[BACK]Return to genfs_vnops.c CVS log [TXT][DIR] Up to [cvs.NetBSD.org] / src / sys / miscfs / genfs

Annotation of src/sys/miscfs/genfs/genfs_vnops.c, Revision 1.11.4.7

1.11.4.7! perseant    1: /*     $NetBSD: genfs_vnops.c,v 1.11.4.6 1999/08/09 00:05:54 chs Exp $ */
1.6       fvdl        2:
                      3: /*
                      4:  * Copyright (c) 1982, 1986, 1989, 1993
                      5:  *     The Regents of the University of California.  All rights reserved.
                      6:  *
                      7:  * Redistribution and use in source and binary forms, with or without
                      8:  * modification, are permitted provided that the following conditions
                      9:  * are met:
                     10:  * 1. Redistributions of source code must retain the above copyright
                     11:  *    notice, this list of conditions and the following disclaimer.
                     12:  * 2. Redistributions in binary form must reproduce the above copyright
                     13:  *    notice, this list of conditions and the following disclaimer in the
                     14:  *    documentation and/or other materials provided with the distribution.
                     15:  * 3. All advertising materials mentioning features or use of this software
                     16:  *    must display the following acknowledgement:
                     17:  *     This product includes software developed by the University of
                     18:  *     California, Berkeley and its contributors.
                     19:  * 4. Neither the name of the University nor the names of its contributors
                     20:  *    may be used to endorse or promote products derived from this software
                     21:  *    without specific prior written permission.
                     22:  *
                     23:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     24:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     25:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     26:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     27:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     28:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     29:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     30:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     31:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     32:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     33:  * SUCH DAMAGE.
                     34:  *
                     35:  */
1.5       perry      36:
1.8       thorpej    37: #include "opt_nfsserver.h"
                     38:
1.1       mycroft    39: #include <sys/param.h>
                     40: #include <sys/systm.h>
1.6       fvdl       41: #include <sys/proc.h>
1.1       mycroft    42: #include <sys/kernel.h>
                     43: #include <sys/mount.h>
                     44: #include <sys/namei.h>
                     45: #include <sys/vnode.h>
                     46: #include <sys/malloc.h>
1.3       mycroft    47: #include <sys/poll.h>
1.1       mycroft    48:
                     49: #include <miscfs/genfs/genfs.h>
1.6       fvdl       50: #include <miscfs/specfs/specdev.h>
1.1       mycroft    51:
1.11.4.1  chs        52: #include <vm/vm.h>
                     53: #include <uvm/uvm.h>
1.11.4.7! perseant   54: #include <uvm/uvm_pager.h>
1.11.4.1  chs        55:
1.8       thorpej    56: #ifdef NFSSERVER
                     57: #include <nfs/rpcv2.h>
                     58: #include <nfs/nfsproto.h>
                     59: #include <nfs/nfs.h>
                     60: #include <nfs/nqnfs.h>
                     61: #include <nfs/nfs_var.h>
                     62: #endif
                     63:
1.1       mycroft    64: int
1.3       mycroft    65: genfs_poll(v)
1.1       mycroft    66:        void *v;
                     67: {
1.3       mycroft    68:        struct vop_poll_args /* {
1.1       mycroft    69:                struct vnode *a_vp;
1.3       mycroft    70:                int a_events;
1.1       mycroft    71:                struct proc *a_p;
                     72:        } */ *ap = v;
                     73:
1.3       mycroft    74:        return (ap->a_events & (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM));
1.1       mycroft    75: }
                     76:
                     77: int
                     78: genfs_fsync(v)
                     79:        void *v;
                     80: {
                     81:        struct vop_fsync_args /* {
                     82:                struct vnode *a_vp;
                     83:                struct ucred *a_cred;
1.7       kleink     84:                int a_flags;
1.1       mycroft    85:                struct proc *a_p;
                     86:        } */ *ap = v;
                     87:        register struct vnode *vp = ap->a_vp;
1.11      mycroft    88:        int wait;
1.1       mycroft    89:
1.11      mycroft    90:        wait = (ap->a_flags & FSYNC_WAIT) != 0;
                     91:        vflushbuf(vp, wait);
                     92:        if ((ap->a_flags & FSYNC_DATAONLY) != 0)
1.7       kleink     93:                return (0);
1.11      mycroft    94:        else
                     95:                return (VOP_UPDATE(ap->a_vp, NULL, NULL, wait));
1.1       mycroft    96: }
                     97:
                     98: int
1.4       kleink     99: genfs_seek(v)
                    100:        void *v;
                    101: {
                    102:        struct vop_seek_args /* {
                    103:                struct vnode *a_vp;
                    104:                off_t a_oldoff;
                    105:                off_t a_newoff;
                    106:                struct ucred *a_ucred;
                    107:        } */ *ap = v;
                    108:
                    109:        if (ap->a_newoff < 0)
                    110:                return (EINVAL);
                    111:
                    112:        return (0);
                    113: }
                    114:
                    115: int
1.1       mycroft   116: genfs_abortop(v)
                    117:        void *v;
                    118: {
                    119:        struct vop_abortop_args /* {
                    120:                struct vnode *a_dvp;
                    121:                struct componentname *a_cnp;
                    122:        } */ *ap = v;
                    123:
                    124:        if ((ap->a_cnp->cn_flags & (HASBUF | SAVESTART)) == HASBUF)
                    125:                FREE(ap->a_cnp->cn_pnbuf, M_NAMEI);
                    126:        return (0);
                    127: }
                    128:
                    129: /*ARGSUSED*/
                    130: int
                    131: genfs_badop(v)
                    132:        void *v;
                    133: {
                    134:
                    135:        panic("genfs: bad op");
                    136: }
                    137:
                    138: /*ARGSUSED*/
                    139: int
                    140: genfs_nullop(v)
                    141:        void *v;
                    142: {
                    143:
                    144:        return (0);
1.10      kleink    145: }
                    146:
                    147: /*ARGSUSED*/
                    148: int
                    149: genfs_einval(v)
                    150:        void *v;
                    151: {
                    152:
                    153:        return (EINVAL);
1.1       mycroft   154: }
                    155:
                    156: /*ARGSUSED*/
                    157: int
                    158: genfs_eopnotsupp(v)
                    159:        void *v;
                    160: {
                    161:
                    162:        return (EOPNOTSUPP);
                    163: }
                    164:
1.11.4.5  thorpej   165: /*
                    166:  * Called when an fs doesn't support a particular vop but the vop needs to
                    167:  * vrele, vput, or vunlock passed in vnodes.
