freebsd-dev/sys/dev/vinum/vinumrequest.c
Jeff Roberson 17661e5ac4 - Add an interlock argument to BUF_LOCK and BUF_TIMELOCK.
- Remove the buftimelock mutex and acquire the buf's interlock to protect
   these fields instead.
 - Hold the vnode interlock while locking bufs on the clean/dirty queues.
   This reduces some cases from one BUF_LOCK with a LK_NOWAIT and another
   BUF_LOCK with a LK_TIMEFAIL to a single lock.

Reviewed by:	arch, mckusick
2003-02-25 03:37:48 +00:00

1106 lines
38 KiB
C

/*-
* Copyright (c) 1997, 1998, 1999
* Nan Yang Computer Services Limited. All rights reserved.
*
* Parts copyright (c) 1997, 1998 Cybernet Corporation, NetMAX project.
*
* Written by Greg Lehey
*
* This software is distributed under the so-called ``Berkeley
* License'':
*
* 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 Nan Yang Computer
* Services Limited.
* 4. Neither the name of the Company 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 ``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 company 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.
*
* $Id: vinumrequest.c,v 1.32 2001/05/23 23:04:38 grog Exp grog $
* $FreeBSD$
*/
#include <dev/vinum/vinumhdr.h>
#include <dev/vinum/request.h>
#include <sys/resourcevar.h>
enum requeststatus bre(struct request *rq,
int plexno,
daddr_t * diskstart,
daddr_t diskend);
enum requeststatus bre5(struct request *rq,
int plexno,
daddr_t * diskstart,
daddr_t diskend);
enum requeststatus build_read_request(struct request *rq, int volplexno);
enum requeststatus build_write_request(struct request *rq);
enum requeststatus build_rq_buffer(struct rqelement *rqe, struct plex *plex);
int find_alternate_sd(struct request *rq);
int check_range_covered(struct request *);
void complete_rqe(struct buf *bp);
void complete_raid5_write(struct rqelement *);
int abortrequest(struct request *rq, int error);
void sdio_done(struct buf *bp);
int vinum_bounds_check(struct buf *bp, struct volume *vol);
caddr_t allocdatabuf(struct rqelement *rqe);
void freedatabuf(struct rqelement *rqe);
#ifdef VINUMDEBUG
struct rqinfo rqinfo[RQINFO_SIZE];
struct rqinfo *rqip = rqinfo;
void
logrq(enum rqinfo_type type, union rqinfou info, struct buf *ubp)
{
int s = splhigh();
microtime(&rqip->timestamp); /* when did this happen? */
rqip->type = type;
rqip->bp = ubp; /* user buffer */
switch (type) {
case loginfo_user_bp:
case loginfo_user_bpl:
case loginfo_sdio: /* subdisk I/O */
case loginfo_sdiol: /* subdisk I/O launch */
case loginfo_sdiodone: /* subdisk I/O complete */
bcopy(info.bp, &rqip->info.b, sizeof(struct buf));
rqip->devmajor = major(info.bp->b_dev);
rqip->devminor = minor(info.bp->b_dev);
break;
case loginfo_iodone:
case loginfo_rqe:
case loginfo_raid5_data:
case loginfo_raid5_parity:
bcopy(info.rqe, &rqip->info.rqe, sizeof(struct rqelement));
rqip->devmajor = major(info.rqe->b.b_dev);
rqip->devminor = minor(info.rqe->b.b_dev);
break;
case loginfo_lockwait:
case loginfo_lock:
case loginfo_unlock:
bcopy(info.lockinfo, &rqip->info.lockinfo, sizeof(struct rangelock));
break;
case loginfo_unused:
break;
}
rqip++;
if (rqip >= &rqinfo[RQINFO_SIZE]) /* wrap around */
rqip = rqinfo;
splx(s);
}
#endif
void
vinumstrategy(struct bio *biop)
{
struct buf *bp = (struct buf *) biop;
int volno;
struct volume *vol = NULL;
switch (DEVTYPE(bp->b_dev)) {
case VINUM_SD_TYPE:
case VINUM_RAWSD_TYPE:
sdio(bp);
return;
/*
* In fact, vinum doesn't handle drives: they're
* handled directly by the disk drivers
*/
case VINUM_DRIVE_TYPE:
default:
bp->b_error = EIO; /* I/O error */
bp->b_io.bio_flags |= BIO_ERROR;
bufdone(bp);
return;
case VINUM_VOLUME_TYPE: /* volume I/O */
volno = Volno(bp->b_dev);
vol = &VOL[volno];
if (vol->state != volume_up) { /* can't access this volume */
bp->b_error = EIO; /* I/O error */
bp->b_io.bio_flags |= BIO_ERROR;
bufdone(bp);
return;
}
if (vinum_bounds_check(bp, vol) <= 0) { /* don't like them bounds */
bufdone(bp);
return;
}
/* FALLTHROUGH */
/*
* Plex I/O is pretty much the same as volume I/O
* for a single plex. Indicate this by passing a NULL
* pointer (set above) for the volume
*/
case VINUM_PLEX_TYPE:
case VINUM_RAWPLEX_TYPE:
bp->b_resid = bp->b_bcount; /* transfer everything */
vinumstart(bp, 0);
return;
}
}
/*
* Start a transfer. Return -1 on error, 0 if OK,
* 1 if we need to retry. Parameter reviveok is
* set when doing transfers for revives: it allows
* transfers to be started immediately when a
* revive is in progress. During revive, normal
* transfers are queued if they share address
* space with a currently active revive operation.
