)bp->b_data;
dlp <= (struct disklabel *)
((char *)bp->b_data + 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_DONE;
bp->b_iocmd = BIO_WRITE;
#ifdef __alpha__
alpha_fix_srm_checksum(bp);
#endif
DEV_STRATEGY(bp, 1);
error = bufwait(bp);
goto done;
}
}
error = ESRCH;
done:
#else
bzero(bp->b_data, lp->d_secsize);
dlp = (struct disklabel *)bp->b_data;
*dlp = *lp;
bp->b_flags &= ~B_INVAL;
bp->b_iocmd = BIO_WRITE;
DEV_STRATEGY(bp, 1);
error = bufwait(bp);
#endif
bp->b_flags |= B_INVAL | B_AGE;
brelse(bp);
return (error);
}
/*
* 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.
* The message should be completed with at least a newline. There is no
* trailing space.
*/
void
diskerr(bp, what, blkdone, lp)
struct bio *bp;
char *what;
int blkdone;
register struct disklabel *lp;
{
int part = dkpart(bp->bio_dev);
char partname[2];
char *sname;
daddr_t sn;
*partname = '\0';
sname = bp->bio_dev->si_name;
printf("%s%s: %s %sing fsbn ", sname, partname, what,
bp->bio_cmd == BIO_READ ? "read" : "writ");
sn = bp->bio_blkno;
if (bp->bio_bcount <= DEV_BSIZE)
printf("%ld", (long)sn);
else {
if (blkdone >= 0) {
sn += blkdone;
printf("%ld of ", (long)sn);
}
printf("%ld-%ld", (long)bp->bio_blkno,
(long)(bp->bio_blkno + (bp->bio_bcount - 1) / DEV_BSIZE));
}
if (lp && (blkdone >= 0 || bp->bio_bcount <= lp->d_secsize)) {
sn += lp->d_partitions[part].p_offset;
/*
* XXX should add slice offset and not print the slice,
* but we don't know the slice pointer.
* XXX should print bp->b_pblkno so that this will work
* independent of slices, labels and bad sector remapping,
* but some drivers don't set bp->b_pblkno.
*/
printf(" (%s bn %ld; cn %ld", sname, (long)sn,
(long)(sn / lp->d_secpercyl));
sn %= (long)lp->d_secpercyl;
printf(" tn %ld sn %ld)", (long)(sn / lp->d_nsectors),
(long)(sn % lp->d_nsectors));
}
}