                    168:  */
                    169: int
                    170: genfs_eopnotsupp_rele(v)
                    171:        void *v;
                    172: {
                    173:        struct vop_generic_args /*
                    174:                struct vnodeop_desc *a_desc;
                    175:                / * other random data follows, presumably * /
                    176:        } */ *ap = v;
                    177:        struct vnodeop_desc *desc = ap->a_desc;
                    178:        struct vnode *vp;
                    179:        int flags, i, j, offset;
                    180:
                    181:        flags = desc->vdesc_flags;
                    182:        for (i = 0; i < VDESC_MAX_VPS; flags >>=1, i++) {
                    183:                if ((offset = desc->vdesc_vp_offsets[i]) == VDESC_NO_OFFSET)
                    184:                        break;  /* stop at end of list */
                    185:                if ((j = flags & VDESC_VP0_WILLPUT)) {
                    186:                        vp = *VOPARG_OFFSETTO(struct vnode**,offset,ap);
                    187:                        switch (j) {
                    188:                        case VDESC_VP0_WILLPUT:
                    189:                                vput(vp);
                    190:                                break;
                    191:                        case VDESC_VP0_WILLUNLOCK:
                    192:                                VOP_UNLOCK(vp, 0);
                    193:                                break;
                    194:                        case VDESC_VP0_WILLRELE:
                    195:                                vrele(vp);
                    196:                                break;
                    197:                        }
                    198:                }
                    199:        }
                    200:
                    201:        return (EOPNOTSUPP);
                    202: }
                    203:
1.1       mycroft   204: /*ARGSUSED*/
                    205: int
                    206: genfs_ebadf(v)
                    207:        void *v;
                    208: {
                    209:
                    210:        return (EBADF);
1.9       matthias  211: }
                    212:
                    213: /* ARGSUSED */
                    214: int
                    215: genfs_enoioctl(v)
                    216:        void *v;
                    217: {
                    218:
                    219:        return (ENOTTY);
1.6       fvdl      220: }
                    221:
                    222:
                    223: /*
                    224:  * Eliminate all activity associated with  the requested vnode
                    225:  * and with all vnodes aliased to the requested vnode.
                    226:  */
                    227: int
                    228: genfs_revoke(v)
                    229:        void *v;
                    230: {
                    231:        struct vop_revoke_args /* {
                    232:                struct vnode *a_vp;
                    233:                int a_flags;
                    234:        } */ *ap = v;
                    235:        struct vnode *vp, *vq;
                    236:        struct proc *p = curproc;       /* XXX */
                    237:
                    238: #ifdef DIAGNOSTIC
                    239:        if ((ap->a_flags & REVOKEALL) == 0)
                    240:                panic("genfs_revoke: not revokeall");
                    241: #endif
                    242:
                    243:        vp = ap->a_vp;
                    244:        simple_lock(&vp->v_interlock);
                    245:
                    246:        if (vp->v_flag & VALIASED) {
                    247:                /*
                    248:                 * If a vgone (or vclean) is already in progress,
                    249:                 * wait until it is done and return.
                    250:                 */
                    251:                if (vp->v_flag & VXLOCK) {
                    252:                        vp->v_flag |= VXWANT;
                    253:                        simple_unlock(&vp->v_interlock);
                    254:                        tsleep((caddr_t)vp, PINOD, "vop_revokeall", 0);
                    255:                        return (0);
                    256:                }
                    257:                /*
                    258:                 * Ensure that vp will not be vgone'd while we
                    259:                 * are eliminating its aliases.
                    260:                 */
                    261:                vp->v_flag |= VXLOCK;
                    262:                simple_unlock(&vp->v_interlock);
                    263:                while (vp->v_flag & VALIASED) {
                    264:                        simple_lock(&spechash_slock);
                    265:                        for (vq = *vp->v_hashchain; vq; vq = vq->v_specnext) {
                    266:                                if (vq->v_rdev != vp->v_rdev ||
                    267:                                    vq->v_type != vp->v_type || vp == vq)
                    268:                                        continue;
                    269:                                simple_unlock(&spechash_slock);
                    270:                                vgone(vq);
                    271:                                break;
                    272:                        }
                    273:                        if (vq == NULLVP)
                    274:                                simple_unlock(&spechash_slock);
                    275:                }
                    276:                /*
                    277:                 * Remove the lock so that vgone below will
                    278:                 * really eliminate the vnode after which time
                    279:                 * vgone will awaken any sleepers.