*/
int
vinumstart(struct buf *bp, int reviveok)
{
int plexno;
int maxplex; /* maximum number of plexes to handle */
struct volume *vol;
struct request *rq; /* build up our request here */
enum requeststatus status;
#ifdef VINUMDEBUG
if (debug & DEBUG_LASTREQS)
logrq(loginfo_user_bp, (union rqinfou) bp, bp);
#endif
if ((bp->b_bcount % DEV_BSIZE) != 0) { /* bad length */
bp->b_error = EINVAL; /* invalid size */
bp->b_io.bio_flags |= BIO_ERROR;
bufdone(bp);
return -1;
}
rq = (struct request *) Malloc(sizeof(struct request)); /* allocate a request struct */
if (rq == NULL) { /* can't do it */
bp->b_error = ENOMEM; /* can't get memory */
bp->b_io.bio_flags |= BIO_ERROR;
bufdone(bp);
return -1;
}
bzero(rq, sizeof(struct request));
/*
* Note the volume ID. This can be NULL, which
* the request building functions use as an
* indication for single plex I/O.
*/
rq->bp = bp; /* and the user buffer struct */
if (DEVTYPE(bp->b_dev) == VINUM_VOLUME_TYPE) { /* it's a volume, */
rq->volplex.volno = Volno(bp->b_dev); /* get the volume number */
vol = &VOL[rq->volplex.volno]; /* and point to it */
vol->active++; /* one more active request */
maxplex = vol->plexes; /* consider all its plexes */
} else {
vol = NULL; /* no volume */
rq->volplex.plexno = Plexno(bp->b_dev); /* point to the plex */
rq->isplex = 1; /* note that it's a plex */
maxplex = 1; /* just the one plex */
}
if (bp->b_iocmd == BIO_READ) {
/*
* This is a read request. Decide
* which plex to read from.
*
* There's a potential race condition here,
* since we're not locked, and we could end
* up multiply incrementing the round-robin
* counter. This doesn't have any serious
* effects, however.
*/
if (vol != NULL) {
plexno = vol->preferred_plex; /* get the plex to use */
if (plexno < 0) { /* round robin */
plexno = vol->last_plex_read;
vol->last_plex_read++;
if (vol->last_plex_read >= vol->plexes) /* got the the end? */
vol->last_plex_read = 0; /* wrap around */
}
status = build_read_request(rq, plexno); /* build a request */
} else {
daddr_t diskaddr = bp->b_blkno; /* start offset of transfer */
status = bre(rq, /* build a request list */
rq->volplex.plexno,
&diskaddr,
diskaddr + (bp->b_bcount / DEV_BSIZE));
}
if (status > REQUEST_RECOVERED) { /* can't satisfy it */
if (status == REQUEST_DOWN) { /* not enough subdisks */
bp->b_error = EIO; /* I/O error */
bp->b_io.bio_flags |= BIO_ERROR;
}
bufdone(bp);
freerq(rq);
return -1;
}
return launch_requests(rq, reviveok); /* now start the requests if we can */
} else
/*
* This is a write operation. We write to all plexes. If this is
* a RAID-4 or RAID-5 plex, we must also update the parity stripe.
*/
{
if (vol != NULL) {
if ((vol->plexes > 0) /* multiple plex */
||(isparity((&PLEX[vol->plex[0]])))) { /* or RAID-[45], */
rq->save_data = bp->b_data; /* save the data buffer address */
bp->b_data = Malloc(bp->b_bufsize);
bcopy(rq->save_data, bp->b_data, bp->b_bufsize); /* make a copy */
rq->flags |= XFR_COPYBUF; /* and note that we did it */
}
status = build_write_request(rq);
} else { /* plex I/O */
daddr_t diskstart;
diskstart = bp->b_blkno; /* start offset of transfer */
status = bre(rq,
Plexno(bp->b_dev),
&diskstart,
bp->b_blkno + (bp->b_bcount / DEV_BSIZE)); /* build requests for the plex */
}
if (status > REQUEST_RECOVERED) { /* can't satisfy it */
if (status == REQUEST_DOWN) { /* not enough subdisks */
bp->b_error = EIO; /* I/O error */
bp->b_io.bio_flags |= BIO_ERROR;
}
if (rq->flags & XFR_COPYBUF) {
Free(bp->b_data);
bp->b_data = rq->save_data;
}
bufdone(bp);
freerq(rq);
return -1;
}
return launch_requests(rq, reviveok); /* now start the requests if we can */
}
}
/*
* Call the low-level strategy routines to
* perform the requests in a struct request
*/
int
launch_requests(struct request *rq, int reviveok)
{
struct rqgroup *rqg;
int rqno; /* loop index */
struct rqelement *rqe; /* current element */
struct drive *drive;
int rcount; /* request count */
/*
* First find out whether we're reviving, and the
* request contains a conflict. If so, we hang
* the request off plex->waitlist of the first
* plex we find which is reviving
*/
if ((rq->flags & XFR_REVIVECONFLICT) /* possible revive conflict */
&&(!reviveok)) { /* and we don't want to do it now, */
struct sd *sd;
struct request *waitlist; /* point to the waitlist */
sd = &SD[rq->sdno];
if (sd->waitlist != NULL) { /* something there already, */
waitlist = sd->waitlist;
while (waitlist->next != NULL) /* find the end */
waitlist = waitlist->next;
waitlist->next = rq; /* hook our request there */
} else
sd->waitlist = rq; /* hook our request at the front */
#ifdef VINUMDEBUG
if (debug & DEBUG_REVIVECONFLICT)
log(LOG_DEBUG,
"Revive conflict sd %d: %p\n%s dev %d.%d, offset 0x%llx, length %ld\n",
rq->sdno,
rq,
rq->bp->b_iocmd == BIO_READ ? "Read" : "Write",
major(rq->bp->b_dev),
minor(rq->bp->b_dev),
(long long)rq->bp->b_blkno,
rq->bp->b_bcount);
#endif
return 0; /* and get out of here */
}
rq->active = 0; /* nothing yet */
#ifdef VINUMDEBUG
if (debug & DEBUG_ADDRESSES)
log(LOG_DEBUG,
"Request: %p\n%s dev %d.%d, offset 0x%llx, length %ld\n",
rq,
rq->bp->b_iocmd == BIO_READ ? "Read" : "Write",
major(rq->bp->b_dev),
minor(rq->bp->b_dev),
(long long)rq->bp->b_blkno,
rq->bp->b_bcount);
vinum_conf.lastrq = rq;
vinum_conf.lastbuf = rq->bp;
if (debug & DEBUG_LASTREQS)
logrq(loginfo_user_bpl, (union rqinfou) rq->bp, rq->bp);
#endif
/*
* We used to have an splbio() here anyway, out
* of superstition. With the division of labour
* below (first count the requests, then issue
* them), it looks as if we don't need this
* splbio() protection. In fact, as dillon
* points out, there's a race condition
* incrementing and decrementing rq->active and
* rqg->active. This splbio() didn't help
* there, because the device strategy routine
* can sleep. Solve this by putting shorter
* duration locks on the code.
*/
/*
* This loop happens without any participation
* of the bottom half, so it requires no
* protection.
*/
for (rqg = rq->rqg; rqg != NULL; rqg = rqg->next) { /* through the whole request chain */
rqg->active = rqg->count; /* they're all active */
for (rqno = 0; rqno < rqg->count; rqno++) {
rqe = &rqg->rqe[rqno];
if (rqe->flags & XFR_BAD_SUBDISK) /* this subdisk is bad, */
rqg->active--; /* one less active request */
}
if (rqg->active) /* we have at least one active request, */
rq->active++; /* one more active request group */
}
/*
* Now fire off the requests. In this loop the
* bottom half could be completing requests
* before we finish, so we need splbio() protection.
*/
for (rqg = rq->rqg; rqg != NULL;) { /* through the whole request chain */
if (rqg->lockbase >= 0) /* this rqg needs a lock first */
rqg->lock = lockrange(rqg->lockbase, rqg->rq->bp, &PLEX[rqg->plexno]);
rcount = rqg->count;
for (rqno = 0; rqno < rcount;) {
rqe = &rqg->rqe[rqno];
/*
* Point to next rqg before the bottom end
* changes the structures.
*/
if (++rqno >= rcount)
rqg = rqg->next;
if ((rqe->flags & XFR_BAD_SUBDISK) == 0) { /* this subdisk is good, */
drive = &DRIVE[rqe->driveno]; /* look at drive */
drive->active++;
if (drive->active >= drive->maxactive)
drive->maxactive = drive->active;
vinum_conf.active++;
if (vinum_conf.active >= vinum_conf.maxactive)
vinum_conf.maxactive = vinum_conf.active;
#ifdef VINUMDEBUG
if (debug & DEBUG_ADDRESSES)
log(LOG_DEBUG,
" %s dev %d.%d, sd %d, offset 0x%x, devoffset 0x%llx, length %ld\n",
rqe->b.b_iocmd == BIO_READ ? "Read" : "Write",
major(rqe->b.b_dev),
minor(rqe->b.b_dev),
rqe->sdno,
(u_int) (rqe->b.b_blkno - SD[rqe->sdno].driveoffset),
(long long)rqe->b.b_blkno,
rqe->b.b_bcount);
if (debug & DEBUG_LASTREQS)
logrq(loginfo_rqe, (union rqinfou) rqe, rq->bp);
#endif
/* fire off the request */
DEV_STRATEGY(&rqe->b);
}
}
}
return 0;
}
/*
* define the low-level requests needed to perform a
* high-level I/O operation for a specific plex 'plexno'.