                    280:                 */
                    281:                simple_lock(&vp->v_interlock);
                    282:                vp->v_flag &= ~VXLOCK;
                    283:        }
                    284:        vgonel(vp, p);
                    285:        return (0);
                    286: }
                    287:
                    288: /*
1.11.4.5  thorpej   289:  * Lock the node.
1.6       fvdl      290:  */
                    291: int
1.11.4.5  thorpej   292: genfs_lock(v)
1.6       fvdl      293:        void *v;
                    294: {
                    295:        struct vop_lock_args /* {
                    296:                struct vnode *a_vp;
                    297:                int a_flags;
                    298:                struct proc *a_p;
                    299:        } */ *ap = v;
                    300:        struct vnode *vp = ap->a_vp;
                    301:
1.11.4.5  thorpej   302:        return (lockmgr(&vp->v_lock, ap->a_flags, &vp->v_interlock));
1.6       fvdl      303: }
                    304:
                    305: /*
1.11.4.5  thorpej   306:  * Unlock the node.
1.6       fvdl      307:  */
                    308: int
1.11.4.5  thorpej   309: genfs_unlock(v)
1.6       fvdl      310:        void *v;
                    311: {
                    312:        struct vop_unlock_args /* {
                    313:                struct vnode *a_vp;
                    314:                int a_flags;
                    315:                struct proc *a_p;
                    316:        } */ *ap = v;
                    317:        struct vnode *vp = ap->a_vp;
                    318:
1.11.4.5  thorpej   319:        return (lockmgr(&vp->v_lock, ap->a_flags | LK_RELEASE,
                    320:                &vp->v_interlock));
1.6       fvdl      321: }
                    322:
                    323: /*
1.11.4.5  thorpej   324:  * Return whether or not the node is locked.
1.6       fvdl      325:  */
                    326: int
1.11.4.5  thorpej   327: genfs_islocked(v)
1.6       fvdl      328:        void *v;
                    329: {
                    330:        struct vop_islocked_args /* {
                    331:                struct vnode *a_vp;
                    332:        } */ *ap = v;
                    333:        struct vnode *vp = ap->a_vp;
                    334:
1.11.4.5  thorpej   335:        return (lockstatus(&vp->v_lock));
                    336: }
                    337:
                    338: /*
                    339:  * Stubs to use when there is no locking to be done on the underlying object.
                    340:  */
                    341: int
                    342: genfs_nolock(v)
                    343:        void *v;
                    344: {
                    345:        struct vop_lock_args /* {
                    346:                struct vnode *a_vp;
                    347:                int a_flags;
                    348:                struct proc *a_p;
                    349:        } */ *ap = v;
                    350:
                    351:        /*
                    352:         * Since we are not using the lock manager, we must clear
                    353:         * the interlock here.
                    354:         */
                    355:        if (ap->a_flags & LK_INTERLOCK)
                    356:                simple_unlock(&ap->a_vp->v_interlock);
                    357:        return (0);
                    358: }
                    359:
                    360: int
                    361: genfs_nounlock(v)
                    362:        void *v;
                    363: {
                    364:        return (0);
                    365: }
                    366:
                    367: int
                    368: genfs_noislocked(v)
                    369:        void *v;
                    370: {
                    371:        return (0);
1.8       thorpej   372: }
                    373:
                    374: /*
                    375:  * Local lease check for NFS servers.  Just set up args and let
                    376:  * nqsrv_getlease() do the rest.  If NFSSERVER is not in the kernel,
                    377:  * this is a null operation.
                    378:  */
                    379: int
                    380: genfs_lease_check(v)
                    381:        void *v;
                    382: {
                    383: #ifdef NFSSERVER
                    384:        struct vop_lease_args /* {
                    385:                struct vnode *a_vp;
                    386:                struct proc *a_p;
                    387:                struct ucred *a_cred;
                    388:                int a_flag;
                    389:        } */ *ap = v;
                    390:        u_int32_t duration = 0;
                    391:        int cache;
                    392:        u_quad_t frev;
                    393:
                    394:        (void) nqsrv_getlease(ap->a_vp, &duration, ND_CHECK | ap->a_flag,
                    395:            NQLOCALSLP, ap->a_p, (struct mbuf *)0, &cache, &frev, ap->a_cred);
                    396:        return (0);
                    397: #else
                    398:        return (0);
                    399: #endif /* NFSSERVER */
1.11.4.1  chs       400: }
                    401:
                    402: /*
                    403:  * generic VM getpages routine.
                    404:  * Return PG_BUSY pages for the given range,
                    405:  * reading from backing store if necessary.