*
* Return REQUEST_OK if all subdisks involved in the request are up,
* REQUEST_DOWN if some subdisks are not up, and REQUEST_EOF if the
* request is at least partially outside the bounds of the subdisks.
*
* Modify the pointer *diskstart to point to the end address. On
* read, return on the first bad subdisk, so that the caller
* (build_read_request) can try alternatives.
*
* On entry to this routine, the rqg structures are not assigned. The
* assignment is performed by expandrq(). Strictly speaking, the
* elements rqe->sdno of all entries should be set to -1, since 0
* (from bzero) is a valid subdisk number. We avoid this problem by
* initializing the ones we use, and not looking at the others (index
* >= rqg->requests).
*/
enum requeststatus
bre(struct request *rq,
int plexno,
daddr_t * diskaddr,
daddr_t diskend)
{
int sdno;
struct sd *sd;
struct rqgroup *rqg;
struct buf *bp; /* user's bp */
struct plex *plex;
enum requeststatus status; /* return value */
daddr_t plexoffset; /* offset of transfer in plex */
daddr_t stripebase; /* base address of stripe (1st subdisk) */
daddr_t stripeoffset; /* offset in stripe */
daddr_t blockoffset; /* offset in stripe on subdisk */
struct rqelement *rqe; /* point to this request information */
daddr_t diskstart = *diskaddr; /* remember where this transfer starts */
enum requeststatus s; /* temp return value */
bp = rq->bp; /* buffer pointer */
status = REQUEST_OK; /* return value: OK until proven otherwise */
plex = &PLEX[plexno]; /* point to the plex */
switch (plex->organization) {
case plex_concat:
sd = NULL; /* (keep compiler quiet) */
for (sdno = 0; sdno < plex->subdisks; sdno++) {
sd = &SD[plex->sdnos[sdno]];
if (*diskaddr < sd->plexoffset) /* we must have a hole, */
status = REQUEST_DEGRADED; /* note the fact */
if (*diskaddr < (sd->plexoffset + sd->sectors)) { /* the request starts in this subdisk */
rqg = allocrqg(rq, 1); /* space for the request */
if (rqg == NULL) { /* malloc failed */
bp->b_error = ENOMEM;
bp->b_io.bio_flags |= BIO_ERROR;
return REQUEST_ENOMEM;
}
rqg->plexno = plexno;
rqe = &rqg->rqe[0]; /* point to the element */
rqe->rqg = rqg; /* group */
rqe->sdno = sd->sdno; /* put in the subdisk number */
plexoffset = *diskaddr; /* start offset in plex */
rqe->sdoffset = plexoffset - sd->plexoffset; /* start offset in subdisk */
rqe->useroffset = plexoffset - diskstart; /* start offset in user buffer */
rqe->dataoffset = 0;
rqe->datalen = min(diskend - *diskaddr, /* number of sectors to transfer in this sd */
sd->sectors - rqe->sdoffset);
rqe->groupoffset = 0; /* no groups for concatenated plexes */
rqe->grouplen = 0;
rqe->buflen = rqe->datalen; /* buffer length is data buffer length */
rqe->flags = 0;
rqe->driveno = sd->driveno;
if (sd->state != sd_up) { /* *now* we find the sd is down */
s = checksdstate(sd, rq, *diskaddr, diskend); /* do we need to change state? */
if (s == REQUEST_DOWN) { /* down? */
rqe->flags = XFR_BAD_SUBDISK; /* yup */
if (rq->bp->b_iocmd == BIO_READ) /* read request, */
return REQUEST_DEGRADED; /* give up here */
/*
* If we're writing, don't give up
* because of a bad subdisk. Go
* through to the bitter end, but note
* which ones we can't access.
*/
status = REQUEST_DEGRADED; /* can't do it all */
}
}
*diskaddr += rqe->datalen; /* bump the address */
if (build_rq_buffer(rqe, plex)) { /* build the buffer */
deallocrqg(rqg);
bp->b_error = ENOMEM;
bp->b_io.bio_flags |= BIO_ERROR;
return REQUEST_ENOMEM; /* can't do it */
}
}
if (*diskaddr == diskend) /* we're finished, */
break; /* get out of here */
}
/*
* We've got to the end of the plex. Have we got to the end of
* the transfer? It would seem that having an offset beyond the
* end of the subdisk is an error, but in fact it can happen if
* the volume has another plex of different size. There's a valid
* question as to why you would want to do this, but currently
* it's allowed.
*
* In a previous version, I returned REQUEST_DOWN here. I think
* REQUEST_EOF is more appropriate now.
*/
if (diskend > sd->sectors + sd->plexoffset) /* pointing beyond EOF? */
status = REQUEST_EOF;
break;
case plex_striped:
{
while (*diskaddr < diskend) { /* until we get it all sorted out */
if (*diskaddr >= plex->length) /* beyond the end of the plex */
return REQUEST_EOF; /* can't continue */
/* The offset of the start address from the start of the stripe. */
stripeoffset = *diskaddr % (plex->stripesize * plex->subdisks);
/* The plex-relative address of the start of the stripe. */
stripebase = *diskaddr - stripeoffset;
/* The number of the subdisk in which the start is located. */
sdno = stripeoffset / plex->stripesize;
/* The offset from the beginning of the stripe on this subdisk. */
blockoffset = stripeoffset % plex->stripesize;
sd = &SD[plex->sdnos[sdno]]; /* the subdisk in question */
rqg = allocrqg(rq, 1); /* space for the request */
if (rqg == NULL) { /* malloc failed */
bp->b_error = ENOMEM;
bp->b_io.bio_flags |= BIO_ERROR;
return REQUEST_ENOMEM;
}
rqg->plexno = plexno;
rqe = &rqg->rqe[0]; /* point to the element */
rqe->rqg = rqg;
rqe->sdoffset = stripebase / plex->subdisks + blockoffset; /* start offset in this subdisk */
rqe->useroffset = *diskaddr - diskstart; /* The offset of the start in the user buffer */
rqe->dataoffset = 0;
rqe->datalen = min(diskend - *diskaddr, /* the amount remaining to transfer */
plex->stripesize - blockoffset); /* and the amount left in this stripe */
rqe->groupoffset = 0; /* no groups for striped plexes */
rqe->grouplen = 0;
rqe->buflen = rqe->datalen; /* buffer length is data buffer length */
rqe->flags = 0;
rqe->sdno = sd->sdno; /* put in the subdisk number */
rqe->driveno = sd->driveno;
if (sd->state != sd_up) { /* *now* we find the sd is down */
s = checksdstate(sd, rq, *diskaddr, diskend); /* do we need to change state? */
if (s == REQUEST_DOWN) { /* down? */
rqe->flags = XFR_BAD_SUBDISK; /* yup */
if (rq->bp->b_iocmd == BIO_READ) /* read request, */
return REQUEST_DEGRADED; /* give up here */
/*
* If we're writing, don't give up
* because of a bad subdisk. Go through
* to the bitter end, but note which
* ones we can't access.
*/
status = REQUEST_DEGRADED; /* can't do it all */
}
}
/*
* It would seem that having an offset
* beyond the end of the subdisk is an
* error, but in fact it can happen if the
* volume has another plex of different
* size. There's a valid question as to why
* you would want to do this, but currently
* it's allowed.
*/
if (rqe->sdoffset + rqe->datalen > sd->sectors) { /* ends beyond the end of the subdisk? */
rqe->datalen = sd->sectors - rqe->sdoffset; /* truncate */
#ifdef VINUMDEBUG
if (debug & DEBUG_EOFINFO) { /* tell on the request */
log(LOG_DEBUG,
"vinum: EOF on plex %s, sd %s offset %x (user offset %llx)\n",
plex->name,
sd->name,
(u_int) sd->sectors,
(long long)bp->b_blkno);
log(LOG_DEBUG,
"vinum: stripebase %#llx, stripeoffset %#llxx, blockoffset %#llx\n",
(unsigned long long)stripebase,
(unsigned long long)stripeoffset,
(unsigned long long)blockoffset);
}
#endif
}
if (build_rq_buffer(rqe, plex)) { /* build the buffer */
deallocrqg(rqg);
bp->b_error = ENOMEM;
bp->b_io.bio_flags |= BIO_ERROR;
return REQUEST_ENOMEM; /* can't do it */
}
*diskaddr += rqe->datalen; /* look at the remainder */
if ((*diskaddr < diskend) /* didn't finish the request on this stripe */
&&(*diskaddr < plex->length)) { /* and there's more to come */
plex->multiblock++; /* count another one */
if (sdno == plex->subdisks - 1) /* last subdisk, */
plex->multistripe++; /* another stripe as well */
}
}
}
break;
/*
* RAID-4 and RAID-5 are complicated enough to have their own
* function.
*/
case plex_raid4:
case plex_raid5:
status = bre5(rq, plexno, diskaddr, diskend);
break;
default:
log(LOG_ERR, "vinum: invalid plex type %d in bre\n", plex->organization);
status = REQUEST_DOWN; /* can't access it */
}
return status;
}
/*
* Build up a request structure for reading volumes.
* This function is not needed for plex reads, since there's
* no recovery if a plex read can't be satisified.
*/
enum requeststatus
build_read_request(struct request *rq, /* request */
int plexindex)
{ /* index in the volume's plex table */
struct buf *bp;
daddr_t startaddr; /* offset of previous part of transfer */
daddr_t diskaddr; /* offset of current part of transfer */
daddr_t diskend; /* and end offset of transfer */
int plexno; /* plex index in vinum_conf */
struct rqgroup *rqg; /* point to the request we're working on */
struct volume *vol; /* volume in question */
int recovered = 0; /* set if we recover a read */
enum requeststatus status = REQUEST_OK;
int plexmask; /* bit mask of plexes, for recovery */
bp = rq->bp; /* buffer pointer */
diskaddr = bp->b_blkno; /* start offset of transfer */
diskend = diskaddr + (bp->b_bcount / DEV_BSIZE); /* and end offset of transfer */
rqg = &rq->rqg[plexindex]; /* plex request */
vol = &VOL[rq->volplex.volno]; /* point to volume */
while (diskaddr < diskend) { /* build up request components */
startaddr = diskaddr;
status = bre(rq, vol->plex[plexindex], &diskaddr, diskend); /* build up a request */
switch (status) {
case REQUEST_OK:
continue;
case REQUEST_RECOVERED:
/*
* XXX FIXME if we have more than one plex, and we can
* satisfy the request from another, don't use the
* recovered request, since it's more expensive.