                    406:  */
1.11.4.4  chs       407:
1.11.4.1  chs       408: int
                    409: genfs_getpages(v)
                    410:        void *v;
                    411: {
                    412:        struct vop_getpages_args /* {
                    413:                struct vnode *a_vp;
1.11.4.6  chs       414:                voff_t a_offset;
1.11.4.1  chs       415:                vm_page_t *a_m;
                    416:                int *a_count;
                    417:                int a_centeridx;
                    418:                vm_prot_t a_access_type;
                    419:                int a_advice;
                    420:                int a_flags;
                    421:        } */ *ap = v;
                    422:
1.11.4.4  chs       423:        off_t eof, offset, origoffset, startoffset, endoffset;
1.11.4.1  chs       424:        daddr_t lbn, blkno;
1.11.4.4  chs       425:        int s, i, error, npages, npgs, run, ridx, pidx, pcount;
1.11.4.2  chs       426:        int bsize, bshift, dev_bshift, dev_bsize;
1.11.4.1  chs       427:        int flags = ap->a_flags;
1.11.4.2  chs       428:        size_t bytes, iobytes, tailbytes, totalbytes, skipbytes;
                    429:        boolean_t sawhole = FALSE;
1.11.4.1  chs       430:        char *kva;
                    431:        struct buf *bp, *mbp;
                    432:        struct vnode *vp = ap->a_vp;
                    433:        struct uvm_object *uobj = &vp->v_uvm.u_obj;
1.11.4.4  chs       434:        struct vm_page *pgs[16];                        /* XXX 16 */
1.11.4.2  chs       435:        struct ucred *cred = curproc->p_ucred;          /* XXX curproc */
1.11.4.1  chs       436:        UVMHIST_FUNC("genfs_getpages"); UVMHIST_CALLED(ubchist);
                    437:
                    438: #ifdef DIAGNOSTIC
                    439:        if (ap->a_centeridx < 0 || ap->a_centeridx > *ap->a_count) {
                    440:                panic("genfs_getpages: centeridx %d out of range",
                    441:                      ap->a_centeridx);
                    442:        }
                    443:        if (ap->a_offset & (PAGE_SIZE - 1)) {
                    444:                panic("genfs_getpages: offset 0x%x", (int)ap->a_offset);
                    445:        }
1.11.4.4  chs       446:        if (*ap->a_count < 0) {
                    447:                panic("genfs_getpages: count %d < 0", *ap->a_count);
                    448:        }
1.11.4.1  chs       449: #endif
                    450:
                    451:        /*
                    452:         * Bounds-check the request.
                    453:         */
                    454:
1.11.4.4  chs       455:        eof = vp->v_uvm.u_size;
                    456:        if (ap->a_offset >= eof) {
1.11.4.1  chs       457:                if ((flags & PGO_LOCKED) == 0) {
                    458:                        simple_unlock(&uobj->vmobjlock);
                    459:                }
1.11.4.4  chs       460:                UVMHIST_LOG(ubchist, "off 0x%x count %d goes past EOF 0x%x",
                    461:                            (int)ap->a_offset, *ap->a_count, (int)eof,0);
1.11.4.2  chs       462:                return EINVAL;
1.11.4.1  chs       463:        }
                    464:
                    465:        /*
                    466:         * For PGO_LOCKED requests, just return whatever's in memory.
                    467:         */
                    468:
                    469:        if (flags & PGO_LOCKED) {
                    470:                uvn_findpages(uobj, ap->a_offset, ap->a_count, ap->a_m,
                    471:                              UFP_NOWAIT|UFP_NOALLOC|UFP_NORDONLY);
                    472:
1.11.4.2  chs       473:                return ap->a_m[ap->a_centeridx] == NULL ? EBUSY : 0;
1.11.4.1  chs       474:        }
                    475:
1.11.4.4  chs       476:        if (ap->a_offset + ((*ap->a_count - 1) << PAGE_SHIFT) >= eof) {
                    477:                panic("genfs_getpages: non LOCKED req past EOF vp %p", vp);
                    478:        }
                    479:
1.11.4.1  chs       480:        /* vnode is VOP_LOCKed, uobj is locked */
                    481:
                    482:        error = 0;
                    483:
                    484:        /*
1.11.4.4  chs       485:         * find the requested pages and make some simple checks.
                    486:         * leave space in the page array for a whole block.
1.11.4.1  chs       487:         */
                    488:
1.11.4.4  chs       489:        bshift = vp->v_mount->mnt_fs_bshift;
                    490:        bsize = 1 << bshift;
                    491:        dev_bshift = vp->v_mount->mnt_dev_bshift;
                    492:        dev_bsize = 1 << dev_bshift;
                    493:
                    494:        npages = *ap->a_count;
                    495:        origoffset = ap->a_offset;
                    496:        startoffset = origoffset & ~((off_t)bsize - 1);
                    497:        endoffset = round_page((origoffset + (npages << PAGE_SHIFT)
                    498:                                + bsize - 1) & ~((off_t)bsize - 1));
                    499:        ridx = (origoffset - startoffset) >> PAGE_SHIFT;
                    500:
                    501:        memset(pgs, 0, sizeof(pgs));
                    502:        uvn_findpages(uobj, origoffset, &npages, &pgs[ridx], UFP_ALL);
1.11.4.2  chs       503:
1.11.4.1  chs       504:        /*
1.11.4.4  chs       505:         * if PGO_OVERWRITE is set, don't bother reading the pages.
1.11.4.2  chs       506:         * PGO_OVERWRITE also means that the caller guarantees
1.11.4.4  chs       507:         * that the pages already have backing store allocated.
1.11.4.1  chs       508:         */
                    509:
1.11.4.2  chs       510:        if (flags & PGO_OVERWRITE) {
                    511:                UVMHIST_LOG(ubchist, "PGO_OVERWRITE",0,0,0,0);
1.11.4.1  chs       512:
1.11.4.2  chs       513:                /* XXX for now, zero the page if we allocated it */
1.11.4.4  chs       514:                for (i = 0; i < npages; i++) {
                    515:                        struct vm_page *pg = pgs[ridx + i];
                    516:                        if (pg->flags & PG_FAKE) {
                    517:                                uvm_pagezero(pg);
                    518:                                pg->flags &= ~PG_FAKE;
                    519:                        }
1.11.4.2  chs       520:                }
1.11.4.1  chs       521:
1.11.4.2  chs       522:                simple_unlock(&uobj->vmobjlock);
                    523:                goto out;
                    524:        }
1.11.4.1  chs       525:
                    526:        /*
1.11.4.4  chs       527:         * if the pages are already resident, just return them.