*/
recovered = 1;
break;
case REQUEST_ENOMEM:
return status;
/*
* If we get here, our request is not complete. Try
* to fill in the missing parts from another plex.
* This can happen multiple times in this function,
* and we reinitialize the plex mask each time, since
* we could have a hole in our plexes.
*/
case REQUEST_EOF:
case REQUEST_DOWN: /* can't access the plex */
case REQUEST_DEGRADED: /* can't access the plex */
plexmask = ((1 << vol->plexes) - 1) /* all plexes in the volume */
&~(1 << plexindex); /* except for the one we were looking at */
for (plexno = 0; plexno < vol->plexes; plexno++) {
if (plexmask == 0) /* no plexes left to try */
return REQUEST_DOWN; /* failed */
diskaddr = startaddr; /* start at the beginning again */
if (plexmask & (1 << plexno)) { /* we haven't tried this plex yet */
bre(rq, vol->plex[plexno], &diskaddr, diskend); /* try a request */
if (diskaddr > startaddr) { /* we satisfied another part */
recovered = 1; /* we recovered from the problem */
status = REQUEST_OK; /* don't complain about it */
break;
}
}
}
if (diskaddr == startaddr) /* didn't get any further, */
return status;
}
if (recovered)
vol->recovered_reads += recovered; /* adjust our recovery count */
}
return status;
}
/*
* Build up a request structure for writes.
* Return 0 if all subdisks involved in the request are up, 1 if some
* subdisks are not up, and -1 if the request is at least partially
* outside the bounds of the subdisks.
*/
enum requeststatus
build_write_request(struct request *rq)
{ /* request */
struct buf *bp;
daddr_t diskstart; /* offset of current part of transfer */
daddr_t diskend; /* and end offset of transfer */
int plexno; /* plex index in vinum_conf */
struct volume *vol; /* volume in question */
enum requeststatus status;
bp = rq->bp; /* buffer pointer */
vol = &VOL[rq->volplex.volno]; /* point to volume */
diskend = bp->b_blkno + (bp->b_bcount / DEV_BSIZE); /* end offset of transfer */
status = REQUEST_DOWN; /* assume the worst */
for (plexno = 0; plexno < vol->plexes; plexno++) {
diskstart = bp->b_blkno; /* start offset of transfer */
/*
* Build requests for the plex.
* We take the best possible result here (min,
* not max): we're happy if we can write at all
*/
status = min(status, bre(rq,
vol->plex[plexno],
&diskstart,
diskend));
}
return status;
}
/* Fill in the struct buf part of a request element. */
enum requeststatus
build_rq_buffer(struct rqelement *rqe, struct plex *plex)
{
struct sd *sd; /* point to subdisk */
struct volume *vol;
struct buf *bp;
struct buf *ubp; /* user (high level) buffer header */
vol = &VOL[rqe->rqg->rq->volplex.volno];
sd = &SD[rqe->sdno]; /* point to subdisk */
bp = &rqe->b;
ubp = rqe->rqg->rq->bp; /* pointer to user buffer header */
/* Initialize the buf struct */
/* copy these flags from user bp */
bp->b_flags = ubp->b_flags & (B_NOCACHE | B_ASYNC);
bp->b_io.bio_flags = 0;
bp->b_iocmd = ubp->b_iocmd;
#ifdef VINUMDEBUG
if (rqe->flags & XFR_BUFLOCKED) /* paranoia */
panic("build_rq_buffer: rqe already locked"); /* XXX remove this when we're sure */
#endif
BUF_LOCKINIT(bp); /* get a lock for the buffer */
BUF_LOCK(bp, LK_EXCLUSIVE, NULL); /* and lock it */
BUF_KERNPROC(bp);
rqe->flags |= XFR_BUFLOCKED;
bp->b_iodone = complete_rqe;
/*
* You'd think that we wouldn't need to even
* build the request buffer for a dead subdisk,
* but in some cases we need information like
* the user buffer address. Err on the side of
* generosity and supply what we can. That
* obviously doesn't include drive information
* when the drive is dead.