1.11.4.1  chs       528:         */
                    529:
1.11.4.4  chs       530:        for (i = 0; i < npages; i++) {
                    531:                struct vm_page *pg = pgs[ridx + i];
1.11.4.3  chs       532:
1.11.4.4  chs       533:                if ((pg->flags & PG_FAKE) != 0 ||
                    534:                    ((ap->a_access_type & VM_PROT_WRITE) &&
                    535:                      (pg->flags & PG_RDONLY))) {
                    536:                        break;
                    537:                }
                    538:        }
                    539:        if (i == npages) {
                    540:                UVMHIST_LOG(ubchist, "returning cached pages", 0,0,0,0);
1.11.4.2  chs       541:                simple_unlock(&uobj->vmobjlock);
1.11.4.3  chs       542:                goto out;
1.11.4.1  chs       543:        }
                    544:
                    545:        /*
1.11.4.2  chs       546:         * the page wasn't resident and we're not overwriting,
                    547:         * so we're going to have to do some i/o.
1.11.4.4  chs       548:         * find any additional pages needed to cover the expanded range.
1.11.4.1  chs       549:         */
                    550:
1.11.4.4  chs       551:        if (startoffset != origoffset) {
                    552:                UVMHIST_LOG(ubchist, "reset npages start 0x%x end 0x%x",
                    553:                            (int)startoffset, (int)endoffset, 0,0);
                    554:                npages = (endoffset - startoffset) >> PAGE_SHIFT;
                    555:                if (npages == 0) {
                    556:                        panic("XXX getpages npages = 0");
                    557:                }
                    558:                npgs = npages;
                    559:                uvn_findpages(uobj, startoffset, &npgs, pgs, UFP_ALL);
1.11.4.1  chs       560:        }
1.11.4.2  chs       561:        simple_unlock(&uobj->vmobjlock);
1.11.4.1  chs       562:
                    563:        /*
1.11.4.2  chs       564:         * read the desired page(s).
1.11.4.1  chs       565:         */
                    566:
1.11.4.2  chs       567:        totalbytes = npages << PAGE_SHIFT;
1.11.4.4  chs       568:        bytes = min(totalbytes,
                    569:                    (vp->v_uvm.u_size - startoffset + dev_bsize - 1) &
1.11.4.1  chs       570:                    ~(dev_bsize - 1));
1.11.4.2  chs       571:        tailbytes = totalbytes - bytes;
                    572:        skipbytes = 0;
1.11.4.1  chs       573:
1.11.4.2  chs       574:        kva = (void *)uvm_pagermapin(pgs, npages, M_WAITOK);
1.11.4.1  chs       575:
                    576:        s = splbio();
                    577:        mbp = pool_get(&bufpool, PR_WAITOK);
                    578:        splx(s);
                    579:        mbp->b_bufsize = bytes;
1.11.4.4  chs       580:        mbp->b_data = kva;
1.11.4.2  chs       581:        mbp->b_resid = mbp->b_bcount = bytes;
1.11.4.1  chs       582:        mbp->b_flags = B_BUSY|B_READ| (flags & PGO_SYNCIO ? 0 : B_CALL);
                    583:        mbp->b_iodone = uvm_aio_biodone;
                    584:        mbp->b_vp = vp;
                    585:
1.11.4.4  chs       586:        /*
                    587:         * if EOF is in the middle of the last page, zero the part past EOF.
                    588:         */
                    589:
                    590:        if (tailbytes > 0) {
                    591:                memset(kva + bytes, 0, tailbytes);
                    592:        }
                    593:
                    594:        /*
                    595:         * now loop over the pages, reading as needed.
                    596:         */
                    597:
1.11.4.1  chs       598:        bp = NULL;
1.11.4.4  chs       599:        offset = startoffset;
1.11.4.1  chs       600:        for (; bytes > 0; offset += iobytes, bytes -= iobytes) {
1.11.4.2  chs       601:
                    602:                /*
                    603:                 * skip pages which don't need to be read.
                    604:                 */
                    605:
                    606:                pidx = (offset - startoffset) >> PAGE_SHIFT;
                    607:                while ((pgs[pidx]->flags & PG_FAKE) == 0) {
                    608:                        size_t b;
                    609:
                    610:                        if (offset & (PAGE_SIZE - 1)) {
                    611:                                panic("genfs_getpages: skipping from middle "
                    612:                                      "of page");
                    613:                        }
                    614:
                    615:                        b = min(PAGE_SIZE, bytes);
                    616:                        offset += b;
                    617:                        bytes -= b;
                    618:                        skipbytes += b;
                    619:                        pidx++;
1.11.4.4  chs       620:                        UVMHIST_LOG(ubchist, "skipping, new offset 0x%x",
                    621:                                    (int)offset, 0,0,0);
1.11.4.2  chs       622:                        if (bytes == 0) {
                    623:                                goto loopdone;
                    624:                        }
                    625:                }
1.11.4.1  chs       626:
                    627:                /*
                    628:                 * bmap the file to find out the blkno to read from and
1.11.4.4  chs       629:                 * how much we can read in one i/o.  if bmap returns an error,
                    630:                 * skip the rest of the top-level i/o.
1.11.4.1  chs       631:                 */
                    632:
1.11.4.2  chs       633:                lbn = offset >> bshift;
1.11.4.1  chs       634:                error = VOP_BMAP(vp, lbn, NULL, &blkno, &run);
                    635:                if (error) {
                    636:                        UVMHIST_LOG(ubchist, "VOP_BMAP lbn 0x%x -> %d\n",
                    637:                                    lbn, error,0,0);
1.11.4.4  chs       638:                        skipbytes += bytes;
                    639:                        tailbytes = 0;
                    640:                        goto loopdone;
1.11.4.1  chs       641:                }
                    642:
1.11.4.2  chs       643:                /*
1.11.4.4  chs       644:                 * see how many pages can be read with this i/o.
1.11.4.2  chs       645:                 * reduce the i/o size if necessary.
                    646:                 */
1.11.4.1  chs       647:
1.11.4.2  chs       648:                iobytes = min(((lbn + 1 + run) << bshift) - offset, bytes);
                    649:                if (offset + iobytes > round_page(offset)) {
                    650:                        pcount = 1;
                    651:                        while (pidx + pcount < npages &&
                    652:                               pgs[pidx + pcount]->flags & PG_FAKE) {
                    653:                                pcount++;
1.11.4.1  chs       654:                        }
1.11.4.2  chs       655:                        iobytes = min(iobytes, (pcount << PAGE_SHIFT) -
                    656:                                      (offset - trunc_page(offset)));
                    657:                }
1.11.4.1  chs       658:
1.11.4.2  chs       659:                /*
                    660:                 * if this block isn't allocated, zero it instead of reading it.
1.11.4.4  chs       661:                 * if this is a read access, mark the pages we zeroed PG_RDONLY.
1.11.4.2  chs       662:                 */
1.11.4.1  chs       663:
1.11.4.7! perseant  664:                if (blkno < 0) {
1.11.4.2  chs       665:                        UVMHIST_LOG(ubchist, "lbn 0x%x -> HOLE", lbn,0,0,0);
1.11.4.1  chs       666:
1.11.4.2  chs       667:                        sawhole = TRUE;
1.11.4.4  chs       668:                        memset(kva + (offset - startoffset), 0, iobytes);
                    669:
                    670:                        if (ap->a_access_type == VM_PROT_READ) {
                    671:                                int holepages =
                    672:                                        (round_page(offset + iobytes) -
                    673:                                         trunc_page(offset)) >> PAGE_SHIFT;
                    674:                                for (i = 0; i < holepages; i++) {
                    675:                                        pgs[pidx + i]->flags |= PG_RDONLY;
                    676:                                }
                    677:                        }
1.11.4.1  chs       678:                        continue;
                    679:                }
                    680:
                    681:                /*
                    682:                 * allocate a sub-buf for this piece of the i/o
                    683:                 * (or just use mbp if there's only 1 piece),
                    684:                 * and start it going.
                    685:                 */
                    686:
1.11.4.4  chs       687:                if (offset == startoffset && iobytes == bytes) {
1.11.4.1  chs       688:                        bp = mbp;
                    689:                } else {
                    690:                        s = splbio();
                    691:                        bp = pool_get(&bufpool, PR_WAITOK);
                    692:                        splx(s);
1.11.4.4  chs       693:                        bp->b_data = kva + offset - startoffset;
1.11.4.2  chs       694:                        bp->b_resid = bp->b_bcount = iobytes;
1.11.4.1  chs       695:                        bp->b_flags = B_BUSY|B_READ|B_CALL;
                    696:                        bp->b_iodone = uvm_aio_biodone1;
                    697:                        bp->b_vp = vp;
                    698:                }
                    699:                bp->b_lblkno = 0;
                    700:                bp->b_private = mbp;
                    701:
                    702:                /* adjust physical blkno for partial blocks */
                    703:                bp->b_blkno = blkno + ((offset - (lbn << bshift)) >>
                    704:                                       dev_bshift);
1.11.4.2  chs       705:
                    706:                UVMHIST_LOG(ubchist, "bp %p offset 0x%x bcount 0x%x blkno 0x%x",
                    707:                            bp, (int)offset, (int)iobytes, bp->b_blkno);
1.11.4.1  chs       708:
                    709:                VOP_STRATEGY(bp);
                    710:        }
1.11.4.2  chs       711:
1.11.4.4  chs       712: loopdone:
                    713:        s = splbio();
1.11.4.2  chs       714:        if (skipbytes) {
                    715:                mbp->b_resid -= skipbytes;
                    716:                if (mbp->b_resid == 0) {
                    717:                        biodone(mbp);
                    718:                }
                    719:        }
1.11.4.4  chs       720:        splx(s);
1.11.4.1  chs       721:        if ((flags & PGO_SYNCIO) == 0) {
1.11.4.2  chs       722:                UVMHIST_LOG(ubchist, "returning PEND",0,0,0,0);
                    723:                return EINPROGRESS;
1.11.4.1  chs       724:        }
                    725:        if (bp != NULL) {
                    726:                error = biowait(mbp);
                    727:        }
                    728:        s = splbio();
                    729:        pool_put(&bufpool, mbp);
                    730:        splx(s);
1.11.4.4  chs       731:        for (i = 0; i < npages; i++) {
                    732:                UVMHIST_LOG(ubchist, "pgs[%d][0] = 0x%x",
                    733:                            i, *(int *)(kva + (i << PAGE_SHIFT)), 0,0);
                    734:        }
1.11.4.2  chs       735:        uvm_pagermapout((vaddr_t)kva, npages);
                    736:
                    737:        /*
                    738:         * if this we encountered a hole then we have to do a little more work.
                    739:         * for read faults, we must mark the page PG_RDONLY so that future
                    740:         * write accesses to the page will fault again.