*/
if ((rqe->flags & XFR_BAD_SUBDISK) == 0) /* subdisk is accessible, */
bp->b_dev = DRIVE[rqe->driveno].dev; /* drive device */
bp->b_blkno = rqe->sdoffset + sd->driveoffset; /* start address */
bp->b_bcount = rqe->buflen << DEV_BSHIFT; /* number of bytes to transfer */
bp->b_resid = bp->b_bcount; /* and it's still all waiting */
bp->b_bufsize = bp->b_bcount; /* and buffer size */
bp->b_rcred = FSCRED; /* we have the filesystem credentials */
bp->b_wcred = FSCRED; /* we have the filesystem credentials */
if (rqe->flags & XFR_MALLOCED) { /* this operation requires a malloced buffer */
bp->b_data = Malloc(bp->b_bcount); /* get a buffer to put it in */
if (bp->b_data == NULL) { /* failed */
abortrequest(rqe->rqg->rq, ENOMEM);
return REQUEST_ENOMEM; /* no memory */
}
} else
/*
* Point directly to user buffer data. This means
* that we don't need to do anything when we have
* finished the transfer
*/
bp->b_data = ubp->b_data + rqe->useroffset * DEV_BSIZE;
/*
* On a recovery read, we perform an XOR of
* all blocks to the user buffer. To make
* this work, we first clean out the buffer
*/
if ((rqe->flags & (XFR_RECOVERY_READ | XFR_BAD_SUBDISK))
== (XFR_RECOVERY_READ | XFR_BAD_SUBDISK)) { /* bad subdisk of a recovery read */
int length = rqe->grouplen << DEV_BSHIFT; /* and count involved */
char *data = (char *) &rqe->b.b_data[rqe->groupoffset << DEV_BSHIFT]; /* destination */
bzero(data, length); /* clean it out */
}
return 0;
}
/*
* Abort a request: free resources and complete the
* user request with the specified error
*/
int
abortrequest(struct request *rq, int error)
{
struct buf *bp = rq->bp; /* user buffer */
bp->b_error = error;
freerq(rq); /* free everything we're doing */
bp->b_io.bio_flags |= BIO_ERROR;
return error; /* and give up */
}
/*
* Check that our transfer will cover the
* complete address space of the user request.
*
* Return 1 if it can, otherwise 0
*/
int
check_range_covered(struct request *rq)
{
return 1;
}
/* Perform I/O on a subdisk */
void
sdio(struct buf *bp)
{
int s; /* spl */
struct sd *sd;
struct sdbuf *sbp;
daddr_t endoffset;
struct drive *drive;
#ifdef VINUMDEBUG
if (debug & DEBUG_LASTREQS)
logrq(loginfo_sdio, (union rqinfou) bp, bp);
#endif
sd = &SD[Sdno(bp->b_dev)]; /* point to the subdisk */
drive = &DRIVE[sd->driveno];
if (drive->state != drive_up) {
if (sd->state >= sd_crashed) {
if (bp->b_iocmd == BIO_WRITE) /* writing, */
set_sd_state(sd->sdno, sd_stale, setstate_force);
else
set_sd_state(sd->sdno, sd_crashed, setstate_force);
}
bp->b_error = EIO;
bp->b_io.bio_flags |= BIO_ERROR;
bufdone(bp);
return;
}
/*
* We allow access to any kind of subdisk as long as we can expect
* to get the I/O performed.
*/
if (sd->state < sd_empty) { /* nothing to talk to, */
bp->b_error = EIO;
bp->b_io.bio_flags |= BIO_ERROR;
bufdone(bp);
return;
}
/* Get a buffer */
sbp = (struct sdbuf *) Malloc(sizeof(struct sdbuf));
if (sbp == NULL) {
bp->b_error = ENOMEM;
bp->b_io.bio_flags |= BIO_ERROR;
bufdone(bp);
return;
}
bzero(sbp, sizeof(struct sdbuf)); /* start with nothing */
sbp->b.b_flags = bp->b_flags;
sbp->b.b_iocmd = bp->b_iocmd;
sbp->b.b_bufsize = bp->b_bufsize; /* buffer size */
sbp->b.b_bcount = bp->b_bcount; /* number of bytes to transfer */
sbp->b.b_resid = bp->b_resid; /* and amount waiting */
sbp->b.b_dev = DRIVE[sd->driveno].dev; /* device */
sbp->b.b_data = bp->b_data; /* data buffer */
sbp->b.b_blkno = bp->b_blkno + sd->driveoffset;
sbp->b.b_iodone = sdio_done; /* come here on completion */
BUF_LOCKINIT(&sbp->b); /* get a lock for the buffer */
BUF_LOCK(&sbp->b, LK_EXCLUSIVE, NULL); /* and lock it */
BUF_KERNPROC(&sbp->b);
sbp->bp = bp; /* note the address of the original header */
sbp->sdno = sd->sdno; /* note for statistics */
sbp->driveno = sd->driveno;
endoffset = bp->b_blkno + sbp->b.b_bcount / DEV_BSIZE; /* final sector offset */
if (endoffset > sd->sectors) { /* beyond the end */
sbp->b.b_bcount -= (endoffset - sd->sectors) * DEV_BSIZE; /* trim */
if (sbp->b.b_bcount <= 0) { /* nothing to transfer */
bp->b_resid = bp->b_bcount; /* nothing transferred */
bufdone(bp);
BUF_UNLOCK(&sbp->b);
BUF_LOCKFREE(&sbp->b);
Free(sbp);
return;
}
}
#ifdef VINUMDEBUG
if (debug & DEBUG_ADDRESSES)
log(LOG_DEBUG,
" %s dev %d.%d, sd %d, offset 0x%x, devoffset 0x%x, length %ld\n",
sbp->b.b_iocmd == BIO_READ ? "Read" : "Write",
major(sbp->b.b_dev),
minor(sbp->b.b_dev),
sbp->sdno,
(u_int) (sbp->b.b_blkno - SD[sbp->sdno].driveoffset),
(int) sbp->b.b_blkno,
sbp->b.b_bcount);
#endif
s = splbio();
#ifdef VINUMDEBUG
if (debug & DEBUG_LASTREQS)
logrq(loginfo_sdiol, (union rqinfou) &sbp->b, &sbp->b);
#endif
DEV_STRATEGY(&sbp->b);
splx(s);
}
/*
* Simplified version of bounds_check_with_label
* Determine the size of the transfer, and make sure it is
* within the boundaries of the partition. Adjust transfer
* if needed, and signal errors or early completion.