                    741:         * for write faults, we must make sure that the backing store for
                    742:         * the page is completely allocated.
                    743:         */
                    744:
1.11.4.4  chs       745:        if (sawhole && ap->a_access_type == VM_PROT_WRITE) {
                    746:                error = VOP_BALLOC(vp, startoffset, npages << PAGE_SHIFT,
                    747:                                   cred, 0);
                    748:                if (error) {
                    749:                        UVMHIST_LOG(ubchist, "balloc lbn 0x%x -> %d",
                    750:                                    lbn, error,0,0);
                    751:                        goto out;
1.11.4.2  chs       752:                }
                    753:        }
                    754:
                    755:        /*
                    756:         * see if we want to start any readahead.
                    757:         * XXX writeme
                    758:         */
                    759:
                    760:        /*
                    761:         * we're almost done!  release the pages...
                    762:         * for errors, we free the pages.
                    763:         * otherwise we activate them and mark them as valid and clean.
                    764:         * also, unbusy all but the center page.
                    765:         */
1.11.4.1  chs       766:
                    767: out:
                    768:        if (error) {
                    769:                simple_lock(&uobj->vmobjlock);
1.11.4.2  chs       770:                for (i = 0; i < npages; i++) {
                    771:                        UVMHIST_LOG(ubchist, "examining pg %p flags 0x%x",
                    772:                                    pgs[i], pgs[i]->flags, 0,0);
                    773:                        if (pgs[i]->flags & PG_FAKE) {
                    774:                                if (pgs[i]->flags & PG_WANTED) {
                    775:                                        wakeup(pgs[i]);
                    776:                                }
                    777:                                uvm_pagefree(pgs[i]);
                    778:                        }
1.11.4.1  chs       779:                }
                    780:                simple_unlock(&uobj->vmobjlock);
1.11.4.2  chs       781:                UVMHIST_LOG(ubchist, "returning error %d", error,0,0,0);
                    782:                return error;
1.11.4.1  chs       783:        }
1.11.4.2  chs       784:
                    785:        UVMHIST_LOG(ubchist, "succeeding, npages %d", npages,0,0,0);
                    786:        simple_lock(&uobj->vmobjlock);
                    787:        for (i = 0; i < npages; i++) {
1.11.4.4  chs       788:                if (pgs[i] == NULL) {
                    789:                        continue;
                    790:                }
1.11.4.2  chs       791:                UVMHIST_LOG(ubchist, "examining pg %p flags 0x%x",
                    792:                            pgs[i], pgs[i]->flags, 0,0);
                    793:                if (pgs[i]->flags & PG_FAKE) {
                    794:                        UVMHIST_LOG(ubchist, "unfaking pg %p offset 0x%x",
                    795:                                    pgs[i], (int)pgs[i]->offset,0,0);
                    796:                        pgs[i]->flags &= ~(PG_FAKE);
                    797:                        pmap_clear_modify(PMAP_PGARG(pgs[i]));
                    798:                        pmap_clear_reference(PMAP_PGARG(pgs[i]));
                    799:                }
1.11.4.4  chs       800:                if (i < ridx || i >= ridx + *ap->a_count) {
1.11.4.2  chs       801:                        UVMHIST_LOG(ubchist, "unbusy pg %p offset 0x%x",
                    802:                                    pgs[i], (int)pgs[i]->offset,0,0);
                    803:                        /*
                    804:                        KASSERT((pgs[i]->flags & PG_RELEASED) == 0);
                    805:                        */
                    806:
                    807:                        if (pgs[i]->flags & PG_WANTED) {
                    808:                                wakeup(pgs[i]);
                    809:                        }
                    810:                        pgs[i]->flags &= ~(PG_WANTED|PG_BUSY);
                    811:                        UVM_PAGE_OWN(pgs[i], NULL);
                    812:                }
                    813:        }
                    814:        simple_unlock(&uobj->vmobjlock);
1.11.4.4  chs       815:        memcpy(ap->a_m, &pgs[ridx], *ap->a_count * sizeof(struct vm_page *));
1.11.4.2  chs       816:        return 0;
1.11.4.1  chs       817: }
                    818:
                    819: /*
                    820:  * generic VM putpages routine.
                    821:  * Write the given range of pages to backing store.
                    822:  */
                    823: int
                    824: genfs_putpages(v)
                    825:        void *v;
                    826: {
                    827:        struct vop_putpages_args /* {
                    828:                struct vnode *a_vp;
                    829:                struct vm_page **a_m;
                    830:                int a_count;
1.11.4.7! perseant  831:                int a_flags;
1.11.4.1  chs       832:                int *a_rtvals;
                    833:        } */ *ap = v;
                    834:
1.11.4.4  chs       835:        int s, error, npages, bshift, dev_bshift, dev_bsize, run;
                    836:        char * kva;
                    837:        off_t offset, startoffset;
                    838:        size_t bytes, iobytes, skipbytes;
1.11.4.1  chs       839:        daddr_t lbn, blkno;
                    840:        struct vm_page *pg;
                    841:        struct buf *mbp, *bp;
                    842:        struct vnode *vp = ap->a_vp;
                    843:        UVMHIST_FUNC("genfs_putpages"); UVMHIST_CALLED(ubchist);
                    844:
1.11.4.2  chs       845:        error = 0;
1.11.4.4  chs       846:        npages = ap->a_count;
1.11.4.1  chs       847:        bshift = vp->v_mount->mnt_fs_bshift;
                    848:        dev_bshift = vp->v_mount->mnt_dev_bshift;
                    849:        dev_bsize = 1 << dev_bshift;
                    850:
                    851:        pg = ap->a_m[0];
1.11.4.4  chs       852:        startoffset = pg->offset;
                    853:        bytes = min(npages << PAGE_SHIFT,
                    854:                    (vp->v_uvm.u_size - startoffset + dev_bsize - 1) &
                    855:                    ~((off_t)dev_bsize - 1));
                    856:        skipbytes = 0;
1.11.4.1  chs       857:
1.11.4.4  chs       858:        if (bytes == 0) {
                    859:                panic("genfs_putpages: bytes == 0??? vp %p", vp);
                    860:        }
                    861:
                    862:        kva = (void *)uvm_pagermapin(ap->a_m, npages, M_WAITOK);
1.11.4.2  chs       863:
1.11.4.1  chs       864:        s = splbio();
                    865:        vp->v_numoutput++;
                    866:        mbp = pool_get(&bufpool, PR_WAITOK);
1.11.4.4  chs       867:        UVMHIST_LOG(ubchist, "master vp %p bp %p num now %d",
                    868:                    vp, mbp, vp->v_numoutput, 0);
1.11.4.1  chs       869:        splx(s);
1.11.4.4  chs       870:        mbp->b_bufsize = npages << PAGE_SHIFT;
                    871:        mbp->b_data = kva;
1.11.4.2  chs       872:        mbp->b_resid = mbp->b_bcount = bytes;
1.11.4.7! perseant  873:        mbp->b_flags = B_BUSY|B_WRITE| ((ap->a_flags & PGO_SYNCIO) ? 0 : B_CALL) |
1.11.4.1  chs       874:                (curproc == uvm.pagedaemon_proc ? B_PDAEMON : 0);
                    875:        mbp->b_iodone = uvm_aio_biodone;
                    876:        mbp->b_vp = vp;
                    877:
                    878:        bp = NULL;
1.11.4.4  chs       879:        offset = startoffset;
1.11.4.1  chs       880:        for (; bytes > 0; offset += iobytes, bytes -= iobytes) {
                    881:                lbn = offset >> bshift;
                    882:                error = VOP_BMAP(vp, lbn, NULL, &blkno, &run);
                    883:                if (error) {
1.11.4.4  chs       884:                        UVMHIST_LOG(ubchist, "VOP_BMAP() -> %d", error,0,0,0);
1.11.4.1  chs       885:                        goto errout;
                    886:                }
                    887:
1.11.4.4  chs       888:                iobytes = min(((lbn + 1 + run) << bshift) - offset, bytes);
1.11.4.1  chs       889:                if (blkno == (daddr_t)-1) {
1.11.4.4  chs       890:                        skipbytes += iobytes;
                    891:                        continue;
1.11.4.1  chs       892:                }
                    893:
                    894:                /* if it's really one i/o, don't make a second buf */
1.11.4.4  chs       895:                if (offset == startoffset && iobytes == bytes) {
1.11.4.1  chs       896:                        bp = mbp;
                    897:                } else {
                    898:                        s = splbio();
                    899:                        vp->v_numoutput++;
                    900:                        bp = pool_get(&bufpool, PR_WAITOK);
1.11.4.4  chs       901:                        UVMHIST_LOG(ubchist, "vp %p bp %p num now %d",
                    902:                                    vp, bp, vp->v_numoutput, 0);
1.11.4.1  chs       903:                        splx(s);
1.11.4.4  chs       904:                        bp->b_data = kva + offset - pg->offset;
1.11.4.2  chs       905:                        bp->b_resid = bp->b_bcount = iobytes;
1.11.4.1  chs       906:                        bp->b_flags = B_BUSY|B_WRITE|B_CALL;
                    907:                        bp->b_iodone = uvm_aio_biodone1;
                    908:                        bp->b_vp = vp;
                    909:                }
                    910:                bp->b_lblkno = 0;
                    911:                bp->b_private = mbp;
                    912:
                    913:                /* adjust physical blkno for partial blocks */
                    914:                bp->b_blkno = blkno + ((offset - (lbn << bshift)) >>
                    915:                                       dev_bshift);
1.11.4.2  chs       916:                UVMHIST_LOG(ubchist, "vp %p offset 0x%x bcount 0x%x blkno 0x%x",
1.11.4.4  chs       917:                            vp, (int)offset, (int)bp->b_bcount,
                    918:                            (int)bp->b_blkno);
1.11.4.1  chs       919:                VOP_STRATEGY(bp);
                    920:        }
1.11.4.4  chs       921:        s = splbio();
                    922:        if (skipbytes) {
                    923:                mbp->b_resid -= skipbytes;
                    924:                if (mbp->b_resid == 0) {
                    925:                        biodone(mbp);
                    926:                }
                    927:        }
                    928:        splx(s);
1.11.4.7! perseant  929:        if (!(ap->a_flags & PGO_SYNCIO)) {
1.11.4.2  chs       930:                return EINPROGRESS;
1.11.4.1  chs       931:        }
                    932:
                    933: errout:
                    934:        if (bp != NULL) {
                    935:                error = biowait(mbp);
                    936:        }
                    937:        s = splbio();
                    938:        pool_put(&bufpool, mbp);
                    939:        splx(s);
1.11.4.4  chs       940:        uvm_pagermapout((vaddr_t)kva, npages);
1.11.4.1  chs       941:        UVMHIST_LOG(ubchist, "returning, error %d", error,0,0,0);
1.11.4.2  chs       942:        return error;
1.1       mycroft   943: }

CVSweb <webmaster@jp.NetBSD.org>