*
* Volumes are simpler than disk slices: they only contain
* one component (though we call them a, b and c to make
* system utilities happy), and they always take up the
* complete space of the "partition".
*
* I'm still not happy with this: why should the label be
* protected? If it weren't so damned difficult to write
* one in the first pleace (because it's protected), it wouldn't
* be a problem.
*/
int
vinum_bounds_check(struct buf *bp, struct volume *vol)
{
int maxsize = vol->size; /* size of the partition (sectors) */
int size = (bp->b_bcount + DEV_BSIZE - 1) >> DEV_BSHIFT; /* size of this request (sectors) */
/* Would this transfer overwrite the disk label? */
if (bp->b_blkno <= LABELSECTOR /* starts before or at the label */
#if LABELSECTOR != 0
&& bp->b_blkno + size > LABELSECTOR /* and finishes after */
#endif
&& (!(vol->flags & VF_RAW)) /* and it's not raw */
&&(bp->b_iocmd == BIO_WRITE) /* and it's a write */
&&(!vol->flags & (VF_WLABEL | VF_LABELLING))) { /* and we're not allowed to write the label */
bp->b_error = EROFS; /* read-only */
bp->b_io.bio_flags |= BIO_ERROR;
return -1;
}
if (size == 0) /* no transfer specified, */
return 0; /* treat as EOF */
/* beyond partition? */
if (bp->b_blkno < 0 /* negative start */
|| bp->b_blkno + size > maxsize) { /* or goes beyond the end of the partition */
/* if exactly at end of disk, return an EOF */
if (bp->b_blkno == maxsize) {
bp->b_resid = bp->b_bcount;
return 0;
}
/* or truncate if part of it fits */
size = maxsize - bp->b_blkno;
if (size <= 0) { /* nothing to transfer */
bp->b_error = EINVAL;
bp->b_io.bio_flags |= BIO_ERROR;
return -1;
}
bp->b_bcount = size << DEV_BSHIFT;
}
bp->b_pblkno = bp->b_blkno;
return 1;
}
/*
* Allocate a request group and hook
* it in in the list for rq
*/
struct rqgroup *
allocrqg(struct request *rq, int elements)
{
struct rqgroup *rqg; /* the one we're going to allocate */
int size = sizeof(struct rqgroup) + elements * sizeof(struct rqelement);
rqg = (struct rqgroup *) Malloc(size);
if (rqg != NULL) { /* malloc OK, */
if (rq->rqg) /* we already have requests */
rq->lrqg->next = rqg; /* hang it off the end */
else /* first request */
rq->rqg = rqg; /* at the start */
rq->lrqg = rqg; /* this one is the last in the list */
bzero(rqg, size); /* no old junk */
rqg->rq = rq; /* point back to the parent request */
rqg->count = elements; /* number of requests in the group */
rqg->lockbase = -1; /* no lock required yet */
}
return rqg;
}
/*
* Deallocate a request group out of a chain. We do
* this by linear search: the chain is short, this
* almost never happens, and currently it can only
* happen to the first member of the chain.
*/
void
deallocrqg(struct rqgroup *rqg)
{
struct rqgroup *rqgc = rqg->rq->rqg; /* point to the request chain */
if (rqg->lock) /* got a lock? */
unlockrange(rqg->plexno, rqg->lock); /* yes, free it */
if (rqgc == rqg) /* we're first in line */
rqg->rq->rqg = rqg->next; /* unhook ourselves */
else {
while ((rqgc->next != NULL) /* find the group */
&&(rqgc->next != rqg))
rqgc = rqgc->next;
if (rqgc->next == NULL)
log(LOG_ERR,
"vinum deallocrqg: rqg %p not found in request %p\n",
rqg->rq,
rqg);
else
rqgc->next = rqg->next; /* make the chain jump over us */
}
Free(rqg);
}
/* Local Variables: */
/* fill-column: 50 */
/* End